Johan Fourie's blog

I'd rather be a comma than a fullstop

Posts Tagged ‘race

The trouble with bad social science

with 2 comments


I don’t often read Foreign Affairs – the leading political science magazine published by the Council on Foreign Relations – but when, ahead of a 14-hour flight back to South Africa, I found an airport bookshop displaying the March/April issue with the provocative title ‘The Trouble with Race’, I decided that it would make for good airplane reading. Here was one of the world’s most authoritative magazines writing about an issue that affects everyone everywhere, but it is in South Africa during March 2015 that racial injustices has been brought into the limelight. Race, and racial prejudice, is the underlying theme of #RhodesMustFall, a movement by thousands of students across the country agitating for the dismantling of white power structures two decades after apartheid formally ended. Discussions of race and racial exclusion has emerged at campuses across the country. Rhodes will fall, but the future is murky.

And then I spot the article in Foreign Affairs titled ‘Apartheid’s Long Shadow: How Racial Divides Distort South Africa’s Democracy‘. Finally, I thought, some answers to our country’s qualms. From the leading thinkers of Foreign Affairs.

I turn to page 41. The paper is by James Gibson, Professor of Government at Washington University in St. Louis and (to my surprise) Professor Extraordinary in Political Science at Stellenbosch University. It starts brightly: South Africa is a diverse and unequal society. ‘The future of South Africa’s multiracial democracy depends heavily on minimizing animosity and hostility among these groups. Can South Africans of different races continue to get along?’ Wonderful. Answers.

RaceThen Gibson attempts to explain who these peoples are: Black South Africans, the ‘descendants of the original inhabitants of the area’, who constitute different ethnic and linguistic subgroups. ‘Apartheid eliminated such distinctions by grouping all blacks together and, in essence, expelling them from South Africa.’ This is awkward: had Gibson attended a first-year History class (or even my second-year Econ History class), he would know that there existed a group of original inhabitants known as the Khoesan. (Or Khoisan. Actually they are not, in fact, a single group but for the purposes of this post, let’s assume they are. More about them here.) Blacks did arrive south of the Limpopo (around 300 AD) long before whites , but they did not enter vacant land, no more than the Voortrekkers trekked into vacant territory when they crossed the Gariep (as the apartheid apologists wished to believe). The sad truth is that the original inhabitants of South Africa have been mostly wiped out.

And he goes on: ‘A third group is referred to as Coloured, a term that applies to an extremely diverse group of people whose mixed heritage attests to decades of intimate contact between three different peoples: Europeans, certain groups of black Africans, and slaves brought to the area by European colonists from Madagascar and Southeast Asia.’ Again, where are the Khoesan? Surely they can’t be classified as ‘certain groups of black Africans’? (In a paper with Erik Green, I call them the ‘Missing People‘. Gibson’s paper shows why.)

But these factual errors aside, Gibson’s paper is about evolving racial prejudice in South Africa. He argues that race relations are not really improving, despite two decades of democracy. Using the Institute for Justice and Reconciliation’s South African Reconciliation Barometer, a survey among a representative sample of the population, he finds that ‘intergroup prejudice is not the exclusive problem of black people in South Africa, but as a group, they express considerably more prejudice against their fellow South Africans than do the members of the minority races’. And ‘Whites and Coloured South Africans who have more contact with people of others (sic) races are considerably less prejudiced. Among black and Asian South Africans, however, not only does contact with members of other groups not reduce prejudice; it actually seems to increase it.’

These are powerful conclusions on an emotionally-charged topic, yet Gibson provides little empirical justification to support these claims. Yes, he shows that some race groups have higher self-reported prejudice than others, but as any social scientist should know, correlation does not equate causation. To say that black South Africans are more prejudiced suggests that you’ve controlled for other observable characteristics that may also affect prejudice, like income, location, and education to name a few. But perhaps it’s not black South Africans that are more prejudiced. Consider, instead, that it is poor education that cause people to be prejudiced. But because poor education outcomes are highly correlated with race in South Africa, by implication you will also find a high correlation between race and prejudice. Race, as Gibson uses it, is likely to just be a proxy for other things.

Social scientists solve these issues by using statistical techniques such as regression analysis. Even with a standard regression, though, causality is difficult to determine. Usually we use experiments or, because this is often not possible in a social setting, natural experiments. One* excellent example of this is Martin Abel’s work on forced removals during apartheid. Abel, a PhD student at Harvard University, investigates whether the people that were moved during apartheid are more (or less) prejudiced towards people of a different race today than their counterparts who were not moved. He finds:

…that those living close to former resettlement camps have higher levels of trust towards members of their social network and people in general. They are also more trusting towards members of both their own and other ethnic groups. This effect is larger for people born after 1975 who did not witness the forced removals, which suggests that events had long-term effects on social capital. This is also reflected by lower crime rates in these communities. Exploring causal mechanisms, I document that resettlement camp areas are more ethnically diverse and that diversity is positively correlated with measures of social capital only in areas affected by relocation. Interacting with people from different backgrounds and adopting a shared identity as displaced people may explain why relocation communities have higher levels of social capital despite potential short-term conflict over resources.

Abel’s careful statistical design and analysis forces him to draw exactly the opposite conclusion than Gibson does from his simple correlations. Deeper integration makes people less prejudiced, not more. The policy implications are profound.

#RhodesMustFall shows that there is indeed much value to investigating how and why prejudices evolve. But the way James Gibson does it in Foreign Affairs is not the way to do it. We should be skeptical of causal claims based on simple correlations. That Foreign Affairs accepted an article with factual errors and wonky causal claims on a topic of such importance is worrisome. We should expect better science from our social scientists.

*I received a message on Facebook from MIT PhD student Dan de Kadt: I read this article too, and had a similar reaction. Johan, in some other work Melissa Sands and I address the exact hypothesis Gibson proposes… We find that contact does partially explain cross-racial voting (whites who are in contact with blacks are more willing to vote ANC, and blacks who are in contact with whites more willing to move from the ANC). But there’s very little in our data to say that racial attitudes are affected (so little that we don’t report it in the paper). And there’s no evidence that contact has differential effects depending on race (as one assumes Gibson is proposing).


When diversity hurts

with one comment

‘Diverse People Unite’ is the motto on the South African coat of arms. Diversity is a great thing, we believe, because it exposes us to new peoples, new experiences and new ideas. But what if diversity also results in lower productivity? This is the uncomfortable result of a forthcoming paper in the Quarterly Journal of Economics, one of the foremost journals in the field. The paper, written by Jonas Hjort of Columbia Business School, provides evidence that suggests that ethnically diverse teams are less productive than ethnically homogeneous teams. Hjort writes:

A body of literature suggests that ethnic heterogeneity limits economic growth. This paper provides microeconometric evidence on the direct effect of ethnic divisions on productivity. In team production at a plant in Kenya, an upstream worker supplies and distributes flowers to two downstream workers who assemble them into bunches. The plant uses an essentially random rotation process to assign workers to positions, leading to three types of teams: (a) ethnically homogeneous teams, and teams in which (b) one, or (c) both downstream workers belong to a tribe in rivalry with the upstream worker’s tribe. I find strong evidence that upstream workers undersupply non-coethnic downstream workers (vertical discrimination) and shift flowers from non-coethnic to coethnic downstream workers (horizontal discrimination), at the cost of lower own pay and total output. A period of ethnic conflict following Kenya’s 2007 election led to a sharp increase in discrimination. In response, the plant began paying the two downstream workers for their combined output (team pay). This led to a modest output reduction in (a) and (c) teams – as predicted by standard incentive models – but an increase in output in (b) teams, and overall. Workers’ behavior before conflict, during conflict, and under team pay is predicted by a model of taste-based discrimination. My findings suggest that inter-ethnic rivalries lower allocative efficiency in the private sector, that the economic costs of ethnic diversity vary with the political environment, and that in high-cost environments firms are forced to adopt “second best” policies to limit discrimination distortions.

Pushed to the extreme, these findings suggest that, if South Africa’s businesses want to become more productive, then they should forget about the ‘Rainbow Nation’ and only employ people of the same ethnicity, i.e. Xhosas or Sothos or Afrikaners or Zulus. Businesses that opt for diverse teams will grow slower than businesses which employ only people from one ethnic group; in a profit-maximising, Darwinian world, the ethnically homogenous businesses will eventually win against the diverse but unproductive ones. Where businesses already employ different ethnicities, Hjort’s evidence suggests that it would be best to let those of the same ethnicity work together in teams instead of mixing team members across ethnicities. So much for ‘Diverse People Unite’!

Of course, things are not that simple, and it would be silly to conclude from this that diversity is evil. We should remember that most teams in the real world don’t assemble flowers into bunches; teams are often required to provide creative solutions to complex problems where innovative, out-of-the-box thinking is required. It seems reasonable to assume that ethnically diverse teams have a higher probability of dissimilar ideas than ethnically homogenous ones. In economics jargon, perhaps ethnically diverse teams are more prone to economies-of-scope rather than economies-of-scale. And a logical conclusion from Hjort’s work is that, were ethnic rivalries to disappear, the negative impact of diversity on productivity would all but disappear too.

But such results force us to stop and think about the untested assumptions we make. And it raises difficult questions too: even if our own labour and development economists find that diversity hurts our productivity, is there not a moral argument in favour of maintaining ethnically diverse teams? Given our history of forced racial segregation, how do we weight the ethics against the efficiency arguments? My expectation would be that some of the counter-productivity effects that Hjort finds will be mitigated over time by working with people from a different ethnic group, and that this in any event is the only way of addressing the issue at its core. As economists, we should understand the value in delaying instant productivity gains for even greater long-term benefits.

Written by Johan Fourie

October 15, 2014 at 09:39