Esther Duflo recently presented this paper at the Jackson Hole monetary policy conference. In the “perfect world” model of the economy, there is no role for inequality to play in growth: investment depends on marginal returns, not on who has the cash. A good prospective company gets their money, earns their return, and pays back whoever lent. Marginal propensity to save matters, of course, and is affected by inequality, so that in a Kaldor world high inequality leads to higher savings leads to higher growth. But are we forgetting some constraints? Indeed.
Poor countries are a world of special constraints: credit constraints, nonexistent insurance markets, nonoptimal education decisions, etc. Duflo asks, given these constraints, might pro-equality policies also be pro-growth? Or, what types of pro-growth policies might we favor if we take effects on the poorest into special consideration? Unsurprising if you know Duflo’s work, evidence in this paper is drawn strictly from well-identified micro-studies rather than cross-country regressions or the like; I won’t go into this choice further here, except to say that awful quality of aggregate data on the third world alone seems reason enough to me to support Duflo’s position.
A few stylized facts, first. The gap between interest rates for lenders and borrowers in the third world is enormous: 30 to 40 percentage points per year is not strange. This is perhaps due to poor legal systems which necessitate extensive spending on loan recovery. It is not due to high default rates. Marginal returns to investment in small and medium-sized businesses in the third world is often huge: 50 to 100 percent returns have been found in well-identified studies. Misallocation of capital due to trust issues is widespread. Saving is “difficult” due to expensive (as a percentage of annual income) fees on savings accounts and threats of theft with informal savings mechanisms. Insurance markets, such as weather insurance for farmers, often don’t exist, and take-up rates are very low for actuarially fair insurance; the reasons why are numerous. Parents often undereducate children, both because they do not value childhood welfare equal to their own (as in a Becker model), because education systems are very poor, and because they misunderstand returns to education (it is log-linear, not “lumpy”, but parents often think only kids smart enough to get a government job or similar are worth educating). Many of these constraints combine to keep labor mobility across space suboptimally low, primarily by keeping too many peasants in rural areas.
Pro-growth and pro-poor policies can mitigate many of these problems. A number of studies have shown interventions in primary education that improve student outcomes in poor countries. Subsidized crop insurance may lead to higher growth by letting farmers grow crops with higher return by higher variability. Microcredit may not be that useful (if the production function has thresholds, small loans may allow for less effort while not allowing for high marginal return investments that can pull the poor out of poverty) whereas loan schemes to medium-sized businesses with smaller lend-borrow interest rate gaps may be particularly useful. Property right clarification a la Hernando de Soto can allow the poor to invest more freely, escaping poverty traps. The point of this paper is less to provide precise policies than to show where useful future research may lie concerning pro-poor growth.
One final note on Duflo’s RCT-style development work. Duflo (and Banerjee) could win a Nobel this year and nobody should complain: they would fully deserve it. That said, as useful as their work is, there is an obvious bubble in identification-heavy micro-level development work. I don’t think “replicate studies in 10 different villages in 10 countries” is in any way a good form of testing for external validity, and I also don’t think it’s good that the huge majority of young economists going into development are focusing their attention away from the “big picture” questions. There is a middle ground: identification with theoretically-sound structural models. That is, we take rigorous data collection and careful work on causality, and combine it with “big picture” theory that allows us to compare results across regions and to learn how interconnected aspects of the market function. And I think development is already moving in that direction! (Indeed, precisely these comments apply to a lot of experimental economics, but that’s a discussion for another day…)
http://www.kansascityfed.org/publicat/sympos/2011/2011.Duflo.Paper.pdf (Aug 2011 KC Fed Jackson Hole Conference final version)