Behavioural genetics: should we be worried?

Eugenics got a bad name after the second world war. It got associated with pseudo-scientific theories under which people at the bottom of the societal ladder were branded as hopelessly deficient for supposedly inalterable biological reasons. Societies’ less successful were, quite literally, seen as ‘untermensch’ (under-people) and the ‘science’ of heritable poverty, height, and intelligence was used in the public campaigns of the nazis and others to stigmatise gypsies, jews, homosexuals, vagabonds, and others as being biologically deficient and hence a kind of ‘disease’ for which the only ‘cure’ was annihilation or selective breeding.

Modern behavioural geneticists of course are not like the old eugenicists. They are ‘merely’ looking at the relation between genetic proximity between people and how much their height, their intelligence, their mental disorders, their criminal behaviour, and their body size resemble each other. They talk of alleles, single nucleotide polymorphisms, linkage disequilibrium, heritability, phenotypes and CNVs, not the inherent inferiority of this recognisable group of people over that group. They do not advocate selective breeding, unless it is of mice or plants of course. One cannot find a single paper by behavioural geneticists in Nature or Science, where they appear often, that calls for genetic tests to be used for potential migrants or selective baby-bonuses.

And yet, I find this field somewhat creepy in its treatment of social processes. The same techniques and language is used to ‘track genetic diseases’ in plants as is used to track ‘genetic causes of behaviour’. The same techniques of breeding ‘mice with particular traits’ is advocated to ‘find’ the biological basis of what are essentially social outcomes. The same old penchant of looking for supposedly inalienable biological causes of what are changing social constructs is on display.

The literature I am talking about is vast, and the interested reader is invited to look at some reviews of it here (written by insiders of this literature), or else by psychologists here. There is little point in regurgitating those reviews, which are both very informative and open-minded. What I will do below is say why I think you don’t actually have to worry about this crowd: yes, these geneticists are indeed looking for the genetic recipe of the successful human, but their quest is, at the moment, going nowhere.

The view of humanity that emerges from some behavioural genetics is one whereby there is a strong and virtually direct genetic causality of nearly all social outcomes, independent of local environment or social processes. See page 8 of this explanatory note for the latest techniques used in the modern Genetic Complex Trait Analysis (GCTA) studies, involving direct genetic causal effects and an ‘error term’ which captures all the rest. Their statistical models encapsulate the idea that the successful are successful because of their genes, the poor are poor because of their genes, the stupid are stupid because of their genes, the small are small because of their genes, the schizophrenics and autists are thus due to their genes, etc. Everything else is an ‘error term’ that is presumed completely unrelated to genetics.

Let us not mince words: I see this as a dog-breeders view of humanity, where traits can be bred by combining humans with desirable traits dogs with desirable trait and discarding the other dogs. The important questions for the modern genetic dog breeder is then one of finding the particular genes that cause the particular outcome. Once one has found these, the optimal breeding program can be guided more efficiently. Or, less dramatic, particular ways can be found to ‘improve the functioning of this or that gene’ to get the desired outcome. I shudder to think of the fate of the other humans if the recipe for the successful humans is found. So I sincerely hope this recipe is not found. I hope it doesn’t exist either, but that is another matter.

Let me give you some of the good news immediately: according to their own methods, that of linear additive effects of a million genetic variations (single nucleotide polymorphisms, SNP, in their lingo), people can become as tall as a tree or even attain a negative height. The fact that you don’t see this is in their view because of the pure accident that no human has yet been fortunate enough to get all the good genes, nor unfortunate enough to get all the bad ones. In reality it is of course because their methods are not, strictly speaking, correct but merely approximations of deeper complexity. So forget about seeing these geneticists as dealing in the pure scientific truth. If you thought that, you have been duped.

Let me give you even better news: despite grand claims in their papers that behavioural geneticists have ‘explained’ 70% of human variation in intelligence, height, etc., this is not actually the case. They have not found actual genetic recipes at all that explain this amount of variation. Rather, they infer from their techniques that an unknown combination of around a million small genetic influences should be able to explain 70% variation in intelligence, height, etc.

How does this ‘inferring’ work, you may ask? Well, it is a complicated statistical story of latent variables and association matrices, but the simplified version is this: they essentially count how many genetic variations people share over and above what they would randomly share as part of the same population. They then only look at people who are less close to each other than about 2.5%, which means close family is kicked out. So they look at humans who share around +1% or -1% of each other more than expected as part of the same population. They then find, on average, that people who are one percent more alike, are about 0.7% closer to each other in terms of outcomes, whilst those who are less like each other by an extra percent are 0.7% further away from each other. That 0.7% then gets blown up to 70% ‘variance explained’. So in reality you are looking at ‘explaining’ 0.7% of pair-wise variation, and even that is via an unknown combination of influences. In economics we would say that they have hit upon a ‘proof by error-term’. It is basically a con-trick.

What is wrong with that, you may ask? Well, for one it means one doesn’t know the genetic recipe supposedly responsible for outcomes, so the 70% is pure smoke-and-mirrors. Moreover, that 0.7% that the results are thus actually based on can come from a huge variety of environmental factors, such as historical advantages of this or that group making them higher and smarter because they had more training and more resources, not because of their genes. Genetic similarity might simply pick up a shared environmental advantage, such as belonging to particular ethnic or wealth groups, and that then gets hugely amplified. Statistically speaking, their independence assumption, ie that genetic information is not related to any form of relevant shared environment, is the thing that does the real work in their methods and it is ridiculous: how on earth one can have genetic similarity (via shared actual ancestors!) between people without the high likelihood of a relevant shared environment is beyond me.

Does this crowd not realise this problem? Well, again, they pull a couple of smoke-and-mirror tricks to bury this kind of issue. They talk a lot about G*E interactions by which they mean genetic influence that work in particular environments, but they dont really entertain the notion of interactions, nor do they really deal with the more fundamental possibility that one is essentially looking at pure E effects for which genetic information is then nothing more than a marker.

Why don’t they try and account for environmental factors, you might wonder? Well, for one, these techniques come from the world of plant breeding and animal breeding, where of course the researchers perfectly control the shared environment. There is no such thing as an advantage in terms of wealth or educational habits that is passed down from one plant to the other or from this cow to that one. Hence the problems of variation between humans throws up problems that don’t come up with plants.

Second, the business of actually accounting for environmental circumstances is hopelessly complex. After all, this is what economists, demographers, and psychologists have been doing for decades, and ‘we’ have been struggling to prove causality. Where-ever we look, we find incredible complexity and non-linear interactions. It is a nightmare to measure and analyse social data: it is exceptionally hard to measure ancestral wealth, educational habits, the geography in which people have lived in their lives, the people they have interacted with, the shared cultural and political environment they had, etc.

It is not just a lack of available surveys that nail down what you really want to know, it is also the incredibly complicated causal spaghetti that is the problem: with every ‘environmental control’ one would be trying to account for, comes in a whole set of measurement problems and of course the problem that an environmental variable might itself be merely a poor proxy for something else, as when the ownership of an umbrella might say more about the weather in the area than about the wealth of the owner! So it is quite understandable that these geneticists don’t want to have to deal with the issues that have bedevilled social scientists for decades, and thus shove those issues under the carpet. Indeed, even when they throw in a couple of environmental ‘controls’ they are still only scratching the surface of the difficulties. They would essentially have to solve all the problems of empirical social science on top of their genetic puzzles.

On the purely genetic side, there is more good news which is that they can’t find individual genes explaining much of intelligence, obesity, or mental health disorders. They hit false positive after false positive, by which I mean that a genetic variation accidentally found to relate to an undesirable outcome in one population turns out not to be related to the same undesirable outcome in another population. Indeed, the field has now turned to the somewhat hopeless hypothesis that there are no single genes or even small groups of genes responsible for low intelligence and other undesirable outcomes, but that undesirable outcomes are due to thousands of small negative and positive genetic influences on ‘performance’. I would call this the ‘death by a thousand cuts’ hypothesis. They tell themselves they will soon uncover all those thousands of small influences if other people give them enough money to keep going (which they probably will), but the outsider should see this emerging ‘polygenic hypothesis’ as great news. As Evan Charney says, they have been reduced to ‘chasing ghosts’.

Even if the ‘ghostly hordes’ materialise, finding them would make ‘genetic treatments’ or ‘breeding’ a virtual impossibility: the rate of new small genetic variations in each generation of humans would swamp the usefulness of knowing about today’s influences, and one would need a fantastical amount of embryos by potential parents to be able to select on just a few dozen of these small influences. Hence, in a way, the current state of that literature is such that their usefulness for potential breeders is very much in question.

The news gets even better. The supposed genetic causes of social outcomes turn out not to be constant over social strata. This is what the newer studies like the Bates 2011 study on the genetic causes of intelligence are about: they are finding that, unlike for higher socio-economic groups, genetic proximity doesn’t explain much of the variation in intelligence amongst lower socio-economic groups. Better still, those studies themselves appeal to the idea that the ‘real’ drivers of intelligence are learning (such as in university) and that the inability of genetics to explain intelligence amongst lower socio-economic groups is essentially because there is no genetic determinism of intelligence in that group, rather the determinants there are social processes. I am not a religious man, but it is a real ‘thank god’ moment, though of course the behavioural genetic literature as a whole might just shrug such studies off as blips on the road to finding the full genetic roadmap between genes and outcomes. Well, good luck with unpicking thousands of small genetic effects interacting with an ever changing social reality, I would say to them! It is the kind of Research Mountain that makes macro-economics look easy by comparison.

Another piece of good news is that they can’t be right at the group level: there is no way that genetic determinism really can explain, on its own, fast changes in group averages of the outcomes of interest. Unless one believes that aliens with different genes have surreptitiously invaded the planet, genetic determinism cannot explain the spectacular increases in obesity rates in our society, or the large increases in measured intelligence, or the huge decreases in crime rates, or the large changes in measured mental disorders. In essence, the behavioural genetics crowd is fairly impotent in explaining these changes in group behaviour, as of course also noted by other social scientists. The review in 2012 by the psychologist Nisbett and his colleagues into IQ noted this as an “apparent contradiction between strong heritability effects on IQ and strong secular effects on IQ”. Less nicely put: these geneticists cannot explain changing averages. Though some do try, and, as I have argued before in the context of obesity, their attempts to explain a change by something unchanging is going nowhere.

And, as if the good news just didn’t want to dry up, these geneticists find themselves unable in their techniques to deal with genetic variation that is rare, such as SNPs that are shared by less than 10% or 1% of the population. Why not, you may ask? Technically, because they have many more genetic bits than they have humans and rare genetic bits would then start to identify particular humans, meaning one runs out of degrees of freedom as well as missing covariance information between individuals. In simpler terms: they can’t. Why is this great news? Well, because it means whatever is found has to be shared by a large proportion of the population. This knocks out many of the genetic factors most likely to have truly large effects are likely to be rare, as they will presumably be had by people who wont procreate. It simply seems unlikely that 10% of our populations can be exceptionally socially unsuccessful due to this or that shared genetic effect.

Which brings us back to dogs: the whole point of getting to know which genes lead to desirable traits is because one wants to breed particular average traits. The complete inability of the behavioural geneticist crowd to explain changes in group averages of traits is great news, in my opinion. Their inability to find the actual genetic markers that, across populations, would predict who is going to end up on the top of the social scale and who would end up at the bottom, is similarly great news. Their continued penchant to believe they will soon find the magic genetic formula that ties thousands of small genetic effects to thousands of interacting behavioural processes and other genetic effects is almost charming in its naivety. Good luck, I say to them.

So, at the start of my sojourn into this literature, I found the geneticist crowd slightly creepy in its penchant to treat undesirable social outcomes in the same way as it treats diseases. Having seen how little progress they have actually made though and how much less useful they are in coming up with policy levers than social scientists have been, I am breathing a sigh of relief. There will almost surely be some genetic influence on the distribution of social outcomes, and there are interesting puzzles in term of just how those influences go, but I see little danger at the moment that dog breeders will soon get a manual by which to ply their trade on humans. They might find a couple of genes for this or that mental trait, but the story that geneticists will soon know the recipe for creating a socially successful human being, should be seen for the consumptive fantasy that it is and looks to remain for the foreseeable future.


Author: paulfrijters

Professor of Wellbeing and Economics at the London School of Economics, Centre for Economic Performance

%d bloggers like this: