The Genetics of Social Class and Intuition Sumps
What really is our prospect for upper ward mobility after all?
In a ground-breaking series of books and research papers, the economist Gregory Clark has amassed a body of evidence showing the intimate links between social class and genetics. In this pursuit he has quantified a large gap between what people say and what they actually do.
Some years ago your humble servant met Professor Gregory Clark at a conference and struck up a conversation. The mild mannered Scotsman was horrified by yours truly when he googled me. I told him not to pay it much mind and that my Googled history was a gift from a hostile intelligence agency. Unbeknownst to Clark I was very interested in funding his work and would be similarly interested in whatever ties his work could have for modern life. Oh well.
That was, of course, before Clark was canceled and drummed out of giving a talk at the University of Glasgow in 2011. His work, his protestors contended, smacked of eugenics. In this political persecution Clark wasn’t unlike Professor Steve Hsu of Michigan State University who lost his senior position at the research university once his views became more well known. Hsu hasn’t returned to his previous esteem.
Now to be fair I’ve long had doubts as to whether or not scientific inquiry was even possible at major universities, place which are seemingly teeming with varying degrees of spies and sloth. You’ll never lose money betting against the courage of academics.
****
Gavan Tredoux is one of my colleagues at Traitwell, the genetics firm I started a few years ago.
We’ve built a lot of apps and we will be adding more apps as time goes on.
It’s totally free if you’ve done 23andme.com and Ancestry.com or any other assessment. If not we will soon offer testing. Come on over to Traitwell.com and see for yourself.
Now Tredoux has written the following account and I’ve added to it slightly and lightly. We hope you enjoy it.
Sociology and Anthropology are notoriously hostile to genetic arguments today, but it was not always so. When the Sociology Society was first proposed in 1904, the founders were intensely interested in genetics, then just emerging as a quantitative science. Francis Galton lent his support, joining the new society and contributing several articles to their new Sociological Papers. (Galton, 1904, 1905). Victor Branford, William McDougall, Sidney and Beatrice Webb (founders of the LSE), G.B. Shaw, HG Wells, Bertrand Russell and others felt the same way as Galton. There was no political colouring to this position: many of these were Fabians, but others were not. (We’ve talked about the Fabians before.)
Indeed, it isn’t really possible to understand society without taking genetics into account. Which is not to say that genetics would ever exhaust our understanding of society. But it must contribute. Without it, the discipline rests on quicksand. Nevertheless, sociology progressively lost its way. By the post-WWII period, it had sloughed off all genetic concerns, and had become wholly enslaved to certain defunct philosophers. Ditto for anthropology, under the influence of Franz Boas and his disciples—leading Carleton Coon to lament bitterly, but without much effect. (Coon and Chapple 1978).
Greg Clark is neither a sociologist nor an anthropologist but rather an economist. He thinks for himself. For the last twenty years or so he has been conducting his own research program into the interplay between genetics and social class, especially assortative mating. To this he has related a range of outcomes. Two books document his findings: A Farewell to Alms (2007) and The Son Also Rises (2014). (Clark, 2007, 2014). An unusual data set from English sources has been amassed recording occupational status, higher education attainment, simple literacy, house values, directorships of companies and a ‘multiple deprivation’ index. Well over a million marriage records have been harvested from genealogy enthusiasts.
At times these records go back far in time, to around 1600 for some of the sets. They vary a lot in coverage, which is why there are multiple sets used. Some striking findings from this data set are incorporated in the recent paper ‘The inheritance of social status: England, 1600 to 2022’. (Clark 2023). We will have to make do with this until the third book is released, which may contrive another Hemingway pun: For Whom the Bell Curve Tolls? though publishers are said to be skittish about that.
Assortative mating is the tendency of people to marry those who match them in some sense, by social class, age, intelligence, height, ethnicity and other factors. Generally consid- ered, even inbreeding is a kind of crude assortative mating: merely marrying your relatives will tend to match you by default on many characters, because relatives resemble each other more than random members of the population, especially if they have kept up inbreeding for a long time. This is not to say that people don’t occasionally make unusual choices. Princes may marry milkmaids, every now and then. Generally they do not. That is no surprise to readers of Jane Austen, Charles Dickens, William Makepeace Thackeray, or Leo Tolstoy.
Rather than reading novels (with the possible exception of Hemingway), Clark has been trying to quantify exactly how much people do assort when they marry, for a key trait: social class. It has long been known from twin and other family studies that social class is itself heritable. Clark’s data in this study, which includes many family relationships, even to the fourth-cousin level, and extends to 422,374 individuals, shows that persistence in social status and assortment in mating is really strong. Between parents and their direct children, the correlation in social status is calculated directly to be 0.79. With each additional generation, the relationship declines regularly by a factor of 0.79. Yet even fourth-cousins remain related in status. This persistence, which demonstrates that social mobility is modest though certainly always present, is due to assortment. Without that, fourth cousins would barely be related on social class at all. (Hold that thought, for it explains a great deal about class and breeding systems, and we will return to it.)
Assortment is estimated to be about 0.8. This corresponds to an underlying genetic correlation of perhaps 0.57, using a simple additive model developed by RA Fisher as long ago as 1918. (Clark fits this model to his empirical data in log space using weighted least square regression to find both these estimates.) All this may strike some as too rich. How do humans get to be so good at divining the underlying genetic basis for status? It is after all latent and not directly visible.
Data for other phenotypes, such as educational attainment, where direct genetic data is available, suggests lower levels of phenotype assortment and a rather lower level of genetic correlation. It is a problem that greatly exercises Clark in his paper, but it is probably enough for busy people to note that social class is likely a rather special kind of phenotype, composed in practice of a slew of indicators. It is this combination that gets effectively at the latent genetic factor. In the same way that a bundle of twigs is much harder to break than a single twig, a bundle of indicators is much stronger than a single one. The errors will tend to cancel out, so the correlations will be higher. And humans have doubtless been selecting for social class for a very long period of their evolutionary history, so are quite good at it.
It should really come as no surprise that genes are heavily implicated in assortative mating and all aspects of social class. The class setup—often treated as an unfortunate affliction—is best understood as a breeding system, one of several responsible for flexible and responsive regulation of that state of dynamic tension between adaptation and variation that is the essence of evolution. The system has co-evolved with those genes, in the same marvellous manner that culture has in general. We are wired for class in the same way that we are wired for culture, as Mark Pagel has put it. For many this is an uncomfortable conclusion, but they are, like everybody else, out there assortatively mating every day and selecting ‘good school districts’ for their children, whatever they may say to the contrary.
Clark takes pains to point out that some may also be surprised that the tremendous social changes of the last few hundred years have not altered social mobility and status persistence in any significant way, as his comprehensive data set firmly establishes. It is an expectation that should not be taken seriously to begin with. Think of social class as an underlying continuous variable, normally distributed, which we arbitrarily box. Within any generation there will be great variance along that dimension. Succeeding generations will shift the mean of that distribution, in this case upward as living standards improve. There is no reason to expect that the variance will be affected by that. The root of this misconception is the idea that social class is some sort of aberration or affliction to be overcome by social progress, like poverty.
Clark works hard to perform cross-checks on his conclusions using alternate data sources. One obstacle he faces is that in the historical records women do not have occupations until quite recently. This makes it hard to work out assortment for them in those sets. An ingenious solution is to use the correlation between the groom and his father, along with that between the groom and his father-in-law—for both of which there is empirical data since they are men—to deduce the correlation r between the groom and bride. However Clark’s argument, which produces an estimate for r by dividing the known male correlations and simplifying the terms (the precise details are beyond our scope here), relies on the transitivity of the correlation from groom to bride to father-in-law. It is counter-intuitive, but correlations are not in general transitive absent special conditions. We cannot assume that they are transitive: they may or may not be. So Clark’s argument here requires rather more justification than he provides. This scotches his cross-check but does not affect the main flow of his argument.
There is an obvious reaction to Clark’s evidence. It has been applied to evidence of the same kind since 1865. Which is to point out that some (unspecified) third variable could be arranging the association of social status with genetic relatedness. This sort of objection can always be made with observed (as opposed to experimentally manipulated) variables, regardless of the particular subject at hand. It is never accompanied by evidence that such a force exists, let alone an active research program designed to understand its nature and operation. It remains an inchoate possibility, mooted but never produced. The philosopher Daniel Dennett once wrote an entertaining book containing his favourite toolbox of Intuition Pumps, or aids to thinking. The argument above is by contrast an aid to anti-thinking, which we might call an Intuition Sump. A vacuum is strategically placed to draw off thought into the bottomless aether. We are properly concerned with the best deductions we can make from the evidence before us, and the theories which incorporate them. We are not in the business of selecting, from the boundless set of all possible theories, one which pleases us more, regardless of evidence. Serious people must confront Clark’s evidence and account for it, and for genes in general. Habeas corpus and no deus ex machina.
Another common response to consideration of genetics here is the spectre of reduction- ism. Now, scientists do not generally think that reducing complex systems to simpler terms is a bad thing—that is what a model is—but the idea is anathema to many social scientists. This was the impulse that drew the early sociologists away from biology and genetics in the first place. One cannot, they argue, ignore the emergent properties of a system, where the whole is greater than the sum of the parts. Emergence is certainly a real phenomenon. The human body is more than a few gallons of water and some chemicals. But it is nevertheless impossible to understand the human body while being ignorant of water and its properties. The emergent properties of people are implicit in their DNA, which builds all the infrastruc- ture permitting emergence. It has arrived there through a process of natural selection on the instructions through selection on the emergent properties, since different instructions lead to differences in emergent properties. Moreover those instructions go well beyond individuals to construct what Richard Dawkins has called extended phenotypes, well downstream from genes but affected by them nonetheless. We must include mating systems, of which social classes are an example, among those extended phenotypes.
There is of course nothing, Dr. Pangloss, in the process of status formation and persis- tence that guarantees a good outcome. The whole system may easily become ossified and brittle. History is littered with examples, a Golgotha of bones from the faculty lounges of their day, in matching Che Guevara T-shirts. Unexpected shocks shatter the ossified. Flexibility is necessary. Genes must flow where they are needed.
Let us indulge in one of Dennett’s Intuition Pumps. Suppose that Carl Friedrich Gauss II washes up on the shores of Lilliput. Using his unusual combination of first-class math genes he solves a great many problems, but after he is gone, his descendants must carry on. Unfortunately the Lilliputians rigorously practice purely random mating. Within a few generations those marvellous alleles have been dissipated to the winds. They have a positive effect, as each Lilliputian eventually becomes slightly less obtuse, and they are particulate so they don’t just vanish. But to start a fire you need to concentrate heat, not dissipate it. The alleles of Gauss II will certainly be slowly and inexorably selected for, as RA Fisher showed in the 1920s, but much less efficiently than they would be in a system where they are concentrated. Using that marvellous human creation, culture (itself sustained by genes) concentrated collections of those alleles will propagate solutions to problems that others are not capable of solving themselves but can at least understand and exploit. Overall welfare is increased far more rapidly. Lilliput falls to these marauding invaders, who proceed to put them, well, under the pump.
Social mobility is thus under rigorous control, and it appears that this, at least in part, is exactly what the class system regulates, though it may fail in that task. Those who seek to change the world have, aside from the obligation to prove that they will improve it, the more fundamental obligation to understand it. The outcome may well be An Immoveable Feast. Or, perhaps, To Have Not and (Continue to) Have Not.
Carey, Gregory (2002). Human Genetics for the Social Sciences. London: Sage, 2002. Clark, Gregory (2007). A Farewell to Alms. Princeton, NJ: Princeton University Press, 2007.
— (2014). The Son Also Rises: Surnames and the History of Social Mobility. Princeton, NJ: Princeton University Press, 2014.
— (2023). “The inheritance of social status: England, 1600 to 2022”. In: PNAS 120.27
(2023).
Coon, Carleton and Eiliot Chapple (1978). Principles of Anthropology. Reprint. New York:Krieger, 1978.
Darlington, Cyril (1964). Genetics and Man. London: George Allen and Unwin, 1964.
— (1969). The Evolution of Man and Society. New York: Simon and Schuster, 1969. Dawkins, Richard (1982). The Extended Phenotype. The Gene as the Unit of Selection. London: Oxford, 1982.
Fisher, R. A. (1918). “The correlation between relatives on the supposition of Mendelianinheritance”. In: Transactions of the Royal Society of Edinburgh 52 (1918), pp. 399–433. — (1930). The Genetical Theory of Natural Selection. London: Oxford University Press,
1930.
Galton, Francis (1865a). “Hereditary Talent and Character: Part I”. In: Macmillan’s Magazine 12 (68 June 1865), pp. 157–166.