world development vol 33

8/10/2019 World Development Vol 33

1/22

Where is the PovertyEnvironment Nexus?

Evidence from Cambodia, Lao PDR, and Vietnam

SUSMITA DASGUPTA, UWE DEICHMANN,CRAIG MEISNER and DAVID WHEELER *

The World Bank, Washington, DC, USA

Summary.This paper investigates the povertyenvironment nexus at the provincial and districtlevels in Cambodia, Lao PDR, and Vietnam. The analysis focuses on spatial associations betweenpoverty populations and five environmental problems: deforestation, fragile soils, indoor air pollu-tion, contaminated water, and outdoor air pollution. The results suggest that the nexus is quite dif-ferent in each country. We conclude that the nexus concept can provide a useful catalyst forcountry-specific work, but not a general formula for program design. Joint implementation of pov-erty and environment strategies may be cost effective for some environmental problems, but inde-pendent implementation may be preferable in many cases as well. 2005 Elsevier Ltd. All rights reserved.

Key words poverty, pollution, basic needs, deforestation, sanitation, natural resources

1. INTRODUCTION

During the past few years, the publications ofinternational development agencies have high-lighted the importance of the povertyenviron-ment nexus, a set of mutually reinforcing linksbetween poverty and environmental damage(Bojo et al., 2001; Bosch, Hommann, Rubio,Sadoff, & Travers, 2001; Ekbom & Bojo,1999). In this nexus, poverty reduction and envi-ronmental protection are complementary goals.For environmentalists, the nexus concept hasprovided a welcome defense against arguments,based on the Environmental Kuznets Curve,

that the early stages of development areunavoidably marked by conflicts between pov-erty reduction and environmental protection. 1

Common profession of belief in the nexus hasalso smoothed uneasy relations between envi-ronmental specialists and traditional projectofficers in development aid organizations.

Numerous studies have suggested that envi-ronmental damage can have particular signifi-cance for the poor. Recent participatorypoverty assessments, conducted in 14 develop-ing countries of Asia, Africa, and Latin Amer-

ica, reveal a common perception by the poorthat environmental quality is an importantdeterminant of their health, earning capacity,

security, energy supplies, and housing quality(Brocklesby & Hinshelwood, 2001). Rural stud-

ies often observe that poor peoples economicdependence on natural resources makes themparticularly vulnerable to environmental degra-dation (Ambler, 1999; Cavendish, 1999, 2000;Kepe, 1999; Reddy & Chakravarty, 1999).Other studies have assessed the health damagesuffered by poor households that are directlyexposed to pollution of the air, water, and land(Akbar & Lvovsky, 2000; Bosch et al., 2001;Brooks & Sethi, 1997; Mink, 1993; Songsore& McGranahan, 1993; Surjadi, 1993). In addi-tion, environmental disasters and environ-

ment-related conflicts may have regressiveimpacts because the poor are least capable ofcoping with their effects (Albla-Betrand, 1993;Myers & Kent, 1995).

In some cases, poor households them-selves may increase environmental degradation.

*The authors would like to thank Jostein Nygard,

Giovanna Dore, Piet Buys, Kiran Pandey, and Hua

Wang, for their valuable comments and contributions.

This research was supported by the World Banks

Environment Department and its East Asia and PacificUnit for Environment and Social Development. Final

revision accepted: October 11, 2004.

World Development Vol. 33, No. 4, pp. 617638, 2005 2005 Elsevier Ltd. All rights reserved

Printed in Great Britain0305-750X/$ - see front matter

doi:10.1016/j.worlddev.2004.10.003www.elsevier.com/locate/worlddev

617


2/22

Poverty-constrained options may induce thepoor to deplete resources at rates that are incom-patible with long-term sustainability (Holden,1996). In such cases, degraded resources precip-itate a downward spiral, by further reducing

the income of the poor (Cleaver & Schreiber,1994; Dasgupta & Maler, 1994; Durning, 1989;Ekbom & Bojo, 1999; Mink, 1993; Pearce &Warford, 1993; Prakash, 1997; World Bank,1992; World Commission on Environment &Development, 1987). Rapid population growth,coupled with insufficient means or incentives tointensify production, may induce overexploita-tion of fragile lands on steep hillsides, or inva-sion of areas that governments are attemptingto protect for environmental reasons. Again, adownward spiral can ensue (World Bank, 1992).

The existing literature also suggests that thestrength of povertyenvironment linkages maybe affected by factors as diverse as economicpolicies, resource prices, local institutions,property rights, entitlements to natural re-sources, and gender relations (Ambler, 1999;Arnold & Bird, 1999; Barbier, 2000; Dasgupta& Maler, 1994; Dutt & Rao, 1996; Ekbom &Bojo, 1999; Eskeland & Kong, 1998; Heath &Binswanger, 1996; Leach & Mearns, 1991;Roe, 1998). This research suggests that the rel-ative strength of links between poverty and

environment may be very context specific(Bucknall, Kraus, & Pillai, 2000; Chomitz,1999; Ekbom & Bojo, 1999). In a recent theo-retical work, Ezzati, Singer, and Kammen(2001)have demonstrated the implications forthe Environmental Kuznets Curve (EKC). Ina long-run model that allows for many interac-tions between socioeconomic and environmen-tal variables, they show that the conventional,U-shaped EKC describes only one of many po-tential development paths. 2

What does the empirical evidence suggest

about the actual prevalence and importanceof the povertyenvironment nexus and comple-mentary problems? Here the actual record issparse, because the requisite data are often dif-ficult to obtain in developing countries. In prin-ciple, household-level studies can adequatelytest whether environmental problems have adisproportionate impact on the poor. In prac-tice, such tests are rare. For example, somestudies have established a link between povertyand consumption of wood fuel, and at least,one credible study has established the relation-

ship between indoor combustion and health(Ezzati & Kammen, 2001). However, the re-search also suggests the importance of interven-

ing variables such as cooking practices (indoorvs. outdoor) and fuel choice (e.g., charcoalemits far fewer fine particles than wood). Chil-dren die of waterborne disease at higher rates inpoor households, but again, research points to

the significance of intervening variables suchas water source quality and mothers education(Filmer & Pritchett, 1997; Merick, 1985). Rig-orous empirical studies that combine local-areaenvironmental variables (deforestation, out-door air quality, water quality, soil erosion,etc.) with standard household surveys are al-most nonexistent. Similarly, very few local-areastudies relate environmental quality to thenumber and characteristics of poor households.

In povertyenvironment analysis that is rele-vant for policy, the spatial dimension is critical

for two reasons: First, most environmentalproblems are inherently geographical. In princi-ple, different environmental problems should beanalyzed at different regional scales. In the caseof pollution, for example, the theoreticallyappropriate scale is affected by the dispersalcharacteristics of the pollutant and medium:Particulate pollution from cement mills mayonly be dangerous in one urban region; acidrain from sulfur emissions may damage forestshundreds of miles from the source; and eutro-phication from fertilizer runoff may affect

ocean fisheries a thousand miles downstreamfrom the farms that are the source of the prob-lem. In practice, data constraints often dictatethe choice of the scale. In Cambodia, Vietnam,and Lao PDR, for example, appropriate dataare relatively plentiful at the provincial level,scarce at the district level (except in Cambodia),and practically nonexistent for subdistricts.Accordingly, this paper focuses on provincialdata, with extensions to the district level forCambodia. While we readily acknowledge thatmore disaggregated evidence would be desir-

able, we believe that even province-level analy-sis provides a useful first approximation forpovertyenvironment work.

We also recognize the possibility that cross-country externalities may introduce elementsof the nexus at the regional level, as we havenoted in the previous paragraph. If much of pol-lutions impact is felt far downwind or down-stream, failure to find a nexus in the local datamay simply mask trans-boundary effects. Forsome pollutants (e.g., acid rain and deforesta-tion from industrial sulfur emissions), this may

indeed be a problem. However, a large bodyof evidence indicates that local impacts predom-inate for pollutants that most heavily affect the

618 WORLD DEVELOPMENT


3/22

poor. The foremost examples are fine particu-late air pollution, which causes the bulk of car-diorespiratory problems, and fecal coliformpollution of water, which is the major cause ofintestinal disease. Most particulate air pollution

settles locally; locally generated plumes of fe-cal coliform pollution generally persist for sev-eral miles downstream, but would presenttrans-boundary problems only in border areas.

At the provincial level, a minimum criterionfor potential significance of the nexus is dispro-portionate environmental damage in high-pov-erty areas. Correlation does not necessarilyimply causality, of course, but the lack of timeseries data prevents formal testing of structuralcausation models. 3 Nevertheless, some reason-able inferences from positive correlations are

possible. We can also gain an insight from con-trary cases, in which some of a countrys envi-ronmental problems exhibit no spatialcorrelation with poverty across provinces. Atleast three inferences are possible: The coun-trys povertyenvironment nexus may not in-clude these problems; the government mayalready have addressed them effectively; ortheir part of the nexus may only be operativeat the district or subdistrict level. Data permit-ting, further tests could be run at those levels.

Administrative economics provide the second

rationale for spatially disaggregated analysis.An appropriate geographic scale must strikean appropriate balance between the benefits ofdecentralization and the associated costs. Onthe environment side, for example, effective reg-ulation requires local inspection of damagesources (pollution, deforestation, etc.), as wellas more centralized facilities for informationcollection, storage, and analysis. Environmentalmanagement is undoubtedly improved by aknowledge of local conditions, but the marginalcost of administration rises with distance from

administrative centers, because of deterioratingtransport and communications quality. Gener-ally, province- or district-level administrationstrikes the right balance between headquartersscale economies and the cost of dispersed mon-itoring and enforcement operations.

Similar factors govern the choice of adminis-trative scale for poverty-alleviation programs.Headquarters staffing remains important, butsuch programs also require local monitoringinformation and frequent interaction with cli-ents. By the same logic, province- or district-le-

vel administration may be preferable to nationalor subdistrict administration in many cases.

Why are these factors important for analysisof the povertyenvironment nexus? From a pol-icy perspective, the nexus is important only if ithas consequences for the allocation and admin-istration of public resources for alleviation of

poverty and environment problems. If there isno nexus, then optimal policy should treat thetwo sectors as separate, divide the overall bud-get between them by some criteria, and use sep-arate calculations to distribute resourcesamong provincial or district agencies. If pov-ertyenvironment links are strong, on the otherhand, then optimal policy should treat them

jointly in allocating the overall budget. Someresources for environmental improvementshould be allocated to poverty alleviation whenpoverty significantly increases environmental

degradation, and the converse should holdwhen environmental factors significantly in-crease poverty.

In this paper, we use newly available data totest for the existence of the nexus in Cambodia,Lao PDR, and Vietnam. Our main analyticaltools are georeferenced indicator mapping, cor-relation, and regression analysis. 4 We recog-nize that systematic empirical work is onlybeginning in this area; future work may revealthat the regional scale of our analysis is toobroad, and that the true povertyenviron-

ment nexus is more localized in nature. It is alsoentirely possible that different dimensions ofthe nexus are best analyzed at different geo-graphical scales. For example, watershedsmay provide the best spatial units for the studyof relationships linking poverty, soil erosion,and deforestation. 5 At present, the availabledata limit us to the broader exercise that is pre-sented here. In addition, our focus on the com-parative incidence of povertyenvironmentproblems leads us to focus the exercise at acommon geographical scale.

The remainder of the paper is organized asfollows: Section 2 provides an overview of themethods that we employ and relates them tounderlying hypotheses about links betweenpoverty and the environment. Section 3 intro-duces the new regional dataset for Cambodia,Vietnam, and Lao PDR, with particular atten-tion to the data-collection process, coverage,and accuracy. In Section 4, we provide a de-tailed illustration of our approach for Cambo-dia. Section 5 summarizes comparableevidence for Lao PDR and Vietnam, 6 Section

6 discusses some causal implications of the re-sults, and Section 7 concludes the paper.

WHERE IS THE POVERTYENVIRONMENT NEXUS? 619


4/22

2. MAPPING THE PROBLEMS

(a) Absolute poverty

For each country, our provincial or district-

level absolute poverty indices are determinedby the number of inhabitants whose daily con-sumption expenditure cannot support a foodintake of more than 2,000 calories, plus mini-mal nonfood expenditures. By this measure,the national incidence of absolute poverty is36% in Lao PDR, 40% in Cambodia, and37% in Vietnam. 7 To illustrate, Figure 1 dis-plays the regional distribution of absolute pov-erty in Cambodia.

We vary our use of the absolute poverty mea-sure according to the nature of each environ-

mental problem. For example, we employboth the size of the poverty population andthe poverty head-count ratio (or incidence) inour regression analyses of the relationshipslinking poverty to illness from indoor air pollu-tion and polluted water. 8, 9 For our analysis ofdeforestation, we use the density of the povertypopulation per unit area. 10

Even though absolute material poverty pro-vides a useful indicator, we recognize that nosingle measure can capture all the dimensionsassociated with a broader concept of poverty.

For an analysis of the povertyenvironmentnexus, other relevant factors may include lackof access to common property resources, poorhealth, and low levels of education. For exam-

ple, research on China and Indonesia has sug-gested that education reduces pollutiondamage because better-educated communitiesare more willing and able to organize to controlpolluters. 11 As better data become available,

more fully specified models should incorporatesuch factors. In this study, the consequences ofour reliance on a single poverty index areambiguous. Exclusion of poverty dimensionsthat are correlated with material poverty mayresult in overestimation of its impact in our cor-relation and regression analysis. On the otherhand, exclusion of these other dimensions alsosuppresses their potentially significant two-way interactions with environmental variables.

(b) Environmental problems

We consider five critical environmental prob-lems, two related to natural resource degrada-tion and three to pollution. The Greenproblems are deforestation and soil degrada-tion, whereas the Brown problems are in-door air pollution, contaminated water, andoutdoor air pollution.

(i) DeforestationDeforestation serves as a proxy for the loss of

critical ecosystems and biodiversity, as well as

increased risk of soil erosion in steeply slopedareas. To test for a povertyenvironment nexusin this context, we map forested areas and ratesof deforestation by province and district. In

Figure 1. Cambodia: total poverty population by district, 2000. Source: World Food Program, 2001.



5/22

areas where significant forests remain, we assessthe spatial correlation of poverty and deforesta-tion using maps, graphical scatter plots, andregressions.

For the regression analysis, our two principal

variables are the settlement density of the poorpopulation and overall population density. 12

By incorporating both, we can simultaneouslytest the impact of poor and nonpoor house-holds on forest clearing. 13 Insignificance of set-tlement density for the poverty population inthis regression would certainly weaken thenexus argument that poor households clear for-ests more rapidly than others. We cannot testthe converse proposition (exogenously gener-ated deforestation increases poverty) until wehave better information about the dependence

of the poor on forest products. Future researchshould use local data for a more detailed anal-ysis of this potential link.

We also test for the impact of commerciallogging by controlling for differences in treespecies. In our three study countries, some areaexperts have suggested that deforestation is sig-nificantly faster in areas dominated by ever-greens, which are the preferred species forcommercial loggers.

(ii) Fragile soils

Steep hillsides under intensive cultivation areparticularly vulnerable to erosion and soil deg-radation without terracing, and the economicreturn to farming steeply sloped areas is gener-ally lower than the return to cultivating alluvialsoils in river valleys. While these observationsare straightforward, their implications for thepovertyenvironment nexus depend on localpossibilities for migration. In regions wherepeople are relatively free to migrate to areaswith higher expected returns, we would expectsteeply sloped areas to be more sparsely popu-

lated than alluvial plains. If population growthraises the labor intensity of alluvial farming, wewould expect diminishing returns in the low-lands to induce uphill movement by farmers.This movement would be tempered by erosionand soil depletion in the highlands, with a con-sequent drop in the overall marginal productiv-ity of agricultural labor. The remaininghighland farmers should farm larger plots, onaverage, to compensate for poorer soils andto maintain parity in expected income with low-land farmers. Damage to highland soils would

be a resource conservation problem for societyas a whole, but would not have a dispropor-

tionate impact on the poorest farmers if theyremained free to migrate.

A very different picture would emerge, how-ever, if marginalized ethnic groups were iso-lated in highland areas by historical patterns

of separation and discrimination. In this case,population growth and soil degradation in thehighlands might well create a poverty trapthere. By implication, a potential povertyenvi-ronment nexus exists in regions where poorhouseholds are highly concentrated in steeplysloped areas.

(iii) Indoor air pollutionRecent research has suggested that indoor air

pollution from wood fuels is a major cause ofrespiratory disease in developing countries.

Many households use wood or charcoal in Cam-bodia, Lao PDR, and Vietnam, so indoor airpollution may be a significant health problem.Although indoor air monitoring data are notyet available in the region, household surveyshave recorded the use of wood and charcoal.We use regression analysis to test whether pov-erty and wood fuel use are significantly associ-ated, after controlling for area population. Werecognize that our results can only be suggestive,since the impact of wood fuel use depends onwhether burning is indoors or outdoors. Gaug-

ing the true magnitude of the problem will re-quire household-level pollution monitoring andhealth assessment. This should be an importanttopic for future research in our focal countries.

(iv) Access to clean water and sanitationSafe water and adequate sanitation are criti-

cal determinants of health status, particularlyfor children. Ingestion of coliform bacteriafrom contaminated drinking water or food isa prime cause of diarrheal disease, which is inturn a major cause of infant mortality in devel-

oping countries. Although data for SoutheastAsia remain limited, we use the available infor-mation to assess the spatial relationships link-ing poverty, sanitation, and diarrheal disease.At present, many households in the three coun-tries do not have access to safe water or sanita-tion. A povertyenvironment nexus exists if theaffected households are disproportionatelypoor. We use maps, scatter plots, and regres-sions to test for this possibility.

(v) Outdoor air pollution

Outdoor air pollution is primarily an urbanphenomenon, whose severity depends on the



6/22

scale of polluting activities, their pollutionintensity (or pollution per unit of output), andthe characteristics of the urban air shed. Recentresearch has established that exposure to fineparticulates (with diameters of 10 microns

(PM10) or less) is the main cause of pollution-related respiratory disease. Until recently, littlewas known about fine-particulate pollution inSoutheast Asian cities. During the past year,however, the World Health Organization andthe World Bank have used a large internationaldatabase to develop a prediction model forPM10pollution, based on urban population, in-come, fuel use, and local atmospheric charac-teristics (wind, rainfall, temperature, altitude,etc.). 14

We measure the health impact of PM10pollu-

tion using a standard doseresponse func-tion, based on population exposure studies,that relates the incidence of mortality to theairborne concentration of PM10.

15 For eachurban area in Cambodia, Lao PDR, and Viet-nam, we use the World Bank/WHO model topredict the airborne concentration of PM10,transform the concentration to an estimatedmortality incidence using the doseresponsefunction, and multiply the predicted incidenceby the urban population to obtain estimateddeaths from air pollution. We cannot adjust

specifically for the impact of air pollution onthe poor, because the existing doseresponsefunctions do not distinguish between incomegroups. It seems likely that our approachunderestimates health damage to the poor,who are generally unable to afford health carethat can (at least partially) compensate for theeffect of air pollution.

We would, of course, prefer to base our esti-mates on actual monitoring data. However, toour knowledge, previous environmental studieshave not even attempted to estimate air pollu-

tion for cities in the region. We therefore offerthese estimates as a suggestive benchmark forfurther research. Aggregation of the urban-arearesults to the provincial level enables us to testfor a povertyenvironment nexus by assessingthe spatial correlation between poverty andhealth damage from outdoor air pollution.

3. REGIONAL DATASETSFOR CAMBODIA, VIETNAM,

AND LAO PDR

Work on the povertyenvironment nexus re-quires spatially integrated information, but the

underlying problems have traditionally receivedseparate treatment. The spatial dimension ofpoverty has probably received the most atten-tion. In our three focal countries, for example,poverty mapping has been supported by the

World Food Program and the World Bank inCambodia, the International Food Policy Re-search Institute (IFPRI) in Vietnam, and theWorld Bank in Lao PDR.

Support for spatially oriented work on theenvironment has been scattered among agen-cies that have interests in different topics. TheMekong River Commission has supportedwork on deforestation, terrain slopes, erosion,and water quality in our three focal countries.As previously mentioned, the World HealthOrganization and the World Bank have esti-

mated air pollution for cities with over100,000 in population as part of an interna-tional program.

Various household survey exercises have pro-vided critical information for the analysis ofhealth damage from water pollution and indoorair pollution. For Cambodia and Vietnam,USAID has sponsored regionally coded Demo-graphic and Health Surveys (DHS) that includedata on intestinal illness, respiratory disease re-lated to air pollution, and the use of biomassfuels that are major determinants of indoor

air pollution. Population census data for Cam-bodia provide regionally coded information onhouseholds access to safe drinking water andsanitation.

When this study began, these critical informa-tion sources were scattered among internationalagencies, local development research institutes,and government ministries that had sponsoredor participated in the work. Accessing the datainvolved several trips to the three countries bya team of World Bank staff members and con-sultants, including the authors. The World

Banks Spatial Analysis Team incorporatedthe information into the spatially integrateddataset that has been used for this paper.

Most topical components of the data (e.g.,poverty incidence, intestinal illness from waterpollution) reflect sophisticated survey and map-ping work by experienced international teams.National population census surveys have alsoemployed standard sampling methods. For thisanalysis, the major problem has not been thequality of the underlying data, but rather theappropriate level of spatial integration. Some

geophysical data (e.g., forest coverage, terrainmaps) have very high levels of spatial disaggre-gation, while the coverage of some household



7/22

survey data (e.g., the DHS survey for Vietnam)is only sufficient to support reliable estimationat the provincial level. To maintain an inte-grated view, we have analyzed the data at thegreatest common level of disaggregation avail-

able in each country. Accordingly, we haveanalyzed the Cambodian data at the district(subprovincial) level, while limiting the exerciseto the provincial level in Lao PDR and Viet-nam.

4. EVIDENCE FOR CAMBODIA

To provide the most disaggregated view ofthe evidence, we begin with the district-levelanalysis for Cambodia. Figure 1 provides the

best available map of Cambodias poverty pop-ulation at the district level: Poor households areconcentrated along an axis that runs northwestto the border with Thailand. Figure 2, whichdisplays variations in the density of the poverty

Figure 2. Cambodia: settlement density of the poverty population. Source: World Food Program, 2001.

Figure 3. Cambodia: forested area, 1997. Source: Mekong River Commission (MRC), 2001.



8/22

population, suggests that provision of servicesto the poor would have the lowest unit cost inthe southeastern part of the axis.

Figures 3 and 4provide maps of Cambodiasforest cover and rate of deforestation for the

period 199397.Figure 4suggests that defores-tation is a major problem at the margin of thecentral population axis; many contiguous dis-tricts have very high deforestation, and manyareas, one district removed, also have highrates. The other region with rapid deforestationis the sparsely populated northeast. For thecountry as a whole, a comparison of Figures1 and 4suggests that priority areas for povertyalleviation and forest protection are weakly re-lated because many of the core poverty areas

are already deforested. The scatter diagram inFigure 5 confirms this, showing a nearly ran-dom relationship with a rank correlation of0.15.

Our regression results (Table 1) suggest that

overall population pressure is a major determi-nant of deforestation in Cambodia. However,the results also suggest that forest clearing bypoor people is neither more nor less intensivethan forest clearing by the general population.In the Cambodian regressions, introduction ofexplicit controls for species yields no higherdeforestation rate for evergreens, which are re-puted to be more lucrative for loggers.

Figure 6uses the incidence of steeply slopedlands to map the potential for erosion and soil

Figure 4. Cambodia: deforestation rate, 199397. Source: Mekong River Commission (MRC), 2001.

0

20

40

60

80

100

120

140

0 20 40 60 80 100 120 140 160 180 200

Number of poor (rank)

Deforestationrate(rank)

= 0.15

Figure 5. Cambodia: rank scatter: deforestation rate versus poverty population.



9/22

depletion in Cambodia. Distinct highland areasare visible in the northeast, southeast, and par-ticularly the southwest regions of the county.The countrys central population axis, on theother hand, is effectively defined by the low-lands. Regions with intermediate topographyare intermediate in settlement as well.

Comparison of Figures 1 and 6 suggests a

negative relationship between settlement bythe poor and steeply sloped land: Poor peopleare heavily concentrated in lowland areas andreside at a much lower density in highland

areas. The map shows little evidence of largepoverty populations in steeply sloped areas,suggesting relatively few cases of inability tomigrate because of ethnic segmentation anddiscrimination. The scatter inFigure 7confirms

the negative relationship between poverty andsteeply sloped land (simple correlation coeffi-cient: 0.29), and is consistent with a modelof relatively free migration in Cambodia.

Figure 8displays the scatter plot of district-level poverty population versus populationusing fuel wood or charcoal. Obviously, therelationship is very close (the correlation coeffi-cient is 0.70, with much of the remaining varia-tion explained by the plots obvious separationinto two separate sets of points). However, theexistence of a true povertyenvironment nexus

in this context depends on more intensive useof charcoal and wood fuel by poor households.

The results inTable 2do, in fact, indicate astrong association between poverty and woodfuel use after we control for general populationeffects.

Figure 9 suggests a close spatial correlationbetween poverty and lack of access to cleanwater. Regression analysis (Table 3) also sug-gests that poor households have much less accessto safe water than higher-income households inCambodia. The implications for child mortality

are suggested byFigure 10, which displays theregional distribution of childhood deaths inCambodia. Again, the spatial correlation withthe poverty population is evident.

Figure 6. Cambodia: percent of land that is steeply sloped. Source: Mekong River Commission (MRC), 2001.

Table 1. Cambodia: population, poverty anddeforestation

Variable Model 1 Model 2 Model 3

Log(Poor/Forest

cover 93)

0.007 0.007

Log(Population/

Forest cover 93)

0.010 0.011 0.018**

Evergreen 0.052* 0.018 0.020

Deciduous 0.036

Mixed 0.062**

Constant 0.014 0.030 0.039N 369 369 369

R2 0.065 0.056 0.052

Dependent variable: Log(Forest cover 1997/Forest cover1993). Evergreen, deciduous, and mixed forest dummyvariables.*

Significant at the 10% level.** Significant at the 5% level.



10/22

Using the WHO/World Bank model, we pro-ject PM10 pollution levels for urban areas inCambodian cities.Figure 11indicates that esti-

mated pollution levels are generally higher incities located in Cambodias population periph-ery. Using a standard doseresponse model, weestimate the resulting loss of life and aggregate

the results to the provincial level. Our findings,displayed inFigure 12, suggest minimal corre-lation (0.14) between poverty population anddeaths from air pollution.

Figure 13 summarizes the available evidencefor Cambodias poverty population, deforesta-tion, steeply sloped land, indoor air pollution,unsafe water, child mortality, and mortalityfrom outdoor air pollution. The elements ofthe matrix are numerically ranked by severityfor ease of comparison.Figure 14further con-denses the evidence into average rankings for

the first two (Green) indices and the lastthree (Brown) indices. The results seem con-sistent with a povertyenvironment nexus for

Landwithslope>16%(

rank)

= - 0.29

0 20 40 60 80 100 120 140 160 180 200


0

20

40

60

80

100

120

140

160

Figure 7. Cambodia: rank scatter: steeply sloped land versus poverty population. Source: Mekong River Commission

(MRC), 2001.

0

20

40

60

80

100

120

140

160

180

200

P

op.usingwood&c

harcoal

forcooking(rank)

= 0.70

0 20 40 60 80 100 120 140 160 180 200


Figure 8. Cambodia: rank scatter: fuel wood-using population versus poverty population. Source: Population Census,

1998.

Table 2. Cambodia: population, poverty and use of woodfuel and charcoal

Variable Model 1 Model 2

Total population 0.843**

Number of poor 0.292**

Log(Total population) 0.971**

Log(Number of poor) 0.013**

Constant 1101.698** 0.141**

N 180 180

R2 0.979 0.994

Dependent variable: Model 1Population using wood& charcoal; Model 2Log(Population using wood &charcoal).** Significant at the 5% level.



11/22

indoor air pollution and water contamination.However, there is no evident relationship be-tween the spatial distributions of poverty anddeaths from outdoor air pollution. Nor does

there appear to be a significant spatial relation-ship between poverty and either of the Greenindices. On the basis of currently availableevidence, we conclude that the regional pov-ertyenvironment nexus in Cambodia is largelyconfined to household-level problems associ-ated with contaminated air and water.

5. EVIDENCE FOR LAO PDRAND VIETNAM

Similar evidence for Lao PDR in Figure 15suggests a povertyenvironment nexus that issignificantly broader than Cambodias. Across

provinces, Figure 15shows a strong correspon-dence between poverty and environmental deg-radation in all five categoriesdeforestation,erosion potential, indoor air pollution, contam-inated water, and outdoor air pollution. Theassociation is particularly strong for the lowest-and highest-income provinces. When the envi-ronmental rankings are combined into Green

and Brown indices, the association is cleareracross all provinces. We conclude that the re-gional povertyenvironment nexus seems verybroad for Lao PDR, so the potential synergybetween poverty alleviation and environmentalpolicies may be very high. The north- andnortheastern regions of the country appear tobe the main locus for action in this context.

For Vietnam, the evidence in Figures 1619suggests a more limited povertyenvironmentnexus. The spatial correlation with the povertypopulation appears negligible for deforestation,

very weak for sanitation and diarrhea, and neg-ative for outdoor air population. However, thelarge poverty populations in steeply slopedareas suggest that ethnic separation has openedpotential poverty traps. Our evidence alsosuggests a relationship between poverty and in-door air pollution (indexed by cases of acuterespiratory infection).

6. CAUSAL IMPLICATIONS OFHOUSEHOLD-LEVEL RESULTS

Although the lack of time series data pre-vents structural modeling, we believe that

Figure 9. Cambodia: population without access to clean water, 1998. Source: Population Census, 1998.

Table 3. Cambodia: population, poverty and access tosafe water

Variable Model 1 Model 2

Total population 0.241**

Number of poor 1.437**

Log(Total population) 0.587**

Log(Number of poor) 0.186**

Constant 3071.133** 2.398**

N 180 180

R2 0.847 0.664

Dependent variable: Model 1Population without safewater; Model 2Log(Population without safe water).** Significant at the 5% level.



12/22

reasonable inferences about causation are pos-sible for our household-level results. First, weconsider the evidence related to poverty andhouseholds use of polluting wood fuels. Inboth Cambodia and Lao PDR, we find astrong, positive relationship between povertyand use of these fuels. At the same time, we finda generally weak relationship between povertyand deforestation. Since use of wood fuels pro-

motes deforestation, these asymmetric resultssuggest that poverty contributes strongly tohousehold air pollution, but that fuel wood

use (through deforestation) may not contributestrongly to poverty.

Poverty, lack of access to clean water, andintestinal disease are also highly correlated inCambodia and Lao PDR. In this case, it seemsreasonable to infer two-way causation. Ceterisparibus, poverty limits access to clean waterand sanitation. At the same time, sanitation-re-lated diseases exacerbate poverty by reducing

productivity and imposing significant health-care costs on affected households. However,the weaker relationship between poverty and

Figure 10. Cambodia: child deaths, 1998. Source: Population Census, 1998.

Figure 11. Cambodia: urban PM-10 air pollution. Source: World Bank Estimates, 2004.



13/22

sanitation in Vietnam suggests that public

intervention can break this perverse link. Thecritical difference may lie in the Vietnamesegovernments relatively high level of investment

in public health and education (particularly for

women). Extensive literature has documentedthe significance of these factors in reducing dis-ease and mortality, even where access to clean

0

5

10

15

20

25

0 5 10 15 20 25 30


NumberofPM-10d

eaths(rank) = 0.14

Figure 12. Cambodia: rank scatter: PM-10 air pollution deaths versus poverty population.

Province Poor Deforest Slope Wood/

Charcoal

Unsafe

Water

Child

Deaths

PM-10

Deaths

Kampong Chaam 1 1 3 1 1 1 3

Siem Reab 1 2 2 2 1 1 1

Prey Veaeng 1 1 4 1 2 1 3

Kampong Thum 1 3 4 2 1 2 2

Baat Dambang 1 3 2 1 1 1 1

Taakaev 1 1 3 1 1 2 4

Kandaal 2 4 4 1 1 1 2

Kampong Spueu 2 2 1 2 2 2 3

Banteay Mean Chey 2 1 3 2 2 1 1

Kampot 2 2 1 2 2 2 4

Kampong Chhnang 2 3 2 3 2 2 3

Svaay Rieng 2 1 4 2 3 3 4

Pousaat 3 4 1 3 2 2 2

Kracheh 3 4 3 3 3 3 1

Preah Vihear 3 3 2 3 3 3 3

Phnom Penh 3 4 4 1 3 3 1

Kaoh Kong 3 2 1 3 4 4 2

Rotanak Kiri 3 2 1 4 3 3 3

Otdar Mean Chey 4 1 3 4 4 4 _

Stueng Traeng 4 3 2 4 4 3 2Mondol Kiri 4 4 1 4 4 4 4

Krong Preah Sihanouk 4 2 3 3 3 4 1

Krong Kaeb 4 4 4 4 4 4 2

Pailin 4 3 2 4 4 4 4

Note: _ denotes no data for that province.

11stquartile; 22nd quartile; 3 3rd quartile; 4 4th quartile

Figure 13. Cambodia: poverty population and environmental problems.



14/22

water is limited (Merick, 1985; Filmer & Pritch-ett, 1997).

7. SUMMARY AND CONCLUSIONS

In this paper, we have investigated the regio-nal povertyenvironment nexus in Cambodia,Lao PDR, and Vietnam. Our analysis has fo-cused on spatial relations between poverty pop-ulations and environmental problems at theprovincial and district levels (see the summaryinTable 4).

We identify a potential povertyenvironmentnexus in cases where the pattern of regionalsettlement by poor households is strongly asso-ciated with each of five environmental prob-lems: deforestation, fragile soils, indoor air

pollution, contaminated water, and outdoorair pollution. Our results suggest that the nexusis quite different in each country. In Cambodia,

it seems largely confined to household-levelproblems associated with indoor air pollution,contaminated water, and lack of access to ade-quate sanitation. Outdoor air pollution, defor-estation, and fragile lands are not significantlyassociated with poverty at the district level.

We conclude that poor households in Cambo-dia might benefit most strongly from programsthat jointly address poverty and household-level environmental quality. At the same time,all of Cambodias citizens, including the poor,would benefit from more effective measures toreduce the rate of deforestation.

Our results suggest a broader povertyenvi-ronment nexus in Lao PDR, since all fiveenvironmental problems exhibit a spatial corre-lation with poverty. The overlap is particularlystrong in the northern and northeastern regions

of the country. We conclude that the welfare ofthe poor in Lao PDR might be significantlyenhanced by close integration of poverty-allevi-

Province Poor Green Brown

Kampong Chaam 1 3 1

Siem Reab 1 2 1

Prey Veaeng 1 2 1

Kampong Thum 1 4 2

Baat Dambang 1 2 1

Taakaev 1 1 2

Kandaal 2 4 1

Kampong Spueu 2 1 2

Banteay Mean Chey 2 2 2

Kampot 2 1 3

Kampong Chhnang 2 3 3

Svaay Rieng 2 3 2

Pousaat 3 3 2

Kracheh 3 4 3

Preah Vihear 3 3 3

Phnom Penh 3 4 2

Kaoh Kong 3 1 3

Rotanak Kiri 3 1 4

Otdar Mean Chey 4 2 4

Stueng Traeng 4 2 4

Mondol Kiri 4 3 4

Krong Preah Sihanouk 4 2 3

Krong Kaeb 4 4 4

Pailin 4 3 4

Note: Green indicator index (equal weighting): (a) Deforestation

rate, (b) slope greaterthan 16%;

Brown indicator index (equal weighting): (a) Number using wood and

charcoal, (b) number of cases of diarrhea, (c) number without accessto water and toilets, and (d) number of PM-10 air pollution deaths.

Figure 14. Cambodia: poverty population versus Green/Brown environmental problems.



15/22

ation and environmental strategies in all Green

and Brown dimensions. A geographic focus onthe north would appear to be most beneficial.

The case of Vietnam is more eclectic than theother two, suggesting the possibility of a pov-ertyenvironment nexus for fragile soils and in-door air pollution. We conclude that anappropriate povertyenvironment strategy forVietnam might focus on the living conditionsof poor households in steeply sloped areas.

In summary, we find little evidence of a gen-eral povertyenvironment nexus in our threestudy countries. Indoor air pollution is the only

common issue, and its severity depends onheating and cooking practices that are littlestudied as yet. Our evidence suggests that thenexus concept can provide a useful catalystfor country-specific work, but not a general for-mula for program design. Joint implementationof poverty and environment strategies may becost effective for some environmental problems,but independent implementation may be prefer-able in many cases as well.

We recognize that our analysis is far fromexhaustive, and that other environmental prob-

lems may warrant close attention. Possible can-didates include depleted and polluted fisheries,and excessive use of pesticides. Future research

should explore these issues more fully. We also

recognize that sub-district-level analysis mightreveal stronger povertyenvironment links, aswell as providing a better guide for spatial tar-geting of regional programs. For this reason,we hope that future research projects will pro-mote more extensive data collection and analy-sis at the local level.

In addition, we believe that regional coordi-nation of povertyenvironment programs maybe useful, even in some cases where the pov-ertyenvironment nexus does not appear to bestrong in all countries. A good example is pro-

vided by the links between poverty, access tosafe water, and intestinal disease. As we havenoted, the empirical nexus may be weaker inVietnam because government intervention hasalready been effective. A regional program link-ing Vietnamese experts to counterparts in Cam-bodia and Lao PDR could be very appropriatein these circumstances.

Despite these caveats, we believe that ourfindings provide some insights for policy mak-ers who are concerned about the povertyenvi-ronment nexus. Our results suggest that the

nexus is country specific, and institutional fac-tors may play an important role. Data on morecountries would be required for an in-depth

Province Poor Deforest Sloped Wood/

Charcoal

Unsafe

Water

Child

Diarrhea

PM-10

Deaths

Savannakhet 1 4 4 1 1 1 1

Champasack 1 3 4 1 1 1 2

Huaphanh 1 2 1 2 1 1 1Luangphrabang 1 1 1 1 1 1 1

Oudomxay 1 1 2 3 2 2 _

Saravane 2 3 3 2 2 2 2

Khammuane 2 4 3 2 2 1 2

Phongsaly 2 1 1 3 3 4 2

Xiengkhuang 2 2 1 3 2 3 3

Vientiane Municipality 3 1 4 1 4 4 1

Vientiane 3 2 2 2 3 3 4

Luangnamtha 3 1 2 4 3 3 3

Xayabouri 3 2 1 1 1 2 3

Bokeo 4 3 3 4 3 3 _

Attapeu 4 4 3 4 4 2 4

Borikhamxay 4 4 2 3 4 4 3

Sekong 4 3 4 4 4 4 4

Xaysomboon 4 4 4 4 4 4 _

Note: a _denotes no data for that province.

Figure 15. Lao PDR: poverty population and environmental problems.



16/22

analysis of the relationship between a countrysinstitutions and the dimensions of its povertyenvironment nexus. However, the available evi-dence for our three countries does providesome useful insights. In Vietnam, the povertyenvironment nexus (indoor air pollution, fragilelands) is concentrated in northern and westernhighland areas dominated by ethnic minoritypopulations. State policy has focused on out-

door air pollution much more than indoor airpollution; priority has been given to reducinginfant mortality through provision of clean

water; and administrators seem to have man-aged forest clearing to about the same degreein poverty and nonpoverty areas.

Several poorer regions of Lao PDR areinhabited by peoples who are ethnic minoritiesin Vietnam, and the administrative and devel-opment resources of the Lao government ap-pear much more limited than those inVietnam. The result of more limited interven-

tion is apparently a broadening of the nexus,which replicates the crossborder highland prob-lems of Vietnam but also includes water pollu-

Province Poor Deforest Sloped No

Toilets

Child

Diarrhea

Acute

Respiratory

Infections

PM-10

Deaths

Thanh Hoa 1 3 1 1 2 1 1

Nghe An 1 2 1 1 1 1 1Ha Tay 1 4 4 2 1 1 2

An Giang 1 1 4 1 1 1 1

Dak Lak 1 2 2 1 1 1 2

Bac Giang 1 3 4 3 1 4 3

Nam Dinh 1 _ 3 2 1 1

Son La 1 4 1 2 4 2 4

Can Tho 1 1 _ 4 3 3 1

Dong Thap 1 1 _ 4 2 1 2

Thai Binh 1 _ 4 4 1 3

Kien Giang 1 2 4 1 2 2 2

Binh Dinh 1 3 2 1 2 3 1

Ha Tinh 1 4 3 2 3

Quang Ngai 1 2 2 1 3

Quang Nam 2 1 1 1 2 3 2

Phu Tho 2 3 3 3 3 2 2

Gia Lai 2 2 3 1 3

Hai Duong 2 2 4 4 3 1

Soc Trang 2 1 _ 2 1 2 2

Thua Thien - Hue 2 2 2 1 3 4 2

Vinh Phuc 2 3 4 4 3 2 3

Hai Phong city 2 3_

4 2 1 1Lai Chau 2 2 1 2 1 2 4

Binh Thuan 2 2 3 1 4 4 1

Hoa Binh 2 2 2 3 4

Tien Giang 2 4_

4 2 3 2

Ha Giang 2 3 1 2 4

Thai Nguyen 2 3 3 3 3 4 2

Lang Son 2 2 4 2 3 2 4

Note: a blank denotes no data for that province; a _for Slopemeans no land greater

than 16%.

Figure 16. Vietnam: poverty population and environmental problems (top 2 quartile provinces).



17/22

tion, deforestation, and outdoor air pollution.Cambodias public resources and administra-tion also appear weakly developed, but povertyamong highland ethnic minority groups is lesssignificant in this society. With no concentra-tion of poverty in ethnic minority regions,externality-related environmental problemssuch as erosion, deforestation, and outdoorair pollution appear to be more evenly distrib-

uted across poor and nonpoor areas. However,development administration is weaker than inVietnam, and the government has not yet been

successful in addressing any of the household-level environmental problems that are highlycorrelated with poverty (indoor air pollution,contaminated drinking water, lack of accessto sanitation). In Vietnam, by contrast, a rela-tively successful focus on water and sanitationproblems seems to have reduced the provincialpovertyenvironment nexus to indoor air pollu-tion in highland areas where biomass fuels pre-

dominate.Even though our results suggest that the pov-

ertyenvironment nexus can differ substantially

Province Poor Deforest Sloped No

Toilets

Child

Diarrhea

Acute

Respiratory

Infections

PM-10

Deaths

Ben Tre 3 1 _ 4 4

Ha Noi city 3 _ 3 1 1 1

Hung Yen 3 _ 4 4 3 4

Tra Vinh 3 1 _ 1 3 4

Long An 3 4 _ 2 2

Lao Cai 3 3 1 2 3 4 4

Tuyen Quang 3 4 2 3 3 3 4

Quang Binh 3 4 3 3 4

Ca Mau 3 _ 3 4 3 2

Yen Bai 3 3 1 2 3

Quang Ninh 3 4 3 4 1

Bac Ninh 3 _ 4 1 2 4

Lam Dong 3 3 2 2 1 3 2Ninh Binh 3 4 _ 4 3

Vinh Long 3 1 _ 4 2 3 3

Cao Bang 4 4 2 2 4

Phu Yen 4 1 3 1 3

Khanh Hoa 4 1 1 1 1

Quang Tri 4 3 3 3 2 4 3

Ha Nam 4 4 _ 4 4 2 4

Bac Lieu 4 _ 2 4 4 3

Ninh Thuan 4 3 2 1 3

Ho Chi Minh city 4 4 _ 3 2 1

Dong Nai 4 3 4 2 1

Bac Kan 4 4 2 4 4

Kon Tum 4 2 1 3 4

Binh Phuoc 4 4 _ 3 4 3 4

Tay Ninh 4 1 4 4 3

Da Nang city 4 2 3 3 4 1

Ba Ria - Vung Ta 4 1 4 2 1

Binh Duong 4 1 _ 3 4 4 2

Note: a blank denotes no data for that province; a _

for Slopemeans no land greater

than 16%.

Figure 17. Vietnam: poverty population and environmental problems (bottom 2 quartile provinces).



18/22

across countries, we are only beginning to studythe sources of these differences. We have sug-

gested some reasons why the nexus seems tovary in Cambodia, Lao PDR and Vietnam,but the need for more comparative work is

apparent. Since we have no clear basis for iden-tifying country-specific limitations at this point,

it seems advisable to begin future country anal-yses with a broad consideration of potentialpovertyenvironment links. Starting with a

Figure 18. Vietnam: steeply sloped land.

Figure 19. Vietnam: poverty population, 1998. Source: IFPRI, 2001.



19/22

Table 4. Summary of country-wide correlation with poverty

PEN indicator Cambodia Lao PDR

1. Poverty/deforestation rate Weak +ve correlation at the district level Weak +ve correlation at theprovincial level

(prov.:q = 0.46) (district: q = 0.15)** (prov.:q = 0.37)**

2. Poverty/fragile land (slope > 16%) Weakve correlation Weak +ve correlation (prov.:q = 0.29) (district: q = 0.29)** (prov.:q = 0.30)**

3. Poverty/indoor air pollution

(wood/charcoal use)

Strong +ve correlation, supported by

regression model

Strong +ve correlation, support

by linear and log regression mo

(prov.:q = 0.88) (district: q = 0.70)*** (prov.:q = 0.74)***

4. Poverty/access to clean water

(CW), no sanitation (NS) & Diarrhea

Strong +ve correlation, supported by

regression model

Strong (Water) and medium (Sa

+ve correlations, as well as with

Diarrhea casesWater (prov.: qcw = 0.93) (district:

qcw = 0.85)***

Water (prov.: qcw = 0.85)***

Sanitation (prov.: qns = 0.95) (district:

qns = 0.84)***

Sanitation (prov.: qns = 0.43)**

Diarrhea (prov.: q = 0.28)** Diarrhea (prov.: q = 0.75)***

5. Poverty/outdoor air pollution

(number of deaths from PM10)

Weak +ve correlation Strong +ve correlation

(prov.:q = 0.14)* (prov.:q = 0.75)**

Overall povertyenvironment nexus (PEN) PEN largely confined to household-level

problems due to contaminated air & water**Spans most environmental

indices considered***

* Low correlation.** Medium correlation.*** High correlation.


20/22

broad filter also offers the prospect of morecost-effective approaches because it facilitatescomparison across problems that are frequentlyaddressed separately. In our three countries, for

example, simultaneous attention to fragilelands and indoor air pollution facilitates com-parison of intervention costs and potential ben-efits in the two dimensions. 16

NOTES

1. Extensive research has explored the relationship

between environmental degradation and economic

growth. See particularly the special issues of the Journal

of Environment and Development Economics, 2(4), 1997

and Ecological Economics, 25(2), 1998.

2. SeeDasgupta, Laplante, Wang, and Wheeler (2002)

for a related discussion of policy impacts on the EKC in

developing countries.

3. We recognize that structural models will be difficult

to estimate reliability for quite some time, since a

relatively long time series would be necessary to distin-

guish two-way impacts of poverty and environmental

variables from long trends produced by forces such as

demographic change.

4. Henninger and Hammond (2000)make a strong case

for using povertyenvironment maps, which afford

unique insights into the importance of spatial relation-

ships.

5. We are indebted to an anonymous reviewer for this

point.

6. We abbreviate the presentation for Lao PDR and

Vietnam to keep the papers length tractable, and

because our analytical methods are identical to those

used for Cambodia. For a full presentation and discus-

sion of the evidence for Lao PDR and Vietnam, see

Dasgupta, Deichmann, Meisner, and Wheeler (2004).

7. The minimum consumption and poverty estimateshave been produced by the World Bank for Lao PDR,

the World Food Program for Cambodia, and the

International Food Policy Research Institute for Viet-

nam.

8. See the district-level Model 2 results for Cambodia

in Tables 2 and 3. Our model relates an environmental

problem (H) to the poverty count (P), population (N)

and head-count ratio (P/N) as follows:

logHi a0 a1log Pi

Ni

a2logPi a3logNi

a0 a1 a2 logPi a3 a1 logNi:

We use the latter expression for estimation.

9. For Lao PDR, the available data have also enabled

us to test the effect of the poverty gap (the difference

between actual income and the absolute poverty line for

a representative individual). For each province, we have

estimated the total poverty gap by summing across gaps

for all individuals who are estimated to fall below the

poverty line. However, we find the correlation of this

variable with the poverty count (.96) to be so high that

the poverty count seems sufficient for our analysis.

10. See section (b) for a further discussion of the

deforestation model.

11. For a further discussion, seeDasgupta and Wheeler

(1997)and Pargal and Wheeler (1996).

12. Our regression analysis is based on a model of

deforestation in which the representative individual in a

regions population clears a hectares of forest annually.

Forest loss in the region between period 0 and periodt is

therefore represented byFt F0 = aN(a < 0 for defor-estation), where F is the forested area and N is the

regional population. Dividing through by forested area

in period 0 and changing to a logarithmic approxima-

tion, we obtain

Ft F0F0

aN

F0) log

Ft

F0

b0 b1log

N

F0

:

To allow for differential poverty effects, we generalize

this expression to

log Ft

F0

b0 b1log

N

F0

b2log

P

F0

b3logX;

where Pis the regions poverty population and Xrepre-

sents other factors. In this model, b1reflects the average

area cleared by each resident (poor or nonpoor), and b2measures the difference (if any) attributable to poverty.

13. We recognize that the estimated impact of settle-

ment density may be biased by the exclusion of informa-

tion on transport costs and other factors that affect

settlement location, income, and deforestation. However,

our test remains useful if the degree of bias is similar for

poor households and households in general. For a further

discussion, seeCropper, Griffiths, and Mani (1999).

14. SeePandey et al. (forthcoming).



21/22

15. For a further discussion of the relationship between

outdoor air pollution and health, see Holgate, Samet,

Koren, and Maynard (1999)and WHO (2000).

16. Our thanks to an anonymous reviewer for stressing

this implication of the work.

REFERENCES

Akbar, S. & Lvovsky, K. (2000). Indoor air pollution:Energy and health for the poor. ESMAP Newsletter,No. 1. Washington, DC: World Bank.

Albla-Betrand, J. M. (1993). The political economyof large natural disasters. Oxford: Clarendon Press.

Ambler, J. (1999). Attacking poverty while improvingthe environment: Toward win-win policy options.Background technical paper prepared for theSeptember 1999 Forum of Ministers meeting, underthe UNDP-EC Poverty and Environment Initiative;also in Attacking poverty while improving the envi-ronment: Toward winwin policy options (2001),Poverty and Environment Initiative, New York:UNDP.

Arnold, J. E. M., & Bird, P. (1999). Forest and thepovertyenvironment nexus. Paper presented at theUNDP-EC Expert Workshop on Poverty andthe Environment, Brussels, Belgium, January 2021,1999; also in Attacking poverty while improving theenvironment: Toward winwin policy options (2001),Poverty and Environment Initiative, New York:UNDP.

Barbier, E. (2000). The economic linkages between ruralpoverty and land degradation: Some evidence fromAfrica.Agriculture, Ecosystems and Environment, 82,

355370.Bojo, J., Bucknall, J., Hamilton, K., Kishor, N., Kraus,

C., & Pillai, P. (2001). Environment chapter, povertyreduction strategy papers source book. Washington,DC: World Bank.

Bosch, C., Hommann, K., Rubio, G. M., Sadoff, C., &Travers, L. (2001). Water, sanitation and povertychapter, poverty reduction strategy papers sourcebook. Washington, DC: World Bank.

Brocklesby, M. A., & Hinshelwood, E. (2001). Povertyand the environment: What the poor say: Anassessment of povertyenvironment linkages in partic-ipatory poverty assessments. Swansea, UK: Centrefor Development Studies, University of Wales.

Brooks, N., & Sethi, R. (1997). The distribution ofpollution: Community characteristics and exposureto air toxins.Journal of Environmental Economics andManagement, 32, 233250.

Bucknall, J., Kraus, C., & Pillai, P. (2000). Poverty andthe environment, environment strategy background

paper. Washington, DC: World Bank.Cavendish, W. (1999). poverty, inequality and environ-

mental resources: Quantitative analysis of ruralhouseholds. Working Paper Series, 999. Oxford,UK: Centre for the Studies of African Economies,University of Oxford.

Cavendish, W. (2000). Empirical regularities in the

povertyenvironment relationship of African ruralhouseholds. World Development, 28(11), 19792003.

Chomitz, K. (1999). Environmentpoverty connectionsin tropical deforestation. Discussion notes preparedfor the WDR Summer Workshop on Poverty andDevelopment, Washington, DC, July 68, 1999.

Cleaver, K. M., & Schreiber, G. A. (1994). Reversing thespiral: The population, agriculture, and environmentnexus in sub-Saharan Africa. Washington, DC:World Bank.

Cropper, M., Griffiths, C., & Mani, M. (1999). Roads,population pressures, and deforestation in Thailand,19761989. Land Economics, 75(1), 5873.

Dasgupta, S., Deichmann, U., Meisner, C., & Wheeler,D. (2004).The povertyenvironment nexus in Cambo-dia, Lao PDR, and Vietnam. Washington, DC: WorldBank Development Research Group.

Dasgupta, S., Laplante, B., Wang, H., & Wheeler, D.(2002). Confronting the environmental Kuznetscurve. Journal of Economic Perspectives, 16(1).

Dasgupta, P., & Maler, K.-G. (1994). Poverty, institu-tions and the environment-resource base. WorldBank Environment Paper, No. 9. Washington, DC:World Bank.

Dasgupta, S.. & Wheeler, D. (1997). Citizen complaintsas environmental indicators: Evidence from China.World Bank Policy Research Department Working

Paper, No. 1704, January.Durning, A. B. (1989). Poverty and the environment:

Reversing the downward spiral. Worldwatch Paper,No. 92. Washington, DC: Worldwatch Institute.

Dutt, A. K., & Rao, J. M. (1996). Growth, distributionand the environment: Sustainable development inIndia. World Development, 24(2), 287305.

Ecological Economics. (1998). Special issue: The envi-ronmental Kuznets curve. Ecological Economics,

25(2), 143229.Ekbom, A., & Bojo, J. (1999). Poverty and environment:

Evidence of links and integration in the countryassistance strategy process. World Bank AfricaRegion Discussion Paper, No. 4. Washington, DC:

World Bank.Eskeland, G. S., & Kong, C. (1998). Protecting the

environment and the poor: A public goods frame-work and an application to Indonesia. World BankPolicy Research Working Paper, No. 1961. Washing-ton, DC: World Bank.

Ezzati, M., & Kammen, D. (2001). Indoor air pollutionfrom biomass combustion as a risk factor for acuterespiratory infections in Kenya: An exposure-re-sponse study. Lancet, 358(9281), 619624.

Ezzati, M., Singer, B., & Kammen, D. (2001). Towardsan integrated framework for development and envi-ronment policy: The dynamics of environmental

Kuznets curves. World Development, 29(8),14211434.



22/22

Filmer, D., & Pritchett, L. (1997). Child mortality andpublic spending on health: How much does moneymatter? World Bank Policy Research DepartmentWorking Paper, No. 1864, December.

Heath, J., & Binswanger, H. (1996). Natural resourcedegradation effects of poverty and population growth

are largely policy induced: The case of Columbia.Environment and Development Economics, 1(1), 6583.

Henninger, N., & Hammond, A. (2000). Environmentalindicators relevant to poverty reduction: A strategy forthe World Bank. Washington, DC: World ResourcesInstitute.

Holden, S. T. (1996). Adjustment policies, peasanthousehold resource allocation and deforestation inNorthern Zambia: An overview and some policyconclusions. In O. Stokke (Ed.), Forum for Develop-ment Studies, No. 1, 1997. Flekkefjord: NorwegianInstitute of International Affairs.

Holgate, S., Samet, J., Koren, H., & Maynard, R. (Eds.).Air pollution and health. San Diego, CA: AcademicPress.

Journal of Environment and Development Economics.(1997). Special issue: The environmental Kuznetscurve, Journal of Environment and DevelopmentEconomics, 2(4), 357515.

Kepe, T. (1999). Environmental entitlements in Mkam-bati: Livelihoods, social institutions and environ-mental change on the wild coast of the Eastern Cape.Research Report, No. 1, Sussex University, Institutefor Development Studies and PLASS (Program forLand and Agrarian Studies), Sussex, UK.

Leach, M., & Mearns, R. (1991). Poverty and environ-ment in developing countries: An overview study.

Institute of Development Studies, Sussex University,Sussex, UK.

Merick, T. (1985). The effect of piped water on earlychildhood mortality in urban Brazil, 19701976.Demography, 22, 114.

Mink, S. (1993). Poverty, population and the environ-ment. World Bank Discussion Paper, No. 189.Washington, DC: World Bank.

Myers, N., & Kent, J. (1995).Environmental exodusAnemergent crisis in the global arena. Washington, DC:Climate Institute.

Pandey, K. D., Bolt, K., Deichmann, U., Hamilton, K.,Ostro, B., & Wheeler, D. (forthcoming). The humancost of air pollution: New estimates for developingcountries. World Bank Development ResearchGroup Working Paper, Washington, DC.

Pargal, S., & Wheeler, D. (1996). Informal regulation of

industrial pollution in developing countries: Evi-dence from Indonesia. Journal of Political Economy,104(6), 1314+.

Pearce, D. W., & Warford, J. J. (1993). World with-out endeconomics, environment and sustain-able development. New York: Oxford UniversityPress.

Prakash, S. (1997). Poverty and environment linkages inmountains and uplands: Reflections on the PovertyTrap thesis.CREED Working Paper Series, No. 12.London: IIED.

Reddy, S. R. C., & Chakravarty, S. P. (1999). Forestdependence and income distribution in a subsistenceeconomy: Evidence from India. World Development,

27(7), 11411149.Roe, E. (1998). Taking complexity seriously: Policy

analysis, triangulation and sustainable development.Boston: Kluwer Academic Publishers.

Songsore, J., & McGranahan, G. (1993). Environment,wealth and health: Towards an analysis of intra-urban differentials within the greater Accra Metro-politan Area, Ghana.Environment and Urbanization,5(2), 1034.

Surjadi, C. (1993). Respiratory diseases of mothers andchildren and environmental factors among house-holds in Jakarta.Environment and Urbanization, 5(2),7886.

World Bank (1992). World development report 1992Development and the environment. New York: OxfordUniversity Press.

World Commission on Environment and Development(1987). Our common future, report of the worldcommission on environment and development. Oxford:Oxford University Press.

World Health Organization (WHO) 2000. Guidelines forAir Quality, WHO, Geneva. Available from.

http://www.who.int/peh/air/Airqualitygd.htmhttp://www.who.int/peh/air/Airqualitygd.htmhttp://www.who.int/peh/air/Airqualitygd.htmhttp://www.who.int/peh/air/Airqualitygd.htmhttp://www.who.int/peh/air/Airqualitygd.htm

world development vol 33

Documents