Are the vulnerable poor left out of micro-credit? Quamrul Ahsan - PowerPoint PPT Presentation

Are the vulnerable poor left out of micro-credit? Quamrul Ahsan Department of Economics, Economics, University of Bergen Bergen, Norway June 2009 1

1. Introduction The paper is divided into two: theory and empirical application The theoretical question: • Why may the vulnerable poor shun productive borrowing? OR • Do high interest rates deter the vulnerable poor from seeking productive loans? 2

The empirical question In the empirical adaptation of the theoretical question, we ask: • Do the vulnerable poor have a higher propensity for unproductive borrowing? Who are the vulnerable poor ? Those who are exposed to potential consumption fluctuation due to inadequate risk coping ability (Amin et al . 2003 JDE) 3

What determines vulnerability? - lack of assets, both physical and financial - lack of access to social capital, e.g., membership in a large and resourceful social network. In the context of Bangladesh the following groups are identified as among the most vulnerable (Evans 1999, World Development; Amin et al. 2003 (op cit)): - Female headed households - Households headed by a widow - Peasant households 4

2. The model • Motivation Hashemi (1997), Mosley and Hulme (1996): Poor often self-select themselves out of loan market due to the fear that they would not generate sufficient returns to be able to repay. As discussed in Hermes and Lensink (2007), there are politicians as well as members of general public who argue that poor cannot afford high interests, and as such, high interest rates charged by micro-credit institutions are a form of discrimination against the poor. 5

An intertemporal (two-period) model of productive borrowing All individuals maximize expected utility of consumption over two periods: present and future. Assume a rural poor with two alternatives: A1: A fixed wage employment contract. + = ˆ ˆ ˆ u ( c ) u ( c ) U Maximized utility over two periods : 1 2 ˆ c are the optimal consumption in periods 1 and 2 respectively. ˆ c and 2 1 6

A2: The agent borrows an amount L from a micro-lender that is repayable over two periods in equal instalments. Returns from the investment are risky. State 1 State 2 R+ ε R- ε Returns (gross) Probability ½ ½ R+ ε - ℓ = ρ + ε R- ε - ℓ = ρ - ε Net returns 7

If the investment fails to yield adequate returns, the borrower faces repayment problem. In such events, the borrower needs emergency assistance/financial help. Such help may come from friends and/or family. We denote the amount of external help that the agent has access to as B . Assume that the bad state has occurred in the first period: ρ − ε + + ρ − ε − + ρ + ε − 1 u ( B ) { u ( B ) u ( B )} 2 ρ − ε + where ( ) : Period 1 utility, and u B ρ − ε − + ρ + ε − 1 { ( ) ( )} : Period 2 expected utility u B u B . 2 8

Similarly, if the good state occurs in the first period, the agent will save. The agent maximizes expected utility over two periods. We denote * EU maximized expected utility by . * ˆ , the individual will choose to borrow. > U EU If 9

EU * EU * increases with B until * EU B * is reached. For B>B*, expected utility remains constant. B * is the optimal level of external repayment help. B * B 10

The decision to borrow (or not to borrow) * ∂ ( ) EU B > < * 0 for . B B ∂ B ~ ~ Suppose now that there exists a B = B , ( B < B* ) such that ~ * = ˆ ˆ is the optimal utility from the fixed wage EU ( B ) U , where U employment. ~ * < ⇒ ˆ ( ) EU B U Then for all agent with B < B , there will be no demand for microcredit. 11

High interest charges and demand for loans For a given B < B *, expected utility responds negatively to an increases in interest charges, represented by ℓ : * * ∂ ∂ ∂ ρ EU ( B ) EU ( B ) = < . 0 ∂ ∂ ρ ∂ l l + ( ) (-) 12

~ , such Similarly as above, assume that there exist interest payments ℓ = l ~ ~ ~ , there will be no * = ˆ l . Then for all interest charges ℓ > l ( , ) EU B U that demand for credit. 13

Discussions • It follows from the definition of vulnerability that more vulnerable a person is, more limited would be her/his access to external funds ( B ). Hence, the vulnerable will be less likely to seek credit. • Another implication of the results above is that for vulnerable borrowers, the lack of access to sufficient “external funds” will seriously limit the choice of investment opportunities available to them. 14

• Further, those poor who reject microloans for productive investment following a rational decision making process could yet seek credit, albeit under duress, such as to meet emergency needs . • Data from rural India show that a majority of informal loan transactions between rural moneylenders and their poor clients are of this nature, (see references in Hulme and Mosley (1998)). 15

Empirical implementation The main hypothesis that we wish to test is based on the above implication of our model. The claim of the hypothesis is that: • The vulnerable poor have a lower propensity for productive borrowing compared to their less vulnerable counterpart. 16

Estimation strategy We construct a binary variable capturing if loans are used productively or not (– based on self-reported information -), given that one is a loan recipient. The test is based on the Bivariate Probit model that estimates the probability of productive use (as a function of socio-economic and demographic attributes of the loan recipients) jointly with a model of selection. 17

2. Data Household survey conducted in 1994-95 in two groups of villages (each group belonging to a distinct region) in Bangladesh, by the Institute of Development Studies (IDS), University of Sussex. There are four villages in each group. The survey covered 5062 households of which 2495 were from Chandina. Note that the two regions represented here are geographically distinct (about 250 kilometres apart) and have markedly different socio-economic characteristics. 18

We exclude from the sample two villages from Chandina that do not have any functioning credit program. This leaves us with a sample of 3422 households from six villages. 19

Table 1 Average income and landholding of borrowers and non-borrowers Productive Unproductive All Non-borrowers borrowing/borrowers borrowings/borrowers borrowings/borrowers Number of 587 381 968 individual borrowers Average 5224 4312 4864 outstanding borrowing (in Taka) in 1994 Average household 29 20 26 81 landholding (per cent of an acre) Average per capita 6272 4712 5658 7119 household income 20

Table 2 Distribution of borrowers by occupation and gender of household head Productive Unproductive All Total borrowing/borro borrowings/borro borrowings/borro number of wers (number) wers (number) wers (number) households Peasant 74 89 163 498 households Male-headed 568 364 933 3223 households Female- 19 17 36 199 headed households 21

Estimation Question 1: Do the vulnerable poor get credit? A model of selection: Probit regression Dependant variable : Probability of an individual household member being a credit recipient. (Table 6) (List of variables, Table 5) 22

The probability that an individual household member is a micro-credit participant is inversely related to (among other factors) - the size of household landholding , - the size of kinship/social network to which the household belongs the presence of members with higher (post-secondary) education in - the household. 23

In other words, the factors that inversely affect participation probability are also the ones likely to provide insurance against idiosyncratic (income) risks. Female-headed households and peasant households The probability of selection however declines if an individual belongs to a female-headed household , however being a member of a peasant household has no effect on selection. The above indeed supports the hypothesis that the vulnerable of the poor are successfully targeted by microcredit institutions. 24

Do vulnerable poor have a lower propensity for Question 2: productive borrowing? Dependent variable : A binary variable capturing if loans are used productively or not (based on self-reported information), given that one is a loan recipient. The test is based on a bivariate probit model that estimates the probability of productive use (as a function of socio-economic and demographic attributes of the loan recipients) jointly with a model of selection. 25

The bivariate Probit model yields the following result (Table 7): The probability of productive utilization of loans is positively related to - household ownership of land, - higher educational attainment of the household head, - household is long established in the village. - size of kinship network The probability however declines if the borrower belongs to a peasant household. 26

These results corroborate the conjecture that the poor and the vulnerable have a higher propensity to make unproductive use of loans relative to the less vulnerable. Note however that the term “unproductive use” does not imply that vulnerable are also poor investors or that they are inept at finding productive projects. It simply implies that the loans are not intended for income generating activities. 27