Monday, March 11, 2013

Look what the economists do with human diversity data

A colleague in Economics sent me this article:

Ashraf, Quamrul and Oded Galor. 2013. The 'Out of Africa' Hypothesis, Human Genetic Diversity, and Comparative Economic Development. American Economic Review 103:1-46.

Here's the abstract. It gives the jist of the article, if you can't get access:

This research advances and empirically establishes the hypothesis that, in the course of the prehistoric exodus of Homo sapiens out of Africa, variation in migratory distance to various settlements across the globe affected genetic diversity and has had a persistent hump-shaped effect on comparative economic development, reflecting the trade-off between the beneficial and the detrimental effects of diversity on productivity. While the low diversity of Native American populations and the high diversity of African populations have been detrimental for the development of these regions, the intermediate levels of diversity associated with European and Asian populations have been conducive for development. 
Anthropologists recognize this argument: yet another a pseudo-evolutionary craptasmagoria that purports to explain why European and Euro-American populations are inherently superior to African and Native American populations, thereby justifying global inequalities in economic/political power and negating any guilt that dominant societies may have about the legacy of colonialism and genocide. These arguments are always flawed, but it doesn't stop them from cropping up again and again. Or, to quote my new-favorite Jonathan Marks blog post"We have been here before; douchebags are always recruiting Darwin to their stupid views about human society."

The fundamental flaw in the argument is that there is no reasonable mechanism by which diversity affects productivity. Ashraf and Galor claim that some diversity is necessary for economic innovation, because a population's ability to create culturally innovative ideas is tied to its genetic diversity. I would argue our genetically-created ability for flexible behavior, which is shared by all humans, allows for innovation, but in this paper, fewer different alleles = fewer different ideas. Or something. 

Their argument against "too much" diversity is that highly genetically diverse populations are economically inefficient because people who are physically different aren't as willing to cooperate with each other.  (Kin selection was invoked in this context, including the foraging efficiency of groups of closely related spiders vs. spiders who are not related. It made my head hurt.) Of course, anthropologists understand that that physical variation is culturally categorized, as is our understanding of kinship, so genetic diversity and biological relatedness may have nothing to do with how different or similar you consider others in your community.

But even if you accept the mechanisms proposed by the authors, or consider them worth testing, the anthropological data that Ashraf and Galor use to show a hump-shaped relationship between genetic diversity and economic productivity do no such thing. The analysis is extremely problematic, both in the way it measures diversity and in the way it measures economic productivity.

The paper has two arguments, one related to modern economic productivity, and the other related to pre-industrial productivity. I'm only going to discuss the pre-industrial argument here because it's the topic I'm best qualified to tackle. The argument goes like this: prior to the industrial revolution, economic productivity is best measured by population density. These pre-modern economies were agriculturally-based, and surplus tended to be put into population growth. (The authors call this the "Malthusian epoch", a phrase I'm totally stealing.) Of course, in reality, surplus is put into many things, like building pyramids, or irrigation systems, or waging war. But I won't quibble on that point.

Ashraf and Galor set out to prove that the relationship between population density and genetic diversity is hump-shaped, suggesting that moderate amounts of diversity are ideal for population density (their proxy for economic development). This is shown in Figure 3 (p. 22), reproduced below. But before we accept this relationship, we have to ask two questions: Does this graph accurately reflect the diversity of the continents in 1500CE? And, does this graph accurately reflect the economic development of those continents?

Does this graph accurately show the diversity of Europeans, Africans, Asians, and Americans in 1500CE? 

Yes, and no. Ashref and Galor label the dots in Figure 3 with names of modern states. But the genetic data they are using is not generally representative of the people who lived within those borders. The data on genetic homogeneity (a measure of diversity) comes from the HGDP-CEPH Human Genome Diversity Cell Line Panel, as published by Ramachandran et al. (2005). Ramachandran et al., to simplify, showed that genetic diversity is highest in Africa and gets lower the farther you go from the origins of our species. This is clearly true. But Ramachandran and colleagues focused on genetic diversity in isolated communities, because they were interested in the earliest people to enter these regions. But precisely because these samples were picked from isolated communities, they don't say much about diversity in those nations overall at 1500 CE. 

For example, the UK is represented in this sample by one ethnic group: Orcadian Islanders. Now, the Orkney Islands are not without their history of migration and admixture - they've been passed between Scotland and Norway a few times - but this one small population at the upper ends of nowhere can hardly be considered representative of the genetic diversity of the United Kingdom as a whole in 1500 CE, with its history of Roman, Anglo-Saxon, Viking, and Norman influx. Similarly, the nation of Zaire, which today has some 250 ethnic groups, is represented in this data by a single community of Mbuti Pygmies. Most problematic, the entirety of the Americas is represented by only four data points, and the far-right anchor comes from the Karitiana people of Brazil, an indigenous group with only 350 members!

By 1500 CE, there had been a number of major migrations that spread genes across the most populous (that is, least isolated) areas of the world. An estimated 8% of men in Asia carry a Y chromosome believed to come from Ghengis Khan (Zerjal et al. 2003). Cosmopolitan cities existed on all four continents, and the genetic diversity in those cities would have been higher than that in the isolated societies sampled for Ramachandran et al.'s analysis. The city of Teotihuacan in central Mexico, for example, had one neighborhood consisting of Oaxacan merchants. The Crusades led to frequent trade between Europe, North Africa, and western Asia.

By using the genetic diversity of isolated populations to represent the genetic diversity of large states, Ashref and Galor are failing to consider the genetic diversity of the people who were most involved in highly-developed economic systems.

Does this graph accurately show the economic development of Africa, Asia, Europe, and the Americas in 1500CE?

This answer is easier: no. 

First, as mentioned above, the genetic samples tell us about the diversity of small isolated populations, not the diversity of those states as a whole at 1500 CE (or of the people within the modern borders of those states). But the population density estimates (which come from McEvedy and Jones, 1978) are for the states as a whole. To test Ashref and Galor's hypothesis, the genetic diversity of the population should be compared to the estimated population density of those communities (Orcadian genetic diversity compared to the population density of the Orkney Islands, for example). This would be difficult, but not impossible, with archaeological data. It would also utterly destroy the pattern in Figure 3, since the genetic samples from all four continents include some societies that were highly mobile with low population densities (the San, the Bedouin, etc.) and others that were highly sedentary and lived in cities (Japanese, Mayan etc.). 

Second, the population density estimates are problematic. I haven't gotten a copy of McEvedy and Jones (1978) yet, but looking at the data used by Ashref and Galor, the population estimates for the Americas, at a minimum, are too low. There is a long history of underestimating the indigenous population of the Americas, partly because early European records justified colonization efforts on the basis of "empty" land for the taking, and partly because accurate estimates of population at Contact would show such high death rates for the indigenous population that many Euro-descended researchers have shied away from suggesting or accepting them. The data from Brazil is pulling the curve in Figure 3 downward on the right side, a major determinant of its shape. Since McEvedy and Jones's publication in 1978, our understanding of pre-Contact Brazilian history has changed radically. We now know the population density of the Amazon region was far higher than previously thought (for example, Heckenberger et al. 2003). Moving the American data points upward would radically change the shape of the curve. I can't speak to the African population density data, but any errors in the estimates for South Africa (another area where European colonists denied the existence of previous occupants) would also flatten the curve.

A major problem with using population densities as a proxy for economic development, is that by using the population densities within modern national borders, Ashref and Galor are significantly underestimating pre-modern population density in highly-developed areas by lumping them with low-productivity areas. This is less of a problem in Europe, where modern nations tend to be smaller (although it explains Russia's very low density relative to other European nations, or China's relative to Japan), but is more problematic in the New World. Mexico, for example, includes not just the highly populous regions of the Valley of Mexico, but also the low-density regions of the Sonoran and Chihuahuan Deserts, and the Sierra Madres. A more reasonable comparison would be the population density of the UK vs. that of the Aztec Empire, or of the Mali Empire.

Why do we care?

Ashref and Galor argue that biological traits of populations determine (or at least affect) their level of economic prosperity. In doing so, they add their work to the long history of attempts to prove inherent racial differences lead to inequality, rather than inequality being the result of systemic attempts to deny opportunity to others. And they did so in one of the most prestigious journals in their field.

I've never met Ashref or Galor, and I know nothing about them. I have no idea if they harbor personal biases toward any group of people. They may both be saints, dedicating their life to social justice and the pursuit of equality. It doesn't matter. Given the salience of race in Western society, any argument for biological determinism must be held to a high standard of evidence, and the authors must accept responsibility for the political and social implications of their work. They cannot pretend it doesn't matter, or that they're not making an inherently political argument. 

 They walked into a crowded theater and yelled "Fire!" They are responsible for the consequences, whether they did so with malicious intent or not. To yell "Fire!" on the basis of  a manufactured hint of smoke is irresponsible at best, at worst, it is criminal.

I'll leave you with a final quote, this one from Charles Darwin, whose intellectual legacy was so poorly used by Ashref and Galor: “If the misery of the poor be caused not by the laws of nature, but by our institutions, great is our sin.” 

Admitting sin is hard, but blaming the victim is unconscionable


Heckenberger, Michael J., Afukaka Kulkuro, Urissapa Tabata Kulkuro, J. Christian Russell, Morgan Schmidt, Carlos Fausto, Bruna Franchetto. 2003. Amazonia 1492: Pristine Forest of Cultural Parkland? Science 301: 1710-1713.

McEvedy, Colin, and Richard Jones. 1978. Atlas of World Population History. New York: Penguin Books.

Ramachandran, Sohini, Omkar Deshpande, Charles C. Roseman, Noah A. Rosenberg, Marcus W. Feldman, and L. Lucas Cavalli-Sforza. 
2005. Support from the Relationship of Genetic and Geographic Distance in Human Populations for a Serial Founder Effect Originating in Africa. PNAS 102 (44): 15942-15947.

Zerjal, Tatiana et al. 
2003. The Genetic Legacy of the Mongols. American Journal of Human Genetics 72:717-721.

Update: Turns out the anthropologists at Harvard are on this! Check out this open-access article in Current Anthropology reviewing the article.

Guedes, Jade d’Alpoim and others. 
2013  Is Poverty in Our Genes? A Critique of Ashraf and Galor, “The ‘Out of Africa’ Hypothesis, Human Genetic Diversity, and Comparative Economic Development,” American Economic Review (Forthcoming). Current Anthropology 54: 

No comments:

Post a Comment