Hammer Blows or Pinpricks? Stories vs. Statistics in Microfinance

July 22, 2009

It’s been a tough season for the proposition that “microfinance is a proven and cost-effective tool to help the very poor lift themselves out of poverty and improve the lives of their families.” In May came a randomized trial of microcredit in Hyderabad finding no impacts on poverty 15–18 months out. In June came a paper challenging the leading older-generation studies that seemed to show that microcredit had cut poverty in Bangladesh in the 1990s. Now in July we have another randomized trial, of microcredit in Manila, also finding no impact on bottom-line measures of household welfare.

A couple of people, including Tim Ogden, have raised a good question with me in the last few days: What does this mean for microfinance? Has a myth been debunked? Is the whole movement about to implode in a ball of fire? More precisely, will this research perturb the dominant narrative about microfinance in the public mind, about microenterprise as a reliable ladder out of poverty?

I’m reminded of two Mark Twain saws—the one about lies, damned lies, and statistics; and the one about the reports of his death being greatly exaggerated.

Reasons this isn’t quite a David (ahem) and Goliath story:

• The new and randomized trials of credit, though credible, have produced only fragmentary knowledge. On the one hand, that does undercut the assertion that microcredit is a proven anti-poverty tool. On the other, it doesn’t prove that microcredit rarely helps. The Karlan and Zinman Manila sample is predominantly middle class—people who were already out of poverty. Their earlier South Africa study looked at four-month loans to people who (I think) could show pay stubs, and who worked in a system that implicitly expected them to borrow in order to pay for on-the-job training. (See previous post.) The one full-bore study of microcredit for poor people without jobs, in Hyderabad, does not look at outcomes beyond the first 15–18 months.
• The Dupas and Robinson randomized study found benefits from savings—again out of sync with myth, but good news for microfinance more broadly.
• Also appearing in the last few months is the great book, Portfolios of the Poor. It makes no claim that microfinance raises income on average. But it does argue that microfinance is valuable in helping people smooth that income (for a price). To put this more personally, I mentally file all those statistical studies in one chapter of my book, “development as measured impact.” But since our knowledge of the actual impacts in millions of lives is doomed to be limited, I also devote chapters to other notions of success. The ideas in Portfolios of the Poor fit into my “development as freedom” chapter. And into my “development as institution-building” chapter fits the thesis that a chief contribution of microfinance is its enrichment of the institutional fabric of nations.

In other words, the challenge to microfinance per se is far from mortal.

But I do think the new studies pose a test for the story that microfinance promoters have so effectively constructed. Are we approaching a tipping point, the precipice of a sort of Kuhnian revolution in public perception? Will the new research combine with the recent and dramatic demonstrations of the dangers of debt to depose one image of microfinance and seat another in its place? I of course don’t know. But I am launched in this book project in the belief that such change is possible and healthy.

Tweet  

1. Asif Dowla Says:
   July 22nd, 2009 at 12:33 pm

   To see why stories are more important than statistics, read the op-ed by Kristof in NYT http://www.nytimes.com/2009/07.....istof.html

   The scientific study that forms the basis of the column quotes Mother Theresa “If I look at the mass I will never act. If I look at the one, I will.”

2. Daniel Says:
   July 22nd, 2009 at 6:50 pm

   As you imply, offering an individual a line of credit might be considered to serve two key functions:

   1. Reducing credit constraints
   2. Allowing households to better smooth consumption (they can take a loan if times are temporarily tough).

   Even within a simple lifetime expected utility framework, both would be expected to improve welfare. However the first might be expected to increase consumption growth (as individuals can take on more upside risk), the second to decrease it (as individuals are happy to pay to smooth consumption).

   Unfortunately there aren’t any good positive assumptions that can be made about utility functions and temporal preferences, so preferences over the shape of consumption profiles must be inferred from behaviour. Now, if only lots of people were purchasing microcredit: then we might be able to infer that it was a welfare improving intervention

   In most other branches of finance, theory leads practice (often by a long way). My ultimate hope for microfinance RCTs is that by unpacking the various effects of microcredit and other financial products currently offered to the poor, economists might have (even more) useful things to say about product design.

3. anonymous Says:
   July 28th, 2009 at 2:02 pm

   It seems that, with the exception of Karlan’s work, for most these papers the “treatment” is the opportunity to take out a loan–not the effect of the loan itself. In this case, it’s less surprising to me that there’s a small effect. Karlan’s work looks at those who have already self-selected to apply for loans; didn’t he find large effects in his South Africa study?

4. David Roodman Says:
   July 28th, 2009 at 3:09 pm

   It’s true that the studies only look at a) the effect offering a loan to people who b) self-select to apply. And offering a loan should have less effect on average than actually giving one. But I think those qualifiers actually give the study real-world significance. When we contribute to microfinance groups we are funding them to offer loans—they can’t force anyone to take the loans (we hope!). The study asks whether such offers do good on average.

5. Heather Says:
   August 4th, 2009 at 2:29 am

   You and I must have read two very different Spandana studies. The one I read indicated that the treatment area exhibited 30% more new businesses than control area, and that the most profitable decile of businesses in the treatment area was more profitable than the most profitable decile in the control area. Furthermore, the study reported that high-propensity entrepreneurs cut spending on non-durables and temptation goods and diverted this money to durable goods (likely used to launch a business). Low-propensity entrepreneurs increased spending on non-durables and temptation goods and used the loans to smooth their consumption (thus becoming non-liquidity constrained PIH consumers). The report noted no discernable effects on female empowerment, education, and health; however, the positive effects on business and consumption cannot be ignored.

6. David Roodman Says:
   August 4th, 2009 at 7:41 am

   Heather if you haven’t already you might want to check my review of the Hyderabad study, linked to from this post, and see if my fuller description makes it look more familiar. There, I note the effects on business start-ups, profits, and spending patterns. The study found no effect on total household income or spending despite the higher profits for existing businesses. (Maybe the higher profits were mostly offset by lower wages? Maybe some new businesses lost money? Or maybe the profits measure has survivorship bias??) The evidence for consumption smoothing, if I read right, is that 30% of borrowers reported that’s what they were doing with the loans, as opposed to paying off other loans or investing durable goods, say. Having read Portfolios of the Poor, I would have been surprised if the number had been much lower. There’s no doubt that a lot of poor people use the loans to smooth consumption, and that that can be a very good thing. This post was about the big idea that microcredit reliably lifts people out of poverty, and I think it’s accurate to say that the Hyderabad work has yet to demonstrate that.