L II
THE IMPACT OF IRRIGATION FINANCIAL SELF-RELIANCE ON IRRIGATION SYSTEM PERFORMANCE IN THE PHILIPPINES
t I I I I
September 1991
THE IMPACT OF IRRIGATION FINANCIAL SELF-RELIANCE ON IRRIGATION SYSTEM PERFORMANCE IN THE PHILIPPINES
Mark Svendsen1
INTRODUCTION
The disappointing performance of many developing country irrigation sectors has led to the development and application of a variety of measures intended to improve performance of the individual irrigation systems comprising the sector These measures have focussed on improvements to the physical facilities of the irrigation system and to a lesser extent on the institutions involved in system management Recently the latter has often included the creation or strengthening of farmers organizations for this purpose Often efforts have been targeted at a particular level of the system--at the tertiary or on-farm level or more recently at the main systemThese efforts have demonstrated some successes but in general have themselves been disappointing and system operating efficiencies generally remain far below potential Particularly disappointing have been the short effective life span of many of the improvements
Attempts to modify the larger environment in which irrigation systems operate presumably addressing more fundamental underlying causes of poor performance have also been made Policy-level efforts have been directed at changing the levels of funding for irrigationdevelopment and the mix of new system construction and rehabilitation at increasing levels of funding for operations and maintenance (OampM) at upping cost recovery levels and at adjusting the larger set of price incentives that influence farmers production decisions Generally absent have been proposals for policy-level interventions to transform the fundamental character of the public agencies managingirrigation systems into more efficient forms This chapter describes an exception to this rule It examines the impact of a purposive transformation of the main public irrigation agency in the Philippines on the physical performance of the irrigation systems it operatesThe elemental question being asked is leaving aside any proceduralchanges or internal economies achieved do irrigation systems performbetter in a privatized environment
IRRIGATION CHARGING SYSTEMS
Purposes of Charging Systems
A key characteristic of the institutional environment of an irrigation service provider is the mode by which its operations are
- 2 shy
financed This in turn is closely tied to the system employed to charge beneficiaries for irrigation services
As outlined by Carruthers and Clark (1981) a charging system for irrigation water has economic financial and social functions The economic function is to ensure that resources are efficiently used by charging beneficiaries a price equivalent to the value that societyplaces on the resource employed This function works in two waysOne is the effect that water rates have on the construction of future projects When beneficiaries are charged for development costs potential water users with foreknowledge of the rates to be chargedlobby government for the construction of additional irrigationcapacity to serve them2 only to the extent that it is profitable for them Government for its part must examine anticipated revenue flows to itself as a result of the project and evaluate its capacityto repay indebtedness incurred
The other economic effect of a water charging system in theory acts on resource use within and among existing commands sharing a water source When charges are set in accordance with well established principles of marginal cost pricing water is presumed to flow to locations and seasons where marginal returns per unit water are the highest 3
The financial function of a charging system is to cover the costs of the service provided--that is the delivery of irrigation water Costs involved include capital investment costs OampM expenses costs of revenue collection and the cost of negative externalities created Many mechanisms will serve this purpose though often there are unintended or undesired side effects For example benefit capture byartificially restraining output prices below market levels or through forced procurement of output through government purchasing points mayreduce output levels significantly by causing farmers to shift to other crops reduce input levels or abandon unprofitable tenancies
The third function the social one is a mixed bag of policies and actions used to promote income redistribution [and] economic stability or to develop backward areas and encourage investment bybeneficiaries (Carruthers and Clark 1981) This third function mitigates to a major extent the strict application of economic principles which underlies the first two purposes For example it was for social and geopolitical purposes that federally subsidized irrigation development was employed as a mechanism for settling the American West during the first half of the twentieth century With the West settled these policies became the target of increasing criticism on financial and economic efficiency grounds In recent years policies have shifted to reduce significantly the federal role in and subsidies for irrigation development significantly alteringthe previous balance among the three types of criteria
- 3 shy
It should be clear from this brief description of the purposes of irrigation charging systems that cost recovery from projectbeneficiaries is not an end in itself but a way of achieving specificefficiency and equity ends within the national economy Small and coshyauthors (ADB 1986) have summarized these ends in the following set of criteria for usefulness of a charging system They assert that a charging system has appropriate impacts if
1 it results in improved irrigation performance through
(a) more efficient operation and maintenance of irrigationfacilities and
(b) more efficient use of water by farmers and
2 it promotes other objectives of the government by
(a) leading to better irrigation investment decisions (b) easing the governments financial burden and (c) resulting in a more equitable distribution of income
This paper focusses on the impacts subtended by item number one above
Fees and Financial Autonomy
In addition to rules which specify which groups shall bear the costs of providing irrigation service and in what proportions the policies and practices which specify how collections are coursed through the government financial system and the relationship collections bear to irrigation agency income are critical determinants of the agencys operating environment A necessary condition for functionally linking the collection of irrigation service fees and effective irrigation performance is that the agency involved in providing the service be financially autonomous (Small et al 1989 Svendsen 1986) Financial autonomy is defined as a condition where (a) the irrigation agency must rely on user charges for a significantportion of the resources used for OampM and (b) the agency has expenditure control over the use of the funds generated from these charges (Small 1990)
When financial autonomy is present several incentive forces come into play which are otherwise absent First there is incentive for the agency to increase its income Increased income for the agencyimplies maintenance of jobs higher salaries incentive payments and bonuses greater staff mobility and new vehicles quarters and facilities for the staff If fees are levied on an area basis as is usual in developing countries this means that the irrigation agencyhas an a strong vested interest in expanding the area receivingadequate irrigation service increasing fee collection rates and increasing farm incomes
- 4 shy
Second there is incentive to reduce costs Reduced operating costs are often necessary to keep the agencys budget in the black a condition which is usually required to retain independent status In many ways this motive runs counter to the one mentioned above and it is the dynamic tension between these two that creates an efficient and responsive service provider
Working together these two motives generate a demand for better agency relations with cultivators greater accuracy in information collection and record keeping new technology to manage information more effectively better water control and greater farmer involvement in system maintenance and fee collection Thus incentives for greater efficiency in resource use in a context of financial autonomy act on the providers of irrigation service at least as powerfully as the payment of water charges affects consumers of those services Overall a relationship of mutual dependency is established between the two where the agency provides an essential service to farmers and farmers in turn provide the agency with the financial resources it needs to operate
THE PHILIPPINE EXPERIMENT
Irrigation in the Phjlippines
The irrigation sector in the Philippines is divided between two major surface water components--Communal Irrigation Systems (CIS) and National Irrigation Systems (NIS) Groundwater irrigation in the Philippines has always been of secondary importance Communal irrigation is an ancient practice in the Philippines and is the more important segment in terms of net area irrigated currently covering about 48 percent of the nations irrigated area of 1488000 hectares This component of the sector is made up of generally small systems managed by farmers which are usually constructed by them as well Since the mid-seventies the National Irrigation Administration (NIA)has been actively involved in innovative efforts to assist CIS without compromising farmer ownership and operation
The other major component of the irrigation sector NIS comprises about 42 percent of total irrigated area and consists of larger systems developed and operated by NIA Operation of these systems consumes by far the largest share of NIAs operating budgetand constitute its largest potential source of revenue NIAs efforts to deal with these two salient characteristics are described brieflyin the following section
Evolution of a New Institutional Form
In a major departure from regional norms the Philippine Irrigation Department was abolished in 1964 and a public corporation created in its stead During the first decade of its existence the
- 5 shy
National Irrigation Administration (NIA) operated in a way that differed little from regular government departments A major overhaul of its charter in 1974 however led to far-reaching changes in NIAs organizational values structure and operations At the core of these values was the presumption that to be successful NIA must be financially viable taking in more income than it spent By 1979 it had achieved the goal of overall financial viability and in 1981 the last operating subsidy paid out from the national treasury was received
Policy Shift The major thrust of the 1974 charter amendment was to allow NIA to retain all revenues generated by it including irrigationfee collections Heretofore all collections had been immediatelyturned over to the treasury in exchange for an annual appropriationfor operating expenses unrelated to NIAs self-generated revenues The annual appropriation that NIA received had always exceeded by a significant margin the collections that it remitted Accompanyingthis shift however was an agreement that all government operatingsubsidies to NIA were to be phased out over the ensuing five-year period At the end of that period NIAs operating budget would be completely self-financed
NIA Response NIA management responded to these charter changes with a four part strategy aimed at bringing its costs and revenues into balance The strategy comprised actions to
bull Devolve responsibility for certain operational maintenance and fee-collection tasks to farmers
bull Increase corporate revenues by raising fees improvingcollections and generating secondary income from ancillaryactivities
bull Reduce operating costs through a series of minor economies and through major cuts in the personnel budget and
bull Provide financial incentives for superior performance to outstanding field units and to individuals in them
Following earlier successes in organizing farmers in the communal irrigation sector in 1980 NIA began experimenting with waysto organize farmers in its larger systems into effective irrigatorsassociations which could assume responsibility for some canal maintenance water allocation and fee collection functions By 1986 the area under various forms of farmer management had reached about 100000 hectares out of a total of about 600000 hectares in the country Depending on the specific type of devolution reductions in NIAs staffing levels in a sample of affected systems ranged from 13 to 75 percent (Svendsen et al 1989)
Immediately following the 1974 charter amendment NIA obtained permission to increase its fees for irrigation service At the same time fees were indexed for inflation by denominating them in measures of paddy NIA was authorized to collect fees in paddy just as village
- 6 shy
moneylenders do Since that time the agency has made strenuous efforts to increase its fee collections The net effect has been to hold fee revenues per hectare constant in the face of a national rice support price that has steadily declined relative to more generalindices of inflation
At the same time NIA also took steps to reduce its operating expenses Although a number of minor measures of economy were mandated initially the fact that more than three-quarters of the operating budget was devoted to personnel costs meant that any real savings would require reductions in staffing levels Voluntaryreductions were carried out in the late 1970s and early 1980s resulting in a decrease in the number of staff per hectare and a reduction of the personnel share of the budget from 80 percent in 1976 to about 74 percent in 1986
With an eye on its bottom line NIA also instituted a system of performance grants for all field units and the individuals in them termed Viability Incentive Grants To facilitate this each largeirrigation system in the country was made a separate cost center to allow costs and revenues to be accounted for on a system-by-system basis This program provided that once a unit achieved a net excess of revenues over operating costs in a given year a fraction of the surplus would be shared among the units personnel Five of the 11 irrigation regions of the country were receiving these incentive payments by 1986 as were 53 of the 120 individual systems included within the 11 regions
Effects The financial results of these efforts are shown in Table 1 If subsidies are not considered NIA first achieved net profitabilityin 1979 and retained it through the end of the period studied except for a small deficit incurred in 1981 Subsidies were eliminated in 1982 (except for occasional small calamity grants following typhoons) Although revenues have declined in recent years due largely to decreases in interest earnings and construction management fees expenses have declined more rapidly resulting in a series of net positive balances
Achieving a financially viable position is an importantaccomplishment few other irrigation agencies in the developing world have been able to do this However there is some risk that such an achievement occurs at the expense of the quality of service provided to clients Some of the most interesting and important consequences of the new cost recovery policies therefore relate to the physical performance of the irrigation systems NIA operates It is here that the end objectives of the irrigation investments are realized and where the lives of the farmers who till system lands are affected The remainder of the paper will examine and attempt to quantify the impact of these policy changes on physical irrigation systemperformance
- 7 shy
IMPACTS ON SYSTEM PERFORMANCE
Study Methodology
Unfortunately the kinds of changes in hydrologic system outputand the impact on agricultural performance which might be expected to result from improved institutional performance are quite difficult to capture and quantify There are a number of reasons for this First there is the year to year variability of system performance caused byvariable rainfall which feeds rivers fills reservoirs and supplements irrigation water in supplying crop requirements Second there is the difficulty of defining just what performance is and specifying how to measure it Third and most important the regularly collected data from which indicators can be constructed are limited in type number of measuring points and period of record and are sometimes of doubtful reliability
These difficulties notWithstanding an attempt was made to determine the impact that changes in operating procedures staffinglevels and incentive programs had on system performance Because the effects that we are trying to assess resulted from changes that affected all of the systems under NIAs direct authority there are no control systems which can be used as standards We are forced therefore to rely on a comparison of values of selected performance indicators before and after the date of the major structural and procedural changes which is taken to be 1981
To accomplish this secondary data were assembled for 5 systems in Administrative Regions III and VI which had not undergonesignificant phYSical changes during the period of analysis Duringthis process several site visits were made by study team members Time series data collected include service area (SA) and benefitted area (BA) for both wet and dry seasons yields for wet and dry seasons monthly main canal discharge at the system headworks and monthly precipitation The general period of availability for this data is 1966-86 though for systems which began operation after 1966 the period of record is shorter and the records of some systemscontain miSSing values These five systems their 1986 service and benefitted areas and other descriptive data are shown in Table 2
The prinCipal problem with using a before and after approachrather than one that considers comparable systems with and without the innovation is that some of the measured difference in effects mayhave resulted from causes which are independent of the ones beingstudied These causes can be specific in which case they may be relatively easy to identify and accommodate or more general and diffuse and therefore more difficult to control for In the present case--that of changes in the performance of NIA irrigation systemsresulting from the major organizational changes in 1981 two principalexternal factors can be identified which might be expected to affect differences in measured levels of irrigation performance between the
- 8 shy
two time periods These are rainfall and level of use of other agricultural inputs Since our interest is in systems managerial responses to these changes and since there is no reason to believe that the relative magnitudes of the responses are dependent on the size of the system the five systems are treated as equal in the analysis That is changes in measured values relating to the smallest system are considered to be as important as changes in values for the largest with no area weighting applied
Since rainfall can substitute for irrigation water supplies and since it affects the supply of water available in rivers for irrigation it may exert some independent influence on various performance indicators To test the strength of the relationship for the period being analyzed simple correlations were run between rainfall and benefitted area for one system in each regionBenefitted area was used in this analysis because it is the variable deemed most likely to be influenced by year-to-year changes in rainfall Weather data from Cabanatuan City was used for the UPRIIS system which surrounds it and Iloilo City data was used for the nearby Aganan-Santa Barbara system For UPRIIS all of the R2 values for these correlations were less than 0005 suggesting that rainfall has almost no impact on area harvested in this large reservoir-based scheme For Aganan-Santa Barbara wet season rainfall was related to wet season BA (r2 = 016) and to BA during the following dry season (r2 = 024) Signs of the simple correlations were in the expected directions ie wet season rainfall increased BA during both the wet and the subsequent dry seasons These connections are understandable but weak
Another possibility is that there were longer-term differences in rainfall received in the two regions If this were the case a comparison of performance during two different time periods would have to take this difference into account Differences in average precipitation during the two periods were examined for the four stations used in the analysis (see Table 2) In no case were differences in seasonal or annual mean rainfall statisticallysignificant4
bull
Nevertheless in the regression approach adopted to analyze the data rainfall was included in each equation to control for its possible effect on the particular dependent variables being analyzed In doing this wet season rainfall was used in analyzing wet season performance indicators while annual rainfall was used in analyzingdry season data The rationale for this is that while dry season rainfall cannot possibly influence the wet season crop the dry season crop is affected by both the rainfall received directly and the rain falling during the preceding wet season through its effect on river discharge reservoir storage and antecedent soil moisture conditions
The level of agricultural production is also an often-used indicator of an irrigation systems performance Its major weakness
- 9 shy
is that a number of factors other than irrigation service such as labor inputs relative prices and fertilizer use influence it It is necessary therefore either to control for changes in the levels of these inputs or assume that they are constant across the two periods being compared In the present case the most important of these factors is the level of application of chemical fertilizer Because fertilizer use by farmers is responsive to the relative pricesof fertilizer and rice it also includes to some extent input and output price effects Since reliable data on fertilizer use for individual systems were not available estimates derived from FAO fertilizer and cultivated rice area data were used t~ control for the effect of changes in the use of this input over time This variable was included in any of the regression equations in which agriculturalproduction was used as the dependent variable Other factors such as labor use genetic potential of varieties sewn and pesticideapplications are assumed to be constant across the two periods
The analytic approach employed is to fit linear regressionequations to pooled data from the five systems covering a eleven-year period 1976 to 1986 A dummy variable is used to check the impact of pre and post 1981 periods on differences in the dependent variable after the effects of factors such as rainfall and nitrogen fertilizer use have been removed In addition because the dataset was created by pooling data from five different systems a set of 4 site dummies was included in the basic model to control for system-specific differences caused by variables which were not measured For some runs these were replaced with dummies that separated reservoir and non-reservoir systems though equations using the reservoir dummy were consistently inferior to those using the complete set of site dummies Several different dependent variables were created to index the quality of irrigation service and tested using this approachRegression results are given in Tables 3 and 5 and discussed below
Performance Indicators
A variety of indicators have been used in evaluating irrigationperformance in various contexts The selection of appropriate indicators depends on a number of factors including the purpose of the evaluation the audience for its results the way in which the boundaries of the irrigation system are defined and the kind and quality of data available to the evaluators The current analysis is designed to evaluate the impact of a set of management changes on system physical performance The audience for this analysis comprises top-level managers of the irrigation agency and policy-makers at higher government levels Boundary definition is an importantanalytic problem here as evident from the subsequent discussion relating to the choice of the appropriate area values to use in scaling system inputs and outputs This issue is also related to the data quality and availability problems which have already been mentioned
- 10 shy
Three fundamental indicators have been proposed for assessing the effectiveness of irrigation services to cultivators (Svendsen and Small 1989 Abernethy 1990) These are the adequacy of water supplies the equity of their distribution across the command area of the system and the timeliness of the supplies Computation of adequacy measures requires information on the total quantity of water delivered to the system over a season on a per hectare basis Equityand timeliness measures require information on the spatial and temporal distribution respectively of those supplies Where appropriate discharge information is not available proxies can be employed by making suitable assumptions Standards must be selected against which the magnitude of the indicators can be judged
The task in the present case however is somewhat different Here the need is to evaluate changes in selected variables between the pre-1981 and the post-1981 periods Hence the absolute values of variables selected are less important than their relative magnitudesand the statistical significance of the differences in magnitudes between the two periods A distinct limitation is imposed by the data series available for the five sample systems Since discharge and yield data are available only on a whole-system basis it is impossible to develop measures of equity and timeliness directly We will however extend our analysis to a discussion of equity byindirect inference Levine and Coward (1986) have argued that equityought to be considered as the paramount objective in managing largepublic irrigation systems They base their conclusion on an analysisof eight small community-managed systems and five larger public systems including UPRIIS in which equity appears to comprise the most important operational objective in the successful systems It may be appropriate therefore to give success in improving equity of distribution added weight in assessing overall performance
Area Estimates Because measures of system agricultural output and water supplied are typically reduced to a unit area basis before being used much depends on the area values which are used to standardize them Two different area measures are available The first is Service Area (SA) which is defined as the irrigable portion of the command area which is provided with physical facilities for water delivery This represents the area which could conceivably be irrigated in a given season if water supply were not constraining This value may change somewhat from year to year in response to urban encroachment on irrigated command minor remodeling and repair and refinements in area estimates In the present case though the systems selected for analysis were chosen to avoid those which had undergone more extensive rehabilitation or modification
The second measure is Benefitted Area (BA) which is the area billed for payment of irrigation service fees It is the irrigated area harvested which did not have yields so low that the farm was exempted from payment of fees in a given season This threshold value has been approximately 2 tons of paddy per hectare Benefitted area
- 11 shy
varies more than does SA particularly during the dry season when available water supply may seriously constrain the area which can be planted Its magnitude is a function of system managers actions in authorizing the amount of land to be planted in a given season farmers decisions regarding whether to plant or not and the combined ability of system managers and farmersirrigators subsequently to distribute water Both of these area measures will be used to standardize other variables for particular purposes as well as beingcombined to form a separate indicator by themselves
Adequacy The most direct measure of the adequacy of irrigation water supplies to the agricultural system is the quantity of water applied to the system command area on a per unit area basis relative to some standard In this case since our interest is in differences in water adequacy between two time periods and since the systems being assessed have been and continue to be almost entirely devoted to rice cultivation during both cropping seasons depth measures for the two periods may be compared directly assuming the seasonal cropdemand for water to be unchanged Although dry-footed crops can suffer yield losses from overapplication of irrigation water rice is largely insensitive to this effect In addition water can substitute for other inputs that the farmer would otherwise have to provide such as weed control and more careful (and costly) water management We assume therefore that other things being equal larger values of depth applied are better than smaller values in terms of meeting crop water demands and reduce the cost of cultivation At the same time high levels of water adequacy can affect the values of other performance measures--particularly equity
When the regression model is run for quantity of water diverted at the system headworks divided by BA hereafter termed depth we see that the period dummy is negative and significant at the 95 percent confidence level for both wet and dry seasons (see Table 3 equations 1 and 3) Since the overall explanatory power of the wet season model is very weak however we will focus on the dry season in interpreting this result which indicates that after adjusting for rainfall differences significantly less water was delivered to the command per unit of benefitted area following 1981 than before This indicates based on the criteria outlined above that performance in terms of water adequacy deteriorated following financial selfshysufficiency We need to examine this conclusion more carefullyhowever
One difficulty is that the measured quantity of water diverted at the source is largely a function of the supply available in the river rather than of system management This is particularly true during the dry season and in non-reservoir systems Thus while the depth of water supplied to the system is a measure of the adequacy of the systems service it is to some extent beyond the control of the managing agency To better understand the factors behind this decline in water availability we look at simple unadjusted index values for
- 12 shy
several of the key variables Table 4 shows annual values of total volume of water delivered in each season SA and BA in both wet and dry seasons and shows the results of t-tests on the means of a set of indicators before and after 1981 Indicators are used rather than the actual values to weight each of the systems equally regardless of its size The table shows that both the average wet season benefitted area and the average discharge are significantlylower during the second period compared with the first For the dry season too the discharge index is lower after 1981 than before but this difference is not significant At the same time the dry season BA index rose slightly but again the change was not significantSince middotthere is not a clear pattern of relative movement of discharge and BA during the respective seasons no simple interpretation of these index value changes is possible What stands out is that both discharge and benefitted area declined across periods during the wet season while during the dry season there was no significant change in either indicator across the two periods It seems clear that the decline in water adequacy must be evaluated together with other measures of performance in drawing conclusions about the overall impact of the 1981 changes on the quality of system management
Another measured variable per hectare yield can be used as a proxy for water adequacy It has the advantage of partiallyreflecting the impacts of the dimensions of timeliness6 and equity7of distribution as well integrating all three effects into a combined impact on aggregate crop production Table 5 (equations 1 and 3)shows that the period dummy in the yield regressions has a positive sign in both seasons after controlling for nitrogen application and precipitation though the t-values are not significant Treatingyield adjusted in this way as a proxy for quality of irrigationservice leads to the conclusion that by this more comprehensive measure quality of service held constant across the two periods in the dry season Because of the large yield component accounted for byrainfall during the wet season no such judgement is possible for that season however
Equity As noted earlier no reliable data are available for subdivisions of the five sample systems making direct computation of equity measures impossible We can make some judgements about changesin the equity of water distribution however by examining changes in the ratio of two area measures given for each system SA and BA Since SA is the area which theoretically can be supplied with irrigation water by the system and BA s the area which actuallyreceives a quantity of water adequate to produce a remunerative crop the ratio of the two provides a measure of the percentage of the potential service area which was irrigated to a particular standard The larger this pErcentage the more equitable 8 is the distribution This of course assumes that the quantity of water available to the systems is constant across the two periods
- 13 shy
Since the condition of constancy of water supply is not generally satisfied a regression was run in which the total quantityof water diverted at the headworks of each system divided by the systems potential service area SA was included in the regression to control for changes in the water supply available to the system seeTable 3 equations 5 through 8 The average daily rainfall received directly on the system service area during the season was also included as an independent variable The regression was run separately for wet and dry seasons The sign and t-statistic of the period dummy should then tell us whether or not equity as reflected in the BASA ratio increased decreased or remained unchanged across the period divide
Both equations are reasonable good as indicated by the R2 values though the dry season equation is considerably better as would be expected For the wet season both the water delivery term and the rainfall term in equation 5 are of positive sign but are nonshysignificant at the 95 percent confidence level indicating that wet season irrigated area does not change appreciably in response to level of wet season rainfall or the available irrigation water supply The period dummy was negative but not significant indicating that equityof distribution as reflected in the BASA ratio was similar during the two periods
For the dry season the water delivery term in equation 7 is positive and strongly significant indicating a close relationship between the fraction of potential area actually irrigated and the water supply available at the headworks In addition however the period dummy is positive and significant suggesting that once the influence of water supply is removed the BASA ratio was significantly higher in the period following 1981 than it was before
This is an important finding for it reflects significantlyimproved performance in terms of a factor equity of water distribution that is under the control of the managing entity an entity which here comprises both NIA and irrigators associations Interpreted in these terms NIA and allied farmers associations were able to spread a given amount of water more widely across the potential command area of the five sample systems in the period after 1981 than before Moreover they did this in a way that did not decrease average system yields as discussed earlier In making this interpretation we are suggesting that there was some redistribution of water from better-watered areas to fringe areas which would otherwise not have received irrigation water and that this redistribution was a direct response to the change in NIA prioritiesand operating policies and rules occurring around 1981
It is difficult to prove the assertion that water was in fact redistributed with a resulting increase in directly-measured equity Without access to reliable discharge data broken out by systemsection and we can only assume in the absence of a plausible
- 14 shy
alternative explanation that it was such a redistribution that made the increase in the BASA ratio possible In a larger sense it is difficult to prove conclusively that any outcome in a before and after analysis was the result of a particular independent causative factor In this case we have tried to remove the influence of other potential causative factors where we could but the possibilityremains that some combination of unmeasured factors are responsiblefor the difference in the BASA ratio found We do note though that this type of response is exactly the type that would be expected to follow from an emphasis on increased farmer satisfaction and cooperation and increased fee revenues Because the fee schedule is tied to benefitted area the only ways NIA can increase its revenue from that source are to expand benefitted area and to increase collection efficiencies The former depends on redistributing a fixed supply of water over a larger portion of the command while the latter requires that farmers be satisfied with the irrigation service they are receiving and the commitment of the local irrigators association to assist in the task of collecting the amounts due The evidence while not conclusive is highly suggestive that this is exactly what has happened
Efficiency In addition to measures which reflect the levels of adequacy and equity of irrigation service available data allow the calculation of a measure of operating efficiency The term efficiencyusually denotes the relationship between inputs to a process and its outputs often expressed as a ratio The output measure employed here is aggregate system rice output and the input is quantity of irrigation water turned into the system Dividing the first by the second gives a measure of agricultural production per unit water--here termed specific yield This is a highly integrated measure that evaluates the combined efficiency of the irrigation and agricultural processes As such it is a function of the managerial and other inputs supplied both to the irrigation system and to the agricultural operation With respect to one important input to the irrigation system we do know that NIA per hectare field operating expenses were about 29 percent lower in real terms in the 1982-86 period comparedto the 1976-1981 period although this drop may have been partlyoffset by increases in farmer-supplied labor inputs Other things being equal one would thus expect to find a decline in output efficiency
The regression analysis shows positive signs for the period terms in both wet and dry season equations (see Table 5) In the case of the wet season the period dummy in equation 5 is significant but the overall explanatory power of the model is quite low For the dry season (equation 7) the coefficient is positive but non-significant This means that after taking rainfall and fertilizer use into account data do not indicate a lowering of specific yield in the wake of funding reductions and the strong emphasiS on financial viabilitybeginning in 1981 This result provides evidence that the efficiency of the overall irrigation deliveryagricultural production process
- 15 shy
relative to the system water input did not falloff as a result of the changes implemented at least over the short run
Impact magnitude
The preceding analysis has shown us that some indicators of irrigation performance changed significantly following the managerial changes of 1981 while others did not However it has not given us a sense of the size of the changes which occurred To determine the magnitude of these changes the regression model is used to predict the response of the composite system to the managerial changes given a common set of -input and environmental factors To do this averagevalues of the independent variables from the entire eleven-year period 1976 to 1986 are put into the model together with the previously determined coefficients to generate predicted average values of the various dependent variables used in the earlier analysis with and without the period dummy This procedure produces a pair of estimates for each dependent variable under the same conditions--one in which the system responds as it did after the managerial changes were implemented and one in which it responds as it did prior to their introduction The differences between these two values thus indicate the magnitude of the changes occurring in the various indicators of performance discussed above
The results of this exercise are shown in Table 6 The table shows that water availability decreased by about 13 percent in both wet and dry seasons when the period dummy was included and while the coefficients responsible were significant in the earlier analysisthis difference cannot be easily connected with levels of system management as discussed earlier With respect to rice output per hectare although the coefficients were not very significant it is interesting to note that yield increases by 163 kilograms per hectare for the wet season and by 101 kilogram for the dry when the period dummy is included in spite of the reduced water supply available Keep in mind that the predicted yield values have already been adjusted for differences in nitrogen fertilizer use and rainfall This suggests that timeliness and equity of distribution of water supply to farmers may have increased following the changes contributing to the higher predicted yields
Examining the impact of increased equity of distribution bylooking at the ratio of benefitted area to service area we recall that the change was positive and significant for the dry season and negative and not significant for the wet Table 6 shows that the dry season BASA ratio increases by 7 percentage points when the dummy is included a 131 percent increase Other things being equal this should result in a 131 percent increase in system output due to the expansion of ared benefitted This is a major impact on production
- 16 shy
CONCLUSIONS
The Philippine experiment to transform the national irrigation agency into an enterprise has undoubtedly been successful in reducing system operating expenses bringing revenues and costs into line and eliminating the recurrent cost burden imposed by large-scaleirrigation systems on the national budget Evidence presented in this paper indicates that in the process equity of water distribution across systems has also improved In the 5 years following the cessation of operating subsidies from the government an index of equity of distribution improved by about 13 percent At the same time per hectare yields adjusted for rainfall and nitrogen application held constant
There is a strong logical connection between the achievement of financial viability and improved equity of water distribution across the command Because increasing irrigation fees is a politicaldecisionlying largely beyond NIAs control expanding the area which can be billed for service is one of the few revenue increasing measures available to the irrigation agency which does not involve major additional investment In the face of constant or shrinkingwater supplies this is achieved only by redistributing water from areas receiving excessive supplies usually near the head ends of canals and laterals to areas receiving no supplies or inadequatesupplies often located near the tails of canals Although data are not available which would allow the direct examination of this hypotheses the two outcomes are logically consistent with each other
Data also show that per hectare water deliveries declined significantly in the five sample systems after 1981 even thoughrainfall did not differ appreciably between the two periods This decline averaged about 13 percent for both wet and dry seasons and is interpreted as a decline in water availability in the supplying rivers rather than a conscious reduction in withdrawals by system managers Such declines could result from changes in watershed runoff characteristics as caused by deforestation or from increased upstream abstractions from supplying rivers
Improved water distribution tends to increase the area served system agricultural output and NIA service fee revenue Reduced water supplies to the system tend to reduce these things Specificyield defined as system paddy output per unit water held roughly constant across the two periods indicating that the two effects mayhave offset each other
After adjusting for rainfall and nitrogen application perhectare yields increased only marginally in the post-1981 period Area served on the other hand increased by about 13 percent after adjusting for water supply availability indicating that the area benefitted by irrigation in the sample systems increased by about the same percentage Even if yields on this additional area are less than
- 17 shy
average yields for the system this still represents a sizeable increase in system agricultural output as a result of the change in management structure the increase coming not from higher yields but from expanded area under irrigation
The evidence assembled here suggests that there are significantfinancial and economic benefits to be had from changes in the basic character of irrigation managing agencies which make them more responsive to their clientele and which impose rational internal financial discipline on the agency The analysis suggests a number of additional questions however One relates to the longer-term impacts of the structural management changes The improvements in water distribution described here are relatively short-term events occurring during the first 5 years of the new management mode Critics have suggested the danger of underinvestment in systemmaintenance over the longer run accompanied by declining yields and benefitted areas and eventual system collapse This possibility needs to be closely monitored A second concern relates to the apparent decline in water supply to these 5 geographically dispersed systems The nature and causes of this decline need to be explored further since if widespread and secular it may represent a serious threat to the stability of Philippine rice production Whether stemming from poor forest management practices or deficient regulation and allocation of surface water resources or other unidentified factors it is an issue that deserves serious and urgent consideration
A third risk is that the incentive structure set up by NIA to guide and stimulate the performance of field units overemphasizes revenue generation at the expense of irrigation service provision to farmers The evidence presented here supports the view that these two objectives are mutually reinforcing under policies and conditions which have been established in the Philippines More detailed crossshysectional studies based on primary flow measurement data would add confidence to this conclusion and help to specify the conditions under which this effect occurs This could be extremely important in transferring the results of the Philippine experiment to other countries
A final risk is that outside intervention well meaning or otherwise will destroy the basis of NIAs financial autonomy or will impose external pressures or constraints on NIAs decision-making that will subvert the management practices which have been so painstakinglydeveloped and implemented Among these are calls for NIA to be subsumed again within the government department structure in the interests of better coordination with agriculture attempts byexternal financing agencies to arbitrarily increase NIAs expenditures on OampM on the assumption that this will increase system agricultural output or intervention by Philippine legislative bodies to restore operating subsidies to NIA with attached strings leading back to legislators home districts Pressures such as these will cut short a process of experimentation and improvement that seems promising
- 18 shy
enough to date to warrant its continuation Having developed the capacity to establish targets and implement and manage change NIA is in a strong position to modify its objectives to better achieve larger social purposes established for it It is critical to recognize however that this must happen within the context of financing policies that mandate financial autonomy for NIA if the fundamental institutional commitment to manage is to be preserved
The author would like to thank Leslie Small and JeremyBerkoff for helpful comments on an earlier unpublishedversion of this paper and Charles Rogers for his careful and creative help with the analysis
- 19 shy
BIBLIOGRAPHY
Abernethy Charles L 1990 Indicators of the performance of irrigation water distribution systems International Irrigation Management Institute Colombo Sri Lanka Mimeo
Asian Development Bank 1986 Irrigation service fees Proceedingsof the Regional Seminar on Irrigation Service Fees Manila Asian Development Bank
Carruthers Ian and Colin Clark 1981 Economics of IrrigationLiverpool Liverpool University Press Third Edition
Levine G and EW Coward Jr 1986 Irrigation water distribution implications for design and operation AGREP Division WorkingPaper 125 vol 1 World Bank Agriculture and Rural Development Department
Small Les E 1989 User charges in irrigation potentials and limitations Irrigation and drainage vol 3 no 2125-142
Small Les 1990 Irrigation service fees in Asia IrrigationManagement Network 9013 London Overseas DevelopmentInstitute
Svendsen Mark and Les Small 1989 A framework for assessing irrigation system performance Paper prepared for the Symposium on Performance Evaluation 23 November International IrrigationManagement Institute Sri Lanka
Table I--National Irrigation Administration revenues and expenditures in constant prices 1976-86
Item 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986
(peso million 1972)
Revenues Irrigation fees collected 1273 1483 17 13 1831 2070 1668 1699 1893 1728 2129 2546 Other income 715 737 2420 5591 2631 5990 7783 6638 5699 4932 2934
Total direct revenue 1988 2220 4133 7422 4701 7658 9482 8531 7427 061 5480
Expenses in 1972 pricesTotal expenses 4825 5716 5039 6329 3821 77 55 6166 4749 4348 4259 4959
Excess (deficit) (2837(3496) (906) 1093 877 (097) 3316 3782 3079 2802 521 N 0
Subsidies Government operation and
maintenance subsidies 2521 2741 2799 1817 1398 633 0 0 0 0 0 Calamity fund payments 548 0 0 0 0 0 0 0 119 0 142
Total subsidy 3069 2741 2799 1817 1398 633 0 0 119 0 142
Total excess (deficit) 231 (754) 1893 2910 2275 536 3316 3782 3198 2802 663
Source IFPRI analysis of NIA data
- 21 shy
Table 2--Descriptive characteristics of selected MIA systeasa
------- Region III -------- ----- Region VI ----shyUPRIIS Angatii Sto Sibalom- Aganan-
Haasim Thomasc San Jose Sta Barbara
Average service area (hal 102272 31462 3522 5282 8703 Average irrigated area (hal
Wet season 83768 23454 3007 4410 8300 Dry season
Average benefited area Wet season
(hal 64587
77 605
27639
22908
1 781
3007
2801
4369
2770
7698 Dry season
Average rainfall Wet season
(mml d 62478
1 685 5
27396
8576
1 781
3051 0
2769
24731
2997
20001 Dry season 756 333 322 2828 3025
Average discharge (Llsec) Wet season 46501 14792 1692 2353 4984 Dry season 78091 22812 2014 1276 2315
Average water delivery (mmday) Wet season 522 548 487 462 571 Dry season
Average yield (mtha) 1089 715 995 398 686
Wet season 345 419 322 395 435 Dry season
Avg yield per unit water 34 03
(kgm ) 451 412 399 426
Wet season 0373 0440 0373 0538 0443 Dry season 0248 0400 0279 0690 0428
t-statistic difference in mean rainfall 1978-81 1982-86e
Wet season 0432 0713 -0567 1169 1169 Dry season 0519 -0230 -0523 1187 1187 Annual 0460 0707 -0686 1445 1445
~ Summary numbers are averages for the period 1982-1986 except as noted Water delivery discharge and yield per unit discharge are 4-year averages 1982-1985
c Water delivery discharge and yield per unit discharge are 4-year averages d 1983-1986
For Angat 5 years are 1981-85 For St Thomas 1979-83 For Sibalom 1971-75 e No significant differences at 95 confidence
- 22 shy
Table 3--Results of pooled regressions water delivery indicators
Dependent variable Independent variable - Water Delivery in mmBenefited Ha - --- BenefitService Area Ratio --shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season-shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 6799 (998)
6189 (1019)
13102 (1093)
2411 (1 54)
0743 ( 1186)
0800 (1165)
0129 (131 )
0258 (329)
System 2 indicator 0173 (029)
-3743 (-363)
-0063 (-219)
0200 (505)
System 3 indicator 0698 (085)
0163 (012)
0085 (208)
-0046 (-097)
System 4 indicator -0514 (-l 07)
-7011 (-803)
0080 (340)
0169 (l87)
System 5 indicator 0177 (033)
-5049 (-538)
0174 (603)
0051 (058)
Reservoir indicator 0327 (078)
5703 (543)
-0126 (-510)
0156 (261)
Period indicatora -0808 (-240)
-0771 (-229)
-1214 (-209)
-1309 (-1 62)
-0036 (-192)
-0025 (-112)
0071 (239)
0030 (077)
Water delivery mmb
Wet season
Dry season
Precipitation in mm Wet season
Annual
Average daily rainfall Wet season
Dry season
-46E-4 (-125)
-1 9E-4 (-082)
-43E-4 (-077)
20E-3 (385)
0015 (1 59)
-0001 ( -004)
0023 (215)
0001 (021)
0070 (613)
-0017 (-058)
0046 (371)
0010 (033)
R2 (adj) number of observations degrees of freedom
0156 43 36
0133 43 39
0711 41 34
0434 41 37
0693 42 34
0556 42 37
0865 40 32
0730 40 35
Coefficient (t statistic) indicates significance at 95 confidence ~ 1982-1986 1
per ha of service area
Ta
ble
--I
nd
ices
of
se
v
ice a
rea
ib
en
ef
ited
a
rea
a
nd
a
vera
ge sea
so
na
l d
isch
arg
e
Ave
rage
A
vera
ge
tshy19
77
1978
19
79
1980
19
81
1982
19
83
1984
19
85
1986
19
77-8
1 19
82-8
6 S
tati
stic
a
(ind
ex
aver
age
1983
-198
5 =
100)
Ser
vice
are
a UP
RI IS
in
dex
882
91
9
920
91
5
951
95
1
100
0 10
00
100
0 10
00
917
99
0
bull5
53
Ang
at-M
aasi
m R
95
9
994
99
7
996
99
6
996
10
00
100
0 10
00
100
0 98
8
999
1
63
Sto
To
mas
10
63
106
3 10
1 9
10
29
103
0 99
8
100
0 10
00
100
0 11
10
104
1 10
22
-08
9
Siba
lom
-San
Jos
e 95
0
872
94
2
933
94
2
942
10
28
102
8 94
4
101
7
928
99
2
292
A
gana
n-St
a B
arba
ra
108
3 10
91
106
0 10
30
961
10
05
996
10
08
996
99
6
104
5 10
00
-21
1 A
vera
ge
987
98
8
981
98
1
916
97
8
100
5 10
07
988
10
25
984
10
01
123
Wet
seas
on b
enef
ited
are
a in
dex
UPR
IIS
110
0 10
07
114
4 10
55
113
8 11
74
951
10
71
971
ll
58
10
89
106
7 -0
47
Ang
at-M
aasi
m R
97
9
974
92
0
983
10
28
100
8 99
7
102
1 98
2
931
97
7
988
0
54
Sto
To
mas
11
63
115
9 11
23
107
5 10
80
103
9 98
1
978
10
41
103
9 11
20
101
6 -4
88
Siba
lom
-San
Jos
e 11
35
103
8 10
11
931
93
4
906
10
07
975
10
1S
998
10
11
981
-0
80
Aga
nan-
Sta
Bar
bara
10
85
110
5 10
74
106
8 10
01
104
2 99
8
101
5
987
61
0
106
7 93
0
-1 8
4
Ave
rage
10
92
105
6 10
54
102
4 10
36
103
4 98
7
101
3 10
00
947
10
53
996
-2
32
Dry
sea
Son
bene
fite
d ar
ea
inde
x U
PRIIS
14
04
155
0 15
S0
155
8 16
17
128
0 57
2
114
S 15
74
152
3 12
38
-09
4 A
ngat
-Maa
sim
R
903
93
0
103
2 10
61
104
2 10
69
988
99
2
102
1 99
6
993
10
13
061
S
to
Tom
as
105
7 12
27
122
5
961
99
9
115
9 10
1 7
91
0
107
3 12
1 2
10
94
107
4 -0
28
Si
balo
m-S
an J
ose
Aga
nan-
Sta
Bar
bara
66
5
95S
62
6
632
67
4
111
4
501
10
S1
412
11
51
766
93
3
856
94
3
107
0 94
8
107
4 11
09
111
6
158
6 58
8
987
97
6
110
4 5
35
083
N
w
Ave
rage
89
6
964
11
19
103
7 10
44
110
9 10
1 7
89
8
108
5 12
97
101
7 10
81
082
Wet
seas
on d
isch
arge
in
dex
UPR
IIS
132
9 72
7
142
5 11
88
120
4 98
8
105
8 10
63
879
96
4
117
5 99
1
-16
5 A
ngat
-Maa
sim
R
129
7 13
52
134
5 12
58
127
0 11
70
120
3 62
7
131
3 10
68
-18
9
Sto
To
mas
14
71
155
0 14
1 6
10
40
725
12
35
112
3 14
79
103
1 -4
48
Siba
lom
-San
Jos
e 96
8
733
11
1 5
92
2
907
47
1
102
3 85
7
ll2
0
422
92
9
779
-1
09
Aga
nan-
Sta
Bar
bara
87
9
863
68
1
925
96
8
110
7 70
5
871
87
7
00
9
Ave
rage
11
49
105
7 13
61
115
0 10
58
853
10
43
963
99
4
SO3
11
55
940
-2
85
Dry
sea
son
disc
harg
e in
dex
UPR
IIS
425
13
09
153
3 14
28
180
6 14
S0
125
8 56
3
117
9 14
66
130
0 11
89
-04
3
Ang
at-M
aasi
m R
12
26
161
6 15
30
139
2 12
85
100
8 10
33
958
14
41
107
1 -3
80
S
to
Tom
as
116
4 14
82
155
7 79
9
971
12
30
126
5 14
01
106
6 -2
44
Si
balo
m-S
an J
ose
486
71
5
782
62
9
789
64
5
116
9 11
86
801
65
3
918
2
36
Aga
nan-
Sta
Bar
bara
A
vera
ge
425
10
46
133
6 62
0
118
4 81
5
116
1 94
7
112
5 89
7
922
12
20
991
88
2
108
7 12
36
119
2 71
8
114
0 10
37
105
5 3
23
-07
5
a bull
indi
cate
s si
gnif
ican
ce a
t 95
per
cent
con
fide
nce
- 24 shy
Table 5--Results of pooled regressions yield indicators
------------------------- Dependent variables -------------------------shyIndependent variable ---------- Yield MTHa ----------- -- Specific Yield 1000 KgM
3 H20 shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season -shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 1967 (286)
3747 (580)
1991 (293)
2800 (460 )
0084 (041)
0441 (229)
0049 (023 )
0876 (343)
System 2 indicator 0400 (155)
0525 (226)
0057 (085)
0115 (1 95)
System 3 indicator 0466 (139 )
-0029 (-010)
-0016 (-018)
0006 (008)
System 4 indicator 0980 (470)
0090 (049)
0176 (333)
0391 (827)
System 5 indicator 1 221 (570)
0256 ( 138)
0104 (174)
0224 (440)
Reservoir indicator -1033 (-578)
-0063 (-039)
-0146 (-310)
-0326 (-569)
Period indicatora
Precipitation in mm Wet season
Annual
0163 (1 03)
-29E-4 (-182)
0231 (1 42)
-58E-4 (-606)
0161 (118)
92E-5 (072)
0199 043 )
-64pound-5 (-084)
0082 (200)
57E-6 (013)
0092 (223)
-61E-5 (-232)
0032 (093)
29E-7 (001)
0052 (112)
-12pound-4 (-414)
Nitrogen applicationb 0025 (254)
0021 (211)
0025 (274)
0020 (227)
0003 (1 05)
0002 (043)
0003 (101)
-0002 (-052)
R2 (adj) 0553 0516 0251 0198 0309 0262 0713 0445 number of observations 50 50 49 49 42 42 40 40 degrees of freedom 42 45 41 44 34 37 32 35
CoeffiCient (t statistic) indicates significance at 95 confidence ~ 1982-1986 = 1
kgha
- 25 shy
Table 6--Suggmry of estiates perforance indicators
Variable Predicted Predicted Period Dependent Variable
Description (units)
Period Coefficient
1976-86 IndicatorO
1976-86 Indicatorl
Effecta (X)
BenefittedService area ratio
Wet season ratio -0036 0853 0817 -425 Dry Season 0071 0539 0610 1313
Delivery Wet Dry
season season
nm water per benefitted ha
-0808 -1 214
6003 9083
5195 7869
-1346 -1337
Rice Yield Wet season metric tons 0163 3578 3741 456 Dry season per hectare 0161 3905 4066 412
Specific Yield Wet season kg rice per m3 0082 0348 0430 2353 Dry season water de 1 ivered 0032 0378 0409 837
a percentage changes relative to period=O values
- 26 shy
ENDNOTES
IResearch Fellow International Food Policy Research Institute and Senior Irrigation Specialist International Irrigation Management Institute
2Presumable farmers also base private investment decisions on such foreknowledge of costs and commonly do so in the case of small tubewells In addition user groups may invest in small communal schemes independently after examining projected costs and benefits Most commonly however new irrigation capacity is created as a result of public investment decisions made by government and thus the political process comes into play
3This principle applies only when water charges are levied volumetrically--an exceedingly rare situation in the developing worldshy-though the argument is frequently invoked as though it applies to all allocative situations
4For Cabanatuan City the longer period of record available ie 1904 to 1986 was divided into two equal periods and tested for difference in means to test the observation by farmers that rainfall today is lower than it used to be Mean precipitation after 1947 was actually about 3 percent higher than before but the difference was not significant (t = 07)
5The procedure used was to take time series data on total nitrogenfertilizer consumption in the Philippines multiply it by the ratio of N fertilizer used on rice to total N consumption and divide the product by the rice harvested area to obtain an estimate of N use perhectare Values of the ratio of N use on rice to total N use were taken from IRRI (1988 150) with interpolation used to fill in gaps in the series
6This assumes that timeliness is evaluated against a standard based on the physiological demand for water generated by the rice crop and the varying sensitivity of the crop to stress at different growth stages In such a case a poor timeliness rating for irrigationservice will result in reductions in yield
7Because rice yields are insensitive to water applications that exceed PET values a redistribution of water from water excess to water short portions of the command area will usually result in higher aggregate output and higher average yields Exceptional scenarios can be constructed in which this general rule would be violated but such scenarios are unlikely to play out in practice
- 27 shy
aThere are a variety of standards for judging equity that could be used here The most common one is equality of per hectare water deliveries In the present case the implicit standard is equality of water supply utility in producing a crop Thus if two areas received equal amounts of water on a per hectare basis but one had such high seepage and percolation losses that the crop failed to produce a harvest (and was therefore exempted from irrigation fee payment) the water supply to the combined area would be judged inequitable Because such radical differences in seepage and percolation rates are unlikely on puddled rice soils in the lowland Philippines and because the definition of benefitted area accommodates a wide range of acceptable yields these two definitions are largely indistinguishable at the current limits of data resolution
THE IMPACT OF IRRIGATION FINANCIAL SELF-RELIANCE ON IRRIGATION SYSTEM PERFORMANCE IN THE PHILIPPINES
Mark Svendsen1
INTRODUCTION
The disappointing performance of many developing country irrigation sectors has led to the development and application of a variety of measures intended to improve performance of the individual irrigation systems comprising the sector These measures have focussed on improvements to the physical facilities of the irrigation system and to a lesser extent on the institutions involved in system management Recently the latter has often included the creation or strengthening of farmers organizations for this purpose Often efforts have been targeted at a particular level of the system--at the tertiary or on-farm level or more recently at the main systemThese efforts have demonstrated some successes but in general have themselves been disappointing and system operating efficiencies generally remain far below potential Particularly disappointing have been the short effective life span of many of the improvements
Attempts to modify the larger environment in which irrigation systems operate presumably addressing more fundamental underlying causes of poor performance have also been made Policy-level efforts have been directed at changing the levels of funding for irrigationdevelopment and the mix of new system construction and rehabilitation at increasing levels of funding for operations and maintenance (OampM) at upping cost recovery levels and at adjusting the larger set of price incentives that influence farmers production decisions Generally absent have been proposals for policy-level interventions to transform the fundamental character of the public agencies managingirrigation systems into more efficient forms This chapter describes an exception to this rule It examines the impact of a purposive transformation of the main public irrigation agency in the Philippines on the physical performance of the irrigation systems it operatesThe elemental question being asked is leaving aside any proceduralchanges or internal economies achieved do irrigation systems performbetter in a privatized environment
IRRIGATION CHARGING SYSTEMS
Purposes of Charging Systems
A key characteristic of the institutional environment of an irrigation service provider is the mode by which its operations are
- 2 shy
financed This in turn is closely tied to the system employed to charge beneficiaries for irrigation services
As outlined by Carruthers and Clark (1981) a charging system for irrigation water has economic financial and social functions The economic function is to ensure that resources are efficiently used by charging beneficiaries a price equivalent to the value that societyplaces on the resource employed This function works in two waysOne is the effect that water rates have on the construction of future projects When beneficiaries are charged for development costs potential water users with foreknowledge of the rates to be chargedlobby government for the construction of additional irrigationcapacity to serve them2 only to the extent that it is profitable for them Government for its part must examine anticipated revenue flows to itself as a result of the project and evaluate its capacityto repay indebtedness incurred
The other economic effect of a water charging system in theory acts on resource use within and among existing commands sharing a water source When charges are set in accordance with well established principles of marginal cost pricing water is presumed to flow to locations and seasons where marginal returns per unit water are the highest 3
The financial function of a charging system is to cover the costs of the service provided--that is the delivery of irrigation water Costs involved include capital investment costs OampM expenses costs of revenue collection and the cost of negative externalities created Many mechanisms will serve this purpose though often there are unintended or undesired side effects For example benefit capture byartificially restraining output prices below market levels or through forced procurement of output through government purchasing points mayreduce output levels significantly by causing farmers to shift to other crops reduce input levels or abandon unprofitable tenancies
The third function the social one is a mixed bag of policies and actions used to promote income redistribution [and] economic stability or to develop backward areas and encourage investment bybeneficiaries (Carruthers and Clark 1981) This third function mitigates to a major extent the strict application of economic principles which underlies the first two purposes For example it was for social and geopolitical purposes that federally subsidized irrigation development was employed as a mechanism for settling the American West during the first half of the twentieth century With the West settled these policies became the target of increasing criticism on financial and economic efficiency grounds In recent years policies have shifted to reduce significantly the federal role in and subsidies for irrigation development significantly alteringthe previous balance among the three types of criteria
- 3 shy
It should be clear from this brief description of the purposes of irrigation charging systems that cost recovery from projectbeneficiaries is not an end in itself but a way of achieving specificefficiency and equity ends within the national economy Small and coshyauthors (ADB 1986) have summarized these ends in the following set of criteria for usefulness of a charging system They assert that a charging system has appropriate impacts if
1 it results in improved irrigation performance through
(a) more efficient operation and maintenance of irrigationfacilities and
(b) more efficient use of water by farmers and
2 it promotes other objectives of the government by
(a) leading to better irrigation investment decisions (b) easing the governments financial burden and (c) resulting in a more equitable distribution of income
This paper focusses on the impacts subtended by item number one above
Fees and Financial Autonomy
In addition to rules which specify which groups shall bear the costs of providing irrigation service and in what proportions the policies and practices which specify how collections are coursed through the government financial system and the relationship collections bear to irrigation agency income are critical determinants of the agencys operating environment A necessary condition for functionally linking the collection of irrigation service fees and effective irrigation performance is that the agency involved in providing the service be financially autonomous (Small et al 1989 Svendsen 1986) Financial autonomy is defined as a condition where (a) the irrigation agency must rely on user charges for a significantportion of the resources used for OampM and (b) the agency has expenditure control over the use of the funds generated from these charges (Small 1990)
When financial autonomy is present several incentive forces come into play which are otherwise absent First there is incentive for the agency to increase its income Increased income for the agencyimplies maintenance of jobs higher salaries incentive payments and bonuses greater staff mobility and new vehicles quarters and facilities for the staff If fees are levied on an area basis as is usual in developing countries this means that the irrigation agencyhas an a strong vested interest in expanding the area receivingadequate irrigation service increasing fee collection rates and increasing farm incomes
- 4 shy
Second there is incentive to reduce costs Reduced operating costs are often necessary to keep the agencys budget in the black a condition which is usually required to retain independent status In many ways this motive runs counter to the one mentioned above and it is the dynamic tension between these two that creates an efficient and responsive service provider
Working together these two motives generate a demand for better agency relations with cultivators greater accuracy in information collection and record keeping new technology to manage information more effectively better water control and greater farmer involvement in system maintenance and fee collection Thus incentives for greater efficiency in resource use in a context of financial autonomy act on the providers of irrigation service at least as powerfully as the payment of water charges affects consumers of those services Overall a relationship of mutual dependency is established between the two where the agency provides an essential service to farmers and farmers in turn provide the agency with the financial resources it needs to operate
THE PHILIPPINE EXPERIMENT
Irrigation in the Phjlippines
The irrigation sector in the Philippines is divided between two major surface water components--Communal Irrigation Systems (CIS) and National Irrigation Systems (NIS) Groundwater irrigation in the Philippines has always been of secondary importance Communal irrigation is an ancient practice in the Philippines and is the more important segment in terms of net area irrigated currently covering about 48 percent of the nations irrigated area of 1488000 hectares This component of the sector is made up of generally small systems managed by farmers which are usually constructed by them as well Since the mid-seventies the National Irrigation Administration (NIA)has been actively involved in innovative efforts to assist CIS without compromising farmer ownership and operation
The other major component of the irrigation sector NIS comprises about 42 percent of total irrigated area and consists of larger systems developed and operated by NIA Operation of these systems consumes by far the largest share of NIAs operating budgetand constitute its largest potential source of revenue NIAs efforts to deal with these two salient characteristics are described brieflyin the following section
Evolution of a New Institutional Form
In a major departure from regional norms the Philippine Irrigation Department was abolished in 1964 and a public corporation created in its stead During the first decade of its existence the
- 5 shy
National Irrigation Administration (NIA) operated in a way that differed little from regular government departments A major overhaul of its charter in 1974 however led to far-reaching changes in NIAs organizational values structure and operations At the core of these values was the presumption that to be successful NIA must be financially viable taking in more income than it spent By 1979 it had achieved the goal of overall financial viability and in 1981 the last operating subsidy paid out from the national treasury was received
Policy Shift The major thrust of the 1974 charter amendment was to allow NIA to retain all revenues generated by it including irrigationfee collections Heretofore all collections had been immediatelyturned over to the treasury in exchange for an annual appropriationfor operating expenses unrelated to NIAs self-generated revenues The annual appropriation that NIA received had always exceeded by a significant margin the collections that it remitted Accompanyingthis shift however was an agreement that all government operatingsubsidies to NIA were to be phased out over the ensuing five-year period At the end of that period NIAs operating budget would be completely self-financed
NIA Response NIA management responded to these charter changes with a four part strategy aimed at bringing its costs and revenues into balance The strategy comprised actions to
bull Devolve responsibility for certain operational maintenance and fee-collection tasks to farmers
bull Increase corporate revenues by raising fees improvingcollections and generating secondary income from ancillaryactivities
bull Reduce operating costs through a series of minor economies and through major cuts in the personnel budget and
bull Provide financial incentives for superior performance to outstanding field units and to individuals in them
Following earlier successes in organizing farmers in the communal irrigation sector in 1980 NIA began experimenting with waysto organize farmers in its larger systems into effective irrigatorsassociations which could assume responsibility for some canal maintenance water allocation and fee collection functions By 1986 the area under various forms of farmer management had reached about 100000 hectares out of a total of about 600000 hectares in the country Depending on the specific type of devolution reductions in NIAs staffing levels in a sample of affected systems ranged from 13 to 75 percent (Svendsen et al 1989)
Immediately following the 1974 charter amendment NIA obtained permission to increase its fees for irrigation service At the same time fees were indexed for inflation by denominating them in measures of paddy NIA was authorized to collect fees in paddy just as village
- 6 shy
moneylenders do Since that time the agency has made strenuous efforts to increase its fee collections The net effect has been to hold fee revenues per hectare constant in the face of a national rice support price that has steadily declined relative to more generalindices of inflation
At the same time NIA also took steps to reduce its operating expenses Although a number of minor measures of economy were mandated initially the fact that more than three-quarters of the operating budget was devoted to personnel costs meant that any real savings would require reductions in staffing levels Voluntaryreductions were carried out in the late 1970s and early 1980s resulting in a decrease in the number of staff per hectare and a reduction of the personnel share of the budget from 80 percent in 1976 to about 74 percent in 1986
With an eye on its bottom line NIA also instituted a system of performance grants for all field units and the individuals in them termed Viability Incentive Grants To facilitate this each largeirrigation system in the country was made a separate cost center to allow costs and revenues to be accounted for on a system-by-system basis This program provided that once a unit achieved a net excess of revenues over operating costs in a given year a fraction of the surplus would be shared among the units personnel Five of the 11 irrigation regions of the country were receiving these incentive payments by 1986 as were 53 of the 120 individual systems included within the 11 regions
Effects The financial results of these efforts are shown in Table 1 If subsidies are not considered NIA first achieved net profitabilityin 1979 and retained it through the end of the period studied except for a small deficit incurred in 1981 Subsidies were eliminated in 1982 (except for occasional small calamity grants following typhoons) Although revenues have declined in recent years due largely to decreases in interest earnings and construction management fees expenses have declined more rapidly resulting in a series of net positive balances
Achieving a financially viable position is an importantaccomplishment few other irrigation agencies in the developing world have been able to do this However there is some risk that such an achievement occurs at the expense of the quality of service provided to clients Some of the most interesting and important consequences of the new cost recovery policies therefore relate to the physical performance of the irrigation systems NIA operates It is here that the end objectives of the irrigation investments are realized and where the lives of the farmers who till system lands are affected The remainder of the paper will examine and attempt to quantify the impact of these policy changes on physical irrigation systemperformance
- 7 shy
IMPACTS ON SYSTEM PERFORMANCE
Study Methodology
Unfortunately the kinds of changes in hydrologic system outputand the impact on agricultural performance which might be expected to result from improved institutional performance are quite difficult to capture and quantify There are a number of reasons for this First there is the year to year variability of system performance caused byvariable rainfall which feeds rivers fills reservoirs and supplements irrigation water in supplying crop requirements Second there is the difficulty of defining just what performance is and specifying how to measure it Third and most important the regularly collected data from which indicators can be constructed are limited in type number of measuring points and period of record and are sometimes of doubtful reliability
These difficulties notWithstanding an attempt was made to determine the impact that changes in operating procedures staffinglevels and incentive programs had on system performance Because the effects that we are trying to assess resulted from changes that affected all of the systems under NIAs direct authority there are no control systems which can be used as standards We are forced therefore to rely on a comparison of values of selected performance indicators before and after the date of the major structural and procedural changes which is taken to be 1981
To accomplish this secondary data were assembled for 5 systems in Administrative Regions III and VI which had not undergonesignificant phYSical changes during the period of analysis Duringthis process several site visits were made by study team members Time series data collected include service area (SA) and benefitted area (BA) for both wet and dry seasons yields for wet and dry seasons monthly main canal discharge at the system headworks and monthly precipitation The general period of availability for this data is 1966-86 though for systems which began operation after 1966 the period of record is shorter and the records of some systemscontain miSSing values These five systems their 1986 service and benefitted areas and other descriptive data are shown in Table 2
The prinCipal problem with using a before and after approachrather than one that considers comparable systems with and without the innovation is that some of the measured difference in effects mayhave resulted from causes which are independent of the ones beingstudied These causes can be specific in which case they may be relatively easy to identify and accommodate or more general and diffuse and therefore more difficult to control for In the present case--that of changes in the performance of NIA irrigation systemsresulting from the major organizational changes in 1981 two principalexternal factors can be identified which might be expected to affect differences in measured levels of irrigation performance between the
- 8 shy
two time periods These are rainfall and level of use of other agricultural inputs Since our interest is in systems managerial responses to these changes and since there is no reason to believe that the relative magnitudes of the responses are dependent on the size of the system the five systems are treated as equal in the analysis That is changes in measured values relating to the smallest system are considered to be as important as changes in values for the largest with no area weighting applied
Since rainfall can substitute for irrigation water supplies and since it affects the supply of water available in rivers for irrigation it may exert some independent influence on various performance indicators To test the strength of the relationship for the period being analyzed simple correlations were run between rainfall and benefitted area for one system in each regionBenefitted area was used in this analysis because it is the variable deemed most likely to be influenced by year-to-year changes in rainfall Weather data from Cabanatuan City was used for the UPRIIS system which surrounds it and Iloilo City data was used for the nearby Aganan-Santa Barbara system For UPRIIS all of the R2 values for these correlations were less than 0005 suggesting that rainfall has almost no impact on area harvested in this large reservoir-based scheme For Aganan-Santa Barbara wet season rainfall was related to wet season BA (r2 = 016) and to BA during the following dry season (r2 = 024) Signs of the simple correlations were in the expected directions ie wet season rainfall increased BA during both the wet and the subsequent dry seasons These connections are understandable but weak
Another possibility is that there were longer-term differences in rainfall received in the two regions If this were the case a comparison of performance during two different time periods would have to take this difference into account Differences in average precipitation during the two periods were examined for the four stations used in the analysis (see Table 2) In no case were differences in seasonal or annual mean rainfall statisticallysignificant4
bull
Nevertheless in the regression approach adopted to analyze the data rainfall was included in each equation to control for its possible effect on the particular dependent variables being analyzed In doing this wet season rainfall was used in analyzing wet season performance indicators while annual rainfall was used in analyzingdry season data The rationale for this is that while dry season rainfall cannot possibly influence the wet season crop the dry season crop is affected by both the rainfall received directly and the rain falling during the preceding wet season through its effect on river discharge reservoir storage and antecedent soil moisture conditions
The level of agricultural production is also an often-used indicator of an irrigation systems performance Its major weakness
- 9 shy
is that a number of factors other than irrigation service such as labor inputs relative prices and fertilizer use influence it It is necessary therefore either to control for changes in the levels of these inputs or assume that they are constant across the two periods being compared In the present case the most important of these factors is the level of application of chemical fertilizer Because fertilizer use by farmers is responsive to the relative pricesof fertilizer and rice it also includes to some extent input and output price effects Since reliable data on fertilizer use for individual systems were not available estimates derived from FAO fertilizer and cultivated rice area data were used t~ control for the effect of changes in the use of this input over time This variable was included in any of the regression equations in which agriculturalproduction was used as the dependent variable Other factors such as labor use genetic potential of varieties sewn and pesticideapplications are assumed to be constant across the two periods
The analytic approach employed is to fit linear regressionequations to pooled data from the five systems covering a eleven-year period 1976 to 1986 A dummy variable is used to check the impact of pre and post 1981 periods on differences in the dependent variable after the effects of factors such as rainfall and nitrogen fertilizer use have been removed In addition because the dataset was created by pooling data from five different systems a set of 4 site dummies was included in the basic model to control for system-specific differences caused by variables which were not measured For some runs these were replaced with dummies that separated reservoir and non-reservoir systems though equations using the reservoir dummy were consistently inferior to those using the complete set of site dummies Several different dependent variables were created to index the quality of irrigation service and tested using this approachRegression results are given in Tables 3 and 5 and discussed below
Performance Indicators
A variety of indicators have been used in evaluating irrigationperformance in various contexts The selection of appropriate indicators depends on a number of factors including the purpose of the evaluation the audience for its results the way in which the boundaries of the irrigation system are defined and the kind and quality of data available to the evaluators The current analysis is designed to evaluate the impact of a set of management changes on system physical performance The audience for this analysis comprises top-level managers of the irrigation agency and policy-makers at higher government levels Boundary definition is an importantanalytic problem here as evident from the subsequent discussion relating to the choice of the appropriate area values to use in scaling system inputs and outputs This issue is also related to the data quality and availability problems which have already been mentioned
- 10 shy
Three fundamental indicators have been proposed for assessing the effectiveness of irrigation services to cultivators (Svendsen and Small 1989 Abernethy 1990) These are the adequacy of water supplies the equity of their distribution across the command area of the system and the timeliness of the supplies Computation of adequacy measures requires information on the total quantity of water delivered to the system over a season on a per hectare basis Equityand timeliness measures require information on the spatial and temporal distribution respectively of those supplies Where appropriate discharge information is not available proxies can be employed by making suitable assumptions Standards must be selected against which the magnitude of the indicators can be judged
The task in the present case however is somewhat different Here the need is to evaluate changes in selected variables between the pre-1981 and the post-1981 periods Hence the absolute values of variables selected are less important than their relative magnitudesand the statistical significance of the differences in magnitudes between the two periods A distinct limitation is imposed by the data series available for the five sample systems Since discharge and yield data are available only on a whole-system basis it is impossible to develop measures of equity and timeliness directly We will however extend our analysis to a discussion of equity byindirect inference Levine and Coward (1986) have argued that equityought to be considered as the paramount objective in managing largepublic irrigation systems They base their conclusion on an analysisof eight small community-managed systems and five larger public systems including UPRIIS in which equity appears to comprise the most important operational objective in the successful systems It may be appropriate therefore to give success in improving equity of distribution added weight in assessing overall performance
Area Estimates Because measures of system agricultural output and water supplied are typically reduced to a unit area basis before being used much depends on the area values which are used to standardize them Two different area measures are available The first is Service Area (SA) which is defined as the irrigable portion of the command area which is provided with physical facilities for water delivery This represents the area which could conceivably be irrigated in a given season if water supply were not constraining This value may change somewhat from year to year in response to urban encroachment on irrigated command minor remodeling and repair and refinements in area estimates In the present case though the systems selected for analysis were chosen to avoid those which had undergone more extensive rehabilitation or modification
The second measure is Benefitted Area (BA) which is the area billed for payment of irrigation service fees It is the irrigated area harvested which did not have yields so low that the farm was exempted from payment of fees in a given season This threshold value has been approximately 2 tons of paddy per hectare Benefitted area
- 11 shy
varies more than does SA particularly during the dry season when available water supply may seriously constrain the area which can be planted Its magnitude is a function of system managers actions in authorizing the amount of land to be planted in a given season farmers decisions regarding whether to plant or not and the combined ability of system managers and farmersirrigators subsequently to distribute water Both of these area measures will be used to standardize other variables for particular purposes as well as beingcombined to form a separate indicator by themselves
Adequacy The most direct measure of the adequacy of irrigation water supplies to the agricultural system is the quantity of water applied to the system command area on a per unit area basis relative to some standard In this case since our interest is in differences in water adequacy between two time periods and since the systems being assessed have been and continue to be almost entirely devoted to rice cultivation during both cropping seasons depth measures for the two periods may be compared directly assuming the seasonal cropdemand for water to be unchanged Although dry-footed crops can suffer yield losses from overapplication of irrigation water rice is largely insensitive to this effect In addition water can substitute for other inputs that the farmer would otherwise have to provide such as weed control and more careful (and costly) water management We assume therefore that other things being equal larger values of depth applied are better than smaller values in terms of meeting crop water demands and reduce the cost of cultivation At the same time high levels of water adequacy can affect the values of other performance measures--particularly equity
When the regression model is run for quantity of water diverted at the system headworks divided by BA hereafter termed depth we see that the period dummy is negative and significant at the 95 percent confidence level for both wet and dry seasons (see Table 3 equations 1 and 3) Since the overall explanatory power of the wet season model is very weak however we will focus on the dry season in interpreting this result which indicates that after adjusting for rainfall differences significantly less water was delivered to the command per unit of benefitted area following 1981 than before This indicates based on the criteria outlined above that performance in terms of water adequacy deteriorated following financial selfshysufficiency We need to examine this conclusion more carefullyhowever
One difficulty is that the measured quantity of water diverted at the source is largely a function of the supply available in the river rather than of system management This is particularly true during the dry season and in non-reservoir systems Thus while the depth of water supplied to the system is a measure of the adequacy of the systems service it is to some extent beyond the control of the managing agency To better understand the factors behind this decline in water availability we look at simple unadjusted index values for
- 12 shy
several of the key variables Table 4 shows annual values of total volume of water delivered in each season SA and BA in both wet and dry seasons and shows the results of t-tests on the means of a set of indicators before and after 1981 Indicators are used rather than the actual values to weight each of the systems equally regardless of its size The table shows that both the average wet season benefitted area and the average discharge are significantlylower during the second period compared with the first For the dry season too the discharge index is lower after 1981 than before but this difference is not significant At the same time the dry season BA index rose slightly but again the change was not significantSince middotthere is not a clear pattern of relative movement of discharge and BA during the respective seasons no simple interpretation of these index value changes is possible What stands out is that both discharge and benefitted area declined across periods during the wet season while during the dry season there was no significant change in either indicator across the two periods It seems clear that the decline in water adequacy must be evaluated together with other measures of performance in drawing conclusions about the overall impact of the 1981 changes on the quality of system management
Another measured variable per hectare yield can be used as a proxy for water adequacy It has the advantage of partiallyreflecting the impacts of the dimensions of timeliness6 and equity7of distribution as well integrating all three effects into a combined impact on aggregate crop production Table 5 (equations 1 and 3)shows that the period dummy in the yield regressions has a positive sign in both seasons after controlling for nitrogen application and precipitation though the t-values are not significant Treatingyield adjusted in this way as a proxy for quality of irrigationservice leads to the conclusion that by this more comprehensive measure quality of service held constant across the two periods in the dry season Because of the large yield component accounted for byrainfall during the wet season no such judgement is possible for that season however
Equity As noted earlier no reliable data are available for subdivisions of the five sample systems making direct computation of equity measures impossible We can make some judgements about changesin the equity of water distribution however by examining changes in the ratio of two area measures given for each system SA and BA Since SA is the area which theoretically can be supplied with irrigation water by the system and BA s the area which actuallyreceives a quantity of water adequate to produce a remunerative crop the ratio of the two provides a measure of the percentage of the potential service area which was irrigated to a particular standard The larger this pErcentage the more equitable 8 is the distribution This of course assumes that the quantity of water available to the systems is constant across the two periods
- 13 shy
Since the condition of constancy of water supply is not generally satisfied a regression was run in which the total quantityof water diverted at the headworks of each system divided by the systems potential service area SA was included in the regression to control for changes in the water supply available to the system seeTable 3 equations 5 through 8 The average daily rainfall received directly on the system service area during the season was also included as an independent variable The regression was run separately for wet and dry seasons The sign and t-statistic of the period dummy should then tell us whether or not equity as reflected in the BASA ratio increased decreased or remained unchanged across the period divide
Both equations are reasonable good as indicated by the R2 values though the dry season equation is considerably better as would be expected For the wet season both the water delivery term and the rainfall term in equation 5 are of positive sign but are nonshysignificant at the 95 percent confidence level indicating that wet season irrigated area does not change appreciably in response to level of wet season rainfall or the available irrigation water supply The period dummy was negative but not significant indicating that equityof distribution as reflected in the BASA ratio was similar during the two periods
For the dry season the water delivery term in equation 7 is positive and strongly significant indicating a close relationship between the fraction of potential area actually irrigated and the water supply available at the headworks In addition however the period dummy is positive and significant suggesting that once the influence of water supply is removed the BASA ratio was significantly higher in the period following 1981 than it was before
This is an important finding for it reflects significantlyimproved performance in terms of a factor equity of water distribution that is under the control of the managing entity an entity which here comprises both NIA and irrigators associations Interpreted in these terms NIA and allied farmers associations were able to spread a given amount of water more widely across the potential command area of the five sample systems in the period after 1981 than before Moreover they did this in a way that did not decrease average system yields as discussed earlier In making this interpretation we are suggesting that there was some redistribution of water from better-watered areas to fringe areas which would otherwise not have received irrigation water and that this redistribution was a direct response to the change in NIA prioritiesand operating policies and rules occurring around 1981
It is difficult to prove the assertion that water was in fact redistributed with a resulting increase in directly-measured equity Without access to reliable discharge data broken out by systemsection and we can only assume in the absence of a plausible
- 14 shy
alternative explanation that it was such a redistribution that made the increase in the BASA ratio possible In a larger sense it is difficult to prove conclusively that any outcome in a before and after analysis was the result of a particular independent causative factor In this case we have tried to remove the influence of other potential causative factors where we could but the possibilityremains that some combination of unmeasured factors are responsiblefor the difference in the BASA ratio found We do note though that this type of response is exactly the type that would be expected to follow from an emphasis on increased farmer satisfaction and cooperation and increased fee revenues Because the fee schedule is tied to benefitted area the only ways NIA can increase its revenue from that source are to expand benefitted area and to increase collection efficiencies The former depends on redistributing a fixed supply of water over a larger portion of the command while the latter requires that farmers be satisfied with the irrigation service they are receiving and the commitment of the local irrigators association to assist in the task of collecting the amounts due The evidence while not conclusive is highly suggestive that this is exactly what has happened
Efficiency In addition to measures which reflect the levels of adequacy and equity of irrigation service available data allow the calculation of a measure of operating efficiency The term efficiencyusually denotes the relationship between inputs to a process and its outputs often expressed as a ratio The output measure employed here is aggregate system rice output and the input is quantity of irrigation water turned into the system Dividing the first by the second gives a measure of agricultural production per unit water--here termed specific yield This is a highly integrated measure that evaluates the combined efficiency of the irrigation and agricultural processes As such it is a function of the managerial and other inputs supplied both to the irrigation system and to the agricultural operation With respect to one important input to the irrigation system we do know that NIA per hectare field operating expenses were about 29 percent lower in real terms in the 1982-86 period comparedto the 1976-1981 period although this drop may have been partlyoffset by increases in farmer-supplied labor inputs Other things being equal one would thus expect to find a decline in output efficiency
The regression analysis shows positive signs for the period terms in both wet and dry season equations (see Table 5) In the case of the wet season the period dummy in equation 5 is significant but the overall explanatory power of the model is quite low For the dry season (equation 7) the coefficient is positive but non-significant This means that after taking rainfall and fertilizer use into account data do not indicate a lowering of specific yield in the wake of funding reductions and the strong emphasiS on financial viabilitybeginning in 1981 This result provides evidence that the efficiency of the overall irrigation deliveryagricultural production process
- 15 shy
relative to the system water input did not falloff as a result of the changes implemented at least over the short run
Impact magnitude
The preceding analysis has shown us that some indicators of irrigation performance changed significantly following the managerial changes of 1981 while others did not However it has not given us a sense of the size of the changes which occurred To determine the magnitude of these changes the regression model is used to predict the response of the composite system to the managerial changes given a common set of -input and environmental factors To do this averagevalues of the independent variables from the entire eleven-year period 1976 to 1986 are put into the model together with the previously determined coefficients to generate predicted average values of the various dependent variables used in the earlier analysis with and without the period dummy This procedure produces a pair of estimates for each dependent variable under the same conditions--one in which the system responds as it did after the managerial changes were implemented and one in which it responds as it did prior to their introduction The differences between these two values thus indicate the magnitude of the changes occurring in the various indicators of performance discussed above
The results of this exercise are shown in Table 6 The table shows that water availability decreased by about 13 percent in both wet and dry seasons when the period dummy was included and while the coefficients responsible were significant in the earlier analysisthis difference cannot be easily connected with levels of system management as discussed earlier With respect to rice output per hectare although the coefficients were not very significant it is interesting to note that yield increases by 163 kilograms per hectare for the wet season and by 101 kilogram for the dry when the period dummy is included in spite of the reduced water supply available Keep in mind that the predicted yield values have already been adjusted for differences in nitrogen fertilizer use and rainfall This suggests that timeliness and equity of distribution of water supply to farmers may have increased following the changes contributing to the higher predicted yields
Examining the impact of increased equity of distribution bylooking at the ratio of benefitted area to service area we recall that the change was positive and significant for the dry season and negative and not significant for the wet Table 6 shows that the dry season BASA ratio increases by 7 percentage points when the dummy is included a 131 percent increase Other things being equal this should result in a 131 percent increase in system output due to the expansion of ared benefitted This is a major impact on production
- 16 shy
CONCLUSIONS
The Philippine experiment to transform the national irrigation agency into an enterprise has undoubtedly been successful in reducing system operating expenses bringing revenues and costs into line and eliminating the recurrent cost burden imposed by large-scaleirrigation systems on the national budget Evidence presented in this paper indicates that in the process equity of water distribution across systems has also improved In the 5 years following the cessation of operating subsidies from the government an index of equity of distribution improved by about 13 percent At the same time per hectare yields adjusted for rainfall and nitrogen application held constant
There is a strong logical connection between the achievement of financial viability and improved equity of water distribution across the command Because increasing irrigation fees is a politicaldecisionlying largely beyond NIAs control expanding the area which can be billed for service is one of the few revenue increasing measures available to the irrigation agency which does not involve major additional investment In the face of constant or shrinkingwater supplies this is achieved only by redistributing water from areas receiving excessive supplies usually near the head ends of canals and laterals to areas receiving no supplies or inadequatesupplies often located near the tails of canals Although data are not available which would allow the direct examination of this hypotheses the two outcomes are logically consistent with each other
Data also show that per hectare water deliveries declined significantly in the five sample systems after 1981 even thoughrainfall did not differ appreciably between the two periods This decline averaged about 13 percent for both wet and dry seasons and is interpreted as a decline in water availability in the supplying rivers rather than a conscious reduction in withdrawals by system managers Such declines could result from changes in watershed runoff characteristics as caused by deforestation or from increased upstream abstractions from supplying rivers
Improved water distribution tends to increase the area served system agricultural output and NIA service fee revenue Reduced water supplies to the system tend to reduce these things Specificyield defined as system paddy output per unit water held roughly constant across the two periods indicating that the two effects mayhave offset each other
After adjusting for rainfall and nitrogen application perhectare yields increased only marginally in the post-1981 period Area served on the other hand increased by about 13 percent after adjusting for water supply availability indicating that the area benefitted by irrigation in the sample systems increased by about the same percentage Even if yields on this additional area are less than
- 17 shy
average yields for the system this still represents a sizeable increase in system agricultural output as a result of the change in management structure the increase coming not from higher yields but from expanded area under irrigation
The evidence assembled here suggests that there are significantfinancial and economic benefits to be had from changes in the basic character of irrigation managing agencies which make them more responsive to their clientele and which impose rational internal financial discipline on the agency The analysis suggests a number of additional questions however One relates to the longer-term impacts of the structural management changes The improvements in water distribution described here are relatively short-term events occurring during the first 5 years of the new management mode Critics have suggested the danger of underinvestment in systemmaintenance over the longer run accompanied by declining yields and benefitted areas and eventual system collapse This possibility needs to be closely monitored A second concern relates to the apparent decline in water supply to these 5 geographically dispersed systems The nature and causes of this decline need to be explored further since if widespread and secular it may represent a serious threat to the stability of Philippine rice production Whether stemming from poor forest management practices or deficient regulation and allocation of surface water resources or other unidentified factors it is an issue that deserves serious and urgent consideration
A third risk is that the incentive structure set up by NIA to guide and stimulate the performance of field units overemphasizes revenue generation at the expense of irrigation service provision to farmers The evidence presented here supports the view that these two objectives are mutually reinforcing under policies and conditions which have been established in the Philippines More detailed crossshysectional studies based on primary flow measurement data would add confidence to this conclusion and help to specify the conditions under which this effect occurs This could be extremely important in transferring the results of the Philippine experiment to other countries
A final risk is that outside intervention well meaning or otherwise will destroy the basis of NIAs financial autonomy or will impose external pressures or constraints on NIAs decision-making that will subvert the management practices which have been so painstakinglydeveloped and implemented Among these are calls for NIA to be subsumed again within the government department structure in the interests of better coordination with agriculture attempts byexternal financing agencies to arbitrarily increase NIAs expenditures on OampM on the assumption that this will increase system agricultural output or intervention by Philippine legislative bodies to restore operating subsidies to NIA with attached strings leading back to legislators home districts Pressures such as these will cut short a process of experimentation and improvement that seems promising
- 18 shy
enough to date to warrant its continuation Having developed the capacity to establish targets and implement and manage change NIA is in a strong position to modify its objectives to better achieve larger social purposes established for it It is critical to recognize however that this must happen within the context of financing policies that mandate financial autonomy for NIA if the fundamental institutional commitment to manage is to be preserved
The author would like to thank Leslie Small and JeremyBerkoff for helpful comments on an earlier unpublishedversion of this paper and Charles Rogers for his careful and creative help with the analysis
- 19 shy
BIBLIOGRAPHY
Abernethy Charles L 1990 Indicators of the performance of irrigation water distribution systems International Irrigation Management Institute Colombo Sri Lanka Mimeo
Asian Development Bank 1986 Irrigation service fees Proceedingsof the Regional Seminar on Irrigation Service Fees Manila Asian Development Bank
Carruthers Ian and Colin Clark 1981 Economics of IrrigationLiverpool Liverpool University Press Third Edition
Levine G and EW Coward Jr 1986 Irrigation water distribution implications for design and operation AGREP Division WorkingPaper 125 vol 1 World Bank Agriculture and Rural Development Department
Small Les E 1989 User charges in irrigation potentials and limitations Irrigation and drainage vol 3 no 2125-142
Small Les 1990 Irrigation service fees in Asia IrrigationManagement Network 9013 London Overseas DevelopmentInstitute
Svendsen Mark and Les Small 1989 A framework for assessing irrigation system performance Paper prepared for the Symposium on Performance Evaluation 23 November International IrrigationManagement Institute Sri Lanka
Table I--National Irrigation Administration revenues and expenditures in constant prices 1976-86
Item 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986
(peso million 1972)
Revenues Irrigation fees collected 1273 1483 17 13 1831 2070 1668 1699 1893 1728 2129 2546 Other income 715 737 2420 5591 2631 5990 7783 6638 5699 4932 2934
Total direct revenue 1988 2220 4133 7422 4701 7658 9482 8531 7427 061 5480
Expenses in 1972 pricesTotal expenses 4825 5716 5039 6329 3821 77 55 6166 4749 4348 4259 4959
Excess (deficit) (2837(3496) (906) 1093 877 (097) 3316 3782 3079 2802 521 N 0
Subsidies Government operation and
maintenance subsidies 2521 2741 2799 1817 1398 633 0 0 0 0 0 Calamity fund payments 548 0 0 0 0 0 0 0 119 0 142
Total subsidy 3069 2741 2799 1817 1398 633 0 0 119 0 142
Total excess (deficit) 231 (754) 1893 2910 2275 536 3316 3782 3198 2802 663
Source IFPRI analysis of NIA data
- 21 shy
Table 2--Descriptive characteristics of selected MIA systeasa
------- Region III -------- ----- Region VI ----shyUPRIIS Angatii Sto Sibalom- Aganan-
Haasim Thomasc San Jose Sta Barbara
Average service area (hal 102272 31462 3522 5282 8703 Average irrigated area (hal
Wet season 83768 23454 3007 4410 8300 Dry season
Average benefited area Wet season
(hal 64587
77 605
27639
22908
1 781
3007
2801
4369
2770
7698 Dry season
Average rainfall Wet season
(mml d 62478
1 685 5
27396
8576
1 781
3051 0
2769
24731
2997
20001 Dry season 756 333 322 2828 3025
Average discharge (Llsec) Wet season 46501 14792 1692 2353 4984 Dry season 78091 22812 2014 1276 2315
Average water delivery (mmday) Wet season 522 548 487 462 571 Dry season
Average yield (mtha) 1089 715 995 398 686
Wet season 345 419 322 395 435 Dry season
Avg yield per unit water 34 03
(kgm ) 451 412 399 426
Wet season 0373 0440 0373 0538 0443 Dry season 0248 0400 0279 0690 0428
t-statistic difference in mean rainfall 1978-81 1982-86e
Wet season 0432 0713 -0567 1169 1169 Dry season 0519 -0230 -0523 1187 1187 Annual 0460 0707 -0686 1445 1445
~ Summary numbers are averages for the period 1982-1986 except as noted Water delivery discharge and yield per unit discharge are 4-year averages 1982-1985
c Water delivery discharge and yield per unit discharge are 4-year averages d 1983-1986
For Angat 5 years are 1981-85 For St Thomas 1979-83 For Sibalom 1971-75 e No significant differences at 95 confidence
- 22 shy
Table 3--Results of pooled regressions water delivery indicators
Dependent variable Independent variable - Water Delivery in mmBenefited Ha - --- BenefitService Area Ratio --shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season-shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 6799 (998)
6189 (1019)
13102 (1093)
2411 (1 54)
0743 ( 1186)
0800 (1165)
0129 (131 )
0258 (329)
System 2 indicator 0173 (029)
-3743 (-363)
-0063 (-219)
0200 (505)
System 3 indicator 0698 (085)
0163 (012)
0085 (208)
-0046 (-097)
System 4 indicator -0514 (-l 07)
-7011 (-803)
0080 (340)
0169 (l87)
System 5 indicator 0177 (033)
-5049 (-538)
0174 (603)
0051 (058)
Reservoir indicator 0327 (078)
5703 (543)
-0126 (-510)
0156 (261)
Period indicatora -0808 (-240)
-0771 (-229)
-1214 (-209)
-1309 (-1 62)
-0036 (-192)
-0025 (-112)
0071 (239)
0030 (077)
Water delivery mmb
Wet season
Dry season
Precipitation in mm Wet season
Annual
Average daily rainfall Wet season
Dry season
-46E-4 (-125)
-1 9E-4 (-082)
-43E-4 (-077)
20E-3 (385)
0015 (1 59)
-0001 ( -004)
0023 (215)
0001 (021)
0070 (613)
-0017 (-058)
0046 (371)
0010 (033)
R2 (adj) number of observations degrees of freedom
0156 43 36
0133 43 39
0711 41 34
0434 41 37
0693 42 34
0556 42 37
0865 40 32
0730 40 35
Coefficient (t statistic) indicates significance at 95 confidence ~ 1982-1986 1
per ha of service area
Ta
ble
--I
nd
ices
of
se
v
ice a
rea
ib
en
ef
ited
a
rea
a
nd
a
vera
ge sea
so
na
l d
isch
arg
e
Ave
rage
A
vera
ge
tshy19
77
1978
19
79
1980
19
81
1982
19
83
1984
19
85
1986
19
77-8
1 19
82-8
6 S
tati
stic
a
(ind
ex
aver
age
1983
-198
5 =
100)
Ser
vice
are
a UP
RI IS
in
dex
882
91
9
920
91
5
951
95
1
100
0 10
00
100
0 10
00
917
99
0
bull5
53
Ang
at-M
aasi
m R
95
9
994
99
7
996
99
6
996
10
00
100
0 10
00
100
0 98
8
999
1
63
Sto
To
mas
10
63
106
3 10
1 9
10
29
103
0 99
8
100
0 10
00
100
0 11
10
104
1 10
22
-08
9
Siba
lom
-San
Jos
e 95
0
872
94
2
933
94
2
942
10
28
102
8 94
4
101
7
928
99
2
292
A
gana
n-St
a B
arba
ra
108
3 10
91
106
0 10
30
961
10
05
996
10
08
996
99
6
104
5 10
00
-21
1 A
vera
ge
987
98
8
981
98
1
916
97
8
100
5 10
07
988
10
25
984
10
01
123
Wet
seas
on b
enef
ited
are
a in
dex
UPR
IIS
110
0 10
07
114
4 10
55
113
8 11
74
951
10
71
971
ll
58
10
89
106
7 -0
47
Ang
at-M
aasi
m R
97
9
974
92
0
983
10
28
100
8 99
7
102
1 98
2
931
97
7
988
0
54
Sto
To
mas
11
63
115
9 11
23
107
5 10
80
103
9 98
1
978
10
41
103
9 11
20
101
6 -4
88
Siba
lom
-San
Jos
e 11
35
103
8 10
11
931
93
4
906
10
07
975
10
1S
998
10
11
981
-0
80
Aga
nan-
Sta
Bar
bara
10
85
110
5 10
74
106
8 10
01
104
2 99
8
101
5
987
61
0
106
7 93
0
-1 8
4
Ave
rage
10
92
105
6 10
54
102
4 10
36
103
4 98
7
101
3 10
00
947
10
53
996
-2
32
Dry
sea
Son
bene
fite
d ar
ea
inde
x U
PRIIS
14
04
155
0 15
S0
155
8 16
17
128
0 57
2
114
S 15
74
152
3 12
38
-09
4 A
ngat
-Maa
sim
R
903
93
0
103
2 10
61
104
2 10
69
988
99
2
102
1 99
6
993
10
13
061
S
to
Tom
as
105
7 12
27
122
5
961
99
9
115
9 10
1 7
91
0
107
3 12
1 2
10
94
107
4 -0
28
Si
balo
m-S
an J
ose
Aga
nan-
Sta
Bar
bara
66
5
95S
62
6
632
67
4
111
4
501
10
S1
412
11
51
766
93
3
856
94
3
107
0 94
8
107
4 11
09
111
6
158
6 58
8
987
97
6
110
4 5
35
083
N
w
Ave
rage
89
6
964
11
19
103
7 10
44
110
9 10
1 7
89
8
108
5 12
97
101
7 10
81
082
Wet
seas
on d
isch
arge
in
dex
UPR
IIS
132
9 72
7
142
5 11
88
120
4 98
8
105
8 10
63
879
96
4
117
5 99
1
-16
5 A
ngat
-Maa
sim
R
129
7 13
52
134
5 12
58
127
0 11
70
120
3 62
7
131
3 10
68
-18
9
Sto
To
mas
14
71
155
0 14
1 6
10
40
725
12
35
112
3 14
79
103
1 -4
48
Siba
lom
-San
Jos
e 96
8
733
11
1 5
92
2
907
47
1
102
3 85
7
ll2
0
422
92
9
779
-1
09
Aga
nan-
Sta
Bar
bara
87
9
863
68
1
925
96
8
110
7 70
5
871
87
7
00
9
Ave
rage
11
49
105
7 13
61
115
0 10
58
853
10
43
963
99
4
SO3
11
55
940
-2
85
Dry
sea
son
disc
harg
e in
dex
UPR
IIS
425
13
09
153
3 14
28
180
6 14
S0
125
8 56
3
117
9 14
66
130
0 11
89
-04
3
Ang
at-M
aasi
m R
12
26
161
6 15
30
139
2 12
85
100
8 10
33
958
14
41
107
1 -3
80
S
to
Tom
as
116
4 14
82
155
7 79
9
971
12
30
126
5 14
01
106
6 -2
44
Si
balo
m-S
an J
ose
486
71
5
782
62
9
789
64
5
116
9 11
86
801
65
3
918
2
36
Aga
nan-
Sta
Bar
bara
A
vera
ge
425
10
46
133
6 62
0
118
4 81
5
116
1 94
7
112
5 89
7
922
12
20
991
88
2
108
7 12
36
119
2 71
8
114
0 10
37
105
5 3
23
-07
5
a bull
indi
cate
s si
gnif
ican
ce a
t 95
per
cent
con
fide
nce
- 24 shy
Table 5--Results of pooled regressions yield indicators
------------------------- Dependent variables -------------------------shyIndependent variable ---------- Yield MTHa ----------- -- Specific Yield 1000 KgM
3 H20 shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season -shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 1967 (286)
3747 (580)
1991 (293)
2800 (460 )
0084 (041)
0441 (229)
0049 (023 )
0876 (343)
System 2 indicator 0400 (155)
0525 (226)
0057 (085)
0115 (1 95)
System 3 indicator 0466 (139 )
-0029 (-010)
-0016 (-018)
0006 (008)
System 4 indicator 0980 (470)
0090 (049)
0176 (333)
0391 (827)
System 5 indicator 1 221 (570)
0256 ( 138)
0104 (174)
0224 (440)
Reservoir indicator -1033 (-578)
-0063 (-039)
-0146 (-310)
-0326 (-569)
Period indicatora
Precipitation in mm Wet season
Annual
0163 (1 03)
-29E-4 (-182)
0231 (1 42)
-58E-4 (-606)
0161 (118)
92E-5 (072)
0199 043 )
-64pound-5 (-084)
0082 (200)
57E-6 (013)
0092 (223)
-61E-5 (-232)
0032 (093)
29E-7 (001)
0052 (112)
-12pound-4 (-414)
Nitrogen applicationb 0025 (254)
0021 (211)
0025 (274)
0020 (227)
0003 (1 05)
0002 (043)
0003 (101)
-0002 (-052)
R2 (adj) 0553 0516 0251 0198 0309 0262 0713 0445 number of observations 50 50 49 49 42 42 40 40 degrees of freedom 42 45 41 44 34 37 32 35
CoeffiCient (t statistic) indicates significance at 95 confidence ~ 1982-1986 = 1
kgha
- 25 shy
Table 6--Suggmry of estiates perforance indicators
Variable Predicted Predicted Period Dependent Variable
Description (units)
Period Coefficient
1976-86 IndicatorO
1976-86 Indicatorl
Effecta (X)
BenefittedService area ratio
Wet season ratio -0036 0853 0817 -425 Dry Season 0071 0539 0610 1313
Delivery Wet Dry
season season
nm water per benefitted ha
-0808 -1 214
6003 9083
5195 7869
-1346 -1337
Rice Yield Wet season metric tons 0163 3578 3741 456 Dry season per hectare 0161 3905 4066 412
Specific Yield Wet season kg rice per m3 0082 0348 0430 2353 Dry season water de 1 ivered 0032 0378 0409 837
a percentage changes relative to period=O values
- 26 shy
ENDNOTES
IResearch Fellow International Food Policy Research Institute and Senior Irrigation Specialist International Irrigation Management Institute
2Presumable farmers also base private investment decisions on such foreknowledge of costs and commonly do so in the case of small tubewells In addition user groups may invest in small communal schemes independently after examining projected costs and benefits Most commonly however new irrigation capacity is created as a result of public investment decisions made by government and thus the political process comes into play
3This principle applies only when water charges are levied volumetrically--an exceedingly rare situation in the developing worldshy-though the argument is frequently invoked as though it applies to all allocative situations
4For Cabanatuan City the longer period of record available ie 1904 to 1986 was divided into two equal periods and tested for difference in means to test the observation by farmers that rainfall today is lower than it used to be Mean precipitation after 1947 was actually about 3 percent higher than before but the difference was not significant (t = 07)
5The procedure used was to take time series data on total nitrogenfertilizer consumption in the Philippines multiply it by the ratio of N fertilizer used on rice to total N consumption and divide the product by the rice harvested area to obtain an estimate of N use perhectare Values of the ratio of N use on rice to total N use were taken from IRRI (1988 150) with interpolation used to fill in gaps in the series
6This assumes that timeliness is evaluated against a standard based on the physiological demand for water generated by the rice crop and the varying sensitivity of the crop to stress at different growth stages In such a case a poor timeliness rating for irrigationservice will result in reductions in yield
7Because rice yields are insensitive to water applications that exceed PET values a redistribution of water from water excess to water short portions of the command area will usually result in higher aggregate output and higher average yields Exceptional scenarios can be constructed in which this general rule would be violated but such scenarios are unlikely to play out in practice
- 27 shy
aThere are a variety of standards for judging equity that could be used here The most common one is equality of per hectare water deliveries In the present case the implicit standard is equality of water supply utility in producing a crop Thus if two areas received equal amounts of water on a per hectare basis but one had such high seepage and percolation losses that the crop failed to produce a harvest (and was therefore exempted from irrigation fee payment) the water supply to the combined area would be judged inequitable Because such radical differences in seepage and percolation rates are unlikely on puddled rice soils in the lowland Philippines and because the definition of benefitted area accommodates a wide range of acceptable yields these two definitions are largely indistinguishable at the current limits of data resolution
- 2 shy
financed This in turn is closely tied to the system employed to charge beneficiaries for irrigation services
As outlined by Carruthers and Clark (1981) a charging system for irrigation water has economic financial and social functions The economic function is to ensure that resources are efficiently used by charging beneficiaries a price equivalent to the value that societyplaces on the resource employed This function works in two waysOne is the effect that water rates have on the construction of future projects When beneficiaries are charged for development costs potential water users with foreknowledge of the rates to be chargedlobby government for the construction of additional irrigationcapacity to serve them2 only to the extent that it is profitable for them Government for its part must examine anticipated revenue flows to itself as a result of the project and evaluate its capacityto repay indebtedness incurred
The other economic effect of a water charging system in theory acts on resource use within and among existing commands sharing a water source When charges are set in accordance with well established principles of marginal cost pricing water is presumed to flow to locations and seasons where marginal returns per unit water are the highest 3
The financial function of a charging system is to cover the costs of the service provided--that is the delivery of irrigation water Costs involved include capital investment costs OampM expenses costs of revenue collection and the cost of negative externalities created Many mechanisms will serve this purpose though often there are unintended or undesired side effects For example benefit capture byartificially restraining output prices below market levels or through forced procurement of output through government purchasing points mayreduce output levels significantly by causing farmers to shift to other crops reduce input levels or abandon unprofitable tenancies
The third function the social one is a mixed bag of policies and actions used to promote income redistribution [and] economic stability or to develop backward areas and encourage investment bybeneficiaries (Carruthers and Clark 1981) This third function mitigates to a major extent the strict application of economic principles which underlies the first two purposes For example it was for social and geopolitical purposes that federally subsidized irrigation development was employed as a mechanism for settling the American West during the first half of the twentieth century With the West settled these policies became the target of increasing criticism on financial and economic efficiency grounds In recent years policies have shifted to reduce significantly the federal role in and subsidies for irrigation development significantly alteringthe previous balance among the three types of criteria
- 3 shy
It should be clear from this brief description of the purposes of irrigation charging systems that cost recovery from projectbeneficiaries is not an end in itself but a way of achieving specificefficiency and equity ends within the national economy Small and coshyauthors (ADB 1986) have summarized these ends in the following set of criteria for usefulness of a charging system They assert that a charging system has appropriate impacts if
1 it results in improved irrigation performance through
(a) more efficient operation and maintenance of irrigationfacilities and
(b) more efficient use of water by farmers and
2 it promotes other objectives of the government by
(a) leading to better irrigation investment decisions (b) easing the governments financial burden and (c) resulting in a more equitable distribution of income
This paper focusses on the impacts subtended by item number one above
Fees and Financial Autonomy
In addition to rules which specify which groups shall bear the costs of providing irrigation service and in what proportions the policies and practices which specify how collections are coursed through the government financial system and the relationship collections bear to irrigation agency income are critical determinants of the agencys operating environment A necessary condition for functionally linking the collection of irrigation service fees and effective irrigation performance is that the agency involved in providing the service be financially autonomous (Small et al 1989 Svendsen 1986) Financial autonomy is defined as a condition where (a) the irrigation agency must rely on user charges for a significantportion of the resources used for OampM and (b) the agency has expenditure control over the use of the funds generated from these charges (Small 1990)
When financial autonomy is present several incentive forces come into play which are otherwise absent First there is incentive for the agency to increase its income Increased income for the agencyimplies maintenance of jobs higher salaries incentive payments and bonuses greater staff mobility and new vehicles quarters and facilities for the staff If fees are levied on an area basis as is usual in developing countries this means that the irrigation agencyhas an a strong vested interest in expanding the area receivingadequate irrigation service increasing fee collection rates and increasing farm incomes
- 4 shy
Second there is incentive to reduce costs Reduced operating costs are often necessary to keep the agencys budget in the black a condition which is usually required to retain independent status In many ways this motive runs counter to the one mentioned above and it is the dynamic tension between these two that creates an efficient and responsive service provider
Working together these two motives generate a demand for better agency relations with cultivators greater accuracy in information collection and record keeping new technology to manage information more effectively better water control and greater farmer involvement in system maintenance and fee collection Thus incentives for greater efficiency in resource use in a context of financial autonomy act on the providers of irrigation service at least as powerfully as the payment of water charges affects consumers of those services Overall a relationship of mutual dependency is established between the two where the agency provides an essential service to farmers and farmers in turn provide the agency with the financial resources it needs to operate
THE PHILIPPINE EXPERIMENT
Irrigation in the Phjlippines
The irrigation sector in the Philippines is divided between two major surface water components--Communal Irrigation Systems (CIS) and National Irrigation Systems (NIS) Groundwater irrigation in the Philippines has always been of secondary importance Communal irrigation is an ancient practice in the Philippines and is the more important segment in terms of net area irrigated currently covering about 48 percent of the nations irrigated area of 1488000 hectares This component of the sector is made up of generally small systems managed by farmers which are usually constructed by them as well Since the mid-seventies the National Irrigation Administration (NIA)has been actively involved in innovative efforts to assist CIS without compromising farmer ownership and operation
The other major component of the irrigation sector NIS comprises about 42 percent of total irrigated area and consists of larger systems developed and operated by NIA Operation of these systems consumes by far the largest share of NIAs operating budgetand constitute its largest potential source of revenue NIAs efforts to deal with these two salient characteristics are described brieflyin the following section
Evolution of a New Institutional Form
In a major departure from regional norms the Philippine Irrigation Department was abolished in 1964 and a public corporation created in its stead During the first decade of its existence the
- 5 shy
National Irrigation Administration (NIA) operated in a way that differed little from regular government departments A major overhaul of its charter in 1974 however led to far-reaching changes in NIAs organizational values structure and operations At the core of these values was the presumption that to be successful NIA must be financially viable taking in more income than it spent By 1979 it had achieved the goal of overall financial viability and in 1981 the last operating subsidy paid out from the national treasury was received
Policy Shift The major thrust of the 1974 charter amendment was to allow NIA to retain all revenues generated by it including irrigationfee collections Heretofore all collections had been immediatelyturned over to the treasury in exchange for an annual appropriationfor operating expenses unrelated to NIAs self-generated revenues The annual appropriation that NIA received had always exceeded by a significant margin the collections that it remitted Accompanyingthis shift however was an agreement that all government operatingsubsidies to NIA were to be phased out over the ensuing five-year period At the end of that period NIAs operating budget would be completely self-financed
NIA Response NIA management responded to these charter changes with a four part strategy aimed at bringing its costs and revenues into balance The strategy comprised actions to
bull Devolve responsibility for certain operational maintenance and fee-collection tasks to farmers
bull Increase corporate revenues by raising fees improvingcollections and generating secondary income from ancillaryactivities
bull Reduce operating costs through a series of minor economies and through major cuts in the personnel budget and
bull Provide financial incentives for superior performance to outstanding field units and to individuals in them
Following earlier successes in organizing farmers in the communal irrigation sector in 1980 NIA began experimenting with waysto organize farmers in its larger systems into effective irrigatorsassociations which could assume responsibility for some canal maintenance water allocation and fee collection functions By 1986 the area under various forms of farmer management had reached about 100000 hectares out of a total of about 600000 hectares in the country Depending on the specific type of devolution reductions in NIAs staffing levels in a sample of affected systems ranged from 13 to 75 percent (Svendsen et al 1989)
Immediately following the 1974 charter amendment NIA obtained permission to increase its fees for irrigation service At the same time fees were indexed for inflation by denominating them in measures of paddy NIA was authorized to collect fees in paddy just as village
- 6 shy
moneylenders do Since that time the agency has made strenuous efforts to increase its fee collections The net effect has been to hold fee revenues per hectare constant in the face of a national rice support price that has steadily declined relative to more generalindices of inflation
At the same time NIA also took steps to reduce its operating expenses Although a number of minor measures of economy were mandated initially the fact that more than three-quarters of the operating budget was devoted to personnel costs meant that any real savings would require reductions in staffing levels Voluntaryreductions were carried out in the late 1970s and early 1980s resulting in a decrease in the number of staff per hectare and a reduction of the personnel share of the budget from 80 percent in 1976 to about 74 percent in 1986
With an eye on its bottom line NIA also instituted a system of performance grants for all field units and the individuals in them termed Viability Incentive Grants To facilitate this each largeirrigation system in the country was made a separate cost center to allow costs and revenues to be accounted for on a system-by-system basis This program provided that once a unit achieved a net excess of revenues over operating costs in a given year a fraction of the surplus would be shared among the units personnel Five of the 11 irrigation regions of the country were receiving these incentive payments by 1986 as were 53 of the 120 individual systems included within the 11 regions
Effects The financial results of these efforts are shown in Table 1 If subsidies are not considered NIA first achieved net profitabilityin 1979 and retained it through the end of the period studied except for a small deficit incurred in 1981 Subsidies were eliminated in 1982 (except for occasional small calamity grants following typhoons) Although revenues have declined in recent years due largely to decreases in interest earnings and construction management fees expenses have declined more rapidly resulting in a series of net positive balances
Achieving a financially viable position is an importantaccomplishment few other irrigation agencies in the developing world have been able to do this However there is some risk that such an achievement occurs at the expense of the quality of service provided to clients Some of the most interesting and important consequences of the new cost recovery policies therefore relate to the physical performance of the irrigation systems NIA operates It is here that the end objectives of the irrigation investments are realized and where the lives of the farmers who till system lands are affected The remainder of the paper will examine and attempt to quantify the impact of these policy changes on physical irrigation systemperformance
- 7 shy
IMPACTS ON SYSTEM PERFORMANCE
Study Methodology
Unfortunately the kinds of changes in hydrologic system outputand the impact on agricultural performance which might be expected to result from improved institutional performance are quite difficult to capture and quantify There are a number of reasons for this First there is the year to year variability of system performance caused byvariable rainfall which feeds rivers fills reservoirs and supplements irrigation water in supplying crop requirements Second there is the difficulty of defining just what performance is and specifying how to measure it Third and most important the regularly collected data from which indicators can be constructed are limited in type number of measuring points and period of record and are sometimes of doubtful reliability
These difficulties notWithstanding an attempt was made to determine the impact that changes in operating procedures staffinglevels and incentive programs had on system performance Because the effects that we are trying to assess resulted from changes that affected all of the systems under NIAs direct authority there are no control systems which can be used as standards We are forced therefore to rely on a comparison of values of selected performance indicators before and after the date of the major structural and procedural changes which is taken to be 1981
To accomplish this secondary data were assembled for 5 systems in Administrative Regions III and VI which had not undergonesignificant phYSical changes during the period of analysis Duringthis process several site visits were made by study team members Time series data collected include service area (SA) and benefitted area (BA) for both wet and dry seasons yields for wet and dry seasons monthly main canal discharge at the system headworks and monthly precipitation The general period of availability for this data is 1966-86 though for systems which began operation after 1966 the period of record is shorter and the records of some systemscontain miSSing values These five systems their 1986 service and benefitted areas and other descriptive data are shown in Table 2
The prinCipal problem with using a before and after approachrather than one that considers comparable systems with and without the innovation is that some of the measured difference in effects mayhave resulted from causes which are independent of the ones beingstudied These causes can be specific in which case they may be relatively easy to identify and accommodate or more general and diffuse and therefore more difficult to control for In the present case--that of changes in the performance of NIA irrigation systemsresulting from the major organizational changes in 1981 two principalexternal factors can be identified which might be expected to affect differences in measured levels of irrigation performance between the
- 8 shy
two time periods These are rainfall and level of use of other agricultural inputs Since our interest is in systems managerial responses to these changes and since there is no reason to believe that the relative magnitudes of the responses are dependent on the size of the system the five systems are treated as equal in the analysis That is changes in measured values relating to the smallest system are considered to be as important as changes in values for the largest with no area weighting applied
Since rainfall can substitute for irrigation water supplies and since it affects the supply of water available in rivers for irrigation it may exert some independent influence on various performance indicators To test the strength of the relationship for the period being analyzed simple correlations were run between rainfall and benefitted area for one system in each regionBenefitted area was used in this analysis because it is the variable deemed most likely to be influenced by year-to-year changes in rainfall Weather data from Cabanatuan City was used for the UPRIIS system which surrounds it and Iloilo City data was used for the nearby Aganan-Santa Barbara system For UPRIIS all of the R2 values for these correlations were less than 0005 suggesting that rainfall has almost no impact on area harvested in this large reservoir-based scheme For Aganan-Santa Barbara wet season rainfall was related to wet season BA (r2 = 016) and to BA during the following dry season (r2 = 024) Signs of the simple correlations were in the expected directions ie wet season rainfall increased BA during both the wet and the subsequent dry seasons These connections are understandable but weak
Another possibility is that there were longer-term differences in rainfall received in the two regions If this were the case a comparison of performance during two different time periods would have to take this difference into account Differences in average precipitation during the two periods were examined for the four stations used in the analysis (see Table 2) In no case were differences in seasonal or annual mean rainfall statisticallysignificant4
bull
Nevertheless in the regression approach adopted to analyze the data rainfall was included in each equation to control for its possible effect on the particular dependent variables being analyzed In doing this wet season rainfall was used in analyzing wet season performance indicators while annual rainfall was used in analyzingdry season data The rationale for this is that while dry season rainfall cannot possibly influence the wet season crop the dry season crop is affected by both the rainfall received directly and the rain falling during the preceding wet season through its effect on river discharge reservoir storage and antecedent soil moisture conditions
The level of agricultural production is also an often-used indicator of an irrigation systems performance Its major weakness
- 9 shy
is that a number of factors other than irrigation service such as labor inputs relative prices and fertilizer use influence it It is necessary therefore either to control for changes in the levels of these inputs or assume that they are constant across the two periods being compared In the present case the most important of these factors is the level of application of chemical fertilizer Because fertilizer use by farmers is responsive to the relative pricesof fertilizer and rice it also includes to some extent input and output price effects Since reliable data on fertilizer use for individual systems were not available estimates derived from FAO fertilizer and cultivated rice area data were used t~ control for the effect of changes in the use of this input over time This variable was included in any of the regression equations in which agriculturalproduction was used as the dependent variable Other factors such as labor use genetic potential of varieties sewn and pesticideapplications are assumed to be constant across the two periods
The analytic approach employed is to fit linear regressionequations to pooled data from the five systems covering a eleven-year period 1976 to 1986 A dummy variable is used to check the impact of pre and post 1981 periods on differences in the dependent variable after the effects of factors such as rainfall and nitrogen fertilizer use have been removed In addition because the dataset was created by pooling data from five different systems a set of 4 site dummies was included in the basic model to control for system-specific differences caused by variables which were not measured For some runs these were replaced with dummies that separated reservoir and non-reservoir systems though equations using the reservoir dummy were consistently inferior to those using the complete set of site dummies Several different dependent variables were created to index the quality of irrigation service and tested using this approachRegression results are given in Tables 3 and 5 and discussed below
Performance Indicators
A variety of indicators have been used in evaluating irrigationperformance in various contexts The selection of appropriate indicators depends on a number of factors including the purpose of the evaluation the audience for its results the way in which the boundaries of the irrigation system are defined and the kind and quality of data available to the evaluators The current analysis is designed to evaluate the impact of a set of management changes on system physical performance The audience for this analysis comprises top-level managers of the irrigation agency and policy-makers at higher government levels Boundary definition is an importantanalytic problem here as evident from the subsequent discussion relating to the choice of the appropriate area values to use in scaling system inputs and outputs This issue is also related to the data quality and availability problems which have already been mentioned
- 10 shy
Three fundamental indicators have been proposed for assessing the effectiveness of irrigation services to cultivators (Svendsen and Small 1989 Abernethy 1990) These are the adequacy of water supplies the equity of their distribution across the command area of the system and the timeliness of the supplies Computation of adequacy measures requires information on the total quantity of water delivered to the system over a season on a per hectare basis Equityand timeliness measures require information on the spatial and temporal distribution respectively of those supplies Where appropriate discharge information is not available proxies can be employed by making suitable assumptions Standards must be selected against which the magnitude of the indicators can be judged
The task in the present case however is somewhat different Here the need is to evaluate changes in selected variables between the pre-1981 and the post-1981 periods Hence the absolute values of variables selected are less important than their relative magnitudesand the statistical significance of the differences in magnitudes between the two periods A distinct limitation is imposed by the data series available for the five sample systems Since discharge and yield data are available only on a whole-system basis it is impossible to develop measures of equity and timeliness directly We will however extend our analysis to a discussion of equity byindirect inference Levine and Coward (1986) have argued that equityought to be considered as the paramount objective in managing largepublic irrigation systems They base their conclusion on an analysisof eight small community-managed systems and five larger public systems including UPRIIS in which equity appears to comprise the most important operational objective in the successful systems It may be appropriate therefore to give success in improving equity of distribution added weight in assessing overall performance
Area Estimates Because measures of system agricultural output and water supplied are typically reduced to a unit area basis before being used much depends on the area values which are used to standardize them Two different area measures are available The first is Service Area (SA) which is defined as the irrigable portion of the command area which is provided with physical facilities for water delivery This represents the area which could conceivably be irrigated in a given season if water supply were not constraining This value may change somewhat from year to year in response to urban encroachment on irrigated command minor remodeling and repair and refinements in area estimates In the present case though the systems selected for analysis were chosen to avoid those which had undergone more extensive rehabilitation or modification
The second measure is Benefitted Area (BA) which is the area billed for payment of irrigation service fees It is the irrigated area harvested which did not have yields so low that the farm was exempted from payment of fees in a given season This threshold value has been approximately 2 tons of paddy per hectare Benefitted area
- 11 shy
varies more than does SA particularly during the dry season when available water supply may seriously constrain the area which can be planted Its magnitude is a function of system managers actions in authorizing the amount of land to be planted in a given season farmers decisions regarding whether to plant or not and the combined ability of system managers and farmersirrigators subsequently to distribute water Both of these area measures will be used to standardize other variables for particular purposes as well as beingcombined to form a separate indicator by themselves
Adequacy The most direct measure of the adequacy of irrigation water supplies to the agricultural system is the quantity of water applied to the system command area on a per unit area basis relative to some standard In this case since our interest is in differences in water adequacy between two time periods and since the systems being assessed have been and continue to be almost entirely devoted to rice cultivation during both cropping seasons depth measures for the two periods may be compared directly assuming the seasonal cropdemand for water to be unchanged Although dry-footed crops can suffer yield losses from overapplication of irrigation water rice is largely insensitive to this effect In addition water can substitute for other inputs that the farmer would otherwise have to provide such as weed control and more careful (and costly) water management We assume therefore that other things being equal larger values of depth applied are better than smaller values in terms of meeting crop water demands and reduce the cost of cultivation At the same time high levels of water adequacy can affect the values of other performance measures--particularly equity
When the regression model is run for quantity of water diverted at the system headworks divided by BA hereafter termed depth we see that the period dummy is negative and significant at the 95 percent confidence level for both wet and dry seasons (see Table 3 equations 1 and 3) Since the overall explanatory power of the wet season model is very weak however we will focus on the dry season in interpreting this result which indicates that after adjusting for rainfall differences significantly less water was delivered to the command per unit of benefitted area following 1981 than before This indicates based on the criteria outlined above that performance in terms of water adequacy deteriorated following financial selfshysufficiency We need to examine this conclusion more carefullyhowever
One difficulty is that the measured quantity of water diverted at the source is largely a function of the supply available in the river rather than of system management This is particularly true during the dry season and in non-reservoir systems Thus while the depth of water supplied to the system is a measure of the adequacy of the systems service it is to some extent beyond the control of the managing agency To better understand the factors behind this decline in water availability we look at simple unadjusted index values for
- 12 shy
several of the key variables Table 4 shows annual values of total volume of water delivered in each season SA and BA in both wet and dry seasons and shows the results of t-tests on the means of a set of indicators before and after 1981 Indicators are used rather than the actual values to weight each of the systems equally regardless of its size The table shows that both the average wet season benefitted area and the average discharge are significantlylower during the second period compared with the first For the dry season too the discharge index is lower after 1981 than before but this difference is not significant At the same time the dry season BA index rose slightly but again the change was not significantSince middotthere is not a clear pattern of relative movement of discharge and BA during the respective seasons no simple interpretation of these index value changes is possible What stands out is that both discharge and benefitted area declined across periods during the wet season while during the dry season there was no significant change in either indicator across the two periods It seems clear that the decline in water adequacy must be evaluated together with other measures of performance in drawing conclusions about the overall impact of the 1981 changes on the quality of system management
Another measured variable per hectare yield can be used as a proxy for water adequacy It has the advantage of partiallyreflecting the impacts of the dimensions of timeliness6 and equity7of distribution as well integrating all three effects into a combined impact on aggregate crop production Table 5 (equations 1 and 3)shows that the period dummy in the yield regressions has a positive sign in both seasons after controlling for nitrogen application and precipitation though the t-values are not significant Treatingyield adjusted in this way as a proxy for quality of irrigationservice leads to the conclusion that by this more comprehensive measure quality of service held constant across the two periods in the dry season Because of the large yield component accounted for byrainfall during the wet season no such judgement is possible for that season however
Equity As noted earlier no reliable data are available for subdivisions of the five sample systems making direct computation of equity measures impossible We can make some judgements about changesin the equity of water distribution however by examining changes in the ratio of two area measures given for each system SA and BA Since SA is the area which theoretically can be supplied with irrigation water by the system and BA s the area which actuallyreceives a quantity of water adequate to produce a remunerative crop the ratio of the two provides a measure of the percentage of the potential service area which was irrigated to a particular standard The larger this pErcentage the more equitable 8 is the distribution This of course assumes that the quantity of water available to the systems is constant across the two periods
- 13 shy
Since the condition of constancy of water supply is not generally satisfied a regression was run in which the total quantityof water diverted at the headworks of each system divided by the systems potential service area SA was included in the regression to control for changes in the water supply available to the system seeTable 3 equations 5 through 8 The average daily rainfall received directly on the system service area during the season was also included as an independent variable The regression was run separately for wet and dry seasons The sign and t-statistic of the period dummy should then tell us whether or not equity as reflected in the BASA ratio increased decreased or remained unchanged across the period divide
Both equations are reasonable good as indicated by the R2 values though the dry season equation is considerably better as would be expected For the wet season both the water delivery term and the rainfall term in equation 5 are of positive sign but are nonshysignificant at the 95 percent confidence level indicating that wet season irrigated area does not change appreciably in response to level of wet season rainfall or the available irrigation water supply The period dummy was negative but not significant indicating that equityof distribution as reflected in the BASA ratio was similar during the two periods
For the dry season the water delivery term in equation 7 is positive and strongly significant indicating a close relationship between the fraction of potential area actually irrigated and the water supply available at the headworks In addition however the period dummy is positive and significant suggesting that once the influence of water supply is removed the BASA ratio was significantly higher in the period following 1981 than it was before
This is an important finding for it reflects significantlyimproved performance in terms of a factor equity of water distribution that is under the control of the managing entity an entity which here comprises both NIA and irrigators associations Interpreted in these terms NIA and allied farmers associations were able to spread a given amount of water more widely across the potential command area of the five sample systems in the period after 1981 than before Moreover they did this in a way that did not decrease average system yields as discussed earlier In making this interpretation we are suggesting that there was some redistribution of water from better-watered areas to fringe areas which would otherwise not have received irrigation water and that this redistribution was a direct response to the change in NIA prioritiesand operating policies and rules occurring around 1981
It is difficult to prove the assertion that water was in fact redistributed with a resulting increase in directly-measured equity Without access to reliable discharge data broken out by systemsection and we can only assume in the absence of a plausible
- 14 shy
alternative explanation that it was such a redistribution that made the increase in the BASA ratio possible In a larger sense it is difficult to prove conclusively that any outcome in a before and after analysis was the result of a particular independent causative factor In this case we have tried to remove the influence of other potential causative factors where we could but the possibilityremains that some combination of unmeasured factors are responsiblefor the difference in the BASA ratio found We do note though that this type of response is exactly the type that would be expected to follow from an emphasis on increased farmer satisfaction and cooperation and increased fee revenues Because the fee schedule is tied to benefitted area the only ways NIA can increase its revenue from that source are to expand benefitted area and to increase collection efficiencies The former depends on redistributing a fixed supply of water over a larger portion of the command while the latter requires that farmers be satisfied with the irrigation service they are receiving and the commitment of the local irrigators association to assist in the task of collecting the amounts due The evidence while not conclusive is highly suggestive that this is exactly what has happened
Efficiency In addition to measures which reflect the levels of adequacy and equity of irrigation service available data allow the calculation of a measure of operating efficiency The term efficiencyusually denotes the relationship between inputs to a process and its outputs often expressed as a ratio The output measure employed here is aggregate system rice output and the input is quantity of irrigation water turned into the system Dividing the first by the second gives a measure of agricultural production per unit water--here termed specific yield This is a highly integrated measure that evaluates the combined efficiency of the irrigation and agricultural processes As such it is a function of the managerial and other inputs supplied both to the irrigation system and to the agricultural operation With respect to one important input to the irrigation system we do know that NIA per hectare field operating expenses were about 29 percent lower in real terms in the 1982-86 period comparedto the 1976-1981 period although this drop may have been partlyoffset by increases in farmer-supplied labor inputs Other things being equal one would thus expect to find a decline in output efficiency
The regression analysis shows positive signs for the period terms in both wet and dry season equations (see Table 5) In the case of the wet season the period dummy in equation 5 is significant but the overall explanatory power of the model is quite low For the dry season (equation 7) the coefficient is positive but non-significant This means that after taking rainfall and fertilizer use into account data do not indicate a lowering of specific yield in the wake of funding reductions and the strong emphasiS on financial viabilitybeginning in 1981 This result provides evidence that the efficiency of the overall irrigation deliveryagricultural production process
- 15 shy
relative to the system water input did not falloff as a result of the changes implemented at least over the short run
Impact magnitude
The preceding analysis has shown us that some indicators of irrigation performance changed significantly following the managerial changes of 1981 while others did not However it has not given us a sense of the size of the changes which occurred To determine the magnitude of these changes the regression model is used to predict the response of the composite system to the managerial changes given a common set of -input and environmental factors To do this averagevalues of the independent variables from the entire eleven-year period 1976 to 1986 are put into the model together with the previously determined coefficients to generate predicted average values of the various dependent variables used in the earlier analysis with and without the period dummy This procedure produces a pair of estimates for each dependent variable under the same conditions--one in which the system responds as it did after the managerial changes were implemented and one in which it responds as it did prior to their introduction The differences between these two values thus indicate the magnitude of the changes occurring in the various indicators of performance discussed above
The results of this exercise are shown in Table 6 The table shows that water availability decreased by about 13 percent in both wet and dry seasons when the period dummy was included and while the coefficients responsible were significant in the earlier analysisthis difference cannot be easily connected with levels of system management as discussed earlier With respect to rice output per hectare although the coefficients were not very significant it is interesting to note that yield increases by 163 kilograms per hectare for the wet season and by 101 kilogram for the dry when the period dummy is included in spite of the reduced water supply available Keep in mind that the predicted yield values have already been adjusted for differences in nitrogen fertilizer use and rainfall This suggests that timeliness and equity of distribution of water supply to farmers may have increased following the changes contributing to the higher predicted yields
Examining the impact of increased equity of distribution bylooking at the ratio of benefitted area to service area we recall that the change was positive and significant for the dry season and negative and not significant for the wet Table 6 shows that the dry season BASA ratio increases by 7 percentage points when the dummy is included a 131 percent increase Other things being equal this should result in a 131 percent increase in system output due to the expansion of ared benefitted This is a major impact on production
- 16 shy
CONCLUSIONS
The Philippine experiment to transform the national irrigation agency into an enterprise has undoubtedly been successful in reducing system operating expenses bringing revenues and costs into line and eliminating the recurrent cost burden imposed by large-scaleirrigation systems on the national budget Evidence presented in this paper indicates that in the process equity of water distribution across systems has also improved In the 5 years following the cessation of operating subsidies from the government an index of equity of distribution improved by about 13 percent At the same time per hectare yields adjusted for rainfall and nitrogen application held constant
There is a strong logical connection between the achievement of financial viability and improved equity of water distribution across the command Because increasing irrigation fees is a politicaldecisionlying largely beyond NIAs control expanding the area which can be billed for service is one of the few revenue increasing measures available to the irrigation agency which does not involve major additional investment In the face of constant or shrinkingwater supplies this is achieved only by redistributing water from areas receiving excessive supplies usually near the head ends of canals and laterals to areas receiving no supplies or inadequatesupplies often located near the tails of canals Although data are not available which would allow the direct examination of this hypotheses the two outcomes are logically consistent with each other
Data also show that per hectare water deliveries declined significantly in the five sample systems after 1981 even thoughrainfall did not differ appreciably between the two periods This decline averaged about 13 percent for both wet and dry seasons and is interpreted as a decline in water availability in the supplying rivers rather than a conscious reduction in withdrawals by system managers Such declines could result from changes in watershed runoff characteristics as caused by deforestation or from increased upstream abstractions from supplying rivers
Improved water distribution tends to increase the area served system agricultural output and NIA service fee revenue Reduced water supplies to the system tend to reduce these things Specificyield defined as system paddy output per unit water held roughly constant across the two periods indicating that the two effects mayhave offset each other
After adjusting for rainfall and nitrogen application perhectare yields increased only marginally in the post-1981 period Area served on the other hand increased by about 13 percent after adjusting for water supply availability indicating that the area benefitted by irrigation in the sample systems increased by about the same percentage Even if yields on this additional area are less than
- 17 shy
average yields for the system this still represents a sizeable increase in system agricultural output as a result of the change in management structure the increase coming not from higher yields but from expanded area under irrigation
The evidence assembled here suggests that there are significantfinancial and economic benefits to be had from changes in the basic character of irrigation managing agencies which make them more responsive to their clientele and which impose rational internal financial discipline on the agency The analysis suggests a number of additional questions however One relates to the longer-term impacts of the structural management changes The improvements in water distribution described here are relatively short-term events occurring during the first 5 years of the new management mode Critics have suggested the danger of underinvestment in systemmaintenance over the longer run accompanied by declining yields and benefitted areas and eventual system collapse This possibility needs to be closely monitored A second concern relates to the apparent decline in water supply to these 5 geographically dispersed systems The nature and causes of this decline need to be explored further since if widespread and secular it may represent a serious threat to the stability of Philippine rice production Whether stemming from poor forest management practices or deficient regulation and allocation of surface water resources or other unidentified factors it is an issue that deserves serious and urgent consideration
A third risk is that the incentive structure set up by NIA to guide and stimulate the performance of field units overemphasizes revenue generation at the expense of irrigation service provision to farmers The evidence presented here supports the view that these two objectives are mutually reinforcing under policies and conditions which have been established in the Philippines More detailed crossshysectional studies based on primary flow measurement data would add confidence to this conclusion and help to specify the conditions under which this effect occurs This could be extremely important in transferring the results of the Philippine experiment to other countries
A final risk is that outside intervention well meaning or otherwise will destroy the basis of NIAs financial autonomy or will impose external pressures or constraints on NIAs decision-making that will subvert the management practices which have been so painstakinglydeveloped and implemented Among these are calls for NIA to be subsumed again within the government department structure in the interests of better coordination with agriculture attempts byexternal financing agencies to arbitrarily increase NIAs expenditures on OampM on the assumption that this will increase system agricultural output or intervention by Philippine legislative bodies to restore operating subsidies to NIA with attached strings leading back to legislators home districts Pressures such as these will cut short a process of experimentation and improvement that seems promising
- 18 shy
enough to date to warrant its continuation Having developed the capacity to establish targets and implement and manage change NIA is in a strong position to modify its objectives to better achieve larger social purposes established for it It is critical to recognize however that this must happen within the context of financing policies that mandate financial autonomy for NIA if the fundamental institutional commitment to manage is to be preserved
The author would like to thank Leslie Small and JeremyBerkoff for helpful comments on an earlier unpublishedversion of this paper and Charles Rogers for his careful and creative help with the analysis
- 19 shy
BIBLIOGRAPHY
Abernethy Charles L 1990 Indicators of the performance of irrigation water distribution systems International Irrigation Management Institute Colombo Sri Lanka Mimeo
Asian Development Bank 1986 Irrigation service fees Proceedingsof the Regional Seminar on Irrigation Service Fees Manila Asian Development Bank
Carruthers Ian and Colin Clark 1981 Economics of IrrigationLiverpool Liverpool University Press Third Edition
Levine G and EW Coward Jr 1986 Irrigation water distribution implications for design and operation AGREP Division WorkingPaper 125 vol 1 World Bank Agriculture and Rural Development Department
Small Les E 1989 User charges in irrigation potentials and limitations Irrigation and drainage vol 3 no 2125-142
Small Les 1990 Irrigation service fees in Asia IrrigationManagement Network 9013 London Overseas DevelopmentInstitute
Svendsen Mark and Les Small 1989 A framework for assessing irrigation system performance Paper prepared for the Symposium on Performance Evaluation 23 November International IrrigationManagement Institute Sri Lanka
Table I--National Irrigation Administration revenues and expenditures in constant prices 1976-86
Item 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986
(peso million 1972)
Revenues Irrigation fees collected 1273 1483 17 13 1831 2070 1668 1699 1893 1728 2129 2546 Other income 715 737 2420 5591 2631 5990 7783 6638 5699 4932 2934
Total direct revenue 1988 2220 4133 7422 4701 7658 9482 8531 7427 061 5480
Expenses in 1972 pricesTotal expenses 4825 5716 5039 6329 3821 77 55 6166 4749 4348 4259 4959
Excess (deficit) (2837(3496) (906) 1093 877 (097) 3316 3782 3079 2802 521 N 0
Subsidies Government operation and
maintenance subsidies 2521 2741 2799 1817 1398 633 0 0 0 0 0 Calamity fund payments 548 0 0 0 0 0 0 0 119 0 142
Total subsidy 3069 2741 2799 1817 1398 633 0 0 119 0 142
Total excess (deficit) 231 (754) 1893 2910 2275 536 3316 3782 3198 2802 663
Source IFPRI analysis of NIA data
- 21 shy
Table 2--Descriptive characteristics of selected MIA systeasa
------- Region III -------- ----- Region VI ----shyUPRIIS Angatii Sto Sibalom- Aganan-
Haasim Thomasc San Jose Sta Barbara
Average service area (hal 102272 31462 3522 5282 8703 Average irrigated area (hal
Wet season 83768 23454 3007 4410 8300 Dry season
Average benefited area Wet season
(hal 64587
77 605
27639
22908
1 781
3007
2801
4369
2770
7698 Dry season
Average rainfall Wet season
(mml d 62478
1 685 5
27396
8576
1 781
3051 0
2769
24731
2997
20001 Dry season 756 333 322 2828 3025
Average discharge (Llsec) Wet season 46501 14792 1692 2353 4984 Dry season 78091 22812 2014 1276 2315
Average water delivery (mmday) Wet season 522 548 487 462 571 Dry season
Average yield (mtha) 1089 715 995 398 686
Wet season 345 419 322 395 435 Dry season
Avg yield per unit water 34 03
(kgm ) 451 412 399 426
Wet season 0373 0440 0373 0538 0443 Dry season 0248 0400 0279 0690 0428
t-statistic difference in mean rainfall 1978-81 1982-86e
Wet season 0432 0713 -0567 1169 1169 Dry season 0519 -0230 -0523 1187 1187 Annual 0460 0707 -0686 1445 1445
~ Summary numbers are averages for the period 1982-1986 except as noted Water delivery discharge and yield per unit discharge are 4-year averages 1982-1985
c Water delivery discharge and yield per unit discharge are 4-year averages d 1983-1986
For Angat 5 years are 1981-85 For St Thomas 1979-83 For Sibalom 1971-75 e No significant differences at 95 confidence
- 22 shy
Table 3--Results of pooled regressions water delivery indicators
Dependent variable Independent variable - Water Delivery in mmBenefited Ha - --- BenefitService Area Ratio --shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season-shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 6799 (998)
6189 (1019)
13102 (1093)
2411 (1 54)
0743 ( 1186)
0800 (1165)
0129 (131 )
0258 (329)
System 2 indicator 0173 (029)
-3743 (-363)
-0063 (-219)
0200 (505)
System 3 indicator 0698 (085)
0163 (012)
0085 (208)
-0046 (-097)
System 4 indicator -0514 (-l 07)
-7011 (-803)
0080 (340)
0169 (l87)
System 5 indicator 0177 (033)
-5049 (-538)
0174 (603)
0051 (058)
Reservoir indicator 0327 (078)
5703 (543)
-0126 (-510)
0156 (261)
Period indicatora -0808 (-240)
-0771 (-229)
-1214 (-209)
-1309 (-1 62)
-0036 (-192)
-0025 (-112)
0071 (239)
0030 (077)
Water delivery mmb
Wet season
Dry season
Precipitation in mm Wet season
Annual
Average daily rainfall Wet season
Dry season
-46E-4 (-125)
-1 9E-4 (-082)
-43E-4 (-077)
20E-3 (385)
0015 (1 59)
-0001 ( -004)
0023 (215)
0001 (021)
0070 (613)
-0017 (-058)
0046 (371)
0010 (033)
R2 (adj) number of observations degrees of freedom
0156 43 36
0133 43 39
0711 41 34
0434 41 37
0693 42 34
0556 42 37
0865 40 32
0730 40 35
Coefficient (t statistic) indicates significance at 95 confidence ~ 1982-1986 1
per ha of service area
Ta
ble
--I
nd
ices
of
se
v
ice a
rea
ib
en
ef
ited
a
rea
a
nd
a
vera
ge sea
so
na
l d
isch
arg
e
Ave
rage
A
vera
ge
tshy19
77
1978
19
79
1980
19
81
1982
19
83
1984
19
85
1986
19
77-8
1 19
82-8
6 S
tati
stic
a
(ind
ex
aver
age
1983
-198
5 =
100)
Ser
vice
are
a UP
RI IS
in
dex
882
91
9
920
91
5
951
95
1
100
0 10
00
100
0 10
00
917
99
0
bull5
53
Ang
at-M
aasi
m R
95
9
994
99
7
996
99
6
996
10
00
100
0 10
00
100
0 98
8
999
1
63
Sto
To
mas
10
63
106
3 10
1 9
10
29
103
0 99
8
100
0 10
00
100
0 11
10
104
1 10
22
-08
9
Siba
lom
-San
Jos
e 95
0
872
94
2
933
94
2
942
10
28
102
8 94
4
101
7
928
99
2
292
A
gana
n-St
a B
arba
ra
108
3 10
91
106
0 10
30
961
10
05
996
10
08
996
99
6
104
5 10
00
-21
1 A
vera
ge
987
98
8
981
98
1
916
97
8
100
5 10
07
988
10
25
984
10
01
123
Wet
seas
on b
enef
ited
are
a in
dex
UPR
IIS
110
0 10
07
114
4 10
55
113
8 11
74
951
10
71
971
ll
58
10
89
106
7 -0
47
Ang
at-M
aasi
m R
97
9
974
92
0
983
10
28
100
8 99
7
102
1 98
2
931
97
7
988
0
54
Sto
To
mas
11
63
115
9 11
23
107
5 10
80
103
9 98
1
978
10
41
103
9 11
20
101
6 -4
88
Siba
lom
-San
Jos
e 11
35
103
8 10
11
931
93
4
906
10
07
975
10
1S
998
10
11
981
-0
80
Aga
nan-
Sta
Bar
bara
10
85
110
5 10
74
106
8 10
01
104
2 99
8
101
5
987
61
0
106
7 93
0
-1 8
4
Ave
rage
10
92
105
6 10
54
102
4 10
36
103
4 98
7
101
3 10
00
947
10
53
996
-2
32
Dry
sea
Son
bene
fite
d ar
ea
inde
x U
PRIIS
14
04
155
0 15
S0
155
8 16
17
128
0 57
2
114
S 15
74
152
3 12
38
-09
4 A
ngat
-Maa
sim
R
903
93
0
103
2 10
61
104
2 10
69
988
99
2
102
1 99
6
993
10
13
061
S
to
Tom
as
105
7 12
27
122
5
961
99
9
115
9 10
1 7
91
0
107
3 12
1 2
10
94
107
4 -0
28
Si
balo
m-S
an J
ose
Aga
nan-
Sta
Bar
bara
66
5
95S
62
6
632
67
4
111
4
501
10
S1
412
11
51
766
93
3
856
94
3
107
0 94
8
107
4 11
09
111
6
158
6 58
8
987
97
6
110
4 5
35
083
N
w
Ave
rage
89
6
964
11
19
103
7 10
44
110
9 10
1 7
89
8
108
5 12
97
101
7 10
81
082
Wet
seas
on d
isch
arge
in
dex
UPR
IIS
132
9 72
7
142
5 11
88
120
4 98
8
105
8 10
63
879
96
4
117
5 99
1
-16
5 A
ngat
-Maa
sim
R
129
7 13
52
134
5 12
58
127
0 11
70
120
3 62
7
131
3 10
68
-18
9
Sto
To
mas
14
71
155
0 14
1 6
10
40
725
12
35
112
3 14
79
103
1 -4
48
Siba
lom
-San
Jos
e 96
8
733
11
1 5
92
2
907
47
1
102
3 85
7
ll2
0
422
92
9
779
-1
09
Aga
nan-
Sta
Bar
bara
87
9
863
68
1
925
96
8
110
7 70
5
871
87
7
00
9
Ave
rage
11
49
105
7 13
61
115
0 10
58
853
10
43
963
99
4
SO3
11
55
940
-2
85
Dry
sea
son
disc
harg
e in
dex
UPR
IIS
425
13
09
153
3 14
28
180
6 14
S0
125
8 56
3
117
9 14
66
130
0 11
89
-04
3
Ang
at-M
aasi
m R
12
26
161
6 15
30
139
2 12
85
100
8 10
33
958
14
41
107
1 -3
80
S
to
Tom
as
116
4 14
82
155
7 79
9
971
12
30
126
5 14
01
106
6 -2
44
Si
balo
m-S
an J
ose
486
71
5
782
62
9
789
64
5
116
9 11
86
801
65
3
918
2
36
Aga
nan-
Sta
Bar
bara
A
vera
ge
425
10
46
133
6 62
0
118
4 81
5
116
1 94
7
112
5 89
7
922
12
20
991
88
2
108
7 12
36
119
2 71
8
114
0 10
37
105
5 3
23
-07
5
a bull
indi
cate
s si
gnif
ican
ce a
t 95
per
cent
con
fide
nce
- 24 shy
Table 5--Results of pooled regressions yield indicators
------------------------- Dependent variables -------------------------shyIndependent variable ---------- Yield MTHa ----------- -- Specific Yield 1000 KgM
3 H20 shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season -shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 1967 (286)
3747 (580)
1991 (293)
2800 (460 )
0084 (041)
0441 (229)
0049 (023 )
0876 (343)
System 2 indicator 0400 (155)
0525 (226)
0057 (085)
0115 (1 95)
System 3 indicator 0466 (139 )
-0029 (-010)
-0016 (-018)
0006 (008)
System 4 indicator 0980 (470)
0090 (049)
0176 (333)
0391 (827)
System 5 indicator 1 221 (570)
0256 ( 138)
0104 (174)
0224 (440)
Reservoir indicator -1033 (-578)
-0063 (-039)
-0146 (-310)
-0326 (-569)
Period indicatora
Precipitation in mm Wet season
Annual
0163 (1 03)
-29E-4 (-182)
0231 (1 42)
-58E-4 (-606)
0161 (118)
92E-5 (072)
0199 043 )
-64pound-5 (-084)
0082 (200)
57E-6 (013)
0092 (223)
-61E-5 (-232)
0032 (093)
29E-7 (001)
0052 (112)
-12pound-4 (-414)
Nitrogen applicationb 0025 (254)
0021 (211)
0025 (274)
0020 (227)
0003 (1 05)
0002 (043)
0003 (101)
-0002 (-052)
R2 (adj) 0553 0516 0251 0198 0309 0262 0713 0445 number of observations 50 50 49 49 42 42 40 40 degrees of freedom 42 45 41 44 34 37 32 35
CoeffiCient (t statistic) indicates significance at 95 confidence ~ 1982-1986 = 1
kgha
- 25 shy
Table 6--Suggmry of estiates perforance indicators
Variable Predicted Predicted Period Dependent Variable
Description (units)
Period Coefficient
1976-86 IndicatorO
1976-86 Indicatorl
Effecta (X)
BenefittedService area ratio
Wet season ratio -0036 0853 0817 -425 Dry Season 0071 0539 0610 1313
Delivery Wet Dry
season season
nm water per benefitted ha
-0808 -1 214
6003 9083
5195 7869
-1346 -1337
Rice Yield Wet season metric tons 0163 3578 3741 456 Dry season per hectare 0161 3905 4066 412
Specific Yield Wet season kg rice per m3 0082 0348 0430 2353 Dry season water de 1 ivered 0032 0378 0409 837
a percentage changes relative to period=O values
- 26 shy
ENDNOTES
IResearch Fellow International Food Policy Research Institute and Senior Irrigation Specialist International Irrigation Management Institute
2Presumable farmers also base private investment decisions on such foreknowledge of costs and commonly do so in the case of small tubewells In addition user groups may invest in small communal schemes independently after examining projected costs and benefits Most commonly however new irrigation capacity is created as a result of public investment decisions made by government and thus the political process comes into play
3This principle applies only when water charges are levied volumetrically--an exceedingly rare situation in the developing worldshy-though the argument is frequently invoked as though it applies to all allocative situations
4For Cabanatuan City the longer period of record available ie 1904 to 1986 was divided into two equal periods and tested for difference in means to test the observation by farmers that rainfall today is lower than it used to be Mean precipitation after 1947 was actually about 3 percent higher than before but the difference was not significant (t = 07)
5The procedure used was to take time series data on total nitrogenfertilizer consumption in the Philippines multiply it by the ratio of N fertilizer used on rice to total N consumption and divide the product by the rice harvested area to obtain an estimate of N use perhectare Values of the ratio of N use on rice to total N use were taken from IRRI (1988 150) with interpolation used to fill in gaps in the series
6This assumes that timeliness is evaluated against a standard based on the physiological demand for water generated by the rice crop and the varying sensitivity of the crop to stress at different growth stages In such a case a poor timeliness rating for irrigationservice will result in reductions in yield
7Because rice yields are insensitive to water applications that exceed PET values a redistribution of water from water excess to water short portions of the command area will usually result in higher aggregate output and higher average yields Exceptional scenarios can be constructed in which this general rule would be violated but such scenarios are unlikely to play out in practice
- 27 shy
aThere are a variety of standards for judging equity that could be used here The most common one is equality of per hectare water deliveries In the present case the implicit standard is equality of water supply utility in producing a crop Thus if two areas received equal amounts of water on a per hectare basis but one had such high seepage and percolation losses that the crop failed to produce a harvest (and was therefore exempted from irrigation fee payment) the water supply to the combined area would be judged inequitable Because such radical differences in seepage and percolation rates are unlikely on puddled rice soils in the lowland Philippines and because the definition of benefitted area accommodates a wide range of acceptable yields these two definitions are largely indistinguishable at the current limits of data resolution
- 3 shy
It should be clear from this brief description of the purposes of irrigation charging systems that cost recovery from projectbeneficiaries is not an end in itself but a way of achieving specificefficiency and equity ends within the national economy Small and coshyauthors (ADB 1986) have summarized these ends in the following set of criteria for usefulness of a charging system They assert that a charging system has appropriate impacts if
1 it results in improved irrigation performance through
(a) more efficient operation and maintenance of irrigationfacilities and
(b) more efficient use of water by farmers and
2 it promotes other objectives of the government by
(a) leading to better irrigation investment decisions (b) easing the governments financial burden and (c) resulting in a more equitable distribution of income
This paper focusses on the impacts subtended by item number one above
Fees and Financial Autonomy
In addition to rules which specify which groups shall bear the costs of providing irrigation service and in what proportions the policies and practices which specify how collections are coursed through the government financial system and the relationship collections bear to irrigation agency income are critical determinants of the agencys operating environment A necessary condition for functionally linking the collection of irrigation service fees and effective irrigation performance is that the agency involved in providing the service be financially autonomous (Small et al 1989 Svendsen 1986) Financial autonomy is defined as a condition where (a) the irrigation agency must rely on user charges for a significantportion of the resources used for OampM and (b) the agency has expenditure control over the use of the funds generated from these charges (Small 1990)
When financial autonomy is present several incentive forces come into play which are otherwise absent First there is incentive for the agency to increase its income Increased income for the agencyimplies maintenance of jobs higher salaries incentive payments and bonuses greater staff mobility and new vehicles quarters and facilities for the staff If fees are levied on an area basis as is usual in developing countries this means that the irrigation agencyhas an a strong vested interest in expanding the area receivingadequate irrigation service increasing fee collection rates and increasing farm incomes
- 4 shy
Second there is incentive to reduce costs Reduced operating costs are often necessary to keep the agencys budget in the black a condition which is usually required to retain independent status In many ways this motive runs counter to the one mentioned above and it is the dynamic tension between these two that creates an efficient and responsive service provider
Working together these two motives generate a demand for better agency relations with cultivators greater accuracy in information collection and record keeping new technology to manage information more effectively better water control and greater farmer involvement in system maintenance and fee collection Thus incentives for greater efficiency in resource use in a context of financial autonomy act on the providers of irrigation service at least as powerfully as the payment of water charges affects consumers of those services Overall a relationship of mutual dependency is established between the two where the agency provides an essential service to farmers and farmers in turn provide the agency with the financial resources it needs to operate
THE PHILIPPINE EXPERIMENT
Irrigation in the Phjlippines
The irrigation sector in the Philippines is divided between two major surface water components--Communal Irrigation Systems (CIS) and National Irrigation Systems (NIS) Groundwater irrigation in the Philippines has always been of secondary importance Communal irrigation is an ancient practice in the Philippines and is the more important segment in terms of net area irrigated currently covering about 48 percent of the nations irrigated area of 1488000 hectares This component of the sector is made up of generally small systems managed by farmers which are usually constructed by them as well Since the mid-seventies the National Irrigation Administration (NIA)has been actively involved in innovative efforts to assist CIS without compromising farmer ownership and operation
The other major component of the irrigation sector NIS comprises about 42 percent of total irrigated area and consists of larger systems developed and operated by NIA Operation of these systems consumes by far the largest share of NIAs operating budgetand constitute its largest potential source of revenue NIAs efforts to deal with these two salient characteristics are described brieflyin the following section
Evolution of a New Institutional Form
In a major departure from regional norms the Philippine Irrigation Department was abolished in 1964 and a public corporation created in its stead During the first decade of its existence the
- 5 shy
National Irrigation Administration (NIA) operated in a way that differed little from regular government departments A major overhaul of its charter in 1974 however led to far-reaching changes in NIAs organizational values structure and operations At the core of these values was the presumption that to be successful NIA must be financially viable taking in more income than it spent By 1979 it had achieved the goal of overall financial viability and in 1981 the last operating subsidy paid out from the national treasury was received
Policy Shift The major thrust of the 1974 charter amendment was to allow NIA to retain all revenues generated by it including irrigationfee collections Heretofore all collections had been immediatelyturned over to the treasury in exchange for an annual appropriationfor operating expenses unrelated to NIAs self-generated revenues The annual appropriation that NIA received had always exceeded by a significant margin the collections that it remitted Accompanyingthis shift however was an agreement that all government operatingsubsidies to NIA were to be phased out over the ensuing five-year period At the end of that period NIAs operating budget would be completely self-financed
NIA Response NIA management responded to these charter changes with a four part strategy aimed at bringing its costs and revenues into balance The strategy comprised actions to
bull Devolve responsibility for certain operational maintenance and fee-collection tasks to farmers
bull Increase corporate revenues by raising fees improvingcollections and generating secondary income from ancillaryactivities
bull Reduce operating costs through a series of minor economies and through major cuts in the personnel budget and
bull Provide financial incentives for superior performance to outstanding field units and to individuals in them
Following earlier successes in organizing farmers in the communal irrigation sector in 1980 NIA began experimenting with waysto organize farmers in its larger systems into effective irrigatorsassociations which could assume responsibility for some canal maintenance water allocation and fee collection functions By 1986 the area under various forms of farmer management had reached about 100000 hectares out of a total of about 600000 hectares in the country Depending on the specific type of devolution reductions in NIAs staffing levels in a sample of affected systems ranged from 13 to 75 percent (Svendsen et al 1989)
Immediately following the 1974 charter amendment NIA obtained permission to increase its fees for irrigation service At the same time fees were indexed for inflation by denominating them in measures of paddy NIA was authorized to collect fees in paddy just as village
- 6 shy
moneylenders do Since that time the agency has made strenuous efforts to increase its fee collections The net effect has been to hold fee revenues per hectare constant in the face of a national rice support price that has steadily declined relative to more generalindices of inflation
At the same time NIA also took steps to reduce its operating expenses Although a number of minor measures of economy were mandated initially the fact that more than three-quarters of the operating budget was devoted to personnel costs meant that any real savings would require reductions in staffing levels Voluntaryreductions were carried out in the late 1970s and early 1980s resulting in a decrease in the number of staff per hectare and a reduction of the personnel share of the budget from 80 percent in 1976 to about 74 percent in 1986
With an eye on its bottom line NIA also instituted a system of performance grants for all field units and the individuals in them termed Viability Incentive Grants To facilitate this each largeirrigation system in the country was made a separate cost center to allow costs and revenues to be accounted for on a system-by-system basis This program provided that once a unit achieved a net excess of revenues over operating costs in a given year a fraction of the surplus would be shared among the units personnel Five of the 11 irrigation regions of the country were receiving these incentive payments by 1986 as were 53 of the 120 individual systems included within the 11 regions
Effects The financial results of these efforts are shown in Table 1 If subsidies are not considered NIA first achieved net profitabilityin 1979 and retained it through the end of the period studied except for a small deficit incurred in 1981 Subsidies were eliminated in 1982 (except for occasional small calamity grants following typhoons) Although revenues have declined in recent years due largely to decreases in interest earnings and construction management fees expenses have declined more rapidly resulting in a series of net positive balances
Achieving a financially viable position is an importantaccomplishment few other irrigation agencies in the developing world have been able to do this However there is some risk that such an achievement occurs at the expense of the quality of service provided to clients Some of the most interesting and important consequences of the new cost recovery policies therefore relate to the physical performance of the irrigation systems NIA operates It is here that the end objectives of the irrigation investments are realized and where the lives of the farmers who till system lands are affected The remainder of the paper will examine and attempt to quantify the impact of these policy changes on physical irrigation systemperformance
- 7 shy
IMPACTS ON SYSTEM PERFORMANCE
Study Methodology
Unfortunately the kinds of changes in hydrologic system outputand the impact on agricultural performance which might be expected to result from improved institutional performance are quite difficult to capture and quantify There are a number of reasons for this First there is the year to year variability of system performance caused byvariable rainfall which feeds rivers fills reservoirs and supplements irrigation water in supplying crop requirements Second there is the difficulty of defining just what performance is and specifying how to measure it Third and most important the regularly collected data from which indicators can be constructed are limited in type number of measuring points and period of record and are sometimes of doubtful reliability
These difficulties notWithstanding an attempt was made to determine the impact that changes in operating procedures staffinglevels and incentive programs had on system performance Because the effects that we are trying to assess resulted from changes that affected all of the systems under NIAs direct authority there are no control systems which can be used as standards We are forced therefore to rely on a comparison of values of selected performance indicators before and after the date of the major structural and procedural changes which is taken to be 1981
To accomplish this secondary data were assembled for 5 systems in Administrative Regions III and VI which had not undergonesignificant phYSical changes during the period of analysis Duringthis process several site visits were made by study team members Time series data collected include service area (SA) and benefitted area (BA) for both wet and dry seasons yields for wet and dry seasons monthly main canal discharge at the system headworks and monthly precipitation The general period of availability for this data is 1966-86 though for systems which began operation after 1966 the period of record is shorter and the records of some systemscontain miSSing values These five systems their 1986 service and benefitted areas and other descriptive data are shown in Table 2
The prinCipal problem with using a before and after approachrather than one that considers comparable systems with and without the innovation is that some of the measured difference in effects mayhave resulted from causes which are independent of the ones beingstudied These causes can be specific in which case they may be relatively easy to identify and accommodate or more general and diffuse and therefore more difficult to control for In the present case--that of changes in the performance of NIA irrigation systemsresulting from the major organizational changes in 1981 two principalexternal factors can be identified which might be expected to affect differences in measured levels of irrigation performance between the
- 8 shy
two time periods These are rainfall and level of use of other agricultural inputs Since our interest is in systems managerial responses to these changes and since there is no reason to believe that the relative magnitudes of the responses are dependent on the size of the system the five systems are treated as equal in the analysis That is changes in measured values relating to the smallest system are considered to be as important as changes in values for the largest with no area weighting applied
Since rainfall can substitute for irrigation water supplies and since it affects the supply of water available in rivers for irrigation it may exert some independent influence on various performance indicators To test the strength of the relationship for the period being analyzed simple correlations were run between rainfall and benefitted area for one system in each regionBenefitted area was used in this analysis because it is the variable deemed most likely to be influenced by year-to-year changes in rainfall Weather data from Cabanatuan City was used for the UPRIIS system which surrounds it and Iloilo City data was used for the nearby Aganan-Santa Barbara system For UPRIIS all of the R2 values for these correlations were less than 0005 suggesting that rainfall has almost no impact on area harvested in this large reservoir-based scheme For Aganan-Santa Barbara wet season rainfall was related to wet season BA (r2 = 016) and to BA during the following dry season (r2 = 024) Signs of the simple correlations were in the expected directions ie wet season rainfall increased BA during both the wet and the subsequent dry seasons These connections are understandable but weak
Another possibility is that there were longer-term differences in rainfall received in the two regions If this were the case a comparison of performance during two different time periods would have to take this difference into account Differences in average precipitation during the two periods were examined for the four stations used in the analysis (see Table 2) In no case were differences in seasonal or annual mean rainfall statisticallysignificant4
bull
Nevertheless in the regression approach adopted to analyze the data rainfall was included in each equation to control for its possible effect on the particular dependent variables being analyzed In doing this wet season rainfall was used in analyzing wet season performance indicators while annual rainfall was used in analyzingdry season data The rationale for this is that while dry season rainfall cannot possibly influence the wet season crop the dry season crop is affected by both the rainfall received directly and the rain falling during the preceding wet season through its effect on river discharge reservoir storage and antecedent soil moisture conditions
The level of agricultural production is also an often-used indicator of an irrigation systems performance Its major weakness
- 9 shy
is that a number of factors other than irrigation service such as labor inputs relative prices and fertilizer use influence it It is necessary therefore either to control for changes in the levels of these inputs or assume that they are constant across the two periods being compared In the present case the most important of these factors is the level of application of chemical fertilizer Because fertilizer use by farmers is responsive to the relative pricesof fertilizer and rice it also includes to some extent input and output price effects Since reliable data on fertilizer use for individual systems were not available estimates derived from FAO fertilizer and cultivated rice area data were used t~ control for the effect of changes in the use of this input over time This variable was included in any of the regression equations in which agriculturalproduction was used as the dependent variable Other factors such as labor use genetic potential of varieties sewn and pesticideapplications are assumed to be constant across the two periods
The analytic approach employed is to fit linear regressionequations to pooled data from the five systems covering a eleven-year period 1976 to 1986 A dummy variable is used to check the impact of pre and post 1981 periods on differences in the dependent variable after the effects of factors such as rainfall and nitrogen fertilizer use have been removed In addition because the dataset was created by pooling data from five different systems a set of 4 site dummies was included in the basic model to control for system-specific differences caused by variables which were not measured For some runs these were replaced with dummies that separated reservoir and non-reservoir systems though equations using the reservoir dummy were consistently inferior to those using the complete set of site dummies Several different dependent variables were created to index the quality of irrigation service and tested using this approachRegression results are given in Tables 3 and 5 and discussed below
Performance Indicators
A variety of indicators have been used in evaluating irrigationperformance in various contexts The selection of appropriate indicators depends on a number of factors including the purpose of the evaluation the audience for its results the way in which the boundaries of the irrigation system are defined and the kind and quality of data available to the evaluators The current analysis is designed to evaluate the impact of a set of management changes on system physical performance The audience for this analysis comprises top-level managers of the irrigation agency and policy-makers at higher government levels Boundary definition is an importantanalytic problem here as evident from the subsequent discussion relating to the choice of the appropriate area values to use in scaling system inputs and outputs This issue is also related to the data quality and availability problems which have already been mentioned
- 10 shy
Three fundamental indicators have been proposed for assessing the effectiveness of irrigation services to cultivators (Svendsen and Small 1989 Abernethy 1990) These are the adequacy of water supplies the equity of their distribution across the command area of the system and the timeliness of the supplies Computation of adequacy measures requires information on the total quantity of water delivered to the system over a season on a per hectare basis Equityand timeliness measures require information on the spatial and temporal distribution respectively of those supplies Where appropriate discharge information is not available proxies can be employed by making suitable assumptions Standards must be selected against which the magnitude of the indicators can be judged
The task in the present case however is somewhat different Here the need is to evaluate changes in selected variables between the pre-1981 and the post-1981 periods Hence the absolute values of variables selected are less important than their relative magnitudesand the statistical significance of the differences in magnitudes between the two periods A distinct limitation is imposed by the data series available for the five sample systems Since discharge and yield data are available only on a whole-system basis it is impossible to develop measures of equity and timeliness directly We will however extend our analysis to a discussion of equity byindirect inference Levine and Coward (1986) have argued that equityought to be considered as the paramount objective in managing largepublic irrigation systems They base their conclusion on an analysisof eight small community-managed systems and five larger public systems including UPRIIS in which equity appears to comprise the most important operational objective in the successful systems It may be appropriate therefore to give success in improving equity of distribution added weight in assessing overall performance
Area Estimates Because measures of system agricultural output and water supplied are typically reduced to a unit area basis before being used much depends on the area values which are used to standardize them Two different area measures are available The first is Service Area (SA) which is defined as the irrigable portion of the command area which is provided with physical facilities for water delivery This represents the area which could conceivably be irrigated in a given season if water supply were not constraining This value may change somewhat from year to year in response to urban encroachment on irrigated command minor remodeling and repair and refinements in area estimates In the present case though the systems selected for analysis were chosen to avoid those which had undergone more extensive rehabilitation or modification
The second measure is Benefitted Area (BA) which is the area billed for payment of irrigation service fees It is the irrigated area harvested which did not have yields so low that the farm was exempted from payment of fees in a given season This threshold value has been approximately 2 tons of paddy per hectare Benefitted area
- 11 shy
varies more than does SA particularly during the dry season when available water supply may seriously constrain the area which can be planted Its magnitude is a function of system managers actions in authorizing the amount of land to be planted in a given season farmers decisions regarding whether to plant or not and the combined ability of system managers and farmersirrigators subsequently to distribute water Both of these area measures will be used to standardize other variables for particular purposes as well as beingcombined to form a separate indicator by themselves
Adequacy The most direct measure of the adequacy of irrigation water supplies to the agricultural system is the quantity of water applied to the system command area on a per unit area basis relative to some standard In this case since our interest is in differences in water adequacy between two time periods and since the systems being assessed have been and continue to be almost entirely devoted to rice cultivation during both cropping seasons depth measures for the two periods may be compared directly assuming the seasonal cropdemand for water to be unchanged Although dry-footed crops can suffer yield losses from overapplication of irrigation water rice is largely insensitive to this effect In addition water can substitute for other inputs that the farmer would otherwise have to provide such as weed control and more careful (and costly) water management We assume therefore that other things being equal larger values of depth applied are better than smaller values in terms of meeting crop water demands and reduce the cost of cultivation At the same time high levels of water adequacy can affect the values of other performance measures--particularly equity
When the regression model is run for quantity of water diverted at the system headworks divided by BA hereafter termed depth we see that the period dummy is negative and significant at the 95 percent confidence level for both wet and dry seasons (see Table 3 equations 1 and 3) Since the overall explanatory power of the wet season model is very weak however we will focus on the dry season in interpreting this result which indicates that after adjusting for rainfall differences significantly less water was delivered to the command per unit of benefitted area following 1981 than before This indicates based on the criteria outlined above that performance in terms of water adequacy deteriorated following financial selfshysufficiency We need to examine this conclusion more carefullyhowever
One difficulty is that the measured quantity of water diverted at the source is largely a function of the supply available in the river rather than of system management This is particularly true during the dry season and in non-reservoir systems Thus while the depth of water supplied to the system is a measure of the adequacy of the systems service it is to some extent beyond the control of the managing agency To better understand the factors behind this decline in water availability we look at simple unadjusted index values for
- 12 shy
several of the key variables Table 4 shows annual values of total volume of water delivered in each season SA and BA in both wet and dry seasons and shows the results of t-tests on the means of a set of indicators before and after 1981 Indicators are used rather than the actual values to weight each of the systems equally regardless of its size The table shows that both the average wet season benefitted area and the average discharge are significantlylower during the second period compared with the first For the dry season too the discharge index is lower after 1981 than before but this difference is not significant At the same time the dry season BA index rose slightly but again the change was not significantSince middotthere is not a clear pattern of relative movement of discharge and BA during the respective seasons no simple interpretation of these index value changes is possible What stands out is that both discharge and benefitted area declined across periods during the wet season while during the dry season there was no significant change in either indicator across the two periods It seems clear that the decline in water adequacy must be evaluated together with other measures of performance in drawing conclusions about the overall impact of the 1981 changes on the quality of system management
Another measured variable per hectare yield can be used as a proxy for water adequacy It has the advantage of partiallyreflecting the impacts of the dimensions of timeliness6 and equity7of distribution as well integrating all three effects into a combined impact on aggregate crop production Table 5 (equations 1 and 3)shows that the period dummy in the yield regressions has a positive sign in both seasons after controlling for nitrogen application and precipitation though the t-values are not significant Treatingyield adjusted in this way as a proxy for quality of irrigationservice leads to the conclusion that by this more comprehensive measure quality of service held constant across the two periods in the dry season Because of the large yield component accounted for byrainfall during the wet season no such judgement is possible for that season however
Equity As noted earlier no reliable data are available for subdivisions of the five sample systems making direct computation of equity measures impossible We can make some judgements about changesin the equity of water distribution however by examining changes in the ratio of two area measures given for each system SA and BA Since SA is the area which theoretically can be supplied with irrigation water by the system and BA s the area which actuallyreceives a quantity of water adequate to produce a remunerative crop the ratio of the two provides a measure of the percentage of the potential service area which was irrigated to a particular standard The larger this pErcentage the more equitable 8 is the distribution This of course assumes that the quantity of water available to the systems is constant across the two periods
- 13 shy
Since the condition of constancy of water supply is not generally satisfied a regression was run in which the total quantityof water diverted at the headworks of each system divided by the systems potential service area SA was included in the regression to control for changes in the water supply available to the system seeTable 3 equations 5 through 8 The average daily rainfall received directly on the system service area during the season was also included as an independent variable The regression was run separately for wet and dry seasons The sign and t-statistic of the period dummy should then tell us whether or not equity as reflected in the BASA ratio increased decreased or remained unchanged across the period divide
Both equations are reasonable good as indicated by the R2 values though the dry season equation is considerably better as would be expected For the wet season both the water delivery term and the rainfall term in equation 5 are of positive sign but are nonshysignificant at the 95 percent confidence level indicating that wet season irrigated area does not change appreciably in response to level of wet season rainfall or the available irrigation water supply The period dummy was negative but not significant indicating that equityof distribution as reflected in the BASA ratio was similar during the two periods
For the dry season the water delivery term in equation 7 is positive and strongly significant indicating a close relationship between the fraction of potential area actually irrigated and the water supply available at the headworks In addition however the period dummy is positive and significant suggesting that once the influence of water supply is removed the BASA ratio was significantly higher in the period following 1981 than it was before
This is an important finding for it reflects significantlyimproved performance in terms of a factor equity of water distribution that is under the control of the managing entity an entity which here comprises both NIA and irrigators associations Interpreted in these terms NIA and allied farmers associations were able to spread a given amount of water more widely across the potential command area of the five sample systems in the period after 1981 than before Moreover they did this in a way that did not decrease average system yields as discussed earlier In making this interpretation we are suggesting that there was some redistribution of water from better-watered areas to fringe areas which would otherwise not have received irrigation water and that this redistribution was a direct response to the change in NIA prioritiesand operating policies and rules occurring around 1981
It is difficult to prove the assertion that water was in fact redistributed with a resulting increase in directly-measured equity Without access to reliable discharge data broken out by systemsection and we can only assume in the absence of a plausible
- 14 shy
alternative explanation that it was such a redistribution that made the increase in the BASA ratio possible In a larger sense it is difficult to prove conclusively that any outcome in a before and after analysis was the result of a particular independent causative factor In this case we have tried to remove the influence of other potential causative factors where we could but the possibilityremains that some combination of unmeasured factors are responsiblefor the difference in the BASA ratio found We do note though that this type of response is exactly the type that would be expected to follow from an emphasis on increased farmer satisfaction and cooperation and increased fee revenues Because the fee schedule is tied to benefitted area the only ways NIA can increase its revenue from that source are to expand benefitted area and to increase collection efficiencies The former depends on redistributing a fixed supply of water over a larger portion of the command while the latter requires that farmers be satisfied with the irrigation service they are receiving and the commitment of the local irrigators association to assist in the task of collecting the amounts due The evidence while not conclusive is highly suggestive that this is exactly what has happened
Efficiency In addition to measures which reflect the levels of adequacy and equity of irrigation service available data allow the calculation of a measure of operating efficiency The term efficiencyusually denotes the relationship between inputs to a process and its outputs often expressed as a ratio The output measure employed here is aggregate system rice output and the input is quantity of irrigation water turned into the system Dividing the first by the second gives a measure of agricultural production per unit water--here termed specific yield This is a highly integrated measure that evaluates the combined efficiency of the irrigation and agricultural processes As such it is a function of the managerial and other inputs supplied both to the irrigation system and to the agricultural operation With respect to one important input to the irrigation system we do know that NIA per hectare field operating expenses were about 29 percent lower in real terms in the 1982-86 period comparedto the 1976-1981 period although this drop may have been partlyoffset by increases in farmer-supplied labor inputs Other things being equal one would thus expect to find a decline in output efficiency
The regression analysis shows positive signs for the period terms in both wet and dry season equations (see Table 5) In the case of the wet season the period dummy in equation 5 is significant but the overall explanatory power of the model is quite low For the dry season (equation 7) the coefficient is positive but non-significant This means that after taking rainfall and fertilizer use into account data do not indicate a lowering of specific yield in the wake of funding reductions and the strong emphasiS on financial viabilitybeginning in 1981 This result provides evidence that the efficiency of the overall irrigation deliveryagricultural production process
- 15 shy
relative to the system water input did not falloff as a result of the changes implemented at least over the short run
Impact magnitude
The preceding analysis has shown us that some indicators of irrigation performance changed significantly following the managerial changes of 1981 while others did not However it has not given us a sense of the size of the changes which occurred To determine the magnitude of these changes the regression model is used to predict the response of the composite system to the managerial changes given a common set of -input and environmental factors To do this averagevalues of the independent variables from the entire eleven-year period 1976 to 1986 are put into the model together with the previously determined coefficients to generate predicted average values of the various dependent variables used in the earlier analysis with and without the period dummy This procedure produces a pair of estimates for each dependent variable under the same conditions--one in which the system responds as it did after the managerial changes were implemented and one in which it responds as it did prior to their introduction The differences between these two values thus indicate the magnitude of the changes occurring in the various indicators of performance discussed above
The results of this exercise are shown in Table 6 The table shows that water availability decreased by about 13 percent in both wet and dry seasons when the period dummy was included and while the coefficients responsible were significant in the earlier analysisthis difference cannot be easily connected with levels of system management as discussed earlier With respect to rice output per hectare although the coefficients were not very significant it is interesting to note that yield increases by 163 kilograms per hectare for the wet season and by 101 kilogram for the dry when the period dummy is included in spite of the reduced water supply available Keep in mind that the predicted yield values have already been adjusted for differences in nitrogen fertilizer use and rainfall This suggests that timeliness and equity of distribution of water supply to farmers may have increased following the changes contributing to the higher predicted yields
Examining the impact of increased equity of distribution bylooking at the ratio of benefitted area to service area we recall that the change was positive and significant for the dry season and negative and not significant for the wet Table 6 shows that the dry season BASA ratio increases by 7 percentage points when the dummy is included a 131 percent increase Other things being equal this should result in a 131 percent increase in system output due to the expansion of ared benefitted This is a major impact on production
- 16 shy
CONCLUSIONS
The Philippine experiment to transform the national irrigation agency into an enterprise has undoubtedly been successful in reducing system operating expenses bringing revenues and costs into line and eliminating the recurrent cost burden imposed by large-scaleirrigation systems on the national budget Evidence presented in this paper indicates that in the process equity of water distribution across systems has also improved In the 5 years following the cessation of operating subsidies from the government an index of equity of distribution improved by about 13 percent At the same time per hectare yields adjusted for rainfall and nitrogen application held constant
There is a strong logical connection between the achievement of financial viability and improved equity of water distribution across the command Because increasing irrigation fees is a politicaldecisionlying largely beyond NIAs control expanding the area which can be billed for service is one of the few revenue increasing measures available to the irrigation agency which does not involve major additional investment In the face of constant or shrinkingwater supplies this is achieved only by redistributing water from areas receiving excessive supplies usually near the head ends of canals and laterals to areas receiving no supplies or inadequatesupplies often located near the tails of canals Although data are not available which would allow the direct examination of this hypotheses the two outcomes are logically consistent with each other
Data also show that per hectare water deliveries declined significantly in the five sample systems after 1981 even thoughrainfall did not differ appreciably between the two periods This decline averaged about 13 percent for both wet and dry seasons and is interpreted as a decline in water availability in the supplying rivers rather than a conscious reduction in withdrawals by system managers Such declines could result from changes in watershed runoff characteristics as caused by deforestation or from increased upstream abstractions from supplying rivers
Improved water distribution tends to increase the area served system agricultural output and NIA service fee revenue Reduced water supplies to the system tend to reduce these things Specificyield defined as system paddy output per unit water held roughly constant across the two periods indicating that the two effects mayhave offset each other
After adjusting for rainfall and nitrogen application perhectare yields increased only marginally in the post-1981 period Area served on the other hand increased by about 13 percent after adjusting for water supply availability indicating that the area benefitted by irrigation in the sample systems increased by about the same percentage Even if yields on this additional area are less than
- 17 shy
average yields for the system this still represents a sizeable increase in system agricultural output as a result of the change in management structure the increase coming not from higher yields but from expanded area under irrigation
The evidence assembled here suggests that there are significantfinancial and economic benefits to be had from changes in the basic character of irrigation managing agencies which make them more responsive to their clientele and which impose rational internal financial discipline on the agency The analysis suggests a number of additional questions however One relates to the longer-term impacts of the structural management changes The improvements in water distribution described here are relatively short-term events occurring during the first 5 years of the new management mode Critics have suggested the danger of underinvestment in systemmaintenance over the longer run accompanied by declining yields and benefitted areas and eventual system collapse This possibility needs to be closely monitored A second concern relates to the apparent decline in water supply to these 5 geographically dispersed systems The nature and causes of this decline need to be explored further since if widespread and secular it may represent a serious threat to the stability of Philippine rice production Whether stemming from poor forest management practices or deficient regulation and allocation of surface water resources or other unidentified factors it is an issue that deserves serious and urgent consideration
A third risk is that the incentive structure set up by NIA to guide and stimulate the performance of field units overemphasizes revenue generation at the expense of irrigation service provision to farmers The evidence presented here supports the view that these two objectives are mutually reinforcing under policies and conditions which have been established in the Philippines More detailed crossshysectional studies based on primary flow measurement data would add confidence to this conclusion and help to specify the conditions under which this effect occurs This could be extremely important in transferring the results of the Philippine experiment to other countries
A final risk is that outside intervention well meaning or otherwise will destroy the basis of NIAs financial autonomy or will impose external pressures or constraints on NIAs decision-making that will subvert the management practices which have been so painstakinglydeveloped and implemented Among these are calls for NIA to be subsumed again within the government department structure in the interests of better coordination with agriculture attempts byexternal financing agencies to arbitrarily increase NIAs expenditures on OampM on the assumption that this will increase system agricultural output or intervention by Philippine legislative bodies to restore operating subsidies to NIA with attached strings leading back to legislators home districts Pressures such as these will cut short a process of experimentation and improvement that seems promising
- 18 shy
enough to date to warrant its continuation Having developed the capacity to establish targets and implement and manage change NIA is in a strong position to modify its objectives to better achieve larger social purposes established for it It is critical to recognize however that this must happen within the context of financing policies that mandate financial autonomy for NIA if the fundamental institutional commitment to manage is to be preserved
The author would like to thank Leslie Small and JeremyBerkoff for helpful comments on an earlier unpublishedversion of this paper and Charles Rogers for his careful and creative help with the analysis
- 19 shy
BIBLIOGRAPHY
Abernethy Charles L 1990 Indicators of the performance of irrigation water distribution systems International Irrigation Management Institute Colombo Sri Lanka Mimeo
Asian Development Bank 1986 Irrigation service fees Proceedingsof the Regional Seminar on Irrigation Service Fees Manila Asian Development Bank
Carruthers Ian and Colin Clark 1981 Economics of IrrigationLiverpool Liverpool University Press Third Edition
Levine G and EW Coward Jr 1986 Irrigation water distribution implications for design and operation AGREP Division WorkingPaper 125 vol 1 World Bank Agriculture and Rural Development Department
Small Les E 1989 User charges in irrigation potentials and limitations Irrigation and drainage vol 3 no 2125-142
Small Les 1990 Irrigation service fees in Asia IrrigationManagement Network 9013 London Overseas DevelopmentInstitute
Svendsen Mark and Les Small 1989 A framework for assessing irrigation system performance Paper prepared for the Symposium on Performance Evaluation 23 November International IrrigationManagement Institute Sri Lanka
Table I--National Irrigation Administration revenues and expenditures in constant prices 1976-86
Item 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986
(peso million 1972)
Revenues Irrigation fees collected 1273 1483 17 13 1831 2070 1668 1699 1893 1728 2129 2546 Other income 715 737 2420 5591 2631 5990 7783 6638 5699 4932 2934
Total direct revenue 1988 2220 4133 7422 4701 7658 9482 8531 7427 061 5480
Expenses in 1972 pricesTotal expenses 4825 5716 5039 6329 3821 77 55 6166 4749 4348 4259 4959
Excess (deficit) (2837(3496) (906) 1093 877 (097) 3316 3782 3079 2802 521 N 0
Subsidies Government operation and
maintenance subsidies 2521 2741 2799 1817 1398 633 0 0 0 0 0 Calamity fund payments 548 0 0 0 0 0 0 0 119 0 142
Total subsidy 3069 2741 2799 1817 1398 633 0 0 119 0 142
Total excess (deficit) 231 (754) 1893 2910 2275 536 3316 3782 3198 2802 663
Source IFPRI analysis of NIA data
- 21 shy
Table 2--Descriptive characteristics of selected MIA systeasa
------- Region III -------- ----- Region VI ----shyUPRIIS Angatii Sto Sibalom- Aganan-
Haasim Thomasc San Jose Sta Barbara
Average service area (hal 102272 31462 3522 5282 8703 Average irrigated area (hal
Wet season 83768 23454 3007 4410 8300 Dry season
Average benefited area Wet season
(hal 64587
77 605
27639
22908
1 781
3007
2801
4369
2770
7698 Dry season
Average rainfall Wet season
(mml d 62478
1 685 5
27396
8576
1 781
3051 0
2769
24731
2997
20001 Dry season 756 333 322 2828 3025
Average discharge (Llsec) Wet season 46501 14792 1692 2353 4984 Dry season 78091 22812 2014 1276 2315
Average water delivery (mmday) Wet season 522 548 487 462 571 Dry season
Average yield (mtha) 1089 715 995 398 686
Wet season 345 419 322 395 435 Dry season
Avg yield per unit water 34 03
(kgm ) 451 412 399 426
Wet season 0373 0440 0373 0538 0443 Dry season 0248 0400 0279 0690 0428
t-statistic difference in mean rainfall 1978-81 1982-86e
Wet season 0432 0713 -0567 1169 1169 Dry season 0519 -0230 -0523 1187 1187 Annual 0460 0707 -0686 1445 1445
~ Summary numbers are averages for the period 1982-1986 except as noted Water delivery discharge and yield per unit discharge are 4-year averages 1982-1985
c Water delivery discharge and yield per unit discharge are 4-year averages d 1983-1986
For Angat 5 years are 1981-85 For St Thomas 1979-83 For Sibalom 1971-75 e No significant differences at 95 confidence
- 22 shy
Table 3--Results of pooled regressions water delivery indicators
Dependent variable Independent variable - Water Delivery in mmBenefited Ha - --- BenefitService Area Ratio --shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season-shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 6799 (998)
6189 (1019)
13102 (1093)
2411 (1 54)
0743 ( 1186)
0800 (1165)
0129 (131 )
0258 (329)
System 2 indicator 0173 (029)
-3743 (-363)
-0063 (-219)
0200 (505)
System 3 indicator 0698 (085)
0163 (012)
0085 (208)
-0046 (-097)
System 4 indicator -0514 (-l 07)
-7011 (-803)
0080 (340)
0169 (l87)
System 5 indicator 0177 (033)
-5049 (-538)
0174 (603)
0051 (058)
Reservoir indicator 0327 (078)
5703 (543)
-0126 (-510)
0156 (261)
Period indicatora -0808 (-240)
-0771 (-229)
-1214 (-209)
-1309 (-1 62)
-0036 (-192)
-0025 (-112)
0071 (239)
0030 (077)
Water delivery mmb
Wet season
Dry season
Precipitation in mm Wet season
Annual
Average daily rainfall Wet season
Dry season
-46E-4 (-125)
-1 9E-4 (-082)
-43E-4 (-077)
20E-3 (385)
0015 (1 59)
-0001 ( -004)
0023 (215)
0001 (021)
0070 (613)
-0017 (-058)
0046 (371)
0010 (033)
R2 (adj) number of observations degrees of freedom
0156 43 36
0133 43 39
0711 41 34
0434 41 37
0693 42 34
0556 42 37
0865 40 32
0730 40 35
Coefficient (t statistic) indicates significance at 95 confidence ~ 1982-1986 1
per ha of service area
Ta
ble
--I
nd
ices
of
se
v
ice a
rea
ib
en
ef
ited
a
rea
a
nd
a
vera
ge sea
so
na
l d
isch
arg
e
Ave
rage
A
vera
ge
tshy19
77
1978
19
79
1980
19
81
1982
19
83
1984
19
85
1986
19
77-8
1 19
82-8
6 S
tati
stic
a
(ind
ex
aver
age
1983
-198
5 =
100)
Ser
vice
are
a UP
RI IS
in
dex
882
91
9
920
91
5
951
95
1
100
0 10
00
100
0 10
00
917
99
0
bull5
53
Ang
at-M
aasi
m R
95
9
994
99
7
996
99
6
996
10
00
100
0 10
00
100
0 98
8
999
1
63
Sto
To
mas
10
63
106
3 10
1 9
10
29
103
0 99
8
100
0 10
00
100
0 11
10
104
1 10
22
-08
9
Siba
lom
-San
Jos
e 95
0
872
94
2
933
94
2
942
10
28
102
8 94
4
101
7
928
99
2
292
A
gana
n-St
a B
arba
ra
108
3 10
91
106
0 10
30
961
10
05
996
10
08
996
99
6
104
5 10
00
-21
1 A
vera
ge
987
98
8
981
98
1
916
97
8
100
5 10
07
988
10
25
984
10
01
123
Wet
seas
on b
enef
ited
are
a in
dex
UPR
IIS
110
0 10
07
114
4 10
55
113
8 11
74
951
10
71
971
ll
58
10
89
106
7 -0
47
Ang
at-M
aasi
m R
97
9
974
92
0
983
10
28
100
8 99
7
102
1 98
2
931
97
7
988
0
54
Sto
To
mas
11
63
115
9 11
23
107
5 10
80
103
9 98
1
978
10
41
103
9 11
20
101
6 -4
88
Siba
lom
-San
Jos
e 11
35
103
8 10
11
931
93
4
906
10
07
975
10
1S
998
10
11
981
-0
80
Aga
nan-
Sta
Bar
bara
10
85
110
5 10
74
106
8 10
01
104
2 99
8
101
5
987
61
0
106
7 93
0
-1 8
4
Ave
rage
10
92
105
6 10
54
102
4 10
36
103
4 98
7
101
3 10
00
947
10
53
996
-2
32
Dry
sea
Son
bene
fite
d ar
ea
inde
x U
PRIIS
14
04
155
0 15
S0
155
8 16
17
128
0 57
2
114
S 15
74
152
3 12
38
-09
4 A
ngat
-Maa
sim
R
903
93
0
103
2 10
61
104
2 10
69
988
99
2
102
1 99
6
993
10
13
061
S
to
Tom
as
105
7 12
27
122
5
961
99
9
115
9 10
1 7
91
0
107
3 12
1 2
10
94
107
4 -0
28
Si
balo
m-S
an J
ose
Aga
nan-
Sta
Bar
bara
66
5
95S
62
6
632
67
4
111
4
501
10
S1
412
11
51
766
93
3
856
94
3
107
0 94
8
107
4 11
09
111
6
158
6 58
8
987
97
6
110
4 5
35
083
N
w
Ave
rage
89
6
964
11
19
103
7 10
44
110
9 10
1 7
89
8
108
5 12
97
101
7 10
81
082
Wet
seas
on d
isch
arge
in
dex
UPR
IIS
132
9 72
7
142
5 11
88
120
4 98
8
105
8 10
63
879
96
4
117
5 99
1
-16
5 A
ngat
-Maa
sim
R
129
7 13
52
134
5 12
58
127
0 11
70
120
3 62
7
131
3 10
68
-18
9
Sto
To
mas
14
71
155
0 14
1 6
10
40
725
12
35
112
3 14
79
103
1 -4
48
Siba
lom
-San
Jos
e 96
8
733
11
1 5
92
2
907
47
1
102
3 85
7
ll2
0
422
92
9
779
-1
09
Aga
nan-
Sta
Bar
bara
87
9
863
68
1
925
96
8
110
7 70
5
871
87
7
00
9
Ave
rage
11
49
105
7 13
61
115
0 10
58
853
10
43
963
99
4
SO3
11
55
940
-2
85
Dry
sea
son
disc
harg
e in
dex
UPR
IIS
425
13
09
153
3 14
28
180
6 14
S0
125
8 56
3
117
9 14
66
130
0 11
89
-04
3
Ang
at-M
aasi
m R
12
26
161
6 15
30
139
2 12
85
100
8 10
33
958
14
41
107
1 -3
80
S
to
Tom
as
116
4 14
82
155
7 79
9
971
12
30
126
5 14
01
106
6 -2
44
Si
balo
m-S
an J
ose
486
71
5
782
62
9
789
64
5
116
9 11
86
801
65
3
918
2
36
Aga
nan-
Sta
Bar
bara
A
vera
ge
425
10
46
133
6 62
0
118
4 81
5
116
1 94
7
112
5 89
7
922
12
20
991
88
2
108
7 12
36
119
2 71
8
114
0 10
37
105
5 3
23
-07
5
a bull
indi
cate
s si
gnif
ican
ce a
t 95
per
cent
con
fide
nce
- 24 shy
Table 5--Results of pooled regressions yield indicators
------------------------- Dependent variables -------------------------shyIndependent variable ---------- Yield MTHa ----------- -- Specific Yield 1000 KgM
3 H20 shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season -shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 1967 (286)
3747 (580)
1991 (293)
2800 (460 )
0084 (041)
0441 (229)
0049 (023 )
0876 (343)
System 2 indicator 0400 (155)
0525 (226)
0057 (085)
0115 (1 95)
System 3 indicator 0466 (139 )
-0029 (-010)
-0016 (-018)
0006 (008)
System 4 indicator 0980 (470)
0090 (049)
0176 (333)
0391 (827)
System 5 indicator 1 221 (570)
0256 ( 138)
0104 (174)
0224 (440)
Reservoir indicator -1033 (-578)
-0063 (-039)
-0146 (-310)
-0326 (-569)
Period indicatora
Precipitation in mm Wet season
Annual
0163 (1 03)
-29E-4 (-182)
0231 (1 42)
-58E-4 (-606)
0161 (118)
92E-5 (072)
0199 043 )
-64pound-5 (-084)
0082 (200)
57E-6 (013)
0092 (223)
-61E-5 (-232)
0032 (093)
29E-7 (001)
0052 (112)
-12pound-4 (-414)
Nitrogen applicationb 0025 (254)
0021 (211)
0025 (274)
0020 (227)
0003 (1 05)
0002 (043)
0003 (101)
-0002 (-052)
R2 (adj) 0553 0516 0251 0198 0309 0262 0713 0445 number of observations 50 50 49 49 42 42 40 40 degrees of freedom 42 45 41 44 34 37 32 35
CoeffiCient (t statistic) indicates significance at 95 confidence ~ 1982-1986 = 1
kgha
- 25 shy
Table 6--Suggmry of estiates perforance indicators
Variable Predicted Predicted Period Dependent Variable
Description (units)
Period Coefficient
1976-86 IndicatorO
1976-86 Indicatorl
Effecta (X)
BenefittedService area ratio
Wet season ratio -0036 0853 0817 -425 Dry Season 0071 0539 0610 1313
Delivery Wet Dry
season season
nm water per benefitted ha
-0808 -1 214
6003 9083
5195 7869
-1346 -1337
Rice Yield Wet season metric tons 0163 3578 3741 456 Dry season per hectare 0161 3905 4066 412
Specific Yield Wet season kg rice per m3 0082 0348 0430 2353 Dry season water de 1 ivered 0032 0378 0409 837
a percentage changes relative to period=O values
- 26 shy
ENDNOTES
IResearch Fellow International Food Policy Research Institute and Senior Irrigation Specialist International Irrigation Management Institute
2Presumable farmers also base private investment decisions on such foreknowledge of costs and commonly do so in the case of small tubewells In addition user groups may invest in small communal schemes independently after examining projected costs and benefits Most commonly however new irrigation capacity is created as a result of public investment decisions made by government and thus the political process comes into play
3This principle applies only when water charges are levied volumetrically--an exceedingly rare situation in the developing worldshy-though the argument is frequently invoked as though it applies to all allocative situations
4For Cabanatuan City the longer period of record available ie 1904 to 1986 was divided into two equal periods and tested for difference in means to test the observation by farmers that rainfall today is lower than it used to be Mean precipitation after 1947 was actually about 3 percent higher than before but the difference was not significant (t = 07)
5The procedure used was to take time series data on total nitrogenfertilizer consumption in the Philippines multiply it by the ratio of N fertilizer used on rice to total N consumption and divide the product by the rice harvested area to obtain an estimate of N use perhectare Values of the ratio of N use on rice to total N use were taken from IRRI (1988 150) with interpolation used to fill in gaps in the series
6This assumes that timeliness is evaluated against a standard based on the physiological demand for water generated by the rice crop and the varying sensitivity of the crop to stress at different growth stages In such a case a poor timeliness rating for irrigationservice will result in reductions in yield
7Because rice yields are insensitive to water applications that exceed PET values a redistribution of water from water excess to water short portions of the command area will usually result in higher aggregate output and higher average yields Exceptional scenarios can be constructed in which this general rule would be violated but such scenarios are unlikely to play out in practice
- 27 shy
aThere are a variety of standards for judging equity that could be used here The most common one is equality of per hectare water deliveries In the present case the implicit standard is equality of water supply utility in producing a crop Thus if two areas received equal amounts of water on a per hectare basis but one had such high seepage and percolation losses that the crop failed to produce a harvest (and was therefore exempted from irrigation fee payment) the water supply to the combined area would be judged inequitable Because such radical differences in seepage and percolation rates are unlikely on puddled rice soils in the lowland Philippines and because the definition of benefitted area accommodates a wide range of acceptable yields these two definitions are largely indistinguishable at the current limits of data resolution
- 4 shy
Second there is incentive to reduce costs Reduced operating costs are often necessary to keep the agencys budget in the black a condition which is usually required to retain independent status In many ways this motive runs counter to the one mentioned above and it is the dynamic tension between these two that creates an efficient and responsive service provider
Working together these two motives generate a demand for better agency relations with cultivators greater accuracy in information collection and record keeping new technology to manage information more effectively better water control and greater farmer involvement in system maintenance and fee collection Thus incentives for greater efficiency in resource use in a context of financial autonomy act on the providers of irrigation service at least as powerfully as the payment of water charges affects consumers of those services Overall a relationship of mutual dependency is established between the two where the agency provides an essential service to farmers and farmers in turn provide the agency with the financial resources it needs to operate
THE PHILIPPINE EXPERIMENT
Irrigation in the Phjlippines
The irrigation sector in the Philippines is divided between two major surface water components--Communal Irrigation Systems (CIS) and National Irrigation Systems (NIS) Groundwater irrigation in the Philippines has always been of secondary importance Communal irrigation is an ancient practice in the Philippines and is the more important segment in terms of net area irrigated currently covering about 48 percent of the nations irrigated area of 1488000 hectares This component of the sector is made up of generally small systems managed by farmers which are usually constructed by them as well Since the mid-seventies the National Irrigation Administration (NIA)has been actively involved in innovative efforts to assist CIS without compromising farmer ownership and operation
The other major component of the irrigation sector NIS comprises about 42 percent of total irrigated area and consists of larger systems developed and operated by NIA Operation of these systems consumes by far the largest share of NIAs operating budgetand constitute its largest potential source of revenue NIAs efforts to deal with these two salient characteristics are described brieflyin the following section
Evolution of a New Institutional Form
In a major departure from regional norms the Philippine Irrigation Department was abolished in 1964 and a public corporation created in its stead During the first decade of its existence the
- 5 shy
National Irrigation Administration (NIA) operated in a way that differed little from regular government departments A major overhaul of its charter in 1974 however led to far-reaching changes in NIAs organizational values structure and operations At the core of these values was the presumption that to be successful NIA must be financially viable taking in more income than it spent By 1979 it had achieved the goal of overall financial viability and in 1981 the last operating subsidy paid out from the national treasury was received
Policy Shift The major thrust of the 1974 charter amendment was to allow NIA to retain all revenues generated by it including irrigationfee collections Heretofore all collections had been immediatelyturned over to the treasury in exchange for an annual appropriationfor operating expenses unrelated to NIAs self-generated revenues The annual appropriation that NIA received had always exceeded by a significant margin the collections that it remitted Accompanyingthis shift however was an agreement that all government operatingsubsidies to NIA were to be phased out over the ensuing five-year period At the end of that period NIAs operating budget would be completely self-financed
NIA Response NIA management responded to these charter changes with a four part strategy aimed at bringing its costs and revenues into balance The strategy comprised actions to
bull Devolve responsibility for certain operational maintenance and fee-collection tasks to farmers
bull Increase corporate revenues by raising fees improvingcollections and generating secondary income from ancillaryactivities
bull Reduce operating costs through a series of minor economies and through major cuts in the personnel budget and
bull Provide financial incentives for superior performance to outstanding field units and to individuals in them
Following earlier successes in organizing farmers in the communal irrigation sector in 1980 NIA began experimenting with waysto organize farmers in its larger systems into effective irrigatorsassociations which could assume responsibility for some canal maintenance water allocation and fee collection functions By 1986 the area under various forms of farmer management had reached about 100000 hectares out of a total of about 600000 hectares in the country Depending on the specific type of devolution reductions in NIAs staffing levels in a sample of affected systems ranged from 13 to 75 percent (Svendsen et al 1989)
Immediately following the 1974 charter amendment NIA obtained permission to increase its fees for irrigation service At the same time fees were indexed for inflation by denominating them in measures of paddy NIA was authorized to collect fees in paddy just as village
- 6 shy
moneylenders do Since that time the agency has made strenuous efforts to increase its fee collections The net effect has been to hold fee revenues per hectare constant in the face of a national rice support price that has steadily declined relative to more generalindices of inflation
At the same time NIA also took steps to reduce its operating expenses Although a number of minor measures of economy were mandated initially the fact that more than three-quarters of the operating budget was devoted to personnel costs meant that any real savings would require reductions in staffing levels Voluntaryreductions were carried out in the late 1970s and early 1980s resulting in a decrease in the number of staff per hectare and a reduction of the personnel share of the budget from 80 percent in 1976 to about 74 percent in 1986
With an eye on its bottom line NIA also instituted a system of performance grants for all field units and the individuals in them termed Viability Incentive Grants To facilitate this each largeirrigation system in the country was made a separate cost center to allow costs and revenues to be accounted for on a system-by-system basis This program provided that once a unit achieved a net excess of revenues over operating costs in a given year a fraction of the surplus would be shared among the units personnel Five of the 11 irrigation regions of the country were receiving these incentive payments by 1986 as were 53 of the 120 individual systems included within the 11 regions
Effects The financial results of these efforts are shown in Table 1 If subsidies are not considered NIA first achieved net profitabilityin 1979 and retained it through the end of the period studied except for a small deficit incurred in 1981 Subsidies were eliminated in 1982 (except for occasional small calamity grants following typhoons) Although revenues have declined in recent years due largely to decreases in interest earnings and construction management fees expenses have declined more rapidly resulting in a series of net positive balances
Achieving a financially viable position is an importantaccomplishment few other irrigation agencies in the developing world have been able to do this However there is some risk that such an achievement occurs at the expense of the quality of service provided to clients Some of the most interesting and important consequences of the new cost recovery policies therefore relate to the physical performance of the irrigation systems NIA operates It is here that the end objectives of the irrigation investments are realized and where the lives of the farmers who till system lands are affected The remainder of the paper will examine and attempt to quantify the impact of these policy changes on physical irrigation systemperformance
- 7 shy
IMPACTS ON SYSTEM PERFORMANCE
Study Methodology
Unfortunately the kinds of changes in hydrologic system outputand the impact on agricultural performance which might be expected to result from improved institutional performance are quite difficult to capture and quantify There are a number of reasons for this First there is the year to year variability of system performance caused byvariable rainfall which feeds rivers fills reservoirs and supplements irrigation water in supplying crop requirements Second there is the difficulty of defining just what performance is and specifying how to measure it Third and most important the regularly collected data from which indicators can be constructed are limited in type number of measuring points and period of record and are sometimes of doubtful reliability
These difficulties notWithstanding an attempt was made to determine the impact that changes in operating procedures staffinglevels and incentive programs had on system performance Because the effects that we are trying to assess resulted from changes that affected all of the systems under NIAs direct authority there are no control systems which can be used as standards We are forced therefore to rely on a comparison of values of selected performance indicators before and after the date of the major structural and procedural changes which is taken to be 1981
To accomplish this secondary data were assembled for 5 systems in Administrative Regions III and VI which had not undergonesignificant phYSical changes during the period of analysis Duringthis process several site visits were made by study team members Time series data collected include service area (SA) and benefitted area (BA) for both wet and dry seasons yields for wet and dry seasons monthly main canal discharge at the system headworks and monthly precipitation The general period of availability for this data is 1966-86 though for systems which began operation after 1966 the period of record is shorter and the records of some systemscontain miSSing values These five systems their 1986 service and benefitted areas and other descriptive data are shown in Table 2
The prinCipal problem with using a before and after approachrather than one that considers comparable systems with and without the innovation is that some of the measured difference in effects mayhave resulted from causes which are independent of the ones beingstudied These causes can be specific in which case they may be relatively easy to identify and accommodate or more general and diffuse and therefore more difficult to control for In the present case--that of changes in the performance of NIA irrigation systemsresulting from the major organizational changes in 1981 two principalexternal factors can be identified which might be expected to affect differences in measured levels of irrigation performance between the
- 8 shy
two time periods These are rainfall and level of use of other agricultural inputs Since our interest is in systems managerial responses to these changes and since there is no reason to believe that the relative magnitudes of the responses are dependent on the size of the system the five systems are treated as equal in the analysis That is changes in measured values relating to the smallest system are considered to be as important as changes in values for the largest with no area weighting applied
Since rainfall can substitute for irrigation water supplies and since it affects the supply of water available in rivers for irrigation it may exert some independent influence on various performance indicators To test the strength of the relationship for the period being analyzed simple correlations were run between rainfall and benefitted area for one system in each regionBenefitted area was used in this analysis because it is the variable deemed most likely to be influenced by year-to-year changes in rainfall Weather data from Cabanatuan City was used for the UPRIIS system which surrounds it and Iloilo City data was used for the nearby Aganan-Santa Barbara system For UPRIIS all of the R2 values for these correlations were less than 0005 suggesting that rainfall has almost no impact on area harvested in this large reservoir-based scheme For Aganan-Santa Barbara wet season rainfall was related to wet season BA (r2 = 016) and to BA during the following dry season (r2 = 024) Signs of the simple correlations were in the expected directions ie wet season rainfall increased BA during both the wet and the subsequent dry seasons These connections are understandable but weak
Another possibility is that there were longer-term differences in rainfall received in the two regions If this were the case a comparison of performance during two different time periods would have to take this difference into account Differences in average precipitation during the two periods were examined for the four stations used in the analysis (see Table 2) In no case were differences in seasonal or annual mean rainfall statisticallysignificant4
bull
Nevertheless in the regression approach adopted to analyze the data rainfall was included in each equation to control for its possible effect on the particular dependent variables being analyzed In doing this wet season rainfall was used in analyzing wet season performance indicators while annual rainfall was used in analyzingdry season data The rationale for this is that while dry season rainfall cannot possibly influence the wet season crop the dry season crop is affected by both the rainfall received directly and the rain falling during the preceding wet season through its effect on river discharge reservoir storage and antecedent soil moisture conditions
The level of agricultural production is also an often-used indicator of an irrigation systems performance Its major weakness
- 9 shy
is that a number of factors other than irrigation service such as labor inputs relative prices and fertilizer use influence it It is necessary therefore either to control for changes in the levels of these inputs or assume that they are constant across the two periods being compared In the present case the most important of these factors is the level of application of chemical fertilizer Because fertilizer use by farmers is responsive to the relative pricesof fertilizer and rice it also includes to some extent input and output price effects Since reliable data on fertilizer use for individual systems were not available estimates derived from FAO fertilizer and cultivated rice area data were used t~ control for the effect of changes in the use of this input over time This variable was included in any of the regression equations in which agriculturalproduction was used as the dependent variable Other factors such as labor use genetic potential of varieties sewn and pesticideapplications are assumed to be constant across the two periods
The analytic approach employed is to fit linear regressionequations to pooled data from the five systems covering a eleven-year period 1976 to 1986 A dummy variable is used to check the impact of pre and post 1981 periods on differences in the dependent variable after the effects of factors such as rainfall and nitrogen fertilizer use have been removed In addition because the dataset was created by pooling data from five different systems a set of 4 site dummies was included in the basic model to control for system-specific differences caused by variables which were not measured For some runs these were replaced with dummies that separated reservoir and non-reservoir systems though equations using the reservoir dummy were consistently inferior to those using the complete set of site dummies Several different dependent variables were created to index the quality of irrigation service and tested using this approachRegression results are given in Tables 3 and 5 and discussed below
Performance Indicators
A variety of indicators have been used in evaluating irrigationperformance in various contexts The selection of appropriate indicators depends on a number of factors including the purpose of the evaluation the audience for its results the way in which the boundaries of the irrigation system are defined and the kind and quality of data available to the evaluators The current analysis is designed to evaluate the impact of a set of management changes on system physical performance The audience for this analysis comprises top-level managers of the irrigation agency and policy-makers at higher government levels Boundary definition is an importantanalytic problem here as evident from the subsequent discussion relating to the choice of the appropriate area values to use in scaling system inputs and outputs This issue is also related to the data quality and availability problems which have already been mentioned
- 10 shy
Three fundamental indicators have been proposed for assessing the effectiveness of irrigation services to cultivators (Svendsen and Small 1989 Abernethy 1990) These are the adequacy of water supplies the equity of their distribution across the command area of the system and the timeliness of the supplies Computation of adequacy measures requires information on the total quantity of water delivered to the system over a season on a per hectare basis Equityand timeliness measures require information on the spatial and temporal distribution respectively of those supplies Where appropriate discharge information is not available proxies can be employed by making suitable assumptions Standards must be selected against which the magnitude of the indicators can be judged
The task in the present case however is somewhat different Here the need is to evaluate changes in selected variables between the pre-1981 and the post-1981 periods Hence the absolute values of variables selected are less important than their relative magnitudesand the statistical significance of the differences in magnitudes between the two periods A distinct limitation is imposed by the data series available for the five sample systems Since discharge and yield data are available only on a whole-system basis it is impossible to develop measures of equity and timeliness directly We will however extend our analysis to a discussion of equity byindirect inference Levine and Coward (1986) have argued that equityought to be considered as the paramount objective in managing largepublic irrigation systems They base their conclusion on an analysisof eight small community-managed systems and five larger public systems including UPRIIS in which equity appears to comprise the most important operational objective in the successful systems It may be appropriate therefore to give success in improving equity of distribution added weight in assessing overall performance
Area Estimates Because measures of system agricultural output and water supplied are typically reduced to a unit area basis before being used much depends on the area values which are used to standardize them Two different area measures are available The first is Service Area (SA) which is defined as the irrigable portion of the command area which is provided with physical facilities for water delivery This represents the area which could conceivably be irrigated in a given season if water supply were not constraining This value may change somewhat from year to year in response to urban encroachment on irrigated command minor remodeling and repair and refinements in area estimates In the present case though the systems selected for analysis were chosen to avoid those which had undergone more extensive rehabilitation or modification
The second measure is Benefitted Area (BA) which is the area billed for payment of irrigation service fees It is the irrigated area harvested which did not have yields so low that the farm was exempted from payment of fees in a given season This threshold value has been approximately 2 tons of paddy per hectare Benefitted area
- 11 shy
varies more than does SA particularly during the dry season when available water supply may seriously constrain the area which can be planted Its magnitude is a function of system managers actions in authorizing the amount of land to be planted in a given season farmers decisions regarding whether to plant or not and the combined ability of system managers and farmersirrigators subsequently to distribute water Both of these area measures will be used to standardize other variables for particular purposes as well as beingcombined to form a separate indicator by themselves
Adequacy The most direct measure of the adequacy of irrigation water supplies to the agricultural system is the quantity of water applied to the system command area on a per unit area basis relative to some standard In this case since our interest is in differences in water adequacy between two time periods and since the systems being assessed have been and continue to be almost entirely devoted to rice cultivation during both cropping seasons depth measures for the two periods may be compared directly assuming the seasonal cropdemand for water to be unchanged Although dry-footed crops can suffer yield losses from overapplication of irrigation water rice is largely insensitive to this effect In addition water can substitute for other inputs that the farmer would otherwise have to provide such as weed control and more careful (and costly) water management We assume therefore that other things being equal larger values of depth applied are better than smaller values in terms of meeting crop water demands and reduce the cost of cultivation At the same time high levels of water adequacy can affect the values of other performance measures--particularly equity
When the regression model is run for quantity of water diverted at the system headworks divided by BA hereafter termed depth we see that the period dummy is negative and significant at the 95 percent confidence level for both wet and dry seasons (see Table 3 equations 1 and 3) Since the overall explanatory power of the wet season model is very weak however we will focus on the dry season in interpreting this result which indicates that after adjusting for rainfall differences significantly less water was delivered to the command per unit of benefitted area following 1981 than before This indicates based on the criteria outlined above that performance in terms of water adequacy deteriorated following financial selfshysufficiency We need to examine this conclusion more carefullyhowever
One difficulty is that the measured quantity of water diverted at the source is largely a function of the supply available in the river rather than of system management This is particularly true during the dry season and in non-reservoir systems Thus while the depth of water supplied to the system is a measure of the adequacy of the systems service it is to some extent beyond the control of the managing agency To better understand the factors behind this decline in water availability we look at simple unadjusted index values for
- 12 shy
several of the key variables Table 4 shows annual values of total volume of water delivered in each season SA and BA in both wet and dry seasons and shows the results of t-tests on the means of a set of indicators before and after 1981 Indicators are used rather than the actual values to weight each of the systems equally regardless of its size The table shows that both the average wet season benefitted area and the average discharge are significantlylower during the second period compared with the first For the dry season too the discharge index is lower after 1981 than before but this difference is not significant At the same time the dry season BA index rose slightly but again the change was not significantSince middotthere is not a clear pattern of relative movement of discharge and BA during the respective seasons no simple interpretation of these index value changes is possible What stands out is that both discharge and benefitted area declined across periods during the wet season while during the dry season there was no significant change in either indicator across the two periods It seems clear that the decline in water adequacy must be evaluated together with other measures of performance in drawing conclusions about the overall impact of the 1981 changes on the quality of system management
Another measured variable per hectare yield can be used as a proxy for water adequacy It has the advantage of partiallyreflecting the impacts of the dimensions of timeliness6 and equity7of distribution as well integrating all three effects into a combined impact on aggregate crop production Table 5 (equations 1 and 3)shows that the period dummy in the yield regressions has a positive sign in both seasons after controlling for nitrogen application and precipitation though the t-values are not significant Treatingyield adjusted in this way as a proxy for quality of irrigationservice leads to the conclusion that by this more comprehensive measure quality of service held constant across the two periods in the dry season Because of the large yield component accounted for byrainfall during the wet season no such judgement is possible for that season however
Equity As noted earlier no reliable data are available for subdivisions of the five sample systems making direct computation of equity measures impossible We can make some judgements about changesin the equity of water distribution however by examining changes in the ratio of two area measures given for each system SA and BA Since SA is the area which theoretically can be supplied with irrigation water by the system and BA s the area which actuallyreceives a quantity of water adequate to produce a remunerative crop the ratio of the two provides a measure of the percentage of the potential service area which was irrigated to a particular standard The larger this pErcentage the more equitable 8 is the distribution This of course assumes that the quantity of water available to the systems is constant across the two periods
- 13 shy
Since the condition of constancy of water supply is not generally satisfied a regression was run in which the total quantityof water diverted at the headworks of each system divided by the systems potential service area SA was included in the regression to control for changes in the water supply available to the system seeTable 3 equations 5 through 8 The average daily rainfall received directly on the system service area during the season was also included as an independent variable The regression was run separately for wet and dry seasons The sign and t-statistic of the period dummy should then tell us whether or not equity as reflected in the BASA ratio increased decreased or remained unchanged across the period divide
Both equations are reasonable good as indicated by the R2 values though the dry season equation is considerably better as would be expected For the wet season both the water delivery term and the rainfall term in equation 5 are of positive sign but are nonshysignificant at the 95 percent confidence level indicating that wet season irrigated area does not change appreciably in response to level of wet season rainfall or the available irrigation water supply The period dummy was negative but not significant indicating that equityof distribution as reflected in the BASA ratio was similar during the two periods
For the dry season the water delivery term in equation 7 is positive and strongly significant indicating a close relationship between the fraction of potential area actually irrigated and the water supply available at the headworks In addition however the period dummy is positive and significant suggesting that once the influence of water supply is removed the BASA ratio was significantly higher in the period following 1981 than it was before
This is an important finding for it reflects significantlyimproved performance in terms of a factor equity of water distribution that is under the control of the managing entity an entity which here comprises both NIA and irrigators associations Interpreted in these terms NIA and allied farmers associations were able to spread a given amount of water more widely across the potential command area of the five sample systems in the period after 1981 than before Moreover they did this in a way that did not decrease average system yields as discussed earlier In making this interpretation we are suggesting that there was some redistribution of water from better-watered areas to fringe areas which would otherwise not have received irrigation water and that this redistribution was a direct response to the change in NIA prioritiesand operating policies and rules occurring around 1981
It is difficult to prove the assertion that water was in fact redistributed with a resulting increase in directly-measured equity Without access to reliable discharge data broken out by systemsection and we can only assume in the absence of a plausible
- 14 shy
alternative explanation that it was such a redistribution that made the increase in the BASA ratio possible In a larger sense it is difficult to prove conclusively that any outcome in a before and after analysis was the result of a particular independent causative factor In this case we have tried to remove the influence of other potential causative factors where we could but the possibilityremains that some combination of unmeasured factors are responsiblefor the difference in the BASA ratio found We do note though that this type of response is exactly the type that would be expected to follow from an emphasis on increased farmer satisfaction and cooperation and increased fee revenues Because the fee schedule is tied to benefitted area the only ways NIA can increase its revenue from that source are to expand benefitted area and to increase collection efficiencies The former depends on redistributing a fixed supply of water over a larger portion of the command while the latter requires that farmers be satisfied with the irrigation service they are receiving and the commitment of the local irrigators association to assist in the task of collecting the amounts due The evidence while not conclusive is highly suggestive that this is exactly what has happened
Efficiency In addition to measures which reflect the levels of adequacy and equity of irrigation service available data allow the calculation of a measure of operating efficiency The term efficiencyusually denotes the relationship between inputs to a process and its outputs often expressed as a ratio The output measure employed here is aggregate system rice output and the input is quantity of irrigation water turned into the system Dividing the first by the second gives a measure of agricultural production per unit water--here termed specific yield This is a highly integrated measure that evaluates the combined efficiency of the irrigation and agricultural processes As such it is a function of the managerial and other inputs supplied both to the irrigation system and to the agricultural operation With respect to one important input to the irrigation system we do know that NIA per hectare field operating expenses were about 29 percent lower in real terms in the 1982-86 period comparedto the 1976-1981 period although this drop may have been partlyoffset by increases in farmer-supplied labor inputs Other things being equal one would thus expect to find a decline in output efficiency
The regression analysis shows positive signs for the period terms in both wet and dry season equations (see Table 5) In the case of the wet season the period dummy in equation 5 is significant but the overall explanatory power of the model is quite low For the dry season (equation 7) the coefficient is positive but non-significant This means that after taking rainfall and fertilizer use into account data do not indicate a lowering of specific yield in the wake of funding reductions and the strong emphasiS on financial viabilitybeginning in 1981 This result provides evidence that the efficiency of the overall irrigation deliveryagricultural production process
- 15 shy
relative to the system water input did not falloff as a result of the changes implemented at least over the short run
Impact magnitude
The preceding analysis has shown us that some indicators of irrigation performance changed significantly following the managerial changes of 1981 while others did not However it has not given us a sense of the size of the changes which occurred To determine the magnitude of these changes the regression model is used to predict the response of the composite system to the managerial changes given a common set of -input and environmental factors To do this averagevalues of the independent variables from the entire eleven-year period 1976 to 1986 are put into the model together with the previously determined coefficients to generate predicted average values of the various dependent variables used in the earlier analysis with and without the period dummy This procedure produces a pair of estimates for each dependent variable under the same conditions--one in which the system responds as it did after the managerial changes were implemented and one in which it responds as it did prior to their introduction The differences between these two values thus indicate the magnitude of the changes occurring in the various indicators of performance discussed above
The results of this exercise are shown in Table 6 The table shows that water availability decreased by about 13 percent in both wet and dry seasons when the period dummy was included and while the coefficients responsible were significant in the earlier analysisthis difference cannot be easily connected with levels of system management as discussed earlier With respect to rice output per hectare although the coefficients were not very significant it is interesting to note that yield increases by 163 kilograms per hectare for the wet season and by 101 kilogram for the dry when the period dummy is included in spite of the reduced water supply available Keep in mind that the predicted yield values have already been adjusted for differences in nitrogen fertilizer use and rainfall This suggests that timeliness and equity of distribution of water supply to farmers may have increased following the changes contributing to the higher predicted yields
Examining the impact of increased equity of distribution bylooking at the ratio of benefitted area to service area we recall that the change was positive and significant for the dry season and negative and not significant for the wet Table 6 shows that the dry season BASA ratio increases by 7 percentage points when the dummy is included a 131 percent increase Other things being equal this should result in a 131 percent increase in system output due to the expansion of ared benefitted This is a major impact on production
- 16 shy
CONCLUSIONS
The Philippine experiment to transform the national irrigation agency into an enterprise has undoubtedly been successful in reducing system operating expenses bringing revenues and costs into line and eliminating the recurrent cost burden imposed by large-scaleirrigation systems on the national budget Evidence presented in this paper indicates that in the process equity of water distribution across systems has also improved In the 5 years following the cessation of operating subsidies from the government an index of equity of distribution improved by about 13 percent At the same time per hectare yields adjusted for rainfall and nitrogen application held constant
There is a strong logical connection between the achievement of financial viability and improved equity of water distribution across the command Because increasing irrigation fees is a politicaldecisionlying largely beyond NIAs control expanding the area which can be billed for service is one of the few revenue increasing measures available to the irrigation agency which does not involve major additional investment In the face of constant or shrinkingwater supplies this is achieved only by redistributing water from areas receiving excessive supplies usually near the head ends of canals and laterals to areas receiving no supplies or inadequatesupplies often located near the tails of canals Although data are not available which would allow the direct examination of this hypotheses the two outcomes are logically consistent with each other
Data also show that per hectare water deliveries declined significantly in the five sample systems after 1981 even thoughrainfall did not differ appreciably between the two periods This decline averaged about 13 percent for both wet and dry seasons and is interpreted as a decline in water availability in the supplying rivers rather than a conscious reduction in withdrawals by system managers Such declines could result from changes in watershed runoff characteristics as caused by deforestation or from increased upstream abstractions from supplying rivers
Improved water distribution tends to increase the area served system agricultural output and NIA service fee revenue Reduced water supplies to the system tend to reduce these things Specificyield defined as system paddy output per unit water held roughly constant across the two periods indicating that the two effects mayhave offset each other
After adjusting for rainfall and nitrogen application perhectare yields increased only marginally in the post-1981 period Area served on the other hand increased by about 13 percent after adjusting for water supply availability indicating that the area benefitted by irrigation in the sample systems increased by about the same percentage Even if yields on this additional area are less than
- 17 shy
average yields for the system this still represents a sizeable increase in system agricultural output as a result of the change in management structure the increase coming not from higher yields but from expanded area under irrigation
The evidence assembled here suggests that there are significantfinancial and economic benefits to be had from changes in the basic character of irrigation managing agencies which make them more responsive to their clientele and which impose rational internal financial discipline on the agency The analysis suggests a number of additional questions however One relates to the longer-term impacts of the structural management changes The improvements in water distribution described here are relatively short-term events occurring during the first 5 years of the new management mode Critics have suggested the danger of underinvestment in systemmaintenance over the longer run accompanied by declining yields and benefitted areas and eventual system collapse This possibility needs to be closely monitored A second concern relates to the apparent decline in water supply to these 5 geographically dispersed systems The nature and causes of this decline need to be explored further since if widespread and secular it may represent a serious threat to the stability of Philippine rice production Whether stemming from poor forest management practices or deficient regulation and allocation of surface water resources or other unidentified factors it is an issue that deserves serious and urgent consideration
A third risk is that the incentive structure set up by NIA to guide and stimulate the performance of field units overemphasizes revenue generation at the expense of irrigation service provision to farmers The evidence presented here supports the view that these two objectives are mutually reinforcing under policies and conditions which have been established in the Philippines More detailed crossshysectional studies based on primary flow measurement data would add confidence to this conclusion and help to specify the conditions under which this effect occurs This could be extremely important in transferring the results of the Philippine experiment to other countries
A final risk is that outside intervention well meaning or otherwise will destroy the basis of NIAs financial autonomy or will impose external pressures or constraints on NIAs decision-making that will subvert the management practices which have been so painstakinglydeveloped and implemented Among these are calls for NIA to be subsumed again within the government department structure in the interests of better coordination with agriculture attempts byexternal financing agencies to arbitrarily increase NIAs expenditures on OampM on the assumption that this will increase system agricultural output or intervention by Philippine legislative bodies to restore operating subsidies to NIA with attached strings leading back to legislators home districts Pressures such as these will cut short a process of experimentation and improvement that seems promising
- 18 shy
enough to date to warrant its continuation Having developed the capacity to establish targets and implement and manage change NIA is in a strong position to modify its objectives to better achieve larger social purposes established for it It is critical to recognize however that this must happen within the context of financing policies that mandate financial autonomy for NIA if the fundamental institutional commitment to manage is to be preserved
The author would like to thank Leslie Small and JeremyBerkoff for helpful comments on an earlier unpublishedversion of this paper and Charles Rogers for his careful and creative help with the analysis
- 19 shy
BIBLIOGRAPHY
Abernethy Charles L 1990 Indicators of the performance of irrigation water distribution systems International Irrigation Management Institute Colombo Sri Lanka Mimeo
Asian Development Bank 1986 Irrigation service fees Proceedingsof the Regional Seminar on Irrigation Service Fees Manila Asian Development Bank
Carruthers Ian and Colin Clark 1981 Economics of IrrigationLiverpool Liverpool University Press Third Edition
Levine G and EW Coward Jr 1986 Irrigation water distribution implications for design and operation AGREP Division WorkingPaper 125 vol 1 World Bank Agriculture and Rural Development Department
Small Les E 1989 User charges in irrigation potentials and limitations Irrigation and drainage vol 3 no 2125-142
Small Les 1990 Irrigation service fees in Asia IrrigationManagement Network 9013 London Overseas DevelopmentInstitute
Svendsen Mark and Les Small 1989 A framework for assessing irrigation system performance Paper prepared for the Symposium on Performance Evaluation 23 November International IrrigationManagement Institute Sri Lanka
Table I--National Irrigation Administration revenues and expenditures in constant prices 1976-86
Item 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986
(peso million 1972)
Revenues Irrigation fees collected 1273 1483 17 13 1831 2070 1668 1699 1893 1728 2129 2546 Other income 715 737 2420 5591 2631 5990 7783 6638 5699 4932 2934
Total direct revenue 1988 2220 4133 7422 4701 7658 9482 8531 7427 061 5480
Expenses in 1972 pricesTotal expenses 4825 5716 5039 6329 3821 77 55 6166 4749 4348 4259 4959
Excess (deficit) (2837(3496) (906) 1093 877 (097) 3316 3782 3079 2802 521 N 0
Subsidies Government operation and
maintenance subsidies 2521 2741 2799 1817 1398 633 0 0 0 0 0 Calamity fund payments 548 0 0 0 0 0 0 0 119 0 142
Total subsidy 3069 2741 2799 1817 1398 633 0 0 119 0 142
Total excess (deficit) 231 (754) 1893 2910 2275 536 3316 3782 3198 2802 663
Source IFPRI analysis of NIA data
- 21 shy
Table 2--Descriptive characteristics of selected MIA systeasa
------- Region III -------- ----- Region VI ----shyUPRIIS Angatii Sto Sibalom- Aganan-
Haasim Thomasc San Jose Sta Barbara
Average service area (hal 102272 31462 3522 5282 8703 Average irrigated area (hal
Wet season 83768 23454 3007 4410 8300 Dry season
Average benefited area Wet season
(hal 64587
77 605
27639
22908
1 781
3007
2801
4369
2770
7698 Dry season
Average rainfall Wet season
(mml d 62478
1 685 5
27396
8576
1 781
3051 0
2769
24731
2997
20001 Dry season 756 333 322 2828 3025
Average discharge (Llsec) Wet season 46501 14792 1692 2353 4984 Dry season 78091 22812 2014 1276 2315
Average water delivery (mmday) Wet season 522 548 487 462 571 Dry season
Average yield (mtha) 1089 715 995 398 686
Wet season 345 419 322 395 435 Dry season
Avg yield per unit water 34 03
(kgm ) 451 412 399 426
Wet season 0373 0440 0373 0538 0443 Dry season 0248 0400 0279 0690 0428
t-statistic difference in mean rainfall 1978-81 1982-86e
Wet season 0432 0713 -0567 1169 1169 Dry season 0519 -0230 -0523 1187 1187 Annual 0460 0707 -0686 1445 1445
~ Summary numbers are averages for the period 1982-1986 except as noted Water delivery discharge and yield per unit discharge are 4-year averages 1982-1985
c Water delivery discharge and yield per unit discharge are 4-year averages d 1983-1986
For Angat 5 years are 1981-85 For St Thomas 1979-83 For Sibalom 1971-75 e No significant differences at 95 confidence
- 22 shy
Table 3--Results of pooled regressions water delivery indicators
Dependent variable Independent variable - Water Delivery in mmBenefited Ha - --- BenefitService Area Ratio --shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season-shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 6799 (998)
6189 (1019)
13102 (1093)
2411 (1 54)
0743 ( 1186)
0800 (1165)
0129 (131 )
0258 (329)
System 2 indicator 0173 (029)
-3743 (-363)
-0063 (-219)
0200 (505)
System 3 indicator 0698 (085)
0163 (012)
0085 (208)
-0046 (-097)
System 4 indicator -0514 (-l 07)
-7011 (-803)
0080 (340)
0169 (l87)
System 5 indicator 0177 (033)
-5049 (-538)
0174 (603)
0051 (058)
Reservoir indicator 0327 (078)
5703 (543)
-0126 (-510)
0156 (261)
Period indicatora -0808 (-240)
-0771 (-229)
-1214 (-209)
-1309 (-1 62)
-0036 (-192)
-0025 (-112)
0071 (239)
0030 (077)
Water delivery mmb
Wet season
Dry season
Precipitation in mm Wet season
Annual
Average daily rainfall Wet season
Dry season
-46E-4 (-125)
-1 9E-4 (-082)
-43E-4 (-077)
20E-3 (385)
0015 (1 59)
-0001 ( -004)
0023 (215)
0001 (021)
0070 (613)
-0017 (-058)
0046 (371)
0010 (033)
R2 (adj) number of observations degrees of freedom
0156 43 36
0133 43 39
0711 41 34
0434 41 37
0693 42 34
0556 42 37
0865 40 32
0730 40 35
Coefficient (t statistic) indicates significance at 95 confidence ~ 1982-1986 1
per ha of service area
Ta
ble
--I
nd
ices
of
se
v
ice a
rea
ib
en
ef
ited
a
rea
a
nd
a
vera
ge sea
so
na
l d
isch
arg
e
Ave
rage
A
vera
ge
tshy19
77
1978
19
79
1980
19
81
1982
19
83
1984
19
85
1986
19
77-8
1 19
82-8
6 S
tati
stic
a
(ind
ex
aver
age
1983
-198
5 =
100)
Ser
vice
are
a UP
RI IS
in
dex
882
91
9
920
91
5
951
95
1
100
0 10
00
100
0 10
00
917
99
0
bull5
53
Ang
at-M
aasi
m R
95
9
994
99
7
996
99
6
996
10
00
100
0 10
00
100
0 98
8
999
1
63
Sto
To
mas
10
63
106
3 10
1 9
10
29
103
0 99
8
100
0 10
00
100
0 11
10
104
1 10
22
-08
9
Siba
lom
-San
Jos
e 95
0
872
94
2
933
94
2
942
10
28
102
8 94
4
101
7
928
99
2
292
A
gana
n-St
a B
arba
ra
108
3 10
91
106
0 10
30
961
10
05
996
10
08
996
99
6
104
5 10
00
-21
1 A
vera
ge
987
98
8
981
98
1
916
97
8
100
5 10
07
988
10
25
984
10
01
123
Wet
seas
on b
enef
ited
are
a in
dex
UPR
IIS
110
0 10
07
114
4 10
55
113
8 11
74
951
10
71
971
ll
58
10
89
106
7 -0
47
Ang
at-M
aasi
m R
97
9
974
92
0
983
10
28
100
8 99
7
102
1 98
2
931
97
7
988
0
54
Sto
To
mas
11
63
115
9 11
23
107
5 10
80
103
9 98
1
978
10
41
103
9 11
20
101
6 -4
88
Siba
lom
-San
Jos
e 11
35
103
8 10
11
931
93
4
906
10
07
975
10
1S
998
10
11
981
-0
80
Aga
nan-
Sta
Bar
bara
10
85
110
5 10
74
106
8 10
01
104
2 99
8
101
5
987
61
0
106
7 93
0
-1 8
4
Ave
rage
10
92
105
6 10
54
102
4 10
36
103
4 98
7
101
3 10
00
947
10
53
996
-2
32
Dry
sea
Son
bene
fite
d ar
ea
inde
x U
PRIIS
14
04
155
0 15
S0
155
8 16
17
128
0 57
2
114
S 15
74
152
3 12
38
-09
4 A
ngat
-Maa
sim
R
903
93
0
103
2 10
61
104
2 10
69
988
99
2
102
1 99
6
993
10
13
061
S
to
Tom
as
105
7 12
27
122
5
961
99
9
115
9 10
1 7
91
0
107
3 12
1 2
10
94
107
4 -0
28
Si
balo
m-S
an J
ose
Aga
nan-
Sta
Bar
bara
66
5
95S
62
6
632
67
4
111
4
501
10
S1
412
11
51
766
93
3
856
94
3
107
0 94
8
107
4 11
09
111
6
158
6 58
8
987
97
6
110
4 5
35
083
N
w
Ave
rage
89
6
964
11
19
103
7 10
44
110
9 10
1 7
89
8
108
5 12
97
101
7 10
81
082
Wet
seas
on d
isch
arge
in
dex
UPR
IIS
132
9 72
7
142
5 11
88
120
4 98
8
105
8 10
63
879
96
4
117
5 99
1
-16
5 A
ngat
-Maa
sim
R
129
7 13
52
134
5 12
58
127
0 11
70
120
3 62
7
131
3 10
68
-18
9
Sto
To
mas
14
71
155
0 14
1 6
10
40
725
12
35
112
3 14
79
103
1 -4
48
Siba
lom
-San
Jos
e 96
8
733
11
1 5
92
2
907
47
1
102
3 85
7
ll2
0
422
92
9
779
-1
09
Aga
nan-
Sta
Bar
bara
87
9
863
68
1
925
96
8
110
7 70
5
871
87
7
00
9
Ave
rage
11
49
105
7 13
61
115
0 10
58
853
10
43
963
99
4
SO3
11
55
940
-2
85
Dry
sea
son
disc
harg
e in
dex
UPR
IIS
425
13
09
153
3 14
28
180
6 14
S0
125
8 56
3
117
9 14
66
130
0 11
89
-04
3
Ang
at-M
aasi
m R
12
26
161
6 15
30
139
2 12
85
100
8 10
33
958
14
41
107
1 -3
80
S
to
Tom
as
116
4 14
82
155
7 79
9
971
12
30
126
5 14
01
106
6 -2
44
Si
balo
m-S
an J
ose
486
71
5
782
62
9
789
64
5
116
9 11
86
801
65
3
918
2
36
Aga
nan-
Sta
Bar
bara
A
vera
ge
425
10
46
133
6 62
0
118
4 81
5
116
1 94
7
112
5 89
7
922
12
20
991
88
2
108
7 12
36
119
2 71
8
114
0 10
37
105
5 3
23
-07
5
a bull
indi
cate
s si
gnif
ican
ce a
t 95
per
cent
con
fide
nce
- 24 shy
Table 5--Results of pooled regressions yield indicators
------------------------- Dependent variables -------------------------shyIndependent variable ---------- Yield MTHa ----------- -- Specific Yield 1000 KgM
3 H20 shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season -shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 1967 (286)
3747 (580)
1991 (293)
2800 (460 )
0084 (041)
0441 (229)
0049 (023 )
0876 (343)
System 2 indicator 0400 (155)
0525 (226)
0057 (085)
0115 (1 95)
System 3 indicator 0466 (139 )
-0029 (-010)
-0016 (-018)
0006 (008)
System 4 indicator 0980 (470)
0090 (049)
0176 (333)
0391 (827)
System 5 indicator 1 221 (570)
0256 ( 138)
0104 (174)
0224 (440)
Reservoir indicator -1033 (-578)
-0063 (-039)
-0146 (-310)
-0326 (-569)
Period indicatora
Precipitation in mm Wet season
Annual
0163 (1 03)
-29E-4 (-182)
0231 (1 42)
-58E-4 (-606)
0161 (118)
92E-5 (072)
0199 043 )
-64pound-5 (-084)
0082 (200)
57E-6 (013)
0092 (223)
-61E-5 (-232)
0032 (093)
29E-7 (001)
0052 (112)
-12pound-4 (-414)
Nitrogen applicationb 0025 (254)
0021 (211)
0025 (274)
0020 (227)
0003 (1 05)
0002 (043)
0003 (101)
-0002 (-052)
R2 (adj) 0553 0516 0251 0198 0309 0262 0713 0445 number of observations 50 50 49 49 42 42 40 40 degrees of freedom 42 45 41 44 34 37 32 35
CoeffiCient (t statistic) indicates significance at 95 confidence ~ 1982-1986 = 1
kgha
- 25 shy
Table 6--Suggmry of estiates perforance indicators
Variable Predicted Predicted Period Dependent Variable
Description (units)
Period Coefficient
1976-86 IndicatorO
1976-86 Indicatorl
Effecta (X)
BenefittedService area ratio
Wet season ratio -0036 0853 0817 -425 Dry Season 0071 0539 0610 1313
Delivery Wet Dry
season season
nm water per benefitted ha
-0808 -1 214
6003 9083
5195 7869
-1346 -1337
Rice Yield Wet season metric tons 0163 3578 3741 456 Dry season per hectare 0161 3905 4066 412
Specific Yield Wet season kg rice per m3 0082 0348 0430 2353 Dry season water de 1 ivered 0032 0378 0409 837
a percentage changes relative to period=O values
- 26 shy
ENDNOTES
IResearch Fellow International Food Policy Research Institute and Senior Irrigation Specialist International Irrigation Management Institute
2Presumable farmers also base private investment decisions on such foreknowledge of costs and commonly do so in the case of small tubewells In addition user groups may invest in small communal schemes independently after examining projected costs and benefits Most commonly however new irrigation capacity is created as a result of public investment decisions made by government and thus the political process comes into play
3This principle applies only when water charges are levied volumetrically--an exceedingly rare situation in the developing worldshy-though the argument is frequently invoked as though it applies to all allocative situations
4For Cabanatuan City the longer period of record available ie 1904 to 1986 was divided into two equal periods and tested for difference in means to test the observation by farmers that rainfall today is lower than it used to be Mean precipitation after 1947 was actually about 3 percent higher than before but the difference was not significant (t = 07)
5The procedure used was to take time series data on total nitrogenfertilizer consumption in the Philippines multiply it by the ratio of N fertilizer used on rice to total N consumption and divide the product by the rice harvested area to obtain an estimate of N use perhectare Values of the ratio of N use on rice to total N use were taken from IRRI (1988 150) with interpolation used to fill in gaps in the series
6This assumes that timeliness is evaluated against a standard based on the physiological demand for water generated by the rice crop and the varying sensitivity of the crop to stress at different growth stages In such a case a poor timeliness rating for irrigationservice will result in reductions in yield
7Because rice yields are insensitive to water applications that exceed PET values a redistribution of water from water excess to water short portions of the command area will usually result in higher aggregate output and higher average yields Exceptional scenarios can be constructed in which this general rule would be violated but such scenarios are unlikely to play out in practice
- 27 shy
aThere are a variety of standards for judging equity that could be used here The most common one is equality of per hectare water deliveries In the present case the implicit standard is equality of water supply utility in producing a crop Thus if two areas received equal amounts of water on a per hectare basis but one had such high seepage and percolation losses that the crop failed to produce a harvest (and was therefore exempted from irrigation fee payment) the water supply to the combined area would be judged inequitable Because such radical differences in seepage and percolation rates are unlikely on puddled rice soils in the lowland Philippines and because the definition of benefitted area accommodates a wide range of acceptable yields these two definitions are largely indistinguishable at the current limits of data resolution
- 5 shy
National Irrigation Administration (NIA) operated in a way that differed little from regular government departments A major overhaul of its charter in 1974 however led to far-reaching changes in NIAs organizational values structure and operations At the core of these values was the presumption that to be successful NIA must be financially viable taking in more income than it spent By 1979 it had achieved the goal of overall financial viability and in 1981 the last operating subsidy paid out from the national treasury was received
Policy Shift The major thrust of the 1974 charter amendment was to allow NIA to retain all revenues generated by it including irrigationfee collections Heretofore all collections had been immediatelyturned over to the treasury in exchange for an annual appropriationfor operating expenses unrelated to NIAs self-generated revenues The annual appropriation that NIA received had always exceeded by a significant margin the collections that it remitted Accompanyingthis shift however was an agreement that all government operatingsubsidies to NIA were to be phased out over the ensuing five-year period At the end of that period NIAs operating budget would be completely self-financed
NIA Response NIA management responded to these charter changes with a four part strategy aimed at bringing its costs and revenues into balance The strategy comprised actions to
bull Devolve responsibility for certain operational maintenance and fee-collection tasks to farmers
bull Increase corporate revenues by raising fees improvingcollections and generating secondary income from ancillaryactivities
bull Reduce operating costs through a series of minor economies and through major cuts in the personnel budget and
bull Provide financial incentives for superior performance to outstanding field units and to individuals in them
Following earlier successes in organizing farmers in the communal irrigation sector in 1980 NIA began experimenting with waysto organize farmers in its larger systems into effective irrigatorsassociations which could assume responsibility for some canal maintenance water allocation and fee collection functions By 1986 the area under various forms of farmer management had reached about 100000 hectares out of a total of about 600000 hectares in the country Depending on the specific type of devolution reductions in NIAs staffing levels in a sample of affected systems ranged from 13 to 75 percent (Svendsen et al 1989)
Immediately following the 1974 charter amendment NIA obtained permission to increase its fees for irrigation service At the same time fees were indexed for inflation by denominating them in measures of paddy NIA was authorized to collect fees in paddy just as village
- 6 shy
moneylenders do Since that time the agency has made strenuous efforts to increase its fee collections The net effect has been to hold fee revenues per hectare constant in the face of a national rice support price that has steadily declined relative to more generalindices of inflation
At the same time NIA also took steps to reduce its operating expenses Although a number of minor measures of economy were mandated initially the fact that more than three-quarters of the operating budget was devoted to personnel costs meant that any real savings would require reductions in staffing levels Voluntaryreductions were carried out in the late 1970s and early 1980s resulting in a decrease in the number of staff per hectare and a reduction of the personnel share of the budget from 80 percent in 1976 to about 74 percent in 1986
With an eye on its bottom line NIA also instituted a system of performance grants for all field units and the individuals in them termed Viability Incentive Grants To facilitate this each largeirrigation system in the country was made a separate cost center to allow costs and revenues to be accounted for on a system-by-system basis This program provided that once a unit achieved a net excess of revenues over operating costs in a given year a fraction of the surplus would be shared among the units personnel Five of the 11 irrigation regions of the country were receiving these incentive payments by 1986 as were 53 of the 120 individual systems included within the 11 regions
Effects The financial results of these efforts are shown in Table 1 If subsidies are not considered NIA first achieved net profitabilityin 1979 and retained it through the end of the period studied except for a small deficit incurred in 1981 Subsidies were eliminated in 1982 (except for occasional small calamity grants following typhoons) Although revenues have declined in recent years due largely to decreases in interest earnings and construction management fees expenses have declined more rapidly resulting in a series of net positive balances
Achieving a financially viable position is an importantaccomplishment few other irrigation agencies in the developing world have been able to do this However there is some risk that such an achievement occurs at the expense of the quality of service provided to clients Some of the most interesting and important consequences of the new cost recovery policies therefore relate to the physical performance of the irrigation systems NIA operates It is here that the end objectives of the irrigation investments are realized and where the lives of the farmers who till system lands are affected The remainder of the paper will examine and attempt to quantify the impact of these policy changes on physical irrigation systemperformance
- 7 shy
IMPACTS ON SYSTEM PERFORMANCE
Study Methodology
Unfortunately the kinds of changes in hydrologic system outputand the impact on agricultural performance which might be expected to result from improved institutional performance are quite difficult to capture and quantify There are a number of reasons for this First there is the year to year variability of system performance caused byvariable rainfall which feeds rivers fills reservoirs and supplements irrigation water in supplying crop requirements Second there is the difficulty of defining just what performance is and specifying how to measure it Third and most important the regularly collected data from which indicators can be constructed are limited in type number of measuring points and period of record and are sometimes of doubtful reliability
These difficulties notWithstanding an attempt was made to determine the impact that changes in operating procedures staffinglevels and incentive programs had on system performance Because the effects that we are trying to assess resulted from changes that affected all of the systems under NIAs direct authority there are no control systems which can be used as standards We are forced therefore to rely on a comparison of values of selected performance indicators before and after the date of the major structural and procedural changes which is taken to be 1981
To accomplish this secondary data were assembled for 5 systems in Administrative Regions III and VI which had not undergonesignificant phYSical changes during the period of analysis Duringthis process several site visits were made by study team members Time series data collected include service area (SA) and benefitted area (BA) for both wet and dry seasons yields for wet and dry seasons monthly main canal discharge at the system headworks and monthly precipitation The general period of availability for this data is 1966-86 though for systems which began operation after 1966 the period of record is shorter and the records of some systemscontain miSSing values These five systems their 1986 service and benefitted areas and other descriptive data are shown in Table 2
The prinCipal problem with using a before and after approachrather than one that considers comparable systems with and without the innovation is that some of the measured difference in effects mayhave resulted from causes which are independent of the ones beingstudied These causes can be specific in which case they may be relatively easy to identify and accommodate or more general and diffuse and therefore more difficult to control for In the present case--that of changes in the performance of NIA irrigation systemsresulting from the major organizational changes in 1981 two principalexternal factors can be identified which might be expected to affect differences in measured levels of irrigation performance between the
- 8 shy
two time periods These are rainfall and level of use of other agricultural inputs Since our interest is in systems managerial responses to these changes and since there is no reason to believe that the relative magnitudes of the responses are dependent on the size of the system the five systems are treated as equal in the analysis That is changes in measured values relating to the smallest system are considered to be as important as changes in values for the largest with no area weighting applied
Since rainfall can substitute for irrigation water supplies and since it affects the supply of water available in rivers for irrigation it may exert some independent influence on various performance indicators To test the strength of the relationship for the period being analyzed simple correlations were run between rainfall and benefitted area for one system in each regionBenefitted area was used in this analysis because it is the variable deemed most likely to be influenced by year-to-year changes in rainfall Weather data from Cabanatuan City was used for the UPRIIS system which surrounds it and Iloilo City data was used for the nearby Aganan-Santa Barbara system For UPRIIS all of the R2 values for these correlations were less than 0005 suggesting that rainfall has almost no impact on area harvested in this large reservoir-based scheme For Aganan-Santa Barbara wet season rainfall was related to wet season BA (r2 = 016) and to BA during the following dry season (r2 = 024) Signs of the simple correlations were in the expected directions ie wet season rainfall increased BA during both the wet and the subsequent dry seasons These connections are understandable but weak
Another possibility is that there were longer-term differences in rainfall received in the two regions If this were the case a comparison of performance during two different time periods would have to take this difference into account Differences in average precipitation during the two periods were examined for the four stations used in the analysis (see Table 2) In no case were differences in seasonal or annual mean rainfall statisticallysignificant4
bull
Nevertheless in the regression approach adopted to analyze the data rainfall was included in each equation to control for its possible effect on the particular dependent variables being analyzed In doing this wet season rainfall was used in analyzing wet season performance indicators while annual rainfall was used in analyzingdry season data The rationale for this is that while dry season rainfall cannot possibly influence the wet season crop the dry season crop is affected by both the rainfall received directly and the rain falling during the preceding wet season through its effect on river discharge reservoir storage and antecedent soil moisture conditions
The level of agricultural production is also an often-used indicator of an irrigation systems performance Its major weakness
- 9 shy
is that a number of factors other than irrigation service such as labor inputs relative prices and fertilizer use influence it It is necessary therefore either to control for changes in the levels of these inputs or assume that they are constant across the two periods being compared In the present case the most important of these factors is the level of application of chemical fertilizer Because fertilizer use by farmers is responsive to the relative pricesof fertilizer and rice it also includes to some extent input and output price effects Since reliable data on fertilizer use for individual systems were not available estimates derived from FAO fertilizer and cultivated rice area data were used t~ control for the effect of changes in the use of this input over time This variable was included in any of the regression equations in which agriculturalproduction was used as the dependent variable Other factors such as labor use genetic potential of varieties sewn and pesticideapplications are assumed to be constant across the two periods
The analytic approach employed is to fit linear regressionequations to pooled data from the five systems covering a eleven-year period 1976 to 1986 A dummy variable is used to check the impact of pre and post 1981 periods on differences in the dependent variable after the effects of factors such as rainfall and nitrogen fertilizer use have been removed In addition because the dataset was created by pooling data from five different systems a set of 4 site dummies was included in the basic model to control for system-specific differences caused by variables which were not measured For some runs these were replaced with dummies that separated reservoir and non-reservoir systems though equations using the reservoir dummy were consistently inferior to those using the complete set of site dummies Several different dependent variables were created to index the quality of irrigation service and tested using this approachRegression results are given in Tables 3 and 5 and discussed below
Performance Indicators
A variety of indicators have been used in evaluating irrigationperformance in various contexts The selection of appropriate indicators depends on a number of factors including the purpose of the evaluation the audience for its results the way in which the boundaries of the irrigation system are defined and the kind and quality of data available to the evaluators The current analysis is designed to evaluate the impact of a set of management changes on system physical performance The audience for this analysis comprises top-level managers of the irrigation agency and policy-makers at higher government levels Boundary definition is an importantanalytic problem here as evident from the subsequent discussion relating to the choice of the appropriate area values to use in scaling system inputs and outputs This issue is also related to the data quality and availability problems which have already been mentioned
- 10 shy
Three fundamental indicators have been proposed for assessing the effectiveness of irrigation services to cultivators (Svendsen and Small 1989 Abernethy 1990) These are the adequacy of water supplies the equity of their distribution across the command area of the system and the timeliness of the supplies Computation of adequacy measures requires information on the total quantity of water delivered to the system over a season on a per hectare basis Equityand timeliness measures require information on the spatial and temporal distribution respectively of those supplies Where appropriate discharge information is not available proxies can be employed by making suitable assumptions Standards must be selected against which the magnitude of the indicators can be judged
The task in the present case however is somewhat different Here the need is to evaluate changes in selected variables between the pre-1981 and the post-1981 periods Hence the absolute values of variables selected are less important than their relative magnitudesand the statistical significance of the differences in magnitudes between the two periods A distinct limitation is imposed by the data series available for the five sample systems Since discharge and yield data are available only on a whole-system basis it is impossible to develop measures of equity and timeliness directly We will however extend our analysis to a discussion of equity byindirect inference Levine and Coward (1986) have argued that equityought to be considered as the paramount objective in managing largepublic irrigation systems They base their conclusion on an analysisof eight small community-managed systems and five larger public systems including UPRIIS in which equity appears to comprise the most important operational objective in the successful systems It may be appropriate therefore to give success in improving equity of distribution added weight in assessing overall performance
Area Estimates Because measures of system agricultural output and water supplied are typically reduced to a unit area basis before being used much depends on the area values which are used to standardize them Two different area measures are available The first is Service Area (SA) which is defined as the irrigable portion of the command area which is provided with physical facilities for water delivery This represents the area which could conceivably be irrigated in a given season if water supply were not constraining This value may change somewhat from year to year in response to urban encroachment on irrigated command minor remodeling and repair and refinements in area estimates In the present case though the systems selected for analysis were chosen to avoid those which had undergone more extensive rehabilitation or modification
The second measure is Benefitted Area (BA) which is the area billed for payment of irrigation service fees It is the irrigated area harvested which did not have yields so low that the farm was exempted from payment of fees in a given season This threshold value has been approximately 2 tons of paddy per hectare Benefitted area
- 11 shy
varies more than does SA particularly during the dry season when available water supply may seriously constrain the area which can be planted Its magnitude is a function of system managers actions in authorizing the amount of land to be planted in a given season farmers decisions regarding whether to plant or not and the combined ability of system managers and farmersirrigators subsequently to distribute water Both of these area measures will be used to standardize other variables for particular purposes as well as beingcombined to form a separate indicator by themselves
Adequacy The most direct measure of the adequacy of irrigation water supplies to the agricultural system is the quantity of water applied to the system command area on a per unit area basis relative to some standard In this case since our interest is in differences in water adequacy between two time periods and since the systems being assessed have been and continue to be almost entirely devoted to rice cultivation during both cropping seasons depth measures for the two periods may be compared directly assuming the seasonal cropdemand for water to be unchanged Although dry-footed crops can suffer yield losses from overapplication of irrigation water rice is largely insensitive to this effect In addition water can substitute for other inputs that the farmer would otherwise have to provide such as weed control and more careful (and costly) water management We assume therefore that other things being equal larger values of depth applied are better than smaller values in terms of meeting crop water demands and reduce the cost of cultivation At the same time high levels of water adequacy can affect the values of other performance measures--particularly equity
When the regression model is run for quantity of water diverted at the system headworks divided by BA hereafter termed depth we see that the period dummy is negative and significant at the 95 percent confidence level for both wet and dry seasons (see Table 3 equations 1 and 3) Since the overall explanatory power of the wet season model is very weak however we will focus on the dry season in interpreting this result which indicates that after adjusting for rainfall differences significantly less water was delivered to the command per unit of benefitted area following 1981 than before This indicates based on the criteria outlined above that performance in terms of water adequacy deteriorated following financial selfshysufficiency We need to examine this conclusion more carefullyhowever
One difficulty is that the measured quantity of water diverted at the source is largely a function of the supply available in the river rather than of system management This is particularly true during the dry season and in non-reservoir systems Thus while the depth of water supplied to the system is a measure of the adequacy of the systems service it is to some extent beyond the control of the managing agency To better understand the factors behind this decline in water availability we look at simple unadjusted index values for
- 12 shy
several of the key variables Table 4 shows annual values of total volume of water delivered in each season SA and BA in both wet and dry seasons and shows the results of t-tests on the means of a set of indicators before and after 1981 Indicators are used rather than the actual values to weight each of the systems equally regardless of its size The table shows that both the average wet season benefitted area and the average discharge are significantlylower during the second period compared with the first For the dry season too the discharge index is lower after 1981 than before but this difference is not significant At the same time the dry season BA index rose slightly but again the change was not significantSince middotthere is not a clear pattern of relative movement of discharge and BA during the respective seasons no simple interpretation of these index value changes is possible What stands out is that both discharge and benefitted area declined across periods during the wet season while during the dry season there was no significant change in either indicator across the two periods It seems clear that the decline in water adequacy must be evaluated together with other measures of performance in drawing conclusions about the overall impact of the 1981 changes on the quality of system management
Another measured variable per hectare yield can be used as a proxy for water adequacy It has the advantage of partiallyreflecting the impacts of the dimensions of timeliness6 and equity7of distribution as well integrating all three effects into a combined impact on aggregate crop production Table 5 (equations 1 and 3)shows that the period dummy in the yield regressions has a positive sign in both seasons after controlling for nitrogen application and precipitation though the t-values are not significant Treatingyield adjusted in this way as a proxy for quality of irrigationservice leads to the conclusion that by this more comprehensive measure quality of service held constant across the two periods in the dry season Because of the large yield component accounted for byrainfall during the wet season no such judgement is possible for that season however
Equity As noted earlier no reliable data are available for subdivisions of the five sample systems making direct computation of equity measures impossible We can make some judgements about changesin the equity of water distribution however by examining changes in the ratio of two area measures given for each system SA and BA Since SA is the area which theoretically can be supplied with irrigation water by the system and BA s the area which actuallyreceives a quantity of water adequate to produce a remunerative crop the ratio of the two provides a measure of the percentage of the potential service area which was irrigated to a particular standard The larger this pErcentage the more equitable 8 is the distribution This of course assumes that the quantity of water available to the systems is constant across the two periods
- 13 shy
Since the condition of constancy of water supply is not generally satisfied a regression was run in which the total quantityof water diverted at the headworks of each system divided by the systems potential service area SA was included in the regression to control for changes in the water supply available to the system seeTable 3 equations 5 through 8 The average daily rainfall received directly on the system service area during the season was also included as an independent variable The regression was run separately for wet and dry seasons The sign and t-statistic of the period dummy should then tell us whether or not equity as reflected in the BASA ratio increased decreased or remained unchanged across the period divide
Both equations are reasonable good as indicated by the R2 values though the dry season equation is considerably better as would be expected For the wet season both the water delivery term and the rainfall term in equation 5 are of positive sign but are nonshysignificant at the 95 percent confidence level indicating that wet season irrigated area does not change appreciably in response to level of wet season rainfall or the available irrigation water supply The period dummy was negative but not significant indicating that equityof distribution as reflected in the BASA ratio was similar during the two periods
For the dry season the water delivery term in equation 7 is positive and strongly significant indicating a close relationship between the fraction of potential area actually irrigated and the water supply available at the headworks In addition however the period dummy is positive and significant suggesting that once the influence of water supply is removed the BASA ratio was significantly higher in the period following 1981 than it was before
This is an important finding for it reflects significantlyimproved performance in terms of a factor equity of water distribution that is under the control of the managing entity an entity which here comprises both NIA and irrigators associations Interpreted in these terms NIA and allied farmers associations were able to spread a given amount of water more widely across the potential command area of the five sample systems in the period after 1981 than before Moreover they did this in a way that did not decrease average system yields as discussed earlier In making this interpretation we are suggesting that there was some redistribution of water from better-watered areas to fringe areas which would otherwise not have received irrigation water and that this redistribution was a direct response to the change in NIA prioritiesand operating policies and rules occurring around 1981
It is difficult to prove the assertion that water was in fact redistributed with a resulting increase in directly-measured equity Without access to reliable discharge data broken out by systemsection and we can only assume in the absence of a plausible
- 14 shy
alternative explanation that it was such a redistribution that made the increase in the BASA ratio possible In a larger sense it is difficult to prove conclusively that any outcome in a before and after analysis was the result of a particular independent causative factor In this case we have tried to remove the influence of other potential causative factors where we could but the possibilityremains that some combination of unmeasured factors are responsiblefor the difference in the BASA ratio found We do note though that this type of response is exactly the type that would be expected to follow from an emphasis on increased farmer satisfaction and cooperation and increased fee revenues Because the fee schedule is tied to benefitted area the only ways NIA can increase its revenue from that source are to expand benefitted area and to increase collection efficiencies The former depends on redistributing a fixed supply of water over a larger portion of the command while the latter requires that farmers be satisfied with the irrigation service they are receiving and the commitment of the local irrigators association to assist in the task of collecting the amounts due The evidence while not conclusive is highly suggestive that this is exactly what has happened
Efficiency In addition to measures which reflect the levels of adequacy and equity of irrigation service available data allow the calculation of a measure of operating efficiency The term efficiencyusually denotes the relationship between inputs to a process and its outputs often expressed as a ratio The output measure employed here is aggregate system rice output and the input is quantity of irrigation water turned into the system Dividing the first by the second gives a measure of agricultural production per unit water--here termed specific yield This is a highly integrated measure that evaluates the combined efficiency of the irrigation and agricultural processes As such it is a function of the managerial and other inputs supplied both to the irrigation system and to the agricultural operation With respect to one important input to the irrigation system we do know that NIA per hectare field operating expenses were about 29 percent lower in real terms in the 1982-86 period comparedto the 1976-1981 period although this drop may have been partlyoffset by increases in farmer-supplied labor inputs Other things being equal one would thus expect to find a decline in output efficiency
The regression analysis shows positive signs for the period terms in both wet and dry season equations (see Table 5) In the case of the wet season the period dummy in equation 5 is significant but the overall explanatory power of the model is quite low For the dry season (equation 7) the coefficient is positive but non-significant This means that after taking rainfall and fertilizer use into account data do not indicate a lowering of specific yield in the wake of funding reductions and the strong emphasiS on financial viabilitybeginning in 1981 This result provides evidence that the efficiency of the overall irrigation deliveryagricultural production process
- 15 shy
relative to the system water input did not falloff as a result of the changes implemented at least over the short run
Impact magnitude
The preceding analysis has shown us that some indicators of irrigation performance changed significantly following the managerial changes of 1981 while others did not However it has not given us a sense of the size of the changes which occurred To determine the magnitude of these changes the regression model is used to predict the response of the composite system to the managerial changes given a common set of -input and environmental factors To do this averagevalues of the independent variables from the entire eleven-year period 1976 to 1986 are put into the model together with the previously determined coefficients to generate predicted average values of the various dependent variables used in the earlier analysis with and without the period dummy This procedure produces a pair of estimates for each dependent variable under the same conditions--one in which the system responds as it did after the managerial changes were implemented and one in which it responds as it did prior to their introduction The differences between these two values thus indicate the magnitude of the changes occurring in the various indicators of performance discussed above
The results of this exercise are shown in Table 6 The table shows that water availability decreased by about 13 percent in both wet and dry seasons when the period dummy was included and while the coefficients responsible were significant in the earlier analysisthis difference cannot be easily connected with levels of system management as discussed earlier With respect to rice output per hectare although the coefficients were not very significant it is interesting to note that yield increases by 163 kilograms per hectare for the wet season and by 101 kilogram for the dry when the period dummy is included in spite of the reduced water supply available Keep in mind that the predicted yield values have already been adjusted for differences in nitrogen fertilizer use and rainfall This suggests that timeliness and equity of distribution of water supply to farmers may have increased following the changes contributing to the higher predicted yields
Examining the impact of increased equity of distribution bylooking at the ratio of benefitted area to service area we recall that the change was positive and significant for the dry season and negative and not significant for the wet Table 6 shows that the dry season BASA ratio increases by 7 percentage points when the dummy is included a 131 percent increase Other things being equal this should result in a 131 percent increase in system output due to the expansion of ared benefitted This is a major impact on production
- 16 shy
CONCLUSIONS
The Philippine experiment to transform the national irrigation agency into an enterprise has undoubtedly been successful in reducing system operating expenses bringing revenues and costs into line and eliminating the recurrent cost burden imposed by large-scaleirrigation systems on the national budget Evidence presented in this paper indicates that in the process equity of water distribution across systems has also improved In the 5 years following the cessation of operating subsidies from the government an index of equity of distribution improved by about 13 percent At the same time per hectare yields adjusted for rainfall and nitrogen application held constant
There is a strong logical connection between the achievement of financial viability and improved equity of water distribution across the command Because increasing irrigation fees is a politicaldecisionlying largely beyond NIAs control expanding the area which can be billed for service is one of the few revenue increasing measures available to the irrigation agency which does not involve major additional investment In the face of constant or shrinkingwater supplies this is achieved only by redistributing water from areas receiving excessive supplies usually near the head ends of canals and laterals to areas receiving no supplies or inadequatesupplies often located near the tails of canals Although data are not available which would allow the direct examination of this hypotheses the two outcomes are logically consistent with each other
Data also show that per hectare water deliveries declined significantly in the five sample systems after 1981 even thoughrainfall did not differ appreciably between the two periods This decline averaged about 13 percent for both wet and dry seasons and is interpreted as a decline in water availability in the supplying rivers rather than a conscious reduction in withdrawals by system managers Such declines could result from changes in watershed runoff characteristics as caused by deforestation or from increased upstream abstractions from supplying rivers
Improved water distribution tends to increase the area served system agricultural output and NIA service fee revenue Reduced water supplies to the system tend to reduce these things Specificyield defined as system paddy output per unit water held roughly constant across the two periods indicating that the two effects mayhave offset each other
After adjusting for rainfall and nitrogen application perhectare yields increased only marginally in the post-1981 period Area served on the other hand increased by about 13 percent after adjusting for water supply availability indicating that the area benefitted by irrigation in the sample systems increased by about the same percentage Even if yields on this additional area are less than
- 17 shy
average yields for the system this still represents a sizeable increase in system agricultural output as a result of the change in management structure the increase coming not from higher yields but from expanded area under irrigation
The evidence assembled here suggests that there are significantfinancial and economic benefits to be had from changes in the basic character of irrigation managing agencies which make them more responsive to their clientele and which impose rational internal financial discipline on the agency The analysis suggests a number of additional questions however One relates to the longer-term impacts of the structural management changes The improvements in water distribution described here are relatively short-term events occurring during the first 5 years of the new management mode Critics have suggested the danger of underinvestment in systemmaintenance over the longer run accompanied by declining yields and benefitted areas and eventual system collapse This possibility needs to be closely monitored A second concern relates to the apparent decline in water supply to these 5 geographically dispersed systems The nature and causes of this decline need to be explored further since if widespread and secular it may represent a serious threat to the stability of Philippine rice production Whether stemming from poor forest management practices or deficient regulation and allocation of surface water resources or other unidentified factors it is an issue that deserves serious and urgent consideration
A third risk is that the incentive structure set up by NIA to guide and stimulate the performance of field units overemphasizes revenue generation at the expense of irrigation service provision to farmers The evidence presented here supports the view that these two objectives are mutually reinforcing under policies and conditions which have been established in the Philippines More detailed crossshysectional studies based on primary flow measurement data would add confidence to this conclusion and help to specify the conditions under which this effect occurs This could be extremely important in transferring the results of the Philippine experiment to other countries
A final risk is that outside intervention well meaning or otherwise will destroy the basis of NIAs financial autonomy or will impose external pressures or constraints on NIAs decision-making that will subvert the management practices which have been so painstakinglydeveloped and implemented Among these are calls for NIA to be subsumed again within the government department structure in the interests of better coordination with agriculture attempts byexternal financing agencies to arbitrarily increase NIAs expenditures on OampM on the assumption that this will increase system agricultural output or intervention by Philippine legislative bodies to restore operating subsidies to NIA with attached strings leading back to legislators home districts Pressures such as these will cut short a process of experimentation and improvement that seems promising
- 18 shy
enough to date to warrant its continuation Having developed the capacity to establish targets and implement and manage change NIA is in a strong position to modify its objectives to better achieve larger social purposes established for it It is critical to recognize however that this must happen within the context of financing policies that mandate financial autonomy for NIA if the fundamental institutional commitment to manage is to be preserved
The author would like to thank Leslie Small and JeremyBerkoff for helpful comments on an earlier unpublishedversion of this paper and Charles Rogers for his careful and creative help with the analysis
- 19 shy
BIBLIOGRAPHY
Abernethy Charles L 1990 Indicators of the performance of irrigation water distribution systems International Irrigation Management Institute Colombo Sri Lanka Mimeo
Asian Development Bank 1986 Irrigation service fees Proceedingsof the Regional Seminar on Irrigation Service Fees Manila Asian Development Bank
Carruthers Ian and Colin Clark 1981 Economics of IrrigationLiverpool Liverpool University Press Third Edition
Levine G and EW Coward Jr 1986 Irrigation water distribution implications for design and operation AGREP Division WorkingPaper 125 vol 1 World Bank Agriculture and Rural Development Department
Small Les E 1989 User charges in irrigation potentials and limitations Irrigation and drainage vol 3 no 2125-142
Small Les 1990 Irrigation service fees in Asia IrrigationManagement Network 9013 London Overseas DevelopmentInstitute
Svendsen Mark and Les Small 1989 A framework for assessing irrigation system performance Paper prepared for the Symposium on Performance Evaluation 23 November International IrrigationManagement Institute Sri Lanka
Table I--National Irrigation Administration revenues and expenditures in constant prices 1976-86
Item 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986
(peso million 1972)
Revenues Irrigation fees collected 1273 1483 17 13 1831 2070 1668 1699 1893 1728 2129 2546 Other income 715 737 2420 5591 2631 5990 7783 6638 5699 4932 2934
Total direct revenue 1988 2220 4133 7422 4701 7658 9482 8531 7427 061 5480
Expenses in 1972 pricesTotal expenses 4825 5716 5039 6329 3821 77 55 6166 4749 4348 4259 4959
Excess (deficit) (2837(3496) (906) 1093 877 (097) 3316 3782 3079 2802 521 N 0
Subsidies Government operation and
maintenance subsidies 2521 2741 2799 1817 1398 633 0 0 0 0 0 Calamity fund payments 548 0 0 0 0 0 0 0 119 0 142
Total subsidy 3069 2741 2799 1817 1398 633 0 0 119 0 142
Total excess (deficit) 231 (754) 1893 2910 2275 536 3316 3782 3198 2802 663
Source IFPRI analysis of NIA data
- 21 shy
Table 2--Descriptive characteristics of selected MIA systeasa
------- Region III -------- ----- Region VI ----shyUPRIIS Angatii Sto Sibalom- Aganan-
Haasim Thomasc San Jose Sta Barbara
Average service area (hal 102272 31462 3522 5282 8703 Average irrigated area (hal
Wet season 83768 23454 3007 4410 8300 Dry season
Average benefited area Wet season
(hal 64587
77 605
27639
22908
1 781
3007
2801
4369
2770
7698 Dry season
Average rainfall Wet season
(mml d 62478
1 685 5
27396
8576
1 781
3051 0
2769
24731
2997
20001 Dry season 756 333 322 2828 3025
Average discharge (Llsec) Wet season 46501 14792 1692 2353 4984 Dry season 78091 22812 2014 1276 2315
Average water delivery (mmday) Wet season 522 548 487 462 571 Dry season
Average yield (mtha) 1089 715 995 398 686
Wet season 345 419 322 395 435 Dry season
Avg yield per unit water 34 03
(kgm ) 451 412 399 426
Wet season 0373 0440 0373 0538 0443 Dry season 0248 0400 0279 0690 0428
t-statistic difference in mean rainfall 1978-81 1982-86e
Wet season 0432 0713 -0567 1169 1169 Dry season 0519 -0230 -0523 1187 1187 Annual 0460 0707 -0686 1445 1445
~ Summary numbers are averages for the period 1982-1986 except as noted Water delivery discharge and yield per unit discharge are 4-year averages 1982-1985
c Water delivery discharge and yield per unit discharge are 4-year averages d 1983-1986
For Angat 5 years are 1981-85 For St Thomas 1979-83 For Sibalom 1971-75 e No significant differences at 95 confidence
- 22 shy
Table 3--Results of pooled regressions water delivery indicators
Dependent variable Independent variable - Water Delivery in mmBenefited Ha - --- BenefitService Area Ratio --shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season-shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 6799 (998)
6189 (1019)
13102 (1093)
2411 (1 54)
0743 ( 1186)
0800 (1165)
0129 (131 )
0258 (329)
System 2 indicator 0173 (029)
-3743 (-363)
-0063 (-219)
0200 (505)
System 3 indicator 0698 (085)
0163 (012)
0085 (208)
-0046 (-097)
System 4 indicator -0514 (-l 07)
-7011 (-803)
0080 (340)
0169 (l87)
System 5 indicator 0177 (033)
-5049 (-538)
0174 (603)
0051 (058)
Reservoir indicator 0327 (078)
5703 (543)
-0126 (-510)
0156 (261)
Period indicatora -0808 (-240)
-0771 (-229)
-1214 (-209)
-1309 (-1 62)
-0036 (-192)
-0025 (-112)
0071 (239)
0030 (077)
Water delivery mmb
Wet season
Dry season
Precipitation in mm Wet season
Annual
Average daily rainfall Wet season
Dry season
-46E-4 (-125)
-1 9E-4 (-082)
-43E-4 (-077)
20E-3 (385)
0015 (1 59)
-0001 ( -004)
0023 (215)
0001 (021)
0070 (613)
-0017 (-058)
0046 (371)
0010 (033)
R2 (adj) number of observations degrees of freedom
0156 43 36
0133 43 39
0711 41 34
0434 41 37
0693 42 34
0556 42 37
0865 40 32
0730 40 35
Coefficient (t statistic) indicates significance at 95 confidence ~ 1982-1986 1
per ha of service area
Ta
ble
--I
nd
ices
of
se
v
ice a
rea
ib
en
ef
ited
a
rea
a
nd
a
vera
ge sea
so
na
l d
isch
arg
e
Ave
rage
A
vera
ge
tshy19
77
1978
19
79
1980
19
81
1982
19
83
1984
19
85
1986
19
77-8
1 19
82-8
6 S
tati
stic
a
(ind
ex
aver
age
1983
-198
5 =
100)
Ser
vice
are
a UP
RI IS
in
dex
882
91
9
920
91
5
951
95
1
100
0 10
00
100
0 10
00
917
99
0
bull5
53
Ang
at-M
aasi
m R
95
9
994
99
7
996
99
6
996
10
00
100
0 10
00
100
0 98
8
999
1
63
Sto
To
mas
10
63
106
3 10
1 9
10
29
103
0 99
8
100
0 10
00
100
0 11
10
104
1 10
22
-08
9
Siba
lom
-San
Jos
e 95
0
872
94
2
933
94
2
942
10
28
102
8 94
4
101
7
928
99
2
292
A
gana
n-St
a B
arba
ra
108
3 10
91
106
0 10
30
961
10
05
996
10
08
996
99
6
104
5 10
00
-21
1 A
vera
ge
987
98
8
981
98
1
916
97
8
100
5 10
07
988
10
25
984
10
01
123
Wet
seas
on b
enef
ited
are
a in
dex
UPR
IIS
110
0 10
07
114
4 10
55
113
8 11
74
951
10
71
971
ll
58
10
89
106
7 -0
47
Ang
at-M
aasi
m R
97
9
974
92
0
983
10
28
100
8 99
7
102
1 98
2
931
97
7
988
0
54
Sto
To
mas
11
63
115
9 11
23
107
5 10
80
103
9 98
1
978
10
41
103
9 11
20
101
6 -4
88
Siba
lom
-San
Jos
e 11
35
103
8 10
11
931
93
4
906
10
07
975
10
1S
998
10
11
981
-0
80
Aga
nan-
Sta
Bar
bara
10
85
110
5 10
74
106
8 10
01
104
2 99
8
101
5
987
61
0
106
7 93
0
-1 8
4
Ave
rage
10
92
105
6 10
54
102
4 10
36
103
4 98
7
101
3 10
00
947
10
53
996
-2
32
Dry
sea
Son
bene
fite
d ar
ea
inde
x U
PRIIS
14
04
155
0 15
S0
155
8 16
17
128
0 57
2
114
S 15
74
152
3 12
38
-09
4 A
ngat
-Maa
sim
R
903
93
0
103
2 10
61
104
2 10
69
988
99
2
102
1 99
6
993
10
13
061
S
to
Tom
as
105
7 12
27
122
5
961
99
9
115
9 10
1 7
91
0
107
3 12
1 2
10
94
107
4 -0
28
Si
balo
m-S
an J
ose
Aga
nan-
Sta
Bar
bara
66
5
95S
62
6
632
67
4
111
4
501
10
S1
412
11
51
766
93
3
856
94
3
107
0 94
8
107
4 11
09
111
6
158
6 58
8
987
97
6
110
4 5
35
083
N
w
Ave
rage
89
6
964
11
19
103
7 10
44
110
9 10
1 7
89
8
108
5 12
97
101
7 10
81
082
Wet
seas
on d
isch
arge
in
dex
UPR
IIS
132
9 72
7
142
5 11
88
120
4 98
8
105
8 10
63
879
96
4
117
5 99
1
-16
5 A
ngat
-Maa
sim
R
129
7 13
52
134
5 12
58
127
0 11
70
120
3 62
7
131
3 10
68
-18
9
Sto
To
mas
14
71
155
0 14
1 6
10
40
725
12
35
112
3 14
79
103
1 -4
48
Siba
lom
-San
Jos
e 96
8
733
11
1 5
92
2
907
47
1
102
3 85
7
ll2
0
422
92
9
779
-1
09
Aga
nan-
Sta
Bar
bara
87
9
863
68
1
925
96
8
110
7 70
5
871
87
7
00
9
Ave
rage
11
49
105
7 13
61
115
0 10
58
853
10
43
963
99
4
SO3
11
55
940
-2
85
Dry
sea
son
disc
harg
e in
dex
UPR
IIS
425
13
09
153
3 14
28
180
6 14
S0
125
8 56
3
117
9 14
66
130
0 11
89
-04
3
Ang
at-M
aasi
m R
12
26
161
6 15
30
139
2 12
85
100
8 10
33
958
14
41
107
1 -3
80
S
to
Tom
as
116
4 14
82
155
7 79
9
971
12
30
126
5 14
01
106
6 -2
44
Si
balo
m-S
an J
ose
486
71
5
782
62
9
789
64
5
116
9 11
86
801
65
3
918
2
36
Aga
nan-
Sta
Bar
bara
A
vera
ge
425
10
46
133
6 62
0
118
4 81
5
116
1 94
7
112
5 89
7
922
12
20
991
88
2
108
7 12
36
119
2 71
8
114
0 10
37
105
5 3
23
-07
5
a bull
indi
cate
s si
gnif
ican
ce a
t 95
per
cent
con
fide
nce
- 24 shy
Table 5--Results of pooled regressions yield indicators
------------------------- Dependent variables -------------------------shyIndependent variable ---------- Yield MTHa ----------- -- Specific Yield 1000 KgM
3 H20 shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season -shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 1967 (286)
3747 (580)
1991 (293)
2800 (460 )
0084 (041)
0441 (229)
0049 (023 )
0876 (343)
System 2 indicator 0400 (155)
0525 (226)
0057 (085)
0115 (1 95)
System 3 indicator 0466 (139 )
-0029 (-010)
-0016 (-018)
0006 (008)
System 4 indicator 0980 (470)
0090 (049)
0176 (333)
0391 (827)
System 5 indicator 1 221 (570)
0256 ( 138)
0104 (174)
0224 (440)
Reservoir indicator -1033 (-578)
-0063 (-039)
-0146 (-310)
-0326 (-569)
Period indicatora
Precipitation in mm Wet season
Annual
0163 (1 03)
-29E-4 (-182)
0231 (1 42)
-58E-4 (-606)
0161 (118)
92E-5 (072)
0199 043 )
-64pound-5 (-084)
0082 (200)
57E-6 (013)
0092 (223)
-61E-5 (-232)
0032 (093)
29E-7 (001)
0052 (112)
-12pound-4 (-414)
Nitrogen applicationb 0025 (254)
0021 (211)
0025 (274)
0020 (227)
0003 (1 05)
0002 (043)
0003 (101)
-0002 (-052)
R2 (adj) 0553 0516 0251 0198 0309 0262 0713 0445 number of observations 50 50 49 49 42 42 40 40 degrees of freedom 42 45 41 44 34 37 32 35
CoeffiCient (t statistic) indicates significance at 95 confidence ~ 1982-1986 = 1
kgha
- 25 shy
Table 6--Suggmry of estiates perforance indicators
Variable Predicted Predicted Period Dependent Variable
Description (units)
Period Coefficient
1976-86 IndicatorO
1976-86 Indicatorl
Effecta (X)
BenefittedService area ratio
Wet season ratio -0036 0853 0817 -425 Dry Season 0071 0539 0610 1313
Delivery Wet Dry
season season
nm water per benefitted ha
-0808 -1 214
6003 9083
5195 7869
-1346 -1337
Rice Yield Wet season metric tons 0163 3578 3741 456 Dry season per hectare 0161 3905 4066 412
Specific Yield Wet season kg rice per m3 0082 0348 0430 2353 Dry season water de 1 ivered 0032 0378 0409 837
a percentage changes relative to period=O values
- 26 shy
ENDNOTES
IResearch Fellow International Food Policy Research Institute and Senior Irrigation Specialist International Irrigation Management Institute
2Presumable farmers also base private investment decisions on such foreknowledge of costs and commonly do so in the case of small tubewells In addition user groups may invest in small communal schemes independently after examining projected costs and benefits Most commonly however new irrigation capacity is created as a result of public investment decisions made by government and thus the political process comes into play
3This principle applies only when water charges are levied volumetrically--an exceedingly rare situation in the developing worldshy-though the argument is frequently invoked as though it applies to all allocative situations
4For Cabanatuan City the longer period of record available ie 1904 to 1986 was divided into two equal periods and tested for difference in means to test the observation by farmers that rainfall today is lower than it used to be Mean precipitation after 1947 was actually about 3 percent higher than before but the difference was not significant (t = 07)
5The procedure used was to take time series data on total nitrogenfertilizer consumption in the Philippines multiply it by the ratio of N fertilizer used on rice to total N consumption and divide the product by the rice harvested area to obtain an estimate of N use perhectare Values of the ratio of N use on rice to total N use were taken from IRRI (1988 150) with interpolation used to fill in gaps in the series
6This assumes that timeliness is evaluated against a standard based on the physiological demand for water generated by the rice crop and the varying sensitivity of the crop to stress at different growth stages In such a case a poor timeliness rating for irrigationservice will result in reductions in yield
7Because rice yields are insensitive to water applications that exceed PET values a redistribution of water from water excess to water short portions of the command area will usually result in higher aggregate output and higher average yields Exceptional scenarios can be constructed in which this general rule would be violated but such scenarios are unlikely to play out in practice
- 27 shy
aThere are a variety of standards for judging equity that could be used here The most common one is equality of per hectare water deliveries In the present case the implicit standard is equality of water supply utility in producing a crop Thus if two areas received equal amounts of water on a per hectare basis but one had such high seepage and percolation losses that the crop failed to produce a harvest (and was therefore exempted from irrigation fee payment) the water supply to the combined area would be judged inequitable Because such radical differences in seepage and percolation rates are unlikely on puddled rice soils in the lowland Philippines and because the definition of benefitted area accommodates a wide range of acceptable yields these two definitions are largely indistinguishable at the current limits of data resolution
- 6 shy
moneylenders do Since that time the agency has made strenuous efforts to increase its fee collections The net effect has been to hold fee revenues per hectare constant in the face of a national rice support price that has steadily declined relative to more generalindices of inflation
At the same time NIA also took steps to reduce its operating expenses Although a number of minor measures of economy were mandated initially the fact that more than three-quarters of the operating budget was devoted to personnel costs meant that any real savings would require reductions in staffing levels Voluntaryreductions were carried out in the late 1970s and early 1980s resulting in a decrease in the number of staff per hectare and a reduction of the personnel share of the budget from 80 percent in 1976 to about 74 percent in 1986
With an eye on its bottom line NIA also instituted a system of performance grants for all field units and the individuals in them termed Viability Incentive Grants To facilitate this each largeirrigation system in the country was made a separate cost center to allow costs and revenues to be accounted for on a system-by-system basis This program provided that once a unit achieved a net excess of revenues over operating costs in a given year a fraction of the surplus would be shared among the units personnel Five of the 11 irrigation regions of the country were receiving these incentive payments by 1986 as were 53 of the 120 individual systems included within the 11 regions
Effects The financial results of these efforts are shown in Table 1 If subsidies are not considered NIA first achieved net profitabilityin 1979 and retained it through the end of the period studied except for a small deficit incurred in 1981 Subsidies were eliminated in 1982 (except for occasional small calamity grants following typhoons) Although revenues have declined in recent years due largely to decreases in interest earnings and construction management fees expenses have declined more rapidly resulting in a series of net positive balances
Achieving a financially viable position is an importantaccomplishment few other irrigation agencies in the developing world have been able to do this However there is some risk that such an achievement occurs at the expense of the quality of service provided to clients Some of the most interesting and important consequences of the new cost recovery policies therefore relate to the physical performance of the irrigation systems NIA operates It is here that the end objectives of the irrigation investments are realized and where the lives of the farmers who till system lands are affected The remainder of the paper will examine and attempt to quantify the impact of these policy changes on physical irrigation systemperformance
- 7 shy
IMPACTS ON SYSTEM PERFORMANCE
Study Methodology
Unfortunately the kinds of changes in hydrologic system outputand the impact on agricultural performance which might be expected to result from improved institutional performance are quite difficult to capture and quantify There are a number of reasons for this First there is the year to year variability of system performance caused byvariable rainfall which feeds rivers fills reservoirs and supplements irrigation water in supplying crop requirements Second there is the difficulty of defining just what performance is and specifying how to measure it Third and most important the regularly collected data from which indicators can be constructed are limited in type number of measuring points and period of record and are sometimes of doubtful reliability
These difficulties notWithstanding an attempt was made to determine the impact that changes in operating procedures staffinglevels and incentive programs had on system performance Because the effects that we are trying to assess resulted from changes that affected all of the systems under NIAs direct authority there are no control systems which can be used as standards We are forced therefore to rely on a comparison of values of selected performance indicators before and after the date of the major structural and procedural changes which is taken to be 1981
To accomplish this secondary data were assembled for 5 systems in Administrative Regions III and VI which had not undergonesignificant phYSical changes during the period of analysis Duringthis process several site visits were made by study team members Time series data collected include service area (SA) and benefitted area (BA) for both wet and dry seasons yields for wet and dry seasons monthly main canal discharge at the system headworks and monthly precipitation The general period of availability for this data is 1966-86 though for systems which began operation after 1966 the period of record is shorter and the records of some systemscontain miSSing values These five systems their 1986 service and benefitted areas and other descriptive data are shown in Table 2
The prinCipal problem with using a before and after approachrather than one that considers comparable systems with and without the innovation is that some of the measured difference in effects mayhave resulted from causes which are independent of the ones beingstudied These causes can be specific in which case they may be relatively easy to identify and accommodate or more general and diffuse and therefore more difficult to control for In the present case--that of changes in the performance of NIA irrigation systemsresulting from the major organizational changes in 1981 two principalexternal factors can be identified which might be expected to affect differences in measured levels of irrigation performance between the
- 8 shy
two time periods These are rainfall and level of use of other agricultural inputs Since our interest is in systems managerial responses to these changes and since there is no reason to believe that the relative magnitudes of the responses are dependent on the size of the system the five systems are treated as equal in the analysis That is changes in measured values relating to the smallest system are considered to be as important as changes in values for the largest with no area weighting applied
Since rainfall can substitute for irrigation water supplies and since it affects the supply of water available in rivers for irrigation it may exert some independent influence on various performance indicators To test the strength of the relationship for the period being analyzed simple correlations were run between rainfall and benefitted area for one system in each regionBenefitted area was used in this analysis because it is the variable deemed most likely to be influenced by year-to-year changes in rainfall Weather data from Cabanatuan City was used for the UPRIIS system which surrounds it and Iloilo City data was used for the nearby Aganan-Santa Barbara system For UPRIIS all of the R2 values for these correlations were less than 0005 suggesting that rainfall has almost no impact on area harvested in this large reservoir-based scheme For Aganan-Santa Barbara wet season rainfall was related to wet season BA (r2 = 016) and to BA during the following dry season (r2 = 024) Signs of the simple correlations were in the expected directions ie wet season rainfall increased BA during both the wet and the subsequent dry seasons These connections are understandable but weak
Another possibility is that there were longer-term differences in rainfall received in the two regions If this were the case a comparison of performance during two different time periods would have to take this difference into account Differences in average precipitation during the two periods were examined for the four stations used in the analysis (see Table 2) In no case were differences in seasonal or annual mean rainfall statisticallysignificant4
bull
Nevertheless in the regression approach adopted to analyze the data rainfall was included in each equation to control for its possible effect on the particular dependent variables being analyzed In doing this wet season rainfall was used in analyzing wet season performance indicators while annual rainfall was used in analyzingdry season data The rationale for this is that while dry season rainfall cannot possibly influence the wet season crop the dry season crop is affected by both the rainfall received directly and the rain falling during the preceding wet season through its effect on river discharge reservoir storage and antecedent soil moisture conditions
The level of agricultural production is also an often-used indicator of an irrigation systems performance Its major weakness
- 9 shy
is that a number of factors other than irrigation service such as labor inputs relative prices and fertilizer use influence it It is necessary therefore either to control for changes in the levels of these inputs or assume that they are constant across the two periods being compared In the present case the most important of these factors is the level of application of chemical fertilizer Because fertilizer use by farmers is responsive to the relative pricesof fertilizer and rice it also includes to some extent input and output price effects Since reliable data on fertilizer use for individual systems were not available estimates derived from FAO fertilizer and cultivated rice area data were used t~ control for the effect of changes in the use of this input over time This variable was included in any of the regression equations in which agriculturalproduction was used as the dependent variable Other factors such as labor use genetic potential of varieties sewn and pesticideapplications are assumed to be constant across the two periods
The analytic approach employed is to fit linear regressionequations to pooled data from the five systems covering a eleven-year period 1976 to 1986 A dummy variable is used to check the impact of pre and post 1981 periods on differences in the dependent variable after the effects of factors such as rainfall and nitrogen fertilizer use have been removed In addition because the dataset was created by pooling data from five different systems a set of 4 site dummies was included in the basic model to control for system-specific differences caused by variables which were not measured For some runs these were replaced with dummies that separated reservoir and non-reservoir systems though equations using the reservoir dummy were consistently inferior to those using the complete set of site dummies Several different dependent variables were created to index the quality of irrigation service and tested using this approachRegression results are given in Tables 3 and 5 and discussed below
Performance Indicators
A variety of indicators have been used in evaluating irrigationperformance in various contexts The selection of appropriate indicators depends on a number of factors including the purpose of the evaluation the audience for its results the way in which the boundaries of the irrigation system are defined and the kind and quality of data available to the evaluators The current analysis is designed to evaluate the impact of a set of management changes on system physical performance The audience for this analysis comprises top-level managers of the irrigation agency and policy-makers at higher government levels Boundary definition is an importantanalytic problem here as evident from the subsequent discussion relating to the choice of the appropriate area values to use in scaling system inputs and outputs This issue is also related to the data quality and availability problems which have already been mentioned
- 10 shy
Three fundamental indicators have been proposed for assessing the effectiveness of irrigation services to cultivators (Svendsen and Small 1989 Abernethy 1990) These are the adequacy of water supplies the equity of their distribution across the command area of the system and the timeliness of the supplies Computation of adequacy measures requires information on the total quantity of water delivered to the system over a season on a per hectare basis Equityand timeliness measures require information on the spatial and temporal distribution respectively of those supplies Where appropriate discharge information is not available proxies can be employed by making suitable assumptions Standards must be selected against which the magnitude of the indicators can be judged
The task in the present case however is somewhat different Here the need is to evaluate changes in selected variables between the pre-1981 and the post-1981 periods Hence the absolute values of variables selected are less important than their relative magnitudesand the statistical significance of the differences in magnitudes between the two periods A distinct limitation is imposed by the data series available for the five sample systems Since discharge and yield data are available only on a whole-system basis it is impossible to develop measures of equity and timeliness directly We will however extend our analysis to a discussion of equity byindirect inference Levine and Coward (1986) have argued that equityought to be considered as the paramount objective in managing largepublic irrigation systems They base their conclusion on an analysisof eight small community-managed systems and five larger public systems including UPRIIS in which equity appears to comprise the most important operational objective in the successful systems It may be appropriate therefore to give success in improving equity of distribution added weight in assessing overall performance
Area Estimates Because measures of system agricultural output and water supplied are typically reduced to a unit area basis before being used much depends on the area values which are used to standardize them Two different area measures are available The first is Service Area (SA) which is defined as the irrigable portion of the command area which is provided with physical facilities for water delivery This represents the area which could conceivably be irrigated in a given season if water supply were not constraining This value may change somewhat from year to year in response to urban encroachment on irrigated command minor remodeling and repair and refinements in area estimates In the present case though the systems selected for analysis were chosen to avoid those which had undergone more extensive rehabilitation or modification
The second measure is Benefitted Area (BA) which is the area billed for payment of irrigation service fees It is the irrigated area harvested which did not have yields so low that the farm was exempted from payment of fees in a given season This threshold value has been approximately 2 tons of paddy per hectare Benefitted area
- 11 shy
varies more than does SA particularly during the dry season when available water supply may seriously constrain the area which can be planted Its magnitude is a function of system managers actions in authorizing the amount of land to be planted in a given season farmers decisions regarding whether to plant or not and the combined ability of system managers and farmersirrigators subsequently to distribute water Both of these area measures will be used to standardize other variables for particular purposes as well as beingcombined to form a separate indicator by themselves
Adequacy The most direct measure of the adequacy of irrigation water supplies to the agricultural system is the quantity of water applied to the system command area on a per unit area basis relative to some standard In this case since our interest is in differences in water adequacy between two time periods and since the systems being assessed have been and continue to be almost entirely devoted to rice cultivation during both cropping seasons depth measures for the two periods may be compared directly assuming the seasonal cropdemand for water to be unchanged Although dry-footed crops can suffer yield losses from overapplication of irrigation water rice is largely insensitive to this effect In addition water can substitute for other inputs that the farmer would otherwise have to provide such as weed control and more careful (and costly) water management We assume therefore that other things being equal larger values of depth applied are better than smaller values in terms of meeting crop water demands and reduce the cost of cultivation At the same time high levels of water adequacy can affect the values of other performance measures--particularly equity
When the regression model is run for quantity of water diverted at the system headworks divided by BA hereafter termed depth we see that the period dummy is negative and significant at the 95 percent confidence level for both wet and dry seasons (see Table 3 equations 1 and 3) Since the overall explanatory power of the wet season model is very weak however we will focus on the dry season in interpreting this result which indicates that after adjusting for rainfall differences significantly less water was delivered to the command per unit of benefitted area following 1981 than before This indicates based on the criteria outlined above that performance in terms of water adequacy deteriorated following financial selfshysufficiency We need to examine this conclusion more carefullyhowever
One difficulty is that the measured quantity of water diverted at the source is largely a function of the supply available in the river rather than of system management This is particularly true during the dry season and in non-reservoir systems Thus while the depth of water supplied to the system is a measure of the adequacy of the systems service it is to some extent beyond the control of the managing agency To better understand the factors behind this decline in water availability we look at simple unadjusted index values for
- 12 shy
several of the key variables Table 4 shows annual values of total volume of water delivered in each season SA and BA in both wet and dry seasons and shows the results of t-tests on the means of a set of indicators before and after 1981 Indicators are used rather than the actual values to weight each of the systems equally regardless of its size The table shows that both the average wet season benefitted area and the average discharge are significantlylower during the second period compared with the first For the dry season too the discharge index is lower after 1981 than before but this difference is not significant At the same time the dry season BA index rose slightly but again the change was not significantSince middotthere is not a clear pattern of relative movement of discharge and BA during the respective seasons no simple interpretation of these index value changes is possible What stands out is that both discharge and benefitted area declined across periods during the wet season while during the dry season there was no significant change in either indicator across the two periods It seems clear that the decline in water adequacy must be evaluated together with other measures of performance in drawing conclusions about the overall impact of the 1981 changes on the quality of system management
Another measured variable per hectare yield can be used as a proxy for water adequacy It has the advantage of partiallyreflecting the impacts of the dimensions of timeliness6 and equity7of distribution as well integrating all three effects into a combined impact on aggregate crop production Table 5 (equations 1 and 3)shows that the period dummy in the yield regressions has a positive sign in both seasons after controlling for nitrogen application and precipitation though the t-values are not significant Treatingyield adjusted in this way as a proxy for quality of irrigationservice leads to the conclusion that by this more comprehensive measure quality of service held constant across the two periods in the dry season Because of the large yield component accounted for byrainfall during the wet season no such judgement is possible for that season however
Equity As noted earlier no reliable data are available for subdivisions of the five sample systems making direct computation of equity measures impossible We can make some judgements about changesin the equity of water distribution however by examining changes in the ratio of two area measures given for each system SA and BA Since SA is the area which theoretically can be supplied with irrigation water by the system and BA s the area which actuallyreceives a quantity of water adequate to produce a remunerative crop the ratio of the two provides a measure of the percentage of the potential service area which was irrigated to a particular standard The larger this pErcentage the more equitable 8 is the distribution This of course assumes that the quantity of water available to the systems is constant across the two periods
- 13 shy
Since the condition of constancy of water supply is not generally satisfied a regression was run in which the total quantityof water diverted at the headworks of each system divided by the systems potential service area SA was included in the regression to control for changes in the water supply available to the system seeTable 3 equations 5 through 8 The average daily rainfall received directly on the system service area during the season was also included as an independent variable The regression was run separately for wet and dry seasons The sign and t-statistic of the period dummy should then tell us whether or not equity as reflected in the BASA ratio increased decreased or remained unchanged across the period divide
Both equations are reasonable good as indicated by the R2 values though the dry season equation is considerably better as would be expected For the wet season both the water delivery term and the rainfall term in equation 5 are of positive sign but are nonshysignificant at the 95 percent confidence level indicating that wet season irrigated area does not change appreciably in response to level of wet season rainfall or the available irrigation water supply The period dummy was negative but not significant indicating that equityof distribution as reflected in the BASA ratio was similar during the two periods
For the dry season the water delivery term in equation 7 is positive and strongly significant indicating a close relationship between the fraction of potential area actually irrigated and the water supply available at the headworks In addition however the period dummy is positive and significant suggesting that once the influence of water supply is removed the BASA ratio was significantly higher in the period following 1981 than it was before
This is an important finding for it reflects significantlyimproved performance in terms of a factor equity of water distribution that is under the control of the managing entity an entity which here comprises both NIA and irrigators associations Interpreted in these terms NIA and allied farmers associations were able to spread a given amount of water more widely across the potential command area of the five sample systems in the period after 1981 than before Moreover they did this in a way that did not decrease average system yields as discussed earlier In making this interpretation we are suggesting that there was some redistribution of water from better-watered areas to fringe areas which would otherwise not have received irrigation water and that this redistribution was a direct response to the change in NIA prioritiesand operating policies and rules occurring around 1981
It is difficult to prove the assertion that water was in fact redistributed with a resulting increase in directly-measured equity Without access to reliable discharge data broken out by systemsection and we can only assume in the absence of a plausible
- 14 shy
alternative explanation that it was such a redistribution that made the increase in the BASA ratio possible In a larger sense it is difficult to prove conclusively that any outcome in a before and after analysis was the result of a particular independent causative factor In this case we have tried to remove the influence of other potential causative factors where we could but the possibilityremains that some combination of unmeasured factors are responsiblefor the difference in the BASA ratio found We do note though that this type of response is exactly the type that would be expected to follow from an emphasis on increased farmer satisfaction and cooperation and increased fee revenues Because the fee schedule is tied to benefitted area the only ways NIA can increase its revenue from that source are to expand benefitted area and to increase collection efficiencies The former depends on redistributing a fixed supply of water over a larger portion of the command while the latter requires that farmers be satisfied with the irrigation service they are receiving and the commitment of the local irrigators association to assist in the task of collecting the amounts due The evidence while not conclusive is highly suggestive that this is exactly what has happened
Efficiency In addition to measures which reflect the levels of adequacy and equity of irrigation service available data allow the calculation of a measure of operating efficiency The term efficiencyusually denotes the relationship between inputs to a process and its outputs often expressed as a ratio The output measure employed here is aggregate system rice output and the input is quantity of irrigation water turned into the system Dividing the first by the second gives a measure of agricultural production per unit water--here termed specific yield This is a highly integrated measure that evaluates the combined efficiency of the irrigation and agricultural processes As such it is a function of the managerial and other inputs supplied both to the irrigation system and to the agricultural operation With respect to one important input to the irrigation system we do know that NIA per hectare field operating expenses were about 29 percent lower in real terms in the 1982-86 period comparedto the 1976-1981 period although this drop may have been partlyoffset by increases in farmer-supplied labor inputs Other things being equal one would thus expect to find a decline in output efficiency
The regression analysis shows positive signs for the period terms in both wet and dry season equations (see Table 5) In the case of the wet season the period dummy in equation 5 is significant but the overall explanatory power of the model is quite low For the dry season (equation 7) the coefficient is positive but non-significant This means that after taking rainfall and fertilizer use into account data do not indicate a lowering of specific yield in the wake of funding reductions and the strong emphasiS on financial viabilitybeginning in 1981 This result provides evidence that the efficiency of the overall irrigation deliveryagricultural production process
- 15 shy
relative to the system water input did not falloff as a result of the changes implemented at least over the short run
Impact magnitude
The preceding analysis has shown us that some indicators of irrigation performance changed significantly following the managerial changes of 1981 while others did not However it has not given us a sense of the size of the changes which occurred To determine the magnitude of these changes the regression model is used to predict the response of the composite system to the managerial changes given a common set of -input and environmental factors To do this averagevalues of the independent variables from the entire eleven-year period 1976 to 1986 are put into the model together with the previously determined coefficients to generate predicted average values of the various dependent variables used in the earlier analysis with and without the period dummy This procedure produces a pair of estimates for each dependent variable under the same conditions--one in which the system responds as it did after the managerial changes were implemented and one in which it responds as it did prior to their introduction The differences between these two values thus indicate the magnitude of the changes occurring in the various indicators of performance discussed above
The results of this exercise are shown in Table 6 The table shows that water availability decreased by about 13 percent in both wet and dry seasons when the period dummy was included and while the coefficients responsible were significant in the earlier analysisthis difference cannot be easily connected with levels of system management as discussed earlier With respect to rice output per hectare although the coefficients were not very significant it is interesting to note that yield increases by 163 kilograms per hectare for the wet season and by 101 kilogram for the dry when the period dummy is included in spite of the reduced water supply available Keep in mind that the predicted yield values have already been adjusted for differences in nitrogen fertilizer use and rainfall This suggests that timeliness and equity of distribution of water supply to farmers may have increased following the changes contributing to the higher predicted yields
Examining the impact of increased equity of distribution bylooking at the ratio of benefitted area to service area we recall that the change was positive and significant for the dry season and negative and not significant for the wet Table 6 shows that the dry season BASA ratio increases by 7 percentage points when the dummy is included a 131 percent increase Other things being equal this should result in a 131 percent increase in system output due to the expansion of ared benefitted This is a major impact on production
- 16 shy
CONCLUSIONS
The Philippine experiment to transform the national irrigation agency into an enterprise has undoubtedly been successful in reducing system operating expenses bringing revenues and costs into line and eliminating the recurrent cost burden imposed by large-scaleirrigation systems on the national budget Evidence presented in this paper indicates that in the process equity of water distribution across systems has also improved In the 5 years following the cessation of operating subsidies from the government an index of equity of distribution improved by about 13 percent At the same time per hectare yields adjusted for rainfall and nitrogen application held constant
There is a strong logical connection between the achievement of financial viability and improved equity of water distribution across the command Because increasing irrigation fees is a politicaldecisionlying largely beyond NIAs control expanding the area which can be billed for service is one of the few revenue increasing measures available to the irrigation agency which does not involve major additional investment In the face of constant or shrinkingwater supplies this is achieved only by redistributing water from areas receiving excessive supplies usually near the head ends of canals and laterals to areas receiving no supplies or inadequatesupplies often located near the tails of canals Although data are not available which would allow the direct examination of this hypotheses the two outcomes are logically consistent with each other
Data also show that per hectare water deliveries declined significantly in the five sample systems after 1981 even thoughrainfall did not differ appreciably between the two periods This decline averaged about 13 percent for both wet and dry seasons and is interpreted as a decline in water availability in the supplying rivers rather than a conscious reduction in withdrawals by system managers Such declines could result from changes in watershed runoff characteristics as caused by deforestation or from increased upstream abstractions from supplying rivers
Improved water distribution tends to increase the area served system agricultural output and NIA service fee revenue Reduced water supplies to the system tend to reduce these things Specificyield defined as system paddy output per unit water held roughly constant across the two periods indicating that the two effects mayhave offset each other
After adjusting for rainfall and nitrogen application perhectare yields increased only marginally in the post-1981 period Area served on the other hand increased by about 13 percent after adjusting for water supply availability indicating that the area benefitted by irrigation in the sample systems increased by about the same percentage Even if yields on this additional area are less than
- 17 shy
average yields for the system this still represents a sizeable increase in system agricultural output as a result of the change in management structure the increase coming not from higher yields but from expanded area under irrigation
The evidence assembled here suggests that there are significantfinancial and economic benefits to be had from changes in the basic character of irrigation managing agencies which make them more responsive to their clientele and which impose rational internal financial discipline on the agency The analysis suggests a number of additional questions however One relates to the longer-term impacts of the structural management changes The improvements in water distribution described here are relatively short-term events occurring during the first 5 years of the new management mode Critics have suggested the danger of underinvestment in systemmaintenance over the longer run accompanied by declining yields and benefitted areas and eventual system collapse This possibility needs to be closely monitored A second concern relates to the apparent decline in water supply to these 5 geographically dispersed systems The nature and causes of this decline need to be explored further since if widespread and secular it may represent a serious threat to the stability of Philippine rice production Whether stemming from poor forest management practices or deficient regulation and allocation of surface water resources or other unidentified factors it is an issue that deserves serious and urgent consideration
A third risk is that the incentive structure set up by NIA to guide and stimulate the performance of field units overemphasizes revenue generation at the expense of irrigation service provision to farmers The evidence presented here supports the view that these two objectives are mutually reinforcing under policies and conditions which have been established in the Philippines More detailed crossshysectional studies based on primary flow measurement data would add confidence to this conclusion and help to specify the conditions under which this effect occurs This could be extremely important in transferring the results of the Philippine experiment to other countries
A final risk is that outside intervention well meaning or otherwise will destroy the basis of NIAs financial autonomy or will impose external pressures or constraints on NIAs decision-making that will subvert the management practices which have been so painstakinglydeveloped and implemented Among these are calls for NIA to be subsumed again within the government department structure in the interests of better coordination with agriculture attempts byexternal financing agencies to arbitrarily increase NIAs expenditures on OampM on the assumption that this will increase system agricultural output or intervention by Philippine legislative bodies to restore operating subsidies to NIA with attached strings leading back to legislators home districts Pressures such as these will cut short a process of experimentation and improvement that seems promising
- 18 shy
enough to date to warrant its continuation Having developed the capacity to establish targets and implement and manage change NIA is in a strong position to modify its objectives to better achieve larger social purposes established for it It is critical to recognize however that this must happen within the context of financing policies that mandate financial autonomy for NIA if the fundamental institutional commitment to manage is to be preserved
The author would like to thank Leslie Small and JeremyBerkoff for helpful comments on an earlier unpublishedversion of this paper and Charles Rogers for his careful and creative help with the analysis
- 19 shy
BIBLIOGRAPHY
Abernethy Charles L 1990 Indicators of the performance of irrigation water distribution systems International Irrigation Management Institute Colombo Sri Lanka Mimeo
Asian Development Bank 1986 Irrigation service fees Proceedingsof the Regional Seminar on Irrigation Service Fees Manila Asian Development Bank
Carruthers Ian and Colin Clark 1981 Economics of IrrigationLiverpool Liverpool University Press Third Edition
Levine G and EW Coward Jr 1986 Irrigation water distribution implications for design and operation AGREP Division WorkingPaper 125 vol 1 World Bank Agriculture and Rural Development Department
Small Les E 1989 User charges in irrigation potentials and limitations Irrigation and drainage vol 3 no 2125-142
Small Les 1990 Irrigation service fees in Asia IrrigationManagement Network 9013 London Overseas DevelopmentInstitute
Svendsen Mark and Les Small 1989 A framework for assessing irrigation system performance Paper prepared for the Symposium on Performance Evaluation 23 November International IrrigationManagement Institute Sri Lanka
Table I--National Irrigation Administration revenues and expenditures in constant prices 1976-86
Item 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986
(peso million 1972)
Revenues Irrigation fees collected 1273 1483 17 13 1831 2070 1668 1699 1893 1728 2129 2546 Other income 715 737 2420 5591 2631 5990 7783 6638 5699 4932 2934
Total direct revenue 1988 2220 4133 7422 4701 7658 9482 8531 7427 061 5480
Expenses in 1972 pricesTotal expenses 4825 5716 5039 6329 3821 77 55 6166 4749 4348 4259 4959
Excess (deficit) (2837(3496) (906) 1093 877 (097) 3316 3782 3079 2802 521 N 0
Subsidies Government operation and
maintenance subsidies 2521 2741 2799 1817 1398 633 0 0 0 0 0 Calamity fund payments 548 0 0 0 0 0 0 0 119 0 142
Total subsidy 3069 2741 2799 1817 1398 633 0 0 119 0 142
Total excess (deficit) 231 (754) 1893 2910 2275 536 3316 3782 3198 2802 663
Source IFPRI analysis of NIA data
- 21 shy
Table 2--Descriptive characteristics of selected MIA systeasa
------- Region III -------- ----- Region VI ----shyUPRIIS Angatii Sto Sibalom- Aganan-
Haasim Thomasc San Jose Sta Barbara
Average service area (hal 102272 31462 3522 5282 8703 Average irrigated area (hal
Wet season 83768 23454 3007 4410 8300 Dry season
Average benefited area Wet season
(hal 64587
77 605
27639
22908
1 781
3007
2801
4369
2770
7698 Dry season
Average rainfall Wet season
(mml d 62478
1 685 5
27396
8576
1 781
3051 0
2769
24731
2997
20001 Dry season 756 333 322 2828 3025
Average discharge (Llsec) Wet season 46501 14792 1692 2353 4984 Dry season 78091 22812 2014 1276 2315
Average water delivery (mmday) Wet season 522 548 487 462 571 Dry season
Average yield (mtha) 1089 715 995 398 686
Wet season 345 419 322 395 435 Dry season
Avg yield per unit water 34 03
(kgm ) 451 412 399 426
Wet season 0373 0440 0373 0538 0443 Dry season 0248 0400 0279 0690 0428
t-statistic difference in mean rainfall 1978-81 1982-86e
Wet season 0432 0713 -0567 1169 1169 Dry season 0519 -0230 -0523 1187 1187 Annual 0460 0707 -0686 1445 1445
~ Summary numbers are averages for the period 1982-1986 except as noted Water delivery discharge and yield per unit discharge are 4-year averages 1982-1985
c Water delivery discharge and yield per unit discharge are 4-year averages d 1983-1986
For Angat 5 years are 1981-85 For St Thomas 1979-83 For Sibalom 1971-75 e No significant differences at 95 confidence
- 22 shy
Table 3--Results of pooled regressions water delivery indicators
Dependent variable Independent variable - Water Delivery in mmBenefited Ha - --- BenefitService Area Ratio --shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season-shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 6799 (998)
6189 (1019)
13102 (1093)
2411 (1 54)
0743 ( 1186)
0800 (1165)
0129 (131 )
0258 (329)
System 2 indicator 0173 (029)
-3743 (-363)
-0063 (-219)
0200 (505)
System 3 indicator 0698 (085)
0163 (012)
0085 (208)
-0046 (-097)
System 4 indicator -0514 (-l 07)
-7011 (-803)
0080 (340)
0169 (l87)
System 5 indicator 0177 (033)
-5049 (-538)
0174 (603)
0051 (058)
Reservoir indicator 0327 (078)
5703 (543)
-0126 (-510)
0156 (261)
Period indicatora -0808 (-240)
-0771 (-229)
-1214 (-209)
-1309 (-1 62)
-0036 (-192)
-0025 (-112)
0071 (239)
0030 (077)
Water delivery mmb
Wet season
Dry season
Precipitation in mm Wet season
Annual
Average daily rainfall Wet season
Dry season
-46E-4 (-125)
-1 9E-4 (-082)
-43E-4 (-077)
20E-3 (385)
0015 (1 59)
-0001 ( -004)
0023 (215)
0001 (021)
0070 (613)
-0017 (-058)
0046 (371)
0010 (033)
R2 (adj) number of observations degrees of freedom
0156 43 36
0133 43 39
0711 41 34
0434 41 37
0693 42 34
0556 42 37
0865 40 32
0730 40 35
Coefficient (t statistic) indicates significance at 95 confidence ~ 1982-1986 1
per ha of service area
Ta
ble
--I
nd
ices
of
se
v
ice a
rea
ib
en
ef
ited
a
rea
a
nd
a
vera
ge sea
so
na
l d
isch
arg
e
Ave
rage
A
vera
ge
tshy19
77
1978
19
79
1980
19
81
1982
19
83
1984
19
85
1986
19
77-8
1 19
82-8
6 S
tati
stic
a
(ind
ex
aver
age
1983
-198
5 =
100)
Ser
vice
are
a UP
RI IS
in
dex
882
91
9
920
91
5
951
95
1
100
0 10
00
100
0 10
00
917
99
0
bull5
53
Ang
at-M
aasi
m R
95
9
994
99
7
996
99
6
996
10
00
100
0 10
00
100
0 98
8
999
1
63
Sto
To
mas
10
63
106
3 10
1 9
10
29
103
0 99
8
100
0 10
00
100
0 11
10
104
1 10
22
-08
9
Siba
lom
-San
Jos
e 95
0
872
94
2
933
94
2
942
10
28
102
8 94
4
101
7
928
99
2
292
A
gana
n-St
a B
arba
ra
108
3 10
91
106
0 10
30
961
10
05
996
10
08
996
99
6
104
5 10
00
-21
1 A
vera
ge
987
98
8
981
98
1
916
97
8
100
5 10
07
988
10
25
984
10
01
123
Wet
seas
on b
enef
ited
are
a in
dex
UPR
IIS
110
0 10
07
114
4 10
55
113
8 11
74
951
10
71
971
ll
58
10
89
106
7 -0
47
Ang
at-M
aasi
m R
97
9
974
92
0
983
10
28
100
8 99
7
102
1 98
2
931
97
7
988
0
54
Sto
To
mas
11
63
115
9 11
23
107
5 10
80
103
9 98
1
978
10
41
103
9 11
20
101
6 -4
88
Siba
lom
-San
Jos
e 11
35
103
8 10
11
931
93
4
906
10
07
975
10
1S
998
10
11
981
-0
80
Aga
nan-
Sta
Bar
bara
10
85
110
5 10
74
106
8 10
01
104
2 99
8
101
5
987
61
0
106
7 93
0
-1 8
4
Ave
rage
10
92
105
6 10
54
102
4 10
36
103
4 98
7
101
3 10
00
947
10
53
996
-2
32
Dry
sea
Son
bene
fite
d ar
ea
inde
x U
PRIIS
14
04
155
0 15
S0
155
8 16
17
128
0 57
2
114
S 15
74
152
3 12
38
-09
4 A
ngat
-Maa
sim
R
903
93
0
103
2 10
61
104
2 10
69
988
99
2
102
1 99
6
993
10
13
061
S
to
Tom
as
105
7 12
27
122
5
961
99
9
115
9 10
1 7
91
0
107
3 12
1 2
10
94
107
4 -0
28
Si
balo
m-S
an J
ose
Aga
nan-
Sta
Bar
bara
66
5
95S
62
6
632
67
4
111
4
501
10
S1
412
11
51
766
93
3
856
94
3
107
0 94
8
107
4 11
09
111
6
158
6 58
8
987
97
6
110
4 5
35
083
N
w
Ave
rage
89
6
964
11
19
103
7 10
44
110
9 10
1 7
89
8
108
5 12
97
101
7 10
81
082
Wet
seas
on d
isch
arge
in
dex
UPR
IIS
132
9 72
7
142
5 11
88
120
4 98
8
105
8 10
63
879
96
4
117
5 99
1
-16
5 A
ngat
-Maa
sim
R
129
7 13
52
134
5 12
58
127
0 11
70
120
3 62
7
131
3 10
68
-18
9
Sto
To
mas
14
71
155
0 14
1 6
10
40
725
12
35
112
3 14
79
103
1 -4
48
Siba
lom
-San
Jos
e 96
8
733
11
1 5
92
2
907
47
1
102
3 85
7
ll2
0
422
92
9
779
-1
09
Aga
nan-
Sta
Bar
bara
87
9
863
68
1
925
96
8
110
7 70
5
871
87
7
00
9
Ave
rage
11
49
105
7 13
61
115
0 10
58
853
10
43
963
99
4
SO3
11
55
940
-2
85
Dry
sea
son
disc
harg
e in
dex
UPR
IIS
425
13
09
153
3 14
28
180
6 14
S0
125
8 56
3
117
9 14
66
130
0 11
89
-04
3
Ang
at-M
aasi
m R
12
26
161
6 15
30
139
2 12
85
100
8 10
33
958
14
41
107
1 -3
80
S
to
Tom
as
116
4 14
82
155
7 79
9
971
12
30
126
5 14
01
106
6 -2
44
Si
balo
m-S
an J
ose
486
71
5
782
62
9
789
64
5
116
9 11
86
801
65
3
918
2
36
Aga
nan-
Sta
Bar
bara
A
vera
ge
425
10
46
133
6 62
0
118
4 81
5
116
1 94
7
112
5 89
7
922
12
20
991
88
2
108
7 12
36
119
2 71
8
114
0 10
37
105
5 3
23
-07
5
a bull
indi
cate
s si
gnif
ican
ce a
t 95
per
cent
con
fide
nce
- 24 shy
Table 5--Results of pooled regressions yield indicators
------------------------- Dependent variables -------------------------shyIndependent variable ---------- Yield MTHa ----------- -- Specific Yield 1000 KgM
3 H20 shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season -shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 1967 (286)
3747 (580)
1991 (293)
2800 (460 )
0084 (041)
0441 (229)
0049 (023 )
0876 (343)
System 2 indicator 0400 (155)
0525 (226)
0057 (085)
0115 (1 95)
System 3 indicator 0466 (139 )
-0029 (-010)
-0016 (-018)
0006 (008)
System 4 indicator 0980 (470)
0090 (049)
0176 (333)
0391 (827)
System 5 indicator 1 221 (570)
0256 ( 138)
0104 (174)
0224 (440)
Reservoir indicator -1033 (-578)
-0063 (-039)
-0146 (-310)
-0326 (-569)
Period indicatora
Precipitation in mm Wet season
Annual
0163 (1 03)
-29E-4 (-182)
0231 (1 42)
-58E-4 (-606)
0161 (118)
92E-5 (072)
0199 043 )
-64pound-5 (-084)
0082 (200)
57E-6 (013)
0092 (223)
-61E-5 (-232)
0032 (093)
29E-7 (001)
0052 (112)
-12pound-4 (-414)
Nitrogen applicationb 0025 (254)
0021 (211)
0025 (274)
0020 (227)
0003 (1 05)
0002 (043)
0003 (101)
-0002 (-052)
R2 (adj) 0553 0516 0251 0198 0309 0262 0713 0445 number of observations 50 50 49 49 42 42 40 40 degrees of freedom 42 45 41 44 34 37 32 35
CoeffiCient (t statistic) indicates significance at 95 confidence ~ 1982-1986 = 1
kgha
- 25 shy
Table 6--Suggmry of estiates perforance indicators
Variable Predicted Predicted Period Dependent Variable
Description (units)
Period Coefficient
1976-86 IndicatorO
1976-86 Indicatorl
Effecta (X)
BenefittedService area ratio
Wet season ratio -0036 0853 0817 -425 Dry Season 0071 0539 0610 1313
Delivery Wet Dry
season season
nm water per benefitted ha
-0808 -1 214
6003 9083
5195 7869
-1346 -1337
Rice Yield Wet season metric tons 0163 3578 3741 456 Dry season per hectare 0161 3905 4066 412
Specific Yield Wet season kg rice per m3 0082 0348 0430 2353 Dry season water de 1 ivered 0032 0378 0409 837
a percentage changes relative to period=O values
- 26 shy
ENDNOTES
IResearch Fellow International Food Policy Research Institute and Senior Irrigation Specialist International Irrigation Management Institute
2Presumable farmers also base private investment decisions on such foreknowledge of costs and commonly do so in the case of small tubewells In addition user groups may invest in small communal schemes independently after examining projected costs and benefits Most commonly however new irrigation capacity is created as a result of public investment decisions made by government and thus the political process comes into play
3This principle applies only when water charges are levied volumetrically--an exceedingly rare situation in the developing worldshy-though the argument is frequently invoked as though it applies to all allocative situations
4For Cabanatuan City the longer period of record available ie 1904 to 1986 was divided into two equal periods and tested for difference in means to test the observation by farmers that rainfall today is lower than it used to be Mean precipitation after 1947 was actually about 3 percent higher than before but the difference was not significant (t = 07)
5The procedure used was to take time series data on total nitrogenfertilizer consumption in the Philippines multiply it by the ratio of N fertilizer used on rice to total N consumption and divide the product by the rice harvested area to obtain an estimate of N use perhectare Values of the ratio of N use on rice to total N use were taken from IRRI (1988 150) with interpolation used to fill in gaps in the series
6This assumes that timeliness is evaluated against a standard based on the physiological demand for water generated by the rice crop and the varying sensitivity of the crop to stress at different growth stages In such a case a poor timeliness rating for irrigationservice will result in reductions in yield
7Because rice yields are insensitive to water applications that exceed PET values a redistribution of water from water excess to water short portions of the command area will usually result in higher aggregate output and higher average yields Exceptional scenarios can be constructed in which this general rule would be violated but such scenarios are unlikely to play out in practice
- 27 shy
aThere are a variety of standards for judging equity that could be used here The most common one is equality of per hectare water deliveries In the present case the implicit standard is equality of water supply utility in producing a crop Thus if two areas received equal amounts of water on a per hectare basis but one had such high seepage and percolation losses that the crop failed to produce a harvest (and was therefore exempted from irrigation fee payment) the water supply to the combined area would be judged inequitable Because such radical differences in seepage and percolation rates are unlikely on puddled rice soils in the lowland Philippines and because the definition of benefitted area accommodates a wide range of acceptable yields these two definitions are largely indistinguishable at the current limits of data resolution
- 7 shy
IMPACTS ON SYSTEM PERFORMANCE
Study Methodology
Unfortunately the kinds of changes in hydrologic system outputand the impact on agricultural performance which might be expected to result from improved institutional performance are quite difficult to capture and quantify There are a number of reasons for this First there is the year to year variability of system performance caused byvariable rainfall which feeds rivers fills reservoirs and supplements irrigation water in supplying crop requirements Second there is the difficulty of defining just what performance is and specifying how to measure it Third and most important the regularly collected data from which indicators can be constructed are limited in type number of measuring points and period of record and are sometimes of doubtful reliability
These difficulties notWithstanding an attempt was made to determine the impact that changes in operating procedures staffinglevels and incentive programs had on system performance Because the effects that we are trying to assess resulted from changes that affected all of the systems under NIAs direct authority there are no control systems which can be used as standards We are forced therefore to rely on a comparison of values of selected performance indicators before and after the date of the major structural and procedural changes which is taken to be 1981
To accomplish this secondary data were assembled for 5 systems in Administrative Regions III and VI which had not undergonesignificant phYSical changes during the period of analysis Duringthis process several site visits were made by study team members Time series data collected include service area (SA) and benefitted area (BA) for both wet and dry seasons yields for wet and dry seasons monthly main canal discharge at the system headworks and monthly precipitation The general period of availability for this data is 1966-86 though for systems which began operation after 1966 the period of record is shorter and the records of some systemscontain miSSing values These five systems their 1986 service and benefitted areas and other descriptive data are shown in Table 2
The prinCipal problem with using a before and after approachrather than one that considers comparable systems with and without the innovation is that some of the measured difference in effects mayhave resulted from causes which are independent of the ones beingstudied These causes can be specific in which case they may be relatively easy to identify and accommodate or more general and diffuse and therefore more difficult to control for In the present case--that of changes in the performance of NIA irrigation systemsresulting from the major organizational changes in 1981 two principalexternal factors can be identified which might be expected to affect differences in measured levels of irrigation performance between the
- 8 shy
two time periods These are rainfall and level of use of other agricultural inputs Since our interest is in systems managerial responses to these changes and since there is no reason to believe that the relative magnitudes of the responses are dependent on the size of the system the five systems are treated as equal in the analysis That is changes in measured values relating to the smallest system are considered to be as important as changes in values for the largest with no area weighting applied
Since rainfall can substitute for irrigation water supplies and since it affects the supply of water available in rivers for irrigation it may exert some independent influence on various performance indicators To test the strength of the relationship for the period being analyzed simple correlations were run between rainfall and benefitted area for one system in each regionBenefitted area was used in this analysis because it is the variable deemed most likely to be influenced by year-to-year changes in rainfall Weather data from Cabanatuan City was used for the UPRIIS system which surrounds it and Iloilo City data was used for the nearby Aganan-Santa Barbara system For UPRIIS all of the R2 values for these correlations were less than 0005 suggesting that rainfall has almost no impact on area harvested in this large reservoir-based scheme For Aganan-Santa Barbara wet season rainfall was related to wet season BA (r2 = 016) and to BA during the following dry season (r2 = 024) Signs of the simple correlations were in the expected directions ie wet season rainfall increased BA during both the wet and the subsequent dry seasons These connections are understandable but weak
Another possibility is that there were longer-term differences in rainfall received in the two regions If this were the case a comparison of performance during two different time periods would have to take this difference into account Differences in average precipitation during the two periods were examined for the four stations used in the analysis (see Table 2) In no case were differences in seasonal or annual mean rainfall statisticallysignificant4
bull
Nevertheless in the regression approach adopted to analyze the data rainfall was included in each equation to control for its possible effect on the particular dependent variables being analyzed In doing this wet season rainfall was used in analyzing wet season performance indicators while annual rainfall was used in analyzingdry season data The rationale for this is that while dry season rainfall cannot possibly influence the wet season crop the dry season crop is affected by both the rainfall received directly and the rain falling during the preceding wet season through its effect on river discharge reservoir storage and antecedent soil moisture conditions
The level of agricultural production is also an often-used indicator of an irrigation systems performance Its major weakness
- 9 shy
is that a number of factors other than irrigation service such as labor inputs relative prices and fertilizer use influence it It is necessary therefore either to control for changes in the levels of these inputs or assume that they are constant across the two periods being compared In the present case the most important of these factors is the level of application of chemical fertilizer Because fertilizer use by farmers is responsive to the relative pricesof fertilizer and rice it also includes to some extent input and output price effects Since reliable data on fertilizer use for individual systems were not available estimates derived from FAO fertilizer and cultivated rice area data were used t~ control for the effect of changes in the use of this input over time This variable was included in any of the regression equations in which agriculturalproduction was used as the dependent variable Other factors such as labor use genetic potential of varieties sewn and pesticideapplications are assumed to be constant across the two periods
The analytic approach employed is to fit linear regressionequations to pooled data from the five systems covering a eleven-year period 1976 to 1986 A dummy variable is used to check the impact of pre and post 1981 periods on differences in the dependent variable after the effects of factors such as rainfall and nitrogen fertilizer use have been removed In addition because the dataset was created by pooling data from five different systems a set of 4 site dummies was included in the basic model to control for system-specific differences caused by variables which were not measured For some runs these were replaced with dummies that separated reservoir and non-reservoir systems though equations using the reservoir dummy were consistently inferior to those using the complete set of site dummies Several different dependent variables were created to index the quality of irrigation service and tested using this approachRegression results are given in Tables 3 and 5 and discussed below
Performance Indicators
A variety of indicators have been used in evaluating irrigationperformance in various contexts The selection of appropriate indicators depends on a number of factors including the purpose of the evaluation the audience for its results the way in which the boundaries of the irrigation system are defined and the kind and quality of data available to the evaluators The current analysis is designed to evaluate the impact of a set of management changes on system physical performance The audience for this analysis comprises top-level managers of the irrigation agency and policy-makers at higher government levels Boundary definition is an importantanalytic problem here as evident from the subsequent discussion relating to the choice of the appropriate area values to use in scaling system inputs and outputs This issue is also related to the data quality and availability problems which have already been mentioned
- 10 shy
Three fundamental indicators have been proposed for assessing the effectiveness of irrigation services to cultivators (Svendsen and Small 1989 Abernethy 1990) These are the adequacy of water supplies the equity of their distribution across the command area of the system and the timeliness of the supplies Computation of adequacy measures requires information on the total quantity of water delivered to the system over a season on a per hectare basis Equityand timeliness measures require information on the spatial and temporal distribution respectively of those supplies Where appropriate discharge information is not available proxies can be employed by making suitable assumptions Standards must be selected against which the magnitude of the indicators can be judged
The task in the present case however is somewhat different Here the need is to evaluate changes in selected variables between the pre-1981 and the post-1981 periods Hence the absolute values of variables selected are less important than their relative magnitudesand the statistical significance of the differences in magnitudes between the two periods A distinct limitation is imposed by the data series available for the five sample systems Since discharge and yield data are available only on a whole-system basis it is impossible to develop measures of equity and timeliness directly We will however extend our analysis to a discussion of equity byindirect inference Levine and Coward (1986) have argued that equityought to be considered as the paramount objective in managing largepublic irrigation systems They base their conclusion on an analysisof eight small community-managed systems and five larger public systems including UPRIIS in which equity appears to comprise the most important operational objective in the successful systems It may be appropriate therefore to give success in improving equity of distribution added weight in assessing overall performance
Area Estimates Because measures of system agricultural output and water supplied are typically reduced to a unit area basis before being used much depends on the area values which are used to standardize them Two different area measures are available The first is Service Area (SA) which is defined as the irrigable portion of the command area which is provided with physical facilities for water delivery This represents the area which could conceivably be irrigated in a given season if water supply were not constraining This value may change somewhat from year to year in response to urban encroachment on irrigated command minor remodeling and repair and refinements in area estimates In the present case though the systems selected for analysis were chosen to avoid those which had undergone more extensive rehabilitation or modification
The second measure is Benefitted Area (BA) which is the area billed for payment of irrigation service fees It is the irrigated area harvested which did not have yields so low that the farm was exempted from payment of fees in a given season This threshold value has been approximately 2 tons of paddy per hectare Benefitted area
- 11 shy
varies more than does SA particularly during the dry season when available water supply may seriously constrain the area which can be planted Its magnitude is a function of system managers actions in authorizing the amount of land to be planted in a given season farmers decisions regarding whether to plant or not and the combined ability of system managers and farmersirrigators subsequently to distribute water Both of these area measures will be used to standardize other variables for particular purposes as well as beingcombined to form a separate indicator by themselves
Adequacy The most direct measure of the adequacy of irrigation water supplies to the agricultural system is the quantity of water applied to the system command area on a per unit area basis relative to some standard In this case since our interest is in differences in water adequacy between two time periods and since the systems being assessed have been and continue to be almost entirely devoted to rice cultivation during both cropping seasons depth measures for the two periods may be compared directly assuming the seasonal cropdemand for water to be unchanged Although dry-footed crops can suffer yield losses from overapplication of irrigation water rice is largely insensitive to this effect In addition water can substitute for other inputs that the farmer would otherwise have to provide such as weed control and more careful (and costly) water management We assume therefore that other things being equal larger values of depth applied are better than smaller values in terms of meeting crop water demands and reduce the cost of cultivation At the same time high levels of water adequacy can affect the values of other performance measures--particularly equity
When the regression model is run for quantity of water diverted at the system headworks divided by BA hereafter termed depth we see that the period dummy is negative and significant at the 95 percent confidence level for both wet and dry seasons (see Table 3 equations 1 and 3) Since the overall explanatory power of the wet season model is very weak however we will focus on the dry season in interpreting this result which indicates that after adjusting for rainfall differences significantly less water was delivered to the command per unit of benefitted area following 1981 than before This indicates based on the criteria outlined above that performance in terms of water adequacy deteriorated following financial selfshysufficiency We need to examine this conclusion more carefullyhowever
One difficulty is that the measured quantity of water diverted at the source is largely a function of the supply available in the river rather than of system management This is particularly true during the dry season and in non-reservoir systems Thus while the depth of water supplied to the system is a measure of the adequacy of the systems service it is to some extent beyond the control of the managing agency To better understand the factors behind this decline in water availability we look at simple unadjusted index values for
- 12 shy
several of the key variables Table 4 shows annual values of total volume of water delivered in each season SA and BA in both wet and dry seasons and shows the results of t-tests on the means of a set of indicators before and after 1981 Indicators are used rather than the actual values to weight each of the systems equally regardless of its size The table shows that both the average wet season benefitted area and the average discharge are significantlylower during the second period compared with the first For the dry season too the discharge index is lower after 1981 than before but this difference is not significant At the same time the dry season BA index rose slightly but again the change was not significantSince middotthere is not a clear pattern of relative movement of discharge and BA during the respective seasons no simple interpretation of these index value changes is possible What stands out is that both discharge and benefitted area declined across periods during the wet season while during the dry season there was no significant change in either indicator across the two periods It seems clear that the decline in water adequacy must be evaluated together with other measures of performance in drawing conclusions about the overall impact of the 1981 changes on the quality of system management
Another measured variable per hectare yield can be used as a proxy for water adequacy It has the advantage of partiallyreflecting the impacts of the dimensions of timeliness6 and equity7of distribution as well integrating all three effects into a combined impact on aggregate crop production Table 5 (equations 1 and 3)shows that the period dummy in the yield regressions has a positive sign in both seasons after controlling for nitrogen application and precipitation though the t-values are not significant Treatingyield adjusted in this way as a proxy for quality of irrigationservice leads to the conclusion that by this more comprehensive measure quality of service held constant across the two periods in the dry season Because of the large yield component accounted for byrainfall during the wet season no such judgement is possible for that season however
Equity As noted earlier no reliable data are available for subdivisions of the five sample systems making direct computation of equity measures impossible We can make some judgements about changesin the equity of water distribution however by examining changes in the ratio of two area measures given for each system SA and BA Since SA is the area which theoretically can be supplied with irrigation water by the system and BA s the area which actuallyreceives a quantity of water adequate to produce a remunerative crop the ratio of the two provides a measure of the percentage of the potential service area which was irrigated to a particular standard The larger this pErcentage the more equitable 8 is the distribution This of course assumes that the quantity of water available to the systems is constant across the two periods
- 13 shy
Since the condition of constancy of water supply is not generally satisfied a regression was run in which the total quantityof water diverted at the headworks of each system divided by the systems potential service area SA was included in the regression to control for changes in the water supply available to the system seeTable 3 equations 5 through 8 The average daily rainfall received directly on the system service area during the season was also included as an independent variable The regression was run separately for wet and dry seasons The sign and t-statistic of the period dummy should then tell us whether or not equity as reflected in the BASA ratio increased decreased or remained unchanged across the period divide
Both equations are reasonable good as indicated by the R2 values though the dry season equation is considerably better as would be expected For the wet season both the water delivery term and the rainfall term in equation 5 are of positive sign but are nonshysignificant at the 95 percent confidence level indicating that wet season irrigated area does not change appreciably in response to level of wet season rainfall or the available irrigation water supply The period dummy was negative but not significant indicating that equityof distribution as reflected in the BASA ratio was similar during the two periods
For the dry season the water delivery term in equation 7 is positive and strongly significant indicating a close relationship between the fraction of potential area actually irrigated and the water supply available at the headworks In addition however the period dummy is positive and significant suggesting that once the influence of water supply is removed the BASA ratio was significantly higher in the period following 1981 than it was before
This is an important finding for it reflects significantlyimproved performance in terms of a factor equity of water distribution that is under the control of the managing entity an entity which here comprises both NIA and irrigators associations Interpreted in these terms NIA and allied farmers associations were able to spread a given amount of water more widely across the potential command area of the five sample systems in the period after 1981 than before Moreover they did this in a way that did not decrease average system yields as discussed earlier In making this interpretation we are suggesting that there was some redistribution of water from better-watered areas to fringe areas which would otherwise not have received irrigation water and that this redistribution was a direct response to the change in NIA prioritiesand operating policies and rules occurring around 1981
It is difficult to prove the assertion that water was in fact redistributed with a resulting increase in directly-measured equity Without access to reliable discharge data broken out by systemsection and we can only assume in the absence of a plausible
- 14 shy
alternative explanation that it was such a redistribution that made the increase in the BASA ratio possible In a larger sense it is difficult to prove conclusively that any outcome in a before and after analysis was the result of a particular independent causative factor In this case we have tried to remove the influence of other potential causative factors where we could but the possibilityremains that some combination of unmeasured factors are responsiblefor the difference in the BASA ratio found We do note though that this type of response is exactly the type that would be expected to follow from an emphasis on increased farmer satisfaction and cooperation and increased fee revenues Because the fee schedule is tied to benefitted area the only ways NIA can increase its revenue from that source are to expand benefitted area and to increase collection efficiencies The former depends on redistributing a fixed supply of water over a larger portion of the command while the latter requires that farmers be satisfied with the irrigation service they are receiving and the commitment of the local irrigators association to assist in the task of collecting the amounts due The evidence while not conclusive is highly suggestive that this is exactly what has happened
Efficiency In addition to measures which reflect the levels of adequacy and equity of irrigation service available data allow the calculation of a measure of operating efficiency The term efficiencyusually denotes the relationship between inputs to a process and its outputs often expressed as a ratio The output measure employed here is aggregate system rice output and the input is quantity of irrigation water turned into the system Dividing the first by the second gives a measure of agricultural production per unit water--here termed specific yield This is a highly integrated measure that evaluates the combined efficiency of the irrigation and agricultural processes As such it is a function of the managerial and other inputs supplied both to the irrigation system and to the agricultural operation With respect to one important input to the irrigation system we do know that NIA per hectare field operating expenses were about 29 percent lower in real terms in the 1982-86 period comparedto the 1976-1981 period although this drop may have been partlyoffset by increases in farmer-supplied labor inputs Other things being equal one would thus expect to find a decline in output efficiency
The regression analysis shows positive signs for the period terms in both wet and dry season equations (see Table 5) In the case of the wet season the period dummy in equation 5 is significant but the overall explanatory power of the model is quite low For the dry season (equation 7) the coefficient is positive but non-significant This means that after taking rainfall and fertilizer use into account data do not indicate a lowering of specific yield in the wake of funding reductions and the strong emphasiS on financial viabilitybeginning in 1981 This result provides evidence that the efficiency of the overall irrigation deliveryagricultural production process
- 15 shy
relative to the system water input did not falloff as a result of the changes implemented at least over the short run
Impact magnitude
The preceding analysis has shown us that some indicators of irrigation performance changed significantly following the managerial changes of 1981 while others did not However it has not given us a sense of the size of the changes which occurred To determine the magnitude of these changes the regression model is used to predict the response of the composite system to the managerial changes given a common set of -input and environmental factors To do this averagevalues of the independent variables from the entire eleven-year period 1976 to 1986 are put into the model together with the previously determined coefficients to generate predicted average values of the various dependent variables used in the earlier analysis with and without the period dummy This procedure produces a pair of estimates for each dependent variable under the same conditions--one in which the system responds as it did after the managerial changes were implemented and one in which it responds as it did prior to their introduction The differences between these two values thus indicate the magnitude of the changes occurring in the various indicators of performance discussed above
The results of this exercise are shown in Table 6 The table shows that water availability decreased by about 13 percent in both wet and dry seasons when the period dummy was included and while the coefficients responsible were significant in the earlier analysisthis difference cannot be easily connected with levels of system management as discussed earlier With respect to rice output per hectare although the coefficients were not very significant it is interesting to note that yield increases by 163 kilograms per hectare for the wet season and by 101 kilogram for the dry when the period dummy is included in spite of the reduced water supply available Keep in mind that the predicted yield values have already been adjusted for differences in nitrogen fertilizer use and rainfall This suggests that timeliness and equity of distribution of water supply to farmers may have increased following the changes contributing to the higher predicted yields
Examining the impact of increased equity of distribution bylooking at the ratio of benefitted area to service area we recall that the change was positive and significant for the dry season and negative and not significant for the wet Table 6 shows that the dry season BASA ratio increases by 7 percentage points when the dummy is included a 131 percent increase Other things being equal this should result in a 131 percent increase in system output due to the expansion of ared benefitted This is a major impact on production
- 16 shy
CONCLUSIONS
The Philippine experiment to transform the national irrigation agency into an enterprise has undoubtedly been successful in reducing system operating expenses bringing revenues and costs into line and eliminating the recurrent cost burden imposed by large-scaleirrigation systems on the national budget Evidence presented in this paper indicates that in the process equity of water distribution across systems has also improved In the 5 years following the cessation of operating subsidies from the government an index of equity of distribution improved by about 13 percent At the same time per hectare yields adjusted for rainfall and nitrogen application held constant
There is a strong logical connection between the achievement of financial viability and improved equity of water distribution across the command Because increasing irrigation fees is a politicaldecisionlying largely beyond NIAs control expanding the area which can be billed for service is one of the few revenue increasing measures available to the irrigation agency which does not involve major additional investment In the face of constant or shrinkingwater supplies this is achieved only by redistributing water from areas receiving excessive supplies usually near the head ends of canals and laterals to areas receiving no supplies or inadequatesupplies often located near the tails of canals Although data are not available which would allow the direct examination of this hypotheses the two outcomes are logically consistent with each other
Data also show that per hectare water deliveries declined significantly in the five sample systems after 1981 even thoughrainfall did not differ appreciably between the two periods This decline averaged about 13 percent for both wet and dry seasons and is interpreted as a decline in water availability in the supplying rivers rather than a conscious reduction in withdrawals by system managers Such declines could result from changes in watershed runoff characteristics as caused by deforestation or from increased upstream abstractions from supplying rivers
Improved water distribution tends to increase the area served system agricultural output and NIA service fee revenue Reduced water supplies to the system tend to reduce these things Specificyield defined as system paddy output per unit water held roughly constant across the two periods indicating that the two effects mayhave offset each other
After adjusting for rainfall and nitrogen application perhectare yields increased only marginally in the post-1981 period Area served on the other hand increased by about 13 percent after adjusting for water supply availability indicating that the area benefitted by irrigation in the sample systems increased by about the same percentage Even if yields on this additional area are less than
- 17 shy
average yields for the system this still represents a sizeable increase in system agricultural output as a result of the change in management structure the increase coming not from higher yields but from expanded area under irrigation
The evidence assembled here suggests that there are significantfinancial and economic benefits to be had from changes in the basic character of irrigation managing agencies which make them more responsive to their clientele and which impose rational internal financial discipline on the agency The analysis suggests a number of additional questions however One relates to the longer-term impacts of the structural management changes The improvements in water distribution described here are relatively short-term events occurring during the first 5 years of the new management mode Critics have suggested the danger of underinvestment in systemmaintenance over the longer run accompanied by declining yields and benefitted areas and eventual system collapse This possibility needs to be closely monitored A second concern relates to the apparent decline in water supply to these 5 geographically dispersed systems The nature and causes of this decline need to be explored further since if widespread and secular it may represent a serious threat to the stability of Philippine rice production Whether stemming from poor forest management practices or deficient regulation and allocation of surface water resources or other unidentified factors it is an issue that deserves serious and urgent consideration
A third risk is that the incentive structure set up by NIA to guide and stimulate the performance of field units overemphasizes revenue generation at the expense of irrigation service provision to farmers The evidence presented here supports the view that these two objectives are mutually reinforcing under policies and conditions which have been established in the Philippines More detailed crossshysectional studies based on primary flow measurement data would add confidence to this conclusion and help to specify the conditions under which this effect occurs This could be extremely important in transferring the results of the Philippine experiment to other countries
A final risk is that outside intervention well meaning or otherwise will destroy the basis of NIAs financial autonomy or will impose external pressures or constraints on NIAs decision-making that will subvert the management practices which have been so painstakinglydeveloped and implemented Among these are calls for NIA to be subsumed again within the government department structure in the interests of better coordination with agriculture attempts byexternal financing agencies to arbitrarily increase NIAs expenditures on OampM on the assumption that this will increase system agricultural output or intervention by Philippine legislative bodies to restore operating subsidies to NIA with attached strings leading back to legislators home districts Pressures such as these will cut short a process of experimentation and improvement that seems promising
- 18 shy
enough to date to warrant its continuation Having developed the capacity to establish targets and implement and manage change NIA is in a strong position to modify its objectives to better achieve larger social purposes established for it It is critical to recognize however that this must happen within the context of financing policies that mandate financial autonomy for NIA if the fundamental institutional commitment to manage is to be preserved
The author would like to thank Leslie Small and JeremyBerkoff for helpful comments on an earlier unpublishedversion of this paper and Charles Rogers for his careful and creative help with the analysis
- 19 shy
BIBLIOGRAPHY
Abernethy Charles L 1990 Indicators of the performance of irrigation water distribution systems International Irrigation Management Institute Colombo Sri Lanka Mimeo
Asian Development Bank 1986 Irrigation service fees Proceedingsof the Regional Seminar on Irrigation Service Fees Manila Asian Development Bank
Carruthers Ian and Colin Clark 1981 Economics of IrrigationLiverpool Liverpool University Press Third Edition
Levine G and EW Coward Jr 1986 Irrigation water distribution implications for design and operation AGREP Division WorkingPaper 125 vol 1 World Bank Agriculture and Rural Development Department
Small Les E 1989 User charges in irrigation potentials and limitations Irrigation and drainage vol 3 no 2125-142
Small Les 1990 Irrigation service fees in Asia IrrigationManagement Network 9013 London Overseas DevelopmentInstitute
Svendsen Mark and Les Small 1989 A framework for assessing irrigation system performance Paper prepared for the Symposium on Performance Evaluation 23 November International IrrigationManagement Institute Sri Lanka
Table I--National Irrigation Administration revenues and expenditures in constant prices 1976-86
Item 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986
(peso million 1972)
Revenues Irrigation fees collected 1273 1483 17 13 1831 2070 1668 1699 1893 1728 2129 2546 Other income 715 737 2420 5591 2631 5990 7783 6638 5699 4932 2934
Total direct revenue 1988 2220 4133 7422 4701 7658 9482 8531 7427 061 5480
Expenses in 1972 pricesTotal expenses 4825 5716 5039 6329 3821 77 55 6166 4749 4348 4259 4959
Excess (deficit) (2837(3496) (906) 1093 877 (097) 3316 3782 3079 2802 521 N 0
Subsidies Government operation and
maintenance subsidies 2521 2741 2799 1817 1398 633 0 0 0 0 0 Calamity fund payments 548 0 0 0 0 0 0 0 119 0 142
Total subsidy 3069 2741 2799 1817 1398 633 0 0 119 0 142
Total excess (deficit) 231 (754) 1893 2910 2275 536 3316 3782 3198 2802 663
Source IFPRI analysis of NIA data
- 21 shy
Table 2--Descriptive characteristics of selected MIA systeasa
------- Region III -------- ----- Region VI ----shyUPRIIS Angatii Sto Sibalom- Aganan-
Haasim Thomasc San Jose Sta Barbara
Average service area (hal 102272 31462 3522 5282 8703 Average irrigated area (hal
Wet season 83768 23454 3007 4410 8300 Dry season
Average benefited area Wet season
(hal 64587
77 605
27639
22908
1 781
3007
2801
4369
2770
7698 Dry season
Average rainfall Wet season
(mml d 62478
1 685 5
27396
8576
1 781
3051 0
2769
24731
2997
20001 Dry season 756 333 322 2828 3025
Average discharge (Llsec) Wet season 46501 14792 1692 2353 4984 Dry season 78091 22812 2014 1276 2315
Average water delivery (mmday) Wet season 522 548 487 462 571 Dry season
Average yield (mtha) 1089 715 995 398 686
Wet season 345 419 322 395 435 Dry season
Avg yield per unit water 34 03
(kgm ) 451 412 399 426
Wet season 0373 0440 0373 0538 0443 Dry season 0248 0400 0279 0690 0428
t-statistic difference in mean rainfall 1978-81 1982-86e
Wet season 0432 0713 -0567 1169 1169 Dry season 0519 -0230 -0523 1187 1187 Annual 0460 0707 -0686 1445 1445
~ Summary numbers are averages for the period 1982-1986 except as noted Water delivery discharge and yield per unit discharge are 4-year averages 1982-1985
c Water delivery discharge and yield per unit discharge are 4-year averages d 1983-1986
For Angat 5 years are 1981-85 For St Thomas 1979-83 For Sibalom 1971-75 e No significant differences at 95 confidence
- 22 shy
Table 3--Results of pooled regressions water delivery indicators
Dependent variable Independent variable - Water Delivery in mmBenefited Ha - --- BenefitService Area Ratio --shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season-shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 6799 (998)
6189 (1019)
13102 (1093)
2411 (1 54)
0743 ( 1186)
0800 (1165)
0129 (131 )
0258 (329)
System 2 indicator 0173 (029)
-3743 (-363)
-0063 (-219)
0200 (505)
System 3 indicator 0698 (085)
0163 (012)
0085 (208)
-0046 (-097)
System 4 indicator -0514 (-l 07)
-7011 (-803)
0080 (340)
0169 (l87)
System 5 indicator 0177 (033)
-5049 (-538)
0174 (603)
0051 (058)
Reservoir indicator 0327 (078)
5703 (543)
-0126 (-510)
0156 (261)
Period indicatora -0808 (-240)
-0771 (-229)
-1214 (-209)
-1309 (-1 62)
-0036 (-192)
-0025 (-112)
0071 (239)
0030 (077)
Water delivery mmb
Wet season
Dry season
Precipitation in mm Wet season
Annual
Average daily rainfall Wet season
Dry season
-46E-4 (-125)
-1 9E-4 (-082)
-43E-4 (-077)
20E-3 (385)
0015 (1 59)
-0001 ( -004)
0023 (215)
0001 (021)
0070 (613)
-0017 (-058)
0046 (371)
0010 (033)
R2 (adj) number of observations degrees of freedom
0156 43 36
0133 43 39
0711 41 34
0434 41 37
0693 42 34
0556 42 37
0865 40 32
0730 40 35
Coefficient (t statistic) indicates significance at 95 confidence ~ 1982-1986 1
per ha of service area
Ta
ble
--I
nd
ices
of
se
v
ice a
rea
ib
en
ef
ited
a
rea
a
nd
a
vera
ge sea
so
na
l d
isch
arg
e
Ave
rage
A
vera
ge
tshy19
77
1978
19
79
1980
19
81
1982
19
83
1984
19
85
1986
19
77-8
1 19
82-8
6 S
tati
stic
a
(ind
ex
aver
age
1983
-198
5 =
100)
Ser
vice
are
a UP
RI IS
in
dex
882
91
9
920
91
5
951
95
1
100
0 10
00
100
0 10
00
917
99
0
bull5
53
Ang
at-M
aasi
m R
95
9
994
99
7
996
99
6
996
10
00
100
0 10
00
100
0 98
8
999
1
63
Sto
To
mas
10
63
106
3 10
1 9
10
29
103
0 99
8
100
0 10
00
100
0 11
10
104
1 10
22
-08
9
Siba
lom
-San
Jos
e 95
0
872
94
2
933
94
2
942
10
28
102
8 94
4
101
7
928
99
2
292
A
gana
n-St
a B
arba
ra
108
3 10
91
106
0 10
30
961
10
05
996
10
08
996
99
6
104
5 10
00
-21
1 A
vera
ge
987
98
8
981
98
1
916
97
8
100
5 10
07
988
10
25
984
10
01
123
Wet
seas
on b
enef
ited
are
a in
dex
UPR
IIS
110
0 10
07
114
4 10
55
113
8 11
74
951
10
71
971
ll
58
10
89
106
7 -0
47
Ang
at-M
aasi
m R
97
9
974
92
0
983
10
28
100
8 99
7
102
1 98
2
931
97
7
988
0
54
Sto
To
mas
11
63
115
9 11
23
107
5 10
80
103
9 98
1
978
10
41
103
9 11
20
101
6 -4
88
Siba
lom
-San
Jos
e 11
35
103
8 10
11
931
93
4
906
10
07
975
10
1S
998
10
11
981
-0
80
Aga
nan-
Sta
Bar
bara
10
85
110
5 10
74
106
8 10
01
104
2 99
8
101
5
987
61
0
106
7 93
0
-1 8
4
Ave
rage
10
92
105
6 10
54
102
4 10
36
103
4 98
7
101
3 10
00
947
10
53
996
-2
32
Dry
sea
Son
bene
fite
d ar
ea
inde
x U
PRIIS
14
04
155
0 15
S0
155
8 16
17
128
0 57
2
114
S 15
74
152
3 12
38
-09
4 A
ngat
-Maa
sim
R
903
93
0
103
2 10
61
104
2 10
69
988
99
2
102
1 99
6
993
10
13
061
S
to
Tom
as
105
7 12
27
122
5
961
99
9
115
9 10
1 7
91
0
107
3 12
1 2
10
94
107
4 -0
28
Si
balo
m-S
an J
ose
Aga
nan-
Sta
Bar
bara
66
5
95S
62
6
632
67
4
111
4
501
10
S1
412
11
51
766
93
3
856
94
3
107
0 94
8
107
4 11
09
111
6
158
6 58
8
987
97
6
110
4 5
35
083
N
w
Ave
rage
89
6
964
11
19
103
7 10
44
110
9 10
1 7
89
8
108
5 12
97
101
7 10
81
082
Wet
seas
on d
isch
arge
in
dex
UPR
IIS
132
9 72
7
142
5 11
88
120
4 98
8
105
8 10
63
879
96
4
117
5 99
1
-16
5 A
ngat
-Maa
sim
R
129
7 13
52
134
5 12
58
127
0 11
70
120
3 62
7
131
3 10
68
-18
9
Sto
To
mas
14
71
155
0 14
1 6
10
40
725
12
35
112
3 14
79
103
1 -4
48
Siba
lom
-San
Jos
e 96
8
733
11
1 5
92
2
907
47
1
102
3 85
7
ll2
0
422
92
9
779
-1
09
Aga
nan-
Sta
Bar
bara
87
9
863
68
1
925
96
8
110
7 70
5
871
87
7
00
9
Ave
rage
11
49
105
7 13
61
115
0 10
58
853
10
43
963
99
4
SO3
11
55
940
-2
85
Dry
sea
son
disc
harg
e in
dex
UPR
IIS
425
13
09
153
3 14
28
180
6 14
S0
125
8 56
3
117
9 14
66
130
0 11
89
-04
3
Ang
at-M
aasi
m R
12
26
161
6 15
30
139
2 12
85
100
8 10
33
958
14
41
107
1 -3
80
S
to
Tom
as
116
4 14
82
155
7 79
9
971
12
30
126
5 14
01
106
6 -2
44
Si
balo
m-S
an J
ose
486
71
5
782
62
9
789
64
5
116
9 11
86
801
65
3
918
2
36
Aga
nan-
Sta
Bar
bara
A
vera
ge
425
10
46
133
6 62
0
118
4 81
5
116
1 94
7
112
5 89
7
922
12
20
991
88
2
108
7 12
36
119
2 71
8
114
0 10
37
105
5 3
23
-07
5
a bull
indi
cate
s si
gnif
ican
ce a
t 95
per
cent
con
fide
nce
- 24 shy
Table 5--Results of pooled regressions yield indicators
------------------------- Dependent variables -------------------------shyIndependent variable ---------- Yield MTHa ----------- -- Specific Yield 1000 KgM
3 H20 shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season -shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 1967 (286)
3747 (580)
1991 (293)
2800 (460 )
0084 (041)
0441 (229)
0049 (023 )
0876 (343)
System 2 indicator 0400 (155)
0525 (226)
0057 (085)
0115 (1 95)
System 3 indicator 0466 (139 )
-0029 (-010)
-0016 (-018)
0006 (008)
System 4 indicator 0980 (470)
0090 (049)
0176 (333)
0391 (827)
System 5 indicator 1 221 (570)
0256 ( 138)
0104 (174)
0224 (440)
Reservoir indicator -1033 (-578)
-0063 (-039)
-0146 (-310)
-0326 (-569)
Period indicatora
Precipitation in mm Wet season
Annual
0163 (1 03)
-29E-4 (-182)
0231 (1 42)
-58E-4 (-606)
0161 (118)
92E-5 (072)
0199 043 )
-64pound-5 (-084)
0082 (200)
57E-6 (013)
0092 (223)
-61E-5 (-232)
0032 (093)
29E-7 (001)
0052 (112)
-12pound-4 (-414)
Nitrogen applicationb 0025 (254)
0021 (211)
0025 (274)
0020 (227)
0003 (1 05)
0002 (043)
0003 (101)
-0002 (-052)
R2 (adj) 0553 0516 0251 0198 0309 0262 0713 0445 number of observations 50 50 49 49 42 42 40 40 degrees of freedom 42 45 41 44 34 37 32 35
CoeffiCient (t statistic) indicates significance at 95 confidence ~ 1982-1986 = 1
kgha
- 25 shy
Table 6--Suggmry of estiates perforance indicators
Variable Predicted Predicted Period Dependent Variable
Description (units)
Period Coefficient
1976-86 IndicatorO
1976-86 Indicatorl
Effecta (X)
BenefittedService area ratio
Wet season ratio -0036 0853 0817 -425 Dry Season 0071 0539 0610 1313
Delivery Wet Dry
season season
nm water per benefitted ha
-0808 -1 214
6003 9083
5195 7869
-1346 -1337
Rice Yield Wet season metric tons 0163 3578 3741 456 Dry season per hectare 0161 3905 4066 412
Specific Yield Wet season kg rice per m3 0082 0348 0430 2353 Dry season water de 1 ivered 0032 0378 0409 837
a percentage changes relative to period=O values
- 26 shy
ENDNOTES
IResearch Fellow International Food Policy Research Institute and Senior Irrigation Specialist International Irrigation Management Institute
2Presumable farmers also base private investment decisions on such foreknowledge of costs and commonly do so in the case of small tubewells In addition user groups may invest in small communal schemes independently after examining projected costs and benefits Most commonly however new irrigation capacity is created as a result of public investment decisions made by government and thus the political process comes into play
3This principle applies only when water charges are levied volumetrically--an exceedingly rare situation in the developing worldshy-though the argument is frequently invoked as though it applies to all allocative situations
4For Cabanatuan City the longer period of record available ie 1904 to 1986 was divided into two equal periods and tested for difference in means to test the observation by farmers that rainfall today is lower than it used to be Mean precipitation after 1947 was actually about 3 percent higher than before but the difference was not significant (t = 07)
5The procedure used was to take time series data on total nitrogenfertilizer consumption in the Philippines multiply it by the ratio of N fertilizer used on rice to total N consumption and divide the product by the rice harvested area to obtain an estimate of N use perhectare Values of the ratio of N use on rice to total N use were taken from IRRI (1988 150) with interpolation used to fill in gaps in the series
6This assumes that timeliness is evaluated against a standard based on the physiological demand for water generated by the rice crop and the varying sensitivity of the crop to stress at different growth stages In such a case a poor timeliness rating for irrigationservice will result in reductions in yield
7Because rice yields are insensitive to water applications that exceed PET values a redistribution of water from water excess to water short portions of the command area will usually result in higher aggregate output and higher average yields Exceptional scenarios can be constructed in which this general rule would be violated but such scenarios are unlikely to play out in practice
- 27 shy
aThere are a variety of standards for judging equity that could be used here The most common one is equality of per hectare water deliveries In the present case the implicit standard is equality of water supply utility in producing a crop Thus if two areas received equal amounts of water on a per hectare basis but one had such high seepage and percolation losses that the crop failed to produce a harvest (and was therefore exempted from irrigation fee payment) the water supply to the combined area would be judged inequitable Because such radical differences in seepage and percolation rates are unlikely on puddled rice soils in the lowland Philippines and because the definition of benefitted area accommodates a wide range of acceptable yields these two definitions are largely indistinguishable at the current limits of data resolution
- 8 shy
two time periods These are rainfall and level of use of other agricultural inputs Since our interest is in systems managerial responses to these changes and since there is no reason to believe that the relative magnitudes of the responses are dependent on the size of the system the five systems are treated as equal in the analysis That is changes in measured values relating to the smallest system are considered to be as important as changes in values for the largest with no area weighting applied
Since rainfall can substitute for irrigation water supplies and since it affects the supply of water available in rivers for irrigation it may exert some independent influence on various performance indicators To test the strength of the relationship for the period being analyzed simple correlations were run between rainfall and benefitted area for one system in each regionBenefitted area was used in this analysis because it is the variable deemed most likely to be influenced by year-to-year changes in rainfall Weather data from Cabanatuan City was used for the UPRIIS system which surrounds it and Iloilo City data was used for the nearby Aganan-Santa Barbara system For UPRIIS all of the R2 values for these correlations were less than 0005 suggesting that rainfall has almost no impact on area harvested in this large reservoir-based scheme For Aganan-Santa Barbara wet season rainfall was related to wet season BA (r2 = 016) and to BA during the following dry season (r2 = 024) Signs of the simple correlations were in the expected directions ie wet season rainfall increased BA during both the wet and the subsequent dry seasons These connections are understandable but weak
Another possibility is that there were longer-term differences in rainfall received in the two regions If this were the case a comparison of performance during two different time periods would have to take this difference into account Differences in average precipitation during the two periods were examined for the four stations used in the analysis (see Table 2) In no case were differences in seasonal or annual mean rainfall statisticallysignificant4
bull
Nevertheless in the regression approach adopted to analyze the data rainfall was included in each equation to control for its possible effect on the particular dependent variables being analyzed In doing this wet season rainfall was used in analyzing wet season performance indicators while annual rainfall was used in analyzingdry season data The rationale for this is that while dry season rainfall cannot possibly influence the wet season crop the dry season crop is affected by both the rainfall received directly and the rain falling during the preceding wet season through its effect on river discharge reservoir storage and antecedent soil moisture conditions
The level of agricultural production is also an often-used indicator of an irrigation systems performance Its major weakness
- 9 shy
is that a number of factors other than irrigation service such as labor inputs relative prices and fertilizer use influence it It is necessary therefore either to control for changes in the levels of these inputs or assume that they are constant across the two periods being compared In the present case the most important of these factors is the level of application of chemical fertilizer Because fertilizer use by farmers is responsive to the relative pricesof fertilizer and rice it also includes to some extent input and output price effects Since reliable data on fertilizer use for individual systems were not available estimates derived from FAO fertilizer and cultivated rice area data were used t~ control for the effect of changes in the use of this input over time This variable was included in any of the regression equations in which agriculturalproduction was used as the dependent variable Other factors such as labor use genetic potential of varieties sewn and pesticideapplications are assumed to be constant across the two periods
The analytic approach employed is to fit linear regressionequations to pooled data from the five systems covering a eleven-year period 1976 to 1986 A dummy variable is used to check the impact of pre and post 1981 periods on differences in the dependent variable after the effects of factors such as rainfall and nitrogen fertilizer use have been removed In addition because the dataset was created by pooling data from five different systems a set of 4 site dummies was included in the basic model to control for system-specific differences caused by variables which were not measured For some runs these were replaced with dummies that separated reservoir and non-reservoir systems though equations using the reservoir dummy were consistently inferior to those using the complete set of site dummies Several different dependent variables were created to index the quality of irrigation service and tested using this approachRegression results are given in Tables 3 and 5 and discussed below
Performance Indicators
A variety of indicators have been used in evaluating irrigationperformance in various contexts The selection of appropriate indicators depends on a number of factors including the purpose of the evaluation the audience for its results the way in which the boundaries of the irrigation system are defined and the kind and quality of data available to the evaluators The current analysis is designed to evaluate the impact of a set of management changes on system physical performance The audience for this analysis comprises top-level managers of the irrigation agency and policy-makers at higher government levels Boundary definition is an importantanalytic problem here as evident from the subsequent discussion relating to the choice of the appropriate area values to use in scaling system inputs and outputs This issue is also related to the data quality and availability problems which have already been mentioned
- 10 shy
Three fundamental indicators have been proposed for assessing the effectiveness of irrigation services to cultivators (Svendsen and Small 1989 Abernethy 1990) These are the adequacy of water supplies the equity of their distribution across the command area of the system and the timeliness of the supplies Computation of adequacy measures requires information on the total quantity of water delivered to the system over a season on a per hectare basis Equityand timeliness measures require information on the spatial and temporal distribution respectively of those supplies Where appropriate discharge information is not available proxies can be employed by making suitable assumptions Standards must be selected against which the magnitude of the indicators can be judged
The task in the present case however is somewhat different Here the need is to evaluate changes in selected variables between the pre-1981 and the post-1981 periods Hence the absolute values of variables selected are less important than their relative magnitudesand the statistical significance of the differences in magnitudes between the two periods A distinct limitation is imposed by the data series available for the five sample systems Since discharge and yield data are available only on a whole-system basis it is impossible to develop measures of equity and timeliness directly We will however extend our analysis to a discussion of equity byindirect inference Levine and Coward (1986) have argued that equityought to be considered as the paramount objective in managing largepublic irrigation systems They base their conclusion on an analysisof eight small community-managed systems and five larger public systems including UPRIIS in which equity appears to comprise the most important operational objective in the successful systems It may be appropriate therefore to give success in improving equity of distribution added weight in assessing overall performance
Area Estimates Because measures of system agricultural output and water supplied are typically reduced to a unit area basis before being used much depends on the area values which are used to standardize them Two different area measures are available The first is Service Area (SA) which is defined as the irrigable portion of the command area which is provided with physical facilities for water delivery This represents the area which could conceivably be irrigated in a given season if water supply were not constraining This value may change somewhat from year to year in response to urban encroachment on irrigated command minor remodeling and repair and refinements in area estimates In the present case though the systems selected for analysis were chosen to avoid those which had undergone more extensive rehabilitation or modification
The second measure is Benefitted Area (BA) which is the area billed for payment of irrigation service fees It is the irrigated area harvested which did not have yields so low that the farm was exempted from payment of fees in a given season This threshold value has been approximately 2 tons of paddy per hectare Benefitted area
- 11 shy
varies more than does SA particularly during the dry season when available water supply may seriously constrain the area which can be planted Its magnitude is a function of system managers actions in authorizing the amount of land to be planted in a given season farmers decisions regarding whether to plant or not and the combined ability of system managers and farmersirrigators subsequently to distribute water Both of these area measures will be used to standardize other variables for particular purposes as well as beingcombined to form a separate indicator by themselves
Adequacy The most direct measure of the adequacy of irrigation water supplies to the agricultural system is the quantity of water applied to the system command area on a per unit area basis relative to some standard In this case since our interest is in differences in water adequacy between two time periods and since the systems being assessed have been and continue to be almost entirely devoted to rice cultivation during both cropping seasons depth measures for the two periods may be compared directly assuming the seasonal cropdemand for water to be unchanged Although dry-footed crops can suffer yield losses from overapplication of irrigation water rice is largely insensitive to this effect In addition water can substitute for other inputs that the farmer would otherwise have to provide such as weed control and more careful (and costly) water management We assume therefore that other things being equal larger values of depth applied are better than smaller values in terms of meeting crop water demands and reduce the cost of cultivation At the same time high levels of water adequacy can affect the values of other performance measures--particularly equity
When the regression model is run for quantity of water diverted at the system headworks divided by BA hereafter termed depth we see that the period dummy is negative and significant at the 95 percent confidence level for both wet and dry seasons (see Table 3 equations 1 and 3) Since the overall explanatory power of the wet season model is very weak however we will focus on the dry season in interpreting this result which indicates that after adjusting for rainfall differences significantly less water was delivered to the command per unit of benefitted area following 1981 than before This indicates based on the criteria outlined above that performance in terms of water adequacy deteriorated following financial selfshysufficiency We need to examine this conclusion more carefullyhowever
One difficulty is that the measured quantity of water diverted at the source is largely a function of the supply available in the river rather than of system management This is particularly true during the dry season and in non-reservoir systems Thus while the depth of water supplied to the system is a measure of the adequacy of the systems service it is to some extent beyond the control of the managing agency To better understand the factors behind this decline in water availability we look at simple unadjusted index values for
- 12 shy
several of the key variables Table 4 shows annual values of total volume of water delivered in each season SA and BA in both wet and dry seasons and shows the results of t-tests on the means of a set of indicators before and after 1981 Indicators are used rather than the actual values to weight each of the systems equally regardless of its size The table shows that both the average wet season benefitted area and the average discharge are significantlylower during the second period compared with the first For the dry season too the discharge index is lower after 1981 than before but this difference is not significant At the same time the dry season BA index rose slightly but again the change was not significantSince middotthere is not a clear pattern of relative movement of discharge and BA during the respective seasons no simple interpretation of these index value changes is possible What stands out is that both discharge and benefitted area declined across periods during the wet season while during the dry season there was no significant change in either indicator across the two periods It seems clear that the decline in water adequacy must be evaluated together with other measures of performance in drawing conclusions about the overall impact of the 1981 changes on the quality of system management
Another measured variable per hectare yield can be used as a proxy for water adequacy It has the advantage of partiallyreflecting the impacts of the dimensions of timeliness6 and equity7of distribution as well integrating all three effects into a combined impact on aggregate crop production Table 5 (equations 1 and 3)shows that the period dummy in the yield regressions has a positive sign in both seasons after controlling for nitrogen application and precipitation though the t-values are not significant Treatingyield adjusted in this way as a proxy for quality of irrigationservice leads to the conclusion that by this more comprehensive measure quality of service held constant across the two periods in the dry season Because of the large yield component accounted for byrainfall during the wet season no such judgement is possible for that season however
Equity As noted earlier no reliable data are available for subdivisions of the five sample systems making direct computation of equity measures impossible We can make some judgements about changesin the equity of water distribution however by examining changes in the ratio of two area measures given for each system SA and BA Since SA is the area which theoretically can be supplied with irrigation water by the system and BA s the area which actuallyreceives a quantity of water adequate to produce a remunerative crop the ratio of the two provides a measure of the percentage of the potential service area which was irrigated to a particular standard The larger this pErcentage the more equitable 8 is the distribution This of course assumes that the quantity of water available to the systems is constant across the two periods
- 13 shy
Since the condition of constancy of water supply is not generally satisfied a regression was run in which the total quantityof water diverted at the headworks of each system divided by the systems potential service area SA was included in the regression to control for changes in the water supply available to the system seeTable 3 equations 5 through 8 The average daily rainfall received directly on the system service area during the season was also included as an independent variable The regression was run separately for wet and dry seasons The sign and t-statistic of the period dummy should then tell us whether or not equity as reflected in the BASA ratio increased decreased or remained unchanged across the period divide
Both equations are reasonable good as indicated by the R2 values though the dry season equation is considerably better as would be expected For the wet season both the water delivery term and the rainfall term in equation 5 are of positive sign but are nonshysignificant at the 95 percent confidence level indicating that wet season irrigated area does not change appreciably in response to level of wet season rainfall or the available irrigation water supply The period dummy was negative but not significant indicating that equityof distribution as reflected in the BASA ratio was similar during the two periods
For the dry season the water delivery term in equation 7 is positive and strongly significant indicating a close relationship between the fraction of potential area actually irrigated and the water supply available at the headworks In addition however the period dummy is positive and significant suggesting that once the influence of water supply is removed the BASA ratio was significantly higher in the period following 1981 than it was before
This is an important finding for it reflects significantlyimproved performance in terms of a factor equity of water distribution that is under the control of the managing entity an entity which here comprises both NIA and irrigators associations Interpreted in these terms NIA and allied farmers associations were able to spread a given amount of water more widely across the potential command area of the five sample systems in the period after 1981 than before Moreover they did this in a way that did not decrease average system yields as discussed earlier In making this interpretation we are suggesting that there was some redistribution of water from better-watered areas to fringe areas which would otherwise not have received irrigation water and that this redistribution was a direct response to the change in NIA prioritiesand operating policies and rules occurring around 1981
It is difficult to prove the assertion that water was in fact redistributed with a resulting increase in directly-measured equity Without access to reliable discharge data broken out by systemsection and we can only assume in the absence of a plausible
- 14 shy
alternative explanation that it was such a redistribution that made the increase in the BASA ratio possible In a larger sense it is difficult to prove conclusively that any outcome in a before and after analysis was the result of a particular independent causative factor In this case we have tried to remove the influence of other potential causative factors where we could but the possibilityremains that some combination of unmeasured factors are responsiblefor the difference in the BASA ratio found We do note though that this type of response is exactly the type that would be expected to follow from an emphasis on increased farmer satisfaction and cooperation and increased fee revenues Because the fee schedule is tied to benefitted area the only ways NIA can increase its revenue from that source are to expand benefitted area and to increase collection efficiencies The former depends on redistributing a fixed supply of water over a larger portion of the command while the latter requires that farmers be satisfied with the irrigation service they are receiving and the commitment of the local irrigators association to assist in the task of collecting the amounts due The evidence while not conclusive is highly suggestive that this is exactly what has happened
Efficiency In addition to measures which reflect the levels of adequacy and equity of irrigation service available data allow the calculation of a measure of operating efficiency The term efficiencyusually denotes the relationship between inputs to a process and its outputs often expressed as a ratio The output measure employed here is aggregate system rice output and the input is quantity of irrigation water turned into the system Dividing the first by the second gives a measure of agricultural production per unit water--here termed specific yield This is a highly integrated measure that evaluates the combined efficiency of the irrigation and agricultural processes As such it is a function of the managerial and other inputs supplied both to the irrigation system and to the agricultural operation With respect to one important input to the irrigation system we do know that NIA per hectare field operating expenses were about 29 percent lower in real terms in the 1982-86 period comparedto the 1976-1981 period although this drop may have been partlyoffset by increases in farmer-supplied labor inputs Other things being equal one would thus expect to find a decline in output efficiency
The regression analysis shows positive signs for the period terms in both wet and dry season equations (see Table 5) In the case of the wet season the period dummy in equation 5 is significant but the overall explanatory power of the model is quite low For the dry season (equation 7) the coefficient is positive but non-significant This means that after taking rainfall and fertilizer use into account data do not indicate a lowering of specific yield in the wake of funding reductions and the strong emphasiS on financial viabilitybeginning in 1981 This result provides evidence that the efficiency of the overall irrigation deliveryagricultural production process
- 15 shy
relative to the system water input did not falloff as a result of the changes implemented at least over the short run
Impact magnitude
The preceding analysis has shown us that some indicators of irrigation performance changed significantly following the managerial changes of 1981 while others did not However it has not given us a sense of the size of the changes which occurred To determine the magnitude of these changes the regression model is used to predict the response of the composite system to the managerial changes given a common set of -input and environmental factors To do this averagevalues of the independent variables from the entire eleven-year period 1976 to 1986 are put into the model together with the previously determined coefficients to generate predicted average values of the various dependent variables used in the earlier analysis with and without the period dummy This procedure produces a pair of estimates for each dependent variable under the same conditions--one in which the system responds as it did after the managerial changes were implemented and one in which it responds as it did prior to their introduction The differences between these two values thus indicate the magnitude of the changes occurring in the various indicators of performance discussed above
The results of this exercise are shown in Table 6 The table shows that water availability decreased by about 13 percent in both wet and dry seasons when the period dummy was included and while the coefficients responsible were significant in the earlier analysisthis difference cannot be easily connected with levels of system management as discussed earlier With respect to rice output per hectare although the coefficients were not very significant it is interesting to note that yield increases by 163 kilograms per hectare for the wet season and by 101 kilogram for the dry when the period dummy is included in spite of the reduced water supply available Keep in mind that the predicted yield values have already been adjusted for differences in nitrogen fertilizer use and rainfall This suggests that timeliness and equity of distribution of water supply to farmers may have increased following the changes contributing to the higher predicted yields
Examining the impact of increased equity of distribution bylooking at the ratio of benefitted area to service area we recall that the change was positive and significant for the dry season and negative and not significant for the wet Table 6 shows that the dry season BASA ratio increases by 7 percentage points when the dummy is included a 131 percent increase Other things being equal this should result in a 131 percent increase in system output due to the expansion of ared benefitted This is a major impact on production
- 16 shy
CONCLUSIONS
The Philippine experiment to transform the national irrigation agency into an enterprise has undoubtedly been successful in reducing system operating expenses bringing revenues and costs into line and eliminating the recurrent cost burden imposed by large-scaleirrigation systems on the national budget Evidence presented in this paper indicates that in the process equity of water distribution across systems has also improved In the 5 years following the cessation of operating subsidies from the government an index of equity of distribution improved by about 13 percent At the same time per hectare yields adjusted for rainfall and nitrogen application held constant
There is a strong logical connection between the achievement of financial viability and improved equity of water distribution across the command Because increasing irrigation fees is a politicaldecisionlying largely beyond NIAs control expanding the area which can be billed for service is one of the few revenue increasing measures available to the irrigation agency which does not involve major additional investment In the face of constant or shrinkingwater supplies this is achieved only by redistributing water from areas receiving excessive supplies usually near the head ends of canals and laterals to areas receiving no supplies or inadequatesupplies often located near the tails of canals Although data are not available which would allow the direct examination of this hypotheses the two outcomes are logically consistent with each other
Data also show that per hectare water deliveries declined significantly in the five sample systems after 1981 even thoughrainfall did not differ appreciably between the two periods This decline averaged about 13 percent for both wet and dry seasons and is interpreted as a decline in water availability in the supplying rivers rather than a conscious reduction in withdrawals by system managers Such declines could result from changes in watershed runoff characteristics as caused by deforestation or from increased upstream abstractions from supplying rivers
Improved water distribution tends to increase the area served system agricultural output and NIA service fee revenue Reduced water supplies to the system tend to reduce these things Specificyield defined as system paddy output per unit water held roughly constant across the two periods indicating that the two effects mayhave offset each other
After adjusting for rainfall and nitrogen application perhectare yields increased only marginally in the post-1981 period Area served on the other hand increased by about 13 percent after adjusting for water supply availability indicating that the area benefitted by irrigation in the sample systems increased by about the same percentage Even if yields on this additional area are less than
- 17 shy
average yields for the system this still represents a sizeable increase in system agricultural output as a result of the change in management structure the increase coming not from higher yields but from expanded area under irrigation
The evidence assembled here suggests that there are significantfinancial and economic benefits to be had from changes in the basic character of irrigation managing agencies which make them more responsive to their clientele and which impose rational internal financial discipline on the agency The analysis suggests a number of additional questions however One relates to the longer-term impacts of the structural management changes The improvements in water distribution described here are relatively short-term events occurring during the first 5 years of the new management mode Critics have suggested the danger of underinvestment in systemmaintenance over the longer run accompanied by declining yields and benefitted areas and eventual system collapse This possibility needs to be closely monitored A second concern relates to the apparent decline in water supply to these 5 geographically dispersed systems The nature and causes of this decline need to be explored further since if widespread and secular it may represent a serious threat to the stability of Philippine rice production Whether stemming from poor forest management practices or deficient regulation and allocation of surface water resources or other unidentified factors it is an issue that deserves serious and urgent consideration
A third risk is that the incentive structure set up by NIA to guide and stimulate the performance of field units overemphasizes revenue generation at the expense of irrigation service provision to farmers The evidence presented here supports the view that these two objectives are mutually reinforcing under policies and conditions which have been established in the Philippines More detailed crossshysectional studies based on primary flow measurement data would add confidence to this conclusion and help to specify the conditions under which this effect occurs This could be extremely important in transferring the results of the Philippine experiment to other countries
A final risk is that outside intervention well meaning or otherwise will destroy the basis of NIAs financial autonomy or will impose external pressures or constraints on NIAs decision-making that will subvert the management practices which have been so painstakinglydeveloped and implemented Among these are calls for NIA to be subsumed again within the government department structure in the interests of better coordination with agriculture attempts byexternal financing agencies to arbitrarily increase NIAs expenditures on OampM on the assumption that this will increase system agricultural output or intervention by Philippine legislative bodies to restore operating subsidies to NIA with attached strings leading back to legislators home districts Pressures such as these will cut short a process of experimentation and improvement that seems promising
- 18 shy
enough to date to warrant its continuation Having developed the capacity to establish targets and implement and manage change NIA is in a strong position to modify its objectives to better achieve larger social purposes established for it It is critical to recognize however that this must happen within the context of financing policies that mandate financial autonomy for NIA if the fundamental institutional commitment to manage is to be preserved
The author would like to thank Leslie Small and JeremyBerkoff for helpful comments on an earlier unpublishedversion of this paper and Charles Rogers for his careful and creative help with the analysis
- 19 shy
BIBLIOGRAPHY
Abernethy Charles L 1990 Indicators of the performance of irrigation water distribution systems International Irrigation Management Institute Colombo Sri Lanka Mimeo
Asian Development Bank 1986 Irrigation service fees Proceedingsof the Regional Seminar on Irrigation Service Fees Manila Asian Development Bank
Carruthers Ian and Colin Clark 1981 Economics of IrrigationLiverpool Liverpool University Press Third Edition
Levine G and EW Coward Jr 1986 Irrigation water distribution implications for design and operation AGREP Division WorkingPaper 125 vol 1 World Bank Agriculture and Rural Development Department
Small Les E 1989 User charges in irrigation potentials and limitations Irrigation and drainage vol 3 no 2125-142
Small Les 1990 Irrigation service fees in Asia IrrigationManagement Network 9013 London Overseas DevelopmentInstitute
Svendsen Mark and Les Small 1989 A framework for assessing irrigation system performance Paper prepared for the Symposium on Performance Evaluation 23 November International IrrigationManagement Institute Sri Lanka
Table I--National Irrigation Administration revenues and expenditures in constant prices 1976-86
Item 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986
(peso million 1972)
Revenues Irrigation fees collected 1273 1483 17 13 1831 2070 1668 1699 1893 1728 2129 2546 Other income 715 737 2420 5591 2631 5990 7783 6638 5699 4932 2934
Total direct revenue 1988 2220 4133 7422 4701 7658 9482 8531 7427 061 5480
Expenses in 1972 pricesTotal expenses 4825 5716 5039 6329 3821 77 55 6166 4749 4348 4259 4959
Excess (deficit) (2837(3496) (906) 1093 877 (097) 3316 3782 3079 2802 521 N 0
Subsidies Government operation and
maintenance subsidies 2521 2741 2799 1817 1398 633 0 0 0 0 0 Calamity fund payments 548 0 0 0 0 0 0 0 119 0 142
Total subsidy 3069 2741 2799 1817 1398 633 0 0 119 0 142
Total excess (deficit) 231 (754) 1893 2910 2275 536 3316 3782 3198 2802 663
Source IFPRI analysis of NIA data
- 21 shy
Table 2--Descriptive characteristics of selected MIA systeasa
------- Region III -------- ----- Region VI ----shyUPRIIS Angatii Sto Sibalom- Aganan-
Haasim Thomasc San Jose Sta Barbara
Average service area (hal 102272 31462 3522 5282 8703 Average irrigated area (hal
Wet season 83768 23454 3007 4410 8300 Dry season
Average benefited area Wet season
(hal 64587
77 605
27639
22908
1 781
3007
2801
4369
2770
7698 Dry season
Average rainfall Wet season
(mml d 62478
1 685 5
27396
8576
1 781
3051 0
2769
24731
2997
20001 Dry season 756 333 322 2828 3025
Average discharge (Llsec) Wet season 46501 14792 1692 2353 4984 Dry season 78091 22812 2014 1276 2315
Average water delivery (mmday) Wet season 522 548 487 462 571 Dry season
Average yield (mtha) 1089 715 995 398 686
Wet season 345 419 322 395 435 Dry season
Avg yield per unit water 34 03
(kgm ) 451 412 399 426
Wet season 0373 0440 0373 0538 0443 Dry season 0248 0400 0279 0690 0428
t-statistic difference in mean rainfall 1978-81 1982-86e
Wet season 0432 0713 -0567 1169 1169 Dry season 0519 -0230 -0523 1187 1187 Annual 0460 0707 -0686 1445 1445
~ Summary numbers are averages for the period 1982-1986 except as noted Water delivery discharge and yield per unit discharge are 4-year averages 1982-1985
c Water delivery discharge and yield per unit discharge are 4-year averages d 1983-1986
For Angat 5 years are 1981-85 For St Thomas 1979-83 For Sibalom 1971-75 e No significant differences at 95 confidence
- 22 shy
Table 3--Results of pooled regressions water delivery indicators
Dependent variable Independent variable - Water Delivery in mmBenefited Ha - --- BenefitService Area Ratio --shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season-shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 6799 (998)
6189 (1019)
13102 (1093)
2411 (1 54)
0743 ( 1186)
0800 (1165)
0129 (131 )
0258 (329)
System 2 indicator 0173 (029)
-3743 (-363)
-0063 (-219)
0200 (505)
System 3 indicator 0698 (085)
0163 (012)
0085 (208)
-0046 (-097)
System 4 indicator -0514 (-l 07)
-7011 (-803)
0080 (340)
0169 (l87)
System 5 indicator 0177 (033)
-5049 (-538)
0174 (603)
0051 (058)
Reservoir indicator 0327 (078)
5703 (543)
-0126 (-510)
0156 (261)
Period indicatora -0808 (-240)
-0771 (-229)
-1214 (-209)
-1309 (-1 62)
-0036 (-192)
-0025 (-112)
0071 (239)
0030 (077)
Water delivery mmb
Wet season
Dry season
Precipitation in mm Wet season
Annual
Average daily rainfall Wet season
Dry season
-46E-4 (-125)
-1 9E-4 (-082)
-43E-4 (-077)
20E-3 (385)
0015 (1 59)
-0001 ( -004)
0023 (215)
0001 (021)
0070 (613)
-0017 (-058)
0046 (371)
0010 (033)
R2 (adj) number of observations degrees of freedom
0156 43 36
0133 43 39
0711 41 34
0434 41 37
0693 42 34
0556 42 37
0865 40 32
0730 40 35
Coefficient (t statistic) indicates significance at 95 confidence ~ 1982-1986 1
per ha of service area
Ta
ble
--I
nd
ices
of
se
v
ice a
rea
ib
en
ef
ited
a
rea
a
nd
a
vera
ge sea
so
na
l d
isch
arg
e
Ave
rage
A
vera
ge
tshy19
77
1978
19
79
1980
19
81
1982
19
83
1984
19
85
1986
19
77-8
1 19
82-8
6 S
tati
stic
a
(ind
ex
aver
age
1983
-198
5 =
100)
Ser
vice
are
a UP
RI IS
in
dex
882
91
9
920
91
5
951
95
1
100
0 10
00
100
0 10
00
917
99
0
bull5
53
Ang
at-M
aasi
m R
95
9
994
99
7
996
99
6
996
10
00
100
0 10
00
100
0 98
8
999
1
63
Sto
To
mas
10
63
106
3 10
1 9
10
29
103
0 99
8
100
0 10
00
100
0 11
10
104
1 10
22
-08
9
Siba
lom
-San
Jos
e 95
0
872
94
2
933
94
2
942
10
28
102
8 94
4
101
7
928
99
2
292
A
gana
n-St
a B
arba
ra
108
3 10
91
106
0 10
30
961
10
05
996
10
08
996
99
6
104
5 10
00
-21
1 A
vera
ge
987
98
8
981
98
1
916
97
8
100
5 10
07
988
10
25
984
10
01
123
Wet
seas
on b
enef
ited
are
a in
dex
UPR
IIS
110
0 10
07
114
4 10
55
113
8 11
74
951
10
71
971
ll
58
10
89
106
7 -0
47
Ang
at-M
aasi
m R
97
9
974
92
0
983
10
28
100
8 99
7
102
1 98
2
931
97
7
988
0
54
Sto
To
mas
11
63
115
9 11
23
107
5 10
80
103
9 98
1
978
10
41
103
9 11
20
101
6 -4
88
Siba
lom
-San
Jos
e 11
35
103
8 10
11
931
93
4
906
10
07
975
10
1S
998
10
11
981
-0
80
Aga
nan-
Sta
Bar
bara
10
85
110
5 10
74
106
8 10
01
104
2 99
8
101
5
987
61
0
106
7 93
0
-1 8
4
Ave
rage
10
92
105
6 10
54
102
4 10
36
103
4 98
7
101
3 10
00
947
10
53
996
-2
32
Dry
sea
Son
bene
fite
d ar
ea
inde
x U
PRIIS
14
04
155
0 15
S0
155
8 16
17
128
0 57
2
114
S 15
74
152
3 12
38
-09
4 A
ngat
-Maa
sim
R
903
93
0
103
2 10
61
104
2 10
69
988
99
2
102
1 99
6
993
10
13
061
S
to
Tom
as
105
7 12
27
122
5
961
99
9
115
9 10
1 7
91
0
107
3 12
1 2
10
94
107
4 -0
28
Si
balo
m-S
an J
ose
Aga
nan-
Sta
Bar
bara
66
5
95S
62
6
632
67
4
111
4
501
10
S1
412
11
51
766
93
3
856
94
3
107
0 94
8
107
4 11
09
111
6
158
6 58
8
987
97
6
110
4 5
35
083
N
w
Ave
rage
89
6
964
11
19
103
7 10
44
110
9 10
1 7
89
8
108
5 12
97
101
7 10
81
082
Wet
seas
on d
isch
arge
in
dex
UPR
IIS
132
9 72
7
142
5 11
88
120
4 98
8
105
8 10
63
879
96
4
117
5 99
1
-16
5 A
ngat
-Maa
sim
R
129
7 13
52
134
5 12
58
127
0 11
70
120
3 62
7
131
3 10
68
-18
9
Sto
To
mas
14
71
155
0 14
1 6
10
40
725
12
35
112
3 14
79
103
1 -4
48
Siba
lom
-San
Jos
e 96
8
733
11
1 5
92
2
907
47
1
102
3 85
7
ll2
0
422
92
9
779
-1
09
Aga
nan-
Sta
Bar
bara
87
9
863
68
1
925
96
8
110
7 70
5
871
87
7
00
9
Ave
rage
11
49
105
7 13
61
115
0 10
58
853
10
43
963
99
4
SO3
11
55
940
-2
85
Dry
sea
son
disc
harg
e in
dex
UPR
IIS
425
13
09
153
3 14
28
180
6 14
S0
125
8 56
3
117
9 14
66
130
0 11
89
-04
3
Ang
at-M
aasi
m R
12
26
161
6 15
30
139
2 12
85
100
8 10
33
958
14
41
107
1 -3
80
S
to
Tom
as
116
4 14
82
155
7 79
9
971
12
30
126
5 14
01
106
6 -2
44
Si
balo
m-S
an J
ose
486
71
5
782
62
9
789
64
5
116
9 11
86
801
65
3
918
2
36
Aga
nan-
Sta
Bar
bara
A
vera
ge
425
10
46
133
6 62
0
118
4 81
5
116
1 94
7
112
5 89
7
922
12
20
991
88
2
108
7 12
36
119
2 71
8
114
0 10
37
105
5 3
23
-07
5
a bull
indi
cate
s si
gnif
ican
ce a
t 95
per
cent
con
fide
nce
- 24 shy
Table 5--Results of pooled regressions yield indicators
------------------------- Dependent variables -------------------------shyIndependent variable ---------- Yield MTHa ----------- -- Specific Yield 1000 KgM
3 H20 shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season -shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 1967 (286)
3747 (580)
1991 (293)
2800 (460 )
0084 (041)
0441 (229)
0049 (023 )
0876 (343)
System 2 indicator 0400 (155)
0525 (226)
0057 (085)
0115 (1 95)
System 3 indicator 0466 (139 )
-0029 (-010)
-0016 (-018)
0006 (008)
System 4 indicator 0980 (470)
0090 (049)
0176 (333)
0391 (827)
System 5 indicator 1 221 (570)
0256 ( 138)
0104 (174)
0224 (440)
Reservoir indicator -1033 (-578)
-0063 (-039)
-0146 (-310)
-0326 (-569)
Period indicatora
Precipitation in mm Wet season
Annual
0163 (1 03)
-29E-4 (-182)
0231 (1 42)
-58E-4 (-606)
0161 (118)
92E-5 (072)
0199 043 )
-64pound-5 (-084)
0082 (200)
57E-6 (013)
0092 (223)
-61E-5 (-232)
0032 (093)
29E-7 (001)
0052 (112)
-12pound-4 (-414)
Nitrogen applicationb 0025 (254)
0021 (211)
0025 (274)
0020 (227)
0003 (1 05)
0002 (043)
0003 (101)
-0002 (-052)
R2 (adj) 0553 0516 0251 0198 0309 0262 0713 0445 number of observations 50 50 49 49 42 42 40 40 degrees of freedom 42 45 41 44 34 37 32 35
CoeffiCient (t statistic) indicates significance at 95 confidence ~ 1982-1986 = 1
kgha
- 25 shy
Table 6--Suggmry of estiates perforance indicators
Variable Predicted Predicted Period Dependent Variable
Description (units)
Period Coefficient
1976-86 IndicatorO
1976-86 Indicatorl
Effecta (X)
BenefittedService area ratio
Wet season ratio -0036 0853 0817 -425 Dry Season 0071 0539 0610 1313
Delivery Wet Dry
season season
nm water per benefitted ha
-0808 -1 214
6003 9083
5195 7869
-1346 -1337
Rice Yield Wet season metric tons 0163 3578 3741 456 Dry season per hectare 0161 3905 4066 412
Specific Yield Wet season kg rice per m3 0082 0348 0430 2353 Dry season water de 1 ivered 0032 0378 0409 837
a percentage changes relative to period=O values
- 26 shy
ENDNOTES
IResearch Fellow International Food Policy Research Institute and Senior Irrigation Specialist International Irrigation Management Institute
2Presumable farmers also base private investment decisions on such foreknowledge of costs and commonly do so in the case of small tubewells In addition user groups may invest in small communal schemes independently after examining projected costs and benefits Most commonly however new irrigation capacity is created as a result of public investment decisions made by government and thus the political process comes into play
3This principle applies only when water charges are levied volumetrically--an exceedingly rare situation in the developing worldshy-though the argument is frequently invoked as though it applies to all allocative situations
4For Cabanatuan City the longer period of record available ie 1904 to 1986 was divided into two equal periods and tested for difference in means to test the observation by farmers that rainfall today is lower than it used to be Mean precipitation after 1947 was actually about 3 percent higher than before but the difference was not significant (t = 07)
5The procedure used was to take time series data on total nitrogenfertilizer consumption in the Philippines multiply it by the ratio of N fertilizer used on rice to total N consumption and divide the product by the rice harvested area to obtain an estimate of N use perhectare Values of the ratio of N use on rice to total N use were taken from IRRI (1988 150) with interpolation used to fill in gaps in the series
6This assumes that timeliness is evaluated against a standard based on the physiological demand for water generated by the rice crop and the varying sensitivity of the crop to stress at different growth stages In such a case a poor timeliness rating for irrigationservice will result in reductions in yield
7Because rice yields are insensitive to water applications that exceed PET values a redistribution of water from water excess to water short portions of the command area will usually result in higher aggregate output and higher average yields Exceptional scenarios can be constructed in which this general rule would be violated but such scenarios are unlikely to play out in practice
- 27 shy
aThere are a variety of standards for judging equity that could be used here The most common one is equality of per hectare water deliveries In the present case the implicit standard is equality of water supply utility in producing a crop Thus if two areas received equal amounts of water on a per hectare basis but one had such high seepage and percolation losses that the crop failed to produce a harvest (and was therefore exempted from irrigation fee payment) the water supply to the combined area would be judged inequitable Because such radical differences in seepage and percolation rates are unlikely on puddled rice soils in the lowland Philippines and because the definition of benefitted area accommodates a wide range of acceptable yields these two definitions are largely indistinguishable at the current limits of data resolution
- 9 shy
is that a number of factors other than irrigation service such as labor inputs relative prices and fertilizer use influence it It is necessary therefore either to control for changes in the levels of these inputs or assume that they are constant across the two periods being compared In the present case the most important of these factors is the level of application of chemical fertilizer Because fertilizer use by farmers is responsive to the relative pricesof fertilizer and rice it also includes to some extent input and output price effects Since reliable data on fertilizer use for individual systems were not available estimates derived from FAO fertilizer and cultivated rice area data were used t~ control for the effect of changes in the use of this input over time This variable was included in any of the regression equations in which agriculturalproduction was used as the dependent variable Other factors such as labor use genetic potential of varieties sewn and pesticideapplications are assumed to be constant across the two periods
The analytic approach employed is to fit linear regressionequations to pooled data from the five systems covering a eleven-year period 1976 to 1986 A dummy variable is used to check the impact of pre and post 1981 periods on differences in the dependent variable after the effects of factors such as rainfall and nitrogen fertilizer use have been removed In addition because the dataset was created by pooling data from five different systems a set of 4 site dummies was included in the basic model to control for system-specific differences caused by variables which were not measured For some runs these were replaced with dummies that separated reservoir and non-reservoir systems though equations using the reservoir dummy were consistently inferior to those using the complete set of site dummies Several different dependent variables were created to index the quality of irrigation service and tested using this approachRegression results are given in Tables 3 and 5 and discussed below
Performance Indicators
A variety of indicators have been used in evaluating irrigationperformance in various contexts The selection of appropriate indicators depends on a number of factors including the purpose of the evaluation the audience for its results the way in which the boundaries of the irrigation system are defined and the kind and quality of data available to the evaluators The current analysis is designed to evaluate the impact of a set of management changes on system physical performance The audience for this analysis comprises top-level managers of the irrigation agency and policy-makers at higher government levels Boundary definition is an importantanalytic problem here as evident from the subsequent discussion relating to the choice of the appropriate area values to use in scaling system inputs and outputs This issue is also related to the data quality and availability problems which have already been mentioned
- 10 shy
Three fundamental indicators have been proposed for assessing the effectiveness of irrigation services to cultivators (Svendsen and Small 1989 Abernethy 1990) These are the adequacy of water supplies the equity of their distribution across the command area of the system and the timeliness of the supplies Computation of adequacy measures requires information on the total quantity of water delivered to the system over a season on a per hectare basis Equityand timeliness measures require information on the spatial and temporal distribution respectively of those supplies Where appropriate discharge information is not available proxies can be employed by making suitable assumptions Standards must be selected against which the magnitude of the indicators can be judged
The task in the present case however is somewhat different Here the need is to evaluate changes in selected variables between the pre-1981 and the post-1981 periods Hence the absolute values of variables selected are less important than their relative magnitudesand the statistical significance of the differences in magnitudes between the two periods A distinct limitation is imposed by the data series available for the five sample systems Since discharge and yield data are available only on a whole-system basis it is impossible to develop measures of equity and timeliness directly We will however extend our analysis to a discussion of equity byindirect inference Levine and Coward (1986) have argued that equityought to be considered as the paramount objective in managing largepublic irrigation systems They base their conclusion on an analysisof eight small community-managed systems and five larger public systems including UPRIIS in which equity appears to comprise the most important operational objective in the successful systems It may be appropriate therefore to give success in improving equity of distribution added weight in assessing overall performance
Area Estimates Because measures of system agricultural output and water supplied are typically reduced to a unit area basis before being used much depends on the area values which are used to standardize them Two different area measures are available The first is Service Area (SA) which is defined as the irrigable portion of the command area which is provided with physical facilities for water delivery This represents the area which could conceivably be irrigated in a given season if water supply were not constraining This value may change somewhat from year to year in response to urban encroachment on irrigated command minor remodeling and repair and refinements in area estimates In the present case though the systems selected for analysis were chosen to avoid those which had undergone more extensive rehabilitation or modification
The second measure is Benefitted Area (BA) which is the area billed for payment of irrigation service fees It is the irrigated area harvested which did not have yields so low that the farm was exempted from payment of fees in a given season This threshold value has been approximately 2 tons of paddy per hectare Benefitted area
- 11 shy
varies more than does SA particularly during the dry season when available water supply may seriously constrain the area which can be planted Its magnitude is a function of system managers actions in authorizing the amount of land to be planted in a given season farmers decisions regarding whether to plant or not and the combined ability of system managers and farmersirrigators subsequently to distribute water Both of these area measures will be used to standardize other variables for particular purposes as well as beingcombined to form a separate indicator by themselves
Adequacy The most direct measure of the adequacy of irrigation water supplies to the agricultural system is the quantity of water applied to the system command area on a per unit area basis relative to some standard In this case since our interest is in differences in water adequacy between two time periods and since the systems being assessed have been and continue to be almost entirely devoted to rice cultivation during both cropping seasons depth measures for the two periods may be compared directly assuming the seasonal cropdemand for water to be unchanged Although dry-footed crops can suffer yield losses from overapplication of irrigation water rice is largely insensitive to this effect In addition water can substitute for other inputs that the farmer would otherwise have to provide such as weed control and more careful (and costly) water management We assume therefore that other things being equal larger values of depth applied are better than smaller values in terms of meeting crop water demands and reduce the cost of cultivation At the same time high levels of water adequacy can affect the values of other performance measures--particularly equity
When the regression model is run for quantity of water diverted at the system headworks divided by BA hereafter termed depth we see that the period dummy is negative and significant at the 95 percent confidence level for both wet and dry seasons (see Table 3 equations 1 and 3) Since the overall explanatory power of the wet season model is very weak however we will focus on the dry season in interpreting this result which indicates that after adjusting for rainfall differences significantly less water was delivered to the command per unit of benefitted area following 1981 than before This indicates based on the criteria outlined above that performance in terms of water adequacy deteriorated following financial selfshysufficiency We need to examine this conclusion more carefullyhowever
One difficulty is that the measured quantity of water diverted at the source is largely a function of the supply available in the river rather than of system management This is particularly true during the dry season and in non-reservoir systems Thus while the depth of water supplied to the system is a measure of the adequacy of the systems service it is to some extent beyond the control of the managing agency To better understand the factors behind this decline in water availability we look at simple unadjusted index values for
- 12 shy
several of the key variables Table 4 shows annual values of total volume of water delivered in each season SA and BA in both wet and dry seasons and shows the results of t-tests on the means of a set of indicators before and after 1981 Indicators are used rather than the actual values to weight each of the systems equally regardless of its size The table shows that both the average wet season benefitted area and the average discharge are significantlylower during the second period compared with the first For the dry season too the discharge index is lower after 1981 than before but this difference is not significant At the same time the dry season BA index rose slightly but again the change was not significantSince middotthere is not a clear pattern of relative movement of discharge and BA during the respective seasons no simple interpretation of these index value changes is possible What stands out is that both discharge and benefitted area declined across periods during the wet season while during the dry season there was no significant change in either indicator across the two periods It seems clear that the decline in water adequacy must be evaluated together with other measures of performance in drawing conclusions about the overall impact of the 1981 changes on the quality of system management
Another measured variable per hectare yield can be used as a proxy for water adequacy It has the advantage of partiallyreflecting the impacts of the dimensions of timeliness6 and equity7of distribution as well integrating all three effects into a combined impact on aggregate crop production Table 5 (equations 1 and 3)shows that the period dummy in the yield regressions has a positive sign in both seasons after controlling for nitrogen application and precipitation though the t-values are not significant Treatingyield adjusted in this way as a proxy for quality of irrigationservice leads to the conclusion that by this more comprehensive measure quality of service held constant across the two periods in the dry season Because of the large yield component accounted for byrainfall during the wet season no such judgement is possible for that season however
Equity As noted earlier no reliable data are available for subdivisions of the five sample systems making direct computation of equity measures impossible We can make some judgements about changesin the equity of water distribution however by examining changes in the ratio of two area measures given for each system SA and BA Since SA is the area which theoretically can be supplied with irrigation water by the system and BA s the area which actuallyreceives a quantity of water adequate to produce a remunerative crop the ratio of the two provides a measure of the percentage of the potential service area which was irrigated to a particular standard The larger this pErcentage the more equitable 8 is the distribution This of course assumes that the quantity of water available to the systems is constant across the two periods
- 13 shy
Since the condition of constancy of water supply is not generally satisfied a regression was run in which the total quantityof water diverted at the headworks of each system divided by the systems potential service area SA was included in the regression to control for changes in the water supply available to the system seeTable 3 equations 5 through 8 The average daily rainfall received directly on the system service area during the season was also included as an independent variable The regression was run separately for wet and dry seasons The sign and t-statistic of the period dummy should then tell us whether or not equity as reflected in the BASA ratio increased decreased or remained unchanged across the period divide
Both equations are reasonable good as indicated by the R2 values though the dry season equation is considerably better as would be expected For the wet season both the water delivery term and the rainfall term in equation 5 are of positive sign but are nonshysignificant at the 95 percent confidence level indicating that wet season irrigated area does not change appreciably in response to level of wet season rainfall or the available irrigation water supply The period dummy was negative but not significant indicating that equityof distribution as reflected in the BASA ratio was similar during the two periods
For the dry season the water delivery term in equation 7 is positive and strongly significant indicating a close relationship between the fraction of potential area actually irrigated and the water supply available at the headworks In addition however the period dummy is positive and significant suggesting that once the influence of water supply is removed the BASA ratio was significantly higher in the period following 1981 than it was before
This is an important finding for it reflects significantlyimproved performance in terms of a factor equity of water distribution that is under the control of the managing entity an entity which here comprises both NIA and irrigators associations Interpreted in these terms NIA and allied farmers associations were able to spread a given amount of water more widely across the potential command area of the five sample systems in the period after 1981 than before Moreover they did this in a way that did not decrease average system yields as discussed earlier In making this interpretation we are suggesting that there was some redistribution of water from better-watered areas to fringe areas which would otherwise not have received irrigation water and that this redistribution was a direct response to the change in NIA prioritiesand operating policies and rules occurring around 1981
It is difficult to prove the assertion that water was in fact redistributed with a resulting increase in directly-measured equity Without access to reliable discharge data broken out by systemsection and we can only assume in the absence of a plausible
- 14 shy
alternative explanation that it was such a redistribution that made the increase in the BASA ratio possible In a larger sense it is difficult to prove conclusively that any outcome in a before and after analysis was the result of a particular independent causative factor In this case we have tried to remove the influence of other potential causative factors where we could but the possibilityremains that some combination of unmeasured factors are responsiblefor the difference in the BASA ratio found We do note though that this type of response is exactly the type that would be expected to follow from an emphasis on increased farmer satisfaction and cooperation and increased fee revenues Because the fee schedule is tied to benefitted area the only ways NIA can increase its revenue from that source are to expand benefitted area and to increase collection efficiencies The former depends on redistributing a fixed supply of water over a larger portion of the command while the latter requires that farmers be satisfied with the irrigation service they are receiving and the commitment of the local irrigators association to assist in the task of collecting the amounts due The evidence while not conclusive is highly suggestive that this is exactly what has happened
Efficiency In addition to measures which reflect the levels of adequacy and equity of irrigation service available data allow the calculation of a measure of operating efficiency The term efficiencyusually denotes the relationship between inputs to a process and its outputs often expressed as a ratio The output measure employed here is aggregate system rice output and the input is quantity of irrigation water turned into the system Dividing the first by the second gives a measure of agricultural production per unit water--here termed specific yield This is a highly integrated measure that evaluates the combined efficiency of the irrigation and agricultural processes As such it is a function of the managerial and other inputs supplied both to the irrigation system and to the agricultural operation With respect to one important input to the irrigation system we do know that NIA per hectare field operating expenses were about 29 percent lower in real terms in the 1982-86 period comparedto the 1976-1981 period although this drop may have been partlyoffset by increases in farmer-supplied labor inputs Other things being equal one would thus expect to find a decline in output efficiency
The regression analysis shows positive signs for the period terms in both wet and dry season equations (see Table 5) In the case of the wet season the period dummy in equation 5 is significant but the overall explanatory power of the model is quite low For the dry season (equation 7) the coefficient is positive but non-significant This means that after taking rainfall and fertilizer use into account data do not indicate a lowering of specific yield in the wake of funding reductions and the strong emphasiS on financial viabilitybeginning in 1981 This result provides evidence that the efficiency of the overall irrigation deliveryagricultural production process
- 15 shy
relative to the system water input did not falloff as a result of the changes implemented at least over the short run
Impact magnitude
The preceding analysis has shown us that some indicators of irrigation performance changed significantly following the managerial changes of 1981 while others did not However it has not given us a sense of the size of the changes which occurred To determine the magnitude of these changes the regression model is used to predict the response of the composite system to the managerial changes given a common set of -input and environmental factors To do this averagevalues of the independent variables from the entire eleven-year period 1976 to 1986 are put into the model together with the previously determined coefficients to generate predicted average values of the various dependent variables used in the earlier analysis with and without the period dummy This procedure produces a pair of estimates for each dependent variable under the same conditions--one in which the system responds as it did after the managerial changes were implemented and one in which it responds as it did prior to their introduction The differences between these two values thus indicate the magnitude of the changes occurring in the various indicators of performance discussed above
The results of this exercise are shown in Table 6 The table shows that water availability decreased by about 13 percent in both wet and dry seasons when the period dummy was included and while the coefficients responsible were significant in the earlier analysisthis difference cannot be easily connected with levels of system management as discussed earlier With respect to rice output per hectare although the coefficients were not very significant it is interesting to note that yield increases by 163 kilograms per hectare for the wet season and by 101 kilogram for the dry when the period dummy is included in spite of the reduced water supply available Keep in mind that the predicted yield values have already been adjusted for differences in nitrogen fertilizer use and rainfall This suggests that timeliness and equity of distribution of water supply to farmers may have increased following the changes contributing to the higher predicted yields
Examining the impact of increased equity of distribution bylooking at the ratio of benefitted area to service area we recall that the change was positive and significant for the dry season and negative and not significant for the wet Table 6 shows that the dry season BASA ratio increases by 7 percentage points when the dummy is included a 131 percent increase Other things being equal this should result in a 131 percent increase in system output due to the expansion of ared benefitted This is a major impact on production
- 16 shy
CONCLUSIONS
The Philippine experiment to transform the national irrigation agency into an enterprise has undoubtedly been successful in reducing system operating expenses bringing revenues and costs into line and eliminating the recurrent cost burden imposed by large-scaleirrigation systems on the national budget Evidence presented in this paper indicates that in the process equity of water distribution across systems has also improved In the 5 years following the cessation of operating subsidies from the government an index of equity of distribution improved by about 13 percent At the same time per hectare yields adjusted for rainfall and nitrogen application held constant
There is a strong logical connection between the achievement of financial viability and improved equity of water distribution across the command Because increasing irrigation fees is a politicaldecisionlying largely beyond NIAs control expanding the area which can be billed for service is one of the few revenue increasing measures available to the irrigation agency which does not involve major additional investment In the face of constant or shrinkingwater supplies this is achieved only by redistributing water from areas receiving excessive supplies usually near the head ends of canals and laterals to areas receiving no supplies or inadequatesupplies often located near the tails of canals Although data are not available which would allow the direct examination of this hypotheses the two outcomes are logically consistent with each other
Data also show that per hectare water deliveries declined significantly in the five sample systems after 1981 even thoughrainfall did not differ appreciably between the two periods This decline averaged about 13 percent for both wet and dry seasons and is interpreted as a decline in water availability in the supplying rivers rather than a conscious reduction in withdrawals by system managers Such declines could result from changes in watershed runoff characteristics as caused by deforestation or from increased upstream abstractions from supplying rivers
Improved water distribution tends to increase the area served system agricultural output and NIA service fee revenue Reduced water supplies to the system tend to reduce these things Specificyield defined as system paddy output per unit water held roughly constant across the two periods indicating that the two effects mayhave offset each other
After adjusting for rainfall and nitrogen application perhectare yields increased only marginally in the post-1981 period Area served on the other hand increased by about 13 percent after adjusting for water supply availability indicating that the area benefitted by irrigation in the sample systems increased by about the same percentage Even if yields on this additional area are less than
- 17 shy
average yields for the system this still represents a sizeable increase in system agricultural output as a result of the change in management structure the increase coming not from higher yields but from expanded area under irrigation
The evidence assembled here suggests that there are significantfinancial and economic benefits to be had from changes in the basic character of irrigation managing agencies which make them more responsive to their clientele and which impose rational internal financial discipline on the agency The analysis suggests a number of additional questions however One relates to the longer-term impacts of the structural management changes The improvements in water distribution described here are relatively short-term events occurring during the first 5 years of the new management mode Critics have suggested the danger of underinvestment in systemmaintenance over the longer run accompanied by declining yields and benefitted areas and eventual system collapse This possibility needs to be closely monitored A second concern relates to the apparent decline in water supply to these 5 geographically dispersed systems The nature and causes of this decline need to be explored further since if widespread and secular it may represent a serious threat to the stability of Philippine rice production Whether stemming from poor forest management practices or deficient regulation and allocation of surface water resources or other unidentified factors it is an issue that deserves serious and urgent consideration
A third risk is that the incentive structure set up by NIA to guide and stimulate the performance of field units overemphasizes revenue generation at the expense of irrigation service provision to farmers The evidence presented here supports the view that these two objectives are mutually reinforcing under policies and conditions which have been established in the Philippines More detailed crossshysectional studies based on primary flow measurement data would add confidence to this conclusion and help to specify the conditions under which this effect occurs This could be extremely important in transferring the results of the Philippine experiment to other countries
A final risk is that outside intervention well meaning or otherwise will destroy the basis of NIAs financial autonomy or will impose external pressures or constraints on NIAs decision-making that will subvert the management practices which have been so painstakinglydeveloped and implemented Among these are calls for NIA to be subsumed again within the government department structure in the interests of better coordination with agriculture attempts byexternal financing agencies to arbitrarily increase NIAs expenditures on OampM on the assumption that this will increase system agricultural output or intervention by Philippine legislative bodies to restore operating subsidies to NIA with attached strings leading back to legislators home districts Pressures such as these will cut short a process of experimentation and improvement that seems promising
- 18 shy
enough to date to warrant its continuation Having developed the capacity to establish targets and implement and manage change NIA is in a strong position to modify its objectives to better achieve larger social purposes established for it It is critical to recognize however that this must happen within the context of financing policies that mandate financial autonomy for NIA if the fundamental institutional commitment to manage is to be preserved
The author would like to thank Leslie Small and JeremyBerkoff for helpful comments on an earlier unpublishedversion of this paper and Charles Rogers for his careful and creative help with the analysis
- 19 shy
BIBLIOGRAPHY
Abernethy Charles L 1990 Indicators of the performance of irrigation water distribution systems International Irrigation Management Institute Colombo Sri Lanka Mimeo
Asian Development Bank 1986 Irrigation service fees Proceedingsof the Regional Seminar on Irrigation Service Fees Manila Asian Development Bank
Carruthers Ian and Colin Clark 1981 Economics of IrrigationLiverpool Liverpool University Press Third Edition
Levine G and EW Coward Jr 1986 Irrigation water distribution implications for design and operation AGREP Division WorkingPaper 125 vol 1 World Bank Agriculture and Rural Development Department
Small Les E 1989 User charges in irrigation potentials and limitations Irrigation and drainage vol 3 no 2125-142
Small Les 1990 Irrigation service fees in Asia IrrigationManagement Network 9013 London Overseas DevelopmentInstitute
Svendsen Mark and Les Small 1989 A framework for assessing irrigation system performance Paper prepared for the Symposium on Performance Evaluation 23 November International IrrigationManagement Institute Sri Lanka
Table I--National Irrigation Administration revenues and expenditures in constant prices 1976-86
Item 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986
(peso million 1972)
Revenues Irrigation fees collected 1273 1483 17 13 1831 2070 1668 1699 1893 1728 2129 2546 Other income 715 737 2420 5591 2631 5990 7783 6638 5699 4932 2934
Total direct revenue 1988 2220 4133 7422 4701 7658 9482 8531 7427 061 5480
Expenses in 1972 pricesTotal expenses 4825 5716 5039 6329 3821 77 55 6166 4749 4348 4259 4959
Excess (deficit) (2837(3496) (906) 1093 877 (097) 3316 3782 3079 2802 521 N 0
Subsidies Government operation and
maintenance subsidies 2521 2741 2799 1817 1398 633 0 0 0 0 0 Calamity fund payments 548 0 0 0 0 0 0 0 119 0 142
Total subsidy 3069 2741 2799 1817 1398 633 0 0 119 0 142
Total excess (deficit) 231 (754) 1893 2910 2275 536 3316 3782 3198 2802 663
Source IFPRI analysis of NIA data
- 21 shy
Table 2--Descriptive characteristics of selected MIA systeasa
------- Region III -------- ----- Region VI ----shyUPRIIS Angatii Sto Sibalom- Aganan-
Haasim Thomasc San Jose Sta Barbara
Average service area (hal 102272 31462 3522 5282 8703 Average irrigated area (hal
Wet season 83768 23454 3007 4410 8300 Dry season
Average benefited area Wet season
(hal 64587
77 605
27639
22908
1 781
3007
2801
4369
2770
7698 Dry season
Average rainfall Wet season
(mml d 62478
1 685 5
27396
8576
1 781
3051 0
2769
24731
2997
20001 Dry season 756 333 322 2828 3025
Average discharge (Llsec) Wet season 46501 14792 1692 2353 4984 Dry season 78091 22812 2014 1276 2315
Average water delivery (mmday) Wet season 522 548 487 462 571 Dry season
Average yield (mtha) 1089 715 995 398 686
Wet season 345 419 322 395 435 Dry season
Avg yield per unit water 34 03
(kgm ) 451 412 399 426
Wet season 0373 0440 0373 0538 0443 Dry season 0248 0400 0279 0690 0428
t-statistic difference in mean rainfall 1978-81 1982-86e
Wet season 0432 0713 -0567 1169 1169 Dry season 0519 -0230 -0523 1187 1187 Annual 0460 0707 -0686 1445 1445
~ Summary numbers are averages for the period 1982-1986 except as noted Water delivery discharge and yield per unit discharge are 4-year averages 1982-1985
c Water delivery discharge and yield per unit discharge are 4-year averages d 1983-1986
For Angat 5 years are 1981-85 For St Thomas 1979-83 For Sibalom 1971-75 e No significant differences at 95 confidence
- 22 shy
Table 3--Results of pooled regressions water delivery indicators
Dependent variable Independent variable - Water Delivery in mmBenefited Ha - --- BenefitService Area Ratio --shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season-shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 6799 (998)
6189 (1019)
13102 (1093)
2411 (1 54)
0743 ( 1186)
0800 (1165)
0129 (131 )
0258 (329)
System 2 indicator 0173 (029)
-3743 (-363)
-0063 (-219)
0200 (505)
System 3 indicator 0698 (085)
0163 (012)
0085 (208)
-0046 (-097)
System 4 indicator -0514 (-l 07)
-7011 (-803)
0080 (340)
0169 (l87)
System 5 indicator 0177 (033)
-5049 (-538)
0174 (603)
0051 (058)
Reservoir indicator 0327 (078)
5703 (543)
-0126 (-510)
0156 (261)
Period indicatora -0808 (-240)
-0771 (-229)
-1214 (-209)
-1309 (-1 62)
-0036 (-192)
-0025 (-112)
0071 (239)
0030 (077)
Water delivery mmb
Wet season
Dry season
Precipitation in mm Wet season
Annual
Average daily rainfall Wet season
Dry season
-46E-4 (-125)
-1 9E-4 (-082)
-43E-4 (-077)
20E-3 (385)
0015 (1 59)
-0001 ( -004)
0023 (215)
0001 (021)
0070 (613)
-0017 (-058)
0046 (371)
0010 (033)
R2 (adj) number of observations degrees of freedom
0156 43 36
0133 43 39
0711 41 34
0434 41 37
0693 42 34
0556 42 37
0865 40 32
0730 40 35
Coefficient (t statistic) indicates significance at 95 confidence ~ 1982-1986 1
per ha of service area
Ta
ble
--I
nd
ices
of
se
v
ice a
rea
ib
en
ef
ited
a
rea
a
nd
a
vera
ge sea
so
na
l d
isch
arg
e
Ave
rage
A
vera
ge
tshy19
77
1978
19
79
1980
19
81
1982
19
83
1984
19
85
1986
19
77-8
1 19
82-8
6 S
tati
stic
a
(ind
ex
aver
age
1983
-198
5 =
100)
Ser
vice
are
a UP
RI IS
in
dex
882
91
9
920
91
5
951
95
1
100
0 10
00
100
0 10
00
917
99
0
bull5
53
Ang
at-M
aasi
m R
95
9
994
99
7
996
99
6
996
10
00
100
0 10
00
100
0 98
8
999
1
63
Sto
To
mas
10
63
106
3 10
1 9
10
29
103
0 99
8
100
0 10
00
100
0 11
10
104
1 10
22
-08
9
Siba
lom
-San
Jos
e 95
0
872
94
2
933
94
2
942
10
28
102
8 94
4
101
7
928
99
2
292
A
gana
n-St
a B
arba
ra
108
3 10
91
106
0 10
30
961
10
05
996
10
08
996
99
6
104
5 10
00
-21
1 A
vera
ge
987
98
8
981
98
1
916
97
8
100
5 10
07
988
10
25
984
10
01
123
Wet
seas
on b
enef
ited
are
a in
dex
UPR
IIS
110
0 10
07
114
4 10
55
113
8 11
74
951
10
71
971
ll
58
10
89
106
7 -0
47
Ang
at-M
aasi
m R
97
9
974
92
0
983
10
28
100
8 99
7
102
1 98
2
931
97
7
988
0
54
Sto
To
mas
11
63
115
9 11
23
107
5 10
80
103
9 98
1
978
10
41
103
9 11
20
101
6 -4
88
Siba
lom
-San
Jos
e 11
35
103
8 10
11
931
93
4
906
10
07
975
10
1S
998
10
11
981
-0
80
Aga
nan-
Sta
Bar
bara
10
85
110
5 10
74
106
8 10
01
104
2 99
8
101
5
987
61
0
106
7 93
0
-1 8
4
Ave
rage
10
92
105
6 10
54
102
4 10
36
103
4 98
7
101
3 10
00
947
10
53
996
-2
32
Dry
sea
Son
bene
fite
d ar
ea
inde
x U
PRIIS
14
04
155
0 15
S0
155
8 16
17
128
0 57
2
114
S 15
74
152
3 12
38
-09
4 A
ngat
-Maa
sim
R
903
93
0
103
2 10
61
104
2 10
69
988
99
2
102
1 99
6
993
10
13
061
S
to
Tom
as
105
7 12
27
122
5
961
99
9
115
9 10
1 7
91
0
107
3 12
1 2
10
94
107
4 -0
28
Si
balo
m-S
an J
ose
Aga
nan-
Sta
Bar
bara
66
5
95S
62
6
632
67
4
111
4
501
10
S1
412
11
51
766
93
3
856
94
3
107
0 94
8
107
4 11
09
111
6
158
6 58
8
987
97
6
110
4 5
35
083
N
w
Ave
rage
89
6
964
11
19
103
7 10
44
110
9 10
1 7
89
8
108
5 12
97
101
7 10
81
082
Wet
seas
on d
isch
arge
in
dex
UPR
IIS
132
9 72
7
142
5 11
88
120
4 98
8
105
8 10
63
879
96
4
117
5 99
1
-16
5 A
ngat
-Maa
sim
R
129
7 13
52
134
5 12
58
127
0 11
70
120
3 62
7
131
3 10
68
-18
9
Sto
To
mas
14
71
155
0 14
1 6
10
40
725
12
35
112
3 14
79
103
1 -4
48
Siba
lom
-San
Jos
e 96
8
733
11
1 5
92
2
907
47
1
102
3 85
7
ll2
0
422
92
9
779
-1
09
Aga
nan-
Sta
Bar
bara
87
9
863
68
1
925
96
8
110
7 70
5
871
87
7
00
9
Ave
rage
11
49
105
7 13
61
115
0 10
58
853
10
43
963
99
4
SO3
11
55
940
-2
85
Dry
sea
son
disc
harg
e in
dex
UPR
IIS
425
13
09
153
3 14
28
180
6 14
S0
125
8 56
3
117
9 14
66
130
0 11
89
-04
3
Ang
at-M
aasi
m R
12
26
161
6 15
30
139
2 12
85
100
8 10
33
958
14
41
107
1 -3
80
S
to
Tom
as
116
4 14
82
155
7 79
9
971
12
30
126
5 14
01
106
6 -2
44
Si
balo
m-S
an J
ose
486
71
5
782
62
9
789
64
5
116
9 11
86
801
65
3
918
2
36
Aga
nan-
Sta
Bar
bara
A
vera
ge
425
10
46
133
6 62
0
118
4 81
5
116
1 94
7
112
5 89
7
922
12
20
991
88
2
108
7 12
36
119
2 71
8
114
0 10
37
105
5 3
23
-07
5
a bull
indi
cate
s si
gnif
ican
ce a
t 95
per
cent
con
fide
nce
- 24 shy
Table 5--Results of pooled regressions yield indicators
------------------------- Dependent variables -------------------------shyIndependent variable ---------- Yield MTHa ----------- -- Specific Yield 1000 KgM
3 H20 shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season -shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 1967 (286)
3747 (580)
1991 (293)
2800 (460 )
0084 (041)
0441 (229)
0049 (023 )
0876 (343)
System 2 indicator 0400 (155)
0525 (226)
0057 (085)
0115 (1 95)
System 3 indicator 0466 (139 )
-0029 (-010)
-0016 (-018)
0006 (008)
System 4 indicator 0980 (470)
0090 (049)
0176 (333)
0391 (827)
System 5 indicator 1 221 (570)
0256 ( 138)
0104 (174)
0224 (440)
Reservoir indicator -1033 (-578)
-0063 (-039)
-0146 (-310)
-0326 (-569)
Period indicatora
Precipitation in mm Wet season
Annual
0163 (1 03)
-29E-4 (-182)
0231 (1 42)
-58E-4 (-606)
0161 (118)
92E-5 (072)
0199 043 )
-64pound-5 (-084)
0082 (200)
57E-6 (013)
0092 (223)
-61E-5 (-232)
0032 (093)
29E-7 (001)
0052 (112)
-12pound-4 (-414)
Nitrogen applicationb 0025 (254)
0021 (211)
0025 (274)
0020 (227)
0003 (1 05)
0002 (043)
0003 (101)
-0002 (-052)
R2 (adj) 0553 0516 0251 0198 0309 0262 0713 0445 number of observations 50 50 49 49 42 42 40 40 degrees of freedom 42 45 41 44 34 37 32 35
CoeffiCient (t statistic) indicates significance at 95 confidence ~ 1982-1986 = 1
kgha
- 25 shy
Table 6--Suggmry of estiates perforance indicators
Variable Predicted Predicted Period Dependent Variable
Description (units)
Period Coefficient
1976-86 IndicatorO
1976-86 Indicatorl
Effecta (X)
BenefittedService area ratio
Wet season ratio -0036 0853 0817 -425 Dry Season 0071 0539 0610 1313
Delivery Wet Dry
season season
nm water per benefitted ha
-0808 -1 214
6003 9083
5195 7869
-1346 -1337
Rice Yield Wet season metric tons 0163 3578 3741 456 Dry season per hectare 0161 3905 4066 412
Specific Yield Wet season kg rice per m3 0082 0348 0430 2353 Dry season water de 1 ivered 0032 0378 0409 837
a percentage changes relative to period=O values
- 26 shy
ENDNOTES
IResearch Fellow International Food Policy Research Institute and Senior Irrigation Specialist International Irrigation Management Institute
2Presumable farmers also base private investment decisions on such foreknowledge of costs and commonly do so in the case of small tubewells In addition user groups may invest in small communal schemes independently after examining projected costs and benefits Most commonly however new irrigation capacity is created as a result of public investment decisions made by government and thus the political process comes into play
3This principle applies only when water charges are levied volumetrically--an exceedingly rare situation in the developing worldshy-though the argument is frequently invoked as though it applies to all allocative situations
4For Cabanatuan City the longer period of record available ie 1904 to 1986 was divided into two equal periods and tested for difference in means to test the observation by farmers that rainfall today is lower than it used to be Mean precipitation after 1947 was actually about 3 percent higher than before but the difference was not significant (t = 07)
5The procedure used was to take time series data on total nitrogenfertilizer consumption in the Philippines multiply it by the ratio of N fertilizer used on rice to total N consumption and divide the product by the rice harvested area to obtain an estimate of N use perhectare Values of the ratio of N use on rice to total N use were taken from IRRI (1988 150) with interpolation used to fill in gaps in the series
6This assumes that timeliness is evaluated against a standard based on the physiological demand for water generated by the rice crop and the varying sensitivity of the crop to stress at different growth stages In such a case a poor timeliness rating for irrigationservice will result in reductions in yield
7Because rice yields are insensitive to water applications that exceed PET values a redistribution of water from water excess to water short portions of the command area will usually result in higher aggregate output and higher average yields Exceptional scenarios can be constructed in which this general rule would be violated but such scenarios are unlikely to play out in practice
- 27 shy
aThere are a variety of standards for judging equity that could be used here The most common one is equality of per hectare water deliveries In the present case the implicit standard is equality of water supply utility in producing a crop Thus if two areas received equal amounts of water on a per hectare basis but one had such high seepage and percolation losses that the crop failed to produce a harvest (and was therefore exempted from irrigation fee payment) the water supply to the combined area would be judged inequitable Because such radical differences in seepage and percolation rates are unlikely on puddled rice soils in the lowland Philippines and because the definition of benefitted area accommodates a wide range of acceptable yields these two definitions are largely indistinguishable at the current limits of data resolution
- 10 shy
Three fundamental indicators have been proposed for assessing the effectiveness of irrigation services to cultivators (Svendsen and Small 1989 Abernethy 1990) These are the adequacy of water supplies the equity of their distribution across the command area of the system and the timeliness of the supplies Computation of adequacy measures requires information on the total quantity of water delivered to the system over a season on a per hectare basis Equityand timeliness measures require information on the spatial and temporal distribution respectively of those supplies Where appropriate discharge information is not available proxies can be employed by making suitable assumptions Standards must be selected against which the magnitude of the indicators can be judged
The task in the present case however is somewhat different Here the need is to evaluate changes in selected variables between the pre-1981 and the post-1981 periods Hence the absolute values of variables selected are less important than their relative magnitudesand the statistical significance of the differences in magnitudes between the two periods A distinct limitation is imposed by the data series available for the five sample systems Since discharge and yield data are available only on a whole-system basis it is impossible to develop measures of equity and timeliness directly We will however extend our analysis to a discussion of equity byindirect inference Levine and Coward (1986) have argued that equityought to be considered as the paramount objective in managing largepublic irrigation systems They base their conclusion on an analysisof eight small community-managed systems and five larger public systems including UPRIIS in which equity appears to comprise the most important operational objective in the successful systems It may be appropriate therefore to give success in improving equity of distribution added weight in assessing overall performance
Area Estimates Because measures of system agricultural output and water supplied are typically reduced to a unit area basis before being used much depends on the area values which are used to standardize them Two different area measures are available The first is Service Area (SA) which is defined as the irrigable portion of the command area which is provided with physical facilities for water delivery This represents the area which could conceivably be irrigated in a given season if water supply were not constraining This value may change somewhat from year to year in response to urban encroachment on irrigated command minor remodeling and repair and refinements in area estimates In the present case though the systems selected for analysis were chosen to avoid those which had undergone more extensive rehabilitation or modification
The second measure is Benefitted Area (BA) which is the area billed for payment of irrigation service fees It is the irrigated area harvested which did not have yields so low that the farm was exempted from payment of fees in a given season This threshold value has been approximately 2 tons of paddy per hectare Benefitted area
- 11 shy
varies more than does SA particularly during the dry season when available water supply may seriously constrain the area which can be planted Its magnitude is a function of system managers actions in authorizing the amount of land to be planted in a given season farmers decisions regarding whether to plant or not and the combined ability of system managers and farmersirrigators subsequently to distribute water Both of these area measures will be used to standardize other variables for particular purposes as well as beingcombined to form a separate indicator by themselves
Adequacy The most direct measure of the adequacy of irrigation water supplies to the agricultural system is the quantity of water applied to the system command area on a per unit area basis relative to some standard In this case since our interest is in differences in water adequacy between two time periods and since the systems being assessed have been and continue to be almost entirely devoted to rice cultivation during both cropping seasons depth measures for the two periods may be compared directly assuming the seasonal cropdemand for water to be unchanged Although dry-footed crops can suffer yield losses from overapplication of irrigation water rice is largely insensitive to this effect In addition water can substitute for other inputs that the farmer would otherwise have to provide such as weed control and more careful (and costly) water management We assume therefore that other things being equal larger values of depth applied are better than smaller values in terms of meeting crop water demands and reduce the cost of cultivation At the same time high levels of water adequacy can affect the values of other performance measures--particularly equity
When the regression model is run for quantity of water diverted at the system headworks divided by BA hereafter termed depth we see that the period dummy is negative and significant at the 95 percent confidence level for both wet and dry seasons (see Table 3 equations 1 and 3) Since the overall explanatory power of the wet season model is very weak however we will focus on the dry season in interpreting this result which indicates that after adjusting for rainfall differences significantly less water was delivered to the command per unit of benefitted area following 1981 than before This indicates based on the criteria outlined above that performance in terms of water adequacy deteriorated following financial selfshysufficiency We need to examine this conclusion more carefullyhowever
One difficulty is that the measured quantity of water diverted at the source is largely a function of the supply available in the river rather than of system management This is particularly true during the dry season and in non-reservoir systems Thus while the depth of water supplied to the system is a measure of the adequacy of the systems service it is to some extent beyond the control of the managing agency To better understand the factors behind this decline in water availability we look at simple unadjusted index values for
- 12 shy
several of the key variables Table 4 shows annual values of total volume of water delivered in each season SA and BA in both wet and dry seasons and shows the results of t-tests on the means of a set of indicators before and after 1981 Indicators are used rather than the actual values to weight each of the systems equally regardless of its size The table shows that both the average wet season benefitted area and the average discharge are significantlylower during the second period compared with the first For the dry season too the discharge index is lower after 1981 than before but this difference is not significant At the same time the dry season BA index rose slightly but again the change was not significantSince middotthere is not a clear pattern of relative movement of discharge and BA during the respective seasons no simple interpretation of these index value changes is possible What stands out is that both discharge and benefitted area declined across periods during the wet season while during the dry season there was no significant change in either indicator across the two periods It seems clear that the decline in water adequacy must be evaluated together with other measures of performance in drawing conclusions about the overall impact of the 1981 changes on the quality of system management
Another measured variable per hectare yield can be used as a proxy for water adequacy It has the advantage of partiallyreflecting the impacts of the dimensions of timeliness6 and equity7of distribution as well integrating all three effects into a combined impact on aggregate crop production Table 5 (equations 1 and 3)shows that the period dummy in the yield regressions has a positive sign in both seasons after controlling for nitrogen application and precipitation though the t-values are not significant Treatingyield adjusted in this way as a proxy for quality of irrigationservice leads to the conclusion that by this more comprehensive measure quality of service held constant across the two periods in the dry season Because of the large yield component accounted for byrainfall during the wet season no such judgement is possible for that season however
Equity As noted earlier no reliable data are available for subdivisions of the five sample systems making direct computation of equity measures impossible We can make some judgements about changesin the equity of water distribution however by examining changes in the ratio of two area measures given for each system SA and BA Since SA is the area which theoretically can be supplied with irrigation water by the system and BA s the area which actuallyreceives a quantity of water adequate to produce a remunerative crop the ratio of the two provides a measure of the percentage of the potential service area which was irrigated to a particular standard The larger this pErcentage the more equitable 8 is the distribution This of course assumes that the quantity of water available to the systems is constant across the two periods
- 13 shy
Since the condition of constancy of water supply is not generally satisfied a regression was run in which the total quantityof water diverted at the headworks of each system divided by the systems potential service area SA was included in the regression to control for changes in the water supply available to the system seeTable 3 equations 5 through 8 The average daily rainfall received directly on the system service area during the season was also included as an independent variable The regression was run separately for wet and dry seasons The sign and t-statistic of the period dummy should then tell us whether or not equity as reflected in the BASA ratio increased decreased or remained unchanged across the period divide
Both equations are reasonable good as indicated by the R2 values though the dry season equation is considerably better as would be expected For the wet season both the water delivery term and the rainfall term in equation 5 are of positive sign but are nonshysignificant at the 95 percent confidence level indicating that wet season irrigated area does not change appreciably in response to level of wet season rainfall or the available irrigation water supply The period dummy was negative but not significant indicating that equityof distribution as reflected in the BASA ratio was similar during the two periods
For the dry season the water delivery term in equation 7 is positive and strongly significant indicating a close relationship between the fraction of potential area actually irrigated and the water supply available at the headworks In addition however the period dummy is positive and significant suggesting that once the influence of water supply is removed the BASA ratio was significantly higher in the period following 1981 than it was before
This is an important finding for it reflects significantlyimproved performance in terms of a factor equity of water distribution that is under the control of the managing entity an entity which here comprises both NIA and irrigators associations Interpreted in these terms NIA and allied farmers associations were able to spread a given amount of water more widely across the potential command area of the five sample systems in the period after 1981 than before Moreover they did this in a way that did not decrease average system yields as discussed earlier In making this interpretation we are suggesting that there was some redistribution of water from better-watered areas to fringe areas which would otherwise not have received irrigation water and that this redistribution was a direct response to the change in NIA prioritiesand operating policies and rules occurring around 1981
It is difficult to prove the assertion that water was in fact redistributed with a resulting increase in directly-measured equity Without access to reliable discharge data broken out by systemsection and we can only assume in the absence of a plausible
- 14 shy
alternative explanation that it was such a redistribution that made the increase in the BASA ratio possible In a larger sense it is difficult to prove conclusively that any outcome in a before and after analysis was the result of a particular independent causative factor In this case we have tried to remove the influence of other potential causative factors where we could but the possibilityremains that some combination of unmeasured factors are responsiblefor the difference in the BASA ratio found We do note though that this type of response is exactly the type that would be expected to follow from an emphasis on increased farmer satisfaction and cooperation and increased fee revenues Because the fee schedule is tied to benefitted area the only ways NIA can increase its revenue from that source are to expand benefitted area and to increase collection efficiencies The former depends on redistributing a fixed supply of water over a larger portion of the command while the latter requires that farmers be satisfied with the irrigation service they are receiving and the commitment of the local irrigators association to assist in the task of collecting the amounts due The evidence while not conclusive is highly suggestive that this is exactly what has happened
Efficiency In addition to measures which reflect the levels of adequacy and equity of irrigation service available data allow the calculation of a measure of operating efficiency The term efficiencyusually denotes the relationship between inputs to a process and its outputs often expressed as a ratio The output measure employed here is aggregate system rice output and the input is quantity of irrigation water turned into the system Dividing the first by the second gives a measure of agricultural production per unit water--here termed specific yield This is a highly integrated measure that evaluates the combined efficiency of the irrigation and agricultural processes As such it is a function of the managerial and other inputs supplied both to the irrigation system and to the agricultural operation With respect to one important input to the irrigation system we do know that NIA per hectare field operating expenses were about 29 percent lower in real terms in the 1982-86 period comparedto the 1976-1981 period although this drop may have been partlyoffset by increases in farmer-supplied labor inputs Other things being equal one would thus expect to find a decline in output efficiency
The regression analysis shows positive signs for the period terms in both wet and dry season equations (see Table 5) In the case of the wet season the period dummy in equation 5 is significant but the overall explanatory power of the model is quite low For the dry season (equation 7) the coefficient is positive but non-significant This means that after taking rainfall and fertilizer use into account data do not indicate a lowering of specific yield in the wake of funding reductions and the strong emphasiS on financial viabilitybeginning in 1981 This result provides evidence that the efficiency of the overall irrigation deliveryagricultural production process
- 15 shy
relative to the system water input did not falloff as a result of the changes implemented at least over the short run
Impact magnitude
The preceding analysis has shown us that some indicators of irrigation performance changed significantly following the managerial changes of 1981 while others did not However it has not given us a sense of the size of the changes which occurred To determine the magnitude of these changes the regression model is used to predict the response of the composite system to the managerial changes given a common set of -input and environmental factors To do this averagevalues of the independent variables from the entire eleven-year period 1976 to 1986 are put into the model together with the previously determined coefficients to generate predicted average values of the various dependent variables used in the earlier analysis with and without the period dummy This procedure produces a pair of estimates for each dependent variable under the same conditions--one in which the system responds as it did after the managerial changes were implemented and one in which it responds as it did prior to their introduction The differences between these two values thus indicate the magnitude of the changes occurring in the various indicators of performance discussed above
The results of this exercise are shown in Table 6 The table shows that water availability decreased by about 13 percent in both wet and dry seasons when the period dummy was included and while the coefficients responsible were significant in the earlier analysisthis difference cannot be easily connected with levels of system management as discussed earlier With respect to rice output per hectare although the coefficients were not very significant it is interesting to note that yield increases by 163 kilograms per hectare for the wet season and by 101 kilogram for the dry when the period dummy is included in spite of the reduced water supply available Keep in mind that the predicted yield values have already been adjusted for differences in nitrogen fertilizer use and rainfall This suggests that timeliness and equity of distribution of water supply to farmers may have increased following the changes contributing to the higher predicted yields
Examining the impact of increased equity of distribution bylooking at the ratio of benefitted area to service area we recall that the change was positive and significant for the dry season and negative and not significant for the wet Table 6 shows that the dry season BASA ratio increases by 7 percentage points when the dummy is included a 131 percent increase Other things being equal this should result in a 131 percent increase in system output due to the expansion of ared benefitted This is a major impact on production
- 16 shy
CONCLUSIONS
The Philippine experiment to transform the national irrigation agency into an enterprise has undoubtedly been successful in reducing system operating expenses bringing revenues and costs into line and eliminating the recurrent cost burden imposed by large-scaleirrigation systems on the national budget Evidence presented in this paper indicates that in the process equity of water distribution across systems has also improved In the 5 years following the cessation of operating subsidies from the government an index of equity of distribution improved by about 13 percent At the same time per hectare yields adjusted for rainfall and nitrogen application held constant
There is a strong logical connection between the achievement of financial viability and improved equity of water distribution across the command Because increasing irrigation fees is a politicaldecisionlying largely beyond NIAs control expanding the area which can be billed for service is one of the few revenue increasing measures available to the irrigation agency which does not involve major additional investment In the face of constant or shrinkingwater supplies this is achieved only by redistributing water from areas receiving excessive supplies usually near the head ends of canals and laterals to areas receiving no supplies or inadequatesupplies often located near the tails of canals Although data are not available which would allow the direct examination of this hypotheses the two outcomes are logically consistent with each other
Data also show that per hectare water deliveries declined significantly in the five sample systems after 1981 even thoughrainfall did not differ appreciably between the two periods This decline averaged about 13 percent for both wet and dry seasons and is interpreted as a decline in water availability in the supplying rivers rather than a conscious reduction in withdrawals by system managers Such declines could result from changes in watershed runoff characteristics as caused by deforestation or from increased upstream abstractions from supplying rivers
Improved water distribution tends to increase the area served system agricultural output and NIA service fee revenue Reduced water supplies to the system tend to reduce these things Specificyield defined as system paddy output per unit water held roughly constant across the two periods indicating that the two effects mayhave offset each other
After adjusting for rainfall and nitrogen application perhectare yields increased only marginally in the post-1981 period Area served on the other hand increased by about 13 percent after adjusting for water supply availability indicating that the area benefitted by irrigation in the sample systems increased by about the same percentage Even if yields on this additional area are less than
- 17 shy
average yields for the system this still represents a sizeable increase in system agricultural output as a result of the change in management structure the increase coming not from higher yields but from expanded area under irrigation
The evidence assembled here suggests that there are significantfinancial and economic benefits to be had from changes in the basic character of irrigation managing agencies which make them more responsive to their clientele and which impose rational internal financial discipline on the agency The analysis suggests a number of additional questions however One relates to the longer-term impacts of the structural management changes The improvements in water distribution described here are relatively short-term events occurring during the first 5 years of the new management mode Critics have suggested the danger of underinvestment in systemmaintenance over the longer run accompanied by declining yields and benefitted areas and eventual system collapse This possibility needs to be closely monitored A second concern relates to the apparent decline in water supply to these 5 geographically dispersed systems The nature and causes of this decline need to be explored further since if widespread and secular it may represent a serious threat to the stability of Philippine rice production Whether stemming from poor forest management practices or deficient regulation and allocation of surface water resources or other unidentified factors it is an issue that deserves serious and urgent consideration
A third risk is that the incentive structure set up by NIA to guide and stimulate the performance of field units overemphasizes revenue generation at the expense of irrigation service provision to farmers The evidence presented here supports the view that these two objectives are mutually reinforcing under policies and conditions which have been established in the Philippines More detailed crossshysectional studies based on primary flow measurement data would add confidence to this conclusion and help to specify the conditions under which this effect occurs This could be extremely important in transferring the results of the Philippine experiment to other countries
A final risk is that outside intervention well meaning or otherwise will destroy the basis of NIAs financial autonomy or will impose external pressures or constraints on NIAs decision-making that will subvert the management practices which have been so painstakinglydeveloped and implemented Among these are calls for NIA to be subsumed again within the government department structure in the interests of better coordination with agriculture attempts byexternal financing agencies to arbitrarily increase NIAs expenditures on OampM on the assumption that this will increase system agricultural output or intervention by Philippine legislative bodies to restore operating subsidies to NIA with attached strings leading back to legislators home districts Pressures such as these will cut short a process of experimentation and improvement that seems promising
- 18 shy
enough to date to warrant its continuation Having developed the capacity to establish targets and implement and manage change NIA is in a strong position to modify its objectives to better achieve larger social purposes established for it It is critical to recognize however that this must happen within the context of financing policies that mandate financial autonomy for NIA if the fundamental institutional commitment to manage is to be preserved
The author would like to thank Leslie Small and JeremyBerkoff for helpful comments on an earlier unpublishedversion of this paper and Charles Rogers for his careful and creative help with the analysis
- 19 shy
BIBLIOGRAPHY
Abernethy Charles L 1990 Indicators of the performance of irrigation water distribution systems International Irrigation Management Institute Colombo Sri Lanka Mimeo
Asian Development Bank 1986 Irrigation service fees Proceedingsof the Regional Seminar on Irrigation Service Fees Manila Asian Development Bank
Carruthers Ian and Colin Clark 1981 Economics of IrrigationLiverpool Liverpool University Press Third Edition
Levine G and EW Coward Jr 1986 Irrigation water distribution implications for design and operation AGREP Division WorkingPaper 125 vol 1 World Bank Agriculture and Rural Development Department
Small Les E 1989 User charges in irrigation potentials and limitations Irrigation and drainage vol 3 no 2125-142
Small Les 1990 Irrigation service fees in Asia IrrigationManagement Network 9013 London Overseas DevelopmentInstitute
Svendsen Mark and Les Small 1989 A framework for assessing irrigation system performance Paper prepared for the Symposium on Performance Evaluation 23 November International IrrigationManagement Institute Sri Lanka
Table I--National Irrigation Administration revenues and expenditures in constant prices 1976-86
Item 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986
(peso million 1972)
Revenues Irrigation fees collected 1273 1483 17 13 1831 2070 1668 1699 1893 1728 2129 2546 Other income 715 737 2420 5591 2631 5990 7783 6638 5699 4932 2934
Total direct revenue 1988 2220 4133 7422 4701 7658 9482 8531 7427 061 5480
Expenses in 1972 pricesTotal expenses 4825 5716 5039 6329 3821 77 55 6166 4749 4348 4259 4959
Excess (deficit) (2837(3496) (906) 1093 877 (097) 3316 3782 3079 2802 521 N 0
Subsidies Government operation and
maintenance subsidies 2521 2741 2799 1817 1398 633 0 0 0 0 0 Calamity fund payments 548 0 0 0 0 0 0 0 119 0 142
Total subsidy 3069 2741 2799 1817 1398 633 0 0 119 0 142
Total excess (deficit) 231 (754) 1893 2910 2275 536 3316 3782 3198 2802 663
Source IFPRI analysis of NIA data
- 21 shy
Table 2--Descriptive characteristics of selected MIA systeasa
------- Region III -------- ----- Region VI ----shyUPRIIS Angatii Sto Sibalom- Aganan-
Haasim Thomasc San Jose Sta Barbara
Average service area (hal 102272 31462 3522 5282 8703 Average irrigated area (hal
Wet season 83768 23454 3007 4410 8300 Dry season
Average benefited area Wet season
(hal 64587
77 605
27639
22908
1 781
3007
2801
4369
2770
7698 Dry season
Average rainfall Wet season
(mml d 62478
1 685 5
27396
8576
1 781
3051 0
2769
24731
2997
20001 Dry season 756 333 322 2828 3025
Average discharge (Llsec) Wet season 46501 14792 1692 2353 4984 Dry season 78091 22812 2014 1276 2315
Average water delivery (mmday) Wet season 522 548 487 462 571 Dry season
Average yield (mtha) 1089 715 995 398 686
Wet season 345 419 322 395 435 Dry season
Avg yield per unit water 34 03
(kgm ) 451 412 399 426
Wet season 0373 0440 0373 0538 0443 Dry season 0248 0400 0279 0690 0428
t-statistic difference in mean rainfall 1978-81 1982-86e
Wet season 0432 0713 -0567 1169 1169 Dry season 0519 -0230 -0523 1187 1187 Annual 0460 0707 -0686 1445 1445
~ Summary numbers are averages for the period 1982-1986 except as noted Water delivery discharge and yield per unit discharge are 4-year averages 1982-1985
c Water delivery discharge and yield per unit discharge are 4-year averages d 1983-1986
For Angat 5 years are 1981-85 For St Thomas 1979-83 For Sibalom 1971-75 e No significant differences at 95 confidence
- 22 shy
Table 3--Results of pooled regressions water delivery indicators
Dependent variable Independent variable - Water Delivery in mmBenefited Ha - --- BenefitService Area Ratio --shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season-shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 6799 (998)
6189 (1019)
13102 (1093)
2411 (1 54)
0743 ( 1186)
0800 (1165)
0129 (131 )
0258 (329)
System 2 indicator 0173 (029)
-3743 (-363)
-0063 (-219)
0200 (505)
System 3 indicator 0698 (085)
0163 (012)
0085 (208)
-0046 (-097)
System 4 indicator -0514 (-l 07)
-7011 (-803)
0080 (340)
0169 (l87)
System 5 indicator 0177 (033)
-5049 (-538)
0174 (603)
0051 (058)
Reservoir indicator 0327 (078)
5703 (543)
-0126 (-510)
0156 (261)
Period indicatora -0808 (-240)
-0771 (-229)
-1214 (-209)
-1309 (-1 62)
-0036 (-192)
-0025 (-112)
0071 (239)
0030 (077)
Water delivery mmb
Wet season
Dry season
Precipitation in mm Wet season
Annual
Average daily rainfall Wet season
Dry season
-46E-4 (-125)
-1 9E-4 (-082)
-43E-4 (-077)
20E-3 (385)
0015 (1 59)
-0001 ( -004)
0023 (215)
0001 (021)
0070 (613)
-0017 (-058)
0046 (371)
0010 (033)
R2 (adj) number of observations degrees of freedom
0156 43 36
0133 43 39
0711 41 34
0434 41 37
0693 42 34
0556 42 37
0865 40 32
0730 40 35
Coefficient (t statistic) indicates significance at 95 confidence ~ 1982-1986 1
per ha of service area
Ta
ble
--I
nd
ices
of
se
v
ice a
rea
ib
en
ef
ited
a
rea
a
nd
a
vera
ge sea
so
na
l d
isch
arg
e
Ave
rage
A
vera
ge
tshy19
77
1978
19
79
1980
19
81
1982
19
83
1984
19
85
1986
19
77-8
1 19
82-8
6 S
tati
stic
a
(ind
ex
aver
age
1983
-198
5 =
100)
Ser
vice
are
a UP
RI IS
in
dex
882
91
9
920
91
5
951
95
1
100
0 10
00
100
0 10
00
917
99
0
bull5
53
Ang
at-M
aasi
m R
95
9
994
99
7
996
99
6
996
10
00
100
0 10
00
100
0 98
8
999
1
63
Sto
To
mas
10
63
106
3 10
1 9
10
29
103
0 99
8
100
0 10
00
100
0 11
10
104
1 10
22
-08
9
Siba
lom
-San
Jos
e 95
0
872
94
2
933
94
2
942
10
28
102
8 94
4
101
7
928
99
2
292
A
gana
n-St
a B
arba
ra
108
3 10
91
106
0 10
30
961
10
05
996
10
08
996
99
6
104
5 10
00
-21
1 A
vera
ge
987
98
8
981
98
1
916
97
8
100
5 10
07
988
10
25
984
10
01
123
Wet
seas
on b
enef
ited
are
a in
dex
UPR
IIS
110
0 10
07
114
4 10
55
113
8 11
74
951
10
71
971
ll
58
10
89
106
7 -0
47
Ang
at-M
aasi
m R
97
9
974
92
0
983
10
28
100
8 99
7
102
1 98
2
931
97
7
988
0
54
Sto
To
mas
11
63
115
9 11
23
107
5 10
80
103
9 98
1
978
10
41
103
9 11
20
101
6 -4
88
Siba
lom
-San
Jos
e 11
35
103
8 10
11
931
93
4
906
10
07
975
10
1S
998
10
11
981
-0
80
Aga
nan-
Sta
Bar
bara
10
85
110
5 10
74
106
8 10
01
104
2 99
8
101
5
987
61
0
106
7 93
0
-1 8
4
Ave
rage
10
92
105
6 10
54
102
4 10
36
103
4 98
7
101
3 10
00
947
10
53
996
-2
32
Dry
sea
Son
bene
fite
d ar
ea
inde
x U
PRIIS
14
04
155
0 15
S0
155
8 16
17
128
0 57
2
114
S 15
74
152
3 12
38
-09
4 A
ngat
-Maa
sim
R
903
93
0
103
2 10
61
104
2 10
69
988
99
2
102
1 99
6
993
10
13
061
S
to
Tom
as
105
7 12
27
122
5
961
99
9
115
9 10
1 7
91
0
107
3 12
1 2
10
94
107
4 -0
28
Si
balo
m-S
an J
ose
Aga
nan-
Sta
Bar
bara
66
5
95S
62
6
632
67
4
111
4
501
10
S1
412
11
51
766
93
3
856
94
3
107
0 94
8
107
4 11
09
111
6
158
6 58
8
987
97
6
110
4 5
35
083
N
w
Ave
rage
89
6
964
11
19
103
7 10
44
110
9 10
1 7
89
8
108
5 12
97
101
7 10
81
082
Wet
seas
on d
isch
arge
in
dex
UPR
IIS
132
9 72
7
142
5 11
88
120
4 98
8
105
8 10
63
879
96
4
117
5 99
1
-16
5 A
ngat
-Maa
sim
R
129
7 13
52
134
5 12
58
127
0 11
70
120
3 62
7
131
3 10
68
-18
9
Sto
To
mas
14
71
155
0 14
1 6
10
40
725
12
35
112
3 14
79
103
1 -4
48
Siba
lom
-San
Jos
e 96
8
733
11
1 5
92
2
907
47
1
102
3 85
7
ll2
0
422
92
9
779
-1
09
Aga
nan-
Sta
Bar
bara
87
9
863
68
1
925
96
8
110
7 70
5
871
87
7
00
9
Ave
rage
11
49
105
7 13
61
115
0 10
58
853
10
43
963
99
4
SO3
11
55
940
-2
85
Dry
sea
son
disc
harg
e in
dex
UPR
IIS
425
13
09
153
3 14
28
180
6 14
S0
125
8 56
3
117
9 14
66
130
0 11
89
-04
3
Ang
at-M
aasi
m R
12
26
161
6 15
30
139
2 12
85
100
8 10
33
958
14
41
107
1 -3
80
S
to
Tom
as
116
4 14
82
155
7 79
9
971
12
30
126
5 14
01
106
6 -2
44
Si
balo
m-S
an J
ose
486
71
5
782
62
9
789
64
5
116
9 11
86
801
65
3
918
2
36
Aga
nan-
Sta
Bar
bara
A
vera
ge
425
10
46
133
6 62
0
118
4 81
5
116
1 94
7
112
5 89
7
922
12
20
991
88
2
108
7 12
36
119
2 71
8
114
0 10
37
105
5 3
23
-07
5
a bull
indi
cate
s si
gnif
ican
ce a
t 95
per
cent
con
fide
nce
- 24 shy
Table 5--Results of pooled regressions yield indicators
------------------------- Dependent variables -------------------------shyIndependent variable ---------- Yield MTHa ----------- -- Specific Yield 1000 KgM
3 H20 shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season -shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 1967 (286)
3747 (580)
1991 (293)
2800 (460 )
0084 (041)
0441 (229)
0049 (023 )
0876 (343)
System 2 indicator 0400 (155)
0525 (226)
0057 (085)
0115 (1 95)
System 3 indicator 0466 (139 )
-0029 (-010)
-0016 (-018)
0006 (008)
System 4 indicator 0980 (470)
0090 (049)
0176 (333)
0391 (827)
System 5 indicator 1 221 (570)
0256 ( 138)
0104 (174)
0224 (440)
Reservoir indicator -1033 (-578)
-0063 (-039)
-0146 (-310)
-0326 (-569)
Period indicatora
Precipitation in mm Wet season
Annual
0163 (1 03)
-29E-4 (-182)
0231 (1 42)
-58E-4 (-606)
0161 (118)
92E-5 (072)
0199 043 )
-64pound-5 (-084)
0082 (200)
57E-6 (013)
0092 (223)
-61E-5 (-232)
0032 (093)
29E-7 (001)
0052 (112)
-12pound-4 (-414)
Nitrogen applicationb 0025 (254)
0021 (211)
0025 (274)
0020 (227)
0003 (1 05)
0002 (043)
0003 (101)
-0002 (-052)
R2 (adj) 0553 0516 0251 0198 0309 0262 0713 0445 number of observations 50 50 49 49 42 42 40 40 degrees of freedom 42 45 41 44 34 37 32 35
CoeffiCient (t statistic) indicates significance at 95 confidence ~ 1982-1986 = 1
kgha
- 25 shy
Table 6--Suggmry of estiates perforance indicators
Variable Predicted Predicted Period Dependent Variable
Description (units)
Period Coefficient
1976-86 IndicatorO
1976-86 Indicatorl
Effecta (X)
BenefittedService area ratio
Wet season ratio -0036 0853 0817 -425 Dry Season 0071 0539 0610 1313
Delivery Wet Dry
season season
nm water per benefitted ha
-0808 -1 214
6003 9083
5195 7869
-1346 -1337
Rice Yield Wet season metric tons 0163 3578 3741 456 Dry season per hectare 0161 3905 4066 412
Specific Yield Wet season kg rice per m3 0082 0348 0430 2353 Dry season water de 1 ivered 0032 0378 0409 837
a percentage changes relative to period=O values
- 26 shy
ENDNOTES
IResearch Fellow International Food Policy Research Institute and Senior Irrigation Specialist International Irrigation Management Institute
2Presumable farmers also base private investment decisions on such foreknowledge of costs and commonly do so in the case of small tubewells In addition user groups may invest in small communal schemes independently after examining projected costs and benefits Most commonly however new irrigation capacity is created as a result of public investment decisions made by government and thus the political process comes into play
3This principle applies only when water charges are levied volumetrically--an exceedingly rare situation in the developing worldshy-though the argument is frequently invoked as though it applies to all allocative situations
4For Cabanatuan City the longer period of record available ie 1904 to 1986 was divided into two equal periods and tested for difference in means to test the observation by farmers that rainfall today is lower than it used to be Mean precipitation after 1947 was actually about 3 percent higher than before but the difference was not significant (t = 07)
5The procedure used was to take time series data on total nitrogenfertilizer consumption in the Philippines multiply it by the ratio of N fertilizer used on rice to total N consumption and divide the product by the rice harvested area to obtain an estimate of N use perhectare Values of the ratio of N use on rice to total N use were taken from IRRI (1988 150) with interpolation used to fill in gaps in the series
6This assumes that timeliness is evaluated against a standard based on the physiological demand for water generated by the rice crop and the varying sensitivity of the crop to stress at different growth stages In such a case a poor timeliness rating for irrigationservice will result in reductions in yield
7Because rice yields are insensitive to water applications that exceed PET values a redistribution of water from water excess to water short portions of the command area will usually result in higher aggregate output and higher average yields Exceptional scenarios can be constructed in which this general rule would be violated but such scenarios are unlikely to play out in practice
- 27 shy
aThere are a variety of standards for judging equity that could be used here The most common one is equality of per hectare water deliveries In the present case the implicit standard is equality of water supply utility in producing a crop Thus if two areas received equal amounts of water on a per hectare basis but one had such high seepage and percolation losses that the crop failed to produce a harvest (and was therefore exempted from irrigation fee payment) the water supply to the combined area would be judged inequitable Because such radical differences in seepage and percolation rates are unlikely on puddled rice soils in the lowland Philippines and because the definition of benefitted area accommodates a wide range of acceptable yields these two definitions are largely indistinguishable at the current limits of data resolution
- 11 shy
varies more than does SA particularly during the dry season when available water supply may seriously constrain the area which can be planted Its magnitude is a function of system managers actions in authorizing the amount of land to be planted in a given season farmers decisions regarding whether to plant or not and the combined ability of system managers and farmersirrigators subsequently to distribute water Both of these area measures will be used to standardize other variables for particular purposes as well as beingcombined to form a separate indicator by themselves
Adequacy The most direct measure of the adequacy of irrigation water supplies to the agricultural system is the quantity of water applied to the system command area on a per unit area basis relative to some standard In this case since our interest is in differences in water adequacy between two time periods and since the systems being assessed have been and continue to be almost entirely devoted to rice cultivation during both cropping seasons depth measures for the two periods may be compared directly assuming the seasonal cropdemand for water to be unchanged Although dry-footed crops can suffer yield losses from overapplication of irrigation water rice is largely insensitive to this effect In addition water can substitute for other inputs that the farmer would otherwise have to provide such as weed control and more careful (and costly) water management We assume therefore that other things being equal larger values of depth applied are better than smaller values in terms of meeting crop water demands and reduce the cost of cultivation At the same time high levels of water adequacy can affect the values of other performance measures--particularly equity
When the regression model is run for quantity of water diverted at the system headworks divided by BA hereafter termed depth we see that the period dummy is negative and significant at the 95 percent confidence level for both wet and dry seasons (see Table 3 equations 1 and 3) Since the overall explanatory power of the wet season model is very weak however we will focus on the dry season in interpreting this result which indicates that after adjusting for rainfall differences significantly less water was delivered to the command per unit of benefitted area following 1981 than before This indicates based on the criteria outlined above that performance in terms of water adequacy deteriorated following financial selfshysufficiency We need to examine this conclusion more carefullyhowever
One difficulty is that the measured quantity of water diverted at the source is largely a function of the supply available in the river rather than of system management This is particularly true during the dry season and in non-reservoir systems Thus while the depth of water supplied to the system is a measure of the adequacy of the systems service it is to some extent beyond the control of the managing agency To better understand the factors behind this decline in water availability we look at simple unadjusted index values for
- 12 shy
several of the key variables Table 4 shows annual values of total volume of water delivered in each season SA and BA in both wet and dry seasons and shows the results of t-tests on the means of a set of indicators before and after 1981 Indicators are used rather than the actual values to weight each of the systems equally regardless of its size The table shows that both the average wet season benefitted area and the average discharge are significantlylower during the second period compared with the first For the dry season too the discharge index is lower after 1981 than before but this difference is not significant At the same time the dry season BA index rose slightly but again the change was not significantSince middotthere is not a clear pattern of relative movement of discharge and BA during the respective seasons no simple interpretation of these index value changes is possible What stands out is that both discharge and benefitted area declined across periods during the wet season while during the dry season there was no significant change in either indicator across the two periods It seems clear that the decline in water adequacy must be evaluated together with other measures of performance in drawing conclusions about the overall impact of the 1981 changes on the quality of system management
Another measured variable per hectare yield can be used as a proxy for water adequacy It has the advantage of partiallyreflecting the impacts of the dimensions of timeliness6 and equity7of distribution as well integrating all three effects into a combined impact on aggregate crop production Table 5 (equations 1 and 3)shows that the period dummy in the yield regressions has a positive sign in both seasons after controlling for nitrogen application and precipitation though the t-values are not significant Treatingyield adjusted in this way as a proxy for quality of irrigationservice leads to the conclusion that by this more comprehensive measure quality of service held constant across the two periods in the dry season Because of the large yield component accounted for byrainfall during the wet season no such judgement is possible for that season however
Equity As noted earlier no reliable data are available for subdivisions of the five sample systems making direct computation of equity measures impossible We can make some judgements about changesin the equity of water distribution however by examining changes in the ratio of two area measures given for each system SA and BA Since SA is the area which theoretically can be supplied with irrigation water by the system and BA s the area which actuallyreceives a quantity of water adequate to produce a remunerative crop the ratio of the two provides a measure of the percentage of the potential service area which was irrigated to a particular standard The larger this pErcentage the more equitable 8 is the distribution This of course assumes that the quantity of water available to the systems is constant across the two periods
- 13 shy
Since the condition of constancy of water supply is not generally satisfied a regression was run in which the total quantityof water diverted at the headworks of each system divided by the systems potential service area SA was included in the regression to control for changes in the water supply available to the system seeTable 3 equations 5 through 8 The average daily rainfall received directly on the system service area during the season was also included as an independent variable The regression was run separately for wet and dry seasons The sign and t-statistic of the period dummy should then tell us whether or not equity as reflected in the BASA ratio increased decreased or remained unchanged across the period divide
Both equations are reasonable good as indicated by the R2 values though the dry season equation is considerably better as would be expected For the wet season both the water delivery term and the rainfall term in equation 5 are of positive sign but are nonshysignificant at the 95 percent confidence level indicating that wet season irrigated area does not change appreciably in response to level of wet season rainfall or the available irrigation water supply The period dummy was negative but not significant indicating that equityof distribution as reflected in the BASA ratio was similar during the two periods
For the dry season the water delivery term in equation 7 is positive and strongly significant indicating a close relationship between the fraction of potential area actually irrigated and the water supply available at the headworks In addition however the period dummy is positive and significant suggesting that once the influence of water supply is removed the BASA ratio was significantly higher in the period following 1981 than it was before
This is an important finding for it reflects significantlyimproved performance in terms of a factor equity of water distribution that is under the control of the managing entity an entity which here comprises both NIA and irrigators associations Interpreted in these terms NIA and allied farmers associations were able to spread a given amount of water more widely across the potential command area of the five sample systems in the period after 1981 than before Moreover they did this in a way that did not decrease average system yields as discussed earlier In making this interpretation we are suggesting that there was some redistribution of water from better-watered areas to fringe areas which would otherwise not have received irrigation water and that this redistribution was a direct response to the change in NIA prioritiesand operating policies and rules occurring around 1981
It is difficult to prove the assertion that water was in fact redistributed with a resulting increase in directly-measured equity Without access to reliable discharge data broken out by systemsection and we can only assume in the absence of a plausible
- 14 shy
alternative explanation that it was such a redistribution that made the increase in the BASA ratio possible In a larger sense it is difficult to prove conclusively that any outcome in a before and after analysis was the result of a particular independent causative factor In this case we have tried to remove the influence of other potential causative factors where we could but the possibilityremains that some combination of unmeasured factors are responsiblefor the difference in the BASA ratio found We do note though that this type of response is exactly the type that would be expected to follow from an emphasis on increased farmer satisfaction and cooperation and increased fee revenues Because the fee schedule is tied to benefitted area the only ways NIA can increase its revenue from that source are to expand benefitted area and to increase collection efficiencies The former depends on redistributing a fixed supply of water over a larger portion of the command while the latter requires that farmers be satisfied with the irrigation service they are receiving and the commitment of the local irrigators association to assist in the task of collecting the amounts due The evidence while not conclusive is highly suggestive that this is exactly what has happened
Efficiency In addition to measures which reflect the levels of adequacy and equity of irrigation service available data allow the calculation of a measure of operating efficiency The term efficiencyusually denotes the relationship between inputs to a process and its outputs often expressed as a ratio The output measure employed here is aggregate system rice output and the input is quantity of irrigation water turned into the system Dividing the first by the second gives a measure of agricultural production per unit water--here termed specific yield This is a highly integrated measure that evaluates the combined efficiency of the irrigation and agricultural processes As such it is a function of the managerial and other inputs supplied both to the irrigation system and to the agricultural operation With respect to one important input to the irrigation system we do know that NIA per hectare field operating expenses were about 29 percent lower in real terms in the 1982-86 period comparedto the 1976-1981 period although this drop may have been partlyoffset by increases in farmer-supplied labor inputs Other things being equal one would thus expect to find a decline in output efficiency
The regression analysis shows positive signs for the period terms in both wet and dry season equations (see Table 5) In the case of the wet season the period dummy in equation 5 is significant but the overall explanatory power of the model is quite low For the dry season (equation 7) the coefficient is positive but non-significant This means that after taking rainfall and fertilizer use into account data do not indicate a lowering of specific yield in the wake of funding reductions and the strong emphasiS on financial viabilitybeginning in 1981 This result provides evidence that the efficiency of the overall irrigation deliveryagricultural production process
- 15 shy
relative to the system water input did not falloff as a result of the changes implemented at least over the short run
Impact magnitude
The preceding analysis has shown us that some indicators of irrigation performance changed significantly following the managerial changes of 1981 while others did not However it has not given us a sense of the size of the changes which occurred To determine the magnitude of these changes the regression model is used to predict the response of the composite system to the managerial changes given a common set of -input and environmental factors To do this averagevalues of the independent variables from the entire eleven-year period 1976 to 1986 are put into the model together with the previously determined coefficients to generate predicted average values of the various dependent variables used in the earlier analysis with and without the period dummy This procedure produces a pair of estimates for each dependent variable under the same conditions--one in which the system responds as it did after the managerial changes were implemented and one in which it responds as it did prior to their introduction The differences between these two values thus indicate the magnitude of the changes occurring in the various indicators of performance discussed above
The results of this exercise are shown in Table 6 The table shows that water availability decreased by about 13 percent in both wet and dry seasons when the period dummy was included and while the coefficients responsible were significant in the earlier analysisthis difference cannot be easily connected with levels of system management as discussed earlier With respect to rice output per hectare although the coefficients were not very significant it is interesting to note that yield increases by 163 kilograms per hectare for the wet season and by 101 kilogram for the dry when the period dummy is included in spite of the reduced water supply available Keep in mind that the predicted yield values have already been adjusted for differences in nitrogen fertilizer use and rainfall This suggests that timeliness and equity of distribution of water supply to farmers may have increased following the changes contributing to the higher predicted yields
Examining the impact of increased equity of distribution bylooking at the ratio of benefitted area to service area we recall that the change was positive and significant for the dry season and negative and not significant for the wet Table 6 shows that the dry season BASA ratio increases by 7 percentage points when the dummy is included a 131 percent increase Other things being equal this should result in a 131 percent increase in system output due to the expansion of ared benefitted This is a major impact on production
- 16 shy
CONCLUSIONS
The Philippine experiment to transform the national irrigation agency into an enterprise has undoubtedly been successful in reducing system operating expenses bringing revenues and costs into line and eliminating the recurrent cost burden imposed by large-scaleirrigation systems on the national budget Evidence presented in this paper indicates that in the process equity of water distribution across systems has also improved In the 5 years following the cessation of operating subsidies from the government an index of equity of distribution improved by about 13 percent At the same time per hectare yields adjusted for rainfall and nitrogen application held constant
There is a strong logical connection between the achievement of financial viability and improved equity of water distribution across the command Because increasing irrigation fees is a politicaldecisionlying largely beyond NIAs control expanding the area which can be billed for service is one of the few revenue increasing measures available to the irrigation agency which does not involve major additional investment In the face of constant or shrinkingwater supplies this is achieved only by redistributing water from areas receiving excessive supplies usually near the head ends of canals and laterals to areas receiving no supplies or inadequatesupplies often located near the tails of canals Although data are not available which would allow the direct examination of this hypotheses the two outcomes are logically consistent with each other
Data also show that per hectare water deliveries declined significantly in the five sample systems after 1981 even thoughrainfall did not differ appreciably between the two periods This decline averaged about 13 percent for both wet and dry seasons and is interpreted as a decline in water availability in the supplying rivers rather than a conscious reduction in withdrawals by system managers Such declines could result from changes in watershed runoff characteristics as caused by deforestation or from increased upstream abstractions from supplying rivers
Improved water distribution tends to increase the area served system agricultural output and NIA service fee revenue Reduced water supplies to the system tend to reduce these things Specificyield defined as system paddy output per unit water held roughly constant across the two periods indicating that the two effects mayhave offset each other
After adjusting for rainfall and nitrogen application perhectare yields increased only marginally in the post-1981 period Area served on the other hand increased by about 13 percent after adjusting for water supply availability indicating that the area benefitted by irrigation in the sample systems increased by about the same percentage Even if yields on this additional area are less than
- 17 shy
average yields for the system this still represents a sizeable increase in system agricultural output as a result of the change in management structure the increase coming not from higher yields but from expanded area under irrigation
The evidence assembled here suggests that there are significantfinancial and economic benefits to be had from changes in the basic character of irrigation managing agencies which make them more responsive to their clientele and which impose rational internal financial discipline on the agency The analysis suggests a number of additional questions however One relates to the longer-term impacts of the structural management changes The improvements in water distribution described here are relatively short-term events occurring during the first 5 years of the new management mode Critics have suggested the danger of underinvestment in systemmaintenance over the longer run accompanied by declining yields and benefitted areas and eventual system collapse This possibility needs to be closely monitored A second concern relates to the apparent decline in water supply to these 5 geographically dispersed systems The nature and causes of this decline need to be explored further since if widespread and secular it may represent a serious threat to the stability of Philippine rice production Whether stemming from poor forest management practices or deficient regulation and allocation of surface water resources or other unidentified factors it is an issue that deserves serious and urgent consideration
A third risk is that the incentive structure set up by NIA to guide and stimulate the performance of field units overemphasizes revenue generation at the expense of irrigation service provision to farmers The evidence presented here supports the view that these two objectives are mutually reinforcing under policies and conditions which have been established in the Philippines More detailed crossshysectional studies based on primary flow measurement data would add confidence to this conclusion and help to specify the conditions under which this effect occurs This could be extremely important in transferring the results of the Philippine experiment to other countries
A final risk is that outside intervention well meaning or otherwise will destroy the basis of NIAs financial autonomy or will impose external pressures or constraints on NIAs decision-making that will subvert the management practices which have been so painstakinglydeveloped and implemented Among these are calls for NIA to be subsumed again within the government department structure in the interests of better coordination with agriculture attempts byexternal financing agencies to arbitrarily increase NIAs expenditures on OampM on the assumption that this will increase system agricultural output or intervention by Philippine legislative bodies to restore operating subsidies to NIA with attached strings leading back to legislators home districts Pressures such as these will cut short a process of experimentation and improvement that seems promising
- 18 shy
enough to date to warrant its continuation Having developed the capacity to establish targets and implement and manage change NIA is in a strong position to modify its objectives to better achieve larger social purposes established for it It is critical to recognize however that this must happen within the context of financing policies that mandate financial autonomy for NIA if the fundamental institutional commitment to manage is to be preserved
The author would like to thank Leslie Small and JeremyBerkoff for helpful comments on an earlier unpublishedversion of this paper and Charles Rogers for his careful and creative help with the analysis
- 19 shy
BIBLIOGRAPHY
Abernethy Charles L 1990 Indicators of the performance of irrigation water distribution systems International Irrigation Management Institute Colombo Sri Lanka Mimeo
Asian Development Bank 1986 Irrigation service fees Proceedingsof the Regional Seminar on Irrigation Service Fees Manila Asian Development Bank
Carruthers Ian and Colin Clark 1981 Economics of IrrigationLiverpool Liverpool University Press Third Edition
Levine G and EW Coward Jr 1986 Irrigation water distribution implications for design and operation AGREP Division WorkingPaper 125 vol 1 World Bank Agriculture and Rural Development Department
Small Les E 1989 User charges in irrigation potentials and limitations Irrigation and drainage vol 3 no 2125-142
Small Les 1990 Irrigation service fees in Asia IrrigationManagement Network 9013 London Overseas DevelopmentInstitute
Svendsen Mark and Les Small 1989 A framework for assessing irrigation system performance Paper prepared for the Symposium on Performance Evaluation 23 November International IrrigationManagement Institute Sri Lanka
Table I--National Irrigation Administration revenues and expenditures in constant prices 1976-86
Item 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986
(peso million 1972)
Revenues Irrigation fees collected 1273 1483 17 13 1831 2070 1668 1699 1893 1728 2129 2546 Other income 715 737 2420 5591 2631 5990 7783 6638 5699 4932 2934
Total direct revenue 1988 2220 4133 7422 4701 7658 9482 8531 7427 061 5480
Expenses in 1972 pricesTotal expenses 4825 5716 5039 6329 3821 77 55 6166 4749 4348 4259 4959
Excess (deficit) (2837(3496) (906) 1093 877 (097) 3316 3782 3079 2802 521 N 0
Subsidies Government operation and
maintenance subsidies 2521 2741 2799 1817 1398 633 0 0 0 0 0 Calamity fund payments 548 0 0 0 0 0 0 0 119 0 142
Total subsidy 3069 2741 2799 1817 1398 633 0 0 119 0 142
Total excess (deficit) 231 (754) 1893 2910 2275 536 3316 3782 3198 2802 663
Source IFPRI analysis of NIA data
- 21 shy
Table 2--Descriptive characteristics of selected MIA systeasa
------- Region III -------- ----- Region VI ----shyUPRIIS Angatii Sto Sibalom- Aganan-
Haasim Thomasc San Jose Sta Barbara
Average service area (hal 102272 31462 3522 5282 8703 Average irrigated area (hal
Wet season 83768 23454 3007 4410 8300 Dry season
Average benefited area Wet season
(hal 64587
77 605
27639
22908
1 781
3007
2801
4369
2770
7698 Dry season
Average rainfall Wet season
(mml d 62478
1 685 5
27396
8576
1 781
3051 0
2769
24731
2997
20001 Dry season 756 333 322 2828 3025
Average discharge (Llsec) Wet season 46501 14792 1692 2353 4984 Dry season 78091 22812 2014 1276 2315
Average water delivery (mmday) Wet season 522 548 487 462 571 Dry season
Average yield (mtha) 1089 715 995 398 686
Wet season 345 419 322 395 435 Dry season
Avg yield per unit water 34 03
(kgm ) 451 412 399 426
Wet season 0373 0440 0373 0538 0443 Dry season 0248 0400 0279 0690 0428
t-statistic difference in mean rainfall 1978-81 1982-86e
Wet season 0432 0713 -0567 1169 1169 Dry season 0519 -0230 -0523 1187 1187 Annual 0460 0707 -0686 1445 1445
~ Summary numbers are averages for the period 1982-1986 except as noted Water delivery discharge and yield per unit discharge are 4-year averages 1982-1985
c Water delivery discharge and yield per unit discharge are 4-year averages d 1983-1986
For Angat 5 years are 1981-85 For St Thomas 1979-83 For Sibalom 1971-75 e No significant differences at 95 confidence
- 22 shy
Table 3--Results of pooled regressions water delivery indicators
Dependent variable Independent variable - Water Delivery in mmBenefited Ha - --- BenefitService Area Ratio --shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season-shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 6799 (998)
6189 (1019)
13102 (1093)
2411 (1 54)
0743 ( 1186)
0800 (1165)
0129 (131 )
0258 (329)
System 2 indicator 0173 (029)
-3743 (-363)
-0063 (-219)
0200 (505)
System 3 indicator 0698 (085)
0163 (012)
0085 (208)
-0046 (-097)
System 4 indicator -0514 (-l 07)
-7011 (-803)
0080 (340)
0169 (l87)
System 5 indicator 0177 (033)
-5049 (-538)
0174 (603)
0051 (058)
Reservoir indicator 0327 (078)
5703 (543)
-0126 (-510)
0156 (261)
Period indicatora -0808 (-240)
-0771 (-229)
-1214 (-209)
-1309 (-1 62)
-0036 (-192)
-0025 (-112)
0071 (239)
0030 (077)
Water delivery mmb
Wet season
Dry season
Precipitation in mm Wet season
Annual
Average daily rainfall Wet season
Dry season
-46E-4 (-125)
-1 9E-4 (-082)
-43E-4 (-077)
20E-3 (385)
0015 (1 59)
-0001 ( -004)
0023 (215)
0001 (021)
0070 (613)
-0017 (-058)
0046 (371)
0010 (033)
R2 (adj) number of observations degrees of freedom
0156 43 36
0133 43 39
0711 41 34
0434 41 37
0693 42 34
0556 42 37
0865 40 32
0730 40 35
Coefficient (t statistic) indicates significance at 95 confidence ~ 1982-1986 1
per ha of service area
Ta
ble
--I
nd
ices
of
se
v
ice a
rea
ib
en
ef
ited
a
rea
a
nd
a
vera
ge sea
so
na
l d
isch
arg
e
Ave
rage
A
vera
ge
tshy19
77
1978
19
79
1980
19
81
1982
19
83
1984
19
85
1986
19
77-8
1 19
82-8
6 S
tati
stic
a
(ind
ex
aver
age
1983
-198
5 =
100)
Ser
vice
are
a UP
RI IS
in
dex
882
91
9
920
91
5
951
95
1
100
0 10
00
100
0 10
00
917
99
0
bull5
53
Ang
at-M
aasi
m R
95
9
994
99
7
996
99
6
996
10
00
100
0 10
00
100
0 98
8
999
1
63
Sto
To
mas
10
63
106
3 10
1 9
10
29
103
0 99
8
100
0 10
00
100
0 11
10
104
1 10
22
-08
9
Siba
lom
-San
Jos
e 95
0
872
94
2
933
94
2
942
10
28
102
8 94
4
101
7
928
99
2
292
A
gana
n-St
a B
arba
ra
108
3 10
91
106
0 10
30
961
10
05
996
10
08
996
99
6
104
5 10
00
-21
1 A
vera
ge
987
98
8
981
98
1
916
97
8
100
5 10
07
988
10
25
984
10
01
123
Wet
seas
on b
enef
ited
are
a in
dex
UPR
IIS
110
0 10
07
114
4 10
55
113
8 11
74
951
10
71
971
ll
58
10
89
106
7 -0
47
Ang
at-M
aasi
m R
97
9
974
92
0
983
10
28
100
8 99
7
102
1 98
2
931
97
7
988
0
54
Sto
To
mas
11
63
115
9 11
23
107
5 10
80
103
9 98
1
978
10
41
103
9 11
20
101
6 -4
88
Siba
lom
-San
Jos
e 11
35
103
8 10
11
931
93
4
906
10
07
975
10
1S
998
10
11
981
-0
80
Aga
nan-
Sta
Bar
bara
10
85
110
5 10
74
106
8 10
01
104
2 99
8
101
5
987
61
0
106
7 93
0
-1 8
4
Ave
rage
10
92
105
6 10
54
102
4 10
36
103
4 98
7
101
3 10
00
947
10
53
996
-2
32
Dry
sea
Son
bene
fite
d ar
ea
inde
x U
PRIIS
14
04
155
0 15
S0
155
8 16
17
128
0 57
2
114
S 15
74
152
3 12
38
-09
4 A
ngat
-Maa
sim
R
903
93
0
103
2 10
61
104
2 10
69
988
99
2
102
1 99
6
993
10
13
061
S
to
Tom
as
105
7 12
27
122
5
961
99
9
115
9 10
1 7
91
0
107
3 12
1 2
10
94
107
4 -0
28
Si
balo
m-S
an J
ose
Aga
nan-
Sta
Bar
bara
66
5
95S
62
6
632
67
4
111
4
501
10
S1
412
11
51
766
93
3
856
94
3
107
0 94
8
107
4 11
09
111
6
158
6 58
8
987
97
6
110
4 5
35
083
N
w
Ave
rage
89
6
964
11
19
103
7 10
44
110
9 10
1 7
89
8
108
5 12
97
101
7 10
81
082
Wet
seas
on d
isch
arge
in
dex
UPR
IIS
132
9 72
7
142
5 11
88
120
4 98
8
105
8 10
63
879
96
4
117
5 99
1
-16
5 A
ngat
-Maa
sim
R
129
7 13
52
134
5 12
58
127
0 11
70
120
3 62
7
131
3 10
68
-18
9
Sto
To
mas
14
71
155
0 14
1 6
10
40
725
12
35
112
3 14
79
103
1 -4
48
Siba
lom
-San
Jos
e 96
8
733
11
1 5
92
2
907
47
1
102
3 85
7
ll2
0
422
92
9
779
-1
09
Aga
nan-
Sta
Bar
bara
87
9
863
68
1
925
96
8
110
7 70
5
871
87
7
00
9
Ave
rage
11
49
105
7 13
61
115
0 10
58
853
10
43
963
99
4
SO3
11
55
940
-2
85
Dry
sea
son
disc
harg
e in
dex
UPR
IIS
425
13
09
153
3 14
28
180
6 14
S0
125
8 56
3
117
9 14
66
130
0 11
89
-04
3
Ang
at-M
aasi
m R
12
26
161
6 15
30
139
2 12
85
100
8 10
33
958
14
41
107
1 -3
80
S
to
Tom
as
116
4 14
82
155
7 79
9
971
12
30
126
5 14
01
106
6 -2
44
Si
balo
m-S
an J
ose
486
71
5
782
62
9
789
64
5
116
9 11
86
801
65
3
918
2
36
Aga
nan-
Sta
Bar
bara
A
vera
ge
425
10
46
133
6 62
0
118
4 81
5
116
1 94
7
112
5 89
7
922
12
20
991
88
2
108
7 12
36
119
2 71
8
114
0 10
37
105
5 3
23
-07
5
a bull
indi
cate
s si
gnif
ican
ce a
t 95
per
cent
con
fide
nce
- 24 shy
Table 5--Results of pooled regressions yield indicators
------------------------- Dependent variables -------------------------shyIndependent variable ---------- Yield MTHa ----------- -- Specific Yield 1000 KgM
3 H20 shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season -shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 1967 (286)
3747 (580)
1991 (293)
2800 (460 )
0084 (041)
0441 (229)
0049 (023 )
0876 (343)
System 2 indicator 0400 (155)
0525 (226)
0057 (085)
0115 (1 95)
System 3 indicator 0466 (139 )
-0029 (-010)
-0016 (-018)
0006 (008)
System 4 indicator 0980 (470)
0090 (049)
0176 (333)
0391 (827)
System 5 indicator 1 221 (570)
0256 ( 138)
0104 (174)
0224 (440)
Reservoir indicator -1033 (-578)
-0063 (-039)
-0146 (-310)
-0326 (-569)
Period indicatora
Precipitation in mm Wet season
Annual
0163 (1 03)
-29E-4 (-182)
0231 (1 42)
-58E-4 (-606)
0161 (118)
92E-5 (072)
0199 043 )
-64pound-5 (-084)
0082 (200)
57E-6 (013)
0092 (223)
-61E-5 (-232)
0032 (093)
29E-7 (001)
0052 (112)
-12pound-4 (-414)
Nitrogen applicationb 0025 (254)
0021 (211)
0025 (274)
0020 (227)
0003 (1 05)
0002 (043)
0003 (101)
-0002 (-052)
R2 (adj) 0553 0516 0251 0198 0309 0262 0713 0445 number of observations 50 50 49 49 42 42 40 40 degrees of freedom 42 45 41 44 34 37 32 35
CoeffiCient (t statistic) indicates significance at 95 confidence ~ 1982-1986 = 1
kgha
- 25 shy
Table 6--Suggmry of estiates perforance indicators
Variable Predicted Predicted Period Dependent Variable
Description (units)
Period Coefficient
1976-86 IndicatorO
1976-86 Indicatorl
Effecta (X)
BenefittedService area ratio
Wet season ratio -0036 0853 0817 -425 Dry Season 0071 0539 0610 1313
Delivery Wet Dry
season season
nm water per benefitted ha
-0808 -1 214
6003 9083
5195 7869
-1346 -1337
Rice Yield Wet season metric tons 0163 3578 3741 456 Dry season per hectare 0161 3905 4066 412
Specific Yield Wet season kg rice per m3 0082 0348 0430 2353 Dry season water de 1 ivered 0032 0378 0409 837
a percentage changes relative to period=O values
- 26 shy
ENDNOTES
IResearch Fellow International Food Policy Research Institute and Senior Irrigation Specialist International Irrigation Management Institute
2Presumable farmers also base private investment decisions on such foreknowledge of costs and commonly do so in the case of small tubewells In addition user groups may invest in small communal schemes independently after examining projected costs and benefits Most commonly however new irrigation capacity is created as a result of public investment decisions made by government and thus the political process comes into play
3This principle applies only when water charges are levied volumetrically--an exceedingly rare situation in the developing worldshy-though the argument is frequently invoked as though it applies to all allocative situations
4For Cabanatuan City the longer period of record available ie 1904 to 1986 was divided into two equal periods and tested for difference in means to test the observation by farmers that rainfall today is lower than it used to be Mean precipitation after 1947 was actually about 3 percent higher than before but the difference was not significant (t = 07)
5The procedure used was to take time series data on total nitrogenfertilizer consumption in the Philippines multiply it by the ratio of N fertilizer used on rice to total N consumption and divide the product by the rice harvested area to obtain an estimate of N use perhectare Values of the ratio of N use on rice to total N use were taken from IRRI (1988 150) with interpolation used to fill in gaps in the series
6This assumes that timeliness is evaluated against a standard based on the physiological demand for water generated by the rice crop and the varying sensitivity of the crop to stress at different growth stages In such a case a poor timeliness rating for irrigationservice will result in reductions in yield
7Because rice yields are insensitive to water applications that exceed PET values a redistribution of water from water excess to water short portions of the command area will usually result in higher aggregate output and higher average yields Exceptional scenarios can be constructed in which this general rule would be violated but such scenarios are unlikely to play out in practice
- 27 shy
aThere are a variety of standards for judging equity that could be used here The most common one is equality of per hectare water deliveries In the present case the implicit standard is equality of water supply utility in producing a crop Thus if two areas received equal amounts of water on a per hectare basis but one had such high seepage and percolation losses that the crop failed to produce a harvest (and was therefore exempted from irrigation fee payment) the water supply to the combined area would be judged inequitable Because such radical differences in seepage and percolation rates are unlikely on puddled rice soils in the lowland Philippines and because the definition of benefitted area accommodates a wide range of acceptable yields these two definitions are largely indistinguishable at the current limits of data resolution
- 12 shy
several of the key variables Table 4 shows annual values of total volume of water delivered in each season SA and BA in both wet and dry seasons and shows the results of t-tests on the means of a set of indicators before and after 1981 Indicators are used rather than the actual values to weight each of the systems equally regardless of its size The table shows that both the average wet season benefitted area and the average discharge are significantlylower during the second period compared with the first For the dry season too the discharge index is lower after 1981 than before but this difference is not significant At the same time the dry season BA index rose slightly but again the change was not significantSince middotthere is not a clear pattern of relative movement of discharge and BA during the respective seasons no simple interpretation of these index value changes is possible What stands out is that both discharge and benefitted area declined across periods during the wet season while during the dry season there was no significant change in either indicator across the two periods It seems clear that the decline in water adequacy must be evaluated together with other measures of performance in drawing conclusions about the overall impact of the 1981 changes on the quality of system management
Another measured variable per hectare yield can be used as a proxy for water adequacy It has the advantage of partiallyreflecting the impacts of the dimensions of timeliness6 and equity7of distribution as well integrating all three effects into a combined impact on aggregate crop production Table 5 (equations 1 and 3)shows that the period dummy in the yield regressions has a positive sign in both seasons after controlling for nitrogen application and precipitation though the t-values are not significant Treatingyield adjusted in this way as a proxy for quality of irrigationservice leads to the conclusion that by this more comprehensive measure quality of service held constant across the two periods in the dry season Because of the large yield component accounted for byrainfall during the wet season no such judgement is possible for that season however
Equity As noted earlier no reliable data are available for subdivisions of the five sample systems making direct computation of equity measures impossible We can make some judgements about changesin the equity of water distribution however by examining changes in the ratio of two area measures given for each system SA and BA Since SA is the area which theoretically can be supplied with irrigation water by the system and BA s the area which actuallyreceives a quantity of water adequate to produce a remunerative crop the ratio of the two provides a measure of the percentage of the potential service area which was irrigated to a particular standard The larger this pErcentage the more equitable 8 is the distribution This of course assumes that the quantity of water available to the systems is constant across the two periods
- 13 shy
Since the condition of constancy of water supply is not generally satisfied a regression was run in which the total quantityof water diverted at the headworks of each system divided by the systems potential service area SA was included in the regression to control for changes in the water supply available to the system seeTable 3 equations 5 through 8 The average daily rainfall received directly on the system service area during the season was also included as an independent variable The regression was run separately for wet and dry seasons The sign and t-statistic of the period dummy should then tell us whether or not equity as reflected in the BASA ratio increased decreased or remained unchanged across the period divide
Both equations are reasonable good as indicated by the R2 values though the dry season equation is considerably better as would be expected For the wet season both the water delivery term and the rainfall term in equation 5 are of positive sign but are nonshysignificant at the 95 percent confidence level indicating that wet season irrigated area does not change appreciably in response to level of wet season rainfall or the available irrigation water supply The period dummy was negative but not significant indicating that equityof distribution as reflected in the BASA ratio was similar during the two periods
For the dry season the water delivery term in equation 7 is positive and strongly significant indicating a close relationship between the fraction of potential area actually irrigated and the water supply available at the headworks In addition however the period dummy is positive and significant suggesting that once the influence of water supply is removed the BASA ratio was significantly higher in the period following 1981 than it was before
This is an important finding for it reflects significantlyimproved performance in terms of a factor equity of water distribution that is under the control of the managing entity an entity which here comprises both NIA and irrigators associations Interpreted in these terms NIA and allied farmers associations were able to spread a given amount of water more widely across the potential command area of the five sample systems in the period after 1981 than before Moreover they did this in a way that did not decrease average system yields as discussed earlier In making this interpretation we are suggesting that there was some redistribution of water from better-watered areas to fringe areas which would otherwise not have received irrigation water and that this redistribution was a direct response to the change in NIA prioritiesand operating policies and rules occurring around 1981
It is difficult to prove the assertion that water was in fact redistributed with a resulting increase in directly-measured equity Without access to reliable discharge data broken out by systemsection and we can only assume in the absence of a plausible
- 14 shy
alternative explanation that it was such a redistribution that made the increase in the BASA ratio possible In a larger sense it is difficult to prove conclusively that any outcome in a before and after analysis was the result of a particular independent causative factor In this case we have tried to remove the influence of other potential causative factors where we could but the possibilityremains that some combination of unmeasured factors are responsiblefor the difference in the BASA ratio found We do note though that this type of response is exactly the type that would be expected to follow from an emphasis on increased farmer satisfaction and cooperation and increased fee revenues Because the fee schedule is tied to benefitted area the only ways NIA can increase its revenue from that source are to expand benefitted area and to increase collection efficiencies The former depends on redistributing a fixed supply of water over a larger portion of the command while the latter requires that farmers be satisfied with the irrigation service they are receiving and the commitment of the local irrigators association to assist in the task of collecting the amounts due The evidence while not conclusive is highly suggestive that this is exactly what has happened
Efficiency In addition to measures which reflect the levels of adequacy and equity of irrigation service available data allow the calculation of a measure of operating efficiency The term efficiencyusually denotes the relationship between inputs to a process and its outputs often expressed as a ratio The output measure employed here is aggregate system rice output and the input is quantity of irrigation water turned into the system Dividing the first by the second gives a measure of agricultural production per unit water--here termed specific yield This is a highly integrated measure that evaluates the combined efficiency of the irrigation and agricultural processes As such it is a function of the managerial and other inputs supplied both to the irrigation system and to the agricultural operation With respect to one important input to the irrigation system we do know that NIA per hectare field operating expenses were about 29 percent lower in real terms in the 1982-86 period comparedto the 1976-1981 period although this drop may have been partlyoffset by increases in farmer-supplied labor inputs Other things being equal one would thus expect to find a decline in output efficiency
The regression analysis shows positive signs for the period terms in both wet and dry season equations (see Table 5) In the case of the wet season the period dummy in equation 5 is significant but the overall explanatory power of the model is quite low For the dry season (equation 7) the coefficient is positive but non-significant This means that after taking rainfall and fertilizer use into account data do not indicate a lowering of specific yield in the wake of funding reductions and the strong emphasiS on financial viabilitybeginning in 1981 This result provides evidence that the efficiency of the overall irrigation deliveryagricultural production process
- 15 shy
relative to the system water input did not falloff as a result of the changes implemented at least over the short run
Impact magnitude
The preceding analysis has shown us that some indicators of irrigation performance changed significantly following the managerial changes of 1981 while others did not However it has not given us a sense of the size of the changes which occurred To determine the magnitude of these changes the regression model is used to predict the response of the composite system to the managerial changes given a common set of -input and environmental factors To do this averagevalues of the independent variables from the entire eleven-year period 1976 to 1986 are put into the model together with the previously determined coefficients to generate predicted average values of the various dependent variables used in the earlier analysis with and without the period dummy This procedure produces a pair of estimates for each dependent variable under the same conditions--one in which the system responds as it did after the managerial changes were implemented and one in which it responds as it did prior to their introduction The differences between these two values thus indicate the magnitude of the changes occurring in the various indicators of performance discussed above
The results of this exercise are shown in Table 6 The table shows that water availability decreased by about 13 percent in both wet and dry seasons when the period dummy was included and while the coefficients responsible were significant in the earlier analysisthis difference cannot be easily connected with levels of system management as discussed earlier With respect to rice output per hectare although the coefficients were not very significant it is interesting to note that yield increases by 163 kilograms per hectare for the wet season and by 101 kilogram for the dry when the period dummy is included in spite of the reduced water supply available Keep in mind that the predicted yield values have already been adjusted for differences in nitrogen fertilizer use and rainfall This suggests that timeliness and equity of distribution of water supply to farmers may have increased following the changes contributing to the higher predicted yields
Examining the impact of increased equity of distribution bylooking at the ratio of benefitted area to service area we recall that the change was positive and significant for the dry season and negative and not significant for the wet Table 6 shows that the dry season BASA ratio increases by 7 percentage points when the dummy is included a 131 percent increase Other things being equal this should result in a 131 percent increase in system output due to the expansion of ared benefitted This is a major impact on production
- 16 shy
CONCLUSIONS
The Philippine experiment to transform the national irrigation agency into an enterprise has undoubtedly been successful in reducing system operating expenses bringing revenues and costs into line and eliminating the recurrent cost burden imposed by large-scaleirrigation systems on the national budget Evidence presented in this paper indicates that in the process equity of water distribution across systems has also improved In the 5 years following the cessation of operating subsidies from the government an index of equity of distribution improved by about 13 percent At the same time per hectare yields adjusted for rainfall and nitrogen application held constant
There is a strong logical connection between the achievement of financial viability and improved equity of water distribution across the command Because increasing irrigation fees is a politicaldecisionlying largely beyond NIAs control expanding the area which can be billed for service is one of the few revenue increasing measures available to the irrigation agency which does not involve major additional investment In the face of constant or shrinkingwater supplies this is achieved only by redistributing water from areas receiving excessive supplies usually near the head ends of canals and laterals to areas receiving no supplies or inadequatesupplies often located near the tails of canals Although data are not available which would allow the direct examination of this hypotheses the two outcomes are logically consistent with each other
Data also show that per hectare water deliveries declined significantly in the five sample systems after 1981 even thoughrainfall did not differ appreciably between the two periods This decline averaged about 13 percent for both wet and dry seasons and is interpreted as a decline in water availability in the supplying rivers rather than a conscious reduction in withdrawals by system managers Such declines could result from changes in watershed runoff characteristics as caused by deforestation or from increased upstream abstractions from supplying rivers
Improved water distribution tends to increase the area served system agricultural output and NIA service fee revenue Reduced water supplies to the system tend to reduce these things Specificyield defined as system paddy output per unit water held roughly constant across the two periods indicating that the two effects mayhave offset each other
After adjusting for rainfall and nitrogen application perhectare yields increased only marginally in the post-1981 period Area served on the other hand increased by about 13 percent after adjusting for water supply availability indicating that the area benefitted by irrigation in the sample systems increased by about the same percentage Even if yields on this additional area are less than
- 17 shy
average yields for the system this still represents a sizeable increase in system agricultural output as a result of the change in management structure the increase coming not from higher yields but from expanded area under irrigation
The evidence assembled here suggests that there are significantfinancial and economic benefits to be had from changes in the basic character of irrigation managing agencies which make them more responsive to their clientele and which impose rational internal financial discipline on the agency The analysis suggests a number of additional questions however One relates to the longer-term impacts of the structural management changes The improvements in water distribution described here are relatively short-term events occurring during the first 5 years of the new management mode Critics have suggested the danger of underinvestment in systemmaintenance over the longer run accompanied by declining yields and benefitted areas and eventual system collapse This possibility needs to be closely monitored A second concern relates to the apparent decline in water supply to these 5 geographically dispersed systems The nature and causes of this decline need to be explored further since if widespread and secular it may represent a serious threat to the stability of Philippine rice production Whether stemming from poor forest management practices or deficient regulation and allocation of surface water resources or other unidentified factors it is an issue that deserves serious and urgent consideration
A third risk is that the incentive structure set up by NIA to guide and stimulate the performance of field units overemphasizes revenue generation at the expense of irrigation service provision to farmers The evidence presented here supports the view that these two objectives are mutually reinforcing under policies and conditions which have been established in the Philippines More detailed crossshysectional studies based on primary flow measurement data would add confidence to this conclusion and help to specify the conditions under which this effect occurs This could be extremely important in transferring the results of the Philippine experiment to other countries
A final risk is that outside intervention well meaning or otherwise will destroy the basis of NIAs financial autonomy or will impose external pressures or constraints on NIAs decision-making that will subvert the management practices which have been so painstakinglydeveloped and implemented Among these are calls for NIA to be subsumed again within the government department structure in the interests of better coordination with agriculture attempts byexternal financing agencies to arbitrarily increase NIAs expenditures on OampM on the assumption that this will increase system agricultural output or intervention by Philippine legislative bodies to restore operating subsidies to NIA with attached strings leading back to legislators home districts Pressures such as these will cut short a process of experimentation and improvement that seems promising
- 18 shy
enough to date to warrant its continuation Having developed the capacity to establish targets and implement and manage change NIA is in a strong position to modify its objectives to better achieve larger social purposes established for it It is critical to recognize however that this must happen within the context of financing policies that mandate financial autonomy for NIA if the fundamental institutional commitment to manage is to be preserved
The author would like to thank Leslie Small and JeremyBerkoff for helpful comments on an earlier unpublishedversion of this paper and Charles Rogers for his careful and creative help with the analysis
- 19 shy
BIBLIOGRAPHY
Abernethy Charles L 1990 Indicators of the performance of irrigation water distribution systems International Irrigation Management Institute Colombo Sri Lanka Mimeo
Asian Development Bank 1986 Irrigation service fees Proceedingsof the Regional Seminar on Irrigation Service Fees Manila Asian Development Bank
Carruthers Ian and Colin Clark 1981 Economics of IrrigationLiverpool Liverpool University Press Third Edition
Levine G and EW Coward Jr 1986 Irrigation water distribution implications for design and operation AGREP Division WorkingPaper 125 vol 1 World Bank Agriculture and Rural Development Department
Small Les E 1989 User charges in irrigation potentials and limitations Irrigation and drainage vol 3 no 2125-142
Small Les 1990 Irrigation service fees in Asia IrrigationManagement Network 9013 London Overseas DevelopmentInstitute
Svendsen Mark and Les Small 1989 A framework for assessing irrigation system performance Paper prepared for the Symposium on Performance Evaluation 23 November International IrrigationManagement Institute Sri Lanka
Table I--National Irrigation Administration revenues and expenditures in constant prices 1976-86
Item 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986
(peso million 1972)
Revenues Irrigation fees collected 1273 1483 17 13 1831 2070 1668 1699 1893 1728 2129 2546 Other income 715 737 2420 5591 2631 5990 7783 6638 5699 4932 2934
Total direct revenue 1988 2220 4133 7422 4701 7658 9482 8531 7427 061 5480
Expenses in 1972 pricesTotal expenses 4825 5716 5039 6329 3821 77 55 6166 4749 4348 4259 4959
Excess (deficit) (2837(3496) (906) 1093 877 (097) 3316 3782 3079 2802 521 N 0
Subsidies Government operation and
maintenance subsidies 2521 2741 2799 1817 1398 633 0 0 0 0 0 Calamity fund payments 548 0 0 0 0 0 0 0 119 0 142
Total subsidy 3069 2741 2799 1817 1398 633 0 0 119 0 142
Total excess (deficit) 231 (754) 1893 2910 2275 536 3316 3782 3198 2802 663
Source IFPRI analysis of NIA data
- 21 shy
Table 2--Descriptive characteristics of selected MIA systeasa
------- Region III -------- ----- Region VI ----shyUPRIIS Angatii Sto Sibalom- Aganan-
Haasim Thomasc San Jose Sta Barbara
Average service area (hal 102272 31462 3522 5282 8703 Average irrigated area (hal
Wet season 83768 23454 3007 4410 8300 Dry season
Average benefited area Wet season
(hal 64587
77 605
27639
22908
1 781
3007
2801
4369
2770
7698 Dry season
Average rainfall Wet season
(mml d 62478
1 685 5
27396
8576
1 781
3051 0
2769
24731
2997
20001 Dry season 756 333 322 2828 3025
Average discharge (Llsec) Wet season 46501 14792 1692 2353 4984 Dry season 78091 22812 2014 1276 2315
Average water delivery (mmday) Wet season 522 548 487 462 571 Dry season
Average yield (mtha) 1089 715 995 398 686
Wet season 345 419 322 395 435 Dry season
Avg yield per unit water 34 03
(kgm ) 451 412 399 426
Wet season 0373 0440 0373 0538 0443 Dry season 0248 0400 0279 0690 0428
t-statistic difference in mean rainfall 1978-81 1982-86e
Wet season 0432 0713 -0567 1169 1169 Dry season 0519 -0230 -0523 1187 1187 Annual 0460 0707 -0686 1445 1445
~ Summary numbers are averages for the period 1982-1986 except as noted Water delivery discharge and yield per unit discharge are 4-year averages 1982-1985
c Water delivery discharge and yield per unit discharge are 4-year averages d 1983-1986
For Angat 5 years are 1981-85 For St Thomas 1979-83 For Sibalom 1971-75 e No significant differences at 95 confidence
- 22 shy
Table 3--Results of pooled regressions water delivery indicators
Dependent variable Independent variable - Water Delivery in mmBenefited Ha - --- BenefitService Area Ratio --shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season-shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 6799 (998)
6189 (1019)
13102 (1093)
2411 (1 54)
0743 ( 1186)
0800 (1165)
0129 (131 )
0258 (329)
System 2 indicator 0173 (029)
-3743 (-363)
-0063 (-219)
0200 (505)
System 3 indicator 0698 (085)
0163 (012)
0085 (208)
-0046 (-097)
System 4 indicator -0514 (-l 07)
-7011 (-803)
0080 (340)
0169 (l87)
System 5 indicator 0177 (033)
-5049 (-538)
0174 (603)
0051 (058)
Reservoir indicator 0327 (078)
5703 (543)
-0126 (-510)
0156 (261)
Period indicatora -0808 (-240)
-0771 (-229)
-1214 (-209)
-1309 (-1 62)
-0036 (-192)
-0025 (-112)
0071 (239)
0030 (077)
Water delivery mmb
Wet season
Dry season
Precipitation in mm Wet season
Annual
Average daily rainfall Wet season
Dry season
-46E-4 (-125)
-1 9E-4 (-082)
-43E-4 (-077)
20E-3 (385)
0015 (1 59)
-0001 ( -004)
0023 (215)
0001 (021)
0070 (613)
-0017 (-058)
0046 (371)
0010 (033)
R2 (adj) number of observations degrees of freedom
0156 43 36
0133 43 39
0711 41 34
0434 41 37
0693 42 34
0556 42 37
0865 40 32
0730 40 35
Coefficient (t statistic) indicates significance at 95 confidence ~ 1982-1986 1
per ha of service area
Ta
ble
--I
nd
ices
of
se
v
ice a
rea
ib
en
ef
ited
a
rea
a
nd
a
vera
ge sea
so
na
l d
isch
arg
e
Ave
rage
A
vera
ge
tshy19
77
1978
19
79
1980
19
81
1982
19
83
1984
19
85
1986
19
77-8
1 19
82-8
6 S
tati
stic
a
(ind
ex
aver
age
1983
-198
5 =
100)
Ser
vice
are
a UP
RI IS
in
dex
882
91
9
920
91
5
951
95
1
100
0 10
00
100
0 10
00
917
99
0
bull5
53
Ang
at-M
aasi
m R
95
9
994
99
7
996
99
6
996
10
00
100
0 10
00
100
0 98
8
999
1
63
Sto
To
mas
10
63
106
3 10
1 9
10
29
103
0 99
8
100
0 10
00
100
0 11
10
104
1 10
22
-08
9
Siba
lom
-San
Jos
e 95
0
872
94
2
933
94
2
942
10
28
102
8 94
4
101
7
928
99
2
292
A
gana
n-St
a B
arba
ra
108
3 10
91
106
0 10
30
961
10
05
996
10
08
996
99
6
104
5 10
00
-21
1 A
vera
ge
987
98
8
981
98
1
916
97
8
100
5 10
07
988
10
25
984
10
01
123
Wet
seas
on b
enef
ited
are
a in
dex
UPR
IIS
110
0 10
07
114
4 10
55
113
8 11
74
951
10
71
971
ll
58
10
89
106
7 -0
47
Ang
at-M
aasi
m R
97
9
974
92
0
983
10
28
100
8 99
7
102
1 98
2
931
97
7
988
0
54
Sto
To
mas
11
63
115
9 11
23
107
5 10
80
103
9 98
1
978
10
41
103
9 11
20
101
6 -4
88
Siba
lom
-San
Jos
e 11
35
103
8 10
11
931
93
4
906
10
07
975
10
1S
998
10
11
981
-0
80
Aga
nan-
Sta
Bar
bara
10
85
110
5 10
74
106
8 10
01
104
2 99
8
101
5
987
61
0
106
7 93
0
-1 8
4
Ave
rage
10
92
105
6 10
54
102
4 10
36
103
4 98
7
101
3 10
00
947
10
53
996
-2
32
Dry
sea
Son
bene
fite
d ar
ea
inde
x U
PRIIS
14
04
155
0 15
S0
155
8 16
17
128
0 57
2
114
S 15
74
152
3 12
38
-09
4 A
ngat
-Maa
sim
R
903
93
0
103
2 10
61
104
2 10
69
988
99
2
102
1 99
6
993
10
13
061
S
to
Tom
as
105
7 12
27
122
5
961
99
9
115
9 10
1 7
91
0
107
3 12
1 2
10
94
107
4 -0
28
Si
balo
m-S
an J
ose
Aga
nan-
Sta
Bar
bara
66
5
95S
62
6
632
67
4
111
4
501
10
S1
412
11
51
766
93
3
856
94
3
107
0 94
8
107
4 11
09
111
6
158
6 58
8
987
97
6
110
4 5
35
083
N
w
Ave
rage
89
6
964
11
19
103
7 10
44
110
9 10
1 7
89
8
108
5 12
97
101
7 10
81
082
Wet
seas
on d
isch
arge
in
dex
UPR
IIS
132
9 72
7
142
5 11
88
120
4 98
8
105
8 10
63
879
96
4
117
5 99
1
-16
5 A
ngat
-Maa
sim
R
129
7 13
52
134
5 12
58
127
0 11
70
120
3 62
7
131
3 10
68
-18
9
Sto
To
mas
14
71
155
0 14
1 6
10
40
725
12
35
112
3 14
79
103
1 -4
48
Siba
lom
-San
Jos
e 96
8
733
11
1 5
92
2
907
47
1
102
3 85
7
ll2
0
422
92
9
779
-1
09
Aga
nan-
Sta
Bar
bara
87
9
863
68
1
925
96
8
110
7 70
5
871
87
7
00
9
Ave
rage
11
49
105
7 13
61
115
0 10
58
853
10
43
963
99
4
SO3
11
55
940
-2
85
Dry
sea
son
disc
harg
e in
dex
UPR
IIS
425
13
09
153
3 14
28
180
6 14
S0
125
8 56
3
117
9 14
66
130
0 11
89
-04
3
Ang
at-M
aasi
m R
12
26
161
6 15
30
139
2 12
85
100
8 10
33
958
14
41
107
1 -3
80
S
to
Tom
as
116
4 14
82
155
7 79
9
971
12
30
126
5 14
01
106
6 -2
44
Si
balo
m-S
an J
ose
486
71
5
782
62
9
789
64
5
116
9 11
86
801
65
3
918
2
36
Aga
nan-
Sta
Bar
bara
A
vera
ge
425
10
46
133
6 62
0
118
4 81
5
116
1 94
7
112
5 89
7
922
12
20
991
88
2
108
7 12
36
119
2 71
8
114
0 10
37
105
5 3
23
-07
5
a bull
indi
cate
s si
gnif
ican
ce a
t 95
per
cent
con
fide
nce
- 24 shy
Table 5--Results of pooled regressions yield indicators
------------------------- Dependent variables -------------------------shyIndependent variable ---------- Yield MTHa ----------- -- Specific Yield 1000 KgM
3 H20 shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season -shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 1967 (286)
3747 (580)
1991 (293)
2800 (460 )
0084 (041)
0441 (229)
0049 (023 )
0876 (343)
System 2 indicator 0400 (155)
0525 (226)
0057 (085)
0115 (1 95)
System 3 indicator 0466 (139 )
-0029 (-010)
-0016 (-018)
0006 (008)
System 4 indicator 0980 (470)
0090 (049)
0176 (333)
0391 (827)
System 5 indicator 1 221 (570)
0256 ( 138)
0104 (174)
0224 (440)
Reservoir indicator -1033 (-578)
-0063 (-039)
-0146 (-310)
-0326 (-569)
Period indicatora
Precipitation in mm Wet season
Annual
0163 (1 03)
-29E-4 (-182)
0231 (1 42)
-58E-4 (-606)
0161 (118)
92E-5 (072)
0199 043 )
-64pound-5 (-084)
0082 (200)
57E-6 (013)
0092 (223)
-61E-5 (-232)
0032 (093)
29E-7 (001)
0052 (112)
-12pound-4 (-414)
Nitrogen applicationb 0025 (254)
0021 (211)
0025 (274)
0020 (227)
0003 (1 05)
0002 (043)
0003 (101)
-0002 (-052)
R2 (adj) 0553 0516 0251 0198 0309 0262 0713 0445 number of observations 50 50 49 49 42 42 40 40 degrees of freedom 42 45 41 44 34 37 32 35
CoeffiCient (t statistic) indicates significance at 95 confidence ~ 1982-1986 = 1
kgha
- 25 shy
Table 6--Suggmry of estiates perforance indicators
Variable Predicted Predicted Period Dependent Variable
Description (units)
Period Coefficient
1976-86 IndicatorO
1976-86 Indicatorl
Effecta (X)
BenefittedService area ratio
Wet season ratio -0036 0853 0817 -425 Dry Season 0071 0539 0610 1313
Delivery Wet Dry
season season
nm water per benefitted ha
-0808 -1 214
6003 9083
5195 7869
-1346 -1337
Rice Yield Wet season metric tons 0163 3578 3741 456 Dry season per hectare 0161 3905 4066 412
Specific Yield Wet season kg rice per m3 0082 0348 0430 2353 Dry season water de 1 ivered 0032 0378 0409 837
a percentage changes relative to period=O values
- 26 shy
ENDNOTES
IResearch Fellow International Food Policy Research Institute and Senior Irrigation Specialist International Irrigation Management Institute
2Presumable farmers also base private investment decisions on such foreknowledge of costs and commonly do so in the case of small tubewells In addition user groups may invest in small communal schemes independently after examining projected costs and benefits Most commonly however new irrigation capacity is created as a result of public investment decisions made by government and thus the political process comes into play
3This principle applies only when water charges are levied volumetrically--an exceedingly rare situation in the developing worldshy-though the argument is frequently invoked as though it applies to all allocative situations
4For Cabanatuan City the longer period of record available ie 1904 to 1986 was divided into two equal periods and tested for difference in means to test the observation by farmers that rainfall today is lower than it used to be Mean precipitation after 1947 was actually about 3 percent higher than before but the difference was not significant (t = 07)
5The procedure used was to take time series data on total nitrogenfertilizer consumption in the Philippines multiply it by the ratio of N fertilizer used on rice to total N consumption and divide the product by the rice harvested area to obtain an estimate of N use perhectare Values of the ratio of N use on rice to total N use were taken from IRRI (1988 150) with interpolation used to fill in gaps in the series
6This assumes that timeliness is evaluated against a standard based on the physiological demand for water generated by the rice crop and the varying sensitivity of the crop to stress at different growth stages In such a case a poor timeliness rating for irrigationservice will result in reductions in yield
7Because rice yields are insensitive to water applications that exceed PET values a redistribution of water from water excess to water short portions of the command area will usually result in higher aggregate output and higher average yields Exceptional scenarios can be constructed in which this general rule would be violated but such scenarios are unlikely to play out in practice
- 27 shy
aThere are a variety of standards for judging equity that could be used here The most common one is equality of per hectare water deliveries In the present case the implicit standard is equality of water supply utility in producing a crop Thus if two areas received equal amounts of water on a per hectare basis but one had such high seepage and percolation losses that the crop failed to produce a harvest (and was therefore exempted from irrigation fee payment) the water supply to the combined area would be judged inequitable Because such radical differences in seepage and percolation rates are unlikely on puddled rice soils in the lowland Philippines and because the definition of benefitted area accommodates a wide range of acceptable yields these two definitions are largely indistinguishable at the current limits of data resolution
- 13 shy
Since the condition of constancy of water supply is not generally satisfied a regression was run in which the total quantityof water diverted at the headworks of each system divided by the systems potential service area SA was included in the regression to control for changes in the water supply available to the system seeTable 3 equations 5 through 8 The average daily rainfall received directly on the system service area during the season was also included as an independent variable The regression was run separately for wet and dry seasons The sign and t-statistic of the period dummy should then tell us whether or not equity as reflected in the BASA ratio increased decreased or remained unchanged across the period divide
Both equations are reasonable good as indicated by the R2 values though the dry season equation is considerably better as would be expected For the wet season both the water delivery term and the rainfall term in equation 5 are of positive sign but are nonshysignificant at the 95 percent confidence level indicating that wet season irrigated area does not change appreciably in response to level of wet season rainfall or the available irrigation water supply The period dummy was negative but not significant indicating that equityof distribution as reflected in the BASA ratio was similar during the two periods
For the dry season the water delivery term in equation 7 is positive and strongly significant indicating a close relationship between the fraction of potential area actually irrigated and the water supply available at the headworks In addition however the period dummy is positive and significant suggesting that once the influence of water supply is removed the BASA ratio was significantly higher in the period following 1981 than it was before
This is an important finding for it reflects significantlyimproved performance in terms of a factor equity of water distribution that is under the control of the managing entity an entity which here comprises both NIA and irrigators associations Interpreted in these terms NIA and allied farmers associations were able to spread a given amount of water more widely across the potential command area of the five sample systems in the period after 1981 than before Moreover they did this in a way that did not decrease average system yields as discussed earlier In making this interpretation we are suggesting that there was some redistribution of water from better-watered areas to fringe areas which would otherwise not have received irrigation water and that this redistribution was a direct response to the change in NIA prioritiesand operating policies and rules occurring around 1981
It is difficult to prove the assertion that water was in fact redistributed with a resulting increase in directly-measured equity Without access to reliable discharge data broken out by systemsection and we can only assume in the absence of a plausible
- 14 shy
alternative explanation that it was such a redistribution that made the increase in the BASA ratio possible In a larger sense it is difficult to prove conclusively that any outcome in a before and after analysis was the result of a particular independent causative factor In this case we have tried to remove the influence of other potential causative factors where we could but the possibilityremains that some combination of unmeasured factors are responsiblefor the difference in the BASA ratio found We do note though that this type of response is exactly the type that would be expected to follow from an emphasis on increased farmer satisfaction and cooperation and increased fee revenues Because the fee schedule is tied to benefitted area the only ways NIA can increase its revenue from that source are to expand benefitted area and to increase collection efficiencies The former depends on redistributing a fixed supply of water over a larger portion of the command while the latter requires that farmers be satisfied with the irrigation service they are receiving and the commitment of the local irrigators association to assist in the task of collecting the amounts due The evidence while not conclusive is highly suggestive that this is exactly what has happened
Efficiency In addition to measures which reflect the levels of adequacy and equity of irrigation service available data allow the calculation of a measure of operating efficiency The term efficiencyusually denotes the relationship between inputs to a process and its outputs often expressed as a ratio The output measure employed here is aggregate system rice output and the input is quantity of irrigation water turned into the system Dividing the first by the second gives a measure of agricultural production per unit water--here termed specific yield This is a highly integrated measure that evaluates the combined efficiency of the irrigation and agricultural processes As such it is a function of the managerial and other inputs supplied both to the irrigation system and to the agricultural operation With respect to one important input to the irrigation system we do know that NIA per hectare field operating expenses were about 29 percent lower in real terms in the 1982-86 period comparedto the 1976-1981 period although this drop may have been partlyoffset by increases in farmer-supplied labor inputs Other things being equal one would thus expect to find a decline in output efficiency
The regression analysis shows positive signs for the period terms in both wet and dry season equations (see Table 5) In the case of the wet season the period dummy in equation 5 is significant but the overall explanatory power of the model is quite low For the dry season (equation 7) the coefficient is positive but non-significant This means that after taking rainfall and fertilizer use into account data do not indicate a lowering of specific yield in the wake of funding reductions and the strong emphasiS on financial viabilitybeginning in 1981 This result provides evidence that the efficiency of the overall irrigation deliveryagricultural production process
- 15 shy
relative to the system water input did not falloff as a result of the changes implemented at least over the short run
Impact magnitude
The preceding analysis has shown us that some indicators of irrigation performance changed significantly following the managerial changes of 1981 while others did not However it has not given us a sense of the size of the changes which occurred To determine the magnitude of these changes the regression model is used to predict the response of the composite system to the managerial changes given a common set of -input and environmental factors To do this averagevalues of the independent variables from the entire eleven-year period 1976 to 1986 are put into the model together with the previously determined coefficients to generate predicted average values of the various dependent variables used in the earlier analysis with and without the period dummy This procedure produces a pair of estimates for each dependent variable under the same conditions--one in which the system responds as it did after the managerial changes were implemented and one in which it responds as it did prior to their introduction The differences between these two values thus indicate the magnitude of the changes occurring in the various indicators of performance discussed above
The results of this exercise are shown in Table 6 The table shows that water availability decreased by about 13 percent in both wet and dry seasons when the period dummy was included and while the coefficients responsible were significant in the earlier analysisthis difference cannot be easily connected with levels of system management as discussed earlier With respect to rice output per hectare although the coefficients were not very significant it is interesting to note that yield increases by 163 kilograms per hectare for the wet season and by 101 kilogram for the dry when the period dummy is included in spite of the reduced water supply available Keep in mind that the predicted yield values have already been adjusted for differences in nitrogen fertilizer use and rainfall This suggests that timeliness and equity of distribution of water supply to farmers may have increased following the changes contributing to the higher predicted yields
Examining the impact of increased equity of distribution bylooking at the ratio of benefitted area to service area we recall that the change was positive and significant for the dry season and negative and not significant for the wet Table 6 shows that the dry season BASA ratio increases by 7 percentage points when the dummy is included a 131 percent increase Other things being equal this should result in a 131 percent increase in system output due to the expansion of ared benefitted This is a major impact on production
- 16 shy
CONCLUSIONS
The Philippine experiment to transform the national irrigation agency into an enterprise has undoubtedly been successful in reducing system operating expenses bringing revenues and costs into line and eliminating the recurrent cost burden imposed by large-scaleirrigation systems on the national budget Evidence presented in this paper indicates that in the process equity of water distribution across systems has also improved In the 5 years following the cessation of operating subsidies from the government an index of equity of distribution improved by about 13 percent At the same time per hectare yields adjusted for rainfall and nitrogen application held constant
There is a strong logical connection between the achievement of financial viability and improved equity of water distribution across the command Because increasing irrigation fees is a politicaldecisionlying largely beyond NIAs control expanding the area which can be billed for service is one of the few revenue increasing measures available to the irrigation agency which does not involve major additional investment In the face of constant or shrinkingwater supplies this is achieved only by redistributing water from areas receiving excessive supplies usually near the head ends of canals and laterals to areas receiving no supplies or inadequatesupplies often located near the tails of canals Although data are not available which would allow the direct examination of this hypotheses the two outcomes are logically consistent with each other
Data also show that per hectare water deliveries declined significantly in the five sample systems after 1981 even thoughrainfall did not differ appreciably between the two periods This decline averaged about 13 percent for both wet and dry seasons and is interpreted as a decline in water availability in the supplying rivers rather than a conscious reduction in withdrawals by system managers Such declines could result from changes in watershed runoff characteristics as caused by deforestation or from increased upstream abstractions from supplying rivers
Improved water distribution tends to increase the area served system agricultural output and NIA service fee revenue Reduced water supplies to the system tend to reduce these things Specificyield defined as system paddy output per unit water held roughly constant across the two periods indicating that the two effects mayhave offset each other
After adjusting for rainfall and nitrogen application perhectare yields increased only marginally in the post-1981 period Area served on the other hand increased by about 13 percent after adjusting for water supply availability indicating that the area benefitted by irrigation in the sample systems increased by about the same percentage Even if yields on this additional area are less than
- 17 shy
average yields for the system this still represents a sizeable increase in system agricultural output as a result of the change in management structure the increase coming not from higher yields but from expanded area under irrigation
The evidence assembled here suggests that there are significantfinancial and economic benefits to be had from changes in the basic character of irrigation managing agencies which make them more responsive to their clientele and which impose rational internal financial discipline on the agency The analysis suggests a number of additional questions however One relates to the longer-term impacts of the structural management changes The improvements in water distribution described here are relatively short-term events occurring during the first 5 years of the new management mode Critics have suggested the danger of underinvestment in systemmaintenance over the longer run accompanied by declining yields and benefitted areas and eventual system collapse This possibility needs to be closely monitored A second concern relates to the apparent decline in water supply to these 5 geographically dispersed systems The nature and causes of this decline need to be explored further since if widespread and secular it may represent a serious threat to the stability of Philippine rice production Whether stemming from poor forest management practices or deficient regulation and allocation of surface water resources or other unidentified factors it is an issue that deserves serious and urgent consideration
A third risk is that the incentive structure set up by NIA to guide and stimulate the performance of field units overemphasizes revenue generation at the expense of irrigation service provision to farmers The evidence presented here supports the view that these two objectives are mutually reinforcing under policies and conditions which have been established in the Philippines More detailed crossshysectional studies based on primary flow measurement data would add confidence to this conclusion and help to specify the conditions under which this effect occurs This could be extremely important in transferring the results of the Philippine experiment to other countries
A final risk is that outside intervention well meaning or otherwise will destroy the basis of NIAs financial autonomy or will impose external pressures or constraints on NIAs decision-making that will subvert the management practices which have been so painstakinglydeveloped and implemented Among these are calls for NIA to be subsumed again within the government department structure in the interests of better coordination with agriculture attempts byexternal financing agencies to arbitrarily increase NIAs expenditures on OampM on the assumption that this will increase system agricultural output or intervention by Philippine legislative bodies to restore operating subsidies to NIA with attached strings leading back to legislators home districts Pressures such as these will cut short a process of experimentation and improvement that seems promising
- 18 shy
enough to date to warrant its continuation Having developed the capacity to establish targets and implement and manage change NIA is in a strong position to modify its objectives to better achieve larger social purposes established for it It is critical to recognize however that this must happen within the context of financing policies that mandate financial autonomy for NIA if the fundamental institutional commitment to manage is to be preserved
The author would like to thank Leslie Small and JeremyBerkoff for helpful comments on an earlier unpublishedversion of this paper and Charles Rogers for his careful and creative help with the analysis
- 19 shy
BIBLIOGRAPHY
Abernethy Charles L 1990 Indicators of the performance of irrigation water distribution systems International Irrigation Management Institute Colombo Sri Lanka Mimeo
Asian Development Bank 1986 Irrigation service fees Proceedingsof the Regional Seminar on Irrigation Service Fees Manila Asian Development Bank
Carruthers Ian and Colin Clark 1981 Economics of IrrigationLiverpool Liverpool University Press Third Edition
Levine G and EW Coward Jr 1986 Irrigation water distribution implications for design and operation AGREP Division WorkingPaper 125 vol 1 World Bank Agriculture and Rural Development Department
Small Les E 1989 User charges in irrigation potentials and limitations Irrigation and drainage vol 3 no 2125-142
Small Les 1990 Irrigation service fees in Asia IrrigationManagement Network 9013 London Overseas DevelopmentInstitute
Svendsen Mark and Les Small 1989 A framework for assessing irrigation system performance Paper prepared for the Symposium on Performance Evaluation 23 November International IrrigationManagement Institute Sri Lanka
Table I--National Irrigation Administration revenues and expenditures in constant prices 1976-86
Item 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986
(peso million 1972)
Revenues Irrigation fees collected 1273 1483 17 13 1831 2070 1668 1699 1893 1728 2129 2546 Other income 715 737 2420 5591 2631 5990 7783 6638 5699 4932 2934
Total direct revenue 1988 2220 4133 7422 4701 7658 9482 8531 7427 061 5480
Expenses in 1972 pricesTotal expenses 4825 5716 5039 6329 3821 77 55 6166 4749 4348 4259 4959
Excess (deficit) (2837(3496) (906) 1093 877 (097) 3316 3782 3079 2802 521 N 0
Subsidies Government operation and
maintenance subsidies 2521 2741 2799 1817 1398 633 0 0 0 0 0 Calamity fund payments 548 0 0 0 0 0 0 0 119 0 142
Total subsidy 3069 2741 2799 1817 1398 633 0 0 119 0 142
Total excess (deficit) 231 (754) 1893 2910 2275 536 3316 3782 3198 2802 663
Source IFPRI analysis of NIA data
- 21 shy
Table 2--Descriptive characteristics of selected MIA systeasa
------- Region III -------- ----- Region VI ----shyUPRIIS Angatii Sto Sibalom- Aganan-
Haasim Thomasc San Jose Sta Barbara
Average service area (hal 102272 31462 3522 5282 8703 Average irrigated area (hal
Wet season 83768 23454 3007 4410 8300 Dry season
Average benefited area Wet season
(hal 64587
77 605
27639
22908
1 781
3007
2801
4369
2770
7698 Dry season
Average rainfall Wet season
(mml d 62478
1 685 5
27396
8576
1 781
3051 0
2769
24731
2997
20001 Dry season 756 333 322 2828 3025
Average discharge (Llsec) Wet season 46501 14792 1692 2353 4984 Dry season 78091 22812 2014 1276 2315
Average water delivery (mmday) Wet season 522 548 487 462 571 Dry season
Average yield (mtha) 1089 715 995 398 686
Wet season 345 419 322 395 435 Dry season
Avg yield per unit water 34 03
(kgm ) 451 412 399 426
Wet season 0373 0440 0373 0538 0443 Dry season 0248 0400 0279 0690 0428
t-statistic difference in mean rainfall 1978-81 1982-86e
Wet season 0432 0713 -0567 1169 1169 Dry season 0519 -0230 -0523 1187 1187 Annual 0460 0707 -0686 1445 1445
~ Summary numbers are averages for the period 1982-1986 except as noted Water delivery discharge and yield per unit discharge are 4-year averages 1982-1985
c Water delivery discharge and yield per unit discharge are 4-year averages d 1983-1986
For Angat 5 years are 1981-85 For St Thomas 1979-83 For Sibalom 1971-75 e No significant differences at 95 confidence
- 22 shy
Table 3--Results of pooled regressions water delivery indicators
Dependent variable Independent variable - Water Delivery in mmBenefited Ha - --- BenefitService Area Ratio --shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season-shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 6799 (998)
6189 (1019)
13102 (1093)
2411 (1 54)
0743 ( 1186)
0800 (1165)
0129 (131 )
0258 (329)
System 2 indicator 0173 (029)
-3743 (-363)
-0063 (-219)
0200 (505)
System 3 indicator 0698 (085)
0163 (012)
0085 (208)
-0046 (-097)
System 4 indicator -0514 (-l 07)
-7011 (-803)
0080 (340)
0169 (l87)
System 5 indicator 0177 (033)
-5049 (-538)
0174 (603)
0051 (058)
Reservoir indicator 0327 (078)
5703 (543)
-0126 (-510)
0156 (261)
Period indicatora -0808 (-240)
-0771 (-229)
-1214 (-209)
-1309 (-1 62)
-0036 (-192)
-0025 (-112)
0071 (239)
0030 (077)
Water delivery mmb
Wet season
Dry season
Precipitation in mm Wet season
Annual
Average daily rainfall Wet season
Dry season
-46E-4 (-125)
-1 9E-4 (-082)
-43E-4 (-077)
20E-3 (385)
0015 (1 59)
-0001 ( -004)
0023 (215)
0001 (021)
0070 (613)
-0017 (-058)
0046 (371)
0010 (033)
R2 (adj) number of observations degrees of freedom
0156 43 36
0133 43 39
0711 41 34
0434 41 37
0693 42 34
0556 42 37
0865 40 32
0730 40 35
Coefficient (t statistic) indicates significance at 95 confidence ~ 1982-1986 1
per ha of service area
Ta
ble
--I
nd
ices
of
se
v
ice a
rea
ib
en
ef
ited
a
rea
a
nd
a
vera
ge sea
so
na
l d
isch
arg
e
Ave
rage
A
vera
ge
tshy19
77
1978
19
79
1980
19
81
1982
19
83
1984
19
85
1986
19
77-8
1 19
82-8
6 S
tati
stic
a
(ind
ex
aver
age
1983
-198
5 =
100)
Ser
vice
are
a UP
RI IS
in
dex
882
91
9
920
91
5
951
95
1
100
0 10
00
100
0 10
00
917
99
0
bull5
53
Ang
at-M
aasi
m R
95
9
994
99
7
996
99
6
996
10
00
100
0 10
00
100
0 98
8
999
1
63
Sto
To
mas
10
63
106
3 10
1 9
10
29
103
0 99
8
100
0 10
00
100
0 11
10
104
1 10
22
-08
9
Siba
lom
-San
Jos
e 95
0
872
94
2
933
94
2
942
10
28
102
8 94
4
101
7
928
99
2
292
A
gana
n-St
a B
arba
ra
108
3 10
91
106
0 10
30
961
10
05
996
10
08
996
99
6
104
5 10
00
-21
1 A
vera
ge
987
98
8
981
98
1
916
97
8
100
5 10
07
988
10
25
984
10
01
123
Wet
seas
on b
enef
ited
are
a in
dex
UPR
IIS
110
0 10
07
114
4 10
55
113
8 11
74
951
10
71
971
ll
58
10
89
106
7 -0
47
Ang
at-M
aasi
m R
97
9
974
92
0
983
10
28
100
8 99
7
102
1 98
2
931
97
7
988
0
54
Sto
To
mas
11
63
115
9 11
23
107
5 10
80
103
9 98
1
978
10
41
103
9 11
20
101
6 -4
88
Siba
lom
-San
Jos
e 11
35
103
8 10
11
931
93
4
906
10
07
975
10
1S
998
10
11
981
-0
80
Aga
nan-
Sta
Bar
bara
10
85
110
5 10
74
106
8 10
01
104
2 99
8
101
5
987
61
0
106
7 93
0
-1 8
4
Ave
rage
10
92
105
6 10
54
102
4 10
36
103
4 98
7
101
3 10
00
947
10
53
996
-2
32
Dry
sea
Son
bene
fite
d ar
ea
inde
x U
PRIIS
14
04
155
0 15
S0
155
8 16
17
128
0 57
2
114
S 15
74
152
3 12
38
-09
4 A
ngat
-Maa
sim
R
903
93
0
103
2 10
61
104
2 10
69
988
99
2
102
1 99
6
993
10
13
061
S
to
Tom
as
105
7 12
27
122
5
961
99
9
115
9 10
1 7
91
0
107
3 12
1 2
10
94
107
4 -0
28
Si
balo
m-S
an J
ose
Aga
nan-
Sta
Bar
bara
66
5
95S
62
6
632
67
4
111
4
501
10
S1
412
11
51
766
93
3
856
94
3
107
0 94
8
107
4 11
09
111
6
158
6 58
8
987
97
6
110
4 5
35
083
N
w
Ave
rage
89
6
964
11
19
103
7 10
44
110
9 10
1 7
89
8
108
5 12
97
101
7 10
81
082
Wet
seas
on d
isch
arge
in
dex
UPR
IIS
132
9 72
7
142
5 11
88
120
4 98
8
105
8 10
63
879
96
4
117
5 99
1
-16
5 A
ngat
-Maa
sim
R
129
7 13
52
134
5 12
58
127
0 11
70
120
3 62
7
131
3 10
68
-18
9
Sto
To
mas
14
71
155
0 14
1 6
10
40
725
12
35
112
3 14
79
103
1 -4
48
Siba
lom
-San
Jos
e 96
8
733
11
1 5
92
2
907
47
1
102
3 85
7
ll2
0
422
92
9
779
-1
09
Aga
nan-
Sta
Bar
bara
87
9
863
68
1
925
96
8
110
7 70
5
871
87
7
00
9
Ave
rage
11
49
105
7 13
61
115
0 10
58
853
10
43
963
99
4
SO3
11
55
940
-2
85
Dry
sea
son
disc
harg
e in
dex
UPR
IIS
425
13
09
153
3 14
28
180
6 14
S0
125
8 56
3
117
9 14
66
130
0 11
89
-04
3
Ang
at-M
aasi
m R
12
26
161
6 15
30
139
2 12
85
100
8 10
33
958
14
41
107
1 -3
80
S
to
Tom
as
116
4 14
82
155
7 79
9
971
12
30
126
5 14
01
106
6 -2
44
Si
balo
m-S
an J
ose
486
71
5
782
62
9
789
64
5
116
9 11
86
801
65
3
918
2
36
Aga
nan-
Sta
Bar
bara
A
vera
ge
425
10
46
133
6 62
0
118
4 81
5
116
1 94
7
112
5 89
7
922
12
20
991
88
2
108
7 12
36
119
2 71
8
114
0 10
37
105
5 3
23
-07
5
a bull
indi
cate
s si
gnif
ican
ce a
t 95
per
cent
con
fide
nce
- 24 shy
Table 5--Results of pooled regressions yield indicators
------------------------- Dependent variables -------------------------shyIndependent variable ---------- Yield MTHa ----------- -- Specific Yield 1000 KgM
3 H20 shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season -shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 1967 (286)
3747 (580)
1991 (293)
2800 (460 )
0084 (041)
0441 (229)
0049 (023 )
0876 (343)
System 2 indicator 0400 (155)
0525 (226)
0057 (085)
0115 (1 95)
System 3 indicator 0466 (139 )
-0029 (-010)
-0016 (-018)
0006 (008)
System 4 indicator 0980 (470)
0090 (049)
0176 (333)
0391 (827)
System 5 indicator 1 221 (570)
0256 ( 138)
0104 (174)
0224 (440)
Reservoir indicator -1033 (-578)
-0063 (-039)
-0146 (-310)
-0326 (-569)
Period indicatora
Precipitation in mm Wet season
Annual
0163 (1 03)
-29E-4 (-182)
0231 (1 42)
-58E-4 (-606)
0161 (118)
92E-5 (072)
0199 043 )
-64pound-5 (-084)
0082 (200)
57E-6 (013)
0092 (223)
-61E-5 (-232)
0032 (093)
29E-7 (001)
0052 (112)
-12pound-4 (-414)
Nitrogen applicationb 0025 (254)
0021 (211)
0025 (274)
0020 (227)
0003 (1 05)
0002 (043)
0003 (101)
-0002 (-052)
R2 (adj) 0553 0516 0251 0198 0309 0262 0713 0445 number of observations 50 50 49 49 42 42 40 40 degrees of freedom 42 45 41 44 34 37 32 35
CoeffiCient (t statistic) indicates significance at 95 confidence ~ 1982-1986 = 1
kgha
- 25 shy
Table 6--Suggmry of estiates perforance indicators
Variable Predicted Predicted Period Dependent Variable
Description (units)
Period Coefficient
1976-86 IndicatorO
1976-86 Indicatorl
Effecta (X)
BenefittedService area ratio
Wet season ratio -0036 0853 0817 -425 Dry Season 0071 0539 0610 1313
Delivery Wet Dry
season season
nm water per benefitted ha
-0808 -1 214
6003 9083
5195 7869
-1346 -1337
Rice Yield Wet season metric tons 0163 3578 3741 456 Dry season per hectare 0161 3905 4066 412
Specific Yield Wet season kg rice per m3 0082 0348 0430 2353 Dry season water de 1 ivered 0032 0378 0409 837
a percentage changes relative to period=O values
- 26 shy
ENDNOTES
IResearch Fellow International Food Policy Research Institute and Senior Irrigation Specialist International Irrigation Management Institute
2Presumable farmers also base private investment decisions on such foreknowledge of costs and commonly do so in the case of small tubewells In addition user groups may invest in small communal schemes independently after examining projected costs and benefits Most commonly however new irrigation capacity is created as a result of public investment decisions made by government and thus the political process comes into play
3This principle applies only when water charges are levied volumetrically--an exceedingly rare situation in the developing worldshy-though the argument is frequently invoked as though it applies to all allocative situations
4For Cabanatuan City the longer period of record available ie 1904 to 1986 was divided into two equal periods and tested for difference in means to test the observation by farmers that rainfall today is lower than it used to be Mean precipitation after 1947 was actually about 3 percent higher than before but the difference was not significant (t = 07)
5The procedure used was to take time series data on total nitrogenfertilizer consumption in the Philippines multiply it by the ratio of N fertilizer used on rice to total N consumption and divide the product by the rice harvested area to obtain an estimate of N use perhectare Values of the ratio of N use on rice to total N use were taken from IRRI (1988 150) with interpolation used to fill in gaps in the series
6This assumes that timeliness is evaluated against a standard based on the physiological demand for water generated by the rice crop and the varying sensitivity of the crop to stress at different growth stages In such a case a poor timeliness rating for irrigationservice will result in reductions in yield
7Because rice yields are insensitive to water applications that exceed PET values a redistribution of water from water excess to water short portions of the command area will usually result in higher aggregate output and higher average yields Exceptional scenarios can be constructed in which this general rule would be violated but such scenarios are unlikely to play out in practice
- 27 shy
aThere are a variety of standards for judging equity that could be used here The most common one is equality of per hectare water deliveries In the present case the implicit standard is equality of water supply utility in producing a crop Thus if two areas received equal amounts of water on a per hectare basis but one had such high seepage and percolation losses that the crop failed to produce a harvest (and was therefore exempted from irrigation fee payment) the water supply to the combined area would be judged inequitable Because such radical differences in seepage and percolation rates are unlikely on puddled rice soils in the lowland Philippines and because the definition of benefitted area accommodates a wide range of acceptable yields these two definitions are largely indistinguishable at the current limits of data resolution
- 14 shy
alternative explanation that it was such a redistribution that made the increase in the BASA ratio possible In a larger sense it is difficult to prove conclusively that any outcome in a before and after analysis was the result of a particular independent causative factor In this case we have tried to remove the influence of other potential causative factors where we could but the possibilityremains that some combination of unmeasured factors are responsiblefor the difference in the BASA ratio found We do note though that this type of response is exactly the type that would be expected to follow from an emphasis on increased farmer satisfaction and cooperation and increased fee revenues Because the fee schedule is tied to benefitted area the only ways NIA can increase its revenue from that source are to expand benefitted area and to increase collection efficiencies The former depends on redistributing a fixed supply of water over a larger portion of the command while the latter requires that farmers be satisfied with the irrigation service they are receiving and the commitment of the local irrigators association to assist in the task of collecting the amounts due The evidence while not conclusive is highly suggestive that this is exactly what has happened
Efficiency In addition to measures which reflect the levels of adequacy and equity of irrigation service available data allow the calculation of a measure of operating efficiency The term efficiencyusually denotes the relationship between inputs to a process and its outputs often expressed as a ratio The output measure employed here is aggregate system rice output and the input is quantity of irrigation water turned into the system Dividing the first by the second gives a measure of agricultural production per unit water--here termed specific yield This is a highly integrated measure that evaluates the combined efficiency of the irrigation and agricultural processes As such it is a function of the managerial and other inputs supplied both to the irrigation system and to the agricultural operation With respect to one important input to the irrigation system we do know that NIA per hectare field operating expenses were about 29 percent lower in real terms in the 1982-86 period comparedto the 1976-1981 period although this drop may have been partlyoffset by increases in farmer-supplied labor inputs Other things being equal one would thus expect to find a decline in output efficiency
The regression analysis shows positive signs for the period terms in both wet and dry season equations (see Table 5) In the case of the wet season the period dummy in equation 5 is significant but the overall explanatory power of the model is quite low For the dry season (equation 7) the coefficient is positive but non-significant This means that after taking rainfall and fertilizer use into account data do not indicate a lowering of specific yield in the wake of funding reductions and the strong emphasiS on financial viabilitybeginning in 1981 This result provides evidence that the efficiency of the overall irrigation deliveryagricultural production process
- 15 shy
relative to the system water input did not falloff as a result of the changes implemented at least over the short run
Impact magnitude
The preceding analysis has shown us that some indicators of irrigation performance changed significantly following the managerial changes of 1981 while others did not However it has not given us a sense of the size of the changes which occurred To determine the magnitude of these changes the regression model is used to predict the response of the composite system to the managerial changes given a common set of -input and environmental factors To do this averagevalues of the independent variables from the entire eleven-year period 1976 to 1986 are put into the model together with the previously determined coefficients to generate predicted average values of the various dependent variables used in the earlier analysis with and without the period dummy This procedure produces a pair of estimates for each dependent variable under the same conditions--one in which the system responds as it did after the managerial changes were implemented and one in which it responds as it did prior to their introduction The differences between these two values thus indicate the magnitude of the changes occurring in the various indicators of performance discussed above
The results of this exercise are shown in Table 6 The table shows that water availability decreased by about 13 percent in both wet and dry seasons when the period dummy was included and while the coefficients responsible were significant in the earlier analysisthis difference cannot be easily connected with levels of system management as discussed earlier With respect to rice output per hectare although the coefficients were not very significant it is interesting to note that yield increases by 163 kilograms per hectare for the wet season and by 101 kilogram for the dry when the period dummy is included in spite of the reduced water supply available Keep in mind that the predicted yield values have already been adjusted for differences in nitrogen fertilizer use and rainfall This suggests that timeliness and equity of distribution of water supply to farmers may have increased following the changes contributing to the higher predicted yields
Examining the impact of increased equity of distribution bylooking at the ratio of benefitted area to service area we recall that the change was positive and significant for the dry season and negative and not significant for the wet Table 6 shows that the dry season BASA ratio increases by 7 percentage points when the dummy is included a 131 percent increase Other things being equal this should result in a 131 percent increase in system output due to the expansion of ared benefitted This is a major impact on production
- 16 shy
CONCLUSIONS
The Philippine experiment to transform the national irrigation agency into an enterprise has undoubtedly been successful in reducing system operating expenses bringing revenues and costs into line and eliminating the recurrent cost burden imposed by large-scaleirrigation systems on the national budget Evidence presented in this paper indicates that in the process equity of water distribution across systems has also improved In the 5 years following the cessation of operating subsidies from the government an index of equity of distribution improved by about 13 percent At the same time per hectare yields adjusted for rainfall and nitrogen application held constant
There is a strong logical connection between the achievement of financial viability and improved equity of water distribution across the command Because increasing irrigation fees is a politicaldecisionlying largely beyond NIAs control expanding the area which can be billed for service is one of the few revenue increasing measures available to the irrigation agency which does not involve major additional investment In the face of constant or shrinkingwater supplies this is achieved only by redistributing water from areas receiving excessive supplies usually near the head ends of canals and laterals to areas receiving no supplies or inadequatesupplies often located near the tails of canals Although data are not available which would allow the direct examination of this hypotheses the two outcomes are logically consistent with each other
Data also show that per hectare water deliveries declined significantly in the five sample systems after 1981 even thoughrainfall did not differ appreciably between the two periods This decline averaged about 13 percent for both wet and dry seasons and is interpreted as a decline in water availability in the supplying rivers rather than a conscious reduction in withdrawals by system managers Such declines could result from changes in watershed runoff characteristics as caused by deforestation or from increased upstream abstractions from supplying rivers
Improved water distribution tends to increase the area served system agricultural output and NIA service fee revenue Reduced water supplies to the system tend to reduce these things Specificyield defined as system paddy output per unit water held roughly constant across the two periods indicating that the two effects mayhave offset each other
After adjusting for rainfall and nitrogen application perhectare yields increased only marginally in the post-1981 period Area served on the other hand increased by about 13 percent after adjusting for water supply availability indicating that the area benefitted by irrigation in the sample systems increased by about the same percentage Even if yields on this additional area are less than
- 17 shy
average yields for the system this still represents a sizeable increase in system agricultural output as a result of the change in management structure the increase coming not from higher yields but from expanded area under irrigation
The evidence assembled here suggests that there are significantfinancial and economic benefits to be had from changes in the basic character of irrigation managing agencies which make them more responsive to their clientele and which impose rational internal financial discipline on the agency The analysis suggests a number of additional questions however One relates to the longer-term impacts of the structural management changes The improvements in water distribution described here are relatively short-term events occurring during the first 5 years of the new management mode Critics have suggested the danger of underinvestment in systemmaintenance over the longer run accompanied by declining yields and benefitted areas and eventual system collapse This possibility needs to be closely monitored A second concern relates to the apparent decline in water supply to these 5 geographically dispersed systems The nature and causes of this decline need to be explored further since if widespread and secular it may represent a serious threat to the stability of Philippine rice production Whether stemming from poor forest management practices or deficient regulation and allocation of surface water resources or other unidentified factors it is an issue that deserves serious and urgent consideration
A third risk is that the incentive structure set up by NIA to guide and stimulate the performance of field units overemphasizes revenue generation at the expense of irrigation service provision to farmers The evidence presented here supports the view that these two objectives are mutually reinforcing under policies and conditions which have been established in the Philippines More detailed crossshysectional studies based on primary flow measurement data would add confidence to this conclusion and help to specify the conditions under which this effect occurs This could be extremely important in transferring the results of the Philippine experiment to other countries
A final risk is that outside intervention well meaning or otherwise will destroy the basis of NIAs financial autonomy or will impose external pressures or constraints on NIAs decision-making that will subvert the management practices which have been so painstakinglydeveloped and implemented Among these are calls for NIA to be subsumed again within the government department structure in the interests of better coordination with agriculture attempts byexternal financing agencies to arbitrarily increase NIAs expenditures on OampM on the assumption that this will increase system agricultural output or intervention by Philippine legislative bodies to restore operating subsidies to NIA with attached strings leading back to legislators home districts Pressures such as these will cut short a process of experimentation and improvement that seems promising
- 18 shy
enough to date to warrant its continuation Having developed the capacity to establish targets and implement and manage change NIA is in a strong position to modify its objectives to better achieve larger social purposes established for it It is critical to recognize however that this must happen within the context of financing policies that mandate financial autonomy for NIA if the fundamental institutional commitment to manage is to be preserved
The author would like to thank Leslie Small and JeremyBerkoff for helpful comments on an earlier unpublishedversion of this paper and Charles Rogers for his careful and creative help with the analysis
- 19 shy
BIBLIOGRAPHY
Abernethy Charles L 1990 Indicators of the performance of irrigation water distribution systems International Irrigation Management Institute Colombo Sri Lanka Mimeo
Asian Development Bank 1986 Irrigation service fees Proceedingsof the Regional Seminar on Irrigation Service Fees Manila Asian Development Bank
Carruthers Ian and Colin Clark 1981 Economics of IrrigationLiverpool Liverpool University Press Third Edition
Levine G and EW Coward Jr 1986 Irrigation water distribution implications for design and operation AGREP Division WorkingPaper 125 vol 1 World Bank Agriculture and Rural Development Department
Small Les E 1989 User charges in irrigation potentials and limitations Irrigation and drainage vol 3 no 2125-142
Small Les 1990 Irrigation service fees in Asia IrrigationManagement Network 9013 London Overseas DevelopmentInstitute
Svendsen Mark and Les Small 1989 A framework for assessing irrigation system performance Paper prepared for the Symposium on Performance Evaluation 23 November International IrrigationManagement Institute Sri Lanka
Table I--National Irrigation Administration revenues and expenditures in constant prices 1976-86
Item 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986
(peso million 1972)
Revenues Irrigation fees collected 1273 1483 17 13 1831 2070 1668 1699 1893 1728 2129 2546 Other income 715 737 2420 5591 2631 5990 7783 6638 5699 4932 2934
Total direct revenue 1988 2220 4133 7422 4701 7658 9482 8531 7427 061 5480
Expenses in 1972 pricesTotal expenses 4825 5716 5039 6329 3821 77 55 6166 4749 4348 4259 4959
Excess (deficit) (2837(3496) (906) 1093 877 (097) 3316 3782 3079 2802 521 N 0
Subsidies Government operation and
maintenance subsidies 2521 2741 2799 1817 1398 633 0 0 0 0 0 Calamity fund payments 548 0 0 0 0 0 0 0 119 0 142
Total subsidy 3069 2741 2799 1817 1398 633 0 0 119 0 142
Total excess (deficit) 231 (754) 1893 2910 2275 536 3316 3782 3198 2802 663
Source IFPRI analysis of NIA data
- 21 shy
Table 2--Descriptive characteristics of selected MIA systeasa
------- Region III -------- ----- Region VI ----shyUPRIIS Angatii Sto Sibalom- Aganan-
Haasim Thomasc San Jose Sta Barbara
Average service area (hal 102272 31462 3522 5282 8703 Average irrigated area (hal
Wet season 83768 23454 3007 4410 8300 Dry season
Average benefited area Wet season
(hal 64587
77 605
27639
22908
1 781
3007
2801
4369
2770
7698 Dry season
Average rainfall Wet season
(mml d 62478
1 685 5
27396
8576
1 781
3051 0
2769
24731
2997
20001 Dry season 756 333 322 2828 3025
Average discharge (Llsec) Wet season 46501 14792 1692 2353 4984 Dry season 78091 22812 2014 1276 2315
Average water delivery (mmday) Wet season 522 548 487 462 571 Dry season
Average yield (mtha) 1089 715 995 398 686
Wet season 345 419 322 395 435 Dry season
Avg yield per unit water 34 03
(kgm ) 451 412 399 426
Wet season 0373 0440 0373 0538 0443 Dry season 0248 0400 0279 0690 0428
t-statistic difference in mean rainfall 1978-81 1982-86e
Wet season 0432 0713 -0567 1169 1169 Dry season 0519 -0230 -0523 1187 1187 Annual 0460 0707 -0686 1445 1445
~ Summary numbers are averages for the period 1982-1986 except as noted Water delivery discharge and yield per unit discharge are 4-year averages 1982-1985
c Water delivery discharge and yield per unit discharge are 4-year averages d 1983-1986
For Angat 5 years are 1981-85 For St Thomas 1979-83 For Sibalom 1971-75 e No significant differences at 95 confidence
- 22 shy
Table 3--Results of pooled regressions water delivery indicators
Dependent variable Independent variable - Water Delivery in mmBenefited Ha - --- BenefitService Area Ratio --shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season-shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 6799 (998)
6189 (1019)
13102 (1093)
2411 (1 54)
0743 ( 1186)
0800 (1165)
0129 (131 )
0258 (329)
System 2 indicator 0173 (029)
-3743 (-363)
-0063 (-219)
0200 (505)
System 3 indicator 0698 (085)
0163 (012)
0085 (208)
-0046 (-097)
System 4 indicator -0514 (-l 07)
-7011 (-803)
0080 (340)
0169 (l87)
System 5 indicator 0177 (033)
-5049 (-538)
0174 (603)
0051 (058)
Reservoir indicator 0327 (078)
5703 (543)
-0126 (-510)
0156 (261)
Period indicatora -0808 (-240)
-0771 (-229)
-1214 (-209)
-1309 (-1 62)
-0036 (-192)
-0025 (-112)
0071 (239)
0030 (077)
Water delivery mmb
Wet season
Dry season
Precipitation in mm Wet season
Annual
Average daily rainfall Wet season
Dry season
-46E-4 (-125)
-1 9E-4 (-082)
-43E-4 (-077)
20E-3 (385)
0015 (1 59)
-0001 ( -004)
0023 (215)
0001 (021)
0070 (613)
-0017 (-058)
0046 (371)
0010 (033)
R2 (adj) number of observations degrees of freedom
0156 43 36
0133 43 39
0711 41 34
0434 41 37
0693 42 34
0556 42 37
0865 40 32
0730 40 35
Coefficient (t statistic) indicates significance at 95 confidence ~ 1982-1986 1
per ha of service area
Ta
ble
--I
nd
ices
of
se
v
ice a
rea
ib
en
ef
ited
a
rea
a
nd
a
vera
ge sea
so
na
l d
isch
arg
e
Ave
rage
A
vera
ge
tshy19
77
1978
19
79
1980
19
81
1982
19
83
1984
19
85
1986
19
77-8
1 19
82-8
6 S
tati
stic
a
(ind
ex
aver
age
1983
-198
5 =
100)
Ser
vice
are
a UP
RI IS
in
dex
882
91
9
920
91
5
951
95
1
100
0 10
00
100
0 10
00
917
99
0
bull5
53
Ang
at-M
aasi
m R
95
9
994
99
7
996
99
6
996
10
00
100
0 10
00
100
0 98
8
999
1
63
Sto
To
mas
10
63
106
3 10
1 9
10
29
103
0 99
8
100
0 10
00
100
0 11
10
104
1 10
22
-08
9
Siba
lom
-San
Jos
e 95
0
872
94
2
933
94
2
942
10
28
102
8 94
4
101
7
928
99
2
292
A
gana
n-St
a B
arba
ra
108
3 10
91
106
0 10
30
961
10
05
996
10
08
996
99
6
104
5 10
00
-21
1 A
vera
ge
987
98
8
981
98
1
916
97
8
100
5 10
07
988
10
25
984
10
01
123
Wet
seas
on b
enef
ited
are
a in
dex
UPR
IIS
110
0 10
07
114
4 10
55
113
8 11
74
951
10
71
971
ll
58
10
89
106
7 -0
47
Ang
at-M
aasi
m R
97
9
974
92
0
983
10
28
100
8 99
7
102
1 98
2
931
97
7
988
0
54
Sto
To
mas
11
63
115
9 11
23
107
5 10
80
103
9 98
1
978
10
41
103
9 11
20
101
6 -4
88
Siba
lom
-San
Jos
e 11
35
103
8 10
11
931
93
4
906
10
07
975
10
1S
998
10
11
981
-0
80
Aga
nan-
Sta
Bar
bara
10
85
110
5 10
74
106
8 10
01
104
2 99
8
101
5
987
61
0
106
7 93
0
-1 8
4
Ave
rage
10
92
105
6 10
54
102
4 10
36
103
4 98
7
101
3 10
00
947
10
53
996
-2
32
Dry
sea
Son
bene
fite
d ar
ea
inde
x U
PRIIS
14
04
155
0 15
S0
155
8 16
17
128
0 57
2
114
S 15
74
152
3 12
38
-09
4 A
ngat
-Maa
sim
R
903
93
0
103
2 10
61
104
2 10
69
988
99
2
102
1 99
6
993
10
13
061
S
to
Tom
as
105
7 12
27
122
5
961
99
9
115
9 10
1 7
91
0
107
3 12
1 2
10
94
107
4 -0
28
Si
balo
m-S
an J
ose
Aga
nan-
Sta
Bar
bara
66
5
95S
62
6
632
67
4
111
4
501
10
S1
412
11
51
766
93
3
856
94
3
107
0 94
8
107
4 11
09
111
6
158
6 58
8
987
97
6
110
4 5
35
083
N
w
Ave
rage
89
6
964
11
19
103
7 10
44
110
9 10
1 7
89
8
108
5 12
97
101
7 10
81
082
Wet
seas
on d
isch
arge
in
dex
UPR
IIS
132
9 72
7
142
5 11
88
120
4 98
8
105
8 10
63
879
96
4
117
5 99
1
-16
5 A
ngat
-Maa
sim
R
129
7 13
52
134
5 12
58
127
0 11
70
120
3 62
7
131
3 10
68
-18
9
Sto
To
mas
14
71
155
0 14
1 6
10
40
725
12
35
112
3 14
79
103
1 -4
48
Siba
lom
-San
Jos
e 96
8
733
11
1 5
92
2
907
47
1
102
3 85
7
ll2
0
422
92
9
779
-1
09
Aga
nan-
Sta
Bar
bara
87
9
863
68
1
925
96
8
110
7 70
5
871
87
7
00
9
Ave
rage
11
49
105
7 13
61
115
0 10
58
853
10
43
963
99
4
SO3
11
55
940
-2
85
Dry
sea
son
disc
harg
e in
dex
UPR
IIS
425
13
09
153
3 14
28
180
6 14
S0
125
8 56
3
117
9 14
66
130
0 11
89
-04
3
Ang
at-M
aasi
m R
12
26
161
6 15
30
139
2 12
85
100
8 10
33
958
14
41
107
1 -3
80
S
to
Tom
as
116
4 14
82
155
7 79
9
971
12
30
126
5 14
01
106
6 -2
44
Si
balo
m-S
an J
ose
486
71
5
782
62
9
789
64
5
116
9 11
86
801
65
3
918
2
36
Aga
nan-
Sta
Bar
bara
A
vera
ge
425
10
46
133
6 62
0
118
4 81
5
116
1 94
7
112
5 89
7
922
12
20
991
88
2
108
7 12
36
119
2 71
8
114
0 10
37
105
5 3
23
-07
5
a bull
indi
cate
s si
gnif
ican
ce a
t 95
per
cent
con
fide
nce
- 24 shy
Table 5--Results of pooled regressions yield indicators
------------------------- Dependent variables -------------------------shyIndependent variable ---------- Yield MTHa ----------- -- Specific Yield 1000 KgM
3 H20 shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season -shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 1967 (286)
3747 (580)
1991 (293)
2800 (460 )
0084 (041)
0441 (229)
0049 (023 )
0876 (343)
System 2 indicator 0400 (155)
0525 (226)
0057 (085)
0115 (1 95)
System 3 indicator 0466 (139 )
-0029 (-010)
-0016 (-018)
0006 (008)
System 4 indicator 0980 (470)
0090 (049)
0176 (333)
0391 (827)
System 5 indicator 1 221 (570)
0256 ( 138)
0104 (174)
0224 (440)
Reservoir indicator -1033 (-578)
-0063 (-039)
-0146 (-310)
-0326 (-569)
Period indicatora
Precipitation in mm Wet season
Annual
0163 (1 03)
-29E-4 (-182)
0231 (1 42)
-58E-4 (-606)
0161 (118)
92E-5 (072)
0199 043 )
-64pound-5 (-084)
0082 (200)
57E-6 (013)
0092 (223)
-61E-5 (-232)
0032 (093)
29E-7 (001)
0052 (112)
-12pound-4 (-414)
Nitrogen applicationb 0025 (254)
0021 (211)
0025 (274)
0020 (227)
0003 (1 05)
0002 (043)
0003 (101)
-0002 (-052)
R2 (adj) 0553 0516 0251 0198 0309 0262 0713 0445 number of observations 50 50 49 49 42 42 40 40 degrees of freedom 42 45 41 44 34 37 32 35
CoeffiCient (t statistic) indicates significance at 95 confidence ~ 1982-1986 = 1
kgha
- 25 shy
Table 6--Suggmry of estiates perforance indicators
Variable Predicted Predicted Period Dependent Variable
Description (units)
Period Coefficient
1976-86 IndicatorO
1976-86 Indicatorl
Effecta (X)
BenefittedService area ratio
Wet season ratio -0036 0853 0817 -425 Dry Season 0071 0539 0610 1313
Delivery Wet Dry
season season
nm water per benefitted ha
-0808 -1 214
6003 9083
5195 7869
-1346 -1337
Rice Yield Wet season metric tons 0163 3578 3741 456 Dry season per hectare 0161 3905 4066 412
Specific Yield Wet season kg rice per m3 0082 0348 0430 2353 Dry season water de 1 ivered 0032 0378 0409 837
a percentage changes relative to period=O values
- 26 shy
ENDNOTES
IResearch Fellow International Food Policy Research Institute and Senior Irrigation Specialist International Irrigation Management Institute
2Presumable farmers also base private investment decisions on such foreknowledge of costs and commonly do so in the case of small tubewells In addition user groups may invest in small communal schemes independently after examining projected costs and benefits Most commonly however new irrigation capacity is created as a result of public investment decisions made by government and thus the political process comes into play
3This principle applies only when water charges are levied volumetrically--an exceedingly rare situation in the developing worldshy-though the argument is frequently invoked as though it applies to all allocative situations
4For Cabanatuan City the longer period of record available ie 1904 to 1986 was divided into two equal periods and tested for difference in means to test the observation by farmers that rainfall today is lower than it used to be Mean precipitation after 1947 was actually about 3 percent higher than before but the difference was not significant (t = 07)
5The procedure used was to take time series data on total nitrogenfertilizer consumption in the Philippines multiply it by the ratio of N fertilizer used on rice to total N consumption and divide the product by the rice harvested area to obtain an estimate of N use perhectare Values of the ratio of N use on rice to total N use were taken from IRRI (1988 150) with interpolation used to fill in gaps in the series
6This assumes that timeliness is evaluated against a standard based on the physiological demand for water generated by the rice crop and the varying sensitivity of the crop to stress at different growth stages In such a case a poor timeliness rating for irrigationservice will result in reductions in yield
7Because rice yields are insensitive to water applications that exceed PET values a redistribution of water from water excess to water short portions of the command area will usually result in higher aggregate output and higher average yields Exceptional scenarios can be constructed in which this general rule would be violated but such scenarios are unlikely to play out in practice
- 27 shy
aThere are a variety of standards for judging equity that could be used here The most common one is equality of per hectare water deliveries In the present case the implicit standard is equality of water supply utility in producing a crop Thus if two areas received equal amounts of water on a per hectare basis but one had such high seepage and percolation losses that the crop failed to produce a harvest (and was therefore exempted from irrigation fee payment) the water supply to the combined area would be judged inequitable Because such radical differences in seepage and percolation rates are unlikely on puddled rice soils in the lowland Philippines and because the definition of benefitted area accommodates a wide range of acceptable yields these two definitions are largely indistinguishable at the current limits of data resolution
- 15 shy
relative to the system water input did not falloff as a result of the changes implemented at least over the short run
Impact magnitude
The preceding analysis has shown us that some indicators of irrigation performance changed significantly following the managerial changes of 1981 while others did not However it has not given us a sense of the size of the changes which occurred To determine the magnitude of these changes the regression model is used to predict the response of the composite system to the managerial changes given a common set of -input and environmental factors To do this averagevalues of the independent variables from the entire eleven-year period 1976 to 1986 are put into the model together with the previously determined coefficients to generate predicted average values of the various dependent variables used in the earlier analysis with and without the period dummy This procedure produces a pair of estimates for each dependent variable under the same conditions--one in which the system responds as it did after the managerial changes were implemented and one in which it responds as it did prior to their introduction The differences between these two values thus indicate the magnitude of the changes occurring in the various indicators of performance discussed above
The results of this exercise are shown in Table 6 The table shows that water availability decreased by about 13 percent in both wet and dry seasons when the period dummy was included and while the coefficients responsible were significant in the earlier analysisthis difference cannot be easily connected with levels of system management as discussed earlier With respect to rice output per hectare although the coefficients were not very significant it is interesting to note that yield increases by 163 kilograms per hectare for the wet season and by 101 kilogram for the dry when the period dummy is included in spite of the reduced water supply available Keep in mind that the predicted yield values have already been adjusted for differences in nitrogen fertilizer use and rainfall This suggests that timeliness and equity of distribution of water supply to farmers may have increased following the changes contributing to the higher predicted yields
Examining the impact of increased equity of distribution bylooking at the ratio of benefitted area to service area we recall that the change was positive and significant for the dry season and negative and not significant for the wet Table 6 shows that the dry season BASA ratio increases by 7 percentage points when the dummy is included a 131 percent increase Other things being equal this should result in a 131 percent increase in system output due to the expansion of ared benefitted This is a major impact on production
- 16 shy
CONCLUSIONS
The Philippine experiment to transform the national irrigation agency into an enterprise has undoubtedly been successful in reducing system operating expenses bringing revenues and costs into line and eliminating the recurrent cost burden imposed by large-scaleirrigation systems on the national budget Evidence presented in this paper indicates that in the process equity of water distribution across systems has also improved In the 5 years following the cessation of operating subsidies from the government an index of equity of distribution improved by about 13 percent At the same time per hectare yields adjusted for rainfall and nitrogen application held constant
There is a strong logical connection between the achievement of financial viability and improved equity of water distribution across the command Because increasing irrigation fees is a politicaldecisionlying largely beyond NIAs control expanding the area which can be billed for service is one of the few revenue increasing measures available to the irrigation agency which does not involve major additional investment In the face of constant or shrinkingwater supplies this is achieved only by redistributing water from areas receiving excessive supplies usually near the head ends of canals and laterals to areas receiving no supplies or inadequatesupplies often located near the tails of canals Although data are not available which would allow the direct examination of this hypotheses the two outcomes are logically consistent with each other
Data also show that per hectare water deliveries declined significantly in the five sample systems after 1981 even thoughrainfall did not differ appreciably between the two periods This decline averaged about 13 percent for both wet and dry seasons and is interpreted as a decline in water availability in the supplying rivers rather than a conscious reduction in withdrawals by system managers Such declines could result from changes in watershed runoff characteristics as caused by deforestation or from increased upstream abstractions from supplying rivers
Improved water distribution tends to increase the area served system agricultural output and NIA service fee revenue Reduced water supplies to the system tend to reduce these things Specificyield defined as system paddy output per unit water held roughly constant across the two periods indicating that the two effects mayhave offset each other
After adjusting for rainfall and nitrogen application perhectare yields increased only marginally in the post-1981 period Area served on the other hand increased by about 13 percent after adjusting for water supply availability indicating that the area benefitted by irrigation in the sample systems increased by about the same percentage Even if yields on this additional area are less than
- 17 shy
average yields for the system this still represents a sizeable increase in system agricultural output as a result of the change in management structure the increase coming not from higher yields but from expanded area under irrigation
The evidence assembled here suggests that there are significantfinancial and economic benefits to be had from changes in the basic character of irrigation managing agencies which make them more responsive to their clientele and which impose rational internal financial discipline on the agency The analysis suggests a number of additional questions however One relates to the longer-term impacts of the structural management changes The improvements in water distribution described here are relatively short-term events occurring during the first 5 years of the new management mode Critics have suggested the danger of underinvestment in systemmaintenance over the longer run accompanied by declining yields and benefitted areas and eventual system collapse This possibility needs to be closely monitored A second concern relates to the apparent decline in water supply to these 5 geographically dispersed systems The nature and causes of this decline need to be explored further since if widespread and secular it may represent a serious threat to the stability of Philippine rice production Whether stemming from poor forest management practices or deficient regulation and allocation of surface water resources or other unidentified factors it is an issue that deserves serious and urgent consideration
A third risk is that the incentive structure set up by NIA to guide and stimulate the performance of field units overemphasizes revenue generation at the expense of irrigation service provision to farmers The evidence presented here supports the view that these two objectives are mutually reinforcing under policies and conditions which have been established in the Philippines More detailed crossshysectional studies based on primary flow measurement data would add confidence to this conclusion and help to specify the conditions under which this effect occurs This could be extremely important in transferring the results of the Philippine experiment to other countries
A final risk is that outside intervention well meaning or otherwise will destroy the basis of NIAs financial autonomy or will impose external pressures or constraints on NIAs decision-making that will subvert the management practices which have been so painstakinglydeveloped and implemented Among these are calls for NIA to be subsumed again within the government department structure in the interests of better coordination with agriculture attempts byexternal financing agencies to arbitrarily increase NIAs expenditures on OampM on the assumption that this will increase system agricultural output or intervention by Philippine legislative bodies to restore operating subsidies to NIA with attached strings leading back to legislators home districts Pressures such as these will cut short a process of experimentation and improvement that seems promising
- 18 shy
enough to date to warrant its continuation Having developed the capacity to establish targets and implement and manage change NIA is in a strong position to modify its objectives to better achieve larger social purposes established for it It is critical to recognize however that this must happen within the context of financing policies that mandate financial autonomy for NIA if the fundamental institutional commitment to manage is to be preserved
The author would like to thank Leslie Small and JeremyBerkoff for helpful comments on an earlier unpublishedversion of this paper and Charles Rogers for his careful and creative help with the analysis
- 19 shy
BIBLIOGRAPHY
Abernethy Charles L 1990 Indicators of the performance of irrigation water distribution systems International Irrigation Management Institute Colombo Sri Lanka Mimeo
Asian Development Bank 1986 Irrigation service fees Proceedingsof the Regional Seminar on Irrigation Service Fees Manila Asian Development Bank
Carruthers Ian and Colin Clark 1981 Economics of IrrigationLiverpool Liverpool University Press Third Edition
Levine G and EW Coward Jr 1986 Irrigation water distribution implications for design and operation AGREP Division WorkingPaper 125 vol 1 World Bank Agriculture and Rural Development Department
Small Les E 1989 User charges in irrigation potentials and limitations Irrigation and drainage vol 3 no 2125-142
Small Les 1990 Irrigation service fees in Asia IrrigationManagement Network 9013 London Overseas DevelopmentInstitute
Svendsen Mark and Les Small 1989 A framework for assessing irrigation system performance Paper prepared for the Symposium on Performance Evaluation 23 November International IrrigationManagement Institute Sri Lanka
Table I--National Irrigation Administration revenues and expenditures in constant prices 1976-86
Item 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986
(peso million 1972)
Revenues Irrigation fees collected 1273 1483 17 13 1831 2070 1668 1699 1893 1728 2129 2546 Other income 715 737 2420 5591 2631 5990 7783 6638 5699 4932 2934
Total direct revenue 1988 2220 4133 7422 4701 7658 9482 8531 7427 061 5480
Expenses in 1972 pricesTotal expenses 4825 5716 5039 6329 3821 77 55 6166 4749 4348 4259 4959
Excess (deficit) (2837(3496) (906) 1093 877 (097) 3316 3782 3079 2802 521 N 0
Subsidies Government operation and
maintenance subsidies 2521 2741 2799 1817 1398 633 0 0 0 0 0 Calamity fund payments 548 0 0 0 0 0 0 0 119 0 142
Total subsidy 3069 2741 2799 1817 1398 633 0 0 119 0 142
Total excess (deficit) 231 (754) 1893 2910 2275 536 3316 3782 3198 2802 663
Source IFPRI analysis of NIA data
- 21 shy
Table 2--Descriptive characteristics of selected MIA systeasa
------- Region III -------- ----- Region VI ----shyUPRIIS Angatii Sto Sibalom- Aganan-
Haasim Thomasc San Jose Sta Barbara
Average service area (hal 102272 31462 3522 5282 8703 Average irrigated area (hal
Wet season 83768 23454 3007 4410 8300 Dry season
Average benefited area Wet season
(hal 64587
77 605
27639
22908
1 781
3007
2801
4369
2770
7698 Dry season
Average rainfall Wet season
(mml d 62478
1 685 5
27396
8576
1 781
3051 0
2769
24731
2997
20001 Dry season 756 333 322 2828 3025
Average discharge (Llsec) Wet season 46501 14792 1692 2353 4984 Dry season 78091 22812 2014 1276 2315
Average water delivery (mmday) Wet season 522 548 487 462 571 Dry season
Average yield (mtha) 1089 715 995 398 686
Wet season 345 419 322 395 435 Dry season
Avg yield per unit water 34 03
(kgm ) 451 412 399 426
Wet season 0373 0440 0373 0538 0443 Dry season 0248 0400 0279 0690 0428
t-statistic difference in mean rainfall 1978-81 1982-86e
Wet season 0432 0713 -0567 1169 1169 Dry season 0519 -0230 -0523 1187 1187 Annual 0460 0707 -0686 1445 1445
~ Summary numbers are averages for the period 1982-1986 except as noted Water delivery discharge and yield per unit discharge are 4-year averages 1982-1985
c Water delivery discharge and yield per unit discharge are 4-year averages d 1983-1986
For Angat 5 years are 1981-85 For St Thomas 1979-83 For Sibalom 1971-75 e No significant differences at 95 confidence
- 22 shy
Table 3--Results of pooled regressions water delivery indicators
Dependent variable Independent variable - Water Delivery in mmBenefited Ha - --- BenefitService Area Ratio --shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season-shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 6799 (998)
6189 (1019)
13102 (1093)
2411 (1 54)
0743 ( 1186)
0800 (1165)
0129 (131 )
0258 (329)
System 2 indicator 0173 (029)
-3743 (-363)
-0063 (-219)
0200 (505)
System 3 indicator 0698 (085)
0163 (012)
0085 (208)
-0046 (-097)
System 4 indicator -0514 (-l 07)
-7011 (-803)
0080 (340)
0169 (l87)
System 5 indicator 0177 (033)
-5049 (-538)
0174 (603)
0051 (058)
Reservoir indicator 0327 (078)
5703 (543)
-0126 (-510)
0156 (261)
Period indicatora -0808 (-240)
-0771 (-229)
-1214 (-209)
-1309 (-1 62)
-0036 (-192)
-0025 (-112)
0071 (239)
0030 (077)
Water delivery mmb
Wet season
Dry season
Precipitation in mm Wet season
Annual
Average daily rainfall Wet season
Dry season
-46E-4 (-125)
-1 9E-4 (-082)
-43E-4 (-077)
20E-3 (385)
0015 (1 59)
-0001 ( -004)
0023 (215)
0001 (021)
0070 (613)
-0017 (-058)
0046 (371)
0010 (033)
R2 (adj) number of observations degrees of freedom
0156 43 36
0133 43 39
0711 41 34
0434 41 37
0693 42 34
0556 42 37
0865 40 32
0730 40 35
Coefficient (t statistic) indicates significance at 95 confidence ~ 1982-1986 1
per ha of service area
Ta
ble
--I
nd
ices
of
se
v
ice a
rea
ib
en
ef
ited
a
rea
a
nd
a
vera
ge sea
so
na
l d
isch
arg
e
Ave
rage
A
vera
ge
tshy19
77
1978
19
79
1980
19
81
1982
19
83
1984
19
85
1986
19
77-8
1 19
82-8
6 S
tati
stic
a
(ind
ex
aver
age
1983
-198
5 =
100)
Ser
vice
are
a UP
RI IS
in
dex
882
91
9
920
91
5
951
95
1
100
0 10
00
100
0 10
00
917
99
0
bull5
53
Ang
at-M
aasi
m R
95
9
994
99
7
996
99
6
996
10
00
100
0 10
00
100
0 98
8
999
1
63
Sto
To
mas
10
63
106
3 10
1 9
10
29
103
0 99
8
100
0 10
00
100
0 11
10
104
1 10
22
-08
9
Siba
lom
-San
Jos
e 95
0
872
94
2
933
94
2
942
10
28
102
8 94
4
101
7
928
99
2
292
A
gana
n-St
a B
arba
ra
108
3 10
91
106
0 10
30
961
10
05
996
10
08
996
99
6
104
5 10
00
-21
1 A
vera
ge
987
98
8
981
98
1
916
97
8
100
5 10
07
988
10
25
984
10
01
123
Wet
seas
on b
enef
ited
are
a in
dex
UPR
IIS
110
0 10
07
114
4 10
55
113
8 11
74
951
10
71
971
ll
58
10
89
106
7 -0
47
Ang
at-M
aasi
m R
97
9
974
92
0
983
10
28
100
8 99
7
102
1 98
2
931
97
7
988
0
54
Sto
To
mas
11
63
115
9 11
23
107
5 10
80
103
9 98
1
978
10
41
103
9 11
20
101
6 -4
88
Siba
lom
-San
Jos
e 11
35
103
8 10
11
931
93
4
906
10
07
975
10
1S
998
10
11
981
-0
80
Aga
nan-
Sta
Bar
bara
10
85
110
5 10
74
106
8 10
01
104
2 99
8
101
5
987
61
0
106
7 93
0
-1 8
4
Ave
rage
10
92
105
6 10
54
102
4 10
36
103
4 98
7
101
3 10
00
947
10
53
996
-2
32
Dry
sea
Son
bene
fite
d ar
ea
inde
x U
PRIIS
14
04
155
0 15
S0
155
8 16
17
128
0 57
2
114
S 15
74
152
3 12
38
-09
4 A
ngat
-Maa
sim
R
903
93
0
103
2 10
61
104
2 10
69
988
99
2
102
1 99
6
993
10
13
061
S
to
Tom
as
105
7 12
27
122
5
961
99
9
115
9 10
1 7
91
0
107
3 12
1 2
10
94
107
4 -0
28
Si
balo
m-S
an J
ose
Aga
nan-
Sta
Bar
bara
66
5
95S
62
6
632
67
4
111
4
501
10
S1
412
11
51
766
93
3
856
94
3
107
0 94
8
107
4 11
09
111
6
158
6 58
8
987
97
6
110
4 5
35
083
N
w
Ave
rage
89
6
964
11
19
103
7 10
44
110
9 10
1 7
89
8
108
5 12
97
101
7 10
81
082
Wet
seas
on d
isch
arge
in
dex
UPR
IIS
132
9 72
7
142
5 11
88
120
4 98
8
105
8 10
63
879
96
4
117
5 99
1
-16
5 A
ngat
-Maa
sim
R
129
7 13
52
134
5 12
58
127
0 11
70
120
3 62
7
131
3 10
68
-18
9
Sto
To
mas
14
71
155
0 14
1 6
10
40
725
12
35
112
3 14
79
103
1 -4
48
Siba
lom
-San
Jos
e 96
8
733
11
1 5
92
2
907
47
1
102
3 85
7
ll2
0
422
92
9
779
-1
09
Aga
nan-
Sta
Bar
bara
87
9
863
68
1
925
96
8
110
7 70
5
871
87
7
00
9
Ave
rage
11
49
105
7 13
61
115
0 10
58
853
10
43
963
99
4
SO3
11
55
940
-2
85
Dry
sea
son
disc
harg
e in
dex
UPR
IIS
425
13
09
153
3 14
28
180
6 14
S0
125
8 56
3
117
9 14
66
130
0 11
89
-04
3
Ang
at-M
aasi
m R
12
26
161
6 15
30
139
2 12
85
100
8 10
33
958
14
41
107
1 -3
80
S
to
Tom
as
116
4 14
82
155
7 79
9
971
12
30
126
5 14
01
106
6 -2
44
Si
balo
m-S
an J
ose
486
71
5
782
62
9
789
64
5
116
9 11
86
801
65
3
918
2
36
Aga
nan-
Sta
Bar
bara
A
vera
ge
425
10
46
133
6 62
0
118
4 81
5
116
1 94
7
112
5 89
7
922
12
20
991
88
2
108
7 12
36
119
2 71
8
114
0 10
37
105
5 3
23
-07
5
a bull
indi
cate
s si
gnif
ican
ce a
t 95
per
cent
con
fide
nce
- 24 shy
Table 5--Results of pooled regressions yield indicators
------------------------- Dependent variables -------------------------shyIndependent variable ---------- Yield MTHa ----------- -- Specific Yield 1000 KgM
3 H20 shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season -shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 1967 (286)
3747 (580)
1991 (293)
2800 (460 )
0084 (041)
0441 (229)
0049 (023 )
0876 (343)
System 2 indicator 0400 (155)
0525 (226)
0057 (085)
0115 (1 95)
System 3 indicator 0466 (139 )
-0029 (-010)
-0016 (-018)
0006 (008)
System 4 indicator 0980 (470)
0090 (049)
0176 (333)
0391 (827)
System 5 indicator 1 221 (570)
0256 ( 138)
0104 (174)
0224 (440)
Reservoir indicator -1033 (-578)
-0063 (-039)
-0146 (-310)
-0326 (-569)
Period indicatora
Precipitation in mm Wet season
Annual
0163 (1 03)
-29E-4 (-182)
0231 (1 42)
-58E-4 (-606)
0161 (118)
92E-5 (072)
0199 043 )
-64pound-5 (-084)
0082 (200)
57E-6 (013)
0092 (223)
-61E-5 (-232)
0032 (093)
29E-7 (001)
0052 (112)
-12pound-4 (-414)
Nitrogen applicationb 0025 (254)
0021 (211)
0025 (274)
0020 (227)
0003 (1 05)
0002 (043)
0003 (101)
-0002 (-052)
R2 (adj) 0553 0516 0251 0198 0309 0262 0713 0445 number of observations 50 50 49 49 42 42 40 40 degrees of freedom 42 45 41 44 34 37 32 35
CoeffiCient (t statistic) indicates significance at 95 confidence ~ 1982-1986 = 1
kgha
- 25 shy
Table 6--Suggmry of estiates perforance indicators
Variable Predicted Predicted Period Dependent Variable
Description (units)
Period Coefficient
1976-86 IndicatorO
1976-86 Indicatorl
Effecta (X)
BenefittedService area ratio
Wet season ratio -0036 0853 0817 -425 Dry Season 0071 0539 0610 1313
Delivery Wet Dry
season season
nm water per benefitted ha
-0808 -1 214
6003 9083
5195 7869
-1346 -1337
Rice Yield Wet season metric tons 0163 3578 3741 456 Dry season per hectare 0161 3905 4066 412
Specific Yield Wet season kg rice per m3 0082 0348 0430 2353 Dry season water de 1 ivered 0032 0378 0409 837
a percentage changes relative to period=O values
- 26 shy
ENDNOTES
IResearch Fellow International Food Policy Research Institute and Senior Irrigation Specialist International Irrigation Management Institute
2Presumable farmers also base private investment decisions on such foreknowledge of costs and commonly do so in the case of small tubewells In addition user groups may invest in small communal schemes independently after examining projected costs and benefits Most commonly however new irrigation capacity is created as a result of public investment decisions made by government and thus the political process comes into play
3This principle applies only when water charges are levied volumetrically--an exceedingly rare situation in the developing worldshy-though the argument is frequently invoked as though it applies to all allocative situations
4For Cabanatuan City the longer period of record available ie 1904 to 1986 was divided into two equal periods and tested for difference in means to test the observation by farmers that rainfall today is lower than it used to be Mean precipitation after 1947 was actually about 3 percent higher than before but the difference was not significant (t = 07)
5The procedure used was to take time series data on total nitrogenfertilizer consumption in the Philippines multiply it by the ratio of N fertilizer used on rice to total N consumption and divide the product by the rice harvested area to obtain an estimate of N use perhectare Values of the ratio of N use on rice to total N use were taken from IRRI (1988 150) with interpolation used to fill in gaps in the series
6This assumes that timeliness is evaluated against a standard based on the physiological demand for water generated by the rice crop and the varying sensitivity of the crop to stress at different growth stages In such a case a poor timeliness rating for irrigationservice will result in reductions in yield
7Because rice yields are insensitive to water applications that exceed PET values a redistribution of water from water excess to water short portions of the command area will usually result in higher aggregate output and higher average yields Exceptional scenarios can be constructed in which this general rule would be violated but such scenarios are unlikely to play out in practice
- 27 shy
aThere are a variety of standards for judging equity that could be used here The most common one is equality of per hectare water deliveries In the present case the implicit standard is equality of water supply utility in producing a crop Thus if two areas received equal amounts of water on a per hectare basis but one had such high seepage and percolation losses that the crop failed to produce a harvest (and was therefore exempted from irrigation fee payment) the water supply to the combined area would be judged inequitable Because such radical differences in seepage and percolation rates are unlikely on puddled rice soils in the lowland Philippines and because the definition of benefitted area accommodates a wide range of acceptable yields these two definitions are largely indistinguishable at the current limits of data resolution
- 16 shy
CONCLUSIONS
The Philippine experiment to transform the national irrigation agency into an enterprise has undoubtedly been successful in reducing system operating expenses bringing revenues and costs into line and eliminating the recurrent cost burden imposed by large-scaleirrigation systems on the national budget Evidence presented in this paper indicates that in the process equity of water distribution across systems has also improved In the 5 years following the cessation of operating subsidies from the government an index of equity of distribution improved by about 13 percent At the same time per hectare yields adjusted for rainfall and nitrogen application held constant
There is a strong logical connection between the achievement of financial viability and improved equity of water distribution across the command Because increasing irrigation fees is a politicaldecisionlying largely beyond NIAs control expanding the area which can be billed for service is one of the few revenue increasing measures available to the irrigation agency which does not involve major additional investment In the face of constant or shrinkingwater supplies this is achieved only by redistributing water from areas receiving excessive supplies usually near the head ends of canals and laterals to areas receiving no supplies or inadequatesupplies often located near the tails of canals Although data are not available which would allow the direct examination of this hypotheses the two outcomes are logically consistent with each other
Data also show that per hectare water deliveries declined significantly in the five sample systems after 1981 even thoughrainfall did not differ appreciably between the two periods This decline averaged about 13 percent for both wet and dry seasons and is interpreted as a decline in water availability in the supplying rivers rather than a conscious reduction in withdrawals by system managers Such declines could result from changes in watershed runoff characteristics as caused by deforestation or from increased upstream abstractions from supplying rivers
Improved water distribution tends to increase the area served system agricultural output and NIA service fee revenue Reduced water supplies to the system tend to reduce these things Specificyield defined as system paddy output per unit water held roughly constant across the two periods indicating that the two effects mayhave offset each other
After adjusting for rainfall and nitrogen application perhectare yields increased only marginally in the post-1981 period Area served on the other hand increased by about 13 percent after adjusting for water supply availability indicating that the area benefitted by irrigation in the sample systems increased by about the same percentage Even if yields on this additional area are less than
- 17 shy
average yields for the system this still represents a sizeable increase in system agricultural output as a result of the change in management structure the increase coming not from higher yields but from expanded area under irrigation
The evidence assembled here suggests that there are significantfinancial and economic benefits to be had from changes in the basic character of irrigation managing agencies which make them more responsive to their clientele and which impose rational internal financial discipline on the agency The analysis suggests a number of additional questions however One relates to the longer-term impacts of the structural management changes The improvements in water distribution described here are relatively short-term events occurring during the first 5 years of the new management mode Critics have suggested the danger of underinvestment in systemmaintenance over the longer run accompanied by declining yields and benefitted areas and eventual system collapse This possibility needs to be closely monitored A second concern relates to the apparent decline in water supply to these 5 geographically dispersed systems The nature and causes of this decline need to be explored further since if widespread and secular it may represent a serious threat to the stability of Philippine rice production Whether stemming from poor forest management practices or deficient regulation and allocation of surface water resources or other unidentified factors it is an issue that deserves serious and urgent consideration
A third risk is that the incentive structure set up by NIA to guide and stimulate the performance of field units overemphasizes revenue generation at the expense of irrigation service provision to farmers The evidence presented here supports the view that these two objectives are mutually reinforcing under policies and conditions which have been established in the Philippines More detailed crossshysectional studies based on primary flow measurement data would add confidence to this conclusion and help to specify the conditions under which this effect occurs This could be extremely important in transferring the results of the Philippine experiment to other countries
A final risk is that outside intervention well meaning or otherwise will destroy the basis of NIAs financial autonomy or will impose external pressures or constraints on NIAs decision-making that will subvert the management practices which have been so painstakinglydeveloped and implemented Among these are calls for NIA to be subsumed again within the government department structure in the interests of better coordination with agriculture attempts byexternal financing agencies to arbitrarily increase NIAs expenditures on OampM on the assumption that this will increase system agricultural output or intervention by Philippine legislative bodies to restore operating subsidies to NIA with attached strings leading back to legislators home districts Pressures such as these will cut short a process of experimentation and improvement that seems promising
- 18 shy
enough to date to warrant its continuation Having developed the capacity to establish targets and implement and manage change NIA is in a strong position to modify its objectives to better achieve larger social purposes established for it It is critical to recognize however that this must happen within the context of financing policies that mandate financial autonomy for NIA if the fundamental institutional commitment to manage is to be preserved
The author would like to thank Leslie Small and JeremyBerkoff for helpful comments on an earlier unpublishedversion of this paper and Charles Rogers for his careful and creative help with the analysis
- 19 shy
BIBLIOGRAPHY
Abernethy Charles L 1990 Indicators of the performance of irrigation water distribution systems International Irrigation Management Institute Colombo Sri Lanka Mimeo
Asian Development Bank 1986 Irrigation service fees Proceedingsof the Regional Seminar on Irrigation Service Fees Manila Asian Development Bank
Carruthers Ian and Colin Clark 1981 Economics of IrrigationLiverpool Liverpool University Press Third Edition
Levine G and EW Coward Jr 1986 Irrigation water distribution implications for design and operation AGREP Division WorkingPaper 125 vol 1 World Bank Agriculture and Rural Development Department
Small Les E 1989 User charges in irrigation potentials and limitations Irrigation and drainage vol 3 no 2125-142
Small Les 1990 Irrigation service fees in Asia IrrigationManagement Network 9013 London Overseas DevelopmentInstitute
Svendsen Mark and Les Small 1989 A framework for assessing irrigation system performance Paper prepared for the Symposium on Performance Evaluation 23 November International IrrigationManagement Institute Sri Lanka
Table I--National Irrigation Administration revenues and expenditures in constant prices 1976-86
Item 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986
(peso million 1972)
Revenues Irrigation fees collected 1273 1483 17 13 1831 2070 1668 1699 1893 1728 2129 2546 Other income 715 737 2420 5591 2631 5990 7783 6638 5699 4932 2934
Total direct revenue 1988 2220 4133 7422 4701 7658 9482 8531 7427 061 5480
Expenses in 1972 pricesTotal expenses 4825 5716 5039 6329 3821 77 55 6166 4749 4348 4259 4959
Excess (deficit) (2837(3496) (906) 1093 877 (097) 3316 3782 3079 2802 521 N 0
Subsidies Government operation and
maintenance subsidies 2521 2741 2799 1817 1398 633 0 0 0 0 0 Calamity fund payments 548 0 0 0 0 0 0 0 119 0 142
Total subsidy 3069 2741 2799 1817 1398 633 0 0 119 0 142
Total excess (deficit) 231 (754) 1893 2910 2275 536 3316 3782 3198 2802 663
Source IFPRI analysis of NIA data
- 21 shy
Table 2--Descriptive characteristics of selected MIA systeasa
------- Region III -------- ----- Region VI ----shyUPRIIS Angatii Sto Sibalom- Aganan-
Haasim Thomasc San Jose Sta Barbara
Average service area (hal 102272 31462 3522 5282 8703 Average irrigated area (hal
Wet season 83768 23454 3007 4410 8300 Dry season
Average benefited area Wet season
(hal 64587
77 605
27639
22908
1 781
3007
2801
4369
2770
7698 Dry season
Average rainfall Wet season
(mml d 62478
1 685 5
27396
8576
1 781
3051 0
2769
24731
2997
20001 Dry season 756 333 322 2828 3025
Average discharge (Llsec) Wet season 46501 14792 1692 2353 4984 Dry season 78091 22812 2014 1276 2315
Average water delivery (mmday) Wet season 522 548 487 462 571 Dry season
Average yield (mtha) 1089 715 995 398 686
Wet season 345 419 322 395 435 Dry season
Avg yield per unit water 34 03
(kgm ) 451 412 399 426
Wet season 0373 0440 0373 0538 0443 Dry season 0248 0400 0279 0690 0428
t-statistic difference in mean rainfall 1978-81 1982-86e
Wet season 0432 0713 -0567 1169 1169 Dry season 0519 -0230 -0523 1187 1187 Annual 0460 0707 -0686 1445 1445
~ Summary numbers are averages for the period 1982-1986 except as noted Water delivery discharge and yield per unit discharge are 4-year averages 1982-1985
c Water delivery discharge and yield per unit discharge are 4-year averages d 1983-1986
For Angat 5 years are 1981-85 For St Thomas 1979-83 For Sibalom 1971-75 e No significant differences at 95 confidence
- 22 shy
Table 3--Results of pooled regressions water delivery indicators
Dependent variable Independent variable - Water Delivery in mmBenefited Ha - --- BenefitService Area Ratio --shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season-shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 6799 (998)
6189 (1019)
13102 (1093)
2411 (1 54)
0743 ( 1186)
0800 (1165)
0129 (131 )
0258 (329)
System 2 indicator 0173 (029)
-3743 (-363)
-0063 (-219)
0200 (505)
System 3 indicator 0698 (085)
0163 (012)
0085 (208)
-0046 (-097)
System 4 indicator -0514 (-l 07)
-7011 (-803)
0080 (340)
0169 (l87)
System 5 indicator 0177 (033)
-5049 (-538)
0174 (603)
0051 (058)
Reservoir indicator 0327 (078)
5703 (543)
-0126 (-510)
0156 (261)
Period indicatora -0808 (-240)
-0771 (-229)
-1214 (-209)
-1309 (-1 62)
-0036 (-192)
-0025 (-112)
0071 (239)
0030 (077)
Water delivery mmb
Wet season
Dry season
Precipitation in mm Wet season
Annual
Average daily rainfall Wet season
Dry season
-46E-4 (-125)
-1 9E-4 (-082)
-43E-4 (-077)
20E-3 (385)
0015 (1 59)
-0001 ( -004)
0023 (215)
0001 (021)
0070 (613)
-0017 (-058)
0046 (371)
0010 (033)
R2 (adj) number of observations degrees of freedom
0156 43 36
0133 43 39
0711 41 34
0434 41 37
0693 42 34
0556 42 37
0865 40 32
0730 40 35
Coefficient (t statistic) indicates significance at 95 confidence ~ 1982-1986 1
per ha of service area
Ta
ble
--I
nd
ices
of
se
v
ice a
rea
ib
en
ef
ited
a
rea
a
nd
a
vera
ge sea
so
na
l d
isch
arg
e
Ave
rage
A
vera
ge
tshy19
77
1978
19
79
1980
19
81
1982
19
83
1984
19
85
1986
19
77-8
1 19
82-8
6 S
tati
stic
a
(ind
ex
aver
age
1983
-198
5 =
100)
Ser
vice
are
a UP
RI IS
in
dex
882
91
9
920
91
5
951
95
1
100
0 10
00
100
0 10
00
917
99
0
bull5
53
Ang
at-M
aasi
m R
95
9
994
99
7
996
99
6
996
10
00
100
0 10
00
100
0 98
8
999
1
63
Sto
To
mas
10
63
106
3 10
1 9
10
29
103
0 99
8
100
0 10
00
100
0 11
10
104
1 10
22
-08
9
Siba
lom
-San
Jos
e 95
0
872
94
2
933
94
2
942
10
28
102
8 94
4
101
7
928
99
2
292
A
gana
n-St
a B
arba
ra
108
3 10
91
106
0 10
30
961
10
05
996
10
08
996
99
6
104
5 10
00
-21
1 A
vera
ge
987
98
8
981
98
1
916
97
8
100
5 10
07
988
10
25
984
10
01
123
Wet
seas
on b
enef
ited
are
a in
dex
UPR
IIS
110
0 10
07
114
4 10
55
113
8 11
74
951
10
71
971
ll
58
10
89
106
7 -0
47
Ang
at-M
aasi
m R
97
9
974
92
0
983
10
28
100
8 99
7
102
1 98
2
931
97
7
988
0
54
Sto
To
mas
11
63
115
9 11
23
107
5 10
80
103
9 98
1
978
10
41
103
9 11
20
101
6 -4
88
Siba
lom
-San
Jos
e 11
35
103
8 10
11
931
93
4
906
10
07
975
10
1S
998
10
11
981
-0
80
Aga
nan-
Sta
Bar
bara
10
85
110
5 10
74
106
8 10
01
104
2 99
8
101
5
987
61
0
106
7 93
0
-1 8
4
Ave
rage
10
92
105
6 10
54
102
4 10
36
103
4 98
7
101
3 10
00
947
10
53
996
-2
32
Dry
sea
Son
bene
fite
d ar
ea
inde
x U
PRIIS
14
04
155
0 15
S0
155
8 16
17
128
0 57
2
114
S 15
74
152
3 12
38
-09
4 A
ngat
-Maa
sim
R
903
93
0
103
2 10
61
104
2 10
69
988
99
2
102
1 99
6
993
10
13
061
S
to
Tom
as
105
7 12
27
122
5
961
99
9
115
9 10
1 7
91
0
107
3 12
1 2
10
94
107
4 -0
28
Si
balo
m-S
an J
ose
Aga
nan-
Sta
Bar
bara
66
5
95S
62
6
632
67
4
111
4
501
10
S1
412
11
51
766
93
3
856
94
3
107
0 94
8
107
4 11
09
111
6
158
6 58
8
987
97
6
110
4 5
35
083
N
w
Ave
rage
89
6
964
11
19
103
7 10
44
110
9 10
1 7
89
8
108
5 12
97
101
7 10
81
082
Wet
seas
on d
isch
arge
in
dex
UPR
IIS
132
9 72
7
142
5 11
88
120
4 98
8
105
8 10
63
879
96
4
117
5 99
1
-16
5 A
ngat
-Maa
sim
R
129
7 13
52
134
5 12
58
127
0 11
70
120
3 62
7
131
3 10
68
-18
9
Sto
To
mas
14
71
155
0 14
1 6
10
40
725
12
35
112
3 14
79
103
1 -4
48
Siba
lom
-San
Jos
e 96
8
733
11
1 5
92
2
907
47
1
102
3 85
7
ll2
0
422
92
9
779
-1
09
Aga
nan-
Sta
Bar
bara
87
9
863
68
1
925
96
8
110
7 70
5
871
87
7
00
9
Ave
rage
11
49
105
7 13
61
115
0 10
58
853
10
43
963
99
4
SO3
11
55
940
-2
85
Dry
sea
son
disc
harg
e in
dex
UPR
IIS
425
13
09
153
3 14
28
180
6 14
S0
125
8 56
3
117
9 14
66
130
0 11
89
-04
3
Ang
at-M
aasi
m R
12
26
161
6 15
30
139
2 12
85
100
8 10
33
958
14
41
107
1 -3
80
S
to
Tom
as
116
4 14
82
155
7 79
9
971
12
30
126
5 14
01
106
6 -2
44
Si
balo
m-S
an J
ose
486
71
5
782
62
9
789
64
5
116
9 11
86
801
65
3
918
2
36
Aga
nan-
Sta
Bar
bara
A
vera
ge
425
10
46
133
6 62
0
118
4 81
5
116
1 94
7
112
5 89
7
922
12
20
991
88
2
108
7 12
36
119
2 71
8
114
0 10
37
105
5 3
23
-07
5
a bull
indi
cate
s si
gnif
ican
ce a
t 95
per
cent
con
fide
nce
- 24 shy
Table 5--Results of pooled regressions yield indicators
------------------------- Dependent variables -------------------------shyIndependent variable ---------- Yield MTHa ----------- -- Specific Yield 1000 KgM
3 H20 shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season -shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 1967 (286)
3747 (580)
1991 (293)
2800 (460 )
0084 (041)
0441 (229)
0049 (023 )
0876 (343)
System 2 indicator 0400 (155)
0525 (226)
0057 (085)
0115 (1 95)
System 3 indicator 0466 (139 )
-0029 (-010)
-0016 (-018)
0006 (008)
System 4 indicator 0980 (470)
0090 (049)
0176 (333)
0391 (827)
System 5 indicator 1 221 (570)
0256 ( 138)
0104 (174)
0224 (440)
Reservoir indicator -1033 (-578)
-0063 (-039)
-0146 (-310)
-0326 (-569)
Period indicatora
Precipitation in mm Wet season
Annual
0163 (1 03)
-29E-4 (-182)
0231 (1 42)
-58E-4 (-606)
0161 (118)
92E-5 (072)
0199 043 )
-64pound-5 (-084)
0082 (200)
57E-6 (013)
0092 (223)
-61E-5 (-232)
0032 (093)
29E-7 (001)
0052 (112)
-12pound-4 (-414)
Nitrogen applicationb 0025 (254)
0021 (211)
0025 (274)
0020 (227)
0003 (1 05)
0002 (043)
0003 (101)
-0002 (-052)
R2 (adj) 0553 0516 0251 0198 0309 0262 0713 0445 number of observations 50 50 49 49 42 42 40 40 degrees of freedom 42 45 41 44 34 37 32 35
CoeffiCient (t statistic) indicates significance at 95 confidence ~ 1982-1986 = 1
kgha
- 25 shy
Table 6--Suggmry of estiates perforance indicators
Variable Predicted Predicted Period Dependent Variable
Description (units)
Period Coefficient
1976-86 IndicatorO
1976-86 Indicatorl
Effecta (X)
BenefittedService area ratio
Wet season ratio -0036 0853 0817 -425 Dry Season 0071 0539 0610 1313
Delivery Wet Dry
season season
nm water per benefitted ha
-0808 -1 214
6003 9083
5195 7869
-1346 -1337
Rice Yield Wet season metric tons 0163 3578 3741 456 Dry season per hectare 0161 3905 4066 412
Specific Yield Wet season kg rice per m3 0082 0348 0430 2353 Dry season water de 1 ivered 0032 0378 0409 837
a percentage changes relative to period=O values
- 26 shy
ENDNOTES
IResearch Fellow International Food Policy Research Institute and Senior Irrigation Specialist International Irrigation Management Institute
2Presumable farmers also base private investment decisions on such foreknowledge of costs and commonly do so in the case of small tubewells In addition user groups may invest in small communal schemes independently after examining projected costs and benefits Most commonly however new irrigation capacity is created as a result of public investment decisions made by government and thus the political process comes into play
3This principle applies only when water charges are levied volumetrically--an exceedingly rare situation in the developing worldshy-though the argument is frequently invoked as though it applies to all allocative situations
4For Cabanatuan City the longer period of record available ie 1904 to 1986 was divided into two equal periods and tested for difference in means to test the observation by farmers that rainfall today is lower than it used to be Mean precipitation after 1947 was actually about 3 percent higher than before but the difference was not significant (t = 07)
5The procedure used was to take time series data on total nitrogenfertilizer consumption in the Philippines multiply it by the ratio of N fertilizer used on rice to total N consumption and divide the product by the rice harvested area to obtain an estimate of N use perhectare Values of the ratio of N use on rice to total N use were taken from IRRI (1988 150) with interpolation used to fill in gaps in the series
6This assumes that timeliness is evaluated against a standard based on the physiological demand for water generated by the rice crop and the varying sensitivity of the crop to stress at different growth stages In such a case a poor timeliness rating for irrigationservice will result in reductions in yield
7Because rice yields are insensitive to water applications that exceed PET values a redistribution of water from water excess to water short portions of the command area will usually result in higher aggregate output and higher average yields Exceptional scenarios can be constructed in which this general rule would be violated but such scenarios are unlikely to play out in practice
- 27 shy
aThere are a variety of standards for judging equity that could be used here The most common one is equality of per hectare water deliveries In the present case the implicit standard is equality of water supply utility in producing a crop Thus if two areas received equal amounts of water on a per hectare basis but one had such high seepage and percolation losses that the crop failed to produce a harvest (and was therefore exempted from irrigation fee payment) the water supply to the combined area would be judged inequitable Because such radical differences in seepage and percolation rates are unlikely on puddled rice soils in the lowland Philippines and because the definition of benefitted area accommodates a wide range of acceptable yields these two definitions are largely indistinguishable at the current limits of data resolution
- 17 shy
average yields for the system this still represents a sizeable increase in system agricultural output as a result of the change in management structure the increase coming not from higher yields but from expanded area under irrigation
The evidence assembled here suggests that there are significantfinancial and economic benefits to be had from changes in the basic character of irrigation managing agencies which make them more responsive to their clientele and which impose rational internal financial discipline on the agency The analysis suggests a number of additional questions however One relates to the longer-term impacts of the structural management changes The improvements in water distribution described here are relatively short-term events occurring during the first 5 years of the new management mode Critics have suggested the danger of underinvestment in systemmaintenance over the longer run accompanied by declining yields and benefitted areas and eventual system collapse This possibility needs to be closely monitored A second concern relates to the apparent decline in water supply to these 5 geographically dispersed systems The nature and causes of this decline need to be explored further since if widespread and secular it may represent a serious threat to the stability of Philippine rice production Whether stemming from poor forest management practices or deficient regulation and allocation of surface water resources or other unidentified factors it is an issue that deserves serious and urgent consideration
A third risk is that the incentive structure set up by NIA to guide and stimulate the performance of field units overemphasizes revenue generation at the expense of irrigation service provision to farmers The evidence presented here supports the view that these two objectives are mutually reinforcing under policies and conditions which have been established in the Philippines More detailed crossshysectional studies based on primary flow measurement data would add confidence to this conclusion and help to specify the conditions under which this effect occurs This could be extremely important in transferring the results of the Philippine experiment to other countries
A final risk is that outside intervention well meaning or otherwise will destroy the basis of NIAs financial autonomy or will impose external pressures or constraints on NIAs decision-making that will subvert the management practices which have been so painstakinglydeveloped and implemented Among these are calls for NIA to be subsumed again within the government department structure in the interests of better coordination with agriculture attempts byexternal financing agencies to arbitrarily increase NIAs expenditures on OampM on the assumption that this will increase system agricultural output or intervention by Philippine legislative bodies to restore operating subsidies to NIA with attached strings leading back to legislators home districts Pressures such as these will cut short a process of experimentation and improvement that seems promising
- 18 shy
enough to date to warrant its continuation Having developed the capacity to establish targets and implement and manage change NIA is in a strong position to modify its objectives to better achieve larger social purposes established for it It is critical to recognize however that this must happen within the context of financing policies that mandate financial autonomy for NIA if the fundamental institutional commitment to manage is to be preserved
The author would like to thank Leslie Small and JeremyBerkoff for helpful comments on an earlier unpublishedversion of this paper and Charles Rogers for his careful and creative help with the analysis
- 19 shy
BIBLIOGRAPHY
Abernethy Charles L 1990 Indicators of the performance of irrigation water distribution systems International Irrigation Management Institute Colombo Sri Lanka Mimeo
Asian Development Bank 1986 Irrigation service fees Proceedingsof the Regional Seminar on Irrigation Service Fees Manila Asian Development Bank
Carruthers Ian and Colin Clark 1981 Economics of IrrigationLiverpool Liverpool University Press Third Edition
Levine G and EW Coward Jr 1986 Irrigation water distribution implications for design and operation AGREP Division WorkingPaper 125 vol 1 World Bank Agriculture and Rural Development Department
Small Les E 1989 User charges in irrigation potentials and limitations Irrigation and drainage vol 3 no 2125-142
Small Les 1990 Irrigation service fees in Asia IrrigationManagement Network 9013 London Overseas DevelopmentInstitute
Svendsen Mark and Les Small 1989 A framework for assessing irrigation system performance Paper prepared for the Symposium on Performance Evaluation 23 November International IrrigationManagement Institute Sri Lanka
Table I--National Irrigation Administration revenues and expenditures in constant prices 1976-86
Item 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986
(peso million 1972)
Revenues Irrigation fees collected 1273 1483 17 13 1831 2070 1668 1699 1893 1728 2129 2546 Other income 715 737 2420 5591 2631 5990 7783 6638 5699 4932 2934
Total direct revenue 1988 2220 4133 7422 4701 7658 9482 8531 7427 061 5480
Expenses in 1972 pricesTotal expenses 4825 5716 5039 6329 3821 77 55 6166 4749 4348 4259 4959
Excess (deficit) (2837(3496) (906) 1093 877 (097) 3316 3782 3079 2802 521 N 0
Subsidies Government operation and
maintenance subsidies 2521 2741 2799 1817 1398 633 0 0 0 0 0 Calamity fund payments 548 0 0 0 0 0 0 0 119 0 142
Total subsidy 3069 2741 2799 1817 1398 633 0 0 119 0 142
Total excess (deficit) 231 (754) 1893 2910 2275 536 3316 3782 3198 2802 663
Source IFPRI analysis of NIA data
- 21 shy
Table 2--Descriptive characteristics of selected MIA systeasa
------- Region III -------- ----- Region VI ----shyUPRIIS Angatii Sto Sibalom- Aganan-
Haasim Thomasc San Jose Sta Barbara
Average service area (hal 102272 31462 3522 5282 8703 Average irrigated area (hal
Wet season 83768 23454 3007 4410 8300 Dry season
Average benefited area Wet season
(hal 64587
77 605
27639
22908
1 781
3007
2801
4369
2770
7698 Dry season
Average rainfall Wet season
(mml d 62478
1 685 5
27396
8576
1 781
3051 0
2769
24731
2997
20001 Dry season 756 333 322 2828 3025
Average discharge (Llsec) Wet season 46501 14792 1692 2353 4984 Dry season 78091 22812 2014 1276 2315
Average water delivery (mmday) Wet season 522 548 487 462 571 Dry season
Average yield (mtha) 1089 715 995 398 686
Wet season 345 419 322 395 435 Dry season
Avg yield per unit water 34 03
(kgm ) 451 412 399 426
Wet season 0373 0440 0373 0538 0443 Dry season 0248 0400 0279 0690 0428
t-statistic difference in mean rainfall 1978-81 1982-86e
Wet season 0432 0713 -0567 1169 1169 Dry season 0519 -0230 -0523 1187 1187 Annual 0460 0707 -0686 1445 1445
~ Summary numbers are averages for the period 1982-1986 except as noted Water delivery discharge and yield per unit discharge are 4-year averages 1982-1985
c Water delivery discharge and yield per unit discharge are 4-year averages d 1983-1986
For Angat 5 years are 1981-85 For St Thomas 1979-83 For Sibalom 1971-75 e No significant differences at 95 confidence
- 22 shy
Table 3--Results of pooled regressions water delivery indicators
Dependent variable Independent variable - Water Delivery in mmBenefited Ha - --- BenefitService Area Ratio --shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season-shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 6799 (998)
6189 (1019)
13102 (1093)
2411 (1 54)
0743 ( 1186)
0800 (1165)
0129 (131 )
0258 (329)
System 2 indicator 0173 (029)
-3743 (-363)
-0063 (-219)
0200 (505)
System 3 indicator 0698 (085)
0163 (012)
0085 (208)
-0046 (-097)
System 4 indicator -0514 (-l 07)
-7011 (-803)
0080 (340)
0169 (l87)
System 5 indicator 0177 (033)
-5049 (-538)
0174 (603)
0051 (058)
Reservoir indicator 0327 (078)
5703 (543)
-0126 (-510)
0156 (261)
Period indicatora -0808 (-240)
-0771 (-229)
-1214 (-209)
-1309 (-1 62)
-0036 (-192)
-0025 (-112)
0071 (239)
0030 (077)
Water delivery mmb
Wet season
Dry season
Precipitation in mm Wet season
Annual
Average daily rainfall Wet season
Dry season
-46E-4 (-125)
-1 9E-4 (-082)
-43E-4 (-077)
20E-3 (385)
0015 (1 59)
-0001 ( -004)
0023 (215)
0001 (021)
0070 (613)
-0017 (-058)
0046 (371)
0010 (033)
R2 (adj) number of observations degrees of freedom
0156 43 36
0133 43 39
0711 41 34
0434 41 37
0693 42 34
0556 42 37
0865 40 32
0730 40 35
Coefficient (t statistic) indicates significance at 95 confidence ~ 1982-1986 1
per ha of service area
Ta
ble
--I
nd
ices
of
se
v
ice a
rea
ib
en
ef
ited
a
rea
a
nd
a
vera
ge sea
so
na
l d
isch
arg
e
Ave
rage
A
vera
ge
tshy19
77
1978
19
79
1980
19
81
1982
19
83
1984
19
85
1986
19
77-8
1 19
82-8
6 S
tati
stic
a
(ind
ex
aver
age
1983
-198
5 =
100)
Ser
vice
are
a UP
RI IS
in
dex
882
91
9
920
91
5
951
95
1
100
0 10
00
100
0 10
00
917
99
0
bull5
53
Ang
at-M
aasi
m R
95
9
994
99
7
996
99
6
996
10
00
100
0 10
00
100
0 98
8
999
1
63
Sto
To
mas
10
63
106
3 10
1 9
10
29
103
0 99
8
100
0 10
00
100
0 11
10
104
1 10
22
-08
9
Siba
lom
-San
Jos
e 95
0
872
94
2
933
94
2
942
10
28
102
8 94
4
101
7
928
99
2
292
A
gana
n-St
a B
arba
ra
108
3 10
91
106
0 10
30
961
10
05
996
10
08
996
99
6
104
5 10
00
-21
1 A
vera
ge
987
98
8
981
98
1
916
97
8
100
5 10
07
988
10
25
984
10
01
123
Wet
seas
on b
enef
ited
are
a in
dex
UPR
IIS
110
0 10
07
114
4 10
55
113
8 11
74
951
10
71
971
ll
58
10
89
106
7 -0
47
Ang
at-M
aasi
m R
97
9
974
92
0
983
10
28
100
8 99
7
102
1 98
2
931
97
7
988
0
54
Sto
To
mas
11
63
115
9 11
23
107
5 10
80
103
9 98
1
978
10
41
103
9 11
20
101
6 -4
88
Siba
lom
-San
Jos
e 11
35
103
8 10
11
931
93
4
906
10
07
975
10
1S
998
10
11
981
-0
80
Aga
nan-
Sta
Bar
bara
10
85
110
5 10
74
106
8 10
01
104
2 99
8
101
5
987
61
0
106
7 93
0
-1 8
4
Ave
rage
10
92
105
6 10
54
102
4 10
36
103
4 98
7
101
3 10
00
947
10
53
996
-2
32
Dry
sea
Son
bene
fite
d ar
ea
inde
x U
PRIIS
14
04
155
0 15
S0
155
8 16
17
128
0 57
2
114
S 15
74
152
3 12
38
-09
4 A
ngat
-Maa
sim
R
903
93
0
103
2 10
61
104
2 10
69
988
99
2
102
1 99
6
993
10
13
061
S
to
Tom
as
105
7 12
27
122
5
961
99
9
115
9 10
1 7
91
0
107
3 12
1 2
10
94
107
4 -0
28
Si
balo
m-S
an J
ose
Aga
nan-
Sta
Bar
bara
66
5
95S
62
6
632
67
4
111
4
501
10
S1
412
11
51
766
93
3
856
94
3
107
0 94
8
107
4 11
09
111
6
158
6 58
8
987
97
6
110
4 5
35
083
N
w
Ave
rage
89
6
964
11
19
103
7 10
44
110
9 10
1 7
89
8
108
5 12
97
101
7 10
81
082
Wet
seas
on d
isch
arge
in
dex
UPR
IIS
132
9 72
7
142
5 11
88
120
4 98
8
105
8 10
63
879
96
4
117
5 99
1
-16
5 A
ngat
-Maa
sim
R
129
7 13
52
134
5 12
58
127
0 11
70
120
3 62
7
131
3 10
68
-18
9
Sto
To
mas
14
71
155
0 14
1 6
10
40
725
12
35
112
3 14
79
103
1 -4
48
Siba
lom
-San
Jos
e 96
8
733
11
1 5
92
2
907
47
1
102
3 85
7
ll2
0
422
92
9
779
-1
09
Aga
nan-
Sta
Bar
bara
87
9
863
68
1
925
96
8
110
7 70
5
871
87
7
00
9
Ave
rage
11
49
105
7 13
61
115
0 10
58
853
10
43
963
99
4
SO3
11
55
940
-2
85
Dry
sea
son
disc
harg
e in
dex
UPR
IIS
425
13
09
153
3 14
28
180
6 14
S0
125
8 56
3
117
9 14
66
130
0 11
89
-04
3
Ang
at-M
aasi
m R
12
26
161
6 15
30
139
2 12
85
100
8 10
33
958
14
41
107
1 -3
80
S
to
Tom
as
116
4 14
82
155
7 79
9
971
12
30
126
5 14
01
106
6 -2
44
Si
balo
m-S
an J
ose
486
71
5
782
62
9
789
64
5
116
9 11
86
801
65
3
918
2
36
Aga
nan-
Sta
Bar
bara
A
vera
ge
425
10
46
133
6 62
0
118
4 81
5
116
1 94
7
112
5 89
7
922
12
20
991
88
2
108
7 12
36
119
2 71
8
114
0 10
37
105
5 3
23
-07
5
a bull
indi
cate
s si
gnif
ican
ce a
t 95
per
cent
con
fide
nce
- 24 shy
Table 5--Results of pooled regressions yield indicators
------------------------- Dependent variables -------------------------shyIndependent variable ---------- Yield MTHa ----------- -- Specific Yield 1000 KgM
3 H20 shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season -shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 1967 (286)
3747 (580)
1991 (293)
2800 (460 )
0084 (041)
0441 (229)
0049 (023 )
0876 (343)
System 2 indicator 0400 (155)
0525 (226)
0057 (085)
0115 (1 95)
System 3 indicator 0466 (139 )
-0029 (-010)
-0016 (-018)
0006 (008)
System 4 indicator 0980 (470)
0090 (049)
0176 (333)
0391 (827)
System 5 indicator 1 221 (570)
0256 ( 138)
0104 (174)
0224 (440)
Reservoir indicator -1033 (-578)
-0063 (-039)
-0146 (-310)
-0326 (-569)
Period indicatora
Precipitation in mm Wet season
Annual
0163 (1 03)
-29E-4 (-182)
0231 (1 42)
-58E-4 (-606)
0161 (118)
92E-5 (072)
0199 043 )
-64pound-5 (-084)
0082 (200)
57E-6 (013)
0092 (223)
-61E-5 (-232)
0032 (093)
29E-7 (001)
0052 (112)
-12pound-4 (-414)
Nitrogen applicationb 0025 (254)
0021 (211)
0025 (274)
0020 (227)
0003 (1 05)
0002 (043)
0003 (101)
-0002 (-052)
R2 (adj) 0553 0516 0251 0198 0309 0262 0713 0445 number of observations 50 50 49 49 42 42 40 40 degrees of freedom 42 45 41 44 34 37 32 35
CoeffiCient (t statistic) indicates significance at 95 confidence ~ 1982-1986 = 1
kgha
- 25 shy
Table 6--Suggmry of estiates perforance indicators
Variable Predicted Predicted Period Dependent Variable
Description (units)
Period Coefficient
1976-86 IndicatorO
1976-86 Indicatorl
Effecta (X)
BenefittedService area ratio
Wet season ratio -0036 0853 0817 -425 Dry Season 0071 0539 0610 1313
Delivery Wet Dry
season season
nm water per benefitted ha
-0808 -1 214
6003 9083
5195 7869
-1346 -1337
Rice Yield Wet season metric tons 0163 3578 3741 456 Dry season per hectare 0161 3905 4066 412
Specific Yield Wet season kg rice per m3 0082 0348 0430 2353 Dry season water de 1 ivered 0032 0378 0409 837
a percentage changes relative to period=O values
- 26 shy
ENDNOTES
IResearch Fellow International Food Policy Research Institute and Senior Irrigation Specialist International Irrigation Management Institute
2Presumable farmers also base private investment decisions on such foreknowledge of costs and commonly do so in the case of small tubewells In addition user groups may invest in small communal schemes independently after examining projected costs and benefits Most commonly however new irrigation capacity is created as a result of public investment decisions made by government and thus the political process comes into play
3This principle applies only when water charges are levied volumetrically--an exceedingly rare situation in the developing worldshy-though the argument is frequently invoked as though it applies to all allocative situations
4For Cabanatuan City the longer period of record available ie 1904 to 1986 was divided into two equal periods and tested for difference in means to test the observation by farmers that rainfall today is lower than it used to be Mean precipitation after 1947 was actually about 3 percent higher than before but the difference was not significant (t = 07)
5The procedure used was to take time series data on total nitrogenfertilizer consumption in the Philippines multiply it by the ratio of N fertilizer used on rice to total N consumption and divide the product by the rice harvested area to obtain an estimate of N use perhectare Values of the ratio of N use on rice to total N use were taken from IRRI (1988 150) with interpolation used to fill in gaps in the series
6This assumes that timeliness is evaluated against a standard based on the physiological demand for water generated by the rice crop and the varying sensitivity of the crop to stress at different growth stages In such a case a poor timeliness rating for irrigationservice will result in reductions in yield
7Because rice yields are insensitive to water applications that exceed PET values a redistribution of water from water excess to water short portions of the command area will usually result in higher aggregate output and higher average yields Exceptional scenarios can be constructed in which this general rule would be violated but such scenarios are unlikely to play out in practice
- 27 shy
aThere are a variety of standards for judging equity that could be used here The most common one is equality of per hectare water deliveries In the present case the implicit standard is equality of water supply utility in producing a crop Thus if two areas received equal amounts of water on a per hectare basis but one had such high seepage and percolation losses that the crop failed to produce a harvest (and was therefore exempted from irrigation fee payment) the water supply to the combined area would be judged inequitable Because such radical differences in seepage and percolation rates are unlikely on puddled rice soils in the lowland Philippines and because the definition of benefitted area accommodates a wide range of acceptable yields these two definitions are largely indistinguishable at the current limits of data resolution
- 18 shy
enough to date to warrant its continuation Having developed the capacity to establish targets and implement and manage change NIA is in a strong position to modify its objectives to better achieve larger social purposes established for it It is critical to recognize however that this must happen within the context of financing policies that mandate financial autonomy for NIA if the fundamental institutional commitment to manage is to be preserved
The author would like to thank Leslie Small and JeremyBerkoff for helpful comments on an earlier unpublishedversion of this paper and Charles Rogers for his careful and creative help with the analysis
- 19 shy
BIBLIOGRAPHY
Abernethy Charles L 1990 Indicators of the performance of irrigation water distribution systems International Irrigation Management Institute Colombo Sri Lanka Mimeo
Asian Development Bank 1986 Irrigation service fees Proceedingsof the Regional Seminar on Irrigation Service Fees Manila Asian Development Bank
Carruthers Ian and Colin Clark 1981 Economics of IrrigationLiverpool Liverpool University Press Third Edition
Levine G and EW Coward Jr 1986 Irrigation water distribution implications for design and operation AGREP Division WorkingPaper 125 vol 1 World Bank Agriculture and Rural Development Department
Small Les E 1989 User charges in irrigation potentials and limitations Irrigation and drainage vol 3 no 2125-142
Small Les 1990 Irrigation service fees in Asia IrrigationManagement Network 9013 London Overseas DevelopmentInstitute
Svendsen Mark and Les Small 1989 A framework for assessing irrigation system performance Paper prepared for the Symposium on Performance Evaluation 23 November International IrrigationManagement Institute Sri Lanka
Table I--National Irrigation Administration revenues and expenditures in constant prices 1976-86
Item 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986
(peso million 1972)
Revenues Irrigation fees collected 1273 1483 17 13 1831 2070 1668 1699 1893 1728 2129 2546 Other income 715 737 2420 5591 2631 5990 7783 6638 5699 4932 2934
Total direct revenue 1988 2220 4133 7422 4701 7658 9482 8531 7427 061 5480
Expenses in 1972 pricesTotal expenses 4825 5716 5039 6329 3821 77 55 6166 4749 4348 4259 4959
Excess (deficit) (2837(3496) (906) 1093 877 (097) 3316 3782 3079 2802 521 N 0
Subsidies Government operation and
maintenance subsidies 2521 2741 2799 1817 1398 633 0 0 0 0 0 Calamity fund payments 548 0 0 0 0 0 0 0 119 0 142
Total subsidy 3069 2741 2799 1817 1398 633 0 0 119 0 142
Total excess (deficit) 231 (754) 1893 2910 2275 536 3316 3782 3198 2802 663
Source IFPRI analysis of NIA data
- 21 shy
Table 2--Descriptive characteristics of selected MIA systeasa
------- Region III -------- ----- Region VI ----shyUPRIIS Angatii Sto Sibalom- Aganan-
Haasim Thomasc San Jose Sta Barbara
Average service area (hal 102272 31462 3522 5282 8703 Average irrigated area (hal
Wet season 83768 23454 3007 4410 8300 Dry season
Average benefited area Wet season
(hal 64587
77 605
27639
22908
1 781
3007
2801
4369
2770
7698 Dry season
Average rainfall Wet season
(mml d 62478
1 685 5
27396
8576
1 781
3051 0
2769
24731
2997
20001 Dry season 756 333 322 2828 3025
Average discharge (Llsec) Wet season 46501 14792 1692 2353 4984 Dry season 78091 22812 2014 1276 2315
Average water delivery (mmday) Wet season 522 548 487 462 571 Dry season
Average yield (mtha) 1089 715 995 398 686
Wet season 345 419 322 395 435 Dry season
Avg yield per unit water 34 03
(kgm ) 451 412 399 426
Wet season 0373 0440 0373 0538 0443 Dry season 0248 0400 0279 0690 0428
t-statistic difference in mean rainfall 1978-81 1982-86e
Wet season 0432 0713 -0567 1169 1169 Dry season 0519 -0230 -0523 1187 1187 Annual 0460 0707 -0686 1445 1445
~ Summary numbers are averages for the period 1982-1986 except as noted Water delivery discharge and yield per unit discharge are 4-year averages 1982-1985
c Water delivery discharge and yield per unit discharge are 4-year averages d 1983-1986
For Angat 5 years are 1981-85 For St Thomas 1979-83 For Sibalom 1971-75 e No significant differences at 95 confidence
- 22 shy
Table 3--Results of pooled regressions water delivery indicators
Dependent variable Independent variable - Water Delivery in mmBenefited Ha - --- BenefitService Area Ratio --shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season-shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 6799 (998)
6189 (1019)
13102 (1093)
2411 (1 54)
0743 ( 1186)
0800 (1165)
0129 (131 )
0258 (329)
System 2 indicator 0173 (029)
-3743 (-363)
-0063 (-219)
0200 (505)
System 3 indicator 0698 (085)
0163 (012)
0085 (208)
-0046 (-097)
System 4 indicator -0514 (-l 07)
-7011 (-803)
0080 (340)
0169 (l87)
System 5 indicator 0177 (033)
-5049 (-538)
0174 (603)
0051 (058)
Reservoir indicator 0327 (078)
5703 (543)
-0126 (-510)
0156 (261)
Period indicatora -0808 (-240)
-0771 (-229)
-1214 (-209)
-1309 (-1 62)
-0036 (-192)
-0025 (-112)
0071 (239)
0030 (077)
Water delivery mmb
Wet season
Dry season
Precipitation in mm Wet season
Annual
Average daily rainfall Wet season
Dry season
-46E-4 (-125)
-1 9E-4 (-082)
-43E-4 (-077)
20E-3 (385)
0015 (1 59)
-0001 ( -004)
0023 (215)
0001 (021)
0070 (613)
-0017 (-058)
0046 (371)
0010 (033)
R2 (adj) number of observations degrees of freedom
0156 43 36
0133 43 39
0711 41 34
0434 41 37
0693 42 34
0556 42 37
0865 40 32
0730 40 35
Coefficient (t statistic) indicates significance at 95 confidence ~ 1982-1986 1
per ha of service area
Ta
ble
--I
nd
ices
of
se
v
ice a
rea
ib
en
ef
ited
a
rea
a
nd
a
vera
ge sea
so
na
l d
isch
arg
e
Ave
rage
A
vera
ge
tshy19
77
1978
19
79
1980
19
81
1982
19
83
1984
19
85
1986
19
77-8
1 19
82-8
6 S
tati
stic
a
(ind
ex
aver
age
1983
-198
5 =
100)
Ser
vice
are
a UP
RI IS
in
dex
882
91
9
920
91
5
951
95
1
100
0 10
00
100
0 10
00
917
99
0
bull5
53
Ang
at-M
aasi
m R
95
9
994
99
7
996
99
6
996
10
00
100
0 10
00
100
0 98
8
999
1
63
Sto
To
mas
10
63
106
3 10
1 9
10
29
103
0 99
8
100
0 10
00
100
0 11
10
104
1 10
22
-08
9
Siba
lom
-San
Jos
e 95
0
872
94
2
933
94
2
942
10
28
102
8 94
4
101
7
928
99
2
292
A
gana
n-St
a B
arba
ra
108
3 10
91
106
0 10
30
961
10
05
996
10
08
996
99
6
104
5 10
00
-21
1 A
vera
ge
987
98
8
981
98
1
916
97
8
100
5 10
07
988
10
25
984
10
01
123
Wet
seas
on b
enef
ited
are
a in
dex
UPR
IIS
110
0 10
07
114
4 10
55
113
8 11
74
951
10
71
971
ll
58
10
89
106
7 -0
47
Ang
at-M
aasi
m R
97
9
974
92
0
983
10
28
100
8 99
7
102
1 98
2
931
97
7
988
0
54
Sto
To
mas
11
63
115
9 11
23
107
5 10
80
103
9 98
1
978
10
41
103
9 11
20
101
6 -4
88
Siba
lom
-San
Jos
e 11
35
103
8 10
11
931
93
4
906
10
07
975
10
1S
998
10
11
981
-0
80
Aga
nan-
Sta
Bar
bara
10
85
110
5 10
74
106
8 10
01
104
2 99
8
101
5
987
61
0
106
7 93
0
-1 8
4
Ave
rage
10
92
105
6 10
54
102
4 10
36
103
4 98
7
101
3 10
00
947
10
53
996
-2
32
Dry
sea
Son
bene
fite
d ar
ea
inde
x U
PRIIS
14
04
155
0 15
S0
155
8 16
17
128
0 57
2
114
S 15
74
152
3 12
38
-09
4 A
ngat
-Maa
sim
R
903
93
0
103
2 10
61
104
2 10
69
988
99
2
102
1 99
6
993
10
13
061
S
to
Tom
as
105
7 12
27
122
5
961
99
9
115
9 10
1 7
91
0
107
3 12
1 2
10
94
107
4 -0
28
Si
balo
m-S
an J
ose
Aga
nan-
Sta
Bar
bara
66
5
95S
62
6
632
67
4
111
4
501
10
S1
412
11
51
766
93
3
856
94
3
107
0 94
8
107
4 11
09
111
6
158
6 58
8
987
97
6
110
4 5
35
083
N
w
Ave
rage
89
6
964
11
19
103
7 10
44
110
9 10
1 7
89
8
108
5 12
97
101
7 10
81
082
Wet
seas
on d
isch
arge
in
dex
UPR
IIS
132
9 72
7
142
5 11
88
120
4 98
8
105
8 10
63
879
96
4
117
5 99
1
-16
5 A
ngat
-Maa
sim
R
129
7 13
52
134
5 12
58
127
0 11
70
120
3 62
7
131
3 10
68
-18
9
Sto
To
mas
14
71
155
0 14
1 6
10
40
725
12
35
112
3 14
79
103
1 -4
48
Siba
lom
-San
Jos
e 96
8
733
11
1 5
92
2
907
47
1
102
3 85
7
ll2
0
422
92
9
779
-1
09
Aga
nan-
Sta
Bar
bara
87
9
863
68
1
925
96
8
110
7 70
5
871
87
7
00
9
Ave
rage
11
49
105
7 13
61
115
0 10
58
853
10
43
963
99
4
SO3
11
55
940
-2
85
Dry
sea
son
disc
harg
e in
dex
UPR
IIS
425
13
09
153
3 14
28
180
6 14
S0
125
8 56
3
117
9 14
66
130
0 11
89
-04
3
Ang
at-M
aasi
m R
12
26
161
6 15
30
139
2 12
85
100
8 10
33
958
14
41
107
1 -3
80
S
to
Tom
as
116
4 14
82
155
7 79
9
971
12
30
126
5 14
01
106
6 -2
44
Si
balo
m-S
an J
ose
486
71
5
782
62
9
789
64
5
116
9 11
86
801
65
3
918
2
36
Aga
nan-
Sta
Bar
bara
A
vera
ge
425
10
46
133
6 62
0
118
4 81
5
116
1 94
7
112
5 89
7
922
12
20
991
88
2
108
7 12
36
119
2 71
8
114
0 10
37
105
5 3
23
-07
5
a bull
indi
cate
s si
gnif
ican
ce a
t 95
per
cent
con
fide
nce
- 24 shy
Table 5--Results of pooled regressions yield indicators
------------------------- Dependent variables -------------------------shyIndependent variable ---------- Yield MTHa ----------- -- Specific Yield 1000 KgM
3 H20 shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season -shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 1967 (286)
3747 (580)
1991 (293)
2800 (460 )
0084 (041)
0441 (229)
0049 (023 )
0876 (343)
System 2 indicator 0400 (155)
0525 (226)
0057 (085)
0115 (1 95)
System 3 indicator 0466 (139 )
-0029 (-010)
-0016 (-018)
0006 (008)
System 4 indicator 0980 (470)
0090 (049)
0176 (333)
0391 (827)
System 5 indicator 1 221 (570)
0256 ( 138)
0104 (174)
0224 (440)
Reservoir indicator -1033 (-578)
-0063 (-039)
-0146 (-310)
-0326 (-569)
Period indicatora
Precipitation in mm Wet season
Annual
0163 (1 03)
-29E-4 (-182)
0231 (1 42)
-58E-4 (-606)
0161 (118)
92E-5 (072)
0199 043 )
-64pound-5 (-084)
0082 (200)
57E-6 (013)
0092 (223)
-61E-5 (-232)
0032 (093)
29E-7 (001)
0052 (112)
-12pound-4 (-414)
Nitrogen applicationb 0025 (254)
0021 (211)
0025 (274)
0020 (227)
0003 (1 05)
0002 (043)
0003 (101)
-0002 (-052)
R2 (adj) 0553 0516 0251 0198 0309 0262 0713 0445 number of observations 50 50 49 49 42 42 40 40 degrees of freedom 42 45 41 44 34 37 32 35
CoeffiCient (t statistic) indicates significance at 95 confidence ~ 1982-1986 = 1
kgha
- 25 shy
Table 6--Suggmry of estiates perforance indicators
Variable Predicted Predicted Period Dependent Variable
Description (units)
Period Coefficient
1976-86 IndicatorO
1976-86 Indicatorl
Effecta (X)
BenefittedService area ratio
Wet season ratio -0036 0853 0817 -425 Dry Season 0071 0539 0610 1313
Delivery Wet Dry
season season
nm water per benefitted ha
-0808 -1 214
6003 9083
5195 7869
-1346 -1337
Rice Yield Wet season metric tons 0163 3578 3741 456 Dry season per hectare 0161 3905 4066 412
Specific Yield Wet season kg rice per m3 0082 0348 0430 2353 Dry season water de 1 ivered 0032 0378 0409 837
a percentage changes relative to period=O values
- 26 shy
ENDNOTES
IResearch Fellow International Food Policy Research Institute and Senior Irrigation Specialist International Irrigation Management Institute
2Presumable farmers also base private investment decisions on such foreknowledge of costs and commonly do so in the case of small tubewells In addition user groups may invest in small communal schemes independently after examining projected costs and benefits Most commonly however new irrigation capacity is created as a result of public investment decisions made by government and thus the political process comes into play
3This principle applies only when water charges are levied volumetrically--an exceedingly rare situation in the developing worldshy-though the argument is frequently invoked as though it applies to all allocative situations
4For Cabanatuan City the longer period of record available ie 1904 to 1986 was divided into two equal periods and tested for difference in means to test the observation by farmers that rainfall today is lower than it used to be Mean precipitation after 1947 was actually about 3 percent higher than before but the difference was not significant (t = 07)
5The procedure used was to take time series data on total nitrogenfertilizer consumption in the Philippines multiply it by the ratio of N fertilizer used on rice to total N consumption and divide the product by the rice harvested area to obtain an estimate of N use perhectare Values of the ratio of N use on rice to total N use were taken from IRRI (1988 150) with interpolation used to fill in gaps in the series
6This assumes that timeliness is evaluated against a standard based on the physiological demand for water generated by the rice crop and the varying sensitivity of the crop to stress at different growth stages In such a case a poor timeliness rating for irrigationservice will result in reductions in yield
7Because rice yields are insensitive to water applications that exceed PET values a redistribution of water from water excess to water short portions of the command area will usually result in higher aggregate output and higher average yields Exceptional scenarios can be constructed in which this general rule would be violated but such scenarios are unlikely to play out in practice
- 27 shy
aThere are a variety of standards for judging equity that could be used here The most common one is equality of per hectare water deliveries In the present case the implicit standard is equality of water supply utility in producing a crop Thus if two areas received equal amounts of water on a per hectare basis but one had such high seepage and percolation losses that the crop failed to produce a harvest (and was therefore exempted from irrigation fee payment) the water supply to the combined area would be judged inequitable Because such radical differences in seepage and percolation rates are unlikely on puddled rice soils in the lowland Philippines and because the definition of benefitted area accommodates a wide range of acceptable yields these two definitions are largely indistinguishable at the current limits of data resolution
- 19 shy
BIBLIOGRAPHY
Abernethy Charles L 1990 Indicators of the performance of irrigation water distribution systems International Irrigation Management Institute Colombo Sri Lanka Mimeo
Asian Development Bank 1986 Irrigation service fees Proceedingsof the Regional Seminar on Irrigation Service Fees Manila Asian Development Bank
Carruthers Ian and Colin Clark 1981 Economics of IrrigationLiverpool Liverpool University Press Third Edition
Levine G and EW Coward Jr 1986 Irrigation water distribution implications for design and operation AGREP Division WorkingPaper 125 vol 1 World Bank Agriculture and Rural Development Department
Small Les E 1989 User charges in irrigation potentials and limitations Irrigation and drainage vol 3 no 2125-142
Small Les 1990 Irrigation service fees in Asia IrrigationManagement Network 9013 London Overseas DevelopmentInstitute
Svendsen Mark and Les Small 1989 A framework for assessing irrigation system performance Paper prepared for the Symposium on Performance Evaluation 23 November International IrrigationManagement Institute Sri Lanka
Table I--National Irrigation Administration revenues and expenditures in constant prices 1976-86
Item 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986
(peso million 1972)
Revenues Irrigation fees collected 1273 1483 17 13 1831 2070 1668 1699 1893 1728 2129 2546 Other income 715 737 2420 5591 2631 5990 7783 6638 5699 4932 2934
Total direct revenue 1988 2220 4133 7422 4701 7658 9482 8531 7427 061 5480
Expenses in 1972 pricesTotal expenses 4825 5716 5039 6329 3821 77 55 6166 4749 4348 4259 4959
Excess (deficit) (2837(3496) (906) 1093 877 (097) 3316 3782 3079 2802 521 N 0
Subsidies Government operation and
maintenance subsidies 2521 2741 2799 1817 1398 633 0 0 0 0 0 Calamity fund payments 548 0 0 0 0 0 0 0 119 0 142
Total subsidy 3069 2741 2799 1817 1398 633 0 0 119 0 142
Total excess (deficit) 231 (754) 1893 2910 2275 536 3316 3782 3198 2802 663
Source IFPRI analysis of NIA data
- 21 shy
Table 2--Descriptive characteristics of selected MIA systeasa
------- Region III -------- ----- Region VI ----shyUPRIIS Angatii Sto Sibalom- Aganan-
Haasim Thomasc San Jose Sta Barbara
Average service area (hal 102272 31462 3522 5282 8703 Average irrigated area (hal
Wet season 83768 23454 3007 4410 8300 Dry season
Average benefited area Wet season
(hal 64587
77 605
27639
22908
1 781
3007
2801
4369
2770
7698 Dry season
Average rainfall Wet season
(mml d 62478
1 685 5
27396
8576
1 781
3051 0
2769
24731
2997
20001 Dry season 756 333 322 2828 3025
Average discharge (Llsec) Wet season 46501 14792 1692 2353 4984 Dry season 78091 22812 2014 1276 2315
Average water delivery (mmday) Wet season 522 548 487 462 571 Dry season
Average yield (mtha) 1089 715 995 398 686
Wet season 345 419 322 395 435 Dry season
Avg yield per unit water 34 03
(kgm ) 451 412 399 426
Wet season 0373 0440 0373 0538 0443 Dry season 0248 0400 0279 0690 0428
t-statistic difference in mean rainfall 1978-81 1982-86e
Wet season 0432 0713 -0567 1169 1169 Dry season 0519 -0230 -0523 1187 1187 Annual 0460 0707 -0686 1445 1445
~ Summary numbers are averages for the period 1982-1986 except as noted Water delivery discharge and yield per unit discharge are 4-year averages 1982-1985
c Water delivery discharge and yield per unit discharge are 4-year averages d 1983-1986
For Angat 5 years are 1981-85 For St Thomas 1979-83 For Sibalom 1971-75 e No significant differences at 95 confidence
- 22 shy
Table 3--Results of pooled regressions water delivery indicators
Dependent variable Independent variable - Water Delivery in mmBenefited Ha - --- BenefitService Area Ratio --shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season-shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 6799 (998)
6189 (1019)
13102 (1093)
2411 (1 54)
0743 ( 1186)
0800 (1165)
0129 (131 )
0258 (329)
System 2 indicator 0173 (029)
-3743 (-363)
-0063 (-219)
0200 (505)
System 3 indicator 0698 (085)
0163 (012)
0085 (208)
-0046 (-097)
System 4 indicator -0514 (-l 07)
-7011 (-803)
0080 (340)
0169 (l87)
System 5 indicator 0177 (033)
-5049 (-538)
0174 (603)
0051 (058)
Reservoir indicator 0327 (078)
5703 (543)
-0126 (-510)
0156 (261)
Period indicatora -0808 (-240)
-0771 (-229)
-1214 (-209)
-1309 (-1 62)
-0036 (-192)
-0025 (-112)
0071 (239)
0030 (077)
Water delivery mmb
Wet season
Dry season
Precipitation in mm Wet season
Annual
Average daily rainfall Wet season
Dry season
-46E-4 (-125)
-1 9E-4 (-082)
-43E-4 (-077)
20E-3 (385)
0015 (1 59)
-0001 ( -004)
0023 (215)
0001 (021)
0070 (613)
-0017 (-058)
0046 (371)
0010 (033)
R2 (adj) number of observations degrees of freedom
0156 43 36
0133 43 39
0711 41 34
0434 41 37
0693 42 34
0556 42 37
0865 40 32
0730 40 35
Coefficient (t statistic) indicates significance at 95 confidence ~ 1982-1986 1
per ha of service area
Ta
ble
--I
nd
ices
of
se
v
ice a
rea
ib
en
ef
ited
a
rea
a
nd
a
vera
ge sea
so
na
l d
isch
arg
e
Ave
rage
A
vera
ge
tshy19
77
1978
19
79
1980
19
81
1982
19
83
1984
19
85
1986
19
77-8
1 19
82-8
6 S
tati
stic
a
(ind
ex
aver
age
1983
-198
5 =
100)
Ser
vice
are
a UP
RI IS
in
dex
882
91
9
920
91
5
951
95
1
100
0 10
00
100
0 10
00
917
99
0
bull5
53
Ang
at-M
aasi
m R
95
9
994
99
7
996
99
6
996
10
00
100
0 10
00
100
0 98
8
999
1
63
Sto
To
mas
10
63
106
3 10
1 9
10
29
103
0 99
8
100
0 10
00
100
0 11
10
104
1 10
22
-08
9
Siba
lom
-San
Jos
e 95
0
872
94
2
933
94
2
942
10
28
102
8 94
4
101
7
928
99
2
292
A
gana
n-St
a B
arba
ra
108
3 10
91
106
0 10
30
961
10
05
996
10
08
996
99
6
104
5 10
00
-21
1 A
vera
ge
987
98
8
981
98
1
916
97
8
100
5 10
07
988
10
25
984
10
01
123
Wet
seas
on b
enef
ited
are
a in
dex
UPR
IIS
110
0 10
07
114
4 10
55
113
8 11
74
951
10
71
971
ll
58
10
89
106
7 -0
47
Ang
at-M
aasi
m R
97
9
974
92
0
983
10
28
100
8 99
7
102
1 98
2
931
97
7
988
0
54
Sto
To
mas
11
63
115
9 11
23
107
5 10
80
103
9 98
1
978
10
41
103
9 11
20
101
6 -4
88
Siba
lom
-San
Jos
e 11
35
103
8 10
11
931
93
4
906
10
07
975
10
1S
998
10
11
981
-0
80
Aga
nan-
Sta
Bar
bara
10
85
110
5 10
74
106
8 10
01
104
2 99
8
101
5
987
61
0
106
7 93
0
-1 8
4
Ave
rage
10
92
105
6 10
54
102
4 10
36
103
4 98
7
101
3 10
00
947
10
53
996
-2
32
Dry
sea
Son
bene
fite
d ar
ea
inde
x U
PRIIS
14
04
155
0 15
S0
155
8 16
17
128
0 57
2
114
S 15
74
152
3 12
38
-09
4 A
ngat
-Maa
sim
R
903
93
0
103
2 10
61
104
2 10
69
988
99
2
102
1 99
6
993
10
13
061
S
to
Tom
as
105
7 12
27
122
5
961
99
9
115
9 10
1 7
91
0
107
3 12
1 2
10
94
107
4 -0
28
Si
balo
m-S
an J
ose
Aga
nan-
Sta
Bar
bara
66
5
95S
62
6
632
67
4
111
4
501
10
S1
412
11
51
766
93
3
856
94
3
107
0 94
8
107
4 11
09
111
6
158
6 58
8
987
97
6
110
4 5
35
083
N
w
Ave
rage
89
6
964
11
19
103
7 10
44
110
9 10
1 7
89
8
108
5 12
97
101
7 10
81
082
Wet
seas
on d
isch
arge
in
dex
UPR
IIS
132
9 72
7
142
5 11
88
120
4 98
8
105
8 10
63
879
96
4
117
5 99
1
-16
5 A
ngat
-Maa
sim
R
129
7 13
52
134
5 12
58
127
0 11
70
120
3 62
7
131
3 10
68
-18
9
Sto
To
mas
14
71
155
0 14
1 6
10
40
725
12
35
112
3 14
79
103
1 -4
48
Siba
lom
-San
Jos
e 96
8
733
11
1 5
92
2
907
47
1
102
3 85
7
ll2
0
422
92
9
779
-1
09
Aga
nan-
Sta
Bar
bara
87
9
863
68
1
925
96
8
110
7 70
5
871
87
7
00
9
Ave
rage
11
49
105
7 13
61
115
0 10
58
853
10
43
963
99
4
SO3
11
55
940
-2
85
Dry
sea
son
disc
harg
e in
dex
UPR
IIS
425
13
09
153
3 14
28
180
6 14
S0
125
8 56
3
117
9 14
66
130
0 11
89
-04
3
Ang
at-M
aasi
m R
12
26
161
6 15
30
139
2 12
85
100
8 10
33
958
14
41
107
1 -3
80
S
to
Tom
as
116
4 14
82
155
7 79
9
971
12
30
126
5 14
01
106
6 -2
44
Si
balo
m-S
an J
ose
486
71
5
782
62
9
789
64
5
116
9 11
86
801
65
3
918
2
36
Aga
nan-
Sta
Bar
bara
A
vera
ge
425
10
46
133
6 62
0
118
4 81
5
116
1 94
7
112
5 89
7
922
12
20
991
88
2
108
7 12
36
119
2 71
8
114
0 10
37
105
5 3
23
-07
5
a bull
indi
cate
s si
gnif
ican
ce a
t 95
per
cent
con
fide
nce
- 24 shy
Table 5--Results of pooled regressions yield indicators
------------------------- Dependent variables -------------------------shyIndependent variable ---------- Yield MTHa ----------- -- Specific Yield 1000 KgM
3 H20 shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season -shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 1967 (286)
3747 (580)
1991 (293)
2800 (460 )
0084 (041)
0441 (229)
0049 (023 )
0876 (343)
System 2 indicator 0400 (155)
0525 (226)
0057 (085)
0115 (1 95)
System 3 indicator 0466 (139 )
-0029 (-010)
-0016 (-018)
0006 (008)
System 4 indicator 0980 (470)
0090 (049)
0176 (333)
0391 (827)
System 5 indicator 1 221 (570)
0256 ( 138)
0104 (174)
0224 (440)
Reservoir indicator -1033 (-578)
-0063 (-039)
-0146 (-310)
-0326 (-569)
Period indicatora
Precipitation in mm Wet season
Annual
0163 (1 03)
-29E-4 (-182)
0231 (1 42)
-58E-4 (-606)
0161 (118)
92E-5 (072)
0199 043 )
-64pound-5 (-084)
0082 (200)
57E-6 (013)
0092 (223)
-61E-5 (-232)
0032 (093)
29E-7 (001)
0052 (112)
-12pound-4 (-414)
Nitrogen applicationb 0025 (254)
0021 (211)
0025 (274)
0020 (227)
0003 (1 05)
0002 (043)
0003 (101)
-0002 (-052)
R2 (adj) 0553 0516 0251 0198 0309 0262 0713 0445 number of observations 50 50 49 49 42 42 40 40 degrees of freedom 42 45 41 44 34 37 32 35
CoeffiCient (t statistic) indicates significance at 95 confidence ~ 1982-1986 = 1
kgha
- 25 shy
Table 6--Suggmry of estiates perforance indicators
Variable Predicted Predicted Period Dependent Variable
Description (units)
Period Coefficient
1976-86 IndicatorO
1976-86 Indicatorl
Effecta (X)
BenefittedService area ratio
Wet season ratio -0036 0853 0817 -425 Dry Season 0071 0539 0610 1313
Delivery Wet Dry
season season
nm water per benefitted ha
-0808 -1 214
6003 9083
5195 7869
-1346 -1337
Rice Yield Wet season metric tons 0163 3578 3741 456 Dry season per hectare 0161 3905 4066 412
Specific Yield Wet season kg rice per m3 0082 0348 0430 2353 Dry season water de 1 ivered 0032 0378 0409 837
a percentage changes relative to period=O values
- 26 shy
ENDNOTES
IResearch Fellow International Food Policy Research Institute and Senior Irrigation Specialist International Irrigation Management Institute
2Presumable farmers also base private investment decisions on such foreknowledge of costs and commonly do so in the case of small tubewells In addition user groups may invest in small communal schemes independently after examining projected costs and benefits Most commonly however new irrigation capacity is created as a result of public investment decisions made by government and thus the political process comes into play
3This principle applies only when water charges are levied volumetrically--an exceedingly rare situation in the developing worldshy-though the argument is frequently invoked as though it applies to all allocative situations
4For Cabanatuan City the longer period of record available ie 1904 to 1986 was divided into two equal periods and tested for difference in means to test the observation by farmers that rainfall today is lower than it used to be Mean precipitation after 1947 was actually about 3 percent higher than before but the difference was not significant (t = 07)
5The procedure used was to take time series data on total nitrogenfertilizer consumption in the Philippines multiply it by the ratio of N fertilizer used on rice to total N consumption and divide the product by the rice harvested area to obtain an estimate of N use perhectare Values of the ratio of N use on rice to total N use were taken from IRRI (1988 150) with interpolation used to fill in gaps in the series
6This assumes that timeliness is evaluated against a standard based on the physiological demand for water generated by the rice crop and the varying sensitivity of the crop to stress at different growth stages In such a case a poor timeliness rating for irrigationservice will result in reductions in yield
7Because rice yields are insensitive to water applications that exceed PET values a redistribution of water from water excess to water short portions of the command area will usually result in higher aggregate output and higher average yields Exceptional scenarios can be constructed in which this general rule would be violated but such scenarios are unlikely to play out in practice
- 27 shy
aThere are a variety of standards for judging equity that could be used here The most common one is equality of per hectare water deliveries In the present case the implicit standard is equality of water supply utility in producing a crop Thus if two areas received equal amounts of water on a per hectare basis but one had such high seepage and percolation losses that the crop failed to produce a harvest (and was therefore exempted from irrigation fee payment) the water supply to the combined area would be judged inequitable Because such radical differences in seepage and percolation rates are unlikely on puddled rice soils in the lowland Philippines and because the definition of benefitted area accommodates a wide range of acceptable yields these two definitions are largely indistinguishable at the current limits of data resolution
Table I--National Irrigation Administration revenues and expenditures in constant prices 1976-86
Item 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986
(peso million 1972)
Revenues Irrigation fees collected 1273 1483 17 13 1831 2070 1668 1699 1893 1728 2129 2546 Other income 715 737 2420 5591 2631 5990 7783 6638 5699 4932 2934
Total direct revenue 1988 2220 4133 7422 4701 7658 9482 8531 7427 061 5480
Expenses in 1972 pricesTotal expenses 4825 5716 5039 6329 3821 77 55 6166 4749 4348 4259 4959
Excess (deficit) (2837(3496) (906) 1093 877 (097) 3316 3782 3079 2802 521 N 0
Subsidies Government operation and
maintenance subsidies 2521 2741 2799 1817 1398 633 0 0 0 0 0 Calamity fund payments 548 0 0 0 0 0 0 0 119 0 142
Total subsidy 3069 2741 2799 1817 1398 633 0 0 119 0 142
Total excess (deficit) 231 (754) 1893 2910 2275 536 3316 3782 3198 2802 663
Source IFPRI analysis of NIA data
- 21 shy
Table 2--Descriptive characteristics of selected MIA systeasa
------- Region III -------- ----- Region VI ----shyUPRIIS Angatii Sto Sibalom- Aganan-
Haasim Thomasc San Jose Sta Barbara
Average service area (hal 102272 31462 3522 5282 8703 Average irrigated area (hal
Wet season 83768 23454 3007 4410 8300 Dry season
Average benefited area Wet season
(hal 64587
77 605
27639
22908
1 781
3007
2801
4369
2770
7698 Dry season
Average rainfall Wet season
(mml d 62478
1 685 5
27396
8576
1 781
3051 0
2769
24731
2997
20001 Dry season 756 333 322 2828 3025
Average discharge (Llsec) Wet season 46501 14792 1692 2353 4984 Dry season 78091 22812 2014 1276 2315
Average water delivery (mmday) Wet season 522 548 487 462 571 Dry season
Average yield (mtha) 1089 715 995 398 686
Wet season 345 419 322 395 435 Dry season
Avg yield per unit water 34 03
(kgm ) 451 412 399 426
Wet season 0373 0440 0373 0538 0443 Dry season 0248 0400 0279 0690 0428
t-statistic difference in mean rainfall 1978-81 1982-86e
Wet season 0432 0713 -0567 1169 1169 Dry season 0519 -0230 -0523 1187 1187 Annual 0460 0707 -0686 1445 1445
~ Summary numbers are averages for the period 1982-1986 except as noted Water delivery discharge and yield per unit discharge are 4-year averages 1982-1985
c Water delivery discharge and yield per unit discharge are 4-year averages d 1983-1986
For Angat 5 years are 1981-85 For St Thomas 1979-83 For Sibalom 1971-75 e No significant differences at 95 confidence
- 22 shy
Table 3--Results of pooled regressions water delivery indicators
Dependent variable Independent variable - Water Delivery in mmBenefited Ha - --- BenefitService Area Ratio --shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season-shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 6799 (998)
6189 (1019)
13102 (1093)
2411 (1 54)
0743 ( 1186)
0800 (1165)
0129 (131 )
0258 (329)
System 2 indicator 0173 (029)
-3743 (-363)
-0063 (-219)
0200 (505)
System 3 indicator 0698 (085)
0163 (012)
0085 (208)
-0046 (-097)
System 4 indicator -0514 (-l 07)
-7011 (-803)
0080 (340)
0169 (l87)
System 5 indicator 0177 (033)
-5049 (-538)
0174 (603)
0051 (058)
Reservoir indicator 0327 (078)
5703 (543)
-0126 (-510)
0156 (261)
Period indicatora -0808 (-240)
-0771 (-229)
-1214 (-209)
-1309 (-1 62)
-0036 (-192)
-0025 (-112)
0071 (239)
0030 (077)
Water delivery mmb
Wet season
Dry season
Precipitation in mm Wet season
Annual
Average daily rainfall Wet season
Dry season
-46E-4 (-125)
-1 9E-4 (-082)
-43E-4 (-077)
20E-3 (385)
0015 (1 59)
-0001 ( -004)
0023 (215)
0001 (021)
0070 (613)
-0017 (-058)
0046 (371)
0010 (033)
R2 (adj) number of observations degrees of freedom
0156 43 36
0133 43 39
0711 41 34
0434 41 37
0693 42 34
0556 42 37
0865 40 32
0730 40 35
Coefficient (t statistic) indicates significance at 95 confidence ~ 1982-1986 1
per ha of service area
Ta
ble
--I
nd
ices
of
se
v
ice a
rea
ib
en
ef
ited
a
rea
a
nd
a
vera
ge sea
so
na
l d
isch
arg
e
Ave
rage
A
vera
ge
tshy19
77
1978
19
79
1980
19
81
1982
19
83
1984
19
85
1986
19
77-8
1 19
82-8
6 S
tati
stic
a
(ind
ex
aver
age
1983
-198
5 =
100)
Ser
vice
are
a UP
RI IS
in
dex
882
91
9
920
91
5
951
95
1
100
0 10
00
100
0 10
00
917
99
0
bull5
53
Ang
at-M
aasi
m R
95
9
994
99
7
996
99
6
996
10
00
100
0 10
00
100
0 98
8
999
1
63
Sto
To
mas
10
63
106
3 10
1 9
10
29
103
0 99
8
100
0 10
00
100
0 11
10
104
1 10
22
-08
9
Siba
lom
-San
Jos
e 95
0
872
94
2
933
94
2
942
10
28
102
8 94
4
101
7
928
99
2
292
A
gana
n-St
a B
arba
ra
108
3 10
91
106
0 10
30
961
10
05
996
10
08
996
99
6
104
5 10
00
-21
1 A
vera
ge
987
98
8
981
98
1
916
97
8
100
5 10
07
988
10
25
984
10
01
123
Wet
seas
on b
enef
ited
are
a in
dex
UPR
IIS
110
0 10
07
114
4 10
55
113
8 11
74
951
10
71
971
ll
58
10
89
106
7 -0
47
Ang
at-M
aasi
m R
97
9
974
92
0
983
10
28
100
8 99
7
102
1 98
2
931
97
7
988
0
54
Sto
To
mas
11
63
115
9 11
23
107
5 10
80
103
9 98
1
978
10
41
103
9 11
20
101
6 -4
88
Siba
lom
-San
Jos
e 11
35
103
8 10
11
931
93
4
906
10
07
975
10
1S
998
10
11
981
-0
80
Aga
nan-
Sta
Bar
bara
10
85
110
5 10
74
106
8 10
01
104
2 99
8
101
5
987
61
0
106
7 93
0
-1 8
4
Ave
rage
10
92
105
6 10
54
102
4 10
36
103
4 98
7
101
3 10
00
947
10
53
996
-2
32
Dry
sea
Son
bene
fite
d ar
ea
inde
x U
PRIIS
14
04
155
0 15
S0
155
8 16
17
128
0 57
2
114
S 15
74
152
3 12
38
-09
4 A
ngat
-Maa
sim
R
903
93
0
103
2 10
61
104
2 10
69
988
99
2
102
1 99
6
993
10
13
061
S
to
Tom
as
105
7 12
27
122
5
961
99
9
115
9 10
1 7
91
0
107
3 12
1 2
10
94
107
4 -0
28
Si
balo
m-S
an J
ose
Aga
nan-
Sta
Bar
bara
66
5
95S
62
6
632
67
4
111
4
501
10
S1
412
11
51
766
93
3
856
94
3
107
0 94
8
107
4 11
09
111
6
158
6 58
8
987
97
6
110
4 5
35
083
N
w
Ave
rage
89
6
964
11
19
103
7 10
44
110
9 10
1 7
89
8
108
5 12
97
101
7 10
81
082
Wet
seas
on d
isch
arge
in
dex
UPR
IIS
132
9 72
7
142
5 11
88
120
4 98
8
105
8 10
63
879
96
4
117
5 99
1
-16
5 A
ngat
-Maa
sim
R
129
7 13
52
134
5 12
58
127
0 11
70
120
3 62
7
131
3 10
68
-18
9
Sto
To
mas
14
71
155
0 14
1 6
10
40
725
12
35
112
3 14
79
103
1 -4
48
Siba
lom
-San
Jos
e 96
8
733
11
1 5
92
2
907
47
1
102
3 85
7
ll2
0
422
92
9
779
-1
09
Aga
nan-
Sta
Bar
bara
87
9
863
68
1
925
96
8
110
7 70
5
871
87
7
00
9
Ave
rage
11
49
105
7 13
61
115
0 10
58
853
10
43
963
99
4
SO3
11
55
940
-2
85
Dry
sea
son
disc
harg
e in
dex
UPR
IIS
425
13
09
153
3 14
28
180
6 14
S0
125
8 56
3
117
9 14
66
130
0 11
89
-04
3
Ang
at-M
aasi
m R
12
26
161
6 15
30
139
2 12
85
100
8 10
33
958
14
41
107
1 -3
80
S
to
Tom
as
116
4 14
82
155
7 79
9
971
12
30
126
5 14
01
106
6 -2
44
Si
balo
m-S
an J
ose
486
71
5
782
62
9
789
64
5
116
9 11
86
801
65
3
918
2
36
Aga
nan-
Sta
Bar
bara
A
vera
ge
425
10
46
133
6 62
0
118
4 81
5
116
1 94
7
112
5 89
7
922
12
20
991
88
2
108
7 12
36
119
2 71
8
114
0 10
37
105
5 3
23
-07
5
a bull
indi
cate
s si
gnif
ican
ce a
t 95
per
cent
con
fide
nce
- 24 shy
Table 5--Results of pooled regressions yield indicators
------------------------- Dependent variables -------------------------shyIndependent variable ---------- Yield MTHa ----------- -- Specific Yield 1000 KgM
3 H20 shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season -shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 1967 (286)
3747 (580)
1991 (293)
2800 (460 )
0084 (041)
0441 (229)
0049 (023 )
0876 (343)
System 2 indicator 0400 (155)
0525 (226)
0057 (085)
0115 (1 95)
System 3 indicator 0466 (139 )
-0029 (-010)
-0016 (-018)
0006 (008)
System 4 indicator 0980 (470)
0090 (049)
0176 (333)
0391 (827)
System 5 indicator 1 221 (570)
0256 ( 138)
0104 (174)
0224 (440)
Reservoir indicator -1033 (-578)
-0063 (-039)
-0146 (-310)
-0326 (-569)
Period indicatora
Precipitation in mm Wet season
Annual
0163 (1 03)
-29E-4 (-182)
0231 (1 42)
-58E-4 (-606)
0161 (118)
92E-5 (072)
0199 043 )
-64pound-5 (-084)
0082 (200)
57E-6 (013)
0092 (223)
-61E-5 (-232)
0032 (093)
29E-7 (001)
0052 (112)
-12pound-4 (-414)
Nitrogen applicationb 0025 (254)
0021 (211)
0025 (274)
0020 (227)
0003 (1 05)
0002 (043)
0003 (101)
-0002 (-052)
R2 (adj) 0553 0516 0251 0198 0309 0262 0713 0445 number of observations 50 50 49 49 42 42 40 40 degrees of freedom 42 45 41 44 34 37 32 35
CoeffiCient (t statistic) indicates significance at 95 confidence ~ 1982-1986 = 1
kgha
- 25 shy
Table 6--Suggmry of estiates perforance indicators
Variable Predicted Predicted Period Dependent Variable
Description (units)
Period Coefficient
1976-86 IndicatorO
1976-86 Indicatorl
Effecta (X)
BenefittedService area ratio
Wet season ratio -0036 0853 0817 -425 Dry Season 0071 0539 0610 1313
Delivery Wet Dry
season season
nm water per benefitted ha
-0808 -1 214
6003 9083
5195 7869
-1346 -1337
Rice Yield Wet season metric tons 0163 3578 3741 456 Dry season per hectare 0161 3905 4066 412
Specific Yield Wet season kg rice per m3 0082 0348 0430 2353 Dry season water de 1 ivered 0032 0378 0409 837
a percentage changes relative to period=O values
- 26 shy
ENDNOTES
IResearch Fellow International Food Policy Research Institute and Senior Irrigation Specialist International Irrigation Management Institute
2Presumable farmers also base private investment decisions on such foreknowledge of costs and commonly do so in the case of small tubewells In addition user groups may invest in small communal schemes independently after examining projected costs and benefits Most commonly however new irrigation capacity is created as a result of public investment decisions made by government and thus the political process comes into play
3This principle applies only when water charges are levied volumetrically--an exceedingly rare situation in the developing worldshy-though the argument is frequently invoked as though it applies to all allocative situations
4For Cabanatuan City the longer period of record available ie 1904 to 1986 was divided into two equal periods and tested for difference in means to test the observation by farmers that rainfall today is lower than it used to be Mean precipitation after 1947 was actually about 3 percent higher than before but the difference was not significant (t = 07)
5The procedure used was to take time series data on total nitrogenfertilizer consumption in the Philippines multiply it by the ratio of N fertilizer used on rice to total N consumption and divide the product by the rice harvested area to obtain an estimate of N use perhectare Values of the ratio of N use on rice to total N use were taken from IRRI (1988 150) with interpolation used to fill in gaps in the series
6This assumes that timeliness is evaluated against a standard based on the physiological demand for water generated by the rice crop and the varying sensitivity of the crop to stress at different growth stages In such a case a poor timeliness rating for irrigationservice will result in reductions in yield
7Because rice yields are insensitive to water applications that exceed PET values a redistribution of water from water excess to water short portions of the command area will usually result in higher aggregate output and higher average yields Exceptional scenarios can be constructed in which this general rule would be violated but such scenarios are unlikely to play out in practice
- 27 shy
aThere are a variety of standards for judging equity that could be used here The most common one is equality of per hectare water deliveries In the present case the implicit standard is equality of water supply utility in producing a crop Thus if two areas received equal amounts of water on a per hectare basis but one had such high seepage and percolation losses that the crop failed to produce a harvest (and was therefore exempted from irrigation fee payment) the water supply to the combined area would be judged inequitable Because such radical differences in seepage and percolation rates are unlikely on puddled rice soils in the lowland Philippines and because the definition of benefitted area accommodates a wide range of acceptable yields these two definitions are largely indistinguishable at the current limits of data resolution
- 21 shy
Table 2--Descriptive characteristics of selected MIA systeasa
------- Region III -------- ----- Region VI ----shyUPRIIS Angatii Sto Sibalom- Aganan-
Haasim Thomasc San Jose Sta Barbara
Average service area (hal 102272 31462 3522 5282 8703 Average irrigated area (hal
Wet season 83768 23454 3007 4410 8300 Dry season
Average benefited area Wet season
(hal 64587
77 605
27639
22908
1 781
3007
2801
4369
2770
7698 Dry season
Average rainfall Wet season
(mml d 62478
1 685 5
27396
8576
1 781
3051 0
2769
24731
2997
20001 Dry season 756 333 322 2828 3025
Average discharge (Llsec) Wet season 46501 14792 1692 2353 4984 Dry season 78091 22812 2014 1276 2315
Average water delivery (mmday) Wet season 522 548 487 462 571 Dry season
Average yield (mtha) 1089 715 995 398 686
Wet season 345 419 322 395 435 Dry season
Avg yield per unit water 34 03
(kgm ) 451 412 399 426
Wet season 0373 0440 0373 0538 0443 Dry season 0248 0400 0279 0690 0428
t-statistic difference in mean rainfall 1978-81 1982-86e
Wet season 0432 0713 -0567 1169 1169 Dry season 0519 -0230 -0523 1187 1187 Annual 0460 0707 -0686 1445 1445
~ Summary numbers are averages for the period 1982-1986 except as noted Water delivery discharge and yield per unit discharge are 4-year averages 1982-1985
c Water delivery discharge and yield per unit discharge are 4-year averages d 1983-1986
For Angat 5 years are 1981-85 For St Thomas 1979-83 For Sibalom 1971-75 e No significant differences at 95 confidence
- 22 shy
Table 3--Results of pooled regressions water delivery indicators
Dependent variable Independent variable - Water Delivery in mmBenefited Ha - --- BenefitService Area Ratio --shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season-shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 6799 (998)
6189 (1019)
13102 (1093)
2411 (1 54)
0743 ( 1186)
0800 (1165)
0129 (131 )
0258 (329)
System 2 indicator 0173 (029)
-3743 (-363)
-0063 (-219)
0200 (505)
System 3 indicator 0698 (085)
0163 (012)
0085 (208)
-0046 (-097)
System 4 indicator -0514 (-l 07)
-7011 (-803)
0080 (340)
0169 (l87)
System 5 indicator 0177 (033)
-5049 (-538)
0174 (603)
0051 (058)
Reservoir indicator 0327 (078)
5703 (543)
-0126 (-510)
0156 (261)
Period indicatora -0808 (-240)
-0771 (-229)
-1214 (-209)
-1309 (-1 62)
-0036 (-192)
-0025 (-112)
0071 (239)
0030 (077)
Water delivery mmb
Wet season
Dry season
Precipitation in mm Wet season
Annual
Average daily rainfall Wet season
Dry season
-46E-4 (-125)
-1 9E-4 (-082)
-43E-4 (-077)
20E-3 (385)
0015 (1 59)
-0001 ( -004)
0023 (215)
0001 (021)
0070 (613)
-0017 (-058)
0046 (371)
0010 (033)
R2 (adj) number of observations degrees of freedom
0156 43 36
0133 43 39
0711 41 34
0434 41 37
0693 42 34
0556 42 37
0865 40 32
0730 40 35
Coefficient (t statistic) indicates significance at 95 confidence ~ 1982-1986 1
per ha of service area
Ta
ble
--I
nd
ices
of
se
v
ice a
rea
ib
en
ef
ited
a
rea
a
nd
a
vera
ge sea
so
na
l d
isch
arg
e
Ave
rage
A
vera
ge
tshy19
77
1978
19
79
1980
19
81
1982
19
83
1984
19
85
1986
19
77-8
1 19
82-8
6 S
tati
stic
a
(ind
ex
aver
age
1983
-198
5 =
100)
Ser
vice
are
a UP
RI IS
in
dex
882
91
9
920
91
5
951
95
1
100
0 10
00
100
0 10
00
917
99
0
bull5
53
Ang
at-M
aasi
m R
95
9
994
99
7
996
99
6
996
10
00
100
0 10
00
100
0 98
8
999
1
63
Sto
To
mas
10
63
106
3 10
1 9
10
29
103
0 99
8
100
0 10
00
100
0 11
10
104
1 10
22
-08
9
Siba
lom
-San
Jos
e 95
0
872
94
2
933
94
2
942
10
28
102
8 94
4
101
7
928
99
2
292
A
gana
n-St
a B
arba
ra
108
3 10
91
106
0 10
30
961
10
05
996
10
08
996
99
6
104
5 10
00
-21
1 A
vera
ge
987
98
8
981
98
1
916
97
8
100
5 10
07
988
10
25
984
10
01
123
Wet
seas
on b
enef
ited
are
a in
dex
UPR
IIS
110
0 10
07
114
4 10
55
113
8 11
74
951
10
71
971
ll
58
10
89
106
7 -0
47
Ang
at-M
aasi
m R
97
9
974
92
0
983
10
28
100
8 99
7
102
1 98
2
931
97
7
988
0
54
Sto
To
mas
11
63
115
9 11
23
107
5 10
80
103
9 98
1
978
10
41
103
9 11
20
101
6 -4
88
Siba
lom
-San
Jos
e 11
35
103
8 10
11
931
93
4
906
10
07
975
10
1S
998
10
11
981
-0
80
Aga
nan-
Sta
Bar
bara
10
85
110
5 10
74
106
8 10
01
104
2 99
8
101
5
987
61
0
106
7 93
0
-1 8
4
Ave
rage
10
92
105
6 10
54
102
4 10
36
103
4 98
7
101
3 10
00
947
10
53
996
-2
32
Dry
sea
Son
bene
fite
d ar
ea
inde
x U
PRIIS
14
04
155
0 15
S0
155
8 16
17
128
0 57
2
114
S 15
74
152
3 12
38
-09
4 A
ngat
-Maa
sim
R
903
93
0
103
2 10
61
104
2 10
69
988
99
2
102
1 99
6
993
10
13
061
S
to
Tom
as
105
7 12
27
122
5
961
99
9
115
9 10
1 7
91
0
107
3 12
1 2
10
94
107
4 -0
28
Si
balo
m-S
an J
ose
Aga
nan-
Sta
Bar
bara
66
5
95S
62
6
632
67
4
111
4
501
10
S1
412
11
51
766
93
3
856
94
3
107
0 94
8
107
4 11
09
111
6
158
6 58
8
987
97
6
110
4 5
35
083
N
w
Ave
rage
89
6
964
11
19
103
7 10
44
110
9 10
1 7
89
8
108
5 12
97
101
7 10
81
082
Wet
seas
on d
isch
arge
in
dex
UPR
IIS
132
9 72
7
142
5 11
88
120
4 98
8
105
8 10
63
879
96
4
117
5 99
1
-16
5 A
ngat
-Maa
sim
R
129
7 13
52
134
5 12
58
127
0 11
70
120
3 62
7
131
3 10
68
-18
9
Sto
To
mas
14
71
155
0 14
1 6
10
40
725
12
35
112
3 14
79
103
1 -4
48
Siba
lom
-San
Jos
e 96
8
733
11
1 5
92
2
907
47
1
102
3 85
7
ll2
0
422
92
9
779
-1
09
Aga
nan-
Sta
Bar
bara
87
9
863
68
1
925
96
8
110
7 70
5
871
87
7
00
9
Ave
rage
11
49
105
7 13
61
115
0 10
58
853
10
43
963
99
4
SO3
11
55
940
-2
85
Dry
sea
son
disc
harg
e in
dex
UPR
IIS
425
13
09
153
3 14
28
180
6 14
S0
125
8 56
3
117
9 14
66
130
0 11
89
-04
3
Ang
at-M
aasi
m R
12
26
161
6 15
30
139
2 12
85
100
8 10
33
958
14
41
107
1 -3
80
S
to
Tom
as
116
4 14
82
155
7 79
9
971
12
30
126
5 14
01
106
6 -2
44
Si
balo
m-S
an J
ose
486
71
5
782
62
9
789
64
5
116
9 11
86
801
65
3
918
2
36
Aga
nan-
Sta
Bar
bara
A
vera
ge
425
10
46
133
6 62
0
118
4 81
5
116
1 94
7
112
5 89
7
922
12
20
991
88
2
108
7 12
36
119
2 71
8
114
0 10
37
105
5 3
23
-07
5
a bull
indi
cate
s si
gnif
ican
ce a
t 95
per
cent
con
fide
nce
- 24 shy
Table 5--Results of pooled regressions yield indicators
------------------------- Dependent variables -------------------------shyIndependent variable ---------- Yield MTHa ----------- -- Specific Yield 1000 KgM
3 H20 shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season -shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 1967 (286)
3747 (580)
1991 (293)
2800 (460 )
0084 (041)
0441 (229)
0049 (023 )
0876 (343)
System 2 indicator 0400 (155)
0525 (226)
0057 (085)
0115 (1 95)
System 3 indicator 0466 (139 )
-0029 (-010)
-0016 (-018)
0006 (008)
System 4 indicator 0980 (470)
0090 (049)
0176 (333)
0391 (827)
System 5 indicator 1 221 (570)
0256 ( 138)
0104 (174)
0224 (440)
Reservoir indicator -1033 (-578)
-0063 (-039)
-0146 (-310)
-0326 (-569)
Period indicatora
Precipitation in mm Wet season
Annual
0163 (1 03)
-29E-4 (-182)
0231 (1 42)
-58E-4 (-606)
0161 (118)
92E-5 (072)
0199 043 )
-64pound-5 (-084)
0082 (200)
57E-6 (013)
0092 (223)
-61E-5 (-232)
0032 (093)
29E-7 (001)
0052 (112)
-12pound-4 (-414)
Nitrogen applicationb 0025 (254)
0021 (211)
0025 (274)
0020 (227)
0003 (1 05)
0002 (043)
0003 (101)
-0002 (-052)
R2 (adj) 0553 0516 0251 0198 0309 0262 0713 0445 number of observations 50 50 49 49 42 42 40 40 degrees of freedom 42 45 41 44 34 37 32 35
CoeffiCient (t statistic) indicates significance at 95 confidence ~ 1982-1986 = 1
kgha
- 25 shy
Table 6--Suggmry of estiates perforance indicators
Variable Predicted Predicted Period Dependent Variable
Description (units)
Period Coefficient
1976-86 IndicatorO
1976-86 Indicatorl
Effecta (X)
BenefittedService area ratio
Wet season ratio -0036 0853 0817 -425 Dry Season 0071 0539 0610 1313
Delivery Wet Dry
season season
nm water per benefitted ha
-0808 -1 214
6003 9083
5195 7869
-1346 -1337
Rice Yield Wet season metric tons 0163 3578 3741 456 Dry season per hectare 0161 3905 4066 412
Specific Yield Wet season kg rice per m3 0082 0348 0430 2353 Dry season water de 1 ivered 0032 0378 0409 837
a percentage changes relative to period=O values
- 26 shy
ENDNOTES
IResearch Fellow International Food Policy Research Institute and Senior Irrigation Specialist International Irrigation Management Institute
2Presumable farmers also base private investment decisions on such foreknowledge of costs and commonly do so in the case of small tubewells In addition user groups may invest in small communal schemes independently after examining projected costs and benefits Most commonly however new irrigation capacity is created as a result of public investment decisions made by government and thus the political process comes into play
3This principle applies only when water charges are levied volumetrically--an exceedingly rare situation in the developing worldshy-though the argument is frequently invoked as though it applies to all allocative situations
4For Cabanatuan City the longer period of record available ie 1904 to 1986 was divided into two equal periods and tested for difference in means to test the observation by farmers that rainfall today is lower than it used to be Mean precipitation after 1947 was actually about 3 percent higher than before but the difference was not significant (t = 07)
5The procedure used was to take time series data on total nitrogenfertilizer consumption in the Philippines multiply it by the ratio of N fertilizer used on rice to total N consumption and divide the product by the rice harvested area to obtain an estimate of N use perhectare Values of the ratio of N use on rice to total N use were taken from IRRI (1988 150) with interpolation used to fill in gaps in the series
6This assumes that timeliness is evaluated against a standard based on the physiological demand for water generated by the rice crop and the varying sensitivity of the crop to stress at different growth stages In such a case a poor timeliness rating for irrigationservice will result in reductions in yield
7Because rice yields are insensitive to water applications that exceed PET values a redistribution of water from water excess to water short portions of the command area will usually result in higher aggregate output and higher average yields Exceptional scenarios can be constructed in which this general rule would be violated but such scenarios are unlikely to play out in practice
- 27 shy
aThere are a variety of standards for judging equity that could be used here The most common one is equality of per hectare water deliveries In the present case the implicit standard is equality of water supply utility in producing a crop Thus if two areas received equal amounts of water on a per hectare basis but one had such high seepage and percolation losses that the crop failed to produce a harvest (and was therefore exempted from irrigation fee payment) the water supply to the combined area would be judged inequitable Because such radical differences in seepage and percolation rates are unlikely on puddled rice soils in the lowland Philippines and because the definition of benefitted area accommodates a wide range of acceptable yields these two definitions are largely indistinguishable at the current limits of data resolution
- 22 shy
Table 3--Results of pooled regressions water delivery indicators
Dependent variable Independent variable - Water Delivery in mmBenefited Ha - --- BenefitService Area Ratio --shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season-shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 6799 (998)
6189 (1019)
13102 (1093)
2411 (1 54)
0743 ( 1186)
0800 (1165)
0129 (131 )
0258 (329)
System 2 indicator 0173 (029)
-3743 (-363)
-0063 (-219)
0200 (505)
System 3 indicator 0698 (085)
0163 (012)
0085 (208)
-0046 (-097)
System 4 indicator -0514 (-l 07)
-7011 (-803)
0080 (340)
0169 (l87)
System 5 indicator 0177 (033)
-5049 (-538)
0174 (603)
0051 (058)
Reservoir indicator 0327 (078)
5703 (543)
-0126 (-510)
0156 (261)
Period indicatora -0808 (-240)
-0771 (-229)
-1214 (-209)
-1309 (-1 62)
-0036 (-192)
-0025 (-112)
0071 (239)
0030 (077)
Water delivery mmb
Wet season
Dry season
Precipitation in mm Wet season
Annual
Average daily rainfall Wet season
Dry season
-46E-4 (-125)
-1 9E-4 (-082)
-43E-4 (-077)
20E-3 (385)
0015 (1 59)
-0001 ( -004)
0023 (215)
0001 (021)
0070 (613)
-0017 (-058)
0046 (371)
0010 (033)
R2 (adj) number of observations degrees of freedom
0156 43 36
0133 43 39
0711 41 34
0434 41 37
0693 42 34
0556 42 37
0865 40 32
0730 40 35
Coefficient (t statistic) indicates significance at 95 confidence ~ 1982-1986 1
per ha of service area
Ta
ble
--I
nd
ices
of
se
v
ice a
rea
ib
en
ef
ited
a
rea
a
nd
a
vera
ge sea
so
na
l d
isch
arg
e
Ave
rage
A
vera
ge
tshy19
77
1978
19
79
1980
19
81
1982
19
83
1984
19
85
1986
19
77-8
1 19
82-8
6 S
tati
stic
a
(ind
ex
aver
age
1983
-198
5 =
100)
Ser
vice
are
a UP
RI IS
in
dex
882
91
9
920
91
5
951
95
1
100
0 10
00
100
0 10
00
917
99
0
bull5
53
Ang
at-M
aasi
m R
95
9
994
99
7
996
99
6
996
10
00
100
0 10
00
100
0 98
8
999
1
63
Sto
To
mas
10
63
106
3 10
1 9
10
29
103
0 99
8
100
0 10
00
100
0 11
10
104
1 10
22
-08
9
Siba
lom
-San
Jos
e 95
0
872
94
2
933
94
2
942
10
28
102
8 94
4
101
7
928
99
2
292
A
gana
n-St
a B
arba
ra
108
3 10
91
106
0 10
30
961
10
05
996
10
08
996
99
6
104
5 10
00
-21
1 A
vera
ge
987
98
8
981
98
1
916
97
8
100
5 10
07
988
10
25
984
10
01
123
Wet
seas
on b
enef
ited
are
a in
dex
UPR
IIS
110
0 10
07
114
4 10
55
113
8 11
74
951
10
71
971
ll
58
10
89
106
7 -0
47
Ang
at-M
aasi
m R
97
9
974
92
0
983
10
28
100
8 99
7
102
1 98
2
931
97
7
988
0
54
Sto
To
mas
11
63
115
9 11
23
107
5 10
80
103
9 98
1
978
10
41
103
9 11
20
101
6 -4
88
Siba
lom
-San
Jos
e 11
35
103
8 10
11
931
93
4
906
10
07
975
10
1S
998
10
11
981
-0
80
Aga
nan-
Sta
Bar
bara
10
85
110
5 10
74
106
8 10
01
104
2 99
8
101
5
987
61
0
106
7 93
0
-1 8
4
Ave
rage
10
92
105
6 10
54
102
4 10
36
103
4 98
7
101
3 10
00
947
10
53
996
-2
32
Dry
sea
Son
bene
fite
d ar
ea
inde
x U
PRIIS
14
04
155
0 15
S0
155
8 16
17
128
0 57
2
114
S 15
74
152
3 12
38
-09
4 A
ngat
-Maa
sim
R
903
93
0
103
2 10
61
104
2 10
69
988
99
2
102
1 99
6
993
10
13
061
S
to
Tom
as
105
7 12
27
122
5
961
99
9
115
9 10
1 7
91
0
107
3 12
1 2
10
94
107
4 -0
28
Si
balo
m-S
an J
ose
Aga
nan-
Sta
Bar
bara
66
5
95S
62
6
632
67
4
111
4
501
10
S1
412
11
51
766
93
3
856
94
3
107
0 94
8
107
4 11
09
111
6
158
6 58
8
987
97
6
110
4 5
35
083
N
w
Ave
rage
89
6
964
11
19
103
7 10
44
110
9 10
1 7
89
8
108
5 12
97
101
7 10
81
082
Wet
seas
on d
isch
arge
in
dex
UPR
IIS
132
9 72
7
142
5 11
88
120
4 98
8
105
8 10
63
879
96
4
117
5 99
1
-16
5 A
ngat
-Maa
sim
R
129
7 13
52
134
5 12
58
127
0 11
70
120
3 62
7
131
3 10
68
-18
9
Sto
To
mas
14
71
155
0 14
1 6
10
40
725
12
35
112
3 14
79
103
1 -4
48
Siba
lom
-San
Jos
e 96
8
733
11
1 5
92
2
907
47
1
102
3 85
7
ll2
0
422
92
9
779
-1
09
Aga
nan-
Sta
Bar
bara
87
9
863
68
1
925
96
8
110
7 70
5
871
87
7
00
9
Ave
rage
11
49
105
7 13
61
115
0 10
58
853
10
43
963
99
4
SO3
11
55
940
-2
85
Dry
sea
son
disc
harg
e in
dex
UPR
IIS
425
13
09
153
3 14
28
180
6 14
S0
125
8 56
3
117
9 14
66
130
0 11
89
-04
3
Ang
at-M
aasi
m R
12
26
161
6 15
30
139
2 12
85
100
8 10
33
958
14
41
107
1 -3
80
S
to
Tom
as
116
4 14
82
155
7 79
9
971
12
30
126
5 14
01
106
6 -2
44
Si
balo
m-S
an J
ose
486
71
5
782
62
9
789
64
5
116
9 11
86
801
65
3
918
2
36
Aga
nan-
Sta
Bar
bara
A
vera
ge
425
10
46
133
6 62
0
118
4 81
5
116
1 94
7
112
5 89
7
922
12
20
991
88
2
108
7 12
36
119
2 71
8
114
0 10
37
105
5 3
23
-07
5
a bull
indi
cate
s si
gnif
ican
ce a
t 95
per
cent
con
fide
nce
- 24 shy
Table 5--Results of pooled regressions yield indicators
------------------------- Dependent variables -------------------------shyIndependent variable ---------- Yield MTHa ----------- -- Specific Yield 1000 KgM
3 H20 shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season -shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 1967 (286)
3747 (580)
1991 (293)
2800 (460 )
0084 (041)
0441 (229)
0049 (023 )
0876 (343)
System 2 indicator 0400 (155)
0525 (226)
0057 (085)
0115 (1 95)
System 3 indicator 0466 (139 )
-0029 (-010)
-0016 (-018)
0006 (008)
System 4 indicator 0980 (470)
0090 (049)
0176 (333)
0391 (827)
System 5 indicator 1 221 (570)
0256 ( 138)
0104 (174)
0224 (440)
Reservoir indicator -1033 (-578)
-0063 (-039)
-0146 (-310)
-0326 (-569)
Period indicatora
Precipitation in mm Wet season
Annual
0163 (1 03)
-29E-4 (-182)
0231 (1 42)
-58E-4 (-606)
0161 (118)
92E-5 (072)
0199 043 )
-64pound-5 (-084)
0082 (200)
57E-6 (013)
0092 (223)
-61E-5 (-232)
0032 (093)
29E-7 (001)
0052 (112)
-12pound-4 (-414)
Nitrogen applicationb 0025 (254)
0021 (211)
0025 (274)
0020 (227)
0003 (1 05)
0002 (043)
0003 (101)
-0002 (-052)
R2 (adj) 0553 0516 0251 0198 0309 0262 0713 0445 number of observations 50 50 49 49 42 42 40 40 degrees of freedom 42 45 41 44 34 37 32 35
CoeffiCient (t statistic) indicates significance at 95 confidence ~ 1982-1986 = 1
kgha
- 25 shy
Table 6--Suggmry of estiates perforance indicators
Variable Predicted Predicted Period Dependent Variable
Description (units)
Period Coefficient
1976-86 IndicatorO
1976-86 Indicatorl
Effecta (X)
BenefittedService area ratio
Wet season ratio -0036 0853 0817 -425 Dry Season 0071 0539 0610 1313
Delivery Wet Dry
season season
nm water per benefitted ha
-0808 -1 214
6003 9083
5195 7869
-1346 -1337
Rice Yield Wet season metric tons 0163 3578 3741 456 Dry season per hectare 0161 3905 4066 412
Specific Yield Wet season kg rice per m3 0082 0348 0430 2353 Dry season water de 1 ivered 0032 0378 0409 837
a percentage changes relative to period=O values
- 26 shy
ENDNOTES
IResearch Fellow International Food Policy Research Institute and Senior Irrigation Specialist International Irrigation Management Institute
2Presumable farmers also base private investment decisions on such foreknowledge of costs and commonly do so in the case of small tubewells In addition user groups may invest in small communal schemes independently after examining projected costs and benefits Most commonly however new irrigation capacity is created as a result of public investment decisions made by government and thus the political process comes into play
3This principle applies only when water charges are levied volumetrically--an exceedingly rare situation in the developing worldshy-though the argument is frequently invoked as though it applies to all allocative situations
4For Cabanatuan City the longer period of record available ie 1904 to 1986 was divided into two equal periods and tested for difference in means to test the observation by farmers that rainfall today is lower than it used to be Mean precipitation after 1947 was actually about 3 percent higher than before but the difference was not significant (t = 07)
5The procedure used was to take time series data on total nitrogenfertilizer consumption in the Philippines multiply it by the ratio of N fertilizer used on rice to total N consumption and divide the product by the rice harvested area to obtain an estimate of N use perhectare Values of the ratio of N use on rice to total N use were taken from IRRI (1988 150) with interpolation used to fill in gaps in the series
6This assumes that timeliness is evaluated against a standard based on the physiological demand for water generated by the rice crop and the varying sensitivity of the crop to stress at different growth stages In such a case a poor timeliness rating for irrigationservice will result in reductions in yield
7Because rice yields are insensitive to water applications that exceed PET values a redistribution of water from water excess to water short portions of the command area will usually result in higher aggregate output and higher average yields Exceptional scenarios can be constructed in which this general rule would be violated but such scenarios are unlikely to play out in practice
- 27 shy
aThere are a variety of standards for judging equity that could be used here The most common one is equality of per hectare water deliveries In the present case the implicit standard is equality of water supply utility in producing a crop Thus if two areas received equal amounts of water on a per hectare basis but one had such high seepage and percolation losses that the crop failed to produce a harvest (and was therefore exempted from irrigation fee payment) the water supply to the combined area would be judged inequitable Because such radical differences in seepage and percolation rates are unlikely on puddled rice soils in the lowland Philippines and because the definition of benefitted area accommodates a wide range of acceptable yields these two definitions are largely indistinguishable at the current limits of data resolution
Ta
ble
--I
nd
ices
of
se
v
ice a
rea
ib
en
ef
ited
a
rea
a
nd
a
vera
ge sea
so
na
l d
isch
arg
e
Ave
rage
A
vera
ge
tshy19
77
1978
19
79
1980
19
81
1982
19
83
1984
19
85
1986
19
77-8
1 19
82-8
6 S
tati
stic
a
(ind
ex
aver
age
1983
-198
5 =
100)
Ser
vice
are
a UP
RI IS
in
dex
882
91
9
920
91
5
951
95
1
100
0 10
00
100
0 10
00
917
99
0
bull5
53
Ang
at-M
aasi
m R
95
9
994
99
7
996
99
6
996
10
00
100
0 10
00
100
0 98
8
999
1
63
Sto
To
mas
10
63
106
3 10
1 9
10
29
103
0 99
8
100
0 10
00
100
0 11
10
104
1 10
22
-08
9
Siba
lom
-San
Jos
e 95
0
872
94
2
933
94
2
942
10
28
102
8 94
4
101
7
928
99
2
292
A
gana
n-St
a B
arba
ra
108
3 10
91
106
0 10
30
961
10
05
996
10
08
996
99
6
104
5 10
00
-21
1 A
vera
ge
987
98
8
981
98
1
916
97
8
100
5 10
07
988
10
25
984
10
01
123
Wet
seas
on b
enef
ited
are
a in
dex
UPR
IIS
110
0 10
07
114
4 10
55
113
8 11
74
951
10
71
971
ll
58
10
89
106
7 -0
47
Ang
at-M
aasi
m R
97
9
974
92
0
983
10
28
100
8 99
7
102
1 98
2
931
97
7
988
0
54
Sto
To
mas
11
63
115
9 11
23
107
5 10
80
103
9 98
1
978
10
41
103
9 11
20
101
6 -4
88
Siba
lom
-San
Jos
e 11
35
103
8 10
11
931
93
4
906
10
07
975
10
1S
998
10
11
981
-0
80
Aga
nan-
Sta
Bar
bara
10
85
110
5 10
74
106
8 10
01
104
2 99
8
101
5
987
61
0
106
7 93
0
-1 8
4
Ave
rage
10
92
105
6 10
54
102
4 10
36
103
4 98
7
101
3 10
00
947
10
53
996
-2
32
Dry
sea
Son
bene
fite
d ar
ea
inde
x U
PRIIS
14
04
155
0 15
S0
155
8 16
17
128
0 57
2
114
S 15
74
152
3 12
38
-09
4 A
ngat
-Maa
sim
R
903
93
0
103
2 10
61
104
2 10
69
988
99
2
102
1 99
6
993
10
13
061
S
to
Tom
as
105
7 12
27
122
5
961
99
9
115
9 10
1 7
91
0
107
3 12
1 2
10
94
107
4 -0
28
Si
balo
m-S
an J
ose
Aga
nan-
Sta
Bar
bara
66
5
95S
62
6
632
67
4
111
4
501
10
S1
412
11
51
766
93
3
856
94
3
107
0 94
8
107
4 11
09
111
6
158
6 58
8
987
97
6
110
4 5
35
083
N
w
Ave
rage
89
6
964
11
19
103
7 10
44
110
9 10
1 7
89
8
108
5 12
97
101
7 10
81
082
Wet
seas
on d
isch
arge
in
dex
UPR
IIS
132
9 72
7
142
5 11
88
120
4 98
8
105
8 10
63
879
96
4
117
5 99
1
-16
5 A
ngat
-Maa
sim
R
129
7 13
52
134
5 12
58
127
0 11
70
120
3 62
7
131
3 10
68
-18
9
Sto
To
mas
14
71
155
0 14
1 6
10
40
725
12
35
112
3 14
79
103
1 -4
48
Siba
lom
-San
Jos
e 96
8
733
11
1 5
92
2
907
47
1
102
3 85
7
ll2
0
422
92
9
779
-1
09
Aga
nan-
Sta
Bar
bara
87
9
863
68
1
925
96
8
110
7 70
5
871
87
7
00
9
Ave
rage
11
49
105
7 13
61
115
0 10
58
853
10
43
963
99
4
SO3
11
55
940
-2
85
Dry
sea
son
disc
harg
e in
dex
UPR
IIS
425
13
09
153
3 14
28
180
6 14
S0
125
8 56
3
117
9 14
66
130
0 11
89
-04
3
Ang
at-M
aasi
m R
12
26
161
6 15
30
139
2 12
85
100
8 10
33
958
14
41
107
1 -3
80
S
to
Tom
as
116
4 14
82
155
7 79
9
971
12
30
126
5 14
01
106
6 -2
44
Si
balo
m-S
an J
ose
486
71
5
782
62
9
789
64
5
116
9 11
86
801
65
3
918
2
36
Aga
nan-
Sta
Bar
bara
A
vera
ge
425
10
46
133
6 62
0
118
4 81
5
116
1 94
7
112
5 89
7
922
12
20
991
88
2
108
7 12
36
119
2 71
8
114
0 10
37
105
5 3
23
-07
5
a bull
indi
cate
s si
gnif
ican
ce a
t 95
per
cent
con
fide
nce
- 24 shy
Table 5--Results of pooled regressions yield indicators
------------------------- Dependent variables -------------------------shyIndependent variable ---------- Yield MTHa ----------- -- Specific Yield 1000 KgM
3 H20 shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season -shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 1967 (286)
3747 (580)
1991 (293)
2800 (460 )
0084 (041)
0441 (229)
0049 (023 )
0876 (343)
System 2 indicator 0400 (155)
0525 (226)
0057 (085)
0115 (1 95)
System 3 indicator 0466 (139 )
-0029 (-010)
-0016 (-018)
0006 (008)
System 4 indicator 0980 (470)
0090 (049)
0176 (333)
0391 (827)
System 5 indicator 1 221 (570)
0256 ( 138)
0104 (174)
0224 (440)
Reservoir indicator -1033 (-578)
-0063 (-039)
-0146 (-310)
-0326 (-569)
Period indicatora
Precipitation in mm Wet season
Annual
0163 (1 03)
-29E-4 (-182)
0231 (1 42)
-58E-4 (-606)
0161 (118)
92E-5 (072)
0199 043 )
-64pound-5 (-084)
0082 (200)
57E-6 (013)
0092 (223)
-61E-5 (-232)
0032 (093)
29E-7 (001)
0052 (112)
-12pound-4 (-414)
Nitrogen applicationb 0025 (254)
0021 (211)
0025 (274)
0020 (227)
0003 (1 05)
0002 (043)
0003 (101)
-0002 (-052)
R2 (adj) 0553 0516 0251 0198 0309 0262 0713 0445 number of observations 50 50 49 49 42 42 40 40 degrees of freedom 42 45 41 44 34 37 32 35
CoeffiCient (t statistic) indicates significance at 95 confidence ~ 1982-1986 = 1
kgha
- 25 shy
Table 6--Suggmry of estiates perforance indicators
Variable Predicted Predicted Period Dependent Variable
Description (units)
Period Coefficient
1976-86 IndicatorO
1976-86 Indicatorl
Effecta (X)
BenefittedService area ratio
Wet season ratio -0036 0853 0817 -425 Dry Season 0071 0539 0610 1313
Delivery Wet Dry
season season
nm water per benefitted ha
-0808 -1 214
6003 9083
5195 7869
-1346 -1337
Rice Yield Wet season metric tons 0163 3578 3741 456 Dry season per hectare 0161 3905 4066 412
Specific Yield Wet season kg rice per m3 0082 0348 0430 2353 Dry season water de 1 ivered 0032 0378 0409 837
a percentage changes relative to period=O values
- 26 shy
ENDNOTES
IResearch Fellow International Food Policy Research Institute and Senior Irrigation Specialist International Irrigation Management Institute
2Presumable farmers also base private investment decisions on such foreknowledge of costs and commonly do so in the case of small tubewells In addition user groups may invest in small communal schemes independently after examining projected costs and benefits Most commonly however new irrigation capacity is created as a result of public investment decisions made by government and thus the political process comes into play
3This principle applies only when water charges are levied volumetrically--an exceedingly rare situation in the developing worldshy-though the argument is frequently invoked as though it applies to all allocative situations
4For Cabanatuan City the longer period of record available ie 1904 to 1986 was divided into two equal periods and tested for difference in means to test the observation by farmers that rainfall today is lower than it used to be Mean precipitation after 1947 was actually about 3 percent higher than before but the difference was not significant (t = 07)
5The procedure used was to take time series data on total nitrogenfertilizer consumption in the Philippines multiply it by the ratio of N fertilizer used on rice to total N consumption and divide the product by the rice harvested area to obtain an estimate of N use perhectare Values of the ratio of N use on rice to total N use were taken from IRRI (1988 150) with interpolation used to fill in gaps in the series
6This assumes that timeliness is evaluated against a standard based on the physiological demand for water generated by the rice crop and the varying sensitivity of the crop to stress at different growth stages In such a case a poor timeliness rating for irrigationservice will result in reductions in yield
7Because rice yields are insensitive to water applications that exceed PET values a redistribution of water from water excess to water short portions of the command area will usually result in higher aggregate output and higher average yields Exceptional scenarios can be constructed in which this general rule would be violated but such scenarios are unlikely to play out in practice
- 27 shy
aThere are a variety of standards for judging equity that could be used here The most common one is equality of per hectare water deliveries In the present case the implicit standard is equality of water supply utility in producing a crop Thus if two areas received equal amounts of water on a per hectare basis but one had such high seepage and percolation losses that the crop failed to produce a harvest (and was therefore exempted from irrigation fee payment) the water supply to the combined area would be judged inequitable Because such radical differences in seepage and percolation rates are unlikely on puddled rice soils in the lowland Philippines and because the definition of benefitted area accommodates a wide range of acceptable yields these two definitions are largely indistinguishable at the current limits of data resolution
- 24 shy
Table 5--Results of pooled regressions yield indicators
------------------------- Dependent variables -------------------------shyIndependent variable ---------- Yield MTHa ----------- -- Specific Yield 1000 KgM
3 H20 shy
-- Wet Season -- -- Dry Season -- -- Wet Season -- -- Dry Season -shy
Equation Number (1) (2) (3) (4) (5) (6) (7) (8)
Constant 1967 (286)
3747 (580)
1991 (293)
2800 (460 )
0084 (041)
0441 (229)
0049 (023 )
0876 (343)
System 2 indicator 0400 (155)
0525 (226)
0057 (085)
0115 (1 95)
System 3 indicator 0466 (139 )
-0029 (-010)
-0016 (-018)
0006 (008)
System 4 indicator 0980 (470)
0090 (049)
0176 (333)
0391 (827)
System 5 indicator 1 221 (570)
0256 ( 138)
0104 (174)
0224 (440)
Reservoir indicator -1033 (-578)
-0063 (-039)
-0146 (-310)
-0326 (-569)
Period indicatora
Precipitation in mm Wet season
Annual
0163 (1 03)
-29E-4 (-182)
0231 (1 42)
-58E-4 (-606)
0161 (118)
92E-5 (072)
0199 043 )
-64pound-5 (-084)
0082 (200)
57E-6 (013)
0092 (223)
-61E-5 (-232)
0032 (093)
29E-7 (001)
0052 (112)
-12pound-4 (-414)
Nitrogen applicationb 0025 (254)
0021 (211)
0025 (274)
0020 (227)
0003 (1 05)
0002 (043)
0003 (101)
-0002 (-052)
R2 (adj) 0553 0516 0251 0198 0309 0262 0713 0445 number of observations 50 50 49 49 42 42 40 40 degrees of freedom 42 45 41 44 34 37 32 35
CoeffiCient (t statistic) indicates significance at 95 confidence ~ 1982-1986 = 1
kgha
- 25 shy
Table 6--Suggmry of estiates perforance indicators
Variable Predicted Predicted Period Dependent Variable
Description (units)
Period Coefficient
1976-86 IndicatorO
1976-86 Indicatorl
Effecta (X)
BenefittedService area ratio
Wet season ratio -0036 0853 0817 -425 Dry Season 0071 0539 0610 1313
Delivery Wet Dry
season season
nm water per benefitted ha
-0808 -1 214
6003 9083
5195 7869
-1346 -1337
Rice Yield Wet season metric tons 0163 3578 3741 456 Dry season per hectare 0161 3905 4066 412
Specific Yield Wet season kg rice per m3 0082 0348 0430 2353 Dry season water de 1 ivered 0032 0378 0409 837
a percentage changes relative to period=O values
- 26 shy
ENDNOTES
IResearch Fellow International Food Policy Research Institute and Senior Irrigation Specialist International Irrigation Management Institute
2Presumable farmers also base private investment decisions on such foreknowledge of costs and commonly do so in the case of small tubewells In addition user groups may invest in small communal schemes independently after examining projected costs and benefits Most commonly however new irrigation capacity is created as a result of public investment decisions made by government and thus the political process comes into play
3This principle applies only when water charges are levied volumetrically--an exceedingly rare situation in the developing worldshy-though the argument is frequently invoked as though it applies to all allocative situations
4For Cabanatuan City the longer period of record available ie 1904 to 1986 was divided into two equal periods and tested for difference in means to test the observation by farmers that rainfall today is lower than it used to be Mean precipitation after 1947 was actually about 3 percent higher than before but the difference was not significant (t = 07)
5The procedure used was to take time series data on total nitrogenfertilizer consumption in the Philippines multiply it by the ratio of N fertilizer used on rice to total N consumption and divide the product by the rice harvested area to obtain an estimate of N use perhectare Values of the ratio of N use on rice to total N use were taken from IRRI (1988 150) with interpolation used to fill in gaps in the series
6This assumes that timeliness is evaluated against a standard based on the physiological demand for water generated by the rice crop and the varying sensitivity of the crop to stress at different growth stages In such a case a poor timeliness rating for irrigationservice will result in reductions in yield
7Because rice yields are insensitive to water applications that exceed PET values a redistribution of water from water excess to water short portions of the command area will usually result in higher aggregate output and higher average yields Exceptional scenarios can be constructed in which this general rule would be violated but such scenarios are unlikely to play out in practice
- 27 shy
aThere are a variety of standards for judging equity that could be used here The most common one is equality of per hectare water deliveries In the present case the implicit standard is equality of water supply utility in producing a crop Thus if two areas received equal amounts of water on a per hectare basis but one had such high seepage and percolation losses that the crop failed to produce a harvest (and was therefore exempted from irrigation fee payment) the water supply to the combined area would be judged inequitable Because such radical differences in seepage and percolation rates are unlikely on puddled rice soils in the lowland Philippines and because the definition of benefitted area accommodates a wide range of acceptable yields these two definitions are largely indistinguishable at the current limits of data resolution
- 25 shy
Table 6--Suggmry of estiates perforance indicators
Variable Predicted Predicted Period Dependent Variable
Description (units)
Period Coefficient
1976-86 IndicatorO
1976-86 Indicatorl
Effecta (X)
BenefittedService area ratio
Wet season ratio -0036 0853 0817 -425 Dry Season 0071 0539 0610 1313
Delivery Wet Dry
season season
nm water per benefitted ha
-0808 -1 214
6003 9083
5195 7869
-1346 -1337
Rice Yield Wet season metric tons 0163 3578 3741 456 Dry season per hectare 0161 3905 4066 412
Specific Yield Wet season kg rice per m3 0082 0348 0430 2353 Dry season water de 1 ivered 0032 0378 0409 837
a percentage changes relative to period=O values
- 26 shy
ENDNOTES
IResearch Fellow International Food Policy Research Institute and Senior Irrigation Specialist International Irrigation Management Institute
2Presumable farmers also base private investment decisions on such foreknowledge of costs and commonly do so in the case of small tubewells In addition user groups may invest in small communal schemes independently after examining projected costs and benefits Most commonly however new irrigation capacity is created as a result of public investment decisions made by government and thus the political process comes into play
3This principle applies only when water charges are levied volumetrically--an exceedingly rare situation in the developing worldshy-though the argument is frequently invoked as though it applies to all allocative situations
4For Cabanatuan City the longer period of record available ie 1904 to 1986 was divided into two equal periods and tested for difference in means to test the observation by farmers that rainfall today is lower than it used to be Mean precipitation after 1947 was actually about 3 percent higher than before but the difference was not significant (t = 07)
5The procedure used was to take time series data on total nitrogenfertilizer consumption in the Philippines multiply it by the ratio of N fertilizer used on rice to total N consumption and divide the product by the rice harvested area to obtain an estimate of N use perhectare Values of the ratio of N use on rice to total N use were taken from IRRI (1988 150) with interpolation used to fill in gaps in the series
6This assumes that timeliness is evaluated against a standard based on the physiological demand for water generated by the rice crop and the varying sensitivity of the crop to stress at different growth stages In such a case a poor timeliness rating for irrigationservice will result in reductions in yield
7Because rice yields are insensitive to water applications that exceed PET values a redistribution of water from water excess to water short portions of the command area will usually result in higher aggregate output and higher average yields Exceptional scenarios can be constructed in which this general rule would be violated but such scenarios are unlikely to play out in practice
- 27 shy
aThere are a variety of standards for judging equity that could be used here The most common one is equality of per hectare water deliveries In the present case the implicit standard is equality of water supply utility in producing a crop Thus if two areas received equal amounts of water on a per hectare basis but one had such high seepage and percolation losses that the crop failed to produce a harvest (and was therefore exempted from irrigation fee payment) the water supply to the combined area would be judged inequitable Because such radical differences in seepage and percolation rates are unlikely on puddled rice soils in the lowland Philippines and because the definition of benefitted area accommodates a wide range of acceptable yields these two definitions are largely indistinguishable at the current limits of data resolution
- 26 shy
ENDNOTES
IResearch Fellow International Food Policy Research Institute and Senior Irrigation Specialist International Irrigation Management Institute
2Presumable farmers also base private investment decisions on such foreknowledge of costs and commonly do so in the case of small tubewells In addition user groups may invest in small communal schemes independently after examining projected costs and benefits Most commonly however new irrigation capacity is created as a result of public investment decisions made by government and thus the political process comes into play
3This principle applies only when water charges are levied volumetrically--an exceedingly rare situation in the developing worldshy-though the argument is frequently invoked as though it applies to all allocative situations
4For Cabanatuan City the longer period of record available ie 1904 to 1986 was divided into two equal periods and tested for difference in means to test the observation by farmers that rainfall today is lower than it used to be Mean precipitation after 1947 was actually about 3 percent higher than before but the difference was not significant (t = 07)
5The procedure used was to take time series data on total nitrogenfertilizer consumption in the Philippines multiply it by the ratio of N fertilizer used on rice to total N consumption and divide the product by the rice harvested area to obtain an estimate of N use perhectare Values of the ratio of N use on rice to total N use were taken from IRRI (1988 150) with interpolation used to fill in gaps in the series
6This assumes that timeliness is evaluated against a standard based on the physiological demand for water generated by the rice crop and the varying sensitivity of the crop to stress at different growth stages In such a case a poor timeliness rating for irrigationservice will result in reductions in yield
7Because rice yields are insensitive to water applications that exceed PET values a redistribution of water from water excess to water short portions of the command area will usually result in higher aggregate output and higher average yields Exceptional scenarios can be constructed in which this general rule would be violated but such scenarios are unlikely to play out in practice
- 27 shy
aThere are a variety of standards for judging equity that could be used here The most common one is equality of per hectare water deliveries In the present case the implicit standard is equality of water supply utility in producing a crop Thus if two areas received equal amounts of water on a per hectare basis but one had such high seepage and percolation losses that the crop failed to produce a harvest (and was therefore exempted from irrigation fee payment) the water supply to the combined area would be judged inequitable Because such radical differences in seepage and percolation rates are unlikely on puddled rice soils in the lowland Philippines and because the definition of benefitted area accommodates a wide range of acceptable yields these two definitions are largely indistinguishable at the current limits of data resolution
- 27 shy
aThere are a variety of standards for judging equity that could be used here The most common one is equality of per hectare water deliveries In the present case the implicit standard is equality of water supply utility in producing a crop Thus if two areas received equal amounts of water on a per hectare basis but one had such high seepage and percolation losses that the crop failed to produce a harvest (and was therefore exempted from irrigation fee payment) the water supply to the combined area would be judged inequitable Because such radical differences in seepage and percolation rates are unlikely on puddled rice soils in the lowland Philippines and because the definition of benefitted area accommodates a wide range of acceptable yields these two definitions are largely indistinguishable at the current limits of data resolution