Parent-child similarity is a much studied topic. So much studied that various fields have come up with many different terms for the same thing: intergenerational elasticity, intergenerational persistence, intergenerational mobility (antonym) and so on. Of course, most of this research implicitly assumes that this similarity is due to non-genetic factors, mainly parenting factors (the blank slate/nurture assumption). The parental investment theory of economics has been particularly damaging. It is based on the observation that children of large families tend to have worse outcomes later in life, say, lower incomes. This is then interpreted as due to the splitting of parental resources, in the broad sense of attention and money, as the number of children increases. The policy advice was then that people need to have smaller families to get more economic development. In communist China, they went so far as to implement a 1-child policy based on such thinking (it didn’t have much effect though, and it was also based on fears of overpopulation). In reality, the main reason that this pattern exists is that higher human capital parents have smaller families, and their children genetically inherit these traits giving this observation. Genetic confounding.
The same people who study parent-child similarity also tend to be interested in inequality of various sorts, usually the ‘easily’ measurable things like income and education. Decade after decade they unhappily report that, yes, children of more educated parents tend to be more educated themselves, and this is a Bad Thing, and we need to give more money to early interventions That Definitely Will Work This Time. Normally the reply to this is to say akshaualy, parent-child similarity is mostly due to genetic similarity, that similarity between parents and children due to genetics is actually a sign of meritocracy, not something bad.
But there’s another angle, which is to study parental similarity too. Invariably, it has been found that parents are more similar than average in a population. We call this assortative mating (parents are positively sorted). It is quite important for the study of family similarities (that is, behavioral genetics) because models that ignore assortative mating will be biased, usually in the way of giving too little credit to genetics, and too much credit to the family environment. But it is also important for inequality in itself because parental similarity enhances variation in the next generation for both genetic and environmental reasons. This is because it induces a correlation between the genetic and/or environmental impact of one parent and the other parent, pushing the resulting further out each side of the normal distribution (larger standard deviation). Changes in inequality over time, then, could just be due to changes in the degree of assortative mating over time. Furthermore, insofar as assortative mating reflects genetic similarity, the heritability will change because the genetic variance changes with the degree of assortative mating for heritable phenotypes. This actually means that evolution will be faster since the speed of evolution depends, in part, on the heritability. The high degree of human assortative mating compared to other species has been speculated to be an adaptation so that humans could better keep up the evolutionary arms race with faster-reproducing organisms (other animals, bacteria, viruses).
Has assortative mating for human capital related phenotypes changed over time? Well, it is hard to say! There are a number of reasons for this:
- Measuring education, income, occupational status etc. is difficult without substantial error or short-term variation.
- The distributions of education etc. have changed over time.
- Different countries have different systems for measuring education etc., making them hard to compare.
To try to deal with some of these problems, Seb Jensen and I have a new paper just published looking at education specifically:
- Jensen, S., & Kirkegaard, E. (2024). International Cross-Temporal Meta-Analysis of Assortative Mating for Educational Attainment. Evolutionary Psychology, 22(2), 14747049241249072.
Previous studies have found a high degree of assortative mating for educational attainment (r = .56). However, this can be confounded by cohort effects or country effects, where certain nations may have more pronounced assortative mating than others. In addition, method variance regarding how educational attainment is measured may also result in heterogeneity of effect sizes. Effect sizes were gathered from various datasets and from academic literature, resulting in a large collection of effect sizes (k = 1498, n = 9,159,098), spanning 84 different countries. Assortative mating for educational attainment was stronger than what previous literature suggested (r = .66, [.64, .68]), largely due to the fact that assortative mating for educational attainment is stronger when latent methods are used. The strongest predictors of assortative mating for education between countries were individualism (r = −.61, p < .001) and HDI (r = −.56, p < .001). Assortative mating over time was found to vary by region. Capitalist Europe experienced an increase in assortative mating for education, while Communist Europe experienced a decrease. The United States had a non-linear trend in assortative mating for educational attainment, as it decreased from 1875 to 1926, increased from 1926 to 1945, decreased from 1945 to 1958, increased from 1958 to 1977, and decreased from 1977 onwards.
First, measurement. There’s 2 main ways to measure educational attainment: 1) by an ordinal scale (primary school, vocational training, 2-year bachelor etc.), or 2) years of education. These correlate well, but not perfectly. It turns out that it matters which one is used, or whether both are used as indicators for a latent variable. Using the American General Social Survey (GSS), we found spousal correlations of: Categories (ordinal) 0.56, Years of education 0.61, Latent 0.64.
Second, the data. There is a lot of data from around the world. The map gives you an impression:
Studying the changes over time gets even more complicated. It turns out that the assortative mating has been increasing in USA, but only if one uses the ordinal measurement (categories of education), not if one uses years of education or the latent variable:
Why is that? Well, it turns out that the two methods have been getting more strongly correlated over time:
Thus, it seems the reason assortative mating has increased for educational categories is just that these have become better indicated of ‘educability’ over time.
If one tries to control away these method changes, and try to estimate the current degree of assortative mating for education by country, it looks like this:
It seems that assortment is actually weaker in more developed countries, maybe contrary to some expectations. Averaging the numbers by world regions confirms this pattern:
Given that there were some 60+ countries with meta-analytic estimates, one could try to explain these regional differences in term of, say, Hofstede’s cultural dimensions. Or you know, any other plausible sounding dimension one can think of. There’s a lot of options because most things correlate rather well between countries:
AME is the spousal correlation for education as estimated from the meta-analyses. The 3 most correlated variables with this are: individualism -0.61, power distance (a kind of egalitarianism or opposition to hierarchies) 0.53, and the human development index (HDI) -0.56. The scatterplot for individualism looks like this:
Based on this plot, we might note that USA has unusually strong assortative mating for a developed country, which probably relates to their higher social inequality (e.g. Gini of income). This might just be due to racial tendencies in assortative mating, though. (I don’t know why Hungary and Luxembourg are also high.)
Since there are too many predictors relative to the number of cases here, ordinary multiple regression will not really produce sensible results. One cool option in these edge cases is Bayesian model averaging (BMA). In this method, all or a large number of the possible models are fit, and these are then sorted by some kind of model fit metric (adjusted R², BIC or whatever). The coefficients from the top X (say 500) models are then averaged so that one can see which variables tend to present in the presence of others, while trying many different model comparisons (a kind of anti-p-hacking approach). Two rounds of this BMA resulted in the finding that both individualism and power distance cultural values best predict the degree of assortative mating for education, and the other variables don’t seem to do anything (this included the gini for income/wealth). Granted, the sample size was a maximum of 60 countries due to missing data, so the results should be taken with some skepticism. Of course, this kind of analysis cannot tell causality by itself. We can reason, however, that it seems difficult for the degree of assortative mating to affect cultural values, but maybe there is some trickle down effect long-term. To see this, think of cousin marriages and the relationship to culture of this practice, something that Steve Sailer, JayMan, HBDChick and academics like Joseph Henrich have written about for decades.
Going back to the changes over time, there are quite a few apparent ones. If one groups Western countries, one can get this plot:
USA has the most data, and thus has statistical precision to show the most non-linear patterns. It does show a rather strange pattern. However, there’s not much of a long-term change from 1870 to now. Communist countries, however, showed a general decline during communism. This perhaps reflected the communists’ interference with who was allowed to get educated and who was not. If so, then we should expect a rebound in the post-communism (1990-now) period, which is a maybe. In fact, many countries show seemingly odd changes over time:
Israel has a rise and decline. Why? One could come up with some complicated historical narrative about waves of immigrants from this or that area, but one could also just say: given the biases in the data, and the changes in biases over time, and the imperfect nature of the controls for these methodological differences, it is expected that countries show relatively random changes over time and these don’t necessarily mean anything.
What’s the implications of these findings? Well:
No substantial net increase in assortative mating has taken place in the United States. Previous findings of increases (Eika et al., 2019) can be attributed to categories becoming increasingly valid predictors of educational attainment. This lack of a change mirrors prior literature (Gihleb & Lang, 2020) which has also noted that increases in assortative mating for education found in the literature are sensitive to changes in statistical methods. These findings challenge the argument of Murray (2013), who contended that increasing educational assortment in the United States caused increases in class divides in the United States. While it could still be possible that class divides are increasing within the United States, this cannot be ascribed to assortative mating for education due to the relative stability of this metric across time. This concords with other researchers who studied trends in assortative mating in other regions such as England (Clark & Cummins, 2022) and Norway (Heath et al., 1985) and observed stable levels of assortative mating. Any reports of changes in assortative mating for education across time that do not test whether the method of measuring educational attainment affects the results should be viewed with skepticism, as these effects are sensitive to changes in methodology.
Is it inconsistent with Charles Murray’s Coming Apart? Kinda sorta. I haven’t read the book, but from what I can tell based on the Wikipedia summary (and maybe this is a very bad idea given how much The Bell Curve is misrepresented), it seems Murray is talking about how the upper class (educated professionals) are sticking to traditional family values and good conservative morals (while voting increasingly for anti-Christian, socialist politicians), while the under class is increasingly poorly behaved as in single parenthood, unemployment etc. This change seems real enough, but doesn’t necessarily imply any differences in assortative mating over time, as much as it often relates to physical separation. Murray likes to talk about how in the past, people lived in relatively small towns, and the towns had both smart and dull people, and they got along, went to the same schools, even if they were economically unequal. Now a days, the smart people leave such towns (or don’t grow up there at all), move to Manhattan or San Francisco. 80 IQ and IQ 120 people basically don’t meet each other. This leads to quite incorrect perceptions of the true distribution — diversity — of human capital and behaviors. For instance, if most academics have no real experience with actual lower class Black criminals, because all the Blacks they know are educated professionals (affirmative action notwithstanding), they are probably going to be more lenient on crime and hostile to anti-Black attitudes. In fact, academics have come up with a kinda sorta opposite of reality idea called the contact hypothesis. In this model, the more exposure to some outgroup (immigrants, foreigners, other races etc.), the more one will truly understand them and be more friendly to them. You can see where this might apply. Suppose all you know about people from [outgroup] is the propaganda the media tells you, which is that they Very Bad People. Then you meet some of them in real life, and they turn out to be OK People. The false impression given by the media was dispelled. However, what if the media tells you lies, but they are in the opposite direction? Well, then you should see the opinions get more negative, not more positive. There is some evidence for this (Does Exposure to the Refugee Crisis Make Natives More Hostile?):
Although Europe has experienced unprecedented numbers of refugee arrivals in recent years, there exists almost no causal evidence regarding the impact of the refugee crisis on natives’ attitudes, policy preferences, and political engagement. We exploit a natural experiment in the Aegean Sea, where Greek islands close to the Turkish coast experienced a sudden and massive increase in refugee arrivals, while similar islands slightly farther away did not. Leveraging a targeted survey of 2,070 island residents and distance to Turkey as an instrument, we find that direct exposure to refugee arrivals induces sizable and lasting increases in natives’ hostility toward refugees, immigrants, and Muslim minorities; support for restrictive asylum and immigration policies; and political engagement to effect such exclusionary policies. Since refugees only passed through these islands, our findings challenge both standard economic and cultural explanations of anti-immigrant sentiment and show that mere exposure suffices in generating lasting increases in hostility.
I think we can thus reformulate the contact hypothesis in the light of stereotype accuracy and propaganda. When the media gives you a misleading impression of the behaviors of some group, and you then get some real exposure to representative members of this group, your opinions will get closer to the truth, which means they will get more or less hostile depending on the direction of the media bias. I submit this is a fairly obvious model of social interactions. Murray’s fear — in my interpretation — is that due to the lack of exposure to members of different social status levels in the USA, media bias will have an increasingly large effect, causing unnecessary political strife and hostility.
Anyway, I think our new meta-analysis of measured assortative mating for educational attainment will be helpful to many. I think it’s currently the best there is until someone digs up even more datasets in a few years.
