Book review of The Genetic Lottery: Why DNA Matters for Social Equality by Kathryn Paige Harden.
Most people know about the top woke biologists of the past. I am of course thinking here of Richard Lewontin (evolutionary biologist mainly population genetics) now known primarily for his Marxism and fallacy about classification, and Steven Jay Gould (paleontologist and evolutionary theorist), known mostly for his semi-fraudulent (?) book on intelligence research (for a fun review, see Jensen and Rushton), and Steven Rose (neuroscientist), another Marxist of otherwise no particular interest.
Now a days we have a new breed. The extreme denialism of the old — recall they denied results from twins (together or apart), adoption studies, and everything else psychology related — has been replaced with a general acceptance of behavioral genetics, even enthusiastic. Here I am speaking primarily of Eric Turkheimer (who I have written about many times), and his former pupil Kathryn Paige Harden. These are both top tier behavioral geneticists. There is also Freddie Deboer who wrote a rather crazy pro-behavioral genetics book called The Cult of Smart, which Scott Alexander already expertly reviewed. And finally, there are minor characters like Daphne Martschenko, who writes stuff like Genetics and Education: Recent Developments in the Context of an Ugly History and an Uncertain Future. The common thread for all these people is that they have given up the object level beliefs of their predecessors, they now accept twin studies, polygenic scores (well sort of, for Turkheimer), and intelligence research. But they have not given up their political convictions, well, they are toned down somewhat, but not given up. Now it is chiefly aimed at the target that was always the main worry: group differences, in particular, race differences. So a kind of trade has happened where the most of behavioral genetics is now mainstreamish science (e.g. published in Nature Reviews Genetics), but the race differences have been further taboorized. Big players like Robert Plomin and Ian Deary have wisely (for their careers) avoided the topics, with a staunch focus on individual differences, chiefly between Europeans. It is a sensible move. Thanks to them we now have good behavioral genetics and intelligence research departments at Universities of Edinburgh, Colorado, Minnesota, Kings College London, and so on. Progress!
Anyway, so I am preparing my course materials for a course I am teaching on human biodiversity, covering the broad fields of differential psychology, behavioral genetics, genomics, evolutionary psychology and anthropology. In one lecture I mention Paige Harden’s book — The Genetic Lottery: Why DNA Matters for Social Equality — as an example of recent wokies, but of course, one should not mention such things without actually reading the book, so I read it. I can state the conclusion up front: it’s a well written book, it covers the essential parts well and in enough detail to be convincing to a semi-skeptical reader. This is sensible because Paige Harden is of course expecting a lot of woke readers, that she wants to convince because she is on a moral crusade. It’s also a very ideological book, I would guess something like 20% of the material is rather obnoxious woke stuff, while the remaining 80% is solid science. Unfortunately, Paige Harden doesn’t color code the text to let the reader know where one ends and the other begins. So that’s my job! So the layout of the book is as follows:
- PART I. TAKING GENETICS SERIOUSLY 1
- 1 Introduction 3
- 2 The Genetic Lottery 27
- 3 Cookbooks and College 45
- 4 Ancestry and Race 72
- 5 A Lottery of Life Chances 96
- 6 Random Assignment by Nature 110
- 7 The Mystery of How 130
- PART II. TAKING EQUALITY SERIOUSLY 151
- 8 Alternative Possible Worlds 153
- 9 Using Nature to Understand Nurture 174
- 10 Personal Responsibility 193
- 11 Difference without Hierarchy 210
- 12 Anti-Eugenic Science and Policy 231
You can see where things start going woke. The main falsehoods, or lies if you think Kathryn is doing it on purpose (jury is still out IMO), begin in chapter 4. So let’s jump into things!
Ecological Fallacies and Racist Priors
By this point, it is hopefully becoming clear why any claims about “genetic” racial differences in intelligence or educational attainment or criminality or any behavioral trait are scientifically baseless. Because existing large-scale GWASs are based on European ancestry populations, our knowledge about how genetics are related to inequalities in life outcomes is entirely about individual differences between people whose ancestry is entirely European and whose self-identified race is likely White. We can’t assume that genetic associations are working the same way in people with different genetic ancestry, in part for fairly technical reasons concerning how the genome is measured and structured. We can’t “compare” the genetics of different ancestry groups using their polygenic indices. We can’t assume that everyone who has the same race shares the same genetic ancestry. Whether we are talking about complicated social phenotypes like education or relatively uncontroversial physical ones like height—modern molecular genetic studies, like the older twin studies, have told us a whole lot of nothing about the causes of racial inequalities.
But even in the complete absence of any genetic “evidence” for genetic racial differences, people commonly mount an argument that, on its face, might sound reasonable. If (1) individual differences in education within White populations are caused by genetic differences, and (2) Black people in the US, on average, have lower levels of education, then (3) doesn’t it just make intuitive sense that the group difference is also caused—at least a little bit—by genes?
It might seem simple to make the leap from “X causes individual differences within a group” to “differences in average levels of X cause average differences between groups.” But from a statistical perspective, assuming that correlations within a group tell you something about the causes of between-group differences is a leap that only fools would make. It is an ecological fallacy.
It’s helpful to explain the ecological fallacy in the context of an example that has nothing to do with genetics. This will make it clearer, hopefully, that my objection to connecting individual differences to group differences is not simply motivated by the fact that I find the conclusion that racial group differences are “genetic” to be unpalatable. I am not brandishing a politically motivated talking point. I’m making a statistical point that applies anytime we are trying to jump from one level of aggregation to another—not just when we are talking about the inflammatory topic of genes and race.
Okay, the usual lies about no evidence, despite overwhelming evidence from many sources: ancestry regression (admixture analysis), transracial adoption studies, variance argument (see below), and yes, polygenic scores (her objections notwithstanding). I say lies because Paige Harden is well aware of this evidence and yet lies about its nonexistence, pretending all evidence is from polygenic scores, that she attacks. This approach is not new, the APA team did the same thing in 1995. That aside, she begins with what I’ve called the deductivist fallacy. Sure, knowing trait T is P% heritable within groups G1 and G2 does not logically imply that the between group gap has a heritability of P%, or some other specific value. There is no strict inference to make. However, it is not irrelevant either as Paige Harden’s claims. We know they have a formal mathematical relationship. Arthur Jensen set out the reasoning already in his 1973 book, and we can quote him because people rarely understand this point well. First, the background:
The visible environmental differences and the visible physical genetic differences between two racial groups may have no causal connection with IQ; both may be merely correlated with some other factors which directly influence IQ. Since we know from studies of the heritability of individual differences in IQ that genetic factors have comparatively powerful effects and environmental factors have comparatively weak effects, is there probabilistically more reason to hypothesize environmental factors as of greater importance than genetic factors in explaining group differences in IQ? The a priori preference for strictly or preponderantly environmental explanations seems to stem more from ideological than from any logical or scientific considerations. Thus, Jencks (1969, p. 29) writes, ‘While a significant number of black children may well suffer serious prenatal damage, Jensen’s evidence suggests that we should probably look elsewhere to explain racial differences in IQ scores. But it hardly follows that we must look to genes. We might do equally well to look at patterns of child rearing.’ This clearly expresses a preference which could determine one’s research strategy, but the preference would seem to run counter to the probabilities suggested by already established evidence. Genes have already been established as having powerful effects on IQ; individual genetic differences correlate about 0-85 to 0-90 (the square root of the heritability) with IQ.
The simple fact is that one cannot, in any strict, formal sense, infer between-groups heritability from a knowledge of within- groups heritability. This is true even if the heritability of the trait is perfect (i.e., h2 = 1 *00) within each group and there is absolutely no overlap of the phenotypic distributions of the two groups. As a clear example we can point to various kinds of grasses. Grown in complete darkness, their colors will vary from white to pale yellow, without the slightest trace of green. The heritability of the color differences is perfect. The same grasses grown in sunlight vary in color from light green to dark green, and here the heritability is perfect. The large color difference between the white-yellow and the green grass is entirely attributable to a difference in a single environmental factor, in this case an obvious one – the presence or absence of visible light, without which the photosynthesis of chlorophyll, the green element in plants, cannot occur. By the same token, environmental differences can completely obscure genetic differences, even to the extent that the phenotypic and genotypic differences are in reverse directions. A genetically light green strain of grass, for example, will be darker green when grown in direct sunlight than a genetically darker green strain grown in the shade. Then there is the third possibility of two genotypically different strains looking phenotypically exactly alike when grown in the same environment. Under some other environmental conditions, al though they are exactly the same for both strains, the two strains will reveal quite large phenotypic differences. For example, one type of golden rod when grown in the shade is dwarfed, while another type is tall; but both are of medium height when grown in direct sunlight (Thoday, 1969). Thus even phenotypic and environmental similarity are not sufficient for inferring genetic similarity. In principle, there can be the same lack of correlation between phenotypes and genotypes for different subpopulations of the human species, even granted a very high correlation between phenotypes and genotypes within the subpopulation groups. But we must inquire under what actual conditions this is likely to be true.
And the math:
Let us look at a clear-cut example of human differences. Two particular subpopulations on the African continent – the Pygmies and the Watusi – differ in mean stature by 5 to 6 standard deviations (in terms of the variability in height of Europeans), which is about 11 to 13 inches (Martin & Sailer, 1959).4 As far as I can determine no one has ever proven that this is entirely a genetic difference or even that genetic factors are in any way implicated. A completely environmental explanation of the difference in stature has not been ruled out by evidence. Certainly the environments of the two groups differ enormously: Pygmies inhabit the rain forests while the Watusi live on the plains; their living habits and diets differ markedly, as do the kinds of illnesses and misfortunes to which they are liable. Yet, despite these facts, it would be difficult to find anyone who would seriously proffer a non-genetic or environmentalist explanation of this difference in stature between Watusi and Pygmies. Why?
The mean difference between the groups is large, being more than | of the total range within either group. (b) The heritability of height within populations is known to be very high, usually over 0-90. Thus, the total range of environmental variations within populations only accounts for about 10 percent of the phenotypic variance. The distribution of environmental effects will have a standard deviation which is equal to the square root of the environmental variance. Taking the total phenotypic within-group variance of height as 5 inches, the environmental variance would be 0-10×5 inches = 0-5 inches, and the standard deviation (SD) of the environmental effects would be y/o-S = 0*71 inches. This means, in effect, that two genetically identical individuals (e.g., monozygotic twins) who differ by 0-71 inches can be said to differ by 1 SD in the effects of all environmental factors influencing stature. We can then express the mean difference of, say, 10 inches between Watusi and Pygmies in terms of the number of SDs by which they must differ in the environmental factors affecting height within a population, if we are to explain all of the differences in terms of these environmental factors. This amounts to 10 inches/0-71 inches = 14-8 SDs difference between the two groups in the effects of environment. Two normal distributions whose means differ by as much as 14-8 SDs are so extremely far apart that there would be absolutely no overlap between the two groups in the environmental factors affecting height. In other words, the probability is practically infinitesimal that even the very largest environmental differences affecting height within either population could begin to explain the 10 inches difference between the two populations. No naturally occurring environmental effects within either population would alter height more than about 6 SDs (which includes 99-8 percent of the total range of a normal distribution) or 6×0*71 inches = 4*3 inches. Thus, the mean difference between groups is something more than twice as large as the largest differences within groups that could be attributable to naturally occurring environmental effects within the groups. This is therefore so highly improbable, that in order to go on entertaining a strictly environmental hypothesis of the cause of the mean difference in statures one would have to hypothesize that the environments of Watusi and Pygmies differ in some very potent unknown factor (or factors), ‘X’, which is present in one population and not in the other and which affects all individuals in the one population and none in the other. Furthermore, if factor ‘X* does not have an equal or constant effect on all members of the population in which it is present, and if the two populations are genetically identical for stature in accord with the environmental hypothesis, then we should expect to find a lower heritability for stature in the population affected by factor ‘X’, since it is a variable environmental effect which acts in the one population and not in the other.
This is the so-called variance argument. Jensen repeated this argument throughout the years (e.g. in the 1998 book, and the 2005 review) and there doesn’t appear to be any particular way around it, so one has to question the premises. One of the points of our meta-analysis of racial differences in heritability of intelligence was to examine one of the escape routes: lower heritability for lower blacks. The math looks like this in table format:
(This table is from Russell Warne‘s excellent book In The Know.) First, in the rows are the within population heritabilities for some trait. For intelligence, a reasonable estimate for adult intelligence is 80%, next last row. Second, if a gap is seen between the groups for some trait, and we want to know how large the non-genetic factors have to be in order to explain it without recourse to genetics, then this can be calculated mathematically as above. For 80%, this value is 2.2 d. In other words, the non-genetic cause(s) must differ by 2.2 standard deviations between the groups to explain the gap entirely. If we allow for 20% of the gap to be explained by genetics, a comparatively minor admission, then still the non-genetic cause must differ by 2.0 standard deviations. Clearly, it is not easy to get around high within population heritabilities. How large is 2 standard deviations? How much do blacks and whites actually differ in the USA? Again, Warne has us covered (from an in press paper that will surely cause a stir):
OK, so the typical SES gap is 0.66 d, but 2.2 d are needed. Can we just add some more stuff that differs, say parenting styles, and so on until we get to a sum of 2.2? No, because the thing that matters is the multivariate distance (Mahalanobis Distance), which takes into account the fact that these gaps in different aspects of the environment are correlated positively, and thus subadditive. If we added up all the variables we could think of, and made some crazy assumptions — they are completely caused by non-genetic factors, are causal for intelligence at r = 1.00 — then they would maybe get to 1.0 d. This is still only about half-way. Basically, the variance argument tells us that a completely non-genetic cause of the black-white intelligence gap at heritabilities of 80% is impossible. Actually, the Jensen table above makes an implicit assumption, which is even worse for the numbers. The entire non-genetic portion of the variance is assumed to be caused by factors that could differ by group. However, we know from decades of research that the non-genetic portion of variance is mostly non-shared variance, which is largely random and is random biological noise to a large degree, could possibly differ by group. The relevant variance is not the total non-genetic, it is the part of the non-genetic that is something that can differ by groups. If we generously split the non-genetic variance by 2, this would get the variance left-over to 10%, which is a correlation or .32, which then needs to differ by 1.0 / 0.31 = 3.2 standard deviations to cause the observed 1 d gap. Even more impossible.
Anyway, back to the book:
The prior belief that White people enjoy better life outcomes because of their genetics is a perniciously persistent one. In the 1960s, the educational psychologist Arthur Jensen speculated that the educational progress of Black schoolchildren would not be improvable beyond a certain point, and certainly not to the level of White schoolchildren, because of the limits imposed by genetics.29 In the 1990s, Herrnstein and Murray blithely presented their hypothesis that at least part of the reason that Black and Hispanic people in America had lower average IQ test scores than White people was because of the genetic differences between them.30 Today, the “race realist” and “human biodiversity” communities post copies of Nature Genetics articles that they believe make the case that there are genetic differences between races that cause differences in intelligence test scores, impulsive behavior, and economic success.
This is not a true summary of what Arthur Jensen wrote. Jensen 1969 was extremely measured, and yet somehow we have 6 decades of lies about what that article said, so we quote it again again:
The fact that a reasonable hypothesis has not been rigorously proved does not mean that it should be summarily dismissed. It only means that we need more appropriate research for putting it to the test. I believe such definitive research is entirely possible but has not yet been done. So all we are left with are various lines of evidence, no one of which is definitive alone, but which, viewed all together, make it a not unreasonable hypothesis that genetic factors are strongly implicated in the average Negro-white intelligence difference. The preponderance of the evidence is, in my opinion, less consistent with a strictly environmental hypothesis than with a genetic hypothesis, which, of course, does not exclude the influence of environment or its interaction with genetic factors.
Next up, she wants to talk ethics. Actually, she mainly wants to talk ethics with this book, while also providing a good introduction:
[quoting Sam Harris from a podcast with her]:
The real question is what is the cause of all of these disparities? The problem politically at the moment is, when you’re talking about White-Black differences in American society … the only acceptable answer in many quarters to account for these differences is White racism, or systemic racism, institutional racism, some holdover effect from slavery and Jim Crow.… It is deeply unstable, because we will find out things, about differences among groups.
These comments invoke an either/or: it is either systemic racism, which White people presumably have a moral responsibility to address, or genetics, which is presumed to be a fixed and deterministic aspect of biology for which no one should be made to feel responsible. As the feminist philosopher Kate Manne put it in her work on sexism, “The unstated premise here is a version of the ‘ought implies can’ principle—possibly weakened to something like ‘can’t even implies don’t bother.’ ”37 The unstated premise of positioning genetics, as opposed to racism, as a cause of racial disparities is to imply that people, particularly White people atop a racial hierarchy, need not feel morally compelled to do anything about changing disparities if their cause is genes.
The crucial flaw at the heart of this thinking is not that it posits the existence of genetic differences between racial groups. As we’ve seen, race is a poor representation of genetic ancestry, but it is not unrelated to genetic ancestry. Nor is the crucial flaw that it links genetic difference with differences in life outcomes. As I’ll explain in this book, there is a plethora of scientific evidence that our genes matter for shaping our selves, not just for our physical characteristics.
The crucial flaw in this thinking is that it presumes that the existence of genetically caused human differences waives our social responsibility to address inequality. As I will describe in the upcoming chapters, the existence of genetic influence—regardless of how it is distributed across socially defined groups—does not impose a hard boundary on the prospect of social change via social mechanisms, nor does it operate as a “get out of jail free” card for our social responsibilities.
Ultimately, I think it is likely that the upcoming avalanche of genomic data from multi-ancestry populations will show that populations differ minimally, if at all, in the prevalence of genetic variants relevant for psychological characteristics, like the cognitive abilities tested by our current battery of intelligence tests. But, no matter how people differ genetically, no matter how those genetic differences between people are distributed across socially defined racial groups, no matter how strongly those genetic differences influence the development of human characteristics, no matter whether those characteristics are physiological or psychological, we are still not absolved of the responsibility to arrange society to the benefit of all people, not just the tiny slice of global genetic diversity that is people of predominantly European ancestry. And that responsibility must be lived out in our policies. That is, our policies should reflect the truth that, as the evolutionary biologist Theodosius Dobzhansky wrote, “genetic diversity is mankind’s most precious resource, not a regrettable deviation from an ideal state of monotonous sameness.… Nonfulfillment of human potentialities is a waste of human resources.”38
This is her main claim of the book. That genetic differences should be minimized by redistribution or remedial … everything in life (should we deny food to American blacks to avoid them getting higher BMIs?). And continued: the race differences are most definitely not genetic, but if they were, we (Europeans!) are still somehow responsible for evening out any problematic gaps. I mean, OK, that is one weird philosophical view one can take. Harris’ reasoning is very normal on the other hand. The idea is just that if one group is keeping another group down with some behavior in some unfair way, they should stop doing that. If we however find that genetics is the cause of some social gap, then obviously, it is not due to the above average group doing something unfair, and there is thus no moral obligation from this fact to do anything. This is not difficult to understand.
I like her very bold prediction at the end. As a matter of fact, there is a new multi-ancestry GWAS method of sorts, polypred. This allows one to fine-tune models based on differential LD patterns to better predict in non-European populations. I probably won’t spoil anything if I mention that we have already tried this new GWAS variant and it shows the same results as all the other predictors. Sorry, next try!
Also, she keeps making vague claims that ancestry is not strongly related to social race. I mean, this is just false. They are nearly interchangeable. I show this in a new paper currently in review. TL;DR the associations between genetic ancestry and social race are in the area of AUC > .90, R2 > 90%. Odd is that she keeps equivocating with US census categories of race and social race in general. I mean, sure, US categories are made to function for the US, and don’t work perfectly well for random small segments of society. But they work reasonably well, and one can always expand them. This is just lazy US-centrism on her part.
The topic overall is difficult Paige Harden tells us:
It can be very difficult to keep these ideas clearly in mind when we discuss the relationship between genetics and social inequality. This difficulty is not an accident. This difficulty is the result of decades of racist thought that has persistently appropriated biology as part of its ideological toolkit for legitimizing a racial hierarchy. In the chapters to come, then, as I make the case for why we should take genes seriously as a cause of individual differences, it might be helpful to refer back to this chapter, to remind yourself why individual differences and racial differences are not the same thing, and why the existence of genetic differences does not obviate our social responsibility.
Evidently, she has difficulties with the topic. Again, typical woke interpretations of her own issues. She is a poor decoupler. In fact, she tells us elsewhere too that her brother is free from the “neuroticism and ADHD symptoms that plague my daily life”, which we might speculate has some relationship to the writing in this book (and Freddie Deboer too, and Angela Saini).
On to nicer matters:
Thick and Thin Causation
In the course of ordinary social science and medicine, we are quite comfortable calling something a cause, even when (a) we don’t understand the mechanisms by which the cause exerts its effects, (b) the cause is probabilistically but not deterministically associated with effects, and (c) the cause is of uncertain portability across time and space. “All” that is required to assert that you have identified a cause is to demonstrate evidence that the average outcome for a group of people would have been different if they had experienced X instead of Not-X. And the most convincing evidence that you know what might have been is to assign people randomly to X or Not-X. (The word “all” is in scare quotes here, because as any scientist of human behavior and society knows, actually isolating the variable of interest from the web of potential confounds, so that one can make an inference about causation, turns out to be an incredibly difficult and delicate operation.) I’m going to call this a “thin” model of causation.22
We can contrast the “thin” model of causation with the type of “thick” causation we see in monogenic genetic disorders or chromosomal abnormalities. Take Down’s syndrome, for instance. Down’s syndrome is defined by a single, deterministic, portable cause. To have three copies of chromosome 21, instead of two, is the necessary, sufficient, and sole cause of Down’s syndrome. The causal relationship between having three copies of chromosome 21 and Down’s is one-to-one, with the result that forward and reverse inferences work equally well. The cause of Down’s is chromosome 21 trisomy; the effect of chromosome 21 trisomy is Down’s. Having three copies of chromosome 21 doesn’t raise your probability of having Down’s; it is deterministic of the condition. And this causal relationship operates as a “law of nature,” in the sense that we expect the trisomy-Down’s relationship to operate more or less in the same way, regardless of the social milieu into which an individual is born.
The example is unlucky. Down syndrome is actually quite complicated because one doesn’t need to have a complete extra chromosome 21, one can have a partial copy. Depending on which part you get an extra copy of, and thus, which genes in triplet, this matters for the phenotype. In fact, it appears to be somewhat complicated. I think a better example here would have been something like lactose intolerance, alcohol intolerance, or sickle-cell disease.
In 1962, the evolutionary biologist Theodosius Dobzhansky23 wrote that “people vary in ability, energy, health, character, and other socially important traits, and there is good, though not absolutely conclusive, evidence that the variance of all these traits is in part genetically conditioned. Conditioned, mind you, not fixed or predestined.” Dobzhansky was right, and the evidence that has accumulated in the decades since his writing has only served to make the case more conclusive.
It is unfortunate that so much energy has been wasted debating this fact, which was evident to Dobzhansky and others a half century ago, because determining that genes are a cause of social inequality is perhaps the easiest part of the research enterprise. A much more difficult question, to which we will turn our attention in the next chapter, is: How?
Then pray tell, Paige Harden, who denied this fact? Which kind of scientists? Why are you silent on the misbehavior of socialists in science, yet mention 100s of times how much you hate hereditarians? Why should non-experts, or anyone really, trust the socialists to get behavioral genetics right this time around?
As I described in chapter 4, it’s important to remember that everything I’m describing here applies to understanding individual differences within groups. The research tools that I’m describing here (primarily twin studies and polygenic index analyses) can’t tell us anything about the causes of average differences between groups.
OK, repeating the claim from before, but it is of course false. In fact, even other twin analyses allow for some models to be tested, which is more than nothing. Check out this 2000 paper for instance, and the 2005 follow-up. Dalliard provided a renew of much of this Jensen debate in his 2019 meta-post.
Paige Harden has some weird quip with the word eugenics. In fact, as you saw with the final book chapter, she wants an explicit ideologically anti-eugenic ‘science’, modeled after
Lysenkoist anti-capitalist anti-racist science by Kendi et al. She of course also has the axe out for poor old Francis:
Beginning with Hereditary Genius,5 the nineteenth-century book by the father of eugenics, Francis Galton, and continuing through the twenty-first century, with books such as Human Diversity6 by the conservative provocateur Charles Murray, eugenic thinkers have implied a specific set of answers to these questions: First, that the causal chain between genetics and social inequality is short and primarily mediated via the development of intelligence. Second, that the causal chain between genetics and social inequality is best understood at a cellular and organismic level of analysis, with intelligence seen as an inherent property of a person’s brain, rather than as something that develops in a social context. And, third, that the alternative possible worlds where this chain is broken are dystopian, requiring either massive state intrusion into people’s home lives or widespread genetic engineering. In short, the eugenic formulation is that genes cause social class the way genes cause Huntington’s, via mechanisms that are universal, intuitively biological, and difficult (if not impossible) to modify.
Writing like a journalist, it is a sport to come up with new negative monikers for enemies. I don’t know why she wants to label these other beliefs part of some eugenic ideology. It doesn’t matter for eugenics which way genes cause intelligence, one can still breed for them whether it is a direct neurological route, or some increased affinity for reading or stimulation. It also doesn’t matter for eugenics whether social policies can also modify (hopefully raise!) intelligence. Why would we forego improvements in genetics just because we can also improve things using other means? We should of course pursue whatever works.
Since Francis Galton, eugenic thinkers have steadily and successfully engaged in a misinformation campaign, convincing people that the reimagination of society is futile. Their propaganda is this: if genetic differences between people cause differences in their life outcomes, then social change will be possible only by editing people’s genes, not by changing the social world. This was the thesis of a bombshell paper written in the late 1960s by the psychologist Art Jensen: “How much can we boost IQ and scholastic achievement?,” he asked, and he used early research on the heritability of academic achievement to answer very much in the negative.1
Paige Harden is projecting. It was eugenic thinkers who were right from the beginning:
- General intelligence (g) is a real thing (valid construct, in modern jargon)
- Intelligence can be quantified and measured
- Intelligence is mostly heritable and we can estimate this using various family designs such as twins
- Intelligence is causal for life outcomes in general, including social status, criminal activity, innovation and so on
- Non-genetic programs to improve intelligence have largely failed
Yet because of socialists like Paige Harden, these truths took many decades to get established, indeed were suppressed. Researchers were harassed (and I mean bomb threats), research funding denied, and so on. They are responsible for holding back scientific progress. Galton did nothing wrong. Eugenicists never said social change is only possible by editing people’s genes. They said just the opposite, this is an obvious strawman. Jensen was also right about the specific enrichment programs of the time. Not too surprising, he was merely quoting the US government’s own evaluation of its large-scale programs from the 1950s and 1960s. Later programs did not fare much better. Paige Harden herself summarizes this evidence!
Reading such assertions, one might imagine that there is a vast repertoire of policies and interventions that have been proven to be effective at addressing social inequalities in education and health, and that are just waiting in the wings to be deployed, if we can only muster sufficient political will. But, in fact, experts in the fields of education, behavioral intervention, and social policy have repeatedly reminded us that, often, well-intentioned efforts to improve people’s lives fail to make any difference at all, and sometimes make things worse.
In the world of education, one can glimpse the paucity of successful intervention research by perusing the What Works Clearinghouse,4 a resource curated by the Institute of Education Sciences, the research and evaluation arm of the US Department of Education. A review of randomized controlled trials (RCTs) conducted by the IES concluded: “A clear pattern of findings in these IES studies is that the large majority of interventions evaluated produced weak or no positive effects compared to usual school practices.”5 Similarly, a 2019 review of 141 RCTs in the US and the UK found that their average effect size was less than one-tenth of one standard deviation (.06 SDs). Reckoning with this track record, the authors suggested one possible explanation: “The basic research on which educational interventions are based is unreliable.… Interventions that are based on insights gained from unreliable basic research are unlikely to be effective even if they are well designed, successfully implemented, and appropriately trialed.”6
Similarly, a report by the Laura and John Arnold Foundation (now Arnold Ventures), a philanthropic organization dedicated to finding “evidence-based solutions” for social problems, summarized, “Studies have identified a few interventions that are truly effective … but these are exceptions that have emerged from testing a much larger pool. Most, including those thought promising based on initial studies, are found to produce small or no effects.”7 David Yeager, an intervention researcher and one of my faculty colleagues at UT, put it this way: “Nearly all past high school programs—tutoring programs, school redesigns, and more—showed no significant benefits on objective outcomes.”8
The conclusion either that most interventions don’t work, or that no one has ever even studied whether they work, extends beyond academic performance. The developmental psychologist Larry Steinberg reviewed the effects of school-based intervention programs designed to reduce teenagers’ alcohol and drug use, condomless sex, and other behavioral risks. An estimated 90 percent of American adolescents have been forced to sit through at least one such program. Steinberg concluded: “Even the best programs are successful mainly at changing adolescents’ knowledge but not in altering their behavior.” He went on to note that failure isn’t free: “Most taxpayers would be surprised—and rightly angry—to learn that vast expenditures of their dollars are invested in … programs that either do not work … or are, at best, of unproven or unstudied effectiveness.”9
These sorts of conclusions, from interventionists who really and truly want to make a positive difference in the world, should be humbling. They should make us think twice before asserting that we already know what to do to improve people’s lives. They should make us realize that a lack of knowledge and a paucity of data really are parts of the problem. And, they should remind us that understanding human behavior, much less intervening to change it, is a hard problem to solve.
Then why the /”#(“# does she complain 6 times about hereditarian pessimism when her own progressive heroes reach the same conclusions?
Let’s talk Scarr Rowe:
Finally, twin studies have shown that the heritability of child cognitive ability is lowest for children raised in poverty and highest for children from rich homes—particularly in the US, where social safety nets for poor families are weaker than in other countries.12 The causal chain from genes to performing better on an intelligence test is not entirely broken, but it is weakened, when children have few material resources in their homes.
No, and she should know that because it has been years since better evidence became available (Figlio et al, see prior post).
There is the talk about heritability and modifiability, of course with the same examples as always:
This hereditarian pessimism about the possibility of social change, however, is based on a fundamental misunderstanding of the relationship between genetic causes and environmental interventions. As the economist Art Goldberger quipped in the late 1970s, your genetics caused your poor eyesight, but your eyeglasses still work just fine.3 That is, eyeglasses don’t just help with the environmentally caused portion of bad eyesight. They help with all of your eyesight, regardless of whether it is genetically or environmentally caused. In so doing, they serve as an outside intervention that severs the association between one’s myopia genes and having functional vision.
In the early 1980s, Leon Kamin, a psychologist who was a fierce critic of behavioral genetics, pushed back against the intuition that genetically caused inequalities of life outcome were any more morally acceptable than environmentally caused ones. His example was phenylketonuria (PKU), a rare disorder caused by a single-gene mutation that impedes the body’s ability to metabolize a protein building block called phenylalanine. Untreated, PKU causes intellectual disability. But high-income countries now routinely screen newborns for PKU, which is treated with a restricted diet low in phenylalanine.
The treatment of PKU with diet, just like the treatment of myopia with eyeglasses, reminds us of a point I was making earlier in this chapter: genetic causes can have environmental solutions. In fact, despite the fact that PKU has a simple and well-understood genetic etiology, environmental solutions currently remain the only solutions. Gene therapy for PKU is not (yet) a reality.19 And we can contrast the simple etiology of PKU with the genetic architecture of highly polygenic outcomes, like intelligence test scores or educational attainment, which involve thousands upon thousands of genetic variants with tiny effects and unknown mechanisms.20 To make matters even more complicated, many of these variants are also involved in phenotypes that are valued differently by society: many of the same genetic variants associated with higher educational attainment, for instance, are also associated with higher risk for schizophrenia.21 The suggestion from some conservative academics that we might edit children’s genomes to increase their IQs is not just scientifically unfeasible; it is scientifically absurd.22
PKU is a very rare genetic disorder, which has an unusually strong gene-environment interaction. This is obviously not how most stuff works. For this analogy to fit, Paige Harden has to claim that low intelligence groups — of whichever race — have some genetic defect that we just need to discover the right diet to fix! Will she? That would be a powerful but not very woke claim. The fixing of PKU with diet is also incomplete since it is very difficult to stick to this diet. Kids still suffer some 10 IQ disadvantage from imperfect treatment effect. This is surely better than 30+ or more without treatment, but it is not a magic fix.
The eyesight example is bad too. Glasses change how well people with reduced visual acuity can do in real life, but does not actually affect the acuity itself when measured without glasses or contacts. Laser treatment would be a better example, but it has its limits too. The general problem with her understanding — which is extremely common — is that she does not understand the nature of interventions to boost socially desired traits and how this relates to variance estimates. When heritability is 80%, so non-genetics is 20%, maybe 10% is non-random stuff, we don’t have too much variance left to work with. The things people suggest changing with interventions are things that vary in the population, and that means it is already included in the variance estimates. Since the variance estimate for this kind of environmental variation is already small, making changes in this kind of environment will not produce large changes in intelligence or whatever trait. In other words, if interventions to boost intelligence involve changing things that already vary between families, say, reading at home or parental styles, the variance component for this would already be larger. The reason PKU and glasses etc. can be consistent with the variance estimates is that this was a new type of variation that was thus not already included in the analysis. Thus, any intervention to boost intelligence needs to involve changing something that does not already vary between families to be effective. There is a dearth of ideas about this. This reasoning is not limited to intelligence, it holds for any trait one wants to modify. Heritability of stomach ulcers is high (60-70%, also high in pigs and rats), but obviously, we could treat them with a treatment that wasn’t used before and thus did not contribute to the variance estimates. Since we now know to treat everybody with this, and thus, there is little variance in this treatment, it does not contribute to heritabilities, and thus, numbers have not changed for the variance estimates much, but people with genetic liability for ulcers are now much better off.
Her argument against embryo selection is even more bizarre. The schizophrenia risk comes from the education part, not the intelligence part, this can be seen with the non-cognitive ability GWAS by subtraction, reviewed by Scott Alexander.
The last box shows the differences in the genetic correlations with psychiatric disorders by intelligence vs. non-cognitive. This is not surprising at all because we already know that schizophrenics have lower IQs! What is really happening is quite obvious: people who are smart and have mental issues seek out a protected environment away from the real world of work where they can team up with other crazies to do a lot of largely meaningless work. That’s why these disorders are common in academics despite their intelligence. She needs to look in the mirror.
But I mean, it doesn’t matter. Nothing says that one cannot select for a trait that has a negative correlation with another trait we want to avoid. A good example here is myopia, which does have a negative genetic correlation with intelligence and we do want to avoid it. One simply constructs an index based on the two traits, and one can simultaneously select for higher intelligence and lower or neutral myopia. This kind of problem is common in animal breeding research. Paige Harden needs to get out more. She can begin with my 1 hour long video summarizing all this stuff.
What about that educational reform she mentioned? Seems a very useful example:
One example is a study that examined the health consequences of an educational reform that happened in the UK in the middle of the twentieth century: Everyone born on or after September 1, 1957 had to stay in school until their sixteenth birthday.30 People who were born right before this cut-off are not expected to differ in any systematic way from people who were born right after that cut-off. Birthday, then, forms the basis of a type of natural experiment testing the effects of being forced by the government to stay in school an extra year.
On average, extra education improved people’s health: people who experienced the educational reform had smaller body mass index and better lung function in adulthood. But not everyone responded equally. The effects of the school reform were largest for people who had the highest genetic propensity to be overweight, as measured by a polygenic index from a GWAS of obesity. Because it had the biggest effects for the most at-risk people, the educational reform narrowed genetically associated inequalities. For people who didn’t experience the reform, the one-third of people who had the highest genetic risk had a 20 percent greater risk of being overweight or obese than those who had the lowest genetic risk. After the reform, that gap shrank to just 6 percent.
They tested 8 outcomes, twice because 2 different PGSs, they got 2x p < .10, 3x p < .05, and they used some complex instrumental design. I don’t see why one should trust their crazy p values in the joint tests considering their near-null findings for the separate betas. Looks like a coding or math error, while the results look like p-hacking. Anyway, so let’s look at the Clark study:
Schooling and social outcomes correlate strongly. But are these connections causal? Previous papers for England using compulsory schooling to identify causal effects have produced conflicting results. Some found significant effects of schooling on adult longevity and on earnings, others found no effects. Here we measure the consequence of extending compulsory schooling in England to ages 14, 15 and 16 in the years 1919-22, 1947 and 1972. From administrative data these increases in compulsory schooling added 0.43, 0.60 and 0.43 years of education to the affected cohorts. We estimate the effects of these increases in schooling for each cohort on measures of adult longevity, on dwelling values in 1999 (an index of lifetime incomes), and on the the social characteristics of the places where the affected cohorts died. Since we have access to all the vital registration records, and a nearly complete sample of the 1999 electoral register, we find with high precision that all the schooling extensions failed to increase adult longevity (as had been found previously for the 1947 and 1972 extensions), dwelling values, or the social status of the communities people die in. Compulsory schooling ages 14-16 had no effect, at the cohort level, on social outcomes in England.
The plots basically just show straight lines, so I won’t repost them here, but do read the study, which is very cool!
I find it peculiar that Paige Harden first tells us heritability is not related to malleability, then mentions that most interventions fail to find anything when rigorously evaluated, and then proceed to cite some weak evidence for interventions! No lesson learned apparently from the own reviews she quotes. Let me spell it out: the prior on something working when you skim some paper on an intervention is very low, meaning a strong prior centered on 0 effect size, meaning that a paper must provide quite strong evidence to convince you, meaning you want to pay particular attention to the statistics for evidence of p-hacking and fishing trips. Extraordinary claims require extraordinary evidence. An intervention working for educational and social matters is extraordinary. Sometimes, though, a light shines through and she has clarity:
Every policy decision involves trade-offs: Investing in one thing, like the word gap, necessarily involves not spending that time and money on something else, which could have been more effective at producing one’s desired ends. Ultimately, all interventions and policies are built on a model about how the world works: “If I change x, then y will happen.” A model of the world that pretends all people are genetically the same, or that the only thing that people inherit from their parents is their environment, is a wrong model of how the world works. The more often our models of the world are wrong, the more often we will fail in designing interventions and policies that do what they intend to do, and the more often we will face the unintended consequences of not investing in something more effective.
Like p-hacking, the tacit collusion in some areas of the social science to ignore genetic differences between people is not wrong in the way that jaywalking is wrong. Researchers are not taking a victimless shortcut by ignoring something (genetics) that is only marginally relevant to their work. It’s wrong in the way that robbing banks is wrong. It’s stealing. It’s stealing people’s time when researchers work to churn out critically flawed scientific papers, and other researchers chase false leads that will go nowhere. It’s stealing people’s money when taxpayers and private foundations support policies premised on the shakiest of causal foundations. Failing to take genetics seriously is a scientific practice that pervasively undermines our stated goal of understanding society so that we can improve it.
Yes! Exactly right. Now generalize this idea to races and ethnic groups too. There is no difference, just a matter of scope.
Back to polygenics and ancestry:
There are practical problems that need to be overcome before genetic data, in the form of polygenic indices, can be more routinely integrated into policy and intervention research. As of this writing, the biggest practical problem is that, as I explained earlier in the book, we do not have polygenic indices that are statistically useful for studying health and achievement outcomes in people who aren’t of European genetic ancestry. In the United States, more than half of public-school children have a racial identity that is not White and so can be reasonably expected to have at least some non-European genetic ancestry. The children who are often most in need of improved educational interventions, then, are the same ones for whom we have the fewest tools in the genetic toolbox. The statistical geneticist Alicia Martin summarized the problem this way: “To realize the full and equitable potential of [polygenic indices], we must prioritize greater diversity in genetic studies … to ensure that health disparities are not increased for those already most underserved.”38
This is just patently false. Every study investigating polygenic scores in non-Europeans find them to have above zero utility. This finding goes back some years, at least to 2015 where Domingue et al studied blacks and whites in Add Health dataset:
So the scores do work in Africans, not as well as Europeans, but not too bad either. In our later studies, we find about half the validity in blacks compared to whites in the PNC dataset, and in the ABCD dataset. These are both very large samples. There is also a meta-analysis on polygenic scores in general across ancestries from 2018. This stuff is common knowledge, and yet Paige Harden tells us the opposite. Mysterious.
Paige Harden agrees with Charles Murray about the ethics of human moral equality:
The bioethics scholar Erik Parens summarized what I believe are the two core concerns that stoke controversy, even outrage, about connecting genetic differences between people to social inequalities like poverty and homelessness: “By investigating the causes of human differences, people worry, behavioral genetics will undermine our concept of moral equality.… Unfortunately, there is an old and perhaps permanent danger that inquiries into the genetic differences among us will be appropriated to justify inequalities in the distribution of social power” (emphases mine).2
Parens’s summary of why people worry about (some) behavioral genetic findings bears striking similarities to Elizabeth Anderson’s definition of inegalitarianism, which I first mentioned in the introduction.3 She writes: “Inegalitarianism asserted the justice or necessity of basing social order on a hierarchy of human beings, ranked according to intrinsic worth. Inequality referred not so much to distributions of goods as to relations between superior and inferior persons.… Such unequal social relations generate, and were thought to justify, inequalities in the distribution of freedoms, resources, and welfare. This is the core of inegalitarian ideologies of racism, sexism, nationalism, caste, class, and eugenics” (emphases added).
Here, again, we see the same two core concerns. First, to link biological difference to social inequalities is to allege that some people are superior to inferior others—a hierarchical view of human worth that starkly contrasts with the egalitarian idea of human moral equality. And, second, such a hierarchical view of humanity will justify inequalities. Rather than poverty and oppression being problems to be solved, these inequalities will be seen as right and natural consequences of human biological superiority.
These two concerns might seem inescapable. When I say that people differ genetically, and that these genetic differences have consequences for their education and social class, for their income and employment and chances of ending up homeless, it might feel impossible for that statement to be interpreted in any way other than as an assertion about a hierarchy of human worth and the inevitability—rightness, even—of poverty.
As the poet and activist Audre Lorde explained, “Much of Western European history conditions us to see human differences in simplistic opposition to each other: dominant/subordinate, good/bad, up/down, superior/inferior.” As a result, she argues, “too often, we pour the energy needed for recognizing and exploring difference into pretending those differences are insurmountable barriers, or that they do not exist at all.”4 The eugenicist ideology is to claim that genetic differences are insurmountable barriers to equality; too often, the response to eugenicist ideology is to pretend that genetic differences do not exist at all.
The problem with all this stuff is that it is pure cope. No one behaves as if everybody is morally equal. We prefer some people over others. No one prefers pedophiles, murderers, burglars over others. They would chose to avoid them if they could, and punish them for their misdeeds. This everybody is morally equal idea is probably some remnant from Christianity and its slave morality. I don’t need to go much into this topic because someone else already did it better. But the TL;DR is that, sure, under ethical realism, people will differ in their moral worth, and this will be related to their behavior and thus of course also their psychological attributes — sadists murderers bad; philanthropists good — which of course relates to intelligence too, as this is related to antisocial behavior in general. Since all of these things are heritable and differ by group on average, so will the differences in moral worth. I think the real scary conclusion they are trying to avoid is that the legal system should not have due process, and that we have to go back to class privileges (literally, different laws for different classes). I don’t think differences in moral worth, even if heritable, mean we need to do away with these things. I think individualism is already the optimal way of designing a legal system. Bryan Caplan even had a Twitter survey with 3400 subjects where the most commonly chosen option was that “All lives are of equal worth” is a… “Noble lie” (42%). So my views here are not actually extreme, just taboo.
Now to the real deep stuff:
Washington’s quest to reclaim intelligence tests as a tool to combat environmental racism mirrors the efforts of other scholars of color and feminist scholars who have argued that quantitative research tools can be used to challenge multiple forms of injustice. The feminist Ann Oakley, for example, argued that “the feminist case” for abandoning quantitative methods was “ultimately unhelpful to the goal of an emancipatory social science.”21 Similarly, Kevin Cokley and Germine Awad, my colleagues at the University of Texas, affirmed that “some of the ugliest moments in the history of psychology were the result of researchers using quantitative measures to legitimize and codify the prejudices of the day.”22 They went on to argue, however, that, “quantitative methods are not inherently oppressive,” and can, in fact, “be liberating if used by multiculturally competent researchers and scholar-activists committed to social justice.”
Woah man! Give this man another PhD.
Finally, on the attack on meritocracy, Paige Harden does some not too bad conceptual work, like a philosopher, but she ultimately does not get the point:
And, if merit is defined instrumentally, then our definitions of merit cannot be separated from our definition of what constitutes a good society. What is considered “meritorious” is simply what brings about the social consequences that we desire. These desirable social consequences include the efficient allocation of scarce opportunities to the people who are most likely to profit from them, and the allocation of jobs to the people who are most likely to do them well. But as Amartya Sen pointed out in his essay on merit, these are not the only social consequences that might be desirable. We might also conceptualize a good society as one that does not have gaping economic inequalities and that does not allow the members of one racial group to dominate all elite institutions. As a consequence, Sen writes, when we are assessing “what counts as merit,” we are compelled to take into stock whether the rewarding of that sort of merit mitigates or exacerbates the economic inequalities or racial disparities we care about: “The rewarding of merit cannot be done independent of its distributive consequences.”
While meritocracy is not ultimately fair in the sense that different people are rewarded differently based on their lucky genetics and other life outcomes, meritocracy is fair conditional on this initial distribution, which was random. Since nature cannot give us the same genetics (then there would be no evolution!), random assignment is the most fair option possible. Anyway, in practice though, the reason we want meritocracy is that it is efficient. There are some things that humans want, and they want them so much they pay others to do them for them. Some of these things they want more than others. To try to get them done more they are willing to depart with more money to get them done. As a simple result of this, some jobs are paid more, and since jobs are tied to human capital traits — which are heritable — then some people get paid more. Why do we like this? Well, humans largely agree on what is valuable, so this allocation of talent via differential rewards gets the things most people want done the most done faster. The most important things are things like technology and health, which is why people producing such things are paid the most. They produce the most utility to other humans. This is a very fair and effective system. As the goal is to improve our lives the most, there is no other better system. This is why meritocracy, that is, capitalism, rocks, and why communism sucks.