Can Microfinance Reach the Poorest: Evidence from a Community-Managed Microfinance Intervention

Highlights

• We investigate participation in VSLAs – community-managed microfinance.

• We set up and investigate a pipeline to exploit the steps of participation.

• We compare the poverty profile of those staying in the pipeline with those left behind.

• The poorest are attracted by the VSLAs but are less likely to join. The very poor join only to a limited extent.

• VSLAs do provide financial services to some of the poorest households.

Summary

Reaching the poorest is an important objective in many development interventions, and microfinance is no exception. We review performance indicators for effectiveness of targeting described in the literature and suggest a new metric in order to account for extent and severity of poverty as well as the income distribution among the poor. When applying this to a panel dataset from a community-managed microfinance intervention in Northern Malawi, we find regressive targeting: Participants are less poor than the general population in the area. In addition, we provide suggestions as to when and why the poor exit the project.

Introduction

Together with economic growth, poverty reduction is perhaps the most agreed-upon goal for development aid and is also the first of the Millennium Development Goals. While economic growth may eventually trickle down and reduce poverty, it is also generally agreed that interventions and resources must target the poorest members of the population to achieve this goal efficiently. At the same time, however, an increasing number of development interventions require the participants to have a strong capacity for involvement. As this requirement could conflict with the outreach, more information is needed about how the design and implementation of interventions might affect the outreach. In other words: Do the poorest have the ability to participate?

Microfinance, which includes the provision of loans, savings, and insurance, is a particularly interesting concept in this respect. On the one hand, funders and implementers clearly want to reach the poorest members of the population. Targeting is widely used in microfinance as a means to do this; it is used, for example, by the Grameen Bank and BRAC (Bandiera et al., 2011), and there has been an increasing focus on avoiding “mission drift”, whereby programs include richer people (Christen, 2001, Cull et al., 2007, Hermes et al., 2011). At the same time, however, it is commonly believed that microfinance does not reach the poorest households, and this finding is confirmed by early studies (Hulme, 2000, Navajas et al., 2000, Zeller et al., 2006). The reason that is typically given for this is that both microsaving and microcredit require resources, involvement, and skills on the part of participants. Both require that participants have basic financial literacy, and additionally savers need to have sources of monetary income while borrowers need to be able to use their loans productively, keep track of their repayment schedules, and manage the risks associated with taking on debt.

In this paper, we ask whether participants in a microfinance intervention in the northern region of Malawi are poorer or richer than the general population in the same geographical area, and we address four shortcomings that are common in the literature on targeting and outreach in general and on microfinance in particular. First, many microfinance interventions do not actually try to reach the poorest members of the population. For this reason, it is difficult to know whether microfinance simply does not work for this group or whether the poorest just require services that are different from what most microfinance interventions offer. Of the four institutions analyzed by Zeller et al. (2006), for example, only one, an Indian organization, actually aims to reach the poorest members of the population. The microfinance method we study, the highly standardized village savings and loan associations (VSLAs), is designed particularly with the poorest in mind.

Second, the literature usually relies on cross-section data. The result is that any welfare measure for program participants reflects the sum of pre-program welfare and program effects. This is a problem if programs work, because participants may seem better off compared to nonparticipants – not because they were better off initially, but because the program has improved their status. We therefore use panel data: We solicit data on poverty status prior to participation in a VSLA as well as information on participation two years after the startup of a VSLA.

Third, we seek to overcome the simplistic approach to poverty measurement often taken in the literature, where the question asked is frequently: Is the percentage of people below the poverty line higher among participants or nonparticipants? As is well known from the poverty measurement literature, this approach is problematic because it counts poor households that are just below the poverty line the same as very poor households living on half that amount. Strangely, this observation has not found its way to discussions on targeting and outreach. To address this problem, we develop a metric that is sensitive to both the depth of poverty—how far people are below the poverty line—and income distribution among the poor. We base the metric on the Foster–Greer–Thorbecke squared poverty cap.

Finally, it is common for research on outreach to investigate whether participants are poorer than nonparticipants but then to ignore the underlying question of why this is so. We address this question by analyzing the decision to participate in a VSLA as a pipeline, where each active choice that a household must make toward participation comes with the risk of the household “leaking” out of the pipeline. This approach allows us to assess whether there are particular components of the intervention that deter the poorest households from joining.

We find that the participants in VSLAs are richer than the population at large in the same area. Of course, there are participant households that are below the poverty line. Indeed, roughly half of the participants in our study are poor. Among nonparticipant households, however, the percentage below the poverty line is larger, and the targeting is therefore regressive: Participants in VSLAs are less poor than the overall population in the area, when we use measures such as number of meals per day, length of the households’ so-called hungry period, or a proxy metric to measure consumption. The single exception is our estimate of a household’s total consumption,¹ where the results are insignificant but with the point estimate pointing toward progressive targeting. We suspect that this exception may be due to a large measurement error.

Our results are particularly strong when we apply our own poverty metric, which allows for assessing the depth of outreach beyond simply comparing the mean consumption level among VSLA participants and nonparticipants. Asked about their reasons for not joining a VSLA, nonparticipants report a lack of cash to meet the compulsory savings requirements.

Using our sequential panel approach, we find that both poor households and those that are less poor are attracted by the initial awareness campaign for a VSLA, and that the poor households are actually more likely to join, provided that they have received information about the upcoming intervention. At a later time, however, richer households join the VSLAs, and in larger numbers. In other words, the awareness campaign seems to attract a different group of people than those who end up joining. Implementing organizations should keep this in mind when designing interventions. We believe this result about VSLAs can be extended to other types of development interventions that require active participation by households with a certain degree of initial skills and/or resources.

The rest of this article is organized as follows. The next section describes the VSLA intervention. The third section provides an overview of the methods we used, including a review of the existing targeting literature and our suggestion for an improved targeting metric based on the squared poverty gap. The fourth section explains our sequential approach. The three sections after that present the data, our empirical strategy, and our results. The final section discusses our conclusions and provides policy recommendations based on the results.