A Replication in Economics: Does “Genetic Distance” to the US Predict Development?

Douglas Campbell writes:

new study finding that more than half of psychology studies failed to replicate is a very positive step forward for social science. Could a similar study be undertaken in economics, and what would it find? Most empirical economics research is non-experimental, and thus I suspect that most studies would replicate in the sense that if one used the same data and ran the exact same regressions, the results are unlikely to change. However, if one were also to test the robustness of results to additional (or fewer) control variables, differing estimation approaches, or try out-of-sample testing on new data, I suspect less than half would survive. When I was a graduate student, I became frustrated with constantly being assigned to read papers which I felt were clearly wrong. I suspected that the real key to publication for many of these papers was perhaps the right pedigree and a close relationship between theauthors and  the editor and/or referees. While I knew the conventional wisdom that it’s a bad idea to write “comment papers” in economics, eventually I became curious what would happen if I tried to take down a “seminal” paper published in a top journal.

If you’ve read this blog long enough, you’ll know that I’m sympathetic to Campbell’s argument.  Not that this means he’s correct (or, for that matter, that he’s wrong), I’m just letting you know where I stand, where my preconceptions are.

Campbell continues:

One paper I had been assigned to read in several graduate courses, on “The Diffusion of Development,” published in the QJE, a leading economics journal published at Harvard, argued that there is a causal link between a society’s average skull size “genetic distance to the US” and its GDP per capita. The authors were careful to point out that their results didn’t necessarily indicate a direct impact of genetic traits on economic development, but that genetic distance could be a proxy for a whole host of other cultural traits which could impact the transmission of technology. However, in my view this point was undercut by the authors’ assertion that the apparent impact of genetic distance on GDP per capita survives the inclusion of an ostensibly exhaustive list of geographic and cultural controls. This suggests that genetic distance may not merely proxy differences in cultural traits, but has a direct impact on GDP. Thus black Africa may be poor because of its genetic endowment, and white Europe rich for the same reason.

Except, …, wait a second here before we go leaping to conclusions.

Uh oh.  These sorts of arguments (not Campbell’s, but the ones he’s criticizing) are the kind of thing I hate!  See here and here for some examples.  My problems with these arguments are:  (a) correlations being what they are, these studies are typically based on essentially 2 or 3 or 4 data points, not 91 or 132 or whatever is claimed based on the number of countries in the dataset, and (b) I’m suspicious of the whole GDP-as-revealed-virtue thing, given how time-bound such arguments are.

Campbell might agree with me on this, or maybe not, but in any case he has some very specific criticisms:

How exhaustive were those geographic and cultural controls? A coauthor, Ju Hyun Pyun and I noticed that the authors did not even control for latitude or for a dummy for sub-Saharan Africa in their cross-country income regressions, even as they argued that their results were robust to controls for geographic regions. When we included these controls (standard in this literature) in the first regression we ran, the correlation between genetic distance to the US and development disappeared. We also found that genetic distance to the US failed to predict income levels even when we just included two dummy variables, one for Europe and one for sub-Saharan Africa, with no other controls. Thus, the original findings were equivalent to the observation that white Europe is rich and black Africa is poor, with no more explanatory power than that. While we felt our results were perfectly straightforward, it took us seven submissions and four years to publish our results in a minor journal. Meanwhile, results similar to those we had critiqued continued to be published in leading journals, including one of the same journals where our paper was rejected. We often had to contend with the original authors as referees – once as the sole referee. (Pro tip: if writing a paper like this, recommend to the editor that they not choose the hostile original authors as referees.) One editor sided with a creative referee who objected to our paper on the grounds that “There is no reason to interpret the sub-Saharan dummy as a ‘geographic variable’”. The same referee also zinged us for not including the exact same sample as the original paper, even though the data and original sample were not publicly (or privately) available. This was hardly the type of hassle-to-reward ratio which would lead me to write a similar such paper, at least before tenure. Instead, had we decided to write an “extension” paper, using the genetic distance data to predict some other variable, publication would have been facilitated, since the original authors would have been likely referees and would have been happy to see our results published. The incentive structure here could be improved.

Campbell also writes:

In defense of the authors, the paper itself was at least an interesting idea, and they deserve credit for trying to tackle a sensitive topic such as the link between genetics and development. They are certainly not alone in publishing papers that turned out not to be robust (this should happen to any ambitious researcher), particularly so for research using spatial data, notorious for spurious correlation. Empirical researchers are under a lot of pressure to find statistically significant results that are seemingly robust.

And he concludes with some general comments about replication:

Academic economics dearly needs replication studies to become sexy. There are encouraging signs. Thomas Herndon became famous after catching Reinhart and Rogoff’s excel error. There is a new economics replication wiki, and there will be a panel on replication at the AEA meetings. Some journals, such as the AER, require data to be made available online. (Personally, I believe doing empirical research without posting your data online, at least after publication, should be taboo.) Yet, in a field where building close personal relationships is still easily the best path toward publishing and tenure, more needs to be done. One proposal is for someone to calculate new journal rankings which penalize journals which either do not accept comments on the papers they publish, or rarely publish such comments. It would also be helpful if editors, particularly at leading journals, which have substantial market power, would do more to encourage replication. One proposal is that if the QJE or AER wants to encourage comment papers without hurting their own citation ranking, they could start additional journals focused on replication.

P.S. Lots of discussion in comments, including a response by Enrico Spolaore and Romain Wacziarg that begins, ‘Campbell and Pyun’s paper is a completely misguided criticism of our paper “The Diffusion of Development,” published in the Quarterly Journal of Economics in 2009.’

73 thoughts on “A Replication in Economics: Does “Genetic Distance” to the US Predict Development?

  1. Doug Campbell writes: “Most empirical economics research is non-experimental, and thus I suspect that most studies would replicate in the sense that if one used the same data and ran the exact same regressions, the results are unlikely to change.” It has long been well-known that economics research is generally not replicable. I suggest he acquaint himself with:

    (1) Replication in Empirical Economics: The Journal of Money, Credit and Banking Project
    William G Dewald, Jerry Thursby and Richard Anderson
    American Economic Review, 1986, vol. 76, issue 4, pages 587-603

    (2) Replication And Scientific Standards In Applied Economics A Decade After The Journal
    Of Money, Credit And Banking Project” Richard Anderson and William G. Dewald,
    Federal Reserve Bank of St. Louis Review, November/December 1994, Vol. 76, No. 6, pp. 79-83

    (3) [36] B. D. McCullough and H. D. Vinod
    “Verifying the Solution from a Nonlinear Solver: A Case Study,”
    Aerican Economic Review 93(3), 873-892, 2003
    with comments and replies at American Economic Review 94(1), 2004

    (4) B. D. McCullough, Kerry Anne McGeary and Teresa Harrison
    “Lessons from the JMCB Archive,”
    Journal of Money, Credit and Banking 38(4), 1093-1107, 2006

    (5) B. D. McCullough, Kerry Anne McGeary and Teresa D. Harrison
    “Do Economics Journal Archives Promote Replicable Research?”
    Canadian Journal of Economics 41(4), 1406-1420, 2008

    (6) Is Economics Research Replicable? Sixty Published Papers
    from Thirteen Journals Say “Usually Not”, by Andrew C. Chang
    and Phillip Li, Finance and Economics Discussion Series 2015-083.
    Washington: Board of Governors of the Federal Reserve System

  2. Doug Campbell writes: “Most empirical economics research is non-experimental, and thus I suspect that most studies would replicate in the sense that if one used the same data and ran the exact same regressions, the results are unlikely to change.” It has long been well-known that economics research is generally not replicable. I suggest he acquaint himself with:

    (1) Replication in Empirical Economics: The Journal of Money, Credit and Banking Project
    William G Dewald, Jerry Thursby and Richard Anderson
    American Economic Review, 1986, vol. 76, issue 4, pages 587-603

    (2) Replication And Scientific Standards In Applied Economics A Decade After The Journal
    Of Money, Credit And Banking Project” Richard Anderson and William G. Dewald,
    Federal Reserve Bank of St. Louis Review, November/December 1994, Vol. 76, No. 6, pp. 79-83

    (3) [36] B. D. McCullough and H. D. Vinod
    “Verifying the Solution from a Nonlinear Solver: A Case Study,”
    Aerican Economic Review 93(3), 873-892, 2003
    with comments and replies at American Economic Review 94(1), 2004

    (4) B. D. McCullough, Kerry Anne McGeary and Teresa Harrison
    “Lessons from the JMCB Archive,”
    Journal of Money, Credit and Banking 38(4), 1093-1107, 2006

    (5) B. D. McCullough, Kerry Anne McGeary and Teresa D. Harrison
    “Do Economics Journal Archives Promote Replicable Research?”
    Canadian Journal of Economics 41(4), 1406-1420, 2008

    (6) Is Economics Research Replicable? Sixty Published Papers
    from Thirteen Journals Say “Usually Not”, by Andrew C. Chang
    and Phillip Li, Finance and Economics Discussion Series 2015-083.
    Washington: Board of Governors of the Federal Reserve System

  3. Genetic distance is a pretty innocent measure, constructed as it from ‘junk’ DNA with no known phenotypical value. So the comparison with skull sizes is simply ignorant. Set aside the statistical issues, and the comparison with Wade and Ashraf-Galor is also inappropriate. After all, Spolaore & Wacziarg’s pre,ise is that genetic distance is a proxy for cultural distance.

    • “Genetic distance is a pretty innocent measure” seems to slip a lot under the rug — namely that there are many possible ways of calculating “genetic distance.”

      First, the Wikipedia article on genetic distance points out several ways of calculating genetic distance, assuming that “L loci have been studied. However, it doesn’t mention that the the loci studied may give quite different results.

      But the choice of loci can give quite varying versions of “genetic distance” — see, for example, Box 1 (p. 600) of Jobling, MA and Tyler-Smith C (2003), The human Y chromosome: an evolutionary marker comes of age, Nature Review Genetics, 4(8):598-612 (http://www.nature.com/nrg/journal/v4/n8/full/nrg1124.html). This points out that a study using a microchip based on genes known to vary in European populations may miss variation between European and Asian populations and
      between different Asian populations.

      So I urge extreme caution in using “genetic distance” as a proxy for anything, especially if you don’t know which loci are used or how they were chosen.

    • I had an interesting exchange with Dr. Spolaore in 2013:

      In the comments to Alex Tabarrok’s post at Marginal Revolution, I ask:

      Steve Sailer October 7, 2013 at 4:46 pm
      It would be interesting to compare genetic distance to language distance. Does genetic distance provide us any retrospective predictive power that language distance does not? The two measures correlate positively, but there are a number of interesting test cases where they are strikingly divergent, such as Hungary, Finland, and Basque country.

      And Dr. Spolaore, one of the co-authors, replies:
      Enrico Spolaore October 7, 2013 at 7:05 pm
      Good question. In our article “The Diffusion of Development” (QJE, May 2009), Romain Wacziarg and I discuss the relation between genetic distance and linguistic distance, and study the effect of relative genetic distance on income differences when controlling for measures of linguistic distance (and religious distance) (pp. 504-514). We conclude (p. 512): “In summary, using the best available measures of linguistic and religious distance, the effect of genetic distance on income differences is reduced by about 12%, but the effect remains large and significant. Overall, these results are consistent with our interpretation: when we measure some specific differences in vertically transmitted traits, such as in language or religion, we obtain a reduction in the size of the coefficient on genetic distance, suggesting that genetic distance was capturing some of the barrier effects associated with differences in these vertical characteristics. However, the reduction is not large enough to suggest that genetic distance only captures the effect of linguistic and religious distance. On the contrary, the reduction is relatively modest, and the effect of genetic distance remains large and significant even when controlling for linguistic and religious distance. This suggests that language and religion are but two of the many vertical characteristics that differ across populations, and perhaps not the most important barriers to the diffusion of economic development.”

  4. Campbell’s argument is almost assuredly right, but looking at the choropleth map of genetic distance in the appendix, it seems like controlling for sub-saharan africa is also controlling for the variable of interest, genetic distance from the US, to a large extent.

  5. “However, in my view this point was undercut by the authors’ assertion that the apparent impact of genetic distance on GDP per capita survives the inclusion of an ostensibly exhaustive list of geographic and cultural controls. This suggests that genetic distance may not merely proxy differences in cultural traits, but has a direct impact on GDP. Thus black Africa may be poor because of its genetic endowment, and white Europe rich for the same reason”

    A very tendentious and really a libellous reading of Spolaore and Wacziarg. They are arguing that religious and linguistic differences are not the extent of the cultural differences, which seems an unexceptional thing to say !

  6. Good stuff, but I don’t understand why a replication paper that claims to debunk a famous paper should be automatically be believed. Maybe the replicators are correct, maybe not. It’s an open question, that needs third party verification.

  7. Campbell is broadly correct in the sense that specification matters, indeed, that was the purpose of Ed Leamer’s important book Specification Searches. See:
    http://www.anderson.ucla.edu/faculty/edward.leamer/books/specification_searches/specification_searches.htm

    Also his 1983 paper:
    http://www.econ.ucla.edu/workingpapers/wp239.pdf

    To some extent, the profession has listened, and papers typically have to make a strong case that the results survive myriad alternative specifications and robustness checks, hence the 4, 5, ,or 6 rounds of revision required by most top Econ and finance journals.

    A different issue that McCullough has investigated (see above his comment to this post) has to do with “operational error”, either human error in programming or software itself. I believe the latter issue has been unjustly neglected.

  8. There’s almost certainly a causal link between a society’s average skull size and its GDP per capita. It’s not that larger skulls cause prosperity, but larger skulls mean larger, healthier, and better brains (and therefore, for example, higher IQs) which in turn are linked to higher economic productivity. The causal link can be both environmental and genetic, and is probably both.

    • It seems you could also argue that prosperity leads to larger, healthier, and better brains. Though it could be argued that IQ is not the appropriate statistical index for such an assessment (see Weinberger, 2015 for a good exposition of this). Also, it seems likely that genetic influences could further be related to gene expression which is impacted by environmental variables as well.

      • I argue that both genetic differences and prosperity cause larger, healthier, and better brains, which in turn cause greater prosperity, and so on.

        By Weinberger (2015), do you mean this paper? I read it, but I don’t think it has any relevance here. In any case, for my argument to hold, it doesn’t matter how we measure intelligence or other indicators of “better brains.”

    • Speaking of failed replications and skull sizes, Stephen Jay Gould’s famous evidence-free assertion in “The Mismeasure of Man” that a 19th century study of skull sizes would not replicate itself failed to replicate. From a 2011 New York Times editorial:

      Bias and the Beholder
      JUNE 14, 2011

      Stephen Jay Gould, a prominent evolutionary biologist, gained broad public attention for exposing how scientists’ biases can skew their research. In one celebrated case, he charged that a famous study of human skulls in the mid-19th century had been manipulated, probably unconsciously, to support racist notions.

      The skulls had been collected by Samuel George Morton, a physical anthropologist. He measured their cranial capacity by filling them with seeds and later with lead shot. Caucasians had the largest brain volume, followed by Asians, with American Indians and Africans trailing.

      Dr. Gould, who died in 2002, re-analyzed Morton’s results and concluded that he had selectively reported data and manipulated subgroups to fit a preconception that Caucasians had bigger brains than Africans and were, therefore, more intelligent. Dr. Gould found no important differences among the races. He did not measure the skulls himself.

      Now a team of six physical anthropologists has filled almost half the skulls with pellets and concluded that Morton’s data were generally reliable and not manipulated. Although the team acknowledged that Morton often reported results in a “highly racist fashion,” in this case it found no evidence that Morton believed brain size was a measure of intelligence or was trying to prove it.

      The team expressed admiration for Dr. Gould’s body of work in staunch opposition to racism, but, in this case, it accused him of various errors and manipulations that supported his own hypothesis. “Ironically, Gould’s own analysis of Morton is likely the stronger example of a bias influencing results,” the team said. We wish Dr. Gould were here to defend himself. Right now it looks as though he proved his point, just not as he intended.

      http://www.nytimes.com/2011/06/15/opinion/15wed4.html

      • I followed the link given by Steve to get to the paper by the team of six physical anthropologists — largely because I wondered how the “almost half the skulls” they remeasured were chosen. I appears that they were what I have heard called an “opportunistic sample”: namely, those skulls from Morton’s original study that the team was able to locate. I’m not faulting the team for having a non-random sample — accessibility to specimens is an inherent problem in the field. However, since the sample is not a random sample, it can’t really be used to give credible inference to the whole sample. I’m not defending Gould — his arguments sounded theoretical (in an ad hoc sense rather than data driven), so have questionable credibility. But the recent paper (because of the opportunistic sample) also has questionable credibility. It’s a problem in the field. I wish people in the field would be more accepting/open about the uncertainties inherent in data acquisition in the field, and clearly point them out and translate them into uncertainties in their conclusions. Seems like the human tendency to want to make a strong case, even when you can’t really.

        • Lewis et al. were able to compare Morton’s measurements with their own measurements of the same skulls. The correlation was 0.98. That’s evidence enough of the reliability of Morton’s methods. Morton’s skull collection, even when it was intact, was always a convenience sample, containing whatever skulls he was able to get his hands on from around the world, so it was never suitable for straightforward comparisons between populations.

  9. I am all for replication studies and I have doubts about these correlations out of left field. But as one of the above commenters noted, adding controls without understanding what it does is pointless. This attempt is no different from those that claim that education does not matter because its effect disappear if you adjust for job.

    • I think you are misstating the argument about the effects of controlling for job in the education and earnings relationship. The argument is that requisite skills are what matter and that where those skills are gained is not restricted to formal education. Where they are related to barriers to entry such as licensure, etc. they are inextricably confounded since there is no data for those who did not gain the credential and thus entrance into the opportunity for licensure, etc.

      • No, the point is that part of the return to education is the option to access certain jobs. Once you control for job, you are measuring something else (or asking a different question). And that is the point that get lost when people take a model, add X factors, and claim that the conclusion changed.

        • Into what you worry about, two people have the same job, one with Ph.D., one a HS dropout. What would I expect the effect of the education to be? None. Both are there, whatever one has or lacks, the other somehow matched it. Conditional on being in the same job, education shouldn’t have much more to add. If anything, I wouldn’t be surprised if the lower credentialed one has better residual earnings. The fact he made the cut tells you something about him. And this is something I see in my line of work when you look at job level earnings. Zero education effects or reversal in the expected ordering.

        • So, two ‘Senior Software Engineers’, one with a CS PhD and one with a high school diploma, and you would expect the one with the high school diploma to be earning more, as supported by your anecdata.

          Not sure what world you are living in but it isn’t the same one as me.

  10. Nice/interesting comments all around.

    Bruce — interesting links.

    Pseudoramous — “set aside the statistical issues”. But, why should we do that? Sure, if we set aside the statistical issues — which actually go deeper than what is in this post — then this would be a great paper not worthy of mockery. This phrase is internet-meme worthy.

    MFK — I agree with you. Genetic distance is almost co-linear with a small number of geographic variables. This is why the claim that genetic distance is robust to a wide range of geographic and cultural controls, including controls for regions, is not credible. However, also note that within Sub-Saharan Africa, there is actually quite a lot of variation in genetic distance — more so than in the rest of the world combined. Only, within SSA, the more genetically distant you are from the US, the richer you are (also true for Asia). Ethiopia is actually closer genetically in the data than rich Japan and Korea. Also note that including regional controls which include SSA does not kill the significance of latitude or % of land area in the tropics…

    @Pseudoramous again — From the abstract “[genetic distance] has a statistically and economically significant effect on income differences across countries, even controlling for measures of geographical distance, climatic differences, transportation costs, and measures of historical, religious, and linguistic distance.” If genetic distance merely proxies other measures of distance, then why would it be barely affected by a long list of such controls? And, yes, as I wrote above, I’m aware the authors wrote that they favor the story that genetic distance is merely a proxy for cultural differences, but the empirical evidence they provide undercuts this story, in my view. And they also wrote: “we remain largely agnostic about specific mechanisms of technology diffusion, as well as about the specific traits and characteristics
    that create the barriers.” If you parse them in the round, they admit they can’t rule out a direct impact. The evidence they present certainly does not rule out a direct impact, and does not really bolster the idea that genetic distance operates indirectly.

    Jack PQ: Not speaking for Andrew, but I suspect he was likely a bit skeptical of the result to begin with. In any case, the data and code are on my webpage — knock yourself out.

    @JackP (2nd comment): it’s much better that results be tested by other researchers who don’t have an incentive to confirm the results. What happens is that it isn’t hard to produce a large number of robustness checks that have little prospect of overturning the results.

  11. I’m going to repeat: genetic distance is only about the time of last common ancestry between populations, and is measured from ‘junk’ genes which don’t do anything. Spolaore & Wacziarg were NOT purporting to find an effect of genes on GDP.

    • If a variant is neutral, it doesn’t mean that it cannot have phenotypic effects. It just means that population differences in allele frequencies aren’t due to natural selection.

      • It shouldnt even necessarily mean that – Junk DNA could in fact correlate with some other phenotypical variation, as natural selection is a random process. Person A has a mutation of genes 1 to 60 (one estimate of the average number of mutations), gene mutation 32 provides a significant advantage, so all 60 mutations are passed on to all relevant offspring. There is no reason for evolution to “prune” away those mutations except through a what could be a very long process of re-mutation.

        • I’m sorry but it’s pettifoggery to get bogged down worrying about the technicalities of “genetic distance,” which is more or less genealogical distance. It’s 2015 and this stuff is well worked out by now. It turns out that genome analysis gives pretty much the same results as classical genetic markers did in Cavalli-Sforza’s 1994 magnum opus the The History and Geography of Human Genes, whose results weren’t all that different from what Coon came up with from skull sizes and shapes and the like in 1965’s The Living Races of Man.

          A lot of people are under the misconception that Stephen Jay Gould’s 1981 book The Mismeasure of Man is the final word on the subject, but the last 34 years of rapidly advancing science have not been kind to Gould.

  12. ”Pseudoramous — “set aside the statistical issues”. But, why should we do that? Sure, if we set aside the statistical issues — which actually go deeper than what is in this post — then this would be a great paper not worthy of mockery. This phrase is internet-meme worthy.”

    Oh come on, you know what I mean. By “set aside statistical issues” I mean, I will comment on what “genetic distance” is and how it should and should not be interpreted. Regardless of whether the GD/GDP correlation is spurious or not, one should at least interpret correctly what the authors are saying when they purport that GD is robustly correlated with GDPpc.

    ”if genetic distance merely proxies other measures of distance, then why would it be barely affected by a long list of such controls? And, yes, as I wrote above, I’m aware the authors wrote that they favor the story that genetic distance is merely a proxy for cultural differences, but the empirical evidence they provide undercuts this story, in my view.

    Only if you believe ‘culture’ is exhausted by religion and language. They don’t believe so.

    From Spolaore & W. pp 513-514

    “ This opens up the very interesting question of what other vertical traits and characteristics are behind the large effects of genetic distance on income differences, besides language and religion. Although the identification of specific traits and characteristics is beyond the scope of this paper, and is left for further research, some further discussion is in order.

    “ There are several possible (and not mutually exclusive) channels through which relative genetic distance may operate as a barrier to the diffusion of innovations and development. A possibility is that genetic distance creates obstacles to interaction and communication that cannot be overcome through translation technologies (such as those readily available when people speak different languages). For example, genetic distance may reflect biological traits that, for cultural reasons—racism, discrimination, lack of trust—affect people’s willingness to interact with each other. This would be consistent with work by Guiso, Sapienza, and Zingales (2004, revised in 2008) on cultural biases and trade.

    “ Even when people are willing to interact with each other, communication and adaptation of each other’s innovations may be hampered by deep cultural differences (norms, values, habits, etc.) that are not codifiable and translatable from one society to the other. This would be consistent with the evidence in Desmet et al. (2007), already mentioned in the Introduction, showing a strong correlation between genetic distance and answers to 430 questions about norms, values, and cultural characteristics in the World Values Survey, correlations that remain even after controlling for linguistic distance.61 Such characteristics may facilitate the diffusion of innovations across cultures that share a set of common attitudes, while preventing or slowing down such diffusion when societies are more distant across a large range of values and norms.62”

    So their evidence does not “undercut” anything. S&W know very well — unlike Campbell — that genetic distances canNOT have a direct impact on anything. That’s not determined by statistics. That’s determined by genetics. When Ashraf & Galor published their “Out of Africa” paper, they made an absurd causal claim — that genetic diversity has a direct impact on social outcomes. That was derided even by genetic determinists (https://westhunt.wordpress.com/2014/06/18/diversity-galor/) who are perfectly prediposed to believe genes are a powerful driver of social & economic outcomes. Ashraf & Galor were foolish to think some cross-sectional correlations with GDP data could overturn genetical theory !

    • Pseudoeramus — Well, I certainly agree with you that there is more to culture than the variables they controlled for. And, as I wrote in the blog post above, I understand the authors wrote that they believe in an indirect effect. I also agree that the evidence they present doesn’t rule out an indirect effect. My question is what evidence do they present that you believe supports the idea that genetic distance merely proxies cultural differences? Certainly, not the regressions which purport to show that show that genetic distance is barely affected by an ostensibly exhaustive list of geographic and cultural controls.

      • “My question is what evidence do they present that you believe supports the idea that genetic distance merely proxies cultural differences?”

        This — I believe — is the correct representation of the logic behind the SW diffusion paper:

        1) By definition, Fst genetic distance refers to (overwhelmingly) neutral variation, so GD values can NOT have a causal effect on development. This is key. This drives all else. Spolaore & Wacziarg have a lot of interdisciplinary knowledge, so they know this and rely on this starting point. Most economists don’t know anything about population genetics or linguistic phylogenies, so this starting point is less vivid and the scare word ‘genetic’ becomes a distraction.

        2) Genetically more closely related populations are more likely to share cultural traits.

        3) So *assume* genetic distances are a proxy for cultural distances.

        4) But we know GD is an imperfect measure of CD. Control for some _observable_ cultural differences (i.e., languages, religions)

        5) What’s left over is _unobserved_ and possibly unobservable or difficult-to-quantify cultural differences. Some examples were given above.

        I personally think most cross-country regressions suck, plus I’ve never seen a good theory as to why cultural differences should constitute such a barrier to diffusion of technology, production best practises, etc. So I’m not really invested in the idea that culture prevents profit-maximising agents from taking advantage of things which would help them maximise profits. (***)

        But I was kind of put off by the “skull size” & related comments.

        ( *** There might be something to the “diffusion barriers” idea for household decisions, such as fertility. See around 35:00 https://www.youtube.com/watch?v=q0eM5zNZ2jc )

    • Pseudoerasmus– How genetic distance is measured here is irrelevant. Two reasons.

      First, tell us more about this “junk” DNA… http://www.nytimes.com/2015/03/08/magazine/is-most-of-our-dna-garbage.html?_r=0

      In short, the linked article suggests that junk DNA may not be junk after all.

      But a more important point is that distance since two populations separated likely proxies distance in useful DNA and traits. For example, Humans and Chimpanzees parted relatively recently compared to humans and fish, whether you measure this carefully based on junk DNA or whether you confirm by watching chimps go at it at your local zoo (I won’t judge). Thus a measure of genetic distance could be seen as proxying differences in useful traits, which it does.

      • But a more important point is that distance since two populations separated likely proxies distance in useful DNA and traits. For example, Humans and Chimpanzees parted relatively recently compared to humans and fish, whether you measure this carefully based on junk DNA or whether you confirm by watching chimps go at it at your local zoo (I won’t judge). Thus a measure of genetic distance could be seen as proxying differences in useful traits, which it does.

        You’ve got to be joking. Chimps and humans shared a common ancestor millions of years ago. Of course there are non-neutral genetic differences !

        Different human populations also have non-neutral genetic differences. Lactase persistence is one famous example. But that happened under natural selection. It didn’t happen under genetic drift, which is what create genetic distances. The point: the fact that some neutral genes may not be neutral after all doesn’t mean genetic distances measures mostly non-neutral differences ! They don’t.

        Besides, Spolaore-Wacziarg use Cavalli-Sforza data — which are pre-genomic and based on classical markers. You can actually look up the alleles the C-S genetic distance data are constructed from in http://press.princeton.edu/titles/5313.html

        • I am puzzled by your chimp comparison. Every person has mutations which distinguish him or her from another. When two populations are separated long enough, enough mutations accumulate to occur in measurably different frequencies.

          Yet most of these mutations do not come under selective pressure at all. And without selection (i.e., with only genetic drift) it takes much longer to “make” new traits. Chimps and humans have been separated for so long that both forces have operated.

          But human populations have not been separated long enough for drift to make most genetic differences between human populations non-neutral. That’s why the major non-neutral genetic differences (lactase persistence, EPAS1 in Tibetans, etc.) were found to be under (recent) selection.

          If anything, climate and geographic factors are more consistent with selection hyptheses.

        • First, tell us more about this “junk” DNA… In short, the linked article suggests that junk DNA may not be junk after all.

          You are confused. Most of the human _genome_ may not be “junk”. But most human genetic _variation_ is in “junk”. Individuals vary only in some fraction of 1% of their DNA (0.1-0.5%). Most of the variation in that 0.1-0.5% is neutral. And a lot of the non-neutral variation will govern physical features, blood types, and other traits which have no conceivable bearing on economic development. If there are genes which do have an impact on economic development, then they are quantitatively a miniscule portion of the already miniscule fraction, and they do not show up in the Cavalli-Sforza genetic distances used by Spolaore and Wacziarg. Unless of course YOU want to argue that different frequencies of blood types in different populations may have consequences for economic development ???

        • *Sigh* From someone who knows his overview of biology. DNA falls into the following classes:
          — expressed DNA — these are the popular “genes” which create proteins and RNA
          — control DNA — these are switches which turn genes on and off in response to chemical signals
          — structural DNA — these hold the chromosomes together and do similar things
          — filler DNA — this is gibberish in between the rest

          Early uses of the phrase “junk DNA” referred to control, structural, and filler DNA, and even to DNA which creates RNA. Structural and control DNA is clearly not junk, and without it we’d die.

          Modern uses of the phrase “junk DNA” refer only to filler DNA.

          There is a sense in which filler DNA is not junk. It is the result of fighting off virus infections and otherwise-deleterious mutations, mostly. And it may act as a genetic diversity reserve which enables faster future mutation; something which gives our species robustness in the face of severe environmental bottlenecks. But you *could* delete it all with no *short-term* consequences in an individual.

          When we do DNA fingerprinting, we are mostly using the filler DNA. Because all DNA mutates at a fairly fast rate, but mutations in expressed, control or structural DNA usually lead to spontaneous abortion. So the expressed, control and structural DNA is *nearly* the same in every human. But the filler DNA can mutate with no ill effects, so it’s different in lots of places in nearly everyone.

        • TLDR version:

          “Junk” DNA arguably does do something, but it mutates fast and without selective pressure, so you can use it for “DNA fingerprinting” and for determining genealogy.
          Non-junk DNA is conserved by selective pressure so you can’t use it for determining genealogy or “DNA fingerprinting”.

        • Let me translate it in a less opaque way:
          The guy is basically a moron who does not know WTF he is talking about (getting his science from NYT on top of that.)

        • Seeing how he is in the “ranked #1 in Russia” institution, peddling tired crap like “the best explanation for [North-South disparity] is Jared Diamond’s theory,” i confidently predict that he will not last long in Russia.

        • Look, Doug Campbell knew more about biology than you or Steve Sailer did.

          Mr. Campbell, for instance, knew that there was more genetic distance *within* sub-Saharan Africa than there is from anywhere in sub-Saharan Africa to anywhere in the rest of the world (this is a basic and well-known fact). This alone destroys Spolare and Wacziarg’s thesis, because if their argument that “genetic distance” mattered were correct, there would be massive divergence in development within sub-Saharan Africa — more than there is between sub-Saharan Africa and the rest of the world — which there isn’t.

          As a result, he proved himself to be more intelligent and better educated than you. You could do with bothering to learn something. Steve Sailer is yet again confirming my general belief that American Conservatives do not do their research, but merely look to confirm their pre-existing biases.

  13. Campbell and Pyun’s paper is a completely misguided criticism of our paper “The Diffusion of Development,” published in the Quarterly Journal of Economics in 2009. In that paper, which focused on the determinants of income differences, we did control for continental dummies, as well as for latitude and longitude, climate, percentage of land in the tropics, and so on, as any reader of the QJE can check. Moreover, our results hold when we exclude Sub-Saharan Africa and so cannot be driven solely by those countries (see tables A14, A15 and A16 at http://www.anderson.ucla.edu/faculty_pages/romain.wacziarg/downloads/diffusionmisc.pdf). Finally, our results hold even more strongly within Europe, where the issue of controlling for continental effects is obviously moot.

    Campbell and Pyun’s paper is marred by several methodological and conceptual errors – as we pointed out to them repeatedly – and the authors completely misinterpret our contribution. To start with, contrary to Campbell’s claims in Gelman’s blog entry, our results do not suggest a “direct impact on GDP” of “genetic endowments,” for the numerous reasons we have discussed in many papers, including the fact that genetic distance as measured by Cavalli-Sforza and co-authors captures neutral changes not subject to natural selection (pseudoerasmus is correct in his/her comments above), and that our main empirical analysis focuses on barriers to the diffusion of development, running horse races between relative genetic distance and absolute genetic distance in income-difference regressions (a point completely lost on Campbell and Pyun, whose paper only shows regressions on income levels, which are not the focus of our QJE paper).

    Regarding their income-level regressions, Campbell and Pyun’s main claim is that the correlation between genetic distance from the US and income per capita is completely driven by Sub-Saharan Africa. This is obviously false, because the correlation is robust to excluding Sub-Saharan Africa from the regressions (see table A14 at http://www.anderson.ucla.edu/faculty_pages/romain.wacziarg/downloads/diffusionmisc.pdf). This simple fact should be the end of the story: if an empirical relation holds for the whole sample excluding a subgroup, it cannot be driven by that subgroup.

    Moreover, the authors never show that just adding Sub-Saharan Africa or distance to the equator reduces the significance of genetic distance (in fact, it does not). They obtain such a result only when they add several other variables at the same time, all of them also highly correlated with genetic distance. In fact, Campbell and Pyun need additional and highly collinear controls at the same time to get what they would like to show: a significant effect of the Sub-Saharan dummy along with an insignificant effect of genetic distance. Overall, one gets the impression that these two authors only show that subset of regressions that provide this joint result, even if that requires arbitrary subsets of controls, with little conceptual and empirical justification.

    We share Marginal Revolution commentator ohwilleke’s sentiment about this being “a really powerful commentary on the appallingly sad state of professional journals in economics.” It is indeed appalling that a professional journal such as the Journal of International Development would publish an incorrect and misguided frontal attack on somebody else’s work without ever asking for the authors’ input or inviting us to provide a formal reply. Obviously, there were very good reasons for Campbell’s paper’s being repeatedly rejected at much better journals, without any need to invoke a conspiracy between numerous editors (and referees) and us. We wish we were that powerful!

    Finally, we want to make clear that our data on genetic distance were made available on our websites as soon as our work was published in 2009.

    • @Enrico and Romain (Reposted from Marginal Revolution…)

      Well, your enlivened response certainly lived up to your referee reports! In any case, let me say up front that I can certainly understand your desire to defend your work. Nothing about this is personal, and as I wrote over at Gelman’s blog, any ambitious researcher will occasionally have a result that doesn’t pan out. What is perhaps special about this case is that the critique in question appears to also affect your papers from 2013, 2014, and 2015. If not for this, I might have dropped the matter.

      Let’s start with what we all agree on. In the income-level regressions, genetic distance does not predict income per capita after one includes a dummy variable for sub-Saharan Africa and a second dummy for Europe, with no other controls. (Our paper is here: http://dougcampbell.weebly.com/uploads/1/0/2/2/10227422/diffusion_of_development_110514a.pdf). Secondly, in their table 1, Spolaore and Wacziarg only controlled for “absolute difference in latitude with the US” rather than distance from the equator, which is another influential control. Distance from the equator and a dummy for SSA also kill their results. These are two standard controls in the cross-country income literature, and their omission is puzzling. I say we all agree on this because, above, you did not question these results despite being the crux of the matter, and even a very partial and unfriendly referee (who had access to your data and wrote a report longer than our paper!) also validated this.

      The controversy is over whether it is reasonable to include a sub-Saharan Africa dummy as a control, compared to excluding sub-Saharan Africa countries, as you very strongly prefer. My question is, why should we get to exclude data selectively? Africa has more variability in genetic distance than any other continent, so it should be an interesting test case of your theory. In the data, poor east African countries such as Somalia and Ethiopia are closer to the US genetically than developed countries such as Korea and Japan. Within sub-Saharan Africa, as we show in our Figure 3, there is actually a positive correlation between genetic distance from the US and development. So, too, for Asia. If we exclude SSA and Europe, there’s also no correlation between genetic distance and income. Additionally, in the introduction to your paper, you did not write that your results are only robust when you exclude sub-Saharan Africa. Instead, you wrote that “genetic distance…is robust to controlling for a large number of measures of geographical distance…” And that “we control for a vast array of measures of geographic isolation.” A dummy for sub-Saharan Africa could certainly fit this bill.

      Another motivation for including a dummy variable for sub-Saharan Africa could be something like this:

      The largest genetic distances observed worldwide occur between populations that live on different continents. One concern is that genetic distance may simply be picking up the effect of cross-continental barriers to the diffusion of development, that is, continent effects.

      Actually, this isn’t my motivation – I copied the above motivation from your paper. You guys yourselves argued that your results were robust to the inclusion of continent dummies. Only, when you did this, in your country-pairs regressions, you lumped together rich North Africa with sub-Saharan Africa. In addition, as we point out in our paper, you included a sparse number of fixed effects (12 fixed effects for 6 various continent pairs – which have 6 + 5 + 4 + 3 + 2 + 1 = 21 possible combinations and thus require 21 FEs). Thus, you never even include a full set of continent controls, even when you lump in middle-income North Africa combined with sub-Saharan Africa. We show in our Table 3 that your results don’t hold in the country-pairs regressions – which you say you prefer — using regional dummies.

      In response to some of your comments:

      “a point completely lost on Campbell and Pyun, whose paper only shows regressions on income levels, which are not the focus of our QJE paper”

      Try again. Our Table 3 does the country-pair income differenced regressions. This is not really distinct evidence, as the data you manufactured here is highly dependent. These regressions were just included to give your paper an heir of sophistication.

      “Finally, we want to make clear that our data on genetic distance were made available on our websites as soon as our work was published in 2009.”

      The full dataset, including your sample of countries and other variables, are not available. One of your variables was not available at all online. A very partial referee who did have access to your original data sample still criticized us for not using the exact same sample as you. We believe this was disingenuous. I emailed you asking you to share your full data for your regressions, and you did not respond. That’s fine, everyone can miss an email, and, admittedly, I did email also about the genetic distance data specifically and you did respond to that, to your credit. However, this is why it’s advisable to simply post your data and regressions on your webpage. In that case, no one could ever criticize you for not sharing your data. Perhaps you guys can become trend-setters on this point.

      “It is indeed appalling that a professional journal such as the Journal of International Development would publish an incorrect and misguided frontal attack on somebody else’s work without ever asking for the authors’ input or inviting us to provide a formal reply.”

      Don’t worry – we shared with them our previous referee reports. I’ll close by giving you the last word, one of the gems from these reports: “There is no reason to interpret the Sub-Saharan Africa dummy as a geographic variable…”

      Very Respectfully Submitted, Doug Campbell

      • @Doug Campbell (also reposted from Marginal Revolution)

        We would not like to start a detailed back-an-forth with Doug Campbell here, but we must make a few important points in response to his reply.

        First, it is not true that our results in the QJE were robust to continental dummies only because we aggregated North Africa and Sub-Saharan Africa. Our results continue to be robust when we control for Sub-Saharan Africa separately, as shown in table A16 columns 2 and 3 (“CP” stays for Campbell-Puyn, as those tables were prepared in response to Campbell and Puyn’s criticism, and use their definitions of continents). So we do not agree with Campbell’s claim – our QJE results are indeed robust to including a Sub-Saharan Africa dummy, as well as other appropriate geographical controls.

        Second, we did not run income-difference regressions to give our paper “an heir [sic] of sophistication.” We did it because our paper was NOT about a permanent effect of genetic distance on income levels across countries, but about the barriers to “the DIFFUSION of development” (title of the paper!). In fact, our framework predicts that, in response to a major innovation – such as the Industrial Revolution – initially countries farther from the technological frontier tend to be poorer than those closer to the frontier, but over time many of them catch up. So the effect of relative genetic distance on income differences should decrease and possibly disappear over time, as innovations and development spread. We have made that point repeatedly, and documented it in several papers. Alex blogged about one of our more recent papers at http://marginalrevolution.com/marginalrevolution/2013/10/trade-development-and-genetic-distance.html, and showed our Figure 1, documenting a decrease in the effect of genetic distance over time as development spreads. Hence, our income-difference regressions are key to test our hypothesis that relative genetic distance works as a proxy for those barriers to the diffusion of development. (Moreover, the technological frontier also changes over time. It was Britain in the nineteenth century, the US in the twentieth century, but a thousand years ago it was probably China, and so on. So when studying the effect of relative genetic distance on income differences across societies over time, such frontier change must be taken into account).

        Third, the effect obviously takes place on average, and there are important exceptions, such as Japan – a fact that we discussed explicitly in our 2013 Journal of Economic Literature article (http://sites.tufts.edu/enricospolaore/files/2012/08/RootsF.pdf), where we wrote (p. 40): “The diffusion of modern development to East Asia, which started in Japan and spread to nearby societies, is an example of successfully overcoming long-term barriers. Japan is geographically, historically, and genetically distant from the European innovators, but it got the Industrial Revolution relatively early. This is not inconsistent with the existence of historical and cultural barriers across populations, because such barriers operate on average, and it is always possible for some society to develop traits and characteristics that make it closer to the innovator, or to sidestep cultural and historical barriers altogether through historical contingencies. When Japan got the Industrial Revolution, it became a cultural beachhead. South Korea followed, and then industrialization and modernization spread across several societies in East Asia. North Korea, in contrast, is a sad example that very bad policies and institutions can kill growth and development in a society irrespectively of any long-term historical and cultural variables.”

        Fourth, we interpret the effect of genetic distance not as a direct effect of genetics but as a proxy for differences in inter-generationally transmitted traits, such as culture and norms, acting as barriers to the spread of development (in fact, we are currently working to shed more light on the relation between ancestry and different measures of cultural distance across societies – for a short summary: http://www.voxeu.org/article/ancestry-and-culture, and for the longer version: http://www.anderson.ucla.edu/faculty_pages/romain.wacziarg/downloads/ancestry.pdf). Therefore, given our own framework, we DO expect that appropriately controlling for such relevant traits and barriers would greatly reduce or eliminate the measured effect of genetic distance on income – in fact, we would welcome a paper showing that with the appropriate variables and methodology. Unfortunately, Campbell and Pyun’s is not such a paper.

        Finally, it seems obvious to us that continental dummies in income regressions do not necessarily capture the effects of geography, and in that sense they should not be interpreted as pure “geographical” variables. For example, they may capture the impact of historical and cultural differences between societies located in different continents. By the way, definitions of “continents” are themselves culturally constructed concepts – “Europe” for instance is geographically part of Eurasia.

        • @Enrico and Romain

          OK, I understand you guys have to write something to keep some doubt, as few people will read the details of these long posts anyway.

          In any case — we’re moving toward convergence on some issues. Particularly, I see we’re in agreement that the evidence in Table 1 from “Diffusion of Development” isn’t robust. In your Column 7 of Table A14 in your web appendix, you reproduced our results showing that the apparent impact of genetic distance on income is sensitive to controls for latitude and an SSA Dummy. (I’ll post this myself, as I recognize it from the table from the referee who zinged us for not using the same exact sample as you even though you knew full well this was impossible.)

          In this post, you backed off of the claim that we should simply drop the sub-Saharan Africa observations, and argue instead that now you’d like us to ignore the income-level regressions. Now the paper, you say, is all about the country-pair differenced regressions. But, in the abstract of your paper you wrote “We find that genetic distance, a measure associated with the time elapsed since two populations’ last common ancestors, has a statistically and economically significant effect on income differences across countries….” Times seemed to have changed…

          In any case, I see the income-level regressions popped up again in your 2013 paper…. This was a long time after we first emailed you our results in January of 2011, when I offered to buy you lunch, share our results, and collaborate. You then asked to see our data, and we obliged. Then we asked for yours, and … crickets.

          Re: the country pairs regressions. What you did here is manufacture 10,000 data points from a sample of 150 countries by taking the differences of everything with each other. But you don’t remotely have 10,000 independent data points. In fact, you didn’t have 150 independent data points, which was part of the problem initially — both genetic distance and incomes are highly correlated in large geographic clusters. If genetic distance to the US doesn’t predict income differences across countries, then it can’t be the case that relative income differences to the US predict relative income differences generally. Thus, I don’t agree that there is any sense in these regressions once you’ve agreed that your Table 1 isn’t robust. In any case, I see you claim robustness to continent effects, using the data and code you won’t share. See, we found otherwise, in our Table 3: http://dougcampbell.weebly.com/uploads/1/0/2/2/10227422/diffusion_of_development_110514a.pdf

          I don’t know what is driving the difference in our results. The difference between us, however, is clear: we put our data and code on our webpage: http://dougcampbell.weebly.com/.

          Added comment: How about you put your data and code online for each of your 5-6 papers that use this genetic distance data? I fear that some of these other results may similarly be at risk.

        • ” we interpret the effect of genetic distance not as a direct effect of genetics but as a proxy for differences in inter-generationally transmitted traits, such as culture and norms, acting as barriers to the spread of development …”

          With such a vague definition, might as well come out and say that it’s a proxy for migration, both historical and current. Genetics, so-called “junk dna” or otherwise, _is not exogenous_ due to its non-random allocation by migration processes.

  14. “We also found that genetic distance to the US failed to predict income levels even when we just included two dummy variables, one for Europe and one for sub-Saharan Africa, with no other controls.”

    In other words, an extremely simplistic folk anthropology categorization by race into black and white works about as well as sophisticated genetics.

    Interesting.

  15. Genetic distance correlates closely with genealogical distance. Why would it be surprising if, for reasons of nature and nurture, people turn out to tend to be more similar to people who are more closely related to them genealogically?

  16. Genetic distance correlates closely with genealogical distance. Why would it be surprising if, for reasons of nature and nurture, people turn out to tend to be more similar behaviorally to people who are more closely related to them genealogically?

    • Because there’s a lot of psych evidence against that claim?

      People seem to have nearly the same distribution of psychologies no matter who you’re descended from; look at any large set of brothers (all with the same parents) if you don’t believe that.

      Probably due to the way brain-related genes get scrambled by recombination.

      Behavioral similarities seem to be largely due to culture. Now *that* shouldn’t be surprising, because kids are quite explicitly brought up to “behave themselves” in ways which seem appropriate to the surrounding culture, and even if their parents or teachers aren’t doing so, the other kids around them are reinforcing that too.

      Take Steve Sailer as a baby, and raise him in an Aboriginal family in Australia, he will behave like an Aborigine. Which might be an improvement, given the way he’s been displaying himself on this thread.

  17. In general, contemporary intellectuals have a hard time thinking about the implications of genealogy. There’s nothing more fundamentally real than your biological family tree, but genealogy just doesn’t strike intellectuals these days as cool. Genetics is cool, but the interplay of genetics and genealogy almost so often draws a blank.

      • From the paper: “And it is safe to say that Watson and Crick could not have foreseen a day when an analysis of Oprah Winfrey’s DNA would tell her that she was descended from the Kpelle people of the Liberian rainforest.” Why not? I mean, not exactly this result, sure, but in general relation between particles of inheritance and, well, inheritance isn’t it obvious? And why Watson and Crick? Their main contribution is double-helix structure of DNA. Is it so fundamental to the facts of genetics? What if it were to parallel strands or what if there were no strands at all and genes didn’t reside on chromosomes, but were assembled in some shapeless molecular jumbles? He should have probably assigned this “consequences unexpected by field’s pioneers” example to Mendel.

  18. Campbell and Pyun’s paper is a completely misguided criticism of our paper “The Diffusion of Development,” published in the Quarterly Journal of Economics in 2009. In that paper, which focused on the determinants of income differences, we did control for continental dummies, as well as for latitude and longitude, climate, percentage of land in the tropics, and so on, as any reader of the QJE can check. Moreover, our results hold when we exclude Sub-Saharan Africa and so cannot be driven solely by those countries (see tables A14, A15 and A16 at http://www.anderson.ucla.edu/faculty_pages/romain.wacziarg/downloads/diffusionmisc.pdf). Finally, our results hold even more strongly within Europe, where the issue of controlling for continental effects is obviously moot.

    Campbell and Pyun’s paper is marred by several methodological and conceptual errors – as we pointed out to them repeatedly – and the authors completely misinterpret our contribution. To start with, contrary to Campbell’s claims here, our results do not suggest a “direct impact on GDP” of “genetic endowments,” for the numerous reasons we have discussed in many papers, including the fact that genetic distance as measured by Cavalli-Sforza and co-authors captures neutral changes not subject to natural selection (pseudoerasmus is absolutely correct in his/her comments above), and that our main empirical analysis focuses on barriers to the diffusion of development, running horse races between relative genetic distance and absolute genetic distance in income-difference regressions (a point completely lost on Campbell and Pyun, whose paper only shows regressions on income levels, which are not the focus of our QJE paper).

    Regarding their income-level regressions, Campbell and Pyun’s main claim is that the correlation between genetic distance from the US and income per capita is completely driven by Sub-Saharan Africa. This is obviously false, because the correlation is robust to excluding Sub-Saharan Africa from the regressions (see table A14 at http://www.anderson.ucla.edu/faculty_pages/romain.wacziarg/downloads/diffusionmisc.pdf). This simple fact should be the end of the story: if an empirical relation holds for the whole sample excluding a subgroup, it cannot be driven by that subgroup.

    Moreover, the authors never show that just adding Sub-Saharan Africa or distance to the equator reduces the significance of genetic distance (in fact, it does not). They obtain such a result only when they add several other variables at the same time, all of them also highly correlated with genetic distance. In fact, Campbell and Pyun need additional and highly collinear controls at the same time to get what they would like to show: a significant effect of the Sub-Saharan dummy along with an insignificant effect of genetic distance. Overall, one gets the impression that these two authors only show that subset of regressions that provide this joint result, even if that requires arbitrary subsets of controls, with little conceptual and empirical justification.

    We share Marginal Revolution commentator ohwilleke’s sentiment about this being “a really powerful commentary on the appallingly sad state of professional journals in economics.” It is indeed appalling that a professional journal such as the Journal of International Development would publish an incorrect and misguided frontal attack on somebody else’s work without ever asking for the authors’ input or inviting us to provide a formal reply. Obviously, there were very good reasons for Campbell’s paper’s being repeatedly rejected at much better journals, without any need to invoke a conspiracy between numerous editors (and referees) and us. We wish we were that powerful!

    Finally, we want to make clear that our data on genetic distance were made available on our websites as soon as our work was published in 2009.

  19. This is an absolutely awesome post and set of exchanges. It is what the academic enterprise is all about, at its best. One point tho, genealogy has actually gotten a fair amount of attention from economic historians in recent years, in English due mostly to Gregory Clark:

    The Son Also Rises: Surnames and the History of Social Mobility. Princeton University Press, 2014 (with Neil Cummins et al.)
    A Farewell to Alms: A Brief Economic History of the World. Princeton University Press, 2007.

    Which has inspired a growing secondary literature, e.g., Chakraborty, Shankha, Jon C. Thompson, and Etienne B. Yehoue. “Culture in Development.” The World Bank Economic Review (2015): lhv018. http://wber.oxfordjournals.org/content/early/2015/04/08/wber.lhv018.short

  20. I probably don’t understand enough to post, but it seems to me that if removing Sub-Saharan Africa or introducing a dummy for it changes result substantially, it is the evidence of genetic distance being a poor proxy for cultural differences or barriers.

  21. The most important result of this is the evidence that the “leading journals” in economics are not leading by merit, but by, well, genetic relationships: they’re leading because of who they are, not what they are.

    This is unfortunately common in the social sciences. It is not as common in the mathematical or physical sciences.

  22. This research should have been rejected at the grant stage as uninteresting. By a very laborious method it has been proved that white countries are generally richer than non-white countries. Want a medal for that?

    “Finally, our results hold even more strongly within Europe, where the issue of controlling for continental effects is obviously moot.”

    OK, so you’ve additionally proven that the UK and Germany are richer than France and Spain. Couldn’t you have looked this up on Wikipedia or CIA World Factbook?

Leave a Reply

Your email address will not be published. Required fields are marked *