Typical policies to improve the incomes of poor households and their businesses are based on the sustained provision of services – be it microfinance with multiple loan cycles and regular meetings; conditional cash transfers with regular transfers over a period of years; or business training programs which are based on the idea that capital along is not enough – as in the proverb “give a man a fish and he eats for a day, teach a man to fish and he can feed himself for life”.
In a new paper with Suresh de Mel and Chris Woodruff, published online today in Science, we ask whether the much simpler policy of giving a one-time grant to small business owners has any long-term effect. [An aside to World Bank/IFC readers, I am delighted that our library now finally has arranged full-text online access to Science – and the paper is short, so click through!]. We do this by following up in June 2010 and December 2010 with the microenterprises originally interviewed in April 2005 for an experiment in which we gave one-time grants of approximately $100 and $200 to male and female microenterprise owners in Sri Lanka. The short-term results discussed here and here, found high returns to capital for men and zero returns for women. This new work asks whether these returns persist, and whether the grants affect business survival.
A standard Ramsey model would predict that such grants would only have temporary effects, speeding enterprises to their efficient steady state dictated by the owner’s ability and industry they work in. In contrast, such grants may have long-term effects if there are poverty traps, or persistent underinvestment caused by self-control problems, or intra-household inefficiencies.
Key long-term impacts
· The grants help male-owned firms survive: We find that the one-time grants reduce the rate of firm closure by 10.9 percentage points for male-owned microenterprises, relative to 29% of the control group shutting down over 5 years. The grants don’t seem to differentially affect which firms survive, at least in terms of observable characteristics. In contrast, there is no impact on survival of female-owned firms.
· The surviving male-owned firms are still more profitable 5 years later: A one-off $100 grant is estimated to increase firm profits by $6-12 per month for male owners, with this relatively constant over the 5 years post-treatment – there is no evidence of catch-up by the control group, nor is there evidence of dramatic compounding of returns leading to increasing divergence.
· The grants continue not to have any impacts on female-owned microenterprises in the long-run. This appears to be due to a combination of some of the grants not staying in the business but rather being used for household needs, and to the grants which are invested having low-returns, in part because the industries many of these subsistence female microenterprises work in have low efficient scale.
Thus giving a man a fish (or at least money) seems to help him over a long period of time. We have a new Finance and PSD Impact note which discusses these results in more detail. For this blog, I thought I’d focus more on some of the areas of interest for impact evaluation.
Tracking long-term outcomes
One of the critiques of many randomized experiments, including of many of the recent microfinance evaluations, is that periods of one to two years may be too short to see all the benefits of access to capital appear – or conversely in other cases, that temporary large impacts may not persist. However, there are questions about both the incentives of researchers to monitor impacts over longer periods, as well as the feasibility of doing so because of attrition concerns. This paper shows it is possible to track firm outcomes over longer periods – we re-interviewed 90% of the firms in June 2010, and 92% in December 2010. Taking two follow-up surveys meant that we were able to get some firms in one survey round that were not able to be interviewed at the other point in time, so overall we interviewed 94% of firms in at least one of the follow-ups. In-person observation and interviewing of neighbors was used to assess whether the businesses still were in operation for the remaining 6%.
There have been several other recent studies which look at longer-term effects in health and education. Baird et al. and Ozier use the same survey to look at the long-term impacts of one of the earliest randomized experiments - the de-worming intervention in Busia, Kenya. They tracked 65% of youth a decade later, and used intensive tracking methods on a sub-sample to obtain a weighted survey response rate of 84% of 5,569 respondents. Friedman (Willa not Jed) et al. examine the impacts 4-5 years later of a girl’s scholarship program, also in the same region of Kenya. They have follow-up data on 1,385 girls, which is 42% of their initial sample of 3292 students, but use intensive follow-up on a subsample and reweighting to achieve an effective response rate of 79%. Finally, Jensen looks at impacts 3.5 to 4 years after an intervention which provided information on the returns to schooling in the Dominican Republic. Of the original sample of 2,250 students, he is able to verify school attendance outcomes for 91% of the students.
Taken together these set of studies do show the feasibility of doing longer-term follow-ups. Tracking is hard work, and requires advance planning to collect details of contacts who can help find people if they move, but it can be done in some cases at least.
Firm profits are noisy, and our sample is only 387 firms. To get precision in our estimates we therefore pool several rounds of surveys together – in total we interviewed these firms 13 times, providing sufficient information on them to average out any one-off measurement errors or idiosyncratic shocks.
Examining which firms survive
A concern in firm studies when trying to measure impacts on firm outcomes like profits is that the treatments may affect not just levels of survival, but also the characteristics of which firms survive. (Similar issues arise in labor market studies of wage impacts where treatments also affect employment). Then even though treatment is randomly allocated among the full sample, it is not randomly allocated among the survivors. In our case, the concern might be that the grants allow the least profitable firms to survive who would have otherwise failed, thereby reducing the mean profits for the treatment group and causing us to understate the impact on profitability. The Supporting Online Materials of the Science article contain a number of checks to try and ensure this is not occurring in our case.