Assortative Mating: The Next Generation

Assortative Mating: The Next Generation

Everyone's talking about a new paper that analyzes data about assortative mating -- the tendency of people to marry others like themselves -- in the U.S. In recent decades new career and educational opportunities have been made available to women, so this trend has naturally accelerated when it comes to traits related to income; in general, for example, couples share similar educational credentials. The authors calculate that if mating were random rather than assortative, inequality (as measured by the Gini coefficient) would decline about a quarter.

I'm proud to announce that we're two steps ahead of the game here at RealClearPolicy. Last month, in our White Papers & Research section, we linked to this draft of a paper from Henry Harpending and Gregory Cochran. The paper asks: When there's assortative mating for a given trait, and that trait is at least partly genetic, what happens in future generations?

The answer: It depends on exactly how strong the mating pattern is and how genetic the trait is, but at least in theory these mating tendencies can harden into genetic castes quite quickly.

Here's an example the authors provide using population-genetics models. Say that in a given society, all the people who are above average for a given trait abruptly decide to form a mating class, and so do all the people who are below average. If the trait is purely genetic, a single generation is enough to create strikingly different (albeit still overlapping) groups of people. The left column here shows this scenario, while the right column factors in 5 percent random mating between classes (X axis labels are standard deviations):

As the authors concede, that's a very extreme scenario -- but as they also argue, and as the new paper shows, modern trends in college attendance have shoved American society in this direction. Here are some other ways this can play out, again, depending on how genetic the trait is and/or how assortative the mating patterns are. The left column plays out the above scenario for an extra two generations, the middle column is a situation where the tendency is reduced by half (e.g. a trait that's half genetic but assortative mating is still very strong), and the column on the right is a scenario where the tendency is reduced by nine-tenths:

My guess is that assortative mating by income in the U.S. falls somewhere between the middle and right columns (probably closer to the right). Note that in all of these examples, class mobility -- the part where the two bell curves overlap because the children of one class have abilities that place them in the other class -- declines with each successive generation.

There's an opportunity here for someone to take the results of the new paper, use twin, adoption, and genomic studies to estimate how genetic income-related traits are, and work all that data into Harpending and Cochran's models.

Update: I made a spreadsheet pulling together some basic data from the study. If you lump together all people who've been to college, regardless of whether they graduated, 12 percent of all couples were college/college in 1960, compared with 70 percent noncollege/noncollege and 18 percent mixed. Today it's 46 percent college/college, 29 percent noncollege/noncollege, and 27 percent mixed. The rise of college obviously facilitated an entire class of paired college attendees, though it also created a rising number of mixed marriages. Interestingly, however, in roughly two-thirds of mixed marriages (12 percent of all marriages in 1960, 16 percent today), the college half of the couple did not graduate. Of course, a complication with all of this is the increase in out-of-wedlock childbearing; my guess is that focusing on fertile couples rather than married couples would increase the proportion of noncollege/noncollege and possibly mixed couples.

Robert VerBruggen is editor of RealClearPolicy. Twitter: @RAVerBruggen

Show commentsHide Comments

Related Articles