Friday, 30 March 2018

On race

Where Left wing economic ideas have been in eclipse for half a century, their social ideas have been hugely successful in changing attitudes, and backing up those changes with institutions. This is the first of three essays which take stock of this social agenda. They will focus in turn on three defining enemies of the post-1960s "equality agenda": racism, sexism and homophobia.


The modern anti-racist impulse came from the discovery of Nazi crimes in Europe. Ideas of innate racial differences had served to justify Nazi racial hatred. Back home, Americans found the same ideas supporting the segregation of the Jim Crow South. In the struggle to change this and similar systems, the word racism came to cover two concepts:
  1. A belief in innate, natural differences between racial groups.
  2. Hatred, prejudice and discrimination against other races.
Racists believed (1), and their belief was caused by, and served to justify, (2), so it was natural to name a single syndrome for both aspects.

This concept leads people to think that if (1) is true, then (2) is justifiable. This seems to be agreed by both racists and anti-racists.

The problem is that (1) – let's call it the "natural differences" hypothesis – is an empirical claim about the world. It might have turned out to be false. It still may turn out to be basically false. But as time goes on, that is looking less likely. To explain this, I need to talk about the science of race and ethnicity. I claim no deep expertise in genetics. But I do know a bit about both genetics and about the social science of ethnicity, which I hope gives me some standing to discuss both.

The genetics and social science of races

First, racial groups are a kind of ethnic group, and ethnic groups are best seen as political identities. Being, say, white or Hispanic in the US is more like supporting a football team than like being tall or short-sighted. It is an identity sometimes chosen by ourselves, sometimes imposed on us by others. There are many examples of ethnic identities changing with political circumstances. For instance, the US census used to ask people if they were white or black; it now asks if people are white, black or Hispanic. The reason is that with greater immigration from Latin America, being Hispanic became a salient political identity in the US. Politicians campaigned for the Hispanic vote, and Hispanic people worked together to support shared interests. There are similar cases in Africa, where tribal identities are shifted by the demands of politics.

Ethnic group identities are ascribed by descent – you get them from your parents. Racial groups are a particular subset of ethnic groups where the “badges” that categorize people include natural characteristics like skin colour, which relate to relatively long-run differences in ancestry. These characteristics often vary continuously, and people may spend a lot of effort policing the boundaries of the group: 19th century America had a whole terminology of “quadroon” and “octoroon” to describe different mixes of black and white ancestry.

Nevertheless, while the categorization is social, the characteristics are natural and inherited. This means that the natural characteristics which place you into an ethnic group correlate with your ancestry, which in turn correlates with other genetic differences. There are some well-known examples: people of African descent are more likely to carry a genetic variant putting them at risk of sickle cell disease; there are similar cases for Ashkenazi Jews, who went through a small population bottleneck in the first millennium AD.

They might also correlate with genetic differences that cause important psychological characteristics. As twin studies show, almost all psychological characteristics are partly genetic: that is, differences in two individuals’ characteristics partly reflect differences in their DNA. So, there could be differences between racial groups on any of these characteristics.

“Could be:” there’s the rub. We don’t know yet. As is often said, the fact that inter-individual differences have a genetic basis does not imply that intergroup differences are. But that works both ways: the fact that inter-individual differences are partly environmental, equally, does not prove that intergroup differences are environmental.

So, with all this data, why can’t we find out? It turns out that resolving the question is harder than it sounds.

What we don’t yet know

First of all, you can’t just gather a sample of different ethnic groups, test for genetic variant X, and see how this correlates with your psychological or social outcome of interest. To see why not, try a thought experiment. Take a genetic variant for dark skin. Now, let’s sample the whole US population and test whether this variant affects people’s income. Surprise! It has a large effect. Have we discovered a pathway from the gene to the brain’s famous money-making centre – the nucleus Wallstreeticus? Nope. I told you, by hypothesis, the gene only affects skin colour. Of course, people with this gene are more likely to be black; and black people in the US have less income on average; so our gene is associated with lower income. But we knew that already. There is gene-environment correlation. That tells us zilch about causation.

This problem has led geneticists to focus within populations, rather than between them, in their hunt for causal variants. Mostly, they look at white people. Even among white people, they try to control for broad ancestral differences. The most extreme version of this is to look within siblings from the same family. This way you can identify the true effect of a given variant, because genes are handed out randomly from parents to children.

But now we have a second problem: the effect may be different among different ethnic groups. A variant which makes white people taller may not do the same for black people: first because different bits of DNA interact to build a human, and white and black people’s DNA differs on average; second, because DNA takes effect in a social environment, and ethnic groups are exposed to different environments.

This means that while we know a lot, and are learning fast, about what genes affect what individual outcomes, we are much less sure if the effects are the same between ethnic groups.

But we will inevitably learn – unless we decide to only do “white people genetics”, which would be a terrible idea medically and scientifically. And once we can estimate the effects of variant X in different ethnic groups, we can ask counterfactuals like “what if this group had the same proportion of variant X as this other group?” Or in other words, how much does a given gene contribute to intergroup differences?

We might then find that the total contribution of genetics is large – in effect, that the natural differences hypothesis is true.

My personal hunch is that this is rather likely. Nobody else need agree. But there are some straws in the wind:
  • Polymorphisms in the MAOA gene vary between human ethnic groups. They have also been linked to behaviours like aggression, risk-taking and addiction. (In 2007, the press got hold of this and coined the phrase “warrior gene”.) Single genes rarely have large effects, but nevertheless the evidence on MAOA seems to have held up reasonably well.
  • This paper shows that different human populations have been undergoing selection for different characteristics, including height, educational attainment and, delightfully, “self-reported unibrow”.
  • Nations with more individualist cultures have higher prevalence of a serotonin-linked allele. (This is just a cross-country association, not a causal link, so the case is still very much open.)
So it is worth asking: what if "natural differences" is true? What would that mean?

When a concept falls apart

The concept of racism is highly moralized. Most people think racism is bad (me too). Since nobody wants to think of themselves as holding wicked beliefs, the prospect that natural differences might exist has led to some odd reactions:
These arguments are silly. But if natural differences exist, how would it actually matter?

There are two contrasting views here. The first is rather maximalist, and seems to be shared by anti-racist campaigners and racists alike. Both think natural differences would change everything. This leads anti-racists to argue that believing in them is wicked; and racists to support them and draw extreme political conclusions.

The second view is more relaxed. Charles Murray and Herrnstein put it like this in The Bell Curve
We cannot think of a legitimate argument why any encounter between individual whites and blacks need be affected by the knowledge that an aggregate ethnic difference in intelligence is genetic rather than environmental.
In essence, it doesn’t matter what the causes of intergroup differences are; we still have to deal with them, and a genetic pathway is just another way to intervene. I lean towards this second view, but with some qualifications.

Let’s take a relatively extreme case. Suppose that we learn there are genetic differences between white and black people in the US that – with society as it is – cause substantively large differences in IQ, educational attainment and income. (To repeat: I entertain this here as a hypothesis. As of 2018, you are perfectly entitled to believe that this is false.)

In this case, would it be justified to discriminate by race, for example by preventing black people from voting? Of course not. We can see this, just by thinking of the really existing group of people, of all races, who actually are unintelligent. We could easily and directly identify these people, without any genetic whizzery. But we do not think that dumb people should not vote, or should be denied other civil rights.[1]

What about individual discrimination of a more low-level kind? Would it be OK to deny black people job interviews because of their race? No, it would not. It is a core tenet of liberalism that individuals deserve to be evaluated on their merits. We should treat people as people because that is the right thing to do.

OK, so it might be wrong, but surely it would be tempting – and sensible! Individual self-interest might easily lead employers to save time by only interviewing the white guy, right?

Not really. Consider three ways to learn about someone’s intelligence – in the straightforward practical sense of “is this person smart enough to do their job well?”
  1. Make the judgment based on their race.
  2. Ask for a cheek swab and estimate intelligence from the person’s genes.
  3. Make them do an IQ test.
  4. Look at their CV, and/or talk to them in an interview.
The respective merits of the last two options are arguable, based on how much weight you place on IQ tests. But it is a known fact that 2. is much worse than 3., and always will be; and I am pretty sure that 1. is worse than 2. and always will be.

2. is worse than 3. because genetics only explains about half the variation in intelligence – any presently available genetic test explains much less than that, but even the best possible genetic test would only get half. 1. is worse than 2. because there is a lot of genetic variation within-groups – more than there is between groups, for most things.[2] So racist hiring will continue to be a bad idea, as well as wrong.

In short, if we find substantive genetic differences between ethnic groups, it will not change how society should function. But it may change our politics.

The first change would be epistemic. Different ethnic groups get different outcomes in many, many ways, and these differences are often used to argue that society or institutions are racist, i.e. that they treat people differently based on race. But if genetics provide an alternative explanation, then this argument is not valid. This is the same argument at group level that has already been made at individual level. Thoughtful people now accept that e.g. class differences in outcomes from education are not, in themselves, a proof that schools serve poor people worse: you have to do more work than that. The same may apply to ethnic differences.

The second change is more unpredictable. Call it the “James Watson wants his job back” factor. (The co-discoverer of DNA lost his position after suggesting that Africans were less intelligent than other races.) For a generation, to hold such views has been thought wicked, and people who express them have paid severe penalties – the most obvious being Charles Murray, who wrote twenty pages on race in a long and prescient book on inequality, and is still being monstered a generation later.

In this respect a certain shift has become noticeable of late. The argument has moved from “they are racist and wrong” to “they may conceivably be right, but they were arguing in advance of the evidence”.[3] But in science, we do not normally blame people for being the first to be right. I wonder whether the condemnation of these people has become an end in itself.

This may be a minor issue – delayed justice for a few intellectuals – but it could become more than that. If a large component of elite discourse on race – what people have been told with great firmness and authority for years – turns out to have been wrong, it is hard to predict the effect on the public debate on, say, immigration.

The lesson here is not to infuse your science with morality, and make your morality depend on unproven science. If you do so, you may paint yourself into a corner.

I will say more about this in my next post, on racism.

Postscript: a qualification

I should make it clear that if the natural differences hypothesis is true, it is true in a very different sense to that of the eugenics of the 20th century. In particular, “race” in the biological sense is not a box you can put someone in; it is shorthand for being a bit more genetically similar to some people than to others. Very few DNA variations are exclusive to any ethnic group.[4] This means that statements about racial groups and (the genetics of) any characteristic are always probabilistic. They will be claims about averages. In particular, they should be interpreted, for any individual, in the light of other information.

For example, I teach many Chinese students; I think it is possible that people of Asian descent have higher IQs on average than white people, for genetic reasons… but this does not keep me awake at night, because I have a PhD and my students don’t, which is much more informative than my ancestry about my capacity to teach them. Or for a more relevant example: if you know that somebody is an immigrant from country X, you know two things about them, first that they were born in X, second that they came to your country. The second fact may easily tell you much more about their intelligence – even their genetics – than the first.


[1] Well, some people do… but that’s a stupid idea itself.

[2] Race-based judgments could have the advantage of picking up environmental differences between groups which genetics don’t capture, such as social disadvantage. But clearly, this has nothing to do with genetic science.

[3] David Reich’s recent New York Times article does this to James Watson, for example.

[4] Interestingly, those that are are almost always to do with physical appearance.

No comments:

Post a comment