The Jury’s Decision on Microfinance

Microfinance has been one of the most popular approaches to micro-level development over the last twenty years.  It has grown in scope from just 13 million borrowers worldwide in 1997 to around 200 million borrowers today, in large part (I have argued here recently on the World Bank Impact blog) due to its incredibly compelling narrative  of lifting the poor in developing countries up by their own bootstraps.  What hadn’t been done until recently, was a series of rigorous experimental studies to see whether microfinance really worked as practitioners and other enthusiasts have claimed.  Indeed there has existed a great gulf in the perception of impact between microfinance researchers and microfinance practitioners, a gulf that is the subject of a new paper forthcoming in April 2015 in World Development on the research that my co-authors Ram Rasbanshi, Meng Huang and I carried out in Nepal that examines this particular issue.

So the jurors have returned from the deliberation room (six widely different experimental sites around the world) and the verdict has been rendered in the new January 2015 issue of the American Economic Journal: Applied Economics.  This issue, as a collective symposium of research (that was even designed for ease of comparison between studies) represents one of the most remarkable single issues of an economics journal I have witnessed, particularly from the perspective of a development economist, in that the topic is of unquestioned importance to millions, the studies extremely carefully executed, and with convincing external validity internal to the single AEJ issue.  It consists of a symposium of a half-dozen studies in which rigorous randomized controlled trials were carried out by some top development economists, including Esther Duflo, Abhijit Banerjee, Dean Karlan, Alessandro Tarozzi, and Costas Meghir.  It’s a good jury.

The results of these six studies done in Ethiopia, Mongolia, Morocco, Mexico, India, and Bosnia-Herzegovina differ in minor respects (which we would of course expect them to), but are remarkably consistent in major ways, and indeed fairly consistent with our own recent findings in Nepal: Microfinance has significant effects on promoting entrepreneurship, investment in and time allocated to small enterprises, smoothing out fluctuations in income, and enterprise revenues (in some studies).  Business profits seem to increase a little with microfinance.  However, microfinance doesn’t increase household income to any detectable extent, which is what we really want, and it doesn’t seem to produce significant changes in social indicators such as child schooling, something that we’d also like to see happen with microfinance.  It’s not that microfinance is damaging on average, it’s just that it is far from the silver bullet that many had believed it to be a decade ago.

The following is a segment from the introduction summarizing the results, written by Banerjee, Karlan, and Zinman:

“The full picture of the evidence suggests several tentative conclusions. Reassuringly, these echo the conclusions in a previous survey which covers a number of studies not in this issue (Banerjee 2013).

First, there is little evidence of transformative effects. The studies do not find clear evidence, or even much in the way of suggestive evidence, of reductions in poverty or substantial improvements in living standards. Nor is there robust evidence of improvements in social indicators.

Second, the lack of transformative effects is not for lack of trying in the sense of investment in business growth. There is pretty strong evidence that businesses expand, though the extent of expansion may be limited, and there are hints (eyeballing the pattern of positive coefficients across studies) that profits increase. The evidence on why expansion does not produce strong evidence of increases in household living standards is mixed: some studies find evidence suggesting that households trade off business income for wage income, while others suggest that larger businesses are no more profitable, even in level terms, than smaller ones (at least on average; we discuss heterogeneity below).

Third, the lack of transformative effects should not obscure other more modest, but potentially important, effects. The studies find some, if not entirely robust, evidence of effects on occupational choice, business scale, consumption choice, female decision power, and improved risk management. As we stated previously, if microcredit’s promise were increasing freedom of choice it would be closer to delivering on it.

Fourth, just as there is little support for microcredit’s strongest claims, there is also little support for microcredit’s harshest critics. The studies find little evidence of harmful effects, even with individual lending (Bosnia, Mongolia), and even at a high real interest rate (Mexico).

Fifth, the limited analysis of heterogeneous treatment effects in these studies does suggest the possibility of transformative effects—good for some, bad for others— on segments of microlenders’ target populations. Morocco and India find evidence of large positive effects on business profits in the right tail of the distribution, and Morocco finds negative effects on the left tail. Mexico finds an increase in financial decision power among the left tail. Bosnia finds a decrease in teen schooling among lower educated households. Microcredit’s strongest supporters and harshest critics may each be correct for segments of borrowers, if not on average.

Finally, we emphasize that many of these inferences lack precision, at least at the level of individual studies. Many of the null results are part of confidence intervals that contain economically meaningful effect sizes in one or both directions. Most of the null results would lack precision if they were converted into treatment-on-the-treated units. Statistical power still poses a major challenge to microcredit impact studies, although there are many hopeful indicators for future work, including the progress in the India and Morocco studies in predicting take-up, the Mexico study in obtaining a large sample size, and the prospect of pooling data across studies.”

The authors also point out that these studies have little to say on “inframarginal” borrowers, that is borrowers who have already had credit that are not part of the new group of borrowers induced to take credit from the experimental study.  This is a really important point, and the focus of our new World Development paper.  We argue that you can’t measure microfinance impacts only on late-takers because the productivity of loans is likely to originate from two sources: 1) a borrower’s innate productivity and 2) fluctuating opportunities that the economy provides which make the value of borrowed capital higher.  Early takers are thus likely to be more productive borrowers because they are more likely to have productive opportunities at the point that microfinance comes to town rather than the late takers, less productive borrowers for whom taking a microfinance loan makes sense only when an extraordinary opportunity happens to arise later.  That being said, these studies have much to say about the impacts of expanding microfinance in areas in which it already exists, and the results are certainly grounds for approaching new microfinance efforts with a far more cautious and measured enthusiasm.

• ← Border Crisis = Church Opportunity
• The Behavioral Economics of ISIS →