The deep roots of development - Part 1


This coming semester I’m teaching a new version of my graduate course in growth and development. I’ve always taught a course that looks at long-run growth, but in the past that has involved at times a lot of theory (e.g. unified growth models) along with a close look at the literature on institutions as the origin of the take-off embedded in that theory. Some of the most popular posts I’ve every done on this blog were the “Skeptics Guide to Institutions”, which were in part lecture notes for that course. I also folded in material on the transition out of agriculture, something that I find an important part of this transition.

But that version of the course was always a little muddled in focus, and so for this semester my aim is to step back and start over. The plan is to take a more coherent look at the literature on “deep roots” or “comparative development”, which I’ll define in more detail below. As part of that, my plan is to offer a few lectures during the semester that give an overview of areas within that literature, and a bit of their intellectual history, to frame the specific papers that I’m going to have students dig into and the data work I’m going to have them do.

This post is meant to be a rough draft of the first of those overview lectures, introducing the field and the major ideas and themes we’ll be covering. The title of the post includes “Part 1”, which is intended as a commitment device. Part 2 should (no, will!) be about the evidence of persistence in economic outcomes over history, and the spatial pattern of development. Part 3 will be about institutions and colonization, Part 4 on culture and family structure, Part 5 on agricultural conditions, and Part 6 on cultural and genetic diversity.

Timing-wise, those posts are likely to dribble out over the next few months, ideally well ahead of the scheduled date in class, but more realistically in some panicked burst of activity the day before. Once I get the final syllabus worked up, I’ll post that as well, in case people are interested in the bibliography.

What is the deep roots or comparative development literature?

As with any literature, there is no crystal clear definition. But I’d say that one characteristic common to any paper in this literature is that the underlying question is “Why are some places rich and some places poor?”. A second characteristic is that the answer involves deep-rooted characteristics of populations, rather than things like trade policy, macroeconomic stability, research subsidies, or social safety nets.

To be more concrete, I think there are several key stylized facts or concepts that this line of research has either established or takes as given. Most people in working in this area would agree with most of the following:

  1. Differences in development are about differences in populations, not countries. It is not that France, Germany, and the U.S. are rich, it is that places with people descended from ancient European populations tend to be rich. Conversely, places populated with descendants of ancient African populations tend to be poor. Note that this is not a statement about genetics. It is a statement about where the ancestors of current populations lived in the past.

  2. These populations differ substantially in not just the proximate sources of development (e.g. capital stocks and productivity levels) but in more fundamental characteristics like culture and institutions. Attitudes towards trust, patience, importance of family, and the like all show significant variation across populations. The assumption is that these characteristics drive the proximate sources of development, and some evidence is available to back this up.

  3. Differences in the cultural and institutional characteristics of population arose early. By early, I mean well before industrialization, conceivably back to the origins of settled agriculture in 10,000 BCE, and perhaps even before that. This does not necessarily mean that these cultural and institutional differences created differences in living standards in the past; the effect on development may have been latent.

  4. There are differences in the geographic, biological, agricultural, and climatic conditions of the ancestors of current populations. Furthermore, cultural and institutional characteristics are correlated with these conditions (things like variability in rainfall, the presence of frost, the durability of stored food, the disease environment, access to the ocean, and so on). Whether those environmental conditions caused (some of) the differences in culture and institutions is not as clear, but evidence is accumulating that they did.

  5. Things like culture and institutions are incredibly persistent. Even once the underlying source of differences in those things disappears (e.g. immigrants with different ancestral environmental histories come to the same country) their cultural and institutional differences remain, at least for several generations.

  6. While characteristics are persistent, specific historical events (e.g. the slave trade) appear capable of changing them demonstrably. Therefore these events can change the development possibilities of populations. Which events rise to this significance is an empirical question, but colonization is a common origin.

The work I’ll cover in this course can usually be framed as speaking to one or more of these points. But taking them together, we can stitch together a rough story of why some places are rich and some are poor. First, we should rephrase that as “Why are European-descended populations so rich relative to the rest of the world?”. Second, we could answer it this way. They are rich because the environment they originated in favored the spread of cultural and institutional characteristics (e.g. trust in strangers, patience) that were conducive to economic growth. Given that advantage, they were able to avoid (and instead impose) significant negative historical shocks on the rest of the world that arrested or rolled back their development, giving rise to the observed difference in living standards today.

Institutions versus geography?

The words “institutions” and “geography” both show up in the points I just laid out. And that recalls a long-running debate on the relative importance of each. That debate, I think, has played itself out. Not because one side won and another lost, but because we appreciated that it was a stupid debate in the first place. There is nothing mutually exclusive about institutions and geography (and culture, for that matter) as determinants of development. Rather, both matter and interact with one another to create persistence differences.

Let me illustrate what I mean by comparing two excellent papers in this field. The first is Melissa Dell’s paper on the mining mita in Peru. The mita was a forced labor institution that required natives within a specific area to provide work in the Potosi silver mine for Spain. Dell established in her paper that areas today that were once inside the mita have lower development levels (on several measures) than places that were once outside the mita.

The second paper is Marcella Alsan’s paper on the effect of the TseTse fly on African development. She builds a measure of the natural geographic range of the TseTse fly, based on temperature and rainfall. As the TseTse feed mainly on cattle, areas with the fly are less likely to use cattle in agriculture. Alsan shows that across ethnic groups, places exposed to TseTse flies did use fewer cattle, had less of a role for women in agriculture, and were less dense, among other things.

On the surface, Dell’s paper shows the importance of an institution, while Alsan’s shows the importance of geography. And it would be ridiculous to think that one somehow invalidates the findings of the other. Both, though, show that aspects of development are persistently affected by deep roots, in a broad sense. In Dell’s case, there is a specific historical event - the mita - that continues to cast a shadow on development within Peru. Note, however, that her finding is not on the effect of that institution, per se, as the mita ceased to exist centuries ago. Rather, she shows that historical shocks have persistent effects. Further, you can easily read Dell’s paper as proof that geography matters, because the location of the mita was driven by the location of the Potosi mine. Those places are poor because geographically they happened to be close to a major silver lode.

In Alsan, the deep root is the range of the TseTse fly, which affected how ethnic groups within Africa subsisted, with effects on the role of women and type of agriculture. Again, underlying conditions had persistent effects. Further, she shows that the TseTse itself has no direct effect on development (measured by night lights) once you control for ethnic group centralization prior to colonization. One possibility is that the TseTse - through changing density and other characteristics - influenced centralization, and hence influenced development indirectly.

The point is that in neither case is there a bright line between institutions and geography (or culture). Both papers show the deep-rooted elements - in one case a historical institution, in another the geographic range of a pest - have detectable effects on development that can persist up until the present day. Both papers provided evidence that helped build the concepts in points 1-6 above, but neither invalidates the other.

Empirical methodology and data

Aside from the concepts I listed, perhaps the other defining characteristic of this deep roots literature is the methodology used. It is an almost entirely empirical literature. Theory, when it is used, is illustrative rather than structural. Yes, it is easy to see the deep roots literature as an outgrowth of unified growth theory (UGT), which was much more theory-based. Oded Galor, the originator of UGT, is a central figure in the deep roots literature. But to my point, the theory embedded in his papers today is far more used to frame some simple hypotheses than to model history. If you want to think about how this deep roots literature fits with UGT, think of it as trying to figure out the specific conditions that determine why some places hit the take-off point, and when, rather than figuring out the specific mechanics of how a take-off works.

Methodologically, the deep roots literature has ingested much of the logic of the causal revolution in economics, and combined that with two novel sources. The first source is extenseive GIS data, including night lights as well as datasets on inherent agricultural characteristics. This has meant the ability to do sub-country level empirical work at the district or pixel level, massively increasing the power of studies and allowing for better identification by being able to use fixed effects for larger political units and/or doing spatial regression discontinuities.

The second source(s) are bringing what I’ll call cultural datasets into the literature. One prominent one you’ll see is the Ethnographic Atlas, a compendium of anthropological information that was compiled from ethnic-group level studies, and coded into a common set of indicators for different cultural characteristics. For example, Alsan’s paper uses this to look at which groups within Africa relied mainly on cattle, and which had a significant role for women in agriculture, for example. Another is the World Values Survey (among others) that takes contemporary measures of cultural attitudes on things like trust, family, and the like. In both cases, further work has been done to link these cultural measures to maps, so that one can combine the geographic data with the cultural data, often at a sub-country level.

Beyond these, researchers in this area have brought data in on genetic diversity of populations, language relationships across groups, historical city populations, and so on, in order to make their case. One of the prominent features of this literature is an explosion in datasets being brought to bear on relevant questions.

While welcome, there are potential issues with this expansion of data. The origins of the data are not always well understood, as one might expect within the field of economic history, for example. The Ethnographic Atlas, for example, is taken as given. It was coded mainly by a single anthropologist working from studies taken from over about eighty years. One can easily imagine that you could teach an entire class on a single one of the 1200-odd ethinc groups in the dataset, and debate the nuances of the family structure, gender roles, and taboos of that group. The idea of coding that nuance into, say, a five point scale for whether a group is “patriarchal” probably makes most anthropologists break out into hives.

The geographic datasets utilized also tend to be black boxes. A common source is something called the Global Agro-ecological Zones (GAEZ) project. This provides a pixel-level map of the world with numbers of the potential yield of that pixel for oats, for example. How do they arrive at that potential yield? Well, that depends on estimates of rainfall, sunlight, and soil type in that pixel, combined with plant-level parameters for needed water, length of growing period, and the like. What assumptions about how the plant is cultivated are built into the program the GAEZ uses to make that estimate?

You should remain skeptical - in the academic sense - of the work built on these kinds of datasets. I don’t think we have a good feel for how sensitive the existing results are to plausible variation in these sources. That is, one could easily come up with an alternative Ethnographic Atlas based on a new set of anthropologists coding the same source material, or come up with a new GAEZ based on a different set of assumptions from agronomists. Do those line up closely with the existing versions? If so, great. That might give us some assurance that the results we review are robust. But if not, things are very murky. The deep roots results could well fall apart with a slightly different version of the Atlas or GAEZ.

Determinism

Leaving aside the questions about data quality, this literature seems to imply some strong versions of determinism. That is, were the places that are poor today doomed to be poor because of some historic event 500 years ago, or because of geographic conditions present 10,000 years ago?

Here I think it helps to think statistically, rather than in absolutes. And here I do not mean thinking about the uncertainty associated with point estimates of effect sizes, but rather the fact that in no case will any study show you an R-squared of one, meaning perfect prediction. To use a more concrete example, it is plausible that ethnic groups in areas where there were large returns to long-run investment in agricultural production may end up more patient in the long-run as well (as more patient people were selected for). There is a significant relationship of productivity with patience in the data.

There is a significant relationship, but the R-squared is about 10%. That means that within the existing data, it is quite plausible to find pairs of ethnic groups that have different patience levels today even though they had similar productivity levels, or pairs with the same patience, but with different productivity. That’s all consistent with a statistically significant result and an R-squared of 10%.

More to the point on determinism, the results don’t mean that ethnic group A with higher productivity had to have higher patience than group B with lower productivity. Imagine bootstrapping world history 10,000 times, starting over at like 100,000 BCE every time. The significant relationship of productivity and patience means that we’d expect patience to be higher in A than in B in more than half of those runs, but the R-squared of 10% it wouldn’t happen in all of them. It might be only that in 550,000 of the run does A have more patience than B, meaning B is more patient in 450,000 alternative histories. And yes, the positive relationship of productivity and patience could still be true even in those 450,000 histories because the patience of groups C, D, E, etc.. are all changing each time as well.

The fact that the explanatory variable here, agricultual productivity, is taken to be plausibly exogenous to the patience of the groups is immaterial. Lots of other things affect the patience of an ethnic group over and above the productivity level, and hence lots of things could have changed over the course of history to explain patience.

Acknowledging that geography (or biology, or agro-climatic characteristics) are significant for some cultural trait, institution, or level of development does not imply a deterministic relationship exists, and that history is locked in no matter what. R-squares are not 100%!

The findings in this literature do not mean it is impossible to change the development level of countries or groups today. But the deep-rooted nature of development may mean we have to find other policy levers to push, as we cannot go back in time and roll back colonization, or change inherent agricultural productivity, and so forth. This isn’t that different than what you’d find in an applied micro course on, say, education. There you might study the effects of a program to raise graduation rates, and see whether it is effective or not. But note that graduation rates in some schools were not low to begin with because of the lack of this program, they were low because of (probably) poor socio-economic backgrounds, a lack of resources, poor teachers, etc. Even in these worlds we spend a lot of time pushing other policy levers than the ones that created the situation in the first place.

Going forward

Unlike a lot of courses, this material will at times be breathtaking in the questions it asks and answers it proposes. It will seem presumptuous to explain cross-cultural patience as a function of agricultural productivity around the time of the Neolithic Revolution. And I think that researchers within this field do have a tendency to take big swings at big questions, and at times the prose in the papers will reflect that. Don’t let that put you off. The question still remains whether there is a valid empirical link established, and whether that link can be interpreted causally or just as a correlation. I think you’ll see that there is an accumulation of papers establishing strong links between deep roots and current development, and that this is not just a few odd results.

As mentioned above, what I’m hoping to do from here forward (on the blog) is post subsequent parts with material on persistence, institutions and colonization, culture and families, agricultural productivity, and diversity. In the class itself, we’ll be covering specific papers within each section, and I’ll be trying to get my students up to speed on using R’s spatial packages to bring in some of the datasets used and replicating some basic results.

There is a topic page up on my site already with all the existing posts I’ve classified as related to “deep roots”, so if you’re into this material feel free to have a poke around that.

Back to blog

Share and Discuss