Amazing Grace Periods

I was a discussant last week for the second entry in a new seminar series hosted here in Washington by 3ie and IFPRI. The discussee was Erica Field of Duke University, who presented a paper she wrote with Rohini Pande, John Papp, and Natalia Rigol. The paper reports the results of a randomized controlled trial (RCT) on microcredit. I had not focused on it before because it does not attack that bottom-line impact question: it does not compare the fates of people offered microcredit to the fates of those not offered it, in order to tell us whether "microcredit works." Rather, both the treatment and control groups consist of micro-borrowers. The difference is that one group got a two-month grace period. The other had to start repaying at the next weekly meeting, as is the norm.It is well understood that microcreditors require that repayment begin immediately because it keeps repayment rates high. Think of how teaser rates---easy terms early on---led to defaults in the U.S. It is also well understood that many borrowers respond by setting aside a part of their loans to cover those first few payments. Why? Their investments, if indeed they invest, may not earn returns fast enough to cover those early payments. Naturally, the set-asides reduce how much they can invest and earn.This study sets out to assess the trade-offs here. How much would it cost creditors to offer grace periods? How much would it benefit borrowers?There's a lot to like in the paper:

• It discovers a remarkably large effect: deferring repayment by two months on these normally-44-week loans increased investment by a seemingly modest 6% (roughly what would have gone into repayments in the first couple of months) but that translated three years later into profits a third higher and household income about 20% higher. (However the exact results appear sensitive to inclusions or exclusion of a few extreme data points.) The variability of profits is also substantially higher, suggesting that the grace periods are freeing people to make riskier but potentially more profitable investments.
• The flipside of not being a microcredit impact study is that it is interestingly positioned between that kind of research and a separate literature (to use a fancy academic word) on the returns to capital among microentrepreneurs. I.e., it probes this mystery: why are the microcredit impact studies returning zeroes while equally rigorous research finds microentrepreneurs earning returns on additional capital that would inspire envy in any hedge fund manager---like 5%/month (~60%/year)? The hypothesis explored and supported here is that the form of the microcredit contract is an obstacle. It appears that the rates of return to investment are indeed high, so that over three years 6% additional investment compounds into much higher income gains.
• The study comes tantalizingly close to assessing longer-term impacts. Notice the "three years" above. No microcredit impact study in the public domain has tracked outcomes that long, so I see this as a key frontier for research. To be clear, Field et al. show one kind of microcredit being much better than another. They can't tell us with the same rigor what the absolute impacts are. Possibly both are fantastic, the grace-period kind more so. Possibly both are terrible, the grace-period version less so. The best guess based on what we know is that standard microcredit is neutral out to three years, as it apparently is out to one or two; and that giving a grace period puts it well into positive territory. That implication, though tentative, is important.
• Unusually, the paper includes two case studies, which wonderfully enrich the presentation. One is of a woman who invested more in her sari-selling business. The other is of a woman who handed her loan to her husband, who invested in his tailoring business---and repaid the loan 24 weeks late.

• Another important result is that the grace period did increase delinquency rates. So for the lender, this experiment brought cost without benefit. From the point of view of financial viability, microcreditors have been right to insist that repayment start immediately. What I'd like to see more fully analyzed is how the costs compare to the benefits. If the clients are gaining from the grace period more than the lender is losing, then perhaps there's an opportunity to make the world a better place. Perhaps social investors in microfinance institutions (MFIs) could subsidize microcredit-with-grace-periods by accepting lower returns on their investments as long as grace periods were implemented. Or perhaps MFIs can couple grace periods with slightly higher interest rates to offset their losses. (Not obvious though: interest rate caps might stop them; or higher rates might deter the most conservative microentrepreneurs, further raising average delinquency among remaining clients.)
• The story of the tailor made me wonder about the gender dimension. Can the researchers tell us whether male- or female-led businesses were more likely to take those bigger risks, or earn those higher profits, or fall behind more on the loans?
• The people in this study appear to have been substantially better off to begin with than those in the Hyderabad microcredit impact study. There, average household spending per capita was 1,419 rupees; here average household income is 20,544 rupees, which, divided by an average household size of 4.1 people, yields 5,010 rupees/person.
• The paper includes a mathematical model. I think it is good to accompany rigorous empiricism with rigorous theory. As a reader, I always appreciate a theory section because I can skip over it and feel like I'm making progress. :-)