One of the mysteries of development economics is why more people in subsistence agriculture don't migrate to cities where incomes are much, much higher. New data suggests one answer: when they move, their incomes may not go up as much as we thought.

Poverty in poor countries is largely a rural phenomenon. The original engine of development in Arthur Lewis's Nobel-winning work in the 1950s was the movement of people out of farming in pursuit of higher urban incomes, and this same movement away from agriculture remains a focus in work on structural transformation today (see CGD work by Peter Timmer and Selvyn Alkus, and others like Margaret MacMillan and Dani Rodrik's here).

The agricultural productivity gap—or the ratio of value-added per worker in other sectors compared to agriculture—is huge. A couple years ago, Doug Gollin, David Lagakos, and Michael Waugh published a paper in the Quarterly Journal of Economics systematically documenting these gaps across a sample of 151 developing countries, and found that on average, value-added per worker in other sectors was roughly three times higher than in agriculture.

The agricultural productivity gap

Click on citations in the legend to view the cited paper.

Economists see income gaps and begin to tear up at the inefficiencies and arbitrage opportunities. As Gollin et al. wrote, if we take these gaps at face value they suggest that “by reallocating workers out of agriculture, where the value of their marginal product is low, and into other activities, aggregate output would increase even without increasing the amount of inputs employed in production.”

Maybe farmers are just different

Perhaps the most obvious explanation for income differences between farmers and non-farmers is the difference in human capital between workers in each sector. And it's true, education differences are big. But when Gollin et al. try to adjust for these schooling differences—allowing not just for differences in years of schooling, but also for actual learning measured by literacy—the agricultural productivity gap only falls from about 300 percent to about 200 percent. It's still huge.

If the potential income gains are so big, why doesn't everyone move to the city?

Enter a new paper, presented at the American Economic Association conference in Philadelphia last weekend, by Joan Hicks, Marieke Kleemans, Nicholas Li, and Ted Miguel, which attempts to resolve this mystery.

Hicks et al.’s data allows them to go a couple steps further. They measure cognitive ability more carefully than most economic surveys, and find higher-skill farmers select into migration. But perhaps more interestingly, they focus on two long-term longitudinal data sets that track migrants over many years as they move jobs between sectors. This allows them to compare the same worker—with the same education, cognitive skills, and other possibly unmeasurable attributes—in different sectors to estimate the agricultural productivity gap.

What they find is, essentially, nothing.

In Indonesia they find individuals earn about 8 percent more working outside agriculture than from farming, and in Kenya about 6 percent more. When they focus on the rural-urban divide rather than farming versus non-farming, the gap is almost precisely zero in Indonesia and about 17 percent in Kenya. Perhaps 17 percent is not trivial, but it's a very far cry from 200 to 300 percent.

Ruminating on the intellection implications for economists, Dietz Vollrath has an insightful blog post discussing the Hicks et al. paper and the broader literature, provocatively titled “The Return of Peasant Mentality.” As he notes, the modern assumption in development economics is that rural farmers are just like everyone else, i.e., rational maximizing agents, just stuck in a more challenging context. The Hicks et al. evidence suggests instead that they self-selected on measures of cognitive skills. Would we find the same if we measured risk aversion and other psychological attributes? Maybe, as Vollrath speculates, peasants really are different, as an earlier era of development scholars assumed.

So should policymakers turn away from rural-urban migration as an anti-poverty tool? (No.)

The “No Lean Season” initiative in Bangladesh has garnered a lot of attention recently for showing the large welfare benefits from encouraging rural-urban migration. A randomized trial by Gharad Bryan, Shyamal Chowdhury, and Mushfiq Mobarak offered individuals in rural Bangladesh an incentive of about $8.50 to migrate to the city for work during the hungry season. That relatively small incentive induced about 22 percent of people to move temporarily. The result is that family members who stay behind increase consumption by 30 to 35 percent, and eat 550 to 700 more calories per day.

There's a lot of space between 300 percent and zero. Even if the agricultural productivity gap overestimates the gains from moving out of agriculture, the returns to rural-urban migration may still be significant—and well worth promoting through public policy.

Bryan et al.’s RCT in Bangladesh also provides interesting answers to why more people don't migrate spontaneously: they document important roles for risk, subsistence constraints, and for learning about the returns to migration—which induced remigration for years after the treatment incentives are removed once people have gone and seen what they can earn.

chart of the distribution of total earnings for migrants

Reproduced from Bryan, Chowdhury, and Mobarak (Econometrica 2014)

There's also some evidence that the returns to more permanent (as opposed to the seasonal migration in the Bangladesh study) may be higher in other places.

While Hicks et al. find very little in Kenya and Indonesia, an earlier study in Tanzania by Kathleen Beegle, Joachim De Weerdt, and Stefan Dercon find that rural-urban migrants experience consumption gains of around 30 percent (still a far cry from 300 percent). And while the results are slightly less clean for a variety of data reasons, Alan De Brauw, Valerie Mueller, and Tassew Woldehanna find much, much bigger gains from movement when tracking rural-urban migrants in Ethiopia over many years.

At a bare minimum though, this new evidence from Kenya and Indonesia suggests policymakers should be reluctant to assume that leaving the farm is better for people than they realize, and attempt to coerce movement. Here Hicks et al. invoke Tanzania's disastrous forced villagization policy of the 1970s—hopefully not a policy with many contemporary analogs, but still worth ruling out. And they also highlight that positive selection into migration (i.e., movement of higher skilled people, even on unobservable dimensions) reinforces the tendency for this ladder out of poverty to leave the least fortunate behind, which could be read as a plea not to forget rural development and social protection programs during the structural transformation process.

Should this dampen your enthusiasm for international migration? (Also no.)

While poor farmers in many developing countries are “free,” in the narrowest legal sense, to migrate to urban areas and search for a non-farming job, that's not true for international migrants, who face walls and fences and police demanding papers. With those “frictions,” it makes sense that international income gaps are larger and somewhat more robust.

In one of my favorite CGD papers, my colleagues Michael Clemens and Lant Pritchett together with Claudio Montenegro from the World Bank compare immigrants in the US to observationally identical workers (i.e., same education, age, etc.) in their home countries, and find a lower-bound estimate of an earnings increase of over $13,000 per year in PPP dollars from living and working in America.

The same biases that Hicks et al. identify in rural-urban migration apply internationally—higher-earners select into migration—but once that's accounted for, the returns remain large. Using a lottery of Tongans admitted to New Zealand, David McKenzie, Steven Stillman, and John Gibson show that naive non-experimental estimates of the income gains from international migration overstate the truth by twofold. Nevertheless, they show that randomly selected migrants still experience a 263 percent increase in income (!) one year after migrating.

In sum, there is a risk that “why don’t they just move” becomes the “let them eat cake” of 21st-century development economists. The Hicks et al. paper is perhaps a good reminder to focus on removing the barriers to people’s movement, rather than thinking we know what’s best for them. While the gains to migrants in Kenya and Indonesia appear disappointing, the centrality of structural transformation to poverty reduction in history remains uncontested. It’s not always a simple story though, and more research like this will help us understand that messy process better.


Thanks to Doug Gollin for long and patient answers to my questions, to Michael Clemens for helpful suggestions, and to Divyanshi Wadhwa for research assistance.