Description of Borrowers and Nonborrowers: Issue of Self-Selection Bias

One of the most difficult issues a researcher has to address in an impact study is to sort out whether wealth is created due to the program participation or participants were already relatively wealthy when they joined the program, known as selection bias. When people decide to join the program, they first self-select into it and are selected by their peers in the group-lending scheme. This double selection can create a bias in several instances. First, the wealthy or people with more entrepreneurship abilities are likelier to self-select into the program. Second, even if the poor want to join the program, they may not be selected into it by the wealthier peers.

To prevent the wealthy from joining, microfinance programs usually have some built-in mechanisms such as small loan amounts, risk of stigmatization by being in a club for the poor, and, in some cases, explicit participation criteria. In this respect, KB programs do not sufficiently discourage self-selection bias. It is pointed out in the previous section that although one household is limited to one loan of 10,000 rupees, the wealthy households can effectively circumvent the loan size restrictions by including several household members into microfinance programs. Since the interest rate of the KB lending program is relatively low, it attracts non-poor households to join the program. Although KB has criteria that borrowers must have income less than the minimum taxable limit, it is not clear if this stipulation is strictly enforced.

We turn to a simple t-test to assess if self-selection bias is pronounced among KB borrowers. Table 9 shows the results of testing the mean difference between borrowers and nonborrowers on a host of variables. To assess if borrowers and nonborrowers have similar initial wealth status, we paid particular attention to ownership of large assets, which are not likely to change due to microfinance. The results show that KB borrowers own significantly more land, operate larger cultivation areas, and own more production and household assets. Although the significant difference in variable values (e.g., sales or profit from nonagriculture enterprise) could be attributed to participating in KB's lending program, it is questionable if the initial conditions of KB members are far better than nonmember households. Based on the findings presented in Table 9 [ PDF 43.4KB | 1 page ], it could be concluded that selectivity bias is pronounced among KB clients.

As KB borrowers in the sample appear to be initially wealthier than the control group, the nonborrowers, if we run a naïve regression and estimate impact of borrowing from the KB on different wealth indicators, we are very likely to overestimate the impact of microfinance.

In the case of the previous KB study, Montgomery (2005) drew causal linkages in part based on the assumption that the survey design would minimize the selection bias. In this study we use the PSM method to address the selectivity bias and reassess the impacts on key parameters of household well-being.

Download this Discussion Paper [ PDF 198.6KB| 31 pages ].

Post a Comment

We welcome your feedback on this publication. Post a comment. ADBI is not obliged to acknowledge or publish comments and may abridge or edit them before web posting.

Comment(s)

There are [0] comment(s) for this entry. Post a comment.