The election of Trump is a shock to the development world, like all other worlds. Some might wonder whether aid will be cut, but that is among the least of our problems. The intended beneficiaries of aid efforts are also the targets of resurgent racism and xenophobia. Many people in the development world are busy signing open letters protesting xenophobic immigration bans, but we do so as private individuals just like others who don’t work in development. Is there anything about development ideas that could help shed light on the debates triggered by new waves of xenophobia and racism?
There seems to be an unofficial taboo on discussing racism in development, even though our work has a lot to say about how different nationalities or ethnic groups have good or bad development outcomes and seeks to prevent the bad outcomes for unfortunate groups. The taboo probably emerged for some very good reasons. Maybe we don’t want to give any recognition to racist arguments by engaging them directly. But even such a taboo would still require deciding which arguments are racist, so it doesn’t really avoid the issue. We also may be wary of entering a highly politicized world where you demonize opponents as racist, you enforce politically correct language, and you wind up suppressing free speech and academic freedom—as has indeed happened in some other settings confronting racism. But it would also be censorship to refuse to discuss some topics just because they are too politicized in some other settings.
I make the case for the opposite of censorship—there should be more frank discussion on all topics related to racism and xenophobia in development. We should avoid accusations of racism as much as possible and nobody should be self-righteous—including me and this essay, which could be mistaken, unrepresentative, or unfair to others’ work. I simply argue that development ideas (often more in the public intellectual sphere than in the academic sphere) have not been clear enough about what we mean when we talk about different outcomes for different nationalities or ethnic groups, mainly because we fail to articulate a clear theory of groups.
Carelessness About Generalizations
A few days ago, I asked my undergraduate development class to react to two different statements: (1) “The average trust level today for ethnic groups victimized by the slave trade is lower than that for groups that escaped the slave trade,” and (2) “Nigerian immigrants to other societies tend to be untrusting and opportunistic.” The first statement is based on a paper published in a top journal by one of the leading social scientists of African origin, Leonard Wantchekon of Princeton, and co-authored by another well-known academic, Nathan Nunn of Harvard. The second is a quote from a widely discussed book for general audiences on immigration called Exodus by another leading development economist, Paul Collier, and was also based on the original study by Nunn and Wantchekon.
My unscientific, small multinational sample of 40 students did not object to the first statement as it is a description of a statistical finding and nothing more. It seems obvious that there should be nothing wrong with recognizing differences in mean outcomes between well-defined groups. However, many objected to the second statement as too much of a vague generalization about a whole nationality. Collier may have intended the second statement to be just like the first, but the statement’s imprecision suggested something else to some.
In order not to be unfair to Collier or anyone else in this essay, I should confess that I have been on the receiving end of complaints about my own possible ethnic generalizations. In my 2014 book, The Tyranny of Experts, I discussed the work of Stanford economic historian Avner Greif on the Jewish Maghribi traders in eleventh century Cairo, who managed to enforce honest behavior on agents they trusted with large sums of money, even though they lacked courts or other formal redress if the agents cheated them. The enforcement utilized the threat of expelling dishonest traders from any further dealing with the ethnic network of Maghribi traders, which was costly since the Maghribis dominated trade at this time. Guido Tabellini (echoing other culture researchers) describes the Maghribi outcome as similar to values observed in many other historical and modern settings, calling them “limited morality,” which works wonders in enforcing moral behavior among those who belong to the ethnic network, but does not necessarily deter immoral behavior toward those outside the network. Here was my description of the values of “limited morality”:
Under these values, a group member should not cheat fellow group members, but there were no restrictions on cheating outsiders. Everyone behaved according to the same principle, so only insiders could be trusted, not outsiders. In such circumstances, tolerance for different races, religions, or lifestyles was likely low, and outsiders were seldom entitled to the same treatment as the insiders of a given race or religion. In fact, it was acceptable and even expected to cheat or rob them. Obviously, the precept that “all men are created equal” was not widely shared.
Alan Spira of Stanford Medical School wrote me to complain about this passage soon after the book came out in 2014:
On page 135, in your discussion of the Maghribis, you state, “In fact, it was acceptable and even expected to cheat or rob them. Obviously the precept that ‘all men are created equal’ was not widely shared.” I found this disturbing.
You are committed to the fight against racism and promotion of democracy—which is clear from your writings, but this statement repeats an old calumny against the Jewish people regarding being greedy, manipulative, and conniving. In fact, is not acceptable in Judaism to cheat people or rob them. It is a violation of the Ten Commandments which Jews hold sacred. ...Furthermore, in the Old Testament, it is stated that all people are created equal as humans, and descended from a common source…
I contacted Professor Greif who is also at Stanford and he expressed to me that he never wrote nor implied in his writings that the Maghribis were immoral, nor did they endorse the behavior you attributed to them.
My concern is that a reader would, by extension, apply your comments to Jews in general.
Dr. Spira and I had some cordial further discussion in which I tried to clarify that I was describing an abstract theory and had not intended those statements to be a general characterization of Maghribis or Jews, and Spira accepted this. Nor do I believe Collier intended to make offensive generalizations about Nigerians. I think our problem was carelessness about general statements, failing to articulate what we mean when we discuss theories and empirics of group characteristics and outcomes.
Controversy about group generalizations that may be too careless is also a problem of official aid agencies, NGOs, and celebrities doing aid advocacy. An article on a Ghanaian media site in 2010 reported, “The World Bank has apologized for displaying damning images of Ghana at its recent annual meeting in Washington, DC. Some of the pictures have half-naked women breastfeeding their kids and portray a country high on poverty levels. A journalist…told Joy News the World Bank portrayed Ghana as a country full of hungry and miserable people.”
Careless statements can also be created by exaggeration by development advocates or NGOs. The New York Times quoted a development activist during the refugee crisis in 2015 as saying “the entire global south…is on the move” (while actually only 0.5 percent of the global South’s population were refugees). Celebrity activist Bob Geldof said in 2005, “War, famine, plague, and death are the four horsemen of the apocalypse and these days they’re riding hard through the back roads of Africa.”
Generalizations are sometimes created by numerical exaggerations. The U.S. Agency for International Development in 2016 claimed that “violent conflict and instability affect at least…1.5 billion people,” which was three times higher than the number in the World Bank database. World Vision in 2016 went even higher saying, “By 2018, the majority of the extreme poor will live in fragile states” (compared to the World Bank number of 17 percent). A 2014 celebrity song to raise funds to fight Ebola generalized West Africa as “a world of dread and fear” offering “a song of hope where there’s no hope tonight.” Some African musicians fought back with a parody song “Africa for Norway,” in which Africans would contribute funds for radiators to warm up freezing Norwegians.
It is often insightful to think about how you would feel about generalizations if they were applied to your own group. I had this opportunity with another recent study that found that people of Scots-Irish ancestry in the American South are yet another group unusually prone to bad behavior. Development economist Pauline Grosjean showed that the percent of people of Scots-Irish ancestry in Southern counties in 1790 is positively associated with county-level homicide rates today. Her theoretical explanation is that the Scots-Irish (a group of Scots that moved first to Ireland in the 1600s and then to America in the 1700s, many into the Appalachian South) were originally herders, and that herders use the threat of homicide to deter thefts from their herds, and also came from or went to areas lacking formal institutions to punish theft (and homicide). The article quotes the conventional wisdom that “the 18th century Scots-Irish and Scottish Highlanders were more prone to interpersonal violence than other Western European settlers, as they originated from areas that were lagging behind in the ‘civilizing process.’” The study addresses a puzzle of a higher white homicide rate in the South than elsewhere in the United States and so it could be yet another valuable insight into the lasting effect of cultural historical legacies. But it would be much more problematic as a group characterization (which I don’t believe Grosjean intended).
As usual, writing for general audiences goes further than academic studies. A bestselling memoir by a self-identified member of this group, titled Hillbilly Elegy (citing both personal experiences and research), describes an incident when his grandparents destroyed a pharmacy after the pharmacy clerk insulted their son, and then generalized the situation saying, “That’s what Scots-Irish Appalachians do when people mess with your kid.” As it happens, my own (non-violent) grandmother was Scots-Irish from the Appalachian South. I don’t object to Grosjean’s statistical analysis, but I wrote a column severely criticizing Hillbilly Elegy.
Of course, some groups really do have worse average outcomes than others. What are some good and bad ways to characterize those facts?
The first point to make is that it is already progress (but, as we will see, not enough) that we discuss good or bad group outcomes as being the result of some historical circumstances (like the slave trade or history of herders) or game-theoretic interactions between individuals (like Greif’s model). This contrasts with ancient and modern racist views that different groups have different innate tendencies to behave well or badly. Modern views of innate group tendencies will more often cite culture or religion rather than biological race.
Economics has a long tradition of skepticism about explaining outcomes with innate differences. The classical economist John Stuart Mill in his (unsuccessful) fight against nineteenth century racism attacked “the most vulgar” theory of regarding racial “distinctions of human character as innate, and in the main indelible.” He insisted that development differences “between individuals, races, or sexes” are “produced by differences in circumstances.”
Gary Becker and George Stigler in more recent times carried on the same battle against just assuming different people or groups have different innate tastes, which they pointed out simply means “the problem is abandoned.” They recommend instead for economists to “continue to search for differences in prices or incomes to explain any differences or changes in behavior.”
Today’s economics continues the skeptical tradition about not just taking group differences as given, although we rarely talk about it openly in development. Economics deploys methodological individualism, in which we start with how individuals (who we assume to be primordially equal) choose to behave when facing their own circumstances, and then build up group outcomes from there. Individuals may choose to identify with an ethnic group and follow social norms of that group. How individuals choose to interact with each other will cause social norms to emerge within or across ethnic groups. Since social norms involve everyone following everyone else, there are usually multiple equilibria in social norms. Models with multiple equilibria (although not the only possible kind of model) are a good example of how the same people or group could be caught in either a “good” or “bad” equilibrium, depending on the historical starting point.
For example, the raids by slave traders may have caused their victims to rationally distrust others. And once you are in an untrusting social norm where everyone is distrustful, it is hard to get out. So the next generation after the end of the slave trade may still be taught by their parents to rationally distrust others, and so on, with each new generation. The ethnic groups lucky enough to have been far away from slave traders may have been able to start off in a more trusting equilibrium, and then kept passing that along to the next generation. So the ethnic groups victimized by the slave trade (most of them along the coast of West Africa) could still today be caught in an untrusting equilibrium, while ethnic groups who escaped it enjoy an equilibrium of more trust that makes possible greater investment and trade. Nathan Nunn spelled out a story like this in another article on how the slave trade history also predicts lower income among the formerly victimized groups today. Collier drew upon this theory and evidence when he said, “Nigerians radically, deeply, do not trust each other. Opportunism is the result of decades, probably centuries, in which trust would have been quixotic, and it is now ingrained in ordinary behavior.”
These models and results also contradict the old racist canard that ethnic groups collectively decide how to behave. One of the great insights of economics is that many social outcomes are the result of a general equilibrium process of decentralized interaction of individuals, not led or planned or intended by any one person or any group of persons.
Collier and I were offering some (non-racist) explanations that gave insights into some possible group outcomes and yet still managed to (understandably) offend some of our readers, so there is still another problem to be confronted. These models and empirical results were intended for explaining the effect of some development factor X (like slave trade history) on some development factor Y (like trust). The units of observation or illustration happened to be nations or ethnic groups, because that was where the data and the historical actions were. But maybe we did not think enough about whether they are suitable for making some more general characterizations of any ethnic group.
If we really wanted to make more general statements about any ethnic groups in the real world, we would need to confront how fuzzy and endogenous the definitions are of who belongs to what group (sometimes the group definition is even circular as it is based on the outcome being explained, like “poor whites”). There is exogenous human diversity like location of birth, but then how to define groups around this diversity is arbitrary. This fuzziness about definitions is not some flaw that can be fixed with enough effort at measurement—it reflects the reality that there are rarely any objectively correct group definitions as they are endogenous and heterogeneous constructions by their own members and by non-members reacting to them. We would then acknowledge how complex self-identification is with groups (individuals often self-identify with more than one group or kind of group, for example, and the intensity of self-identification varies within groups and over time). The process of outsiders defining groups to which they assign others is equally complex and heterogeneous. The same person can be called black in America, white in Brazil, and colored in South Africa.
These problems don’t automatically discredit all results based on group differences, since measurement error is common in development, but they make group generalizations even more problematic. We don’t grapple enough with how imprecise and possibly non-robust correlations are between group membership and development outcomes. We need to look at many other group definitions and outcomes besides the one we are emphasizing at any moment. We should be more cautious because of the history of exaggerated stereotypes of some outcomes for some groups. And all of these considerations might be influenced by the purpose of the group characterizations—is it to determine who gets aid, who gets to immigrate, or who will be the target of national security or crime policies?
The complexity of defining groups undermines the above-cited group generalizations. Collier shifted from “groups victimized by the slave trade” to “Nigerians” (but northern Nigerians were not victimized by the slave trade, and many other non-Nigerian groups were) and then he shifts to Nigerian immigrants to the United Kingdom. The eighteenth century Scots-Irish prone to interpersonal violence intermarried with many other groups and moved around a lot into and out of the Southern Appalachians, so the people who belong to this group are almost impossible to define—maybe I only gave away my own Southern Scots-Irish Appalachian identity by being offended by accusations of homicidal tendencies.
The biggest problem of all with group generalizations is the individual heterogeneity within each group. To say that one group is innately prone to a worse mean outcome is offensive enough, but what is even more offensive is to say there is little or no variance within the groups: “You people are all the same.” Historically, zero variance was a staple of racist thought. Even celebrated economists like Alfred Marshall believed in zero variance, writing in Principles of Economics in 1890 that “scanty and untrustworthy as is our information about the habits of savage tribes, we know enough of them to be sure that they show a strange uniformity of general character.”
We have finally arrived at the most important issue of all in how we view groups. First, regardless of low or high measured heterogeneity of some indicator, the group uniformity view is ethically offensive in denying the individuality of each person. This view forces on individuals against their will a faceless sameness based on our view of the group to which we decide to assign them. Even if most group mean outcomes were bad and even if intragroup variance on these outcomes really were small, members of that group do not thereby lose their rights to be seen as an individual. If we deny this individual dignity to members of some groups more than to those of other groups, then we indeed violate the ideal of equal respect for all.
Second, zero heterogeneity within groups is usually empirically false. Some theories of strategic interactions leading to social norms did predict everyone would behave the same, but these are just toy models capturing one tiny dimension of a complex world. The result relating today’s individual trust outcomes to the slave trade history of groups explained a grand total of 1 percent of the variation in individual trust outcomes in the Nunn and Wantchekon sample of about 20,000 individuals.
Of course, asserting homogeneity is even more extreme when we are discussing rare behaviors like terrorism or crime. The attempt to exclude from immigration Muslims or citizens of some “terror-prone” Muslim nations is like trying to find the 0.01 percent needle in the 99.99 percent haystack of non-terrorist Muslims (these are real numbers based on estimated worldwide numbers of terrorists and Muslims).
Collier also offers more generalizations about crime-prone groups: “Jamaican culture is among the most violent in the world. For example, murder rates are fifty times higher than in Britain. Guns are normal, so it is unsurprising that Jamaican immigrants brought their gun culture with them. Indeed, the gun culture of the Afro-Caribbean community is now a specific concern of British crime policy.” Although 50 times is a lot, the Jamaican murder rate in 2015 is still only 0.00045 percent of the population. The generalizations about “Jamaicans,” “Jamaican immigrants,” or the “Afro-Caribbean community”—not to mention the fuzzy group definitions yet again—are neglecting the heterogeneity implied by a vast majority of non-homicidal Jamaicans.
Not Enough Talk
So, in conclusion, this partial and limited look at racism and development ideas suggests our problem in development is not that we talk about ethnic group differences too much, it’s that we don’t talk about them enough. I suggest the opposite of policing language to adhere to some universal political correctness, instead suggesting that we all think and argue a lot more about what we mean and do not mean when we discuss development of groups.
My own imperfect effort to do this in my own teaching, research, and reading of others’ work leads me to want to be more precise about language. Do state the mean of development outcome Y for group X, but don’t use vague language that could imply all people from group X are innately prone to outcome Y. (Or if you do believe in innate tendencies, then say so.)
We should also clarify our methodological individualism underlying theories of group outcomes to better emphasize our non-racist explanations of group differences (the same differences that are cited by racists and xenophobes as support for their beliefs) and to emphasize how much our values, our theories, and our empirics all recognize individual heterogeneity within groups. We should acknowledge how endogenous, fuzzy, and badly measured the group definitions are and hold ourselves and other public intellectuals, researchers, aid agencies, and NGOs accountable for exaggerations and stereotypes.
In development, we have more to offer the debate on xenophobia than just signing an open letter.
The author is grateful for comments from Elizabeth Dalton and Angus Deaton, but he alone is responsible for any mistakes in the essay.
William Easterly is a professor of economics at New York University. Follow on Twitter @bill_easterly.