A survey highlights the results of the first randomized studies
Thirty years into the movement, it might seem strange that researchers are still asking whether microfinance reduces poverty. In fact, by the standards used to judge whether drugs are safe and effective in the bloodstreams of people, the safety and effectiveness of microfinance injected into the fabric of villages and barrios remains unproven. Somewhat by chance, 2009 is turning out to be a pivotal year in the study of the impacts of microfinance. A new generation of studies is emerging that promises to give us a clearer view of the effects.
Few doubt that financial services have helped some poor people permanently improve their circumstances. What is unclear is whether microfinance reduces poverty on average. The worry is that microcredit in particular might leave some people worse off, just as in rich countries credit cards and mortgages have.
A Lack of Solid Evidence So Far
Over the years, academics and evaluators have conducted many studies on the impact of microfinance, especially microcredit. Yet the average effect is still unknown because nearly all studies to date share a serious shortcoming: when studying complex social systems such as families and communities, it is extremely hard to use a correlation to prove a causation. If affluence and microcredit go hand-in-hand, does that mean that the better-off borrow more or that borrowing makes people better off?
Much of econometrics consists of devising mathematical techniques to address this challenge. Despite the rigorous mathematics involved, these methods often fail: even when econometricians have done the best they can with the data they have, it has typically not sufficed to prove that A causes B instead of B causing A. Indeed, the complexity of the latest techniques sometimes only makes it harder to detect the failures. In a new paper, Jonathan Morduch and I make this point with respect to what were historically the leading studies of microcredit's impacts. We replicated three noted studies of data from Bangladesh in the 1990s, including one by Morduch. That is, we applied the original methods to the original data in order to scrutinize these influential analyses on our computers.
In the case of the most prestigious of these (the 1998 study by Pitt and Khandker), which Muhammad Yunus has indirectly cited as showing that 5 percent of Grameen borrowers climb out of poverty each year, we arrived at the opposite result of the paper. Seemingly, lending microcredit to women made their families poorer. We ran additional statistical tests to further examine the discrepancy. Our results question key assumptions behind the original claims that microcredit was affecting household spending and not the other way around. The negative relationship we found may arise from women in richer households taking less (or no) microcredit.
Should such findings drive us to nihilism? Should we dismiss all statistical studies of microfinance's impacts? Not quite. One method of study rises above the controversies over causality: randomized controlled trials (RCTs), which randomly offer some people a service while withholding it from others (at least temporarily). What sets RCTs apart, and what made them the gold standard for drug trials, is that they introduce a source of variation into the world that is independent of everything else. If women randomly offered microcredit earn more money years later than those not offered, the only creditable explanation is that microcredit made the difference.
Emerging Randomized Control Trials
The first RCT study on loans to the poor appeared in 2006, though the loans in question might not fit your definitions of microcredit. Yale's Dean Karlan and Dartmouth's Jonathan Zinman worked with a cash lender, not unlike a payday lender in the United States, to randomly "unreject" some applicants whose scores from an automated credit rating system fell below the acceptance threshold. Six to twelve months after they applied for the four-month loans, unrejected applicants were 10 percentage points more likely to have a job, 7 points less likely to be below the poverty line, and 6 points less likely to report that someone in the household had gone to bed hungry in the last month than those rejected.
Karlan and Zinman did a similar study of individual microcredit in Manila; this time, however, the sample was middle class. The study produced strikingly different results, which just appeared in July 2009. The authors found no changes in household income, spending, or diet 1-2 years later. Borrowers did appear to cut back on some types of spending, including paid helpers, health insurance, and home improvements, perhaps because of belt-tightening at the beginning of new, loan-enabled investments.
Another new study does focus on people usually thought of as microcredit targets. Researchers at the Jameel Poverty Action Lab (J-PAL) worked with Spandana to randomize the district-by-district order of Spandana's expansion into the Indian city of Hyderabad. Household spending averaged about $1/day among the study subjects. Following up some 15-18 months after credit was offered, the researchers found no impact on total income, spending, health, or school enrollment rates. They did find that microcredit boosted profits for families that already had a business. And those deemed most likely from the outset to start their first business, as indicated by having more land or more working-age or literate women, in fact did so more in the districts with microcredit offerings.
The brightest spot in the new research has also been the least heralded trial of savings undertaken by two economists just finishing their dissertations. Pascaline Dupas and Jonathan Robinson worked with a cooperative in Kenya to offer savings accounts to vendors in a market town on the road linking Kampala to Nairobi. The study was also original in using diaries rather than a one-off follow-up survey to track subjects' financial transactions. Despite the small sample - 122 people were randomly offered an account and 67 took it - the researchers found significant impacts, on women. The accounts appeared to help them accumulate money for investments such as stock for their stores, leading eventually to greater prosperity. Unclear is whether the accounts, which charge significant fees to discourage withdrawals, helped women save by giving them more control over their own spending impulses, or by giving them a way to deflect family requests for money.
Surveying these new works, it becomes apparent how fragmentary are the insights they offer. We have just one strong study of group credit for the truly poor, for instance, and it only shows us effects about one year out. Claims that microcredit is a proven anti-poverty intervention thus seem dubious. But more RCTs are coming, and they will paint a fuller picture.That said, our judgment of the impacts of microfinance should not stand or fall solely on whether studies prove that it systematically reduces poverty. Can tens of million clients be wrong? In my book, I am also looking at microfinance through Amartya Sen's "development as freedom" lens, which leads to a focus on when microfinance gives people more control over their lives and when less. I also consider the "development as institution-building" perspective, in which the great contribution of microfinance is the enrichment of the institutional fabric of nations. Statistical studies are important, but they do not substitute for deeper reflections on the process of development and how financial services can contribute to it.