EMPIRICAL MEASUREMENTS OF HOUSEHOLDS' ACCESS TO
CREDIT AND CREDIT CONSTRAINTS IN DEVELOPING

COUNTRIES: METHODOLOGICAL ISSUES AND EVIDENCE

Aliou Diagne, Manfred Zeller, and Manohar Sharma

FCND DP No. 90

FCND DISCUSSION PAPER NO. 90

Food Consumption and Nutrition Division

International Food Policy Research Institute
2033 K Street, N.W.

Washington, D.C.  20006 U.S.A.
(202) 862–5600

Fax:  (202) 467–4439

July 2000

FCND Discussion Papers contain preliminary material and research results, and are circulated prior to a full
peer review in order to stimulate discussion and critical comment.  It is expected that most Discussion Papers
will eventually be published in some other form, and that their content may also be revised.





iii

ABSTRACT

This paper presents a new methodological framework for measuring the level of

household access to credit. It provides an analytical framework for examining the

determinants of household credit limits and derives implications on information needed to

examine the extent to which households are credit constrained. Empirical application of

this method involves directly eliciting credit limit information in household surveys.

Illustrations are provided using data from Bangladesh and Malawi. 





v

CONTENTS

Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii

1.  Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

2. Review of Existing Approaches of Measuring Access to Credit and
Credit Constraints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

Detection of Credit Constraint Through Violation of Life-Cycle Hypothesis . . . . 4
Detection of Credit Constraint by Directly Asking Households . . . . . . . . . . . . . . 6

3. Extension of the Direct Method: The Credit Limit Variable . . . . . . . . . . . . . . . . . . . . 8

The Concept of Credit Limit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
The Contract-Theoretic Framework of the Credit Market . . . . . . . . . . . . . . . . . 12
Credit Constraint, Credit Rationing, and Access to Credit . . . . . . . . . . . . . . . . . 15
Access to Credit and Participation in Formal Credit Programs . . . . . . . . . . . . . . 19
Implications for the Measurement of the Impacts of Access to Credit . . . . . . . . 20
“Expectations” and Observability of the Credit Limit . . . . . . . . . . . . . . . . . . . . . 21

4. Extent of Household Access to Formal and Informal Credit in Malawi
    and Bangladesh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

Empirical Distribution of Credit Limits and Unused Credit Lines . . . . . . . . . . . . 31
Access to Credit, Credit Constraints, and Participation in Credit Programs . . . . 34

5. Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39

Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51



vi

TABLES

1 Average maximum credit limit and unused credit line in the informal and
formal market, by country and by gender of respondent . . . . . . . . . . . . . . . . . . . . . 43

2 Percentage of households having access to informal and formal credit: by
membership in group-based credit program and by country . . . . . . . . . . . . . . . . . . 44

3 Percentage of households experiencing a binding credit constraint in the
informal and formal market, by membership in group-based credit program
and by country . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44

FIGURES

1 Distribution of formal and informal credit limits in Bangladesh and Malawi:
Box plot diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

2 Distribution of unused formal and informal credit lines in Bangladesh and Malawi: Box
plot diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

3 Distribution of formal and informal credit limits in Bangladesh and Malawi:
Box plot diagrams by gender of respondents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

4 Distribution of unused formal and informal credit lines in Bangladesh and
Malawi:  Box plot diagrams by gender of respondents . . . . . . . . . . . . . . . . . . . . . . 50



vii

ACKNOWLEDGMENTS

This paper has benefitted from extensive comments from staff members of the

World Bank and the International Food Policy Research Institute. The authors are grateful

to the reviewers for their helpful comments. The paper represents the authors' views, and

should not be attributed to IFPRI or the World Bank.

Aliou Diagne and Manohar Sharma
International Food Policy Research Institute

Manfred Zeller
University of Goettingen, Germany





1

1.  INTRODUCTION

Most policy and research interests regarding rural credit markets revolve around the

perception that poor rural households in developing countries lack adequate access to

credit, which is believed to have significant negative consequences on various aggregate

and household-level outcomes, including technology adoption, agricultural productivity,

food security, nutrition, health, and overall household welfare.

Access to credit affects household welfare outcomes through at least two channels.

First, it alleviates the capital constraints on agricultural households. Expenditures on

agricultural inputs must be incurred during the planting and growth periods of crops, while

returns are received only after the harvest several months later. Therefore, to finance the

purchase of inputs, the farm household must either dip into savings or obtain credit.

Hence, access to credit can significantly increase the ability of poor household with no or

little savings to acquire needed agricultural inputs. Access to credit also reduces the

opportunity costs of capital-intensive assets relative to family labor, thus encouraging

labor-saving technologies and raising labor productivity, a crucial factor for development,

especially in many African countries (Delgado 1995; Zeller et al. 1997).

The second channel through which access to credit affects household welfare is by

increasing its risk-bearing ability and altering its risk-coping strategy. Just the knowledge

that credit will be available to cushion consumption against an income shortfall should a

potentially profitable, but risky, investment turn out badly will induce the household to



2

bear the additional risk. The household may therefore be willing to adopt new, more risky

technologies (Eswaran and Kotwal 1990). A household may also benefit from mere access

to credit, even if it is not borrowing, because with the option of borrowing it can avoid

adopting risk-reducing, but inefficient, income diversification strategies or engaging in

precautionary savings with negative returns.

For these reasons, most developing countries’ governments and donors during the

past 40 years have set up credit programs aimed at improving rural households’ access to

credit. The vast majority of these programs, especially the so-called “agricultural

development banks” that provide credit at subsidized interest rates, have failed both to

achieve their objectives to serve the rural poor and be sustainable credit institutions

(Adams, Graham, and von Pischke 1984; Braverman and Guasch 1986; Adams and Vogel

1985). Meanwhile, most rural households in developing countries continue to rely on the

informal credit market for their intertemporal transfer of resources. They use complex

strategies to increase their productive capacity, share risk, and smooth consumption over

their life cycles. These strategies generally work through self-enforcing informal contracts

among friends, neighbors, and members of the extended family, and are arranged within

networks of informal institutions of diverse natures (Fafchamps 1992; Coate and Ravallion

1993; Lund and Fafchamps 1997). One hypothesis, often advanced by researchers and

policymakers, is that government and NGO-supported credit programs may crowd out the

financial services offered by these informal financial institutions. Hence, understanding

how the nonmarket informal institutions serve households’ demand for financial services



3

and interact with the formal credit institutions set up by governments and NGOs is critical

in identifying policies, institutional designs, and financial services that expand and

complement rather than substitute for the services offered by the existing informal credit

market. An important step for obtaining this policy-relevant information is to quantify the

extent and determinants of households’ access to nonformal and formal credit markets as

well as the severity of their credit constraints.

The objectives of this paper are twofold: to present a new methodological

framework for measuring the level and determinants of household access to credit and

credit constraints, and the effects of access to credit on household behavior and welfare;

and to exemplify the methodology with data collected in Malawi and Bangladesh on the

extent of rural households’ access to credit and credit constraints in developing countries.

Since the novelty of the methodology is mainly the nature of the data collected, the

method of presenting the empirical evidence is almost exclusively tabular and descriptive.

The credit limit is the key variable in the methodology and, to the authors’ knowledge,

data on it have never been collected before. Therefore, the paper discusses data collection

issues regarding this variable. An econometric analysis that formalizes the methodology of

this paper is presented in Diagne (1997, 1999).

The paper is organized as follows.  Section 2 reviews briefly the methodologies that

have been used in the literature to measure household access to credit and credit

constraints. Section 3 presents the new methodology for measuring access to credit and its

effects on household welfare outcomes and discusses the various credit-related concepts



4

underlying the methodology. Section 4 presents the empirical evidence on the extent of

household access to credit and credit constraints in Malawi and Bangladesh. Section 5

concludes the paper with remarks on the policy relevance of the methodology and

empirical evidence.

2. REVIEW OF EXISTING APPROACHES OF MEASURING ACCESS
TO CREDIT AND CREDIT CONSTRAINTS

Broadly speaking, there are two methodologies for measuring household access to

credit and credit constraints. The first and indirect method infers the presence of credit

constraints from violations of the assumptions of the life-cycle or permanent income

hypothesis, while the second collects information directly from household surveys on

whether households perceive themselves to be credit constrained.

DETECTION OF CREDIT CONSTRAINT THROUGH VIOLATION OF LIFE-CYCLE
HYPOTHESIS

One of the testable implications of the simple version of the life-cycle/permanent

income hypothesis (LC/PIH) is that in the absence of liquidity and borrowing constraints,

transitory income shocks should not affect consumption (see, for example, Hall 1978, and

Deaton 1992). Empirical models testing for the presence of credit constraints based on the

life-cycle/permanent income or “consumption-smoothing” hypotheses use household

consumption and income data to look for a significant dependence (or “excess sensitivity”)



5

of consumption on transitory income. Empirical evidence of a significant dependence is

taken as an indication of a borrowing or liquidity constraint. The LC/PIH literature is

extensive and is reviewed by Browning and Lusardi (1996), Besley (1995), Deaton (1992;

1989), Alderman and Paxson (1992), and Gersovitz (1988), among others. In general, the

empirical evidence based on the LC/PIH approach has been inconclusive. At first, one may

think that this is due to the fact that empirical testing of the implications of the LC/PIH

requires repeated observations on the same household, whereas most of the studies are

based on relatively short panels. However, there are more fundamental reasons why the

evidence from the LC/PIH approach for detecting credit constraint has been inconclusive.

First, and perhaps most important, under conditions of uncertainty, violation of the

implications of the LC/PIH can result from prudent or precautionary behavior even if the

household is not credit constrained (Zeldes 1989b; Kimball 1990; Carroll 1991). Several

authors have extended the standard life-cycle model to explicitly incorporate both liquidity

constraint and precautionary behavior, and assess either through simulation or empirical

testing the importance of each effect (Zeldes 1989b; Morduch 1990; Deaton 1991; Paxson

1992). However, due to identification problems, it is nearly impossible to disentangle the

effects of credit constraint and precautionary behavior from the type of income,

consumption, and asset data typically available for these studies (Browning and Lusardi

1996). Second, if conditions of uncertainty are negatively correlated with wealth, then

current income will be negatively correlated with consumption growth, even in the

absence of borrowing constraints (Carroll 1991). Moreover, as Deaton (1990) points out,



6

the effect of negative income shocks on consumption also depends on the initial asset

position of households. Browning and Lusardi (1996, 1832) list several other reasons why

the implications of the LC/PIH may be violated even without a credit constraint.

Furthermore, an important insight from the simulation results of Deaton (1991) is that a

credit-constrained household may still be able to smooth consumption with precautionary

saving and thus not violate any implication of the LC/PIH. Hence, one can conclude that

the violation of an implication of the LC/PIH is neither a sufficient nor necessary condition

for being credit constrained. 

DETECTION OF CREDIT CONSTRAINT BY DIRECTLY ASKING HOUSEHOLDS

The second method of detecting the presence of credit constraint uses information

gained directly from household members on their participation and experiences in the

credit market to determine if they are credit constrained. In practice, households are

classified as credit constrained, based on their responses to several qualitative questions

regarding their loan applications (or lack of) and rejections during a given recall period.

This classification is then used in reduced-form regression equations to analyze the

determinants of the likelihood of a household being credit constrained and the effects of

this likelihood on various household outcomes. The method was first applied by Jappelli

(1990) using data from the U.S. 1983 Survey of Consumer Finances and by Feder et al.

(1990) using data from a household survey in China. The method was subsequently used

by Zeller (1994), Schrieder and Heidhues (1995), and Zeller, Sumiter, and Sharma (1996)



7

with household survey data from Madagascar, Cameroon, and Pakistan; and by Barham

and Boucher (1994) with data from Guatemala. The theoretical justification for the direct

method itself can be found in the extended version of the life-cycle/permanent income

model that explicitly allows for the possibility of a credit constraint. For example, Jappelli

(1990) used the credit constraint inequality in the extended model to derive reduced-form

equations for the determinants of the likelihood of a household being constrained. The

information on household credit-market experiences collected in the direct method is also

an important contribution toward empirically identifying the respective effects of liquidity

constraint and precautionary behavior in life-cycle models (Browning and Lusardi 1996). 

However, despite representing a substantial improvement compared to trying to

detect presence of credit constraint through violation of the LC/PIH, the direct method, as

it has been applied so far, is still incapable of providing a framework that allows the

quantification of the extent to which households are credit constrained or a satisfactory

assessment of the impact of access to credit on household welfare outcomes. Indeed, the

approach is still qualitative in the sense that households are classified as either credit

constrained or not with no information on how severely credit constrained they are.

Similarly, the estimation of the determinants of credit constraint provides only information

on what socioeconomic factors affect the likelihood of a household being credit

constrained. It does not provide the answer to the question of how these socioeconomic

factors affect the extent to which a household is credit constrained. The approach also

does not allow an assessment of the effects of loosening, but not completely relaxing, the



8

credit constraint has on household welfare outcomes. This would valuable information for

policymakers, who are often faced with the need to allocate limited resources to several

competing projects, all of which aim to increase the poor’s welfare. To correct the

shortcomings of the direct method, the next section develops a conceptual framework and

data collection methodology that aims to measure the extent to which households are

credit constrained. 

3. EXTENSION OF THE DIRECT METHOD: THE CREDIT LIMIT VARIABLE

The credit market can be—and often is—analyzed as any other commodity market

would be, using the concept of demand and supply of credit, with the interest rate being

the price of credit. In this framework, the supply and demand curve represents,

respectively, the amount the lender is willing to lend and the amount the borrower is

willing to borrow at exogenously given interest rates. This simplistic Walrasian equilibrium

view of the credit market is, however, not useful for understanding the nature of credit

transactions and their outcomes (Stiglitz and Weiss 1981; Freixas and Rochet 1997). For

one, the assumption of price-taking behavior on the part of the lender is hardly realistic,

especially in a developing country context. For example, village moneylenders are often

known to behave monopolistically toward borrowers, determining interest rates based on

the characteristics of the borrower and on the size of the loans. In the vast majority of

cases of mutual lending between friends and relatives (which dominates the informal credit



9

 Thus, it is extremely rare to find two credit programs charging the same interest rate even if they1

operate in the same geographical areas.

market of Sub-Saharan Africa), lenders choose not to charge interest. Even in the

semiformal credit market dominated by NGOs and government-supported credit programs

targeted at particular sections of the population, lenders set various interest rates

according to criteria related to accessibility, equity, poverty alleviation, or the desire to

increase agricultural output.1

Even without the unrealistic price-taking behavior assumption, more fundamental

problems related to information asymmetries and contract enforcement make the simplistic

Walrasian equilibrium view of the credit market inappropriate for understanding the nature

of credit transactions and their outcomes. These problems are discussed in the next

section.

THE CONCEPT OF CREDIT LIMIT

The inappropriateness of the Walrasian equilibrium view of the credit market can be

seen when the analysis focuses at the level of an individual credit transaction. It is noted at

the outset that credit from any possible source is of limited supply, i.e., lenders are

constrained by factors beyond their control on the maximum amount they can possibly

lend to any potential borrower. This maximum is a function of available resources and is

independent of the interest rate that can be charged and of the likelihood of default. This is

especially true in developing countries, where even established commercial banks are very



10

limited in their ability to raise capital for on-lending. Consequently, any borrower,

however creditworthy, faces a limit on the overall amount he can borrow from any given

source of credit; regardless of the interest rate he is willing to pay, or collateral he is

willing to put up.

Furthermore, due to the possibility of default and lack of effective contract

enforcement mechanisms, lenders have additional incentives to restrict the supply of

credit, even if they have more than enough to meet a given demand and the borrower is

willing to pay a high enough interest rate (Avery 1981; Stiglitz and Weiss 1981). As

shown by Stiglitz and Weiss (1981), the possibility of adverse selection arising from the

asymmetry of information between the lender and the borrower will usually prevent the

lender from using the interest rate as a way to ration credit (i.e., to eliminate excess

demand for credit).

From the borrower’s view, the relevant limit on supply is not the maximum the

lender is able to lend, but rather the maximum the lender is willing to lend. The wedge

between the willing and able represents the extent of the credit rationing that arises mainly

because of information asymmetries. However, whether there is credit rationing or not,

there is a perceived maximum limit that cannot be exceeded when borrowing, regardless

of the interest rate one is willing to pay. This credit limit is the focus of the methodology

for quantifying the extent of household access to credit presented here. 

Before elaborating further on the credit limit, it should be pointed out that there is

nothing new in the concept, as it explicitly or implicitly underlies any discussion of credit



ct & yt & At

11

or liquidity constraints in the LC\PIH literature. However, the formulation of the credit

constraint in the LC\PIH literature focuses on household assets, income, and consumption

variables, and abstracts from all the variables related directly to the credit market. The

former set of variables do affect credit transactions, but only indirectly (except the case of

an asset being used as collateral). The most important credit variable, the actual amount

borrowed, is almost never explicit in a formulation of the credit constraint in the LC\PIH

literature. It is reduced to being simply the residual , where c , y , and A  aret  t   t

household consumption, labor income, and net asset holding, respectively. The credit limit

variable is almost always implicit in the LC/PIH framework. Even in Jappelli’s formulation

where the credit limit variable is explicit, it ends up being eliminated through a reduced-

form specification of its determinants. Consequently, undertaking an empirical analysis

based on this formulation of the credit constraint implies either not collecting or discarding

important information. A conceptual framework that explicitly accounts for the credit limit

is presented in the following sections.

THE CONTRACT-THEORETIC FRAMEWORK OF THE CREDIT MARKET

Because the credit limit is the central concept in this paper’s methodology for

quantifying the extent of household access to credit and the effects that access to credit

has on its welfare outcomes, a model of loan transaction focusing explicitly on the concept



[bmin , bmax] bmin >0

Rl Rl(b) / (1+r)b
Rl

Rl(b , ù ) / (1+r)b + ç (ù ) ç : Ù 6 ú

Rl( )

(bmax , R l(.) ) bmax

Rl

R l : [0 , bmax] 6 ú

b 0 [0 , bmax]

b ( 0 [0 , bmax]

(bmax , R l(.) )

b ( 0 [0 , bmax]

Rb 0 #R b < Rl(b ()

12

 Due to fixed transaction costs, lenders in the formal and semiformal credit markets have also2

minimum loan sizes. That is, the amount to be borrowed is restricted to lie in a interval of the form
, with . Extension to this case is straightforward, and there is no loss of generality in using

0 as a lower bound for loan sizes. 

 We have abstracted from issues regarding collateral in order to simplify the exposition. One3

specification for  is the linear repayment function , where r is the interest rate endogenously
chosen by the lender. A more general specification for  that allows state-contingent repayment of loans
is , where ù  is an element of the set of possible states of nature, Ù, and  is
an exogenous, real-valued random variable. State-contingent repayments of loans is also a very important
feature of the informal credit market as documented by Udry (1995).

 This is not the only sufficient condition for breaching the terms of the contract. Even if R  = R (b ),4                 b  l *

breach of contract may still have occurred if the timing of the repayment (possibly in partial amounts) does
not correspond with what was specified in the contract. To simplify the notation, we did not include timing
issues in the specification of the repayment function .

will be presented. The model follows a contract-theoretic view of loan transaction (e.g.,

Freixas and Rochet 1997). Essentially, the lender chooses the credit limit and the amount

he wants to be repaid; the borrower then chooses the amount to be borrowed within this

range and once the loan is disbursed, whether and when to pay back the loan.

More precisely, the lender chooses the pair  where  is the

maximum amount he is willing to lend, and  is a repayment function

 specifying how much, when, and under what conditions he wants to

be repaid for any given loan size .  The lender then lets the potential2

borrower choose the optimal amount  to borrow. In other words, the

lender offers the contract  to the borrower who accepts or rejects it by his

choice of . The contract is accepted if b  is strictly positive, and rejected if*

b  = 0.  Once the loan is disbursed, the borrower chooses the timing and amount(s) of the*  3

actual repayment(s) . Default occurs when .  Of course, in its choice4



b a
max

bmax

b a
max

bmax

b a
max

b a
max

13

 For simplicity, we abstracted from issues related to how  is determined by conditions in the5

market for on-lending funds (possibility to raise capital from savers or the central bank, determination of
interest paid on savings, regulations from the central bank, etc.). 

of , the lender is constrained himself by the maximum amount he is able to lend to any

borrower, .  5

Some may argue that it is the borrower who first asks for the loan. Then, the lender,

after having received the loan application, decides whether to grant the amount asked or

offer a lesser amount. Furthermore, one can argue that often a loan transaction involves

some bargaining. However, the sequential order in which the loan transaction is started

and whether some bargaining has occurred are details that, in most cases, are not

important for the characterization of the outcome of the loan transaction. What is

important is that, at the end, it is the lender who offers a contract to the borrower who, at

his free will, decides to accept or reject the contract.

Note that the lender’s optimal choice of the credit limit , which is interpreted

here as the supply for credit, is a function of the maximum he is able to lend, . It is

also a function of the lender’s subjective assessment of the likelihood of default and of

other borrower’s characteristics. However, this function is not a supply-for-credit function

in the traditional meaning of the term where, under the assumption of price-taking

behavior, the supply-for-credit function represents the schedule of what the lender is

willing to lend as the market interest rate varies. This traditional supply function for credit

is not defined in this context where the lender himself chooses the interest rate. Similarly,

the optimal interest rate r chosen by the lender is a function of , the lender’s



bmax b a
max

bmax

bmax

14

 In the framework of Stiglitz and Weiss (1981) the optimal interest rate r ( r̂  in their notation) is the6               *

interest rate that maximizes the lender’s expected return from lending to a borrower; taking into account the
possibility of default. It is this expected return which determines  (for a given level of ). Avery (1981)
also uses the same expected return concept to justify the existence of credit limits for borrowers. His
arguments and econometric specification of the determinants of borrowers’ credit limits were based mostly
on interviews with loan officers in New Jersey.

subjective assessment of the likelihood of default, and of other borrower’s characteristics.

The reader is referred to Avery (1981) and Stiglitz and Weiss (1981), respectively, for an

empirical and a formal analysis of how the lender’s assessment of the likelihood of default

affects the optimal choice of both  and r.  On the other hand, the function defining the6

borrower’s optimal choice of loan size b  is a demand-for-credit function in the traditional*

meaning of the term, i.e., the schedule of what the borrower is willing to borrow when the

interest rate varies. The fact that b  is a function of  in addition to being a function of*

the interest rate is a mere reflection of the borrowing constraint. However, because of

imperfections in the enforcement of the loan contract and the resulting adverse selection,

the demand for credit need not be a downward-sloping function of the interest rate.

Hence, as pointed out by Stiglitz and Weiss (1981), lenders cannot use the interest rate as

a way to ration credit. 

CREDIT CONSTRAINT, CREDIT RATIONING, AND ACCESS TO CREDIT

The arguments in the previous sections have postulated that any borrower is

constrained to some extent in his demand for credit from any source. In other words, at

any point in time any potential borrower faces a constraint in the form:



0 # b # bmax .

15

(1)

The question is not whether this constraint exists, but rather how it is formulated.

One advantage of the formulation of the borrowing constraint in equation (1) compared to

the standard formulation in the LC/PIH literature is the fact that the variables directly

related to the credit market are explicit in equation (1). Also, with this formulation there

can be a separate credit constraint for each source of credit (e.g., formal and informal

credits). 

The term credit rationing is often used interchangeably with credit constraint. But

they express two different concepts. Credit rationing occurs because there is a wedge

between what a lender is willing and able to lend and, from the analysis of Stiglitz and

Weiss (1981), this wedge results from the exclusive choice of the lender. In other words, it

is the lender’s decision to further restrict the supply of credit below the amount he is

actually able to lend that causes credit rationing. On the other hand, whether or not a

borrower ends up being effectively credit constrained depends on his optimal loan

demand. Even if, given the constraint set by the lender, the borrower ends up optimally

borrowing an amount strictly less than what the lender is willing to lend, the credit

transaction is still characterized by credit rationing. Furthermore, even though in this

particular case the credit constraint ends up not being binding in the sense that the optimal



16

  Or he may decide to borrow just $3,000 to buy a used car. 7

amount borrowed is strictly less than the maximum amount that could be borrowed, the

borrower may or may not be characterized as being credit constrained.

To illustrate this point, take the example of a borrower who wants to borrow for the

sole purpose of buying an indivisible durable good. To be more precise, assume that he

wants to finance a new car at $15,000. Also assume that the potential lender is rich

enough to be able to lend up to $20,000, but after assessing the borrower’s likelihood of

repaying the loan, he decides that, for whatever purpose, the maximum he is willing to

gamble on this particular borrower is $8,000. Under these circumstances, the borrower

may decide optimally not to borrow at all because any amount borrowed will not be

enough to cover the cost of the car.  In this case, although the borrower ends up7

borrowing an amount strictly less than what the lender is willing to lend him, there has

been credit rationing in the amount of $12,000. Clearly, the borrower is also credit

constrained in the sense that he could not borrow as much as he wanted. Yet, his credit

constraint is not binding because the optimal amount actually borrowed is strictly less than

the $8,000 credit limit. The borrower would not have been credit constrained if the car he

wanted to buy costed only $5,000. But, in this latter case there still would have been

credit rationing in the same amount of $12,000 (it just did not matter for the borrower). In

other words, a borrower may be credit rationed without being credit constrained. On the

other hand, if the maximum the lender is able to lend were in reality $8,000 instead of



bmax

bmax

17

$20,000, then there would be no credit rationing, yet the borrower would be credit

constrained if he wanted to finance a new car at $10,000.

The example illustrates the difference between being credit constrained and being

credit rationed. It also illustrates the ambiguity in the current meaning and use of the

concept of credit constraint. For many, being credit constrained means having a binding

credit constraint, i.e., when the optimal amount borrowed b  is exactly equal to the credit*

limit  (in other words, when the borrower has exhausted his credit line). But, as we

have seen in the example above, this definition does not correspond to the meaning of the

concept of not being able to borrow as much as one wants. It can be seen that the

definition will capture the meaning of the concept only when the loan is sought for the

purpose of acquiring a divisible good; with the additional assumption of monotonic

preference (or local nonsatiation). The formal definition of being credit constrained that

captures or translates the meaning of not being able to borrow as much as one wants is

that the optimal amount borrowed when borrowing under the credit constraint (1) is

strictly less than the optimal amount that would be borrowed if the credit constraint (1)

did not exist. This definition covers both cases of divisible and indivisible goods.

Although a binding credit constraint is not a necessary condition for being credit

constrained, it is still important to know whether or not the credit constraint (1) is

binding at the time credit was sought; and if so, how severe is the binding of the

constraint, i.e., how close to zero is . The lack of access to credit from a given source

of credit can be seen to be an special and extreme case of a severe binding credit



bmax

bmax

bmax

18

constraint, occurring when the credit limit  for that source of credit is zero. By

definition then, one has access to a certain type of credit when its credit limit  for that

type is strictly positive; and one improves one’s access to that type of credit by increasing

 for that type of credit.

It is important to note that with this definition of access to credit, someone may

lack access to credit without being credit constrained. For example, in the empirical

evidence presented below, there were respondents who said they could not obtain any loan

from a formal lender, but gave the reason of not needing any loan or disliking any

borrowing when answering the question, “Why did not you ask for any loan from a formal

lender?” Hence, by definition, these households lack access to formal credit (since b  =max

0), but they are not credit constrained in the formal sector since the optimal amounts they

would have borrowed from a formal lender were the same (zero) with or without the

credit constraint (1). This is also another illustration of the difference between being credit

constrained and having a binding credit constraint (b  = b  in this case). *
max

ACCESS TO CREDIT AND PARTICIPATION IN FORMAL CREDIT PROGRAMS

Access to formal credit is often confused with participation in formal credit

programs. Indeed, the two concepts are often used interchangeably in many credit studies.

The crucial difference between the two concepts lies in the fact that participation in a

credit program is something that households choose to do, while access to a credit



bmax

bmax = 0

bmax > 0

19

 This is not always the case in some credit programs. For example, both the Grameen Bank in8

Bangladesh and the PMERW credit program in Malawi have a rule stipulating that at any point in time only
half of the members of a group or club can obtain credit. The other half of the group will received credit only
when members of the first half have fully paid back their loans.

program is a limiting constraint put upon them (e.g., availability and eligibility criteria of

credit programs). In other words, consistent with the arguments and definitions in the

previous sections, participation is more of a demand-side issue related to the potential

borrower’s choice of the optimal loan size b , while access is more of a supply-side issue*

related to the potential lender’s choice of the credit limit .

The difference between the two concepts may become clearer by remembering from

the arguments above that in any area with a formal credit program, the group of

nonparticipants can be divided into two subgroups, the first consisting of the

nonparticipants who were constrained through eligibility criteria (i.e., they have ),

and the other by those who chose not to participate because their optimal demands for

formal credit were zero, i.e., they have b  = 0. If it is assumed that credit program*

participants are always able to borrow, i.e.,  for participants,  then it is clear that8

the second subgroup of eligible nonparticipant do have access to formal credit. Some may

be uncomfortable to see some nonparticipants classified as having access to formal credit.

However, this is consistent with the fact that program participation is not a separate

spontaneous decision, but is rather part of the demand for formal credit that, under

uncertainty, is a contingent plan (a sequence of acts contingent on the occurrence of

random events) formulated well in advance of the time of possible borrowing. The



ù 0 Ù

ù 0 Ù

bmax > 0

bmax

bmax

20

decision to participate in a credit program is just a contingent act within that contingent

plan (which includes other contingent acts such as how much to ask to borrow, when to

ask it, etc.); and it is implemented only if the demand for credit is not unconditionally zero,

i.e., if b (ù ) > 0 for some , and the events upon which that decision was*

conditioned do occur. Hence, if a nonparticipant does not anticipate wanting formal credit,

i.e., if b (ù ) = 0 for all , there is no need for him to consider joining a formal credit*

program, even if, with a , he could possibly borrow from that program, i.e., by

definition he has access to formal credit.

IMPLICATIONS FOR THE MEASUREMENT OF THE IMPACTS OF ACCESS TO
CREDIT

One of the most important policy and research questions regarding credit markets is

often phrased in terms of how access or improved access to credit translates into change in

household’s outcomes such as agricultural output, income, food security, etc. This

question is central in many decisions regarding government and NGO-supported credit

programs where the economic benefits of providing households access to credit are

compared to the economic costs of setting up these programs and delivering the credits to

the target households. In the previous sections, access to credit and improvement in

access to credit were identified, respectively, with a strictly positive and increasing credit

limit, . Hence, in the framework of this paper the above important policy and research

question is translated as how an increase in  from 0 to a positive value and an increase



bmax

lmax

bmax

21

in  from any strictly positive value translates into change in the household’s behavioral

and welfare outcomes of interest. The methodological implications of using the credit limit

variable to measure the effect of access to formal credit on household welfare outcomes

are discussed in detail in Diagne (1997b), which applies the methodology to data from

Malawi.

“EXPECTATIONS” AND OBSERVABILITY OF THE CREDIT LIMIT

To this point in the analysis, the roles of borrowers and lenders have been kept

distinct. However, in the informal credit market of developing countries, it is not always

possible to classify a household as either a lender or a borrower, because many households

are borrowing and lending at the same time. This is explained by the insurance function

played by the informal credit market (Fafchamps 1992; Coate and Ravallion 1993; Lund

and Fafchamps 1997). Hence, under these circumstances the household demand for credit

cannot be modeled separately from its supply, as one of the motivations for giving out

loans and gifts to other households is to preserve or improve access to credit in case the

need for borrowing arises. In other words, when a household acting as a potential lender is

faced with a demand for credit, its choice of a credit limit  for the borrowing

household is naturally influenced by its desire or not to maintain or increase a strictly

positive , which will be chosen by the borrowing household when there is a need to

borrow from that same household.



bmax

bmax

bmax

bmax

22

These observations suggest that the credit limit a borrower faces depends on both

the lender and borrower’s characteristics and actions. But it also depends on random

events that affect the fortune of lenders and other potential borrowers (who may compete

with the borrower for the same possible credit). For example, one can expect the

occurrence of drought in a rural agriculture-based economy to reduce the supply of

informal credit, while increasing the number of people looking for loans. Hence, the credit

limit  facing a potential borrower is a random variable whose value is determined by

events, some of which are under the borrower’s control, others under the lender’s control,

and still others outside the control of either both. Therefore, the borrowing constraint (1)

is stochastic, and depends on the prevailing state of nature. This implies that whether and

where the credit constraint will bind cannot be known in advance for sure. Thus, forward-

looking households may anticipate possible binding of their credit constraints and take

precautionary measures (e.g., saving and income diversification) to ensure that their

consumption and/or production plans will not suffer much when this occurs (Deaton

1991).

That  depends on random events also implies that its realized value at the times

when borrowing actually takes place cannot be known exactly in advance (i.e., with

probability 1) by either the lender or the borrower. The fact that  cannot be known in

advance by the borrower is clear, since it ultimately is the result of the lender’s choice

(although, as explained above, the borrower can influence that choice to some extent).

The borrower can only form expectations about the likely value of  at the time of



bmax

bmax b a
max

b a
max bmax

b a
max

bmax

bmax

23

actual borrowing. However, formal lenders usually provide enough information about

their loan policies (eligibility criteria, types of projects funded, collateral and down

payment requirements, etc.) to enable potential borrowers to have reasonably accurate

expectations about their  from each source of formal credit. In the cases of NGO- and

government-supported credit programs, they even usually set and announce fixed credit

limits for all potential borrowers.

In general, the potential lender also cannot know in advance of the time of actual

borrowing the exact value of , because his choice will depend on , the maximum

he would be able to lend at that time, which in turn depends on random events outside of

his control. The lender can only form expectations about those random events and about

. But, again for formal lenders,  is usually very small for any single borrower

compared to , so that even unexpected (relatively small) changes in the latter would

not have any influence in the choice of the former. However, for government- and NGO-

supported credit programs, uncertainty about institutional arrangements, program

objectives, target clientele, and external funds from donors can have significant influence

on .

Furthermore, at the time of borrowing, it is only the lender who observes the

realized value of  (which he himself determines), and may or may not have the

opportunity to reveal it to the borrower. For example, the outcome of the loan transaction

might be characterized by 0 < b (ù ) / b(r,b ,ù ) < b (ù ).  That is, if the borrower’s*
max   max

realized optimal choice of loan size is strictly positive, but also strictly less than the



bmax

bmax

bmax

bmax

bmax

bmax

bmax bmax

24

 The borrower’s optimal loan size b  depends on other environmental and borrower variables.9       *

However, in the expression above we have abstracted from these variables to simplify the notation. Also, to
simplify the notation, we use the same notation ù  for both the true (but unknown) state of nature and for a
generic element of Ù, the set of all possible states of nature. There should be no confusion because the
meaning should be obvious from the context.

realized value of , then the lender may never have the chance to tell the borrower his

actual realized choice of .  Clearly, if at a particular time a borrower does not ask for9

a loan from a given source of credit, he will never learn, even in retrospect, about his

realized  from that source of credit at that time (there may be exceptions in the cases

of NGO and government-supported credit programs that set and announce fixed credit

limits for all potential borrowers). However, the potential borrower will always have

expectations on what would have been the likely value of  at that time. Furthermore,

it is precisely the borrower’s expectations about the likely value of  and its variability

that influence his behavior and make him decide in particular whether to seek a loan from

that particular source of credit. For example, in the direct method of detecting credit

constraint discussed in Section 2, the classification of borrowers usually includes a class of

“discouraged borrowers” (e.g., Jappelli 1990). These borrowers did not seek any loan

because either they expected to face zero or very low , or they expected a relative

high cost (including transaction costs) to obtain loans. The discouraged borrowers may

have been wrong in their expectations and could perhaps have obtained worthwhile loans

at reasonable costs. But, whether they are wrong or not, in the end it was the expectations

about their  that determined their behavior, not the realized values of their , which

will remain unknown to them. Even when a borrower seeks a loan from a given source of



bmax

bmax

bmax

bmax(ù )

bmax(ù )

bmax

bmax(ù )

bmax

bmax

bmax bmax

bmax

bmax

25

credit, the realized value of the optimal loan size is largely determined by his expectations

about his  (especially if the borrower has reasonably accurate information that allows

him to accurately predict the location of ).

Because of the above intuitively justified reasons, it is very important that the

realized optimal choice b (ù ) / b(r,b ,ù ) be a function of the random variable , not*
max

of its possible realized value . This allows the borrower’s realized optimal loan

size to be determined not only by the realized value  at time of borrowing, but also

by his expectations about the likely value of  and its variability, made well in advance

of the time of borrowing. Otherwise, allowing b  to depend solely on the possible realized*

value  corresponds to a myopic behavior in the part of the borrower, i.e., the

borrower would be making his borrowing choices not using his subjective assessment of

the relative likelihood of the random events or any relevant information pertaining to those

events that determine the realized value of .

Note also that the fact that the borrower’s optimal loan size depends primarily on

his expectations about  does not mean that in the formulation of the credit constraint

(1),  should be replaced by the borrower’s expected value for . The credit

constraint (1) is a stochastic constraint that must be satisfied for every possible state of

nature ù , regardless of whether the borrower correctly predicts the location of . The

way b  depends on the borrower’s expectations about  is endogenous to the choice*

problem and is governed by the shape of his utility function (assuming that the objective

for borrowing is to maximize the utility of consumption). Although the general functional



bmax

bmax bmax

bmax

bmax

bmax bmax

bmax

bmax(ù )

bmax

26

  However, Lund (1995) managed to interview both borrowers and lenders in her survey in the10

Phillippines.

form of the dependence of b  on  has been justified intuitively here, it can be derived*

rigorously. In general, if the utility function is not too restrictive, under standard

assumptions on expectations formation b  depends not only on the expected value and*

higher moments of , but also on  separately as a random variable (see Diagne

[1994] for examples of univariate and multivariate demand functions derived under

uncertainty). 

The arguments in the previous paragraphs imply that in the analysis of the demand

for credit, the borrower’s expectations about  are much more important in

determining the demanded amounts of credit than the realized values of . However,

from a policy viewpoint, what is of most interest is not the borrower’s response to change

in his expectations about , but his response to change in  itself, since it is the

variable under the lender’s control and that determines access to credit. In the empirical

evidence presented in the next section, it is the borrower’s expected  from different

sources of credit that were collected. The survey did not collect data on  or the

realized values of , which only lenders could provide with reasonable accuracy. The

survey focused on the demand side of the credit market, and for a relatively large survey it

was not feasible to interview the lender for each loan transaction. Moreover, borrowers

may not be willing to identify their informal lenders or may refuse to be interviewed if they

know that the latter are going to be interviewed too.  Hence, the question remains that if10



bmax

bmax

bmax

bmax

bmax

bmax

bmax

bmax

bmax

bmax bmax

bmax

bmax

bmax

27

the only information collected is expected , how one can estimate and evaluate the

impact of  itself (or its realized value) on b  (or on any other household choice or*

outcome variables). For a purely descriptive analysis aimed at comparing households with

regard to the extent of their access to credit (i.e., comparing their ) and binding credit

constraint (i.e., comparing their  with respect to their b ), the inference that can be*

done based solely on expected  and the realized value of b  is very limited.*

Furthermore, the classification of households based solely on a descriptive analysis using

expected  can only reflect the households’ perceptions of their situation on the credit

market as in the direct method of detecting credit constraint households (Jappelli 1990;

Feder et al. 1990).

However, with an econometric analysis it is possible to estimate and evaluate the

impact of  on b  and other household choice or outcome variables based solely on*

expected . For this to be possible, it is necessary to assume that the realized b  and*

other household choice or outcome variables depend only on expected  and not on

higher moments of  and its realized values. This restriction is plausible if  does

not vary much (so that its variance and higher centered moments are close to zero) and the

borrower has reliable information that allows him to predict the location of  with

reasonable accuracy (so that the realized value of  will have little influence on realized

optimal choices). Under this restriction, the assessment of the impacts of  on

household choice and outcomes variables can be done by exploiting the proprieties of the



28

  Even Avery (1981) whose study focused on the credit limit did not have data on it. His econometric11

model was based on a disequilibrium model, where the observed consumer’s debt is the minimum of the credit
limit (supply of credit) and the demand for credit.

mathematical expectation operator, which, as usual, is identified with the borrower’s

expectation process (see Appendix for details). 

4. EXTENT OF HOUSEHOLD ACCESS TO FORMAL AND INFORMAL
CREDIT IN MALAWI AND BANGLADESH

This section presents evidence on the extent of household access to credit in Malawi

and Bangladesh. Following the methodology presented in the previous section, the credit

limit variable is used to assess the extent of households’ access to formal and informal

credit, as well as the binding of their credit constraints. As noted in the previous section,

there is nothing new in the concept of credit limits. However, to the authors’ knowledge,

IFPRI’s household surveys in Bangladesh (1994) and Malawi (1995) were the first to

collect data on the credit limit variable.  For this reason, details will be given on the way11

the credit limit data was collected in the Malawi survey to let the reader be aware of some

of the problems encountered in the process.

The Malawi survey was a year-long, three-round survey (February–December 1995)

that collected detailed information on various activities undertaken by 404 households in

45 villages in five districts of Malawi. A stratified, random-sampling procedure was used

to select the households surveyed. About half of the sample were participants of four



29

 The five credit programs were Malawi Rural Finance Company (MRFC), Promotion of Micro-12

Enterprises for Rural Women (PMERW), the Malawi Mudzi Fund (MMF), and the Malawi Union of Savings
and Credit Cooperatives (MUSCCO). The first three programs are based on group lending while MUSCCO
is a union of individual, membership-based local saving and credit unions.

credit programs.  The second half of the sample was equally divided between past12

participants (mostly from a failed government credit program), and households that had

never participated in any formal credit program. The reader is referred to Diagne, Zeller,

and Mataya (1996) for details on the survey and data collection methodology.

The credit and saving module of the questionnaire was administered to all adult

household members over 17 years old. For the question related to the credit limit, they

were asked the maximum amount they could borrow during the recall period from both

informal and formal sources of credit. If the respondent was involved in a loan transaction

as a borrower, the question was asked for each loan transaction (both for granted and

rejected loan demands). In this case, the credit limit refers to the time of borrowing and to

the lender involved in that particular loan transaction. If the respondent did not ask for a

loan, the question was asked separately for formal and informal sources of credit with no

reference to particular formal or informal lenders. Respondents who were granted loans

were also asked the same general question (i.e., with no reference to particular formal or

informal lenders) in a way that elicited the credit limit they would face if they had wanted

more loans, not just from the same lender, but from the same sector of the credit market

(formal or informal) where they had borrowed. Consequently, for both formal and

informal credit, the maximum formal and informal credit limits of each adult household



30

member were obtained in each round, even if the member was not involved in any loan

transaction. The question on the credit limit in the Bangladesh survey was more or less

similar, except it was asked only to those involved in a loan transaction (see Zeller,

Ahmed, and Sharma 1996).

Given the central importance of the credit limit variable for the methodology of the

study, several control questions were used to verify the consistency of the answers given

by respondents. Such control questions included their program membership status;

whether, if they did receive a loan of the same type, they were given a lesser amount than

asked for, and if yes, how much had they asked; whether they asked for a loan and were

rejected; and why they did not ask for any (or more) loans. In addition, the enumerators

were instructed to use other control questions not included in the questionnaires whenever

there appeared to be inconsistencies in the respondents’ answers.

A good deal of time was spent checking the consistency of answers to these

questions and in relation to other parts of the questionnaire. As a result of these checks,

during the first round of the survey most respondents were visited at least twice to verify

their answers or clarify apparent inconsistencies in their answers. Most inconsistencies

occurred in the first days of the survey, and resulted from some misunderstanding of the

question, which was often interpreted by either the enumerator or the respondent as

asking the maximum you would like to borrow. The misunderstanding was resolved by

instructing the enumerators to explain to respondents, before they answered the question,

the difference between the two questions.



31

 To correct for the oversampling of credit program participants, the summary statistics in the tables13

have been weighted using the strata population weights from the village and national population censuses.

 In Malawi, the duration of the recall period for which the access to credit was explored was October14

1993 to March 1995. For Bangladesh, the recall period was July 1993 to August 1995. 

Finally, as explained in the previous section, what is referred here as the credit limit

is actually the respondents’ perceived or expected credit limit. However, with this

clarification, for the sake of brevity, the reference to the term expected will be dropped in

the following discussion of the empirical evidence. 

EMPIRICAL DISTRIBUTION OF CREDIT LIMITS AND UNUSED CREDIT LINES

Table 1 presents the average informal and formal credit limits with the

corresponding average amounts of unused credit lines, i.e., the average of the differences

between the respective credit limits and amounts borrowed, in Bangladesh and Malawi.13

In the population as whole, the average informal and formal credit limits are, respectively,

$40 and $71 in Bangladesh and $7 and $11 in Malawi.  To put these figures in14

perspective, the 1995 per capita GNP were $240 and $170, respectively (World Tables

1997). These average figures are not good indicators of the levels of access to credit,

however, because of the truncated nature of the credit limit distributions. The box plot

diagrams of the distributions presented in Figure 1 give a better picture of the extent of

access to credit in both countries. The figure shows that in Bangladesh the median formal

and informal credit limits are, respectively, $50 and $13, and 50 percent of the population

can borrow at most $100 from the formal sector and $50 from the informal sector. The



32

ability to borrow is significantly more restricted in Malawi where the median formal and

informal credit limits are, respectively, zero and $3. Moreover, 50 percent of the

Malawian population can borrow at most less than $10 from either sector of the credit

market.

In both countries those who are presently borrowing from the formal sector had

higher median formal credit limits ($75 in Bangladesh and $25 in Malawi). But their

median informal credit limit is higher only in Bangladesh ($25 in Bangladesh and zero in

Malawi). For Malawi, this is likely to be a reflection of the fact that two of the credit

programs studied are targeted to poor women who might have been excluded from the

few existing sources of informal credit because of their socioeconomic situation (see

Figure 3). For Bangladesh, the positive correlation suggests that informal lenders are

willing to lend significantly more to those who are participating in the formal credit market

than to the rest of the population. One explanation is that perhaps formal sector borrowers

are perceived as good credit risks because they are able to negotiate and manage much

bigger loans from formal sources. Furthermore, informal lenders may believe that, if

necessary, they may divert formal loans to repay their informal ones. However, the

distributions of the unused credit lines shown in Figure 2 reveal interesting facts on the

behavior toward informal sources of credit of formal sector borrowers in Bangladesh.

Indeed, although they have higher informal credit limits compared to the rest of the

population, they use less of them. In fact, almost half of them have over $50 of unused

informal credit lines and virtually all of them deplete their formal credit lines. This perhaps



33

indicates that higher informal credit limits are coming from professional moneylenders and

are too expensive as source of loans compared to formal sources. Another explanation is

that despite having higher informal credit limits compared to the rest of the population,

formal sector borrowers prefer the formal sector because the types of investment for

which formal loans are sought are too big to be financed through informal sources. As

Figure 2 shows, in both countries the median unused formal credit lines are zero for the

population as a whole and for formal sector borrowers. 

Finally, with respect to gender differences in access to credit, Figure 3 shows that,

both in Bangladesh and Malawi in the general population, women have higher median

formal credit limits but lower median informal credit limits compared to men. However,

when the comparison is restricted to formal sector borrowers only, men have higher

median credit limits in both sectors of the credit markets. Moreover, Figure 4 shows that

in both countries female formal sector borrowers are more likely to have a binding credit

constraint in both sectors than their male counterparts. This provides a justification for

targeting credit to women since they are more likely to have a binding credit constrained

than men and have good reputations in repaying back their loans.

ACCESS TO CREDIT, CREDIT CONSTRAINTS, AND PARTICIPATION IN
CREDIT PROGRAMS

To get a picture of how participation in a credit program is related to access to

credit at the household level, participating and nonparticipating households have been



34

classified according to the types of access to credit and binding of the credit constraints of

their individual members. A household is said to have access to a type of credit if at least

one of its members has a strictly positive credit limit for that type of credit. Similarly, a

household is classified as credit constrained for a type of credit if at least one of its

members is constrained for that type of credit. Tables 2 and 3 present the outcome of this

classification according to membership status in a credit program. First, consistent with

the conceptual distinction between access to credit and participation in a credit program,

Table 2 shows that 16 percent and 8 percent of households that never participated in any

formal credit program in Bangladesh and Malawi, respectively, did have access to formal

credit during the recall period, i.e., they could obtain a formal loan if they had wanted.

However, more than two-thirds of the households in Bangladesh do not have access to

any type of credit. The corresponding figure for Malawi is 28 percent.

Second, further illustrating the importance of the distinction between access to

credit and participation in a credit program, Table 2 shows that up to 31 percent and 40

percent of households currently participating in formal credit programs in Bangladesh and

Malawi, respectively, did not have access to formal credit during the recall period. This

means that not only did they not receive any formal loan during that period, but also they

could not borrow anything from their own program. Hence, even for participants of credit

programs, their continuous access to formal credit cannot be taken for granted and should

be considered (together with access to informal credit) as subject to random events.

Moreover, from the participating household’s perspective, these fluctuations in its access



35

to formal credit brings some type of uncertainty in its planning and decisionmaking

process that can be more difficult to cope with than the uncertainty faced when not

participating in a credit program. Indeed, unpredictable institutional changes or abrupt

changes in the policy of a credit program can create more damages in the operations of a

profitable business than nature-induced events that, while still uncertain, are often

predictable.

Finally, Table 3 gives information on whether households’ credit constraints were

binding or not. It shows a clear positive impact that participation in credit programs can

have in relaxing households’ credit constraints. Of current members of credit programs, 65

percent and 39 percent, respectively, did not experience any binding informal or formal

credit constraint in Bangladesh and Malawi. In contrast, among those households that had

never been members of any credit program, only 16 percent and 8 percent, respectively,

did not experience any binding credit constraints. Taking Bangladesh as an example, of

those households that had never been members, 69 percent experienced at least once a

credit constraint in the informal and the formal credit market during the recall period of

almost two years. Fifteen percent had binding credit constraints in the formal sector only,

but not in the informal sector. These results point out to the changes in credit access and

occurrence of credit constraints that can be brought about by expanding microfinance

institutions that lend to the rural poor. 

5. CONCLUSION



36

The quantification of the impact of improved access to formal credit on different

groups of households is important for policy purposes, not only because it can serve as a

guide for the allocation of scarce resources to different and numerous development

programs competing for the same funds, but also because it allows one to determine the

relative importance of the various factors that make some households in some

socioeconomic environments benefit more from access to formal credit than others. This

paper has presented a conceptual framework and data collection methodology for

measuring the extent of household access to credit and analyzing the socioeconomic

determinants and impacts of the extent of access to credit.

The paper reviewed first the two main methodologies that have been used up to

now to measure household access to credit and credit constraints and exposed their

shortcomings. The first method tries to detect credit constrained households through tests

of violation of the life-cycle/permanent income hypothesis. Empirical evidence from this

methodology regarding presence or absence of credit constraint has been inconclusive,

mostly because violation of the implications of life cycle/permanent income hypothesis is

neither a sufficient nor necessary condition for being credit constrained. The second

method of measuring access and detecting credit constraint collects household-level credit

market information directly from household surveys to determine whether households are

credit constrained. This classification is then used in reduced-form regression equations to

analyze the determinants of the likelihood of a household being credit constrained and the



37

effects of this likelihood on various household outcomes. Despite representing a

substantial improvement in comparison to the first method, it is still incapable of providing

the framework that allows one to quantify the extent to which households are credit

constrained or to satisfactorily assess the impact of access to credit on household welfare

outcomes. 

The methodology presented in this paper corrects the shortcomings of the direct

method by developing a conceptual framework and data collection methodology that

focus on the concept of credit limit. This focus is justified by the fact that every potential

borrower faces a credit limit because of asymmetries of information between borrowers

and lenders and the imperfect enforcement of loan contracts. Therefore, a household’s

credit limit from any given source of credit is the best measure of its degree of access to

that source of credit.

Furthermore, the changes in household behavioral and welfare outcomes in response

to changes in its credit limit represent the effects of access to credit (or improvement in

access) on those household outcomes. Hence, once data are collected on households’

credit limits, econometric analysis allows one to quantify the determinants of the extent of

households’ access to credit as well as the effects such access has on household welfare

outcomes. This data has been collected for the first time in IFPRI surveys conducted in

Bangladesh and Malawi. The descriptive summary statistics from the two surveys

presented in this paper illustrate the usefulness of using the credit limit to assess the extent

of households’ access to credit and the binding of their credit constraints.



38



Mf(bmax, .)/Mbmax

bmax bmax

Mf(bmax, .)/Mbmax

bmax

ME (b , . |b =x Mx

max M (b , . /M max

b /f bmax .)

max

b

39

 This is a fu Fréchet5

derivative  In fact, in general  is a continuous linear function and not a real number, and th
int
show here, these technical issues need not be discussed.

APPENDIX

 WHEN ONLY  IS OBSERVED

 can be estimated

when 

value is. First, we recall that the usual partial effect estimated in standard regression

analysis is the quantity 

of the dependent variable following small changes in the possible values of the random

variable , which

involves taking a partial derivative with respect to a random variable (a function!). Hence,

Even the interpretation of “marginal change” is not trivial in this context.  However,

because of the underlying metrics, for all practical purposes the two quantities can be

interpreted in the same ways without worrying about all the subtleties involved.

 and

other household choice variables depend only on  and not on higher

moments of 

with the mathematical expectation operator E

generality, we assume a linear dependence and abstract from the other independent



f(bmax, .) = âEbmax

Mf(bmax, .)/Mbmax= âE

Mf(bmax, .)/Mbmax(h) = âEh bmax dbmax

Mf(bmax, .)/Mbmax(dbmax) = âEdbmax Mf(bmax, .)/Mbmax

b * bmax

40

variables, i.e., . Thus, â is the usual partial effect and the coefficient to

be estimated in a standard regression. Because the expectation operator is a linear

function, one has , and for all random variables h,

. In particular, for a “marginal change” in  of size , we

have . In fact,  can be identified with â

and interpreted as the expected change in  following an expected change in . This

proves the claim made in the paper.



41

TABLES





43

Table 1—Average maximum credit limit and unused credit line in the informal and formal market, by country and by
gender of respondent

Credit limit (in US dollar) Male Female All Male Female All Male Female All Male Female All

Bangladesh Malawi

All respondents Formal sector borrowers only All respondents Formal sector borrowers only

Informal credit limit 53 11 40 130 22 69 7 6 7 9 5 6

of which unused 29  7 22 130 22 69 3 3 3 7 3 5

Formal credit limit 71 71 71 159 90 117 10 12 11 60 38 45

of which unused 58 43 52 52 13 28 4 3 3 16 7 10

Source:  Bangladesh: IFPRI, Rural Finance and Food Security Survey; Malawi: IFPRI/Bunda, Rural Finance survey.
Note:  1995 per capita GNP for Bangladesh and Malawi are, respectively, $240 and $170 (World Tables 1997).



44

Table 2—Percentage of households having access to informal and formal credit: by
membership in group-based credit program and by country

Access to credit during recall period s s s s s s a

Bangladesh Malawi

Current Past been Current Past been
member member member member member member

Never Never

No access 24% 64% 69% 10% 17% 28%

Informal only  7%  9% 15% 30% 50% 64%

Formal only  3%  4%  0% 16% 0%  0%

Informal and formal 66% 23% 16% 44% 32%  8%

Source: Bangladesh: IFPRI, Rural Finance and Food Security Survey, 1994; Malawi: IFPRI/Bunda, Rural Finance
survey, 1995.

In Malawi, the duration of the recall period for which the access to credit was explored was October 1993 to Marcha

1995. For Bangladesh, the recall period was July 1993 to August 1995. 

Table 3—Percentage of households experiencing a binding credit constraint in the
informal and formal market, by membership in group-based credit
program and by country

Credit constraint during recall period s s s s s sa

Bangladesh  Malawi

Current Past been Current Past been
member member member member member member

Never Never

Informal and formal 24% 64% 69% 15% 21% 33%

Informal only  3%  4%  0% 12%  0%  0%

Formal only  8%  9% 15% 34% 50% 60%

No binding constraint 65% 23% 16% 39% 29%  8%

Source: Bangladesh: IFPRI, Rural Finance and Food Security Survey, 1994; Malawi: IFPRI/Bunda, Rural Finance
survey, 1995.

In Malawi, the duration of the recall period for which the access to credit was explored was October 1993 to Marcha

1995. For Bangladesh, the recall period was July 1993 to August 1995. 



45

FIGURES





4711382N =

MalawiBangladesh

500

450

400

350

300

250

200

150

100

50

0

-50
4711382N =

MalawiBangladesh

200

150

100

50

0

-50

90220N =

MalawiBangladesh

200

150

100

50

0

-50
90220N =

MalawiBangladesh

500

450

400

350

300

250

200

150

100

50

0

-50

47

Figure 1—Distribution of formal and informal credit limits in Bangladesh and
Malawi: Box plot diagrams

All respondents

Formal sector borrowers only**

Credit limits (US$).
All respondents Formal Informal Formal Informal

Bangladesh Malawi

Mean  71 40 11 7
Median 50 13 0 3
Standard deviation 118 143 33 24
Minimum 0 0 0 0
Maximum 1250 2500 667 800

Source: Bangladesh: IFPRI, Rural Finance and Food Security Survey, 1994; Malawi: IFPRI/Bunda, Rural Finance survey, 1995.
Notes: * Box plot diagrams are interpreted as follow: For each box, 50 percent of cases have values within the box and the solid

horizontal line inside is the median. The length of the box is the interquartile range and the lower boundary (resp upper boundary)
of the box is the 25th (resp 75th) percentile. Finally, the circles are outliers and the stars are extreme values.
** For Bangladesh, the informal credit limit figures do not include the ones for formal sector borrowers who had no informal loan
transaction. 



48

Per capita 1995 GNP for Bangladesh and Malawi are, respectively, $240 and $170 (World Tables 1997).



4711382N =

MalawiBangladesh

500

450

400

350

300

250

200

150

100

50

0

-50
4711382N =

MalawiBangladesh

200

150

100

50

0

-50

90220N =

MalawiBangladesh

500

450

400

350

300

250

200

150

100

50

0

-50
90220N =

MalawiBangladesh

200

150

100

50

0

-50

49

Figure 2—Distribution of unused formal and informal credit lines in Bangladesh
and Malawi: Box plot diagrams

All respondents

Formal sector borrowers only**

Unused credit lines (US$).
 All respondents Formal Informal Formal Informal

Bangladesh Malawi

Mean  52 22 3 3
Median 0 5 0 1
Standard deviation 111 130 16 9
Minimum 0 0 0 0
Maximum 1,163 2,500 438 267

Source: Bangladesh: IFPRI, Rural Finance and Food Security Survey, 1994; Malawi: IFPRI/Bunda, Rural Finance survey, 1995.
Notes: * Box plot diagrams are interpreted as follow: For each box, 50 percent of cases have values within the box and the solid

horizontal line inside is the median. The length of the box is the interquartile range and the lower boundary (resp upper boundary)
of the box is the 25th (resp 75th) percentile. Finally, the circles are outliers and the stars are extreme values.
** For Bangladesh, the informal credit limit figures do not include the ones for formal sector borrowers who had no informal loan
transaction.
Per capita 1995 GNP for Bangladesh and Malawi are, respectively, $240 and $170 (World Tables 1997).



50



214476 257906N =

MalawiBangladesh

500

450

400

350

300

250

200

150

100

50

0

-50

MALE

FEMALE
214476 257906N =

MalawiBangladesh

200

150

100

50

0

-50

MALE

FEMALE

59123 3097N =

MalawiBangladesh

200

150

100

50

0

-50

MALE

FEMALE
59123 3097N =

MalawiBangladesh

500

450

400

350

300

250

200

150

100

50

0

-50

MALE

FEMALE

51

Figure 3—Distribution of formal and informal credit limits in Bangladesh and
Malawi: Box plot diagrams by gender of respondents

All respondents

Formal sector borrowers only**

Bangladesh Malawi
Credit limits (US$) Formal Informal Formal Informal
All respondents Male Female Male Female Male Female Male Female

Mean  71 71 53 11 10 12 7 6
Median 0 75 13 5 0 0 3 2
Standard deviation 139 59 174 22 36 29 15 31
Minimum 0 0 0 0 0 0 0 0
Maximum 1,250 300 2,500 150 667 333 500 800

Source: Bangladesh: IFPRI, Rural Finance and Food Security Survey, 1994; Malawi: IFPRI/Bunda, Rural Finance survey, 1995.
Notes: * Box plot diagrams are interpreted as follow: For each box, 50 percent of cases have values within the box and the solid

horizontal line inside is the median. The length of the box is the interquartile range and the lower boundary (resp upper boundary)
of the box is the 25th (resp 75th) percentile. Finally, the circles are outliers and the stars are extreme values.
** For Bangladesh, the informal credit limit figures do not include the ones for formal sector borrowers who had no informal loan
transaction.
Per capita 1995 GNP for Bangladesh and Malawi are, respectively, $240 and $170 (World Tables 1997).



214476 257906N =

MalawiBangladesh

500

450

400

350

300

250

200

150

100

50

0

-50

MALE

FEMALE
214476 257906N =

Malawi
Bangladesh

200

150

100

50

0

-50

MALE

FEMALE

59123 3097N =

MalawiBangladesh

200

150

100

50

0

-50

MALE

FEMALE
59123 3097N =

MalawiBangladesh

200

150

100

50

0

-50

MALE

FEMALE

52

Figure 4—Distribution of unused formal and informal credit lines in Bangladesh
and Malawi: Box plot diagrams by gender of respondents

All respondents

Formal sector borrowers only**

Bangladesh Malawi
Unused credit lines (US$). Formal Informal Formal Informal
All respondents Male Female Male Female Male Female Male Femal

Mean  58 43 29 7 4 3 3 3
Median 0 25 6 4 0 0 1 0
Standard deviation 130 60 159 21 17 16 6 12
Minimum 0 0 0 0 0 0 0 0
Maximum 1,163 250 2,500 125 438 300 67 267

Source: Bangladesh: IFPRI, Rural Finance and Food Security Survey, 1994; Malawi: IFPRI/Bunda, Rural Finance survey, 1995.
Notes: * Box plot diagrams are interpreted as follow: For each box, 50 percent of cases have values within the box

and the solid horizontal line inside is the median. The length of the box is the interquartile range and the
lower boundary (resp upper boundary) of the box is the 25th (resp 75th) percentile. Finally, the circles are
outliers and the stars are extreme values.
** For Bangladesh, the informal credit limit figures do not include the ones for formal sector borrowers who
had no informal loan transaction.



53

Per capita 1995 GNP for Bangladesh and Malawi are, respectively, $240 and $170 (World Tables 1997).



54

REFERENCES

Adams, D. W., and R. C. Vogel. 1985. Rural financial markets in low-income countries:

Recent controversies and lessons. World Development 14 (4): 477–87.

Adams, D. W., D. H. Graham, and J. D. von Pischke, eds. 1984. Undermining rural

development with cheap credit. Boulder: Westview Press.

Alderman, A., and C. Paxson. 1992. Do the poor insure?: A synthesis of the literature on

risk sharing institutions in developing countries. Discussion Paper 164.  Princeton,

N.J., U.S.A.:  Princeton University, Woodrow Wilson School of Public and

International Affairs.

Avery, R. B. 1981. Estimating credit constraints by switching regression. In Structural

analysis of discrete data with econometric applications, ed. C. F. Manski and D.

McFadden. Cambridge, Mass., U.S.A.: MIT Press.

Barham, B., and S. Boucher. 1994. Credit constraints, credit unions, and small-scale

producers in Guatemala. Mimeo.

Besley, T. 1995. Savings, credit and insurance. In Handbook of development economics,

Vol. III, ed. J. Behrman and T. N. Srinivasan. Volume III. Amsterdam: North

Holland.

Braverman, A., and J. L. Guasch. 1986. Rural credit markets and institutions in

developing countries: Lessons for policy analysis from practice and modern theory.

World Development 14 (10/11): 1253–1267.



55

Browning, M., and A. Lusardi. 1996. Household saving: Micro theories and micro facts.

Journal of Economic Literature 4: 1797–1855.

Carroll, C. D. 1991. Buffer stock saving and the permanent income hypothesis.

Washington, D.C.: Board of Governors of the Federal Reserve System.

Coate, S., and M. Ravallion. 1993. Reciprocity without commitment: Characterization and

performance of informal insurance arrangements. Journal of Development

Economics 40: 1–24.

Deaton, A. 1989. Saving in developing countries: Theory and review. In Proceedings of

the World Bank Annual Conference on Development Economics, ed. S. Fischer and

D. de Tray, 61-96.  Washington, D.C.:  World Bank.

Deaton, A. 1990. Impacts of credit programs on food security and nutrition in Ghana.

International Food Policy Research Institute, Washington, D.C.  Unpublished.

Deaton, A. 1991. Saving and liquidity constraints. Econometrica 59: 1221–1248. 

Deaton, A. 1992. Understanding consumption. Oxford: Oxford University Press. 

Delgado, C. 1995. Africa’s changing agricultural development strategies: Past and

present paradigms as a guide to the future. 2020 Vision Food, Agriculture, and the

Environment Discussion Paper 3. Washington, D.C.: International Food Policy

Research Institute.



56

Diagne, A. 1994. Consumer behavior and welfare measurement under uncertainty: theory

and empirical evidence from Senegal. Unpublished Ph.D. dissertation. Michigan

State University, Department of Agricultural Economics and Department of

Economics.

Diagne, A. 1996. Measuring access to credit and its impacts on household food security:

Some methodological notes. Paper presented at 1996 annual meeting of The

American Agricultural Economic Association (AAEA), San Antonio, Texas.

Diagne, A. 1997. Impacts of access to credit on poverty and food insecurity in Malawi.

Paper presented at the XXIII International Conference of Agricultural Economists,

August, 10–16, Sacramento, Calif., U.S.A.

Diagne, A. 1999. Determinants of household access to and participation in formal and

informal credit markets in Malawi. Food Consumption and Nutrition Division

Discussion Paper 67. International Food Policy Research Institute, Washington,

D.C.

Diagne, A., M. Zeller, and C. Mataya. 1996. Rural financial markets and household food

security: Impacts of access to credit on the socioeconomic situation of rural

households in Malawi. Final report submitted to the Ministry for Women, Children

Affairs, Community Development and Social Welfare, and to the German Agency

for Technical Cooperation in Malawi, Bunda College of Agriculture, University of

Malawi, Lilongwe.  International Food Policy Research Institute, Washington, D.C.



57

Eswaran, M., and A. Kotwal. 1990. Implications of credit constraints for risk behaviour in

less developed economies. Oxford Economic Papers N.-S. 42 (2): 473–482.

Fafchamps, M. 1992. Solidarity networks in pre-industrial societies: Rational peasants

with a moral economy. Economic Development and Cultural Change 41 (1):

147–174.

Feder G., L. J. Lau, J. Y. Lin, and X. Luo. 1990. The relationship between credit and

productivity in Chinese agriculture: A microeconomic model of disequilibrium.

American Journal of Agricultural Economics 72 (5): 1151–1157.

Freixas, X., and J. C. Rochet. 1997. Microeconomics of banking. Cambridge, Mass.,

U.S.A.: Massachusetts Institute of Technology Press.

Gersovitz, M. 1988. Savings and development. In Handbook of development economics,

Vol. I, ed. H. Chenery and T. N. Srinivasan. Amsterdam: North Holland.

Hall, R. 1978. Stochastic implications of the life cycle permanent income hypothesis:

Theory and evidence. Journal of Political Economy 86 (5): 971–987.

Jappelli, T. 1990. Who is credit constrained in the U.S. economy? Quarterly Journal of

Economics CV (1) (February): 219–234.

Kimball, M. S. 1990. Precautionary saving in the small and in the large. Econometrica 58

(1): 53–73.

Lund, S. M. 1995. Informal credit, insurance networks, and risk-sharing: Empirical

evidence from the Philippines. Food Research Institute, Stanford University,

Stanford, Calif., U.S.A. Unpublished.



58

Lund, S. M., and M. Fafchamps. 1997. Risk-sharing networks in rural Philippines. 

Department of Economics, Stanford University, Stanford, Calif., U.S.A.

Unpublished.

Morduch, J. 1990. Risk, production and savings: Theory and evidence from Indian

households. Cambridge, Mass., U.S.A.: Harvard University.  Unpublished.

Paxson, C. 1992. Using weather variability to estimate the response of savings to

transitory income in Thailand. American Economic Review 82 (1): 15–33.

Pitt, M., and S. R. Khandker. 1994. Household and intrahousehold impacts of the

Grameen Bank and similar targeted credit programs in Bangladesh. World Bank

Discussion Paper 320.  Washington, D.C.

Schrieder, G., and H. Heidhues. 1995. Reaching the poor through financial innovations.

Quarterly Journal of International Agriculture 34 (2): 132–148.

Stiglitz, J. E., and A. Weiss. 1981. Credit rationing in markets with imperfect information.

American Economic Review 71 (3): 393–410.

Udry, C. 1995. Risk and saving in northern Nigeria. American Economic Review 85 (5):

1287–1300.

Zeldes, S. 1989a. Consumption and liquidity constraints: An empirical investigation.

Journal of Political Economy 97 (2): 305–346.

Zeldes, S. 1989b. Optimal consumption with stochastic income: Deviations from certainty

equivalence. Quarterly Journal of Economics 104 (2): 305–346.



59

Zeller, M. 1994. Determinants of credit rationing: A study of informal lenders and formal

groups in Madagascar. World Development 22 (12): 1895–1907.

Zeller, M., M. Sharma, and A.U. Ahmed. 1996. Credit for the rural poor: Country case

Bangladesh. Final report submitted to Deutsche Gesellschaft für Technische

Zusammenarbeit (GTZ). International Food Policy Research Institute, Washington,

D.C.

Zeller, M., B. Sumiter, and M. Sharma. 1996. Financial services for the rural poor:

Implications for policy and future research. Final report to the German Agency for

Technical Cooperation (GTZ). International Food Policy Research Institute,

Washington, D.C.

Zeller, M., G. Schrieder, J. von Braun, and F. Heidhues. 1997. Rural finance for food

security for the poor: Implications for research and policy. Food Policy Review

No. 4. Washington, D.C.: International Food Policy Research Institute.

Zeller, M., A. Ahmed, S. Babu, S. Broca, A. Diagne, and M. Sharma. 1996b. Rural

financial policies for food security of the poor: Methodologies for a multicountry

research project. Food Consumption and Nutrition Division Discussion Paper No.

11. International Food Policy Research Institute, Washington, D.C.



FCND DISCUSSION PAPERS

01 Agricultural Technology and Food Policy to Combat Iron Deficiency in Developing Countries, Howarth E.
Bouis, August 1994

02 Determinants of Credit Rationing:  A Study of Informal Lenders and Formal Credit Groups in
Madagascar, Manfred Zeller, October 1994

03 The Extended Family and Intrahousehold Allocation: Inheritance and Investments in Children in the Rural
Philippines, Agnes R. Quisumbing, March 1995

04 Market Development and Food Demand in Rural China, Jikun Huang and Scott Rozelle, June 1995

05 Gender Differences in Agricultural Productivity: A Survey of Empirical Evidence, Agnes R. Quisumbing,
July 1995

06 Gender Differentials in Farm Productivity:  Implications for Household Efficiency and Agricultural Policy,
Harold Alderman, John Hoddinott, Lawrence Haddad, and Christopher Udry, August 1995

07 A Food Demand System Based on Demand for Characteristics: If There Is "Curvature" in the Slutsky
Matrix, What Do the Curves Look Like and Why?, Howarth E. Bouis, December 1995

08 Measuring Food Insecurity: The Frequency and Severity of "Coping Strategies," Daniel G. Maxwell,
December 1995

09 Gender and Poverty: New Evidence from 10 Developing Countries, Agnes R. Quisumbing, Lawrence
Haddad, and Christine Peña, December 1995

10 Women's Economic Advancement Through Agricultural Change: A Review of Donor Experience, Christine
Peña, Patrick Webb, and Lawrence Haddad, February 1996

11 Rural Financial Policies for Food Security of the Poor: Methodologies for a Multicountry Research
Project, Manfred Zeller, Akhter Ahmed, Suresh Babu, Sumiter Broca, Aliou Diagne, and Manohar
Sharma, April 1996

12 Child Development: Vulnerability and Resilience, Patrice L. Engle, Sarah Castle, and Purnima Menon,
April 1996

13 Determinants of Repayment Performance in Credit Groups: The Role of Program Design, Intra-Group
Risk Pooling, and Social Cohesion in Madagascar, Manfred Zeller, May 1996

14 Demand for High-Value Secondary Crops in Developing Countries: The Case of Potatoes in Bangladesh
and Pakistan, Howarth E. Bouis and Gregory Scott, May 1996

15 Repayment Performance in Group-Based Credit Programs in Bangladesh: An Empirical Analysis,
Manohar Sharma and Manfred Zeller, July 1996

16 How Can Safety Nets Do More with Less?  General Issues with Some Evidence from Southern Africa,
Lawrence Haddad and Manfred Zeller, July 1996

17 Remittances, Income Distribution, and Rural Asset Accumulation, Richard H. Adams, Jr., August 1996

18 Care and Nutrition: Concepts and Measurement, Patrice L. Engle, Purnima Menon, and Lawrence Haddad,
August 1996



FCND DISCUSSION PAPERS

19 Food Security and Nutrition Implications of Intrahousehold Bias:  A Review of Literature, Lawrence
Haddad, Christine Peña, Chizuru Nishida, Agnes Quisumbing, and Alison Slack, September 1996

20 Macroeconomic Crises and Poverty Monitoring:  A Case Study for India, Gaurav Datt and Martin
Ravallion, November 1996

21 Livestock Income, Male/Female Animals, and Inequality in Rural Pakistan, Richard H. Adams, Jr.,
November 1996

22 Alternative Approaches to Locating the Food Insecure:  Qualitative and Quantitative Evidence from South
India, Kimberly Chung, Lawrence Haddad, Jayashree Ramakrishna, and Frank Riely, January 1997

23 Better Rich, or Better There?  Grandparent Wealth, Coresidence, and Intrahousehold Allocation, Agnes R.
Quisumbing, January 1997

24 Child Care Practices Associated with Positive and Negative Nutritional Outcomes for Children in
Bangladesh:  A Descriptive Analysis, Shubh K. Kumar Range, Ruchira Naved, and Saroj Bhattarai,
February 1997

25 Water, Health, and Income:  A Review, John Hoddinott, February 1997

26 Why Have Some Indian States Performed Better Than Others at Reducing Rural Poverty?, Gaurav Datt and
Martin Ravallion, March 1997

27 "Bargaining" and Gender Relations:  Within and Beyond the Household, Bina Agarwal, March 1997

28 Developing a Research and Action Agenda for Examining Urbanization and Caregiving:  Examples from
Southern and Eastern Africa, Patrice L. Engle, Purnima Menon, James L. Garrett, and Alison Slack, April
1997

29 Gender, Property Rights, and Natural Resources, Ruth Meinzen-Dick, Lynn R. Brown, Hilary Sims
Feldstein, and Agnes R. Quisumbing, May 1997

30 Plant Breeding: A Long-Term Strategy for the Control of Zinc Deficiency in Vulnerable Populations,
Marie T. Ruel and Howarth E. Bouis, July 1997

31 Is There an Intrahousehold 'Flypaper Effect'?  Evidence from a School Feeding Program, Hanan Jacoby,
August 1997

32 The Determinants of Demand for Micronutrients: An Analysis of Rural Households in Bangladesh,
Howarth E. Bouis and Mary Jane G. Novenario-Reese, August 1997

33 Human Milk—An Invisible Food Resource, Anne Hatløy and Arne Oshaug, August 1997

34 The Impact of Changes in Common Property Resource Management on Intrahousehold Allocation, Philip
Maggs and John Hoddinott, September 1997

35 Market Access by Smallholder Farmers in Malawi: Implications for Technology Adoption, Agricultural
Productivity, and Crop Income, Manfred Zeller, Aliou Diagne, and Charles Mataya, September 1997

36 The GAPVU Cash Transfer Program in Mozambique:  An assessment, Gaurav Datt, Ellen Payongayong,
James L. Garrett, and Marie Ruel, October 1997



FCND DISCUSSION PAPERS

37 Why Do Migrants Remit?  An Analysis for the Dominican Sierra, Bénédicte de la Brière, Alain de Janvry,
Sylvie Lambert, and Elisabeth Sadoulet, October 1997

38 Systematic Client Consultation in Development: The Case of Food Policy Research in Ghana, India,
Kenya, and Mali, Suresh Chandra Babu, Lynn R. Brown, and Bonnie McClafferty, November 1997

39 Whose Education Matters in the Determination of Household Income: Evidence from a Developing
Country, Dean Jolliffe, November 1997

40 Can Qualitative and Quantitative Methods Serve Complementary Purposes for Policy Research?  Evidence
from Accra, Dan Maxwell, January 1998

41 The Political Economy of Urban Food Security in Sub-Saharan Africa, Dan Maxwell, February 1998

42 Farm Productivity and Rural Poverty in India, Gaurav Datt and Martin Ravallion, March 1998

43 How Reliable Are Group Informant Ratings? A Test of Food Security Rating in Honduras, Gilles Bergeron,
Saul Sutkover Morris, and Juan Manuel Medina Banegas, April 1998

44 Can FAO's Measure of Chronic Undernourishment Be Strengthened?, Lisa C. Smith, with a Response by
Logan Naiken, May 1998

45 Does Urban Agriculture Help Prevent Malnutrition?  Evidence from Kampala, Daniel Maxwell, Carol
Levin, and Joanne Csete, June 1998

46 Impact of Access to Credit on Income and Food Security in Malawi, Aliou Diagne, July 1998

47 Poverty in India and Indian States: An Update, Gaurav Datt, July 1998

48 Human Capital, Productivity, and Labor Allocation in Rural Pakistan, Marcel Fafchamps and Agnes R.
Quisumbing, July 1998

49 A Profile of Poverty in Egypt: 1997, Gaurav Datt, Dean Jolliffe, and Manohar Sharma, August 1998.

50 Computational Tools for Poverty Measurement and Analysis, Gaurav Datt, October 1998

51 Urban Challenges to Food and Nutrition Security: A Review of Food Security, Health, and Caregiving in
the Cities, Marie T. Ruel, James L. Garrett, Saul S. Morris, Daniel Maxwell, Arne Oshaug, Patrice Engle,
Purnima Menon, Alison Slack, and Lawrence Haddad, October 1998

52 Testing Nash Bargaining Household Models With Time-Series Data, John Hoddinott and Christopher
Adam, November 1998

53 Agricultural Wages and Food Prices in Egypt: A Governorate-Level Analysis for 1976-1993, Gaurav Datt
and Jennifer Olmsted, November 1998

54 Endogeneity of Schooling in the Wage Function: Evidence from the Rural Philippines, John Maluccio,
November 1998

55 Efficiency in Intrahousehold Resource Allocation, Marcel Fafchamps, December 1998



FCND DISCUSSION PAPERS

56 How Does the Human Rights Perspective Help to Shape the Food and Nutrition Policy Research Agenda?,
Lawrence Haddad and Arne Oshaug, February 1999

57 The Structure of Wages During the Economic Transition in Romania, Emmanuel Skoufias, February 1999

58 Women's Land Rights in the Transition to Individualized Ownership: Implications for the Management of
Tree Resources in Western Ghana, Agnes Quisumbing, Ellen Payongayong, J. B. Aidoo, and Keijiro
Otsuka, February 1999

59 Placement and Outreach of Group-Based Credit Organizations: The Cases of ASA, BRAC, and PROSHIKA
in Bangladesh, Manohar Sharma and Manfred Zeller, March 1999

60 Explaining Child Malnutrition in Developing Countries: A Cross-Country Analysis, Lisa C. Smith and
Lawrence Haddad, April 1999

61 Does Geographic Targeting of Nutrition Interventions Make Sense in Cities? Evidence from Abidjan and
Accra, Saul S. Morris, Carol Levin, Margaret Armar-Klemesu, Daniel Maxwell, and Marie T. Ruel, April
1999

62 Good Care Practices Can Mitigate the Negative Effects of Poverty and Low Maternal Schooling on
Children's Nutritional Status: Evidence from Accra, Marie T. Ruel, Carol E. Levin, Margaret Armar-
Klemesu, Daniel Maxwell, and Saul S. Morris, April 1999

63 Are Urban Poverty and Undernutrition Growing?  Some Newly Assembled Evidence, Lawrence Haddad,
Marie T. Ruel, and James L. Garrett, April 1999

64 Some Urban Facts of Life:  Implications for Research and Policy, Marie T. Ruel, Lawrence Haddad, and
James L. Garrett, April 1999

65 Are Determinants of Rural and Urban Food Security and Nutritional Status Different?  Some Insights from
Mozambique, James L. Garrett and Marie T. Ruel, April 1999

66 Working Women in an Urban Setting: Traders, Vendors, and Food Security in Accra, Carol E. Levin,
Daniel G. Maxwell, Margaret Armar-Klemesu, Marie T. Ruel, Saul S. Morris, and Clement Ahiadeke,
April 1999

67 Determinants of Household Access to and Participation in Formal and Informal Credit Markets in Malawi,
Aliou Diagne, April 1999

68 Early Childhood Nutrition and Academic Achievement: A Longitudinal Analysis, Paul Glewwe, Hanan
Jacoby, and Elizabeth King, May 1999

69 Supply Response of West African Agricultural Households: Implications of Intrahousehold Preference
Heterogeneity, Lisa C. Smith and Jean-Paul Chavas, July 1999

70 Child Health Care Demand in a Developing Country: Unconditional Estimates from the Philippines, Kelly
Hallman, August 1999

71 Social Capital and Income Generation in South Africa, 1993-98, John Maluccio, Lawrence Haddad, and
Julian May, September 1999



FCND DISCUSSION PAPERS

72 Validity of Rapid Estimates of Household Wealth and Income for Health Surveys in Rural Africa, Saul S.
Morris, Calogero Carletto, John Hoddinott, and Luc J. M. Christiaensen, October 1999

73 Social Roles, Human Capital, and the Intrahousehold Division of Labor: Evidence from Pakistan, Marcel
Fafchamps and Agnes R. Quisumbing, October 1999

74 Can Cash Transfer Programs Work in Resource-Poor Countries? The Experience in Mozambique, Jan W.
Low, James L. Garrett, and Vitória Ginja, October 1999

75 Determinants of Poverty in Egypt, 1997, Gaurav Datt and Dean Jolliffe, October 1999

76 Raising Primary School Enrolment in Developing Countries: The Relative Importance of Supply and
Demand, Sudhanshu Handa, November 1999

77 Th