The Return of Peasant Mentality?

Posted by Dietrich Vollrath on June 28, 2017 · 13 mins read

Without intending to, I ended up taking a blog hiatus (blacation?) for about a month. But in that month I built up a whole reserve of speculative energy that I now need to disgorge all at once. So strap in, because this one gets hairy towards the end.

A thing that has interested me since grad school is the clear gap in labor productivity between urban and rural areas, or between non-agricultural workers and agricultural workers. There are two big questions that come up given that fact. Why do those gaps exist? What effect do those gaps have, if any, on development? I’ve taken a couple of swipes at the second question here and here.

In those papers, I did have to think about the first question, why the gaps exist at all. It seems like if there were big labor productivity gaps, then there should be big wage gaps, and if there were big wage gaps, then people would have moved from the low-wage rural areas to high-wage urban areas already. One possibility is that those gaps exist because of differences in human capital of workers between the two sectors. If rural workers have less human capital, then their labor productivity (output per hour, say) will be lower than urban workers, even if the human capital productivity (output per unit of HC) is identical in rural and urban areas. In this case, the labor productivity gap may not represent any distortion or friction, as the return to a unit of human capital is identical across sectors, and therefore would not imply any kind of macro-level inefficiency.

And over the years, I’d say that more and more evidence accumulates that this is the case. Alwyn Young has a recent paper showing that there is a ton of movement back and forth between rural and urban areas in developing countries, and so it seems unlikely that some kind of friction, or moving cost, is keeping people from responding to any gaps in wages that might exist. Gollin, Lagakos, and Waugh find that gaps in human capital productivity are smaller than the gaps in labor productivity, although there is still a gap. Schoellman and Herrendorf look at raw wage gaps - which should be related to labor productivity - across sectors in a number of countries, and find that they are explained almost entirely by differences in human capital.

The one issue with these studies is that they (and me in my earlier papers) are comparing two different groups of people: rural workers and urban workers. There may be unobservable differences in their human capital that we cannot account for using things like education or experience, and so we don’t really know if our measures of human capital productivity are accurate.

Which brings me to a recent working paper by Hamory-Hicks, Kleemans, Li, and Miguel. The authors use longitudinal data from Indonesia and Kenya, meaning they can track the same individuals over time. This allows them to look at people who move from rural to urban areas, and hence have an identical set of human capital in both places, by definition. As their human capital is identical in both locations, if they earn more in one place than another then this indicates a difference in their human capital productivity, and not a difference in some unobservable human capital. So when people move from rural to urban, or urban to rural places in these countries, do their wages change?

In Indonesia, they do not. Once you are effectively comparing the same person in two different places (i.e. using individual fixed effects) the authors find a zero gap in wages. If they compare individuals who switch between agriculture and non-agriculture (as opposed to rural vs urban), there is also effectively no wage gap. The wages of individuals do not change when they move sector of employment or from rural areas to cities.

For Kenya, some small gaps remain, but they are small compared to the raw difference in wages between areas. There seems to be about a 13% wage premium to either living in an urban area or working in non-agriculture, even for an otherwise identical individual. That number, though, is not a statistically significant estimate, so it could be zero (or it could be a 20 or 30% gap, we don’t know).

The HHKLM paper is consistent with the earlier research, in that when you just look at the average wage gap (i.e. labor productivity) between urban and rural areas in Kenya and Indonesia, it is quite large (on the order of a 60-80% wage premium for urban/non-agricultural work), but when you control for human capital, it falls. The HHKLM paper is better at controlling for human capital (because they compare people to themselves), so it validates a lot of this earlier work.

The combination of facts tells you that there is selection out of rural/agricultural work and into urban/non-agricultural work for people with lots of human capital. There is not some distortion that prevents rural people from moving to higher wage positions, apparently, its just that all the really skilled or smart people move off the farm.

In Indonesia, the people who start rural and stay rural have the lowest levels of education. The people who start urban and stay urban have the highest levels of education. People who go either rural-to-urban or urban-to-rural have similar levels of education, and its about in the middle of the other two groups. The same holds for Kenya.

What’s really interesting is that this pattern shows up in the Raven’s Z-scores as well. If you read Garett Jones’ Hive Mind, you’ll know this is a crude, but effective, proxy for IQ. It’s a simple test that you can administer in the field, and correlates well with broader measures of intelligence (it’s a non-verbal pattern recognition test). So it’s not just that people who are lucky enough to get an education in an urban area stay there, and people unlucky enough to miss out on schooling in rural areas stay there. People with better measures of inherent smarts tend to end up in the city, or are in cities to begin with. HHKLM don’t have information on urban residents for Kenya, but the data for the rural stayers and leavers conforms to the Indonesian data.

[Everything that follows is all me riffing on the result in HHKLM. Don’t blame them if it doesn’t make sense.]

So urban places and/or non-agricultural work looks far more productive than rural/agricultural work because the people doing the urban/non-agricultural work appear to have more human capital, both inherent (Raven’s) and acquired (school). And by explaining the gap between these areas this way, it leads back to an concept that was often invoked in the early 20th century to explain the raw productivity gap: the peasant mentality.

As industrialization started, and these raw productivity gaps opened up for the first time, people started to speculate about why they existed if people could move back and forth relatively freely. Chayanov, in 1925, published The Theory of the Peasant Economy, which speculated about peasant farm organization. In it, he suggested that you cannot think of peasant families as acting according to typical economic logic (maximizing consumption or profits), rather they had a “peasant logic” or “peasant mentality”, which was focused on eking out enough to subsist year to year, but without much thought towards expanding, saving, or growing. One of James Scott’s earlier books, The Moral Economy of the Peasant provides a study of peasants in southeast Asia, how their motivations were driven by subsistence constraints, and paints a similar picture to Chayanov.

Both of those authors are sympathetic to the peasants. Others looked at the same situations and get frustrated with the peasants ignorance, stubborness, and unwillingness to experiment with new techniques or technologies. For some, it was the small-town peasant mentality that was holding back economic development. When Ted Schultz published Transformaing Traditional Agriculture in 1965, he attacked both interpretations of the peasant mentality.

First, he said, peasants were very much rational maximizers, just as economics would assume. They might have additional constraints (subsistence needs) and face more uncertainty unlike “normal” people, but they were maximizing their utility nonetheless with respect to those constraints. Schultz said we should be analyzing peasants with the tools of modern economics; we did not need different theories for them. That brought development into mainstream economics.

Second, in questioning the idea that peasants operated on some kind of different set of rules from other people, he attacked the idea that peasants were different at all. They were just like anyone else. The implicit assumption in Schultz’s work is that if you dropped an urban person into one of their villages, they’d make the same choices as the peasants. Peasants were not different, their environment was.

But what if they are different? That seems to be an implication of the HHKLM study. If people are sorting themselves between rural and urban areas, or between agricultural and non-agricultural work, based on some inherent characterstic captured by the Raven score or some other aspect of their human capital, then doesn’t that imply the people in the rural areas are different than the people in urban areas?

It doesn’t mean that rural people are not rational maximizers, but do they come to different conclusions about how to maximize, or what to maximize? If you look at the literature in Jones’ book, you’ll see that things like the Raven score are correlated with other characteristics like patience. Does the sorting of people across locations mean that in rural areas the people are less patient, and so they decide - rationally - to invest less? Are the things that frustrated outside observers of these peasant economies - unwillingness to experiment, ignorance - a result of the selection of low Raven-score, or low human capital individuals, into those areas? Were they right?

If your gut reaction is “No, or course not”, I get it. So was mine. But let’s not confuse a positive (i.e. factual) finding with a normative conclusion. Finding that rural residents score low on the Raven, and that people with low Raven scores select into rural areas, should change nothing about your opinion regarding interventions to aid development/health/etc in these areas. It doesn’t change Schultz’s conclusion that aggregate development probably requires a massive investment in those rural peasant villages. But the HHKLM data do suggest that the nature of those investments, and how you design the incentives to make them, might need to be different than in urban areas. Think of it like tailoring a class to match the learning styles of different kids.

If the HHKLM study is replicated in other contexts - which may be a big “if” - then perhaps we need to walk back some of the changes in thinking that Schultz initiated fifty years ago. Perhaps we should take seriously the idea that peasants are really different, not just in their constraints (which the development literature has explored in great depth), but in their underlying preferences as well (which has been done far less).