labor supply responses to income shocks under credit ... suministrado philippines labor supply...

Poverty Monitoring, Measurement and Analysis(PMMA) Network

Poverty Monitoring, Measurement and Analysis(PMMA) Network

A paper presented during the 4th PEP Research Network General Meeting,June 13-17, 2005, Colombo, Sri Lanka.

Labor Supply Responses to IncomeShocks under Credit Constraints:

Evidence from Bukidnon, Philippines,

Jasmin SuministradoPhilippines

Labor Supply Responses to IncomeShocks under Credit Constraints:

Evidence from Bukidnon, Philippines,

Jasmin SuministradoPhilippines

DRAFT – NOT FOR CITATION

1

Labor supply responses to adverse shocks under credit constraints: Evidence from Bukidnon, Philippines*

Hazel Jean Malapit Action for Economic Reforms (AER)

& School of Economics

University of the Philippines [email protected]

Jade Eric Redoblado Statistics Center


Deanna Margarett Cabungcal-Dolor National College of Public Administration and Governance


Jasmin Suministrado College of Business and Economics De La Salle University, Philippines

[email protected]

DRAFT FINAL REPORT

May 2005

* This work was carried out with the aid of a grant from the Poverty and Economic Policy (PEP) Research Network, financed by the International Development Research Centre (IDRC). The authors would like to acknowledge Chris Scott, Ignacio Francheschelli, A. S. Oyékalé, Bernard Decaluwé, Jean-Yves Duclos, John Cockburn, Agnes Quisumbing, Emmanuel Esguerra, an anonymous referee, and seminar participants at the PMMA Dakar meeting and Université Laval for their helpful comments.


2

1. Introduction

The ability of families to cope with adverse shocks such as crop failure, unemployment, or illness

is an important aspect of vulnerability to poverty. The increasing attention on risk and vulnerability arose

from mounting evidence that shocks inflict permanent effects on human capital formation, nutrition and

incomes. The existence of poverty traps and other forms of persistence has shown that vulnerability to

poverty is in itself a source of deprivation. (Dercon 2001)

Well-being and poverty are the outcome of a complex decision process of households and

individuals given assets and incomes, and faced with risk. On the other hand, vulnerability is an ex ante

concept, determined by the options available to the households and individuals to make a living, the risks

they face, and their ability to handle these risks (Dercon 2001). The ultimate effect of risk on the well-

being of households and individuals depends largely on the coping strategies that may be employed by

the household to protect consumption when adverse shocks occur.

How is consumption insurance achieved in a low-income setting where formal credit and

insurance markets have been observed to be imperfect or missing? As noted by Kochar (1999), it is

widely believed that consumption insurance is achieved through asset transactions, i.e. saving and

dissaving. However, there are a variety of formal and informal mechanisms households may employ to

insure consumption from fluctuations in income. These risk-management strategies include: community

risk-sharing (e.g. reciprocal arrangements, state-contingent remittances), income diversification, adoption

of low-return low-risk crop and asset portfolios, savings depletion, sale of assets, borrowing, and ex post

labor supply adjustments, among others.

Because labor is often the only asset of the poor, this study attempts to measure the extent to

which farm households use labor supplied to off-farm work in the face of adverse shocks and binding

credit constraints. Moreover, this study investigates how this labor supply response differs between

women and men, and on the labor participation of school-age children. While previous research has

concentrated on the ‘added worker effect’ of wives to augment household income when their husbands

become unemployed, this role need not be confined to married women. In fact, the Filipino norm of

maintaining large households may be viewed as a risk-sharing arrangement, where secondary earners,

adults and children, may be called upon to participate in the labor market to maintain household income

when faced with a negative shock to household income.

Intuitively, the smoothing role of the secondary earners’ labor supply should be more important

for the case of poorer households who cannot rely on asset depletion or borrowing to cope with the shock.

The absence of a redistributive system of taxes or transfers, as well as the underdevelopment of insurance


3

and credit markets also contribute to the importance of secondary earners as the primary household

coping mechanism.

This research differs from past studies in its explicit attention to both labor decisions and credit

constraints,1 analyzed within the framework of a collective household model. Contrary to the notion that

households are composed of individuals with common preferences, the collective model allows for

differences in opinion on economic decisions. The relative bargaining strengths of individuals may then

be an important factor in determining how conflicts are resolved within the household. Analyzing labor

supply responses to shocks in this framework allows for the possibility that some members may be

adjusting to income shocks more than others – in particular, members with less bargaining power.

The analysis of how households cope with adverse shocks is especially important in determining

how adjustment costs are distributed within the household. In the collective framework, this may be

linked with the relative bargaining strengths of males versus females, and adults versus children. Over

the long run, the effects of adjustment costs to certain household members may erode the ability of the

household to cope with future shocks, as is the case for example when children sacrifice schooling for

work.

Household responses at the micro level also translate to macro trends in employment, education

and health outcomes, especially when shocks are aggregate in nature (e.g. economic crises and the like).

The increasing volatility in world markets likewise increase the frequency and severity of aggregate

shocks faced by ordinary households. A deeper understanding of how adjustment costs are borne within

the household can inform social protection policy on where interventions are most necessary.

In his analysis of the effect of the East Asian crisis on the employment of women and men in the

Philippines, Lim (2000) finds that women have higher labor force participation rates and longer working

hours relative to men during the East Asian crisis. During this period, he also notes that high school

enrollment rates declined for both males and females, whereas elementary enrollment declined for

females but not for males. Lim (2000) concludes that in crisis times, and specifically in the East Asian

crisis, there was a tendency toward “overworked” females and “underworked” males. He notes that

maintaining and increasing labor market participation of females not previously in the work force

appeared to be an important coping mechanism in the Philippines.

The objective of this paper is to analyze whether women and men increase their market labor

supply in response to adverse shocks and in light of credit constraints. In addition to the labor supply 1 In the labor literature, the increase in household labor supply as a response to fluctuations in household income (e.g. unemployment of the breadwinner, crop failure) is referred to as the ‘added worker effect’. Because the presence of credit constraints limits the set of coping strategies available to households, the ‘added worker effect’ is expected to be stronger when households are unable to borrow to maintain consumption (Cullen and Gruber 1996; Lundberg 1985; Mincer 1962). Labor supply was seldom studied explicitly within the context of credit constraints, with the exception of García-Escribano (2003).


4

response of adult women and men, we are also interested in how child labor is used as a coping strategy.

Children may directly participate in paid labor off-farm, or as unpaid workers at home, to allow adults to

seek outside employment.

In particular, we attempt to answer the following questions: Controlling for the effect of binding

credit constraints, do women and men work more days off-farm when faced with adverse shocks? Are

school-age children more likely to participate in paid or unpaid work in response to adverse shocks and in

light of binding credit constraints? Does relative bargaining power between spouses affect these labor

supply responses?

Our analysis uses the 2003 data from Bukidnon, Philippines, collected by the International Food

Policy Research Institute (IFPRI) and the Research Institute for Mindanao Culture (RIMCU), which

allows us to investigate these issues using two sets of households: (i) ‘original’ households, who are

demographically older and correspond to the same households surveyed two decades ago in 1984-85, and

(ii) ‘split’ households, who are children of original households that are now grown up and have set up

their own households. Comparing our findings for these two groups also allows us to investigate the life

cycle effects of labor responses to adverse shocks.

2. Review of literature

This research builds on two separate strands of literature: (i) the consumption smoothing

literature, and (ii) the literature on the smoothing role of secondary earners.

2.1. Consumption smoothing

The perfect risk-sharing hypothesis implies that, once aggregate shocks are accounted for, the

growth rate of consumption would be independent of any idiosyncratic shock affecting the resources or

income available to the household (Cochrane 1991, Deaton 1992, Townsend 1995, Skoufias and

Quisumbing 2002). Thus, the greater the correlation between household consumption and income, the

less effective the risk-management strategy adopted by the household. This approach has also been used

to assess the role of credit and savings as insurance substitutes, and make inferences on liquidity

constraints2 (Skoufias and Quisumbing 2002).

Although empirical work on consumption smoothing has rejected the full risk-sharing hypothesis

(Cochrane 1991, Townsend 1995), there is evidence that the overall effect of idiosyncratic income shocks

2 One key insight in the simulation results of Deaton (1991) is that a credit-constrained household may still be able to smooth consumption using precautionary savings, thus remaining consistent with the permanent income hypothesis (Skoufias and Quisumbing 2002).


5

on household consumption is not large. This implies that some mechanisms or channels, including those

that in a first best allocation would be considered sub-optimal, absorb most of the shocks. (Garcia-

Escribano 2003)

Research on low-income economies (for example, see Morduch 1995) show that households use a

mix of formal and informal strategies to cope with adverse shocks including: community risk-sharing

(e.g. reciprocal arrangements, state-contingent remittances), income diversification, adoption of low-

return low-risk crop and asset portfolios, savings depletion, sale of assets, borrowing, and ex post labor

supply adjustments, among others. However, different households may have differential access to these

strategies. Poorer households in particular, may be less able to use strategies that rely on initial wealth as

collateral (Skoufias and Quisumbing 2002). On the other hand, it is often possible to adjust labor supply,

regardless of initial wealth.

As noted by Kochar (1999), past research has demonstrated that farm households in developing

countries are able to protect consumption from idiosyncratic shocks but offers little evidence on how this

is achieved. To be able to understand the underlying economic environment, it is important to study how

and to what extent specific mechanisms isolate consumption from the effect of idiosyncratic income

shocks. Much of the work on consumption smoothing has focused on the contribution of assets in

buffering consumption variability (Garcia-Escribano 2003, Kochar 1999). However, these studies may

not be relevant in explaining how consumption insurance is achieved in low-income communities, where

asset levels may be low and access to credit limited.

2.2. Smoothing role of secondary earners

The literature exploring the role of secondary earners in smoothing transitory shocks to the

household head’s earnings may be divided into two. The first set finds evidence of an insurance ef fect of

secondary earners to the extent that it crowds out precautionary savings (Kochar 1995, 1999, Merrigan

and Normadin 1996, Engen and Gruber 2001, Low 1999). Kochar (1995, 1999) concludes that well-

functioning labor markets in Indian villages allow households to increase labor income in response to

crop shocks, reducing the need to resort to asset depletion or borrowing to smooth consumption. Using

UK household data, Merrigan and Normadin (1996) find that precautionary motives are stronger for

households with two earners compared to households with a single earner. Similarly, Engen and Gruber

(2001) find that the effect of an increase in unemployment insurance on wealth holdings is smaller for

married couples than for singles in the US. Lastly, Low (1999) uses numerical methods to show that

precautionary savings in households with a secondary earner is smaller only if the correlation between

shocks to the potential wages of the husband and wife is sufficiently negative.


6

The second set of literature explores the smoothing role of secondary earners through the ‘added

worker effect’. The ‘added worker effect’ refers to the temporary increase in female labor supply

(participation or hours worked) in response to transitory shocks to household income (excluding the

wife’s income). 3 Most studies estimate female employment or female hours worked as a function of the

husband’s labor status together with standard covariates (e.g. labor market characteristics, household

fixed effects). However, some studies have extended the definition of the husband’s (spouse’s) earnings

loss to account for underemployment (Maloney 1987), idiosyncratic earnings shocks other than

unemployment (Garcia-Escribano 2002), and health shocks (Coile 2004).

The presence of liquidity constraints is one of the main arguments put forward in support of the

existence of the ‘added worker effect’ (Mincer 1962; Lundberg 1985; Cullen and Gruber 1996; Finegan

and Margo 1994; Garcia-Escribano 2003). Cullen and Gruber (1996) report evidence that families are

liquidity-constrained during unemployment spells. This finding is consistent with Stephens (2001),

where empirical results for layoffs are consistent with liquidity-constrained households. Similarly,

Garcia-Escribano (2003) finds that households with limited credit access rely on the labor supply of wives

to smooth the husband’s earnings shocks.

The empirical results in the literature investigating the ‘added worker effect’ remains mixed.

Arguments put forward in support for the ‘added worker effect’ include: the substitutability of leisure of

husbands and wives in home production (Ashenfelter 1980; Lundberg 1985; Maloney 1987); an income

effect (Maloney 1987; Prieto and Rodriguez 2000); and, the presence of liquidity constraints (Mincer

1962; Lundberg 1985; Cullen and Gruber 1996; Finegan and Margo 1994; Garcia-Escribano 2003).

On the other hand, other factors that may obscure this effect include: assortative mating in tastes

for work among spouses (Maloney 1991; Lundberg 1985; Cullen and Gruber 1996); the wife’s

employment factors are affected by the same factors causing the husband’s unemployment, or the

‘discouraged worker effect’ (Serneels 2002; Prieto and Rodriguez 2000; Baslevent and Oneran 2001); a

crowding out effect from social insurance programs (Cullen and Gruber 1996; Finegan and Margo 1994);

the value of the unemployment benefit is linked to the wage received by the wife (Cullen and Gruber

1996); complementarity of leisure between spouses and caregiving needs (Coile 2004); and, different

measurement approaches (Lundberg 1985).

Among the knowledge gaps that emerge from this brief review is the consideration of liquidity

constraints. While it has been cited as the driving force for the ‘added worker effect’ in the life cycle

context, few studies explicitly include liquidity constraints in their empirical models. This line of

research is perhaps more relevant for rural areas in developing countries where credit markets are

imperfect and there are little or no unemployment benefits. 3 See Malapit (2003) for a review of literature on the ‘added worker effect’.


7

In addition, only two studies extend the notion of the ‘added worker’ to other family members

(Serneels 2002; Kochar 1999), although in general, the ‘added worker effect’ refers to all potential

secondary earners in the family, including children. This point may have been irrelevant in the developed

country context where households are often nuclear, but it is not so in the case of developing countries. A

number of studies have linked child labor with income shortfalls and credit constraints (Jacoby and

Skoufias 1997, Dehejia and Gatti 2002), emphasizing that parents may be forced to draw on their

children’s labor when other strategies such as credit are not available.

Only a handful of studies on the ‘added worker effect’ use data on developing countrie s,

primarily as a consequence of the dearth of panel data. Such studies would also require analytical

methods more suited to the specific labor market characteristics in the developing country context. Also,

sources of income shocks may be more diverse for agricultural households (not merely unemployment),

and the ‘added worker effect’ is relevant for all potential secondary workers, which include children. An

exception is the work by Kochar (1999), which estimated hours of work responses to idiosyncratic crop

shocks in rural India. Her model distinguishes labor supply by gender, and all household members aged

15 to 45 may contribute to labor income. However, her model does not accommodate credit constraints.

3. Conceptual framework

This section will outline the conceptual underpinnings of this research. We begin with a

discussion of the unitary and collective household models to establish the theoretical relationships we

wish to explore. The next section will discuss the theoretical treatment of permanent versus transitory

shocks and its implications on labor supply. Lastly, we outline the implications of the collective model

on the ‘added worker effect’.

3.1. Unitary model

To simplify the exposition, we begin with the conventional intertemporal family labor supply

model where the household acts as a single decision-making unit (Killingsworth and Heckman 1986).

This approach is adapted to an agricultural setting, where farming is a significant source of household

income. For period t, the farm profit function may be represented by:

(1) ),,,( tttott pxh θππ = ,


8

where oth is own-farm labor hours, tx is a vector of other inputs including hired labor, tp is a

vector of prices, and tθ is the realization of weather and other crop income shocks. To simplify notation,

assume that the household does not distinguish between the labor hours of its members. The household’s

optimization problem is given by:

(2) ∑∞

=−=

0,,),,,(max

t tmt

ott

tchh

zhhcUEUmo β ,

s.t. (2.1) ))(1( 11 ttmttttt chwara −+++= ++ π ,

(2.2) ttmt

ot lhh Ω=++ ,

(2.3) 0≥ta ,

(2.4) 0;0;0;0 ≥≥≥≥ tmt

ott lhhc .

Assuming that utility functions are additively separable over time, the household maximizes the

expected value of time and preference discounted utility (2), subject to the intertemporal budget constraint

(2.1) where profits, πt, is given by equation (1), time-endowment constraint (2.2), credit constraint (2.3),

and the non-negativity constraints (2.4). Total household consumption is given by tc , tz is a vector of

observed and unobserved factors affecting preferences, tw is the market wage, ta is the household’s

assets at the beginning of period t, and 1+tr is the exogenous interest rate in the next period. Farm profits

are defined as the value of output net of all costs excluding family labor, given by (1). The household’s

labor endowment tΩ , may be allocated to own-farm work oth , market work m

th , and non-market work or

leisure tl . For simplicity, the credit constraint takes the form of a non-negativity constraint on assets.4

This model can easily be extended to distinguish the labor hours of members according to gender by

disaggregating hours of work and wages for females and males.

This optimization yields labor supply functions that depend on aggregate consumption, own

wage, hours worked in own-farm, non-labor income, farm profits, earnings of other family members,

marginal utility of wealth (λt), and preference shifters (zt).

(3) ),,,,,( ttttott

mt

mt zwhchh λπ=

4 While there is evidence that households do not draw assets down to zero (Kochar 1999), setting a positive lower limit on assets rather than zero does not significantly alter the optimality conditions.


9

3.2. Collective model

In recent years, there has been growing evidence contradicting the unitary model both in

developed and developing countries.5 An alternative is the collective model, which allows for differences

of opinion on economic decisions among household members. This model suggests that when

disagreement occurs, how it is resolved may depend on the relative bargaining power of individuals

within the household (Manser and Brown 1980, McElroy and Horney 1981, Quisumbing and Maluccio

2003). Time allocation and labor force participation decisions may in fact be the result of previous

bargaining (Quisumbing and Maluccio 2003).

In addition, conflict-resolution within the household is extremely relevant in the choice of coping

strategies. When adverse shocks occur, couples may have differing preferences on which member should

work more, which expenditures should be cut, or which child should work. Because the choice of

strategies undertaken may result in intertemporal trade-offs (e.g. sale of productive assets, or adjustment

of children’s schooling), the collective framework pro vides important implications for policy and

program design.

In a two-person household where preferences are altruistic, so each person (i = m, f ) cares about

the other’s allocation, we can write the single -period maximization problem as:

(4) ),,,,,,()1(),,,,,,(max zhhhhccUzhhhhccU mf

mm

of

omfmf

mf

mm

of

omfmm µµ −+ ,

subject to the same constraints outlined in the unitary case. It is simple to see that this general

model reverts to the unitary model if the individual utility functions are identical (common preferences),

or if the sharing rule µ is equal to zero or one (dictator).

It is likely that this sharing rule is related to the relative bargaining power of individuals within

the household. Letting bm and bf represent proxy measures for bargaining power that influence µ, market

labor supply functions can be expressed as:

(5) )),,(;,,,,,,,( zbbwwhhhcchh fmfmmj

of

omfm

mi

mi µπ=

The effect of individual bargaining power on market labor supply of person i can be interpreted as

the effect of changing the share of household income allocated to each household member, holding

household income constant. The key result of the unitary model, income pooling, implies that the identity

5 See Strauss and Thomas (1995) for a review.


10

of the income earner, or the person in control of the resources is irrelevant. Thus, if income pooling holds,

these effects should be zero:

(6) fmjbh jmi ,,0/ ==∂∂

As suggested by Quisumbing and Maluccio (2003), this provides a straightforward test of the

unitary model by including proxy measures for male and female bargaining power in the estimation of

labor supply functions. Valid proxy measures are culturally relevant factors that reflect bargaining power

and are exogenous to decision-making within marriage. However, a more general test of the unitary

model is to examine whether the effects of the husband’s and wife’s assets are equal (Q uisumbing &

Maluccio 2003). That is, even if the effects of the husband’s and wife’s assets are non -zero, for as long as

they affect labor supply symmetrically, income pooling remains valid. We employ this general test in this

study.

Using anthropological evidence from rural Philippines, Quisumbing (1994) argues that inherited

landholdings are a valid measure of bargaining power since land is usually given as part of the marriage

gift and major asset transfers occur at the time of marriage. Following recent literature, we use assets

brought into the marriage (primarily human capital), as proxy measures of bargaining power (Quisumbing

and Maluccio 2003). One key limitation for using these types of proxy measures is that human capital

variables may also capture, in addition to bargaining power, labor productivity effects. However, we

continue to use human capital in this study in the absence of superior candidate proxy measures.

3.3. Permanent versus transitory shocks

According to the permanent income hypothesis, consumption is constant over the lifecycle and

depends on permanent income. Temporary fluctuations in income are thus smoothed through credit and

savings and should not affect consumption. Following this argument, only permanent shocks should

affect labor decisions.

Contrary to the permanent income hypothesis, the ‘added worker’ hypothesis predicts that

negative transitory shocks to household income, through shocks on farm profits (e.g. crop failure) or

earnings of other family members (e.g. unemployment), will result in a contemporaneous increase in

market hours of work, all other things equal. The theory also implies that the increase in market hours of

work will be temporary, and will no longer be necessary once the shock has subsided.

In his classic article on female labor supply, Mincer (1962) showed that in a given period, the

‘temporary’ reduction in family income due to the husband’s unemployment increases the probability that


11

the wife will participate in the labor market in that period. He emphasized that this effect is expected

when the family have few consumption-smoothing alternatives: “However, if assets are low or not liquid,

and access to the capital market costly or nonexistent, it might be preferable to make the adjustment to a

drop in family income on the money income side rather than on the money expenditure side … a

transitory increase in labor force participation of the wife may well be an alternative to dissaving, asset

decumulation, or increasing debt.” (Mincer 1962)

On the other hand, Heckman and MaCurdy (1980) observed that ‘permanent’ factors resulting to

higher unemployment probability of the husband should increase the labor supply of wives over their

lifetimes, and not only during the periods of unemployment. Thus, in a life-cycle setting, the ‘added

worker effect’ cannot be expected to be large unless in the presence of credit constraints (Lundberg 1985;

Heckman and MaCurdy 1980). Lundberg (1985) notes that without such a constraint, the wealth effect of

a short unemployment spell is likely to be small, and contemporaneous movements in the labor supply of

a married couple will reflect only cross-substitution effects, which are expected to be small.

This research proposes to investigate both types of shocks within a collective framework: those

that are transitory, occurring within the current period, and those shocks that are possibly transitory but

may exhibit some persistence, occurring over the last twenty years. If credit constraints are binding, both

types of shocks are expected to result in labor supply adjustments.

3.4. ‘Added worker effect’ in the collective model

The collective model, where the distribution of consumption and work is determined by relative

bargaining power, has ambiguous implications on the ‘added worker effect.’ Depending on relative

bargaining power, wives may receive a higher or lower share of her husband’s income relative to the

unitary model. If she is entitled to a larger share of her husband’s income, this implies a higher

reservation wage and lower labor market participation. Conversely, if she is entitled to a smaller share of

her husband’s income, it implies that reservation wages for wives are lower than in a unitary setting.

Lower reservation wages implies higher labor market participation, because more job offers become

acceptable.

Also, the ‘added worker effect’ may be larger or smaller depending on relative bargaining power.

For example, a smaller share of the husband’s earnings implies a weaker income effect. However,

depending on the bargaining process in the household, it is also possible for those members with less

bargaining power to absorb a larger share of the cost of the adverse shock. The consumption and leisure

of the more powerful family members may be insured, while the consumption and leisure of the less

powerful family members are the ones that adjust.


12

4. Data description

This study uses 2003 data from Bukidnon, Philippines, which is a resurvey of households from a

four-round panel survey conducted in 1984-85. The household sampling procedure in 1984/85 was

conducted using a quasi-experimental design to compare households that shifted to sugarcane production

and households that did not following the construction of a sugar mill in the province in 1977. The

survey area extended beyond the neighborhood of the sugar mill, to include households that did not have

the opportunity to adopt sugar (due to prohibitive transport costs) but shared a common farming

environment and cultural heritage with sugar-adopting households (Bouis and Haddad 1990). There were

448 households surveyed in all four rounds, and the last three rounds can be aggregated to comprise a full

year.

The 2003 data resurveys 305 of the core 448 households in 1984/85, as well as 257 new

households formed by children from the original households who are now living in separate households.6

From these 562 households, we include 234 original and 229 split households who have both spouses

present. Primarily we use the 2003 data for the labor supply analysis, although we also use variables from

1984/85 to proxy for certain household and farm characteristics that may be suspected of endogeneity.7

4.1. Identifying credit-constrained households

As a general definition, we define a household to be credit-constrained if it would like to borrow,

for whatever purpose, but cannot obtain credit from any source. We shall no longer distinguish between

formal and informal credit sources as they can function equally well in protecting consumption from

income shocks.

One common method of testing for credit constraints is the consumption insurance hypothesis. If

the growth rate of household consumption covaries with the growth rate of household income, then the

household is said to be credit-constrained. However, one cannot simply look at the smoothness of

consumption and know which mechanisms are at work. If labor income can be used to smooth

consumption, consumption will appear to be insured even in the presence of binding credit constraints.

Thus, to identify households that face binding credit constraints, a direct approach based on household

responses to qualitative questions on credit will be necessary.

6 The 2003 survey initially surveyed 311 original households and 261 split households. Of these 572 households, 10 households were dropped due to missing age and/or sex data for at least one of the household members. 7 Split households are assigned the same past household characteristics as the original household where they came from.


13

In the data, the question “If more credit were available for [purpose] in the past 12 months,

would you have used it? Why not?” was included in the Assets, Backyard Production, Family Business,

Farm Production, and Non-Food Expenditures blocks. Based on this question, households responding

“Yes” to the qualitative question are classified as self -reported credit-constrained. We construct an

indicator variable for each block, and then construct summary indicators for production credit constraints

and non-production credit constraints. Households are classified as credit constrained in production if the

household responded “Yes” to the credit constraint question in at least one of the following blocks: Farm

Production, Family Business, or Backyard Production. Households are classified as credit constrained in

non-production if the household responded “Yes” to the credit co nstraint question in either the Assets, or

Non-Food Expenditures blocks.

4.2. Measuring household income shocks

From the theoretical model, labor supply functions depend on a set of variables including farm

profits, non-labor income, and earnings of other household members. Shocks entering through any of

these factors may result in adjustments in market labor supplied for credit-constrained households.

Because our data deals with agricultural households, fluctuations in crop income are significant sources of

household income shocks.

Several approaches may be used to measure crop income shocks. The first alternative is to use

the residual from a profit regression (Kochar 1999). Positive and negative residuals may be treated as

separate shocks, since strategies used by households to respond to positive shocks are expected to be very

different from strategies used to respond to negative shocks. One problem with this approach is that this

residual contains unobserved variables that determine household expectations, as well as measurement

error in profits. Because the profit regression excludes costs of family labor and other family owned

inputs, it also contains unobserved preference shocks that determine leisure choices.

The second alternative is to use standard instrumental variables techniques. This avoids the

problems associated with the first approach if there is an instrument that is correlated with the “true”

idiosyncratic crop shock but not with preference shocks or measurement error in crop profits.

Although the Bukidnon data set provides a wide set of instruments,8 predicted crop income

shocks obtained using instrumental variables techniques did not result to coefficient estimates

significantly different from zero. Alternatively, we include self-reported incidents of adverse shocks

8 Instruments used include rainfall deviations from the long-run average and incidents of crop failure due to drought and pests, as well as their interactions with farm characteristics (e.g. farm size, crop choice), and incidents and duration of illness by household members.


14

occurring between 1984 and 2003. Various sources of shocks are documented including: weather or

environmental shocks affecting crops or livestock (e.g. drought, flooding, pests, diseases); war, civil

conflict, banditry and crime (e.g. theft, military presence); political, social and legal events (e.g.

confiscation of land, land reform), unexpected economic shocks (e.g. unemployment, severe lack of

financing, severe inability to sell inputs); and unexpected events affecting health or welfare of members9

(e.g. death, illness, disablement, divorce, abandonment). Respondents are reminded that the shocks they

report must have been difficult to foresee and must have significantly affected their households.

We construct count data for the number of incidents for each type of shock and distinguish

between two time periods: past shocks are defined as occurring before 2003, while current shocks are

defined as occurring in 2003. Table 1 presents a list of specific shock categories used in the analysis.

4.3. Descriptive statistics

The means and standard deviations for selected variables are presented separately for original and

split households in Table 2. As we expected, the two groups exhibited statistically significant differences

in the means of a majority of the variables, reflecting the life cycle differences between the two sets of

households.

Original households are larger on average, with more prime-aged male members and less prime-

age females than split households. Split households on average have more young children and school-age

children, while original households have more elderly members. Interestingly, the prime-age members of

original households are younger on average compared with prime-age members of split households. It is

possible that children set up their own households after a certain age, while the younger adult children are

more likely to continue living with their parents.

As expected, heads of original households and their spouses are older and less educated than their

counterparts in split households. Based on these averages, it appears that although original households

are ‘older’ in the sense that there are more elderly members and older household heads and spouses, they

actually have a larger pool of prime-age workers.

Original households are also wealthier than split households on average. They own more land,

more rent-earning assets, and more livestock than split households. They are more likely to be engaged in

farming their own land, have higher loans in the past year, and are more likely to welcome more credit for

production purposes. On the other hand, almost half of split households do not farm or own any land.

This could also explain why on average, both males and females in original households work more days

9 Shocks affecting the health and welfare of the household differ from the other shocks in that it can alter the labor endowment of the household. The effect of this type of shock on labor supply is ambiguous.


15

in their own farms compared with split households. While the number of days worked in off farm

employment by males are not statistically different between the two groups, females in split households

work less days on average compared with females in original households.

Because of the longer history of original households, it is expected that they report more incidents

of adverse shocks occurring over the last twenty years compared with split households. On the other

hand, there seem to be no significant difference between the experience of current shocks for original and

split households except for other weather shocks and other welfare shocks. Original households report a

higher incidence of these two shocks during the year, which is plausible because of their greater

involvement in farming and their demographic composition.

5. Empirical analysis

We conduct separate analysis for original versus split households for two reasons. First, because

split households are formed by children of original households, the two groups are not independent,

having shared common characteristics in the past. Second, the two groups of households are at different

points of their life cycle. Original households are expected to have an older demographic composition

compared with the splits, and each group may respond differently to adverse shocks.

First-order conditions from the household’s utility maximization yield market days of work

equations for female and male labor. Because farm households rely primarily on family labor for crop

production, corner solutions (i.e. zero market days of work) are expected to be significant for both

females and males. Thus, market days of work functions may be estimated using Tobit regressions,

where observed days ))(( ⋅mh equal desired days ))(( ⋅∗h when the latter are positive and zero otherwise.

For labor category i in household j, desired market days of work equation is given by:

(7) ijijjijijij VZxh εαθαααα +++′++=∗43210 '''

where xij is a vector of household characteristics and proxy measures of male and female

bargaining power, Zij is a vector of production and demographic shift variables, Vj is a vector of location

dummies, ijθ is a vector of adverse shock variables, and ijε is an error term with mean zero. If credit

constraints are not binding, the sign of 4α is ambiguous, because the set of coping strategies used by the

household to respond to adverse shocks would depend on the accessibility of various coping strategies as


16

well as the bargaining strengths of members. On the other hand, if credit constraints are binding, we

expect 4α to be positive for both persistent past shocks and transitory current shocks.

Because the presence of binding credit constraints narrows the set of coping strategies available

to the household, and consequently increases the importance of labor supply adjustments as a coping

strategy, it is important to incorporate the effect of credit constraints in our analysis of labor supply.

However, since the credit constraint status of the household is endogenous, we cannot simply split the

sample according to the summary indicator variables we have constructed, or include the indicator

variable as a regressor. Instead, we attempt to correct for the presence of binding credit constraints by

first estimating a bivariate probit model of credit constraints:

(8) npnpnpnpnp

probitnp

pppppprobitp

uWkkk

uWkkk

+=>=

+=>=∗

∗

β

β

'where0

,'where0*

*

where probitpk and probit

npk are observable binary outcomes for possibly interrelated underlying

relationships given by *pk and *

npk for production credit constraints and non-production credit constraints,

respectively; pW and npW are variables that explain production and non-production credit constraints;

and, pu and npu are mean zero error terms.

From the bivariate probit estimates, we compute for the inverse Mills’ rat ios and include these as

regressors in the Tobit estimation of the days worked equation for females and males:

(9) ijnpnpppijjijijij IMRIMRVZxh εγγαθαααα +++++′++=∗43210 ''' .

However, since it is likely that the bivariate probit estimates for production credit constraints and

non-production credit constraints are not independent, we computed for the pairwise correlation of the

two inverse Mills’ ratios. The correlation coefficients indicate that the two terms are highly correlated for

split households, but not for original households. Consequently, we include an interaction term for the

two inverse Mills’ ratios for the split households, in addition to the two ratios in equation (9).

Market wage rates & crop profits. Because we are considering multiple-worker households, there

is an empirical issue as to what the relevant market wage is for the household. The conventional approach

to this problem is to take gender- and year- specific village average wages as the wage applicable to broad


17

aggregates of household labor (Rose, 1992; Skoufias, 1994). Kochar (1999) develops an alternative

approach based on the observation that total labor hours in agriculture is the sum of hours spent in distinct

agricultural tasks, with little variation across individuals performing the same task. Thus, wage rates for

aggregate household labor can be calculated as the weighted average of village-year-gender and task-

specific wages, with the share of household time devoted to specific tasks as weights.

However, Kochar (1999) also notes that since observed wages also reflect household decisions on

how much time is spent on each activity, this measure will be endogenous and correlated with unobserved

characteristics affecting market hours. Since our research objectives do not require an explicit measure of

wages, we follow Kochar’s (1999) approach in substituting for market wages its exogenous determinants

(primarily demographic variables) that determine the household’s choice of market activities.

The same approach is used in the treatment of crop profits. The use of instrumental variables

techniques did not result to significant estimates for predicted profits in the Tobit estimation of days

worked. As we noted earlier, however, crop profits may lead to biased estimates due to measurement

errors and unobserved variables. Instead, we include the self-reported incidents of crop failure as

regressors in the labor supply estimation and omit crop profits as a regressor in favor of its exogenous

determinants that determine production decisions. These include farm characteristics, household head

characteristics affecting farm productivity, demographic variables, and location dummies to account for

price levels and level of economic activity.

5.1. Estimating child labor participation

We also wish to explore the effect of adverse shocks on the labor participation of children,

although we cannot perform the same analysis as we have outlined above for the adults. Although the

data technically includes children in the questions on work hours, only very few households report having

children work positive hours. We suspect there may be a significant degree of underreporting because of

the stigma attached to putting children to work. However, we do have information on whether or not a

child has participated in either paid or unpaid work in the last 12 months, although there is no information

on the intensity or type of work undertaken. We use this data to construct our outcome variable of

interest, defined as ‘number of school-age children working’.

Since we are dealing with non-negative count data as our dependent variable, we use a Poisson

procedure to estimate the following equation, with the same set of regressors as in equation (9):

(10) ijnpnpppijjijijij IMRIMRVZxc εγγαθαααα +++++′++=∗43210 ''' .


18

As in the Tobit estimations of labor supply, we again include the inverse Mills’ ratios derived

from a bivariate probit estimation of production and non-production credit constraints, and include the

interaction of these two terms for the sub-sample of split households.

6. Results

6.1. Credit constraint estimates

In our estimation of credit constraints, we include as regressors independent variables that

influence either the demand or supply of credit: household size, number of children below 6, number of

elderly aged 65 and above, age of household head and its square, area of land owned, share of harvest

from sugar, number of adverse shocks occurring before 2003, and a dummy variable =1 if the household

has borrowed at least once in the past year.

The likelihood ratio tests of ρ = 0 for the bivariate probits on credit constraints were highly

significant for both original and split sub-samples, which rejects the null hypothesis that the two

equations are unrelated. Results of the bivariate probits are presented in Table 3.

Original households were less likely to be credit constrained in production activities the more

elderly members there are in the household, and the older the household head. The age variable could be

exerting a positive effect on the credit limit, reflecting the effect of experience, which increases labor

productivity in own production activities. On the other hand, the number of elderly members could also

indicate that original households may be less involved in own-production, reducing the demand for

production credit.

In addition, original households were less likely to be credit constrained in non-production the

more land the household owns, and if the household has ever borrowed money in the past year. Land

owned may be used as collateral and could increase the credit limit available to the household. The

dummy variable indicating past borrowing could proxy for the accessibility of creditors and the

household’s credit worthiness.

As expected, original households involved in sugar production are more likely to be credit

constrained in production, owing to the higher working capital requirement of the sugar crop. Finally,

original households are more likely to be credit constrained in both production and non-production

activities the more adverse shocks it has experienced in the last 20 years. This supports the view that

persistent shocks have lasting effects on household welfare, since shocks occurring in the past continue to

strongly influence current credit constraints.


19

Similar to the findings for original households, split households are likewise more likely to be

credit constrained in both production and non-production activities the more shocks it has experienced in

the past. Also, split households involved in sugar production are more likely to be credit constrained in

production. Finally, split households who have borrowed in the past year are less likely to be credit

constrained in both production and non-production. As we have noted above, this could be capturing

effects of the household’s cred it worthiness and the accessibility of creditors.

The bivariate probits for both sub-samples performed relatively well in predicting actual self-

reported credit status of households. In the sub-sample of original households, the model correctly

predicted the production credit constraint status of 65% of households, and correctly predicted non-

production credit constraint status of 73% of households. In the sub-sample of split households, the

model correctly predicted 70% and 74% of the households’ product ion and non-production credit

constraint status, respectively.

6.2. Labor supply responses

Our findings for the Tobit and Poisson regressions are presented in Table 4. We include the

following independent variable as regressors: past household and farm characteristics to proxy for current

household characteristics (per capita expenditures, household size, per capita land area cultivated and its

square, regional origin, ethnicity, education, age and age squared of household head), demographic

composition of household, proxy measures of bargaining power (education, age and age squared of

husband and wife), present value of assets, and shock variables (past and current), and location dummies.

For original households, we find that on average, males reduce days worked off-farm when faced

with other weather shocks in the same period. On the other hand, females increase days worked in

response to pest infestations before harvest during the current period. These results suggest that females

exhibit an ‘added worker’ effect in response to contemporaneous shocks, while males exhibit a

‘discouraged worker’ effect.

In addition, female labor is also influenced by past shocks. Female household members work

more in the current period in response to the number of deaths in the family over the past two decades.

Death of household members may be considered a permanent shock, which has affected the labor

decisions of women over their life cycle. On the other hand, past shocks from the severe, temporary lack

of financing, have a negative effect on current labor supply of women. One implication of this result is

that, despite the definition of the shock as temporary, it may have had permanent effects on the household

so that it continues to influence current labor decisions. The negative sign is plausible if the underlying


20

reason for the temporary lack of financing is an expansion of production or non-agricultural business,

rather than a downward adjustment in credit limits or a sudden erosion of credit-worthiness.

We also find that adverse shocks did not significantly influence the number of school-age

children participating in paid or unpaid labor for the sub-sample of original households. Because original

households are demographically older households, we expect that they would have less children of

schooling age residing in their households.

On the other hand, the sub-sample of split households present a very different picture compared

with their parents’ households. As to the effect of current shocks on labor supply, w e find that both men

and women decrease their market labor supply in response to current shocks. For the men, the number of

negative economic shocks are significant and negatively related to market days worked off-farm. For the

women, the incidents of pest infestations before harvest during the current period result to a decrease in

days worked off-farm, which is contrary to the response of women in original households. Because of

life cycle differences, it is possible that women in split households are more likely to have younger

children and are thus less able to increase market days worked in response to current adverse shocks. On

the other hand, women with older children may be more flexible in their work decisions because older

children are capable of taking over domestic responsibilities usually performed by adult women.

Another key difference in the labor supply behavior of original versus split households, is that

male labor supply in splits respond positively to the number of deaths in the past, and negatively to the

number of incidents of pest infestation in the past. In the Tobit estimates for original households, past

adverse shocks influence female labor supply, but not male labor supply. Again, this may be consistent

with the inflexibility of women in their domestic responsibilities. In the case of past deaths, males appear

to be performing the ‘added worker’ function in split households. The negative coefficient of past

incidents of pest infestations appears to be consistent with the labor response of males to current adverse

shocks for both sub-samples. To the extent that the current shocks are aggregate in nature, the seeming

‘discouraged worker’ effect may be explained by the sudden lack of demand for male labor.

Most interesting are our findings on the labor participation of children for split households. First,

we find that adverse shocks occurring in the past are more important in explaining the number of children

participating in paid or unpaid work in the current period. These shocks include incidents of pest

infestation, negative political, social and legal events, and incidents of illness of household members. In

addition, more school-age children participate in paid or unpaid work in response to current negative

economic shocks. The significant and positive coefficient estimates for adverse shock variables suggest

that split households are more likely to resort to child labor as a coping strategy versus original

households, partly because split households are more likely than original households to have more

children of schooling age.


21

6.3. Tests of income pooling

We fail to reject income pooling for original households. However, we obtain mixed results for

the tests of income pooling in split households. For the Tobit estimates of male labor supply, we fail to

reject income pooling for split households. This implies that the proxy measures for relative bargaining

power of husbands and wives do not appear to exert an unequal influence on male days worked. On the

other hand, the test on the equality of coefficients for husbands’ versus wives’ education levels were

significant at the 2% level for Tobit estimates of female days worked, and significant at the 1% level for

the number of school-age children working.

In both cases, the husbands’ education level was not statistically different from zero. However,

the wives’ education levels were significant and positive for female days worked, and highly significant

and negative for the number of school-age children working. If higher education levels are associated

with stronger bargaining power, these results suggest a strong preference of females to increasing their

own market labor supply and decreasing the labor supply of school-age children. However, we note that

education levels are also likely to capture labor productivity in addition to bargaining power. If this is the

case, women may be working more days off-farm because they are more likely to receive higher wages.

7. Summary and conclusion

We find that men on average reduce market hours worked in response to current adverse shocks,

for both types of households. To the extent that shocks are aggregate in nature, males may experience

labor demand constraints during the period of the shock. In this case, even if households wish to increase

their market labor supply, there may not be enough opportunities to do so.

On the other hand, women in original households increase market hours worked in response to

current adverse shocks, while women in split households decrease market hours worked. This is despite

the fact that women in original households work more days off-farm on average than women in split

households, so there does not appear to be a labor demand constraint at work for females in original

households. Since split households have younger children on average, women in these households may

be less able to increase their market labor supply because of their child-care responsibilities. Conversely,

women in original households may be more flexible in their work choices because there are older children

who are able to take over household chores or family farm activities. It is also possible that there exist

differences in the characteristics of off-farm employment performed by women in original versus split

households, such that women in split households are also subject to labor demand constraints.


22

We also find evidence for the ‘added worker’ effect for school -age children in split households.

On average, more children participate in paid or unpaid work in response to adverse past and current

shocks. While we do not find the same results for original households, this is not surprising since original

households have fewer children of schooling age on average. Also notable is the highly significant and

negative coefficient of the wife’s education on the number of school -age children working for split

households. This implies that households with better-educated mothers are less likely to put their

children to work, which may be reflecting a preference for keeping children in school. However, our

findings for child labor participation suggest that adverse shocks can inflict permanent effects on

household welfare over time, especially when households are forced to forgo investing on the human

capital of children to be able to maintain current welfare.

Finally, we fail to reject income pooling for original households and obtain mixed results for split

households. However, these results should be interpreted with caution because of some limitations in our

choice of proxy variables for bargaining power. Human capital variables such as years of schooling and

age, as suggested by the literature, are good candidate proxies because they are exogenous to decisions

made within marriage (Quisumbing and Maluccio 2003). At the same time, however, labor market

decisions in farm households are influenced by own-farm productivity, which are very likely to be

correlated with the human capital of household members, especially that of the household head and

spouse. These variables may then be capturing not only the effect of relative bargaining power, but also

its effect on labor productivity.

Although an evaluation of the effectiveness of these labor adjustment strategies is beyond the

scope of this paper, it is clear that households that are disadvantaged with respect to the quality and

quantity of their labor endowments are least likely to cope well with adverse shocks. This implies that

policymakers need not look too far for effective policies that will enable families to protect their welfare

from adverse shocks, as the standard prescriptions of universal education and health care are likely to be

effective in strengthening the labor endowments of households.


23

8. References

Alderman, H., Chiappori, P., Haddad, L., Hoddinott, J., and Kanbur, R. (1995). “Unitary versus

collective models of the household: Is it time to shift the burden of proof?” World Bank Research

Observer, 10 (1): 1-19.

Banerjee, A. and Duflo, E. (2002). “Do Firms Want to Borrow More? Testing Credit Constraints

Using a Directed Lending Program.” Massachusetts Institute of Technology, May 2002

(manuscript).

Bourguignon, F., Browning, M., Chiappori, P., and Lechene, V. (1994). “Incomes and outcomes: A

structural model of intrahousehold allocation,” Journal of Political Economy, 102: 1067-96.

Bouis, H.E., Palabrica-Costello, M., Solon, O., Westbrook, D. and Limbo, A. B. (1998). Gender

Equality and Investments in Adolescents in the Rural Philippines. Research Report 108.

Washington, DC: International Food Policy Research Institute.

Bouis, H. E. and Haddad, L. J. (1990). Agricultural Commercialization, Nutrition and the Rural Poor:

A Study of Philippine Farm Households, Boulder and London: Lynne Riener.

Browning, M., Chiappori, P., and Lechene, V. (2004). “Collec tive and unitary models: A

clarification,” Center for Applied Microeconometrics.

Carter, M. R. (1989). “The Impact of Credit on Peasant Productivity and Differentiation in

Nicaragua,” Journal of Development Economics 31: 13-36.

Chiappori, P. (1988). “Ratio nal household labor supply,” Econometrica, 56: 63-89.

Chiappori, P. (1992). “Collective labor supply and welfare,” Journal of Political Economy, 100: 437-

67.

Coile, C. (2004). “Health shocks and couples’ labor supply decisions,”NBER Working Paper No.

10810.

Cragg, J. (1971). “Some statistical models for limited dependent variables with application to the

demand for durable goods,” Econometrica 39: 829-844.

Cullen, J.B. and Gruber, J. (1996). “Spousal labor supply as insurance: Does unemployment

insurance crowd out the added worker effect?” NBER Working Paper No. 5608.

Daunfeldt, S. (2001). “Intra -household allocation of time to household production activities –

Evidence from Swedish household data,” Umea Economic Studies Working Paper No. 555.

Deaton, A. (1991). “Saving and liquidity constraints,” Econometrica 59: 1221–1248.

Dehejia, R. and Gatti, R. (2002). “Child labor: The role of income variability and access to credit

across countries,” NBER Working Paper No. 9018.


24

Dela Cruz, R. (2003). “Sugar output bre aks slump, meets targets for first time in 20 years,” The

Manila Times, 1 September 2003.

http://www.manilatimes.net/national/2003/sept/01/business/20030901bus5.html

Dercon, S. (2001). “Assessing vulnerability to poverty,” Paper prepared for DfID.

Diagne, A. and Zeller, M. (2000). “Determinants of access to credit and its impact on income and

food security in Malawi,” Research Report, Washington D.C.: International Food Policy

Research Institute, Washington, D.C.

Elson, D. (1994). “Macro, Meso and Micro..”

Engen, E. and Gruber, J. (2001) “Unemployment insurance and precautionary saving.” Journal of

Monetary Economics, 47(3): 545-579.

Fitzgerald J, Gottschalk, P. and Moffitt, R. An analysis of sample attrition in panel data. Journal of

Hum Resources 33(2):251-299, 1998.

García-Escribano, M. (2003) “Does spousal labor smooth fluctuations in husband’s earnings? The

role of liquidity constraints.” Ph.D. Dissertation, University of Chicago.

Gilligan, D., Harrower, S., Quisumbing, A. and Sharma, M. (2004). “How accurate are reports of

credit constraints? Reconciling theory with respondent’s claims in Bukidnon, Philippines,” Draft

manuscript.

Jacoby, H. and Skoufias, E. (1997). “Risk, financial markets and human capital in a developing

country,” Review of Economics and Statistics, 64: 311-335.

Jappelli, T. (1990) “Who is credit constrained in the U.S?” Quarterly Journal of Economics

(February): 219-234.

Killingsworth, M. R. and Heckman, J. J. (1986). “Female labor supply: A survey” in Handbook of

Labor Economics, Vol. 1, edited by Ashenfelter, O. C. and Layard, R. Amsterdam: Elsevier

Science Publishers, 103-204.

Kochar, A. (1995). “Explaining household vulnerability to idiosyncratic income shocks,” American

Economic Review 85 (2) 159-164.

Kochar, A. (1997). Does lack of access to formal credit constrain agricultural production? Evidence

form the Land Tenancy Market in Rural India American Journal of Agricultural Economics

79:754–63

Kochar, A. (1999). “ Smoothing consumption by smoothing income: Hours-of-work responses to

idiosyncratic agricultural shocks in rural India,” Review of Economics and Statistics 81 (1): 50-

61.

Kuismanen, M. (1997). “Labor supply, unemployment and income taxation: An empirical

application,” Unpublished manuscript.


25

Lim, J. (2000). “The effects of the East Asian crisis on the employment of women and men: The

Philippine case,” World Development, 28 (7): 1285-1306.

Low, H. (1999). “Self insurance and unemployment benefit in a life -cycle model of labour supply and

savings,” Institute for Fiscal Stu dies Working Paper No. W98/24.

Lundberg, S. (1985). “The added worker effect,” Journal of Labor Economic 3(1): 11-37.

Malapit, H. J. (2003). “Review of literature on the ‘added worker effect’,” Unpublished manuscript.

Malapit, H. J., Clemente, T. and Yunzal, C. (2003). “Does violent conflict make chronic poverty

more likely? The Mindanao experience,” Paper presented at the International Conference,

“Staying Poor: Chronic Poverty and Development Policy,” University of Manchester, 7-9 April

2003.

Maloney, T. (1991). “Unobserved variables and the elusive added worker effect,” Economica 58

(230): 173-187.

Manser, M. and Brown, M. (1980). “Marriage and household decision making: A bargaining

analysis,” International Economic Review, 21: 31-44.

McElroy, M. B. and Horney, M. J. (1981). “ Nash -bargained household decisions: Toward a

generalization of the theory of demand,” International Economic Review, 22: 333-49.

Merrigan, P. and Normandin, M. (1996). “Precautionary saving motives: An assessment from UK

time series of cross- sections,” The Economic Journal 106 (438): 1193-1208..

Mincer, J. (1962). “Labor force participation of married women: A study of labor supply,” in Aspects

of Labor Economics, Princeton, N.J.: Princeton University Press, 63-97.

Morduch, J. (1995). “Income smoothing and consumption smoothing,” Journal of Economics

Perspectives, 9(Summer): 103-114.

Morduch, J. (1990). “Risk, production and savings: Theory and evidence from Indian households.

Harvard University, Cambridge, Mass., U.S.A. Mimeo

National Economic and Development Authority (2001). The Medium-Term Philippine Development

Plan 2001-2004, Republic of the Philippines, National Economic and Development Authority.

National Statistics and Coordination Board (2000). Philippine Poverty Profile,

[http://www.nscb.gov.ph].

Paxson, C. (1992). “Using weather variability to estimate the response of savings to transitory income

in Thailand,” American Economic Review 82 (1): 15-33.

Quisumbing, A. R. (1994). “Intergenerational Transfers in Philippin e Rice Villages: Gender

differences in traditional inheritance customs,” Journal of Development Economics 43: 167-195.


26

Quisumbing, A. R. and Maluccio, J. A. (2003). “Resources at marriage and intrahousehold allocation:

Evidence from Bangladesh, Ethiopia, Indonesia, and South Africa,” Oxford Bulletin of

Economics and Statistics, 65(3): 283-327.

Quisumbing, A. R. and Montillo-Burton, L. (2002). “The long-run effects of access to financial

services on asset accumulation, economic mobility, and the evolution of well-being: Revisiting

agricultural commercialization in Bukidnon, 1984-2003” Project proposal.

Rose, E. (1992). “Ex ante and ex post labor supply responses to risk in a low income area,”

manuscript, University of Washington.

Skoufias, E. (1994). “Inter temporal substitution in labor supply: Micro evidence from rural India,”

University of Colorado at Boulder, Institute of Behavioral Science, Population Program. Working

Paper No. WP-94-7.

Skoufias, E. and Quisumbing, A. R. (2002). “Consumption Insurance an d Vulnerability to Poverty: A

Synthesis of the Evidence from Bangladesh, Ethiopia, Mali, Mexico and Russia,” Unpublished

manuscript.

StataCorp (2002). Stata Statistical Software: Release 7.0, Stata Corporation, College Station, TX.

Strauss, J. and Beegle, K. (1996). “Intrahousehold allocations: a review of theories, empirical

evidence, and policy issues,” Michigan State University International Development Working

Paper No. 62, East Lansing, MI.

Strauss, J. and Thomas, D. (1995). “Human resources: empirical modeling of household and family

decisions,” in Srinivasan, T. N. and Behrman, J. (eds.), Handbook of Development Economics,

North-Holland, Amsterdam.

Townsend, R. (1995). “Consumption Insurance: An Evaluation of Risk -Bearing Systems in Low-

Income Economies,” Journal of Economic Perspectives 9(3): 83-102.

Zeldes (1989). “Consumption and liquidity constraints: An empirical investigation,” Journal of

Political Economy, 97(2): 305-46.

Zeller, M. and Sharma, M. (1998). Rural finance and poverty alleviation. Washington, D.C.: IFPRI.


27

Table 1: 2003 Shock Categories and Variable Names

Variable Name*

Weather or environmental shocks

ndroughtp/03 Droughtnpestshp/03 Pests or diseases that affected crops before they were harvested (bugs/rats)

Too much rain or floodToo humidEarthquakeLandslidesErosionHigh windsPests or diseases that led to storage lossesCrop loss due to firesPests or diseases that affected livestock (livestock death)Livestock death due to heatOverall bad harvest season

ncivwarp/03 War, civil conflict, banditry or crime shocks

Destruction, confiscation or theft of tools or inputs for productionTheft of cashTheft of stored cropsDestruction or theft of housingDestruction or theft of consumer goodsMilitary presence (reduced mobility/ increased tension)

nnegsocp/03 Negative political, social or legal events

Confiscation of landConfiscation of other assetsLand reformResettlement, villagization or forced migrationForced contributions or arbitrary taxationImprisonment for political reasonsDiscrimination for political reasonsDiscrimination for social or ethnic reasonsContract dispute or default affecting access to landContract dispute or default affecting access to other inputsContract dispute or default affecting sale of products

Negative economic shock

ncapitalp** Severe, temporary lack of financing/ capitalSevere, temporary lack of access to inputsIncrease in input pricesDecrease in output pricesSevere, temporary lack of demand or inability to sell agricultural productsSevere, temporary lack of demand or inability to sell non-agricultural productsUnable to obtain labour at key crop cycle timesBreak down of processing servicesBreak down in transportation servicesUnemploymentUnexpected change in government regulation concerning income generating sources

Shock regarding health or welfare of householdDeath of husbandDeath of wifeOther death , specify _________Illness of husbandIllness of wifeOther illness, specify_________HospitalizationDisablement of working adult household membersDisablement of other household membersDivorceAbandonmentDisputes with extended family members regarding landDisputes with extended family regarding other assetsCo-op failed due to mismanagementUnexpected change in government regulation concerning eligibility for programatic assistance

*Shock variables with suffix "-p" refer to past shocks, occurring between 1985 and 2002; shock variables with suffix "-03" refer to current shocks occuring in 2003.**ncapital03 was excluded because more detailed variables are available for current period credit constraints.

Other negative economic shocks

nothnegecp/03

Death

Illness

ndeathp/03

nillnessp/03

Other weather shocksnothweathp/03

Other welfare shocks

nothwelfp/03

Shock Categories


28

Table 2: Descriptive Statistics of Original and Split Households

Original Households Split Households N=305 N=257

Variable label Mean Std. Dev. Mean Std. Dev. Demographic characteristics no of hh members 6.380 3.340 4.774 2.182 *** no of children aged <6 in hh 0.495 0.753 1.280 0.901 *** no of children aged 6-15 0.833 1.104 1.008 1.196 * no of elderly aged >65 in hh 0.170 0.559 0.012 0.108 *** no of school-age children participating in paid/unpaid work 0.364 0.762 0.311 0.753 no of male household members aged 15-45 1.980 1.953 1.319 0.824 *** no of female household members aged 15-45 1.141 1.191 1.304 0.791 * ave age of male household members aged 15-45 19.632 11.427 29.357 7.406 *** ave age of female household members aged 15-45 17.414 14.081 27.741 5.548 *** no of males aged 15-45 w/ elem educ 0.003 0.057 0.000 0.000 no of females aged 15-45 w/ elem educ 0.000 0.000 0.000 0.000 no of males aged 15-45 w/ secondary educ 0.790 1.119 0.494 0.691 *** no of females aged 15-45 w/ secondary educ 0.554 0.789 0.739 0.833 *** no of males aged 15-45 w/ higher educ 0.302 0.674 0.218 0.467 * no of females aged 15-45 w/ higher educ 0.282 0.573 0.245 0.490 HH characteristics total hectares of land owned 2.418 5.379 0.307 1.704 *** =1 if HH does not farm & does not own land 0.213 0.410 0.463 0.500 *** present value of rent earning assets 14,019 79,589 1,001 6,353 *** net present value of all animals owned by hh 3,797 5,604 2,227 4,512 *** =1 if ever loaned in last 12 months 0.786 0.412 0.723 0.449 * total amount borrowed during past 12 months-all sources 32,726 101,465 10,834 26,281 *** =1 if would like more credit for agri/non-agri prod'n 0.384 0.487 0.296 0.457 ** =1 if would like more credit for non-production 0.272 0.446 0.249 0.433 Work variables days worked in own farm by male hh members 40.570 84.825 18.638 43.552 *** days worked in own farm by female hh members 13.062 40.717 4.403 19.973 *** days worked in all off-farm employment by male hh members 146.666 208.002 143.588 140.682 days worked in all off-farm employment by female hh members 60.525 119.586 40.453 85.790 ** Human capital of spouses highest grade attained by male spouse in hh 15.446 9.426 21.089 7.432 *** highest grade attained by female spouse in hh 17.826 6.474 23.354 6.073 *** age of male spouse in hh 48.433 18.684 31.136 7.783 *** age of female spouse in hh 50.210 10.828 28.891 6.537 ***


29

Table 2: Descriptive Statistics of Original and Split Households, cont’d

Original Households Split Households N=305 N=257

Variable label Mean Std. Dev. Mean Std. Dev. Past shocks no of incidents of drought before 2003 0.387 0.488 0.109 0.312 *** no of incidents of pest infestation before harvest before 2003 0.256 0.437 0.117 0.345 *** no of incidents of other weather shocks before 2003 0.157 0.407 0.062 0.242 *** no of incidents of civil war before 2003 0.128 0.354 0.039 0.194 *** no of incidents of negative political social or legal events before 2003 0.072 0.272 0.012 0.108 *** no of incidents of severe lack of financing before 2003 0.066 0.248 0.035 0.184 no of incidents of other negative economic shocks before 2003 0.052 0.223 0.016 0.124 ** no of incidents of death before 2003 0.246 0.475 0.039 0.231 *** no of incidents of illness before 2003 0.328 0.548 0.276 0.521 no of incidents of other welfare shocks before 2003 0.075 0.277 0.019 0.138 *** no of incidents of all types of adverse shocks before 2003 1.767 1.331 0.724 0.938 *** Current shocks no of incidents of drought in 2003 0.003 0.057 0.012 0.108 no of incidents of pest infestation before harvest in 2003 0.026 0.160 0.012 0.108 no of incidents of other weather shocks in 2003 0.039 0.195 0.008 0.088 ** no of incidents of civil war in 2003 0.013 0.114 0.004 0.062 no of incidents of negative political social or legal events in 2003 0.007 0.081 0.000 0.000 no of incidents of other negative economic shocks in 2003 0.003 0.057 0.016 0.124 no of incidents of death in 2003 0.026 0.160 0.016 0.124 no of incidents of illness in 2003 0.069 0.266 0.093 0.305 no of incidents of other welfare shocks in 2003 0.020 0.139 0.004 0.062 * no of incidents of all types of adverse shocks in 2003 0.207 0.466 0.163 0.420 Note: Means of the two groups were tested using a t-test with equal variances, P > |t|; *** p-value was significant at the 1% level, ** p-value was significant at the 5% level, * p-value was significant at the 10% level.


30

Table 3: Bivariate Probit Results

BIVARIATE PROBIT ORIGINAL HHS SPLIT HHS

[1] [2] [3] [4]

Variables prod'n credit

constraints

non-prod'n credit

constraints prod'n credit

constraints

non-prod'n credit

constraints no of hh members 0.0074 0.0075 0.0320 0.0017 no of children aged <6 in hh 0.0758 0.0053 -0.0331 0.0879 no of elderly aged>65 in hh -0.3358 * 0.2233 0.5059 0.7294 age of household head -0.2863 ** -0.0070 0.1101 0.1026 age of household head squared 0.0027 ** 0.0000 -0.0016 -0.0014 highest grade attained by household head 0.0155 -0.0074 -0.0044 0.0052 total hectares of land owned -0.0042 -0.0535 ** -0.0069 0.0275 percent hectare-months planted to sugar 0.4474 ** 0.2050 0.5392 ** -0.2133 no of shocks experienced in 1984-2002 0.2401 *** 0.2655 *** 0.1989 ** 0.2480 *** =1 if borrowed during past year -0.2364 -0.6408 *** -0.6925 *** -0.3979 * constant 6.4432 * 0.0971 -1.7592 -2.3803 /athrho 0.6529 *** 0.7724 *** rho 0.5736 0.6483


31

Table 4: Tobit and Poisson Regression Results

TOBIT POISSON

ORIGINAL HHS SPLIT HHS ORIGINAL HHS SPLIT HHS

[5] [6] [7] [8] [9] [10]

male days

worked

female days

worked male days

worked

female days

worked

no. of children working


Past HH characteristics per capita expenditures in 1984/85 -31.4207 -16.0901 -27.7507 * 8.2121 -0.5424 * -0.1646 no of hh members in 1984/85 18.2355 -8.0219 26.9838 -16.0134 0.5837 * 0.2121 per capita farm size in 1984/85 -113.0139 ** -194.2571 *** 0.4856 45.6857 0.6888 1.5089 per capita farm size squared in 1984/85 8.4831 22.9782 -7.2065 -14.4727 -0.0936 -0.5437 * years of schooling of hhh in 1984/85 3.5551 4.0559 3.1051 1.5158 0.0830 0.1146 age of hhh in 1984/85 -1.9018 -18.1100 ** 1.2131 1.1074 0.0021 -0.0496 ** age squared of hhh in 1984/85 0.0016 0.0210 ** -0.0013 -0.0011 0.0000 0.0000 ** =1 if hhh in 1984/85 born in misamis -9.7809 -60.6558 44.3201 79.1614 * -0.0716 0.7492 =1 if hhh in 1984/85 cebuano 28.6653 -10.5405 2.1162 47.3391 -0.1618 -0.6134

Demographic & HH characteristics no of children aged 6 -15 -8.7545 -59.9720 *** -15.1399 28.6281 * 0.8328 *** 0.7946 *** no of children aged <6 in hh -30.2848 -64.1777 ** -24.0593 -2.5449 -0.2929 -0.1822 no of elderly aged>65 in hh -57.5544 47.1121 -0.3698 no of hh members living in 45.1424 *** 49.6043 *** 18.3385 -40.9267 ** 0.1573 -0.0297 no of prime age males in hh, aged 15-45 4.8843 -76.8056 *** 35.4406 110.4736 *** -0.0174 0.2754 no of prime age males in hh, aged 15-45 -65.6872 ** -14.6084 -4.7626 137.1148 *** -0.0548 -2.6628 ** ave age of males in hh aged 15-45 6.2185 -3.5047 14.9997 * 1.8139 -0.0421 0.5237 * ave age of males in hh aged 15-45 squared -0.0683 0.1310 -0.2930 * 0.1090 0.0007 -0.0080 * ave age of females in hh aged 15-45 2.5117 -3.5612 13.6021 24.2218 -0.0225 -0.6479 ave age of females in hh aged 15-45 squared -0.0522 0.1429 -0.2270 -0.4093 0.0001 0.0083 no. of prime-age males w/ secondary educ 21.5689 17.9124 -27.4671 23.2811 -0.0247 -0.0929 no. of prime-age males w/ higher educ -25.7577 40.8579 5.2113 17.0708 -0.4618 -2.1184 * no. of prime-age females w/ secondary educ 8.1010 46.0571 7.7202 -79.6364 *** -0.1028 2.3074 *** no. of prime-age females w/ higher educ 25.2156 101.7759 ** -57.2829 -64.6888 0.0837 4.9493 *** present value of rent-earning assets 0.0001 0.0004 -0.0028 * 0.0038 * 0.0000 0.0000 Proxy measures of bargaining power highest grade attained by male spouse in HH 2.3924 -4.4313 2.1707 -4.8445 -0.0080 0.1244 highest grade attained by female spouse in HH 2.3996 -1.0014 3.4932 12.3641 ** -0.0672 -0.3515 *** age of male spouse in HH 69.4904 369.3440 ** -9.0412 -33.5536 -0.0859 0.9512 age of female spouse in HH -12.8720 97.3013 ** -8.5639 -13.8324 0.4171 0.7218 age of male spouse squared -0.5740 -3.4099 ** 0.1920 0.3821 0.0018 -0.0129 age of female spouse squared 0.1204 -0.9555 ** 0.1148 0.3728 -0.0056 -0.0078


32

Table 4: Tobit and Poisson Regression Results, cont’d

TOBIT POISSON

ORIGINAL HHS SPLIT HHS ORIGINAL HHS SPLIT HHS

[5] [6] [7] [8] [9] [10]

male days

worked

female days

worked male days

worked

female days

worked



Past shocks no of drought before 2003 -4.5238 -41.7140 -51.4486 49.6813 -0.1430 0.6491 no of pest infestation before 2003 18.1167 -1.0514 -98.5071 *** 7.9717 0.1326 1.1790 ** no of other weather shocks before 2003 -56.9227 -33.9766 -37.2229 91.8356 0.4803 0.1247 no of civil wars before 2003 5.9917 -31.3237 -38.1839 -70.6091 -0.5454 -0.2168 no of negative social events before 2003 15.0383 -3.4000 43.9519 -4.7771 -0.1968 2.6210 ** no of severe lack of financing before 2003 17.7266 -110.0320 * -78.3927 41.7350 -0.4564 0.6906 no of other negative econ shocks before 2003 75.6872 -21.2379 -65.5733 87.0959 -1.4004 -2.1039 no of deaths before 2003 -53.4198 76.3628 * 107.3691 ** -32.3007 0.5181 1.0342 no of illnesses before 2003 -36.3117 -11.0693 -4.4105 45.5390 0.0124 0.9726 * Current shocks no of pest infestation in 2003 14.2106 211.5482 * -41.7415 -591.4888 ** 0.2405 -13.6223 no of other weather shocks in 2003 -213.9112 *** 27.8218 -47.5264 100.4389 0.5258 -13.6985 no of other negative econ shocks in 2003 229.2142 327.8972 -189.8046 ** 94.4359 0.8476 1.9312 * no of illness in 2003 -36.5765 66.1011 1.6634 25.3626 0.5170 0.5129 no of other welfare shocks in 2003 -28.4363 1.8667 -0.0708 -15.2734

Location dummies municipality 2 263.3871 *** -14.8553 81.9669 -97.1261 1.9813 ** 2.0478 * municipality 3 183.5376 *** 125.0758 19.8459 -37.2893 -0.5592 -0.4872 municipality 4 270.2139 *** 147.3831 * 33.1714 -106.1018 0.4943 0.0978 municipality 5 206.4264 *** 91.3752 52.8182 -114.3237 -0.2246 0.0127 municipality 6 161.5172 ** 122.1637 -7.1637 -74.2235 -0.4641 0.7419 municipality 7 103.0087 226.0637 ** 35.6036 -34.6085 -14.7001 -0.2454 municipality 8 290.4355 *** 108.4794 88.7551 80.2073 -15.2834 -0.4995 municipality 9 284.8045 *** 16.6581 64.7278 -35.6633 0.2718 0.7953 municipality 10 191.7814 ** 189.6115 * -5.5223 -63.4467 -0.1558 1.1021

IMR prod 109.6165 -10.6356 -59.5116 390.4491 * -0.6935 3.1760 IMR non-prod -69.2770 -34.4906 -340.4812 ** 360.8727 0.0816 6.0264 IMR prod*IMR non-prod 97.2624 -259.2658 * -2.5144 constant -1579.5410 -8441.9500 *** -108.1757 -905.5864 -7.5906 -21.5611 _se 193.1729 171.0693 127.6588 143.8535 Pseudo R2 0.0614 0.0982 0.0380 0.0868 0.4797 0.512 Obs. Summary 263 263 252 252 263 252 Left Censored 79 171 37 171 Uncensored 184 92 215 81

F-tests (p-values) husband's educ=wife's educ 0.9987 0.5649 0.7908 0.0130 ** 0.3680 0.0004 *** husband's age=wife's age 0.5194 0.1126 0.9897 0.7209 0.7401 0.8489 husband's age squared=wife's age squared 0.5434 0.1088 0.8943 0.9914 0.5922 0.7691

Note: Values are coefficent estimates; *significant at 10% level; **significant at 5% level; ***significant at 1% level.

labor supply responses to income shocks under credit ... suministrado philippines labor supply...

Documents