1. The Economics Production Function: Over the last few years, papers on microenterprises generally shared a couple of remarkable--given the general narrative--findings: microenterprises (on average) didn't grow no matter what you did to try to boost them, and women-owned microenterprises performed worse than male-owned ones. Those findings led to plenty of yowls from practitioners whose work, livelihoods and in some cases core beliefs were based on the opposite. In many conversations I had, I got the impression that people outside the profession believed that economists would publish these findings and then move on. But that perception really misunderstands the motivations of economists and the way the field works. Economists don't leave puzzles alone once they find them--the field pursues them relentlessly. The best session I attended this weekend was based on the particular puzzle of why female-owned microenterprises are less profitable. Natalia Rigol presented work following up on an earlier studies that documented the profitability gender gap, finding that the source of the gap is mostly due to lower returns from female-owned enterprises where there was another (male-owned) enterprise in the household. Those male-owned enterprises were in more profitable industries (something documented in the original studies), so the households were making quite rational decisions to allocate additional funds to the more profitable business (and making it look as if the female-owned business had 0 or negative returns). In households where there was only a female-owned business there is no gap in returns to capital. Leonardo Iacovone and David McKenzie presented on efforts in Mexico and Togo, respectively, to provide training to help women entrepreneurs improve their businesses with positive results--in both cases seemingly based on personal initiative training rather than business skills. And Gisella Nagy presented results (unfortunately there's nothing yet to point to on this one) that women tailors in Ghana show lower profitability than male tailors because there are more women tailors which drives down prices they can get in the market. This last finding is particularly important because it suggests that part of the way forward for microcredit aimed at building women's businesses is to do a much better job targeting, or as I've called it elsewhere, abandoning the vaccine (everyone gets one!) model of microcredit for an antibiotic (only people who really need it get one!) model. And all of that is just a very small sample of work being done on the puzzle of heterogeneity of returns to microenterprises and what can be done about it. I'm now sorely tempted to write an overview on all these studies, but dammit I really want to get to "subsistence retail."
2. Causal Inference is Hard: Those two topics aren't orthogonal to each other of course. One way they are joined together is my common theme about how hard causal inference is for the average person, and in particular for the subsistence (or just above) operator of a microentrprise (whether farming or retail). That's what I kept thinking about when reading this new post from David McKenzie on "Statistical Power and the Funnel of Attribution". David is writing for economists trying to write convincing papers, but this point "Failure to see impacts on your ultimate outcome need not mean the program has no effect, just that the funnel of attribution is long and narrows" is equally important for the people being treated. If the funnel of attribution is long and narrows, then its approaching impossible for the individual (not gifted with a large sample size or a deep understanding of statistics) to figure out which of their actions actually matter. There is a connection to AEA here. As I was perusing the poster displays (also known as "the saddest place on earth") I kept hearing people arguing with Jacob Cosman, the creator of a poster about how the opening of new restaurants in a neighborhood affects the behavior of existing restaurants. The answer: a very precisely estimated no effect at all. (Here's a link to an old version of the paper with somewhat different results.) Economists walking by simply couldn't believe this and were constantly suggesting to the author things he must have done wrong. I was amused. My strong prior is that a person would not open one of these restaurants unless they believed that their restaurant was unique (otherwise, you would believe that your restaurant would quickly fail like the 90+% of other small restaurants and you wouldn't open in the first place). So when another new restaurant arrives, you don't actually see it as a threat that needs a response. You are, after all, different! But even if you did think you needed to respond, how would you possibly know what the right response was? Do prices matter? Menus? Advertising? Item descriptions? Coupons? The funnel of attribution on all of these is so long and imprecise we should assume that individual entrepreneurs have no idea what to do even if they wanted to do something. Which ultimately brings us back to why it's so hard to get microentrepreneurs to change their behavior in a lasting way, and why personal initiative training may work much better than business skills. Personal initiative training teaches you that what you do matters, even if you can't tell, while business skills training teaches you to do something even though you can't tell that it matters.