Via Matthew Yglesias:
Among college-educated people, in particular, the tendency is not so much to marry within your community as to marry within your educational cohort. Every once in a while you do see a college graduate married to a high school dropout (my parents, for example), but it’s quite rare. Since incomes are normally measured on the household level for statistical purposes, it matters quite a bit to big-picture national trends. In particular, “assortative mating” of this kind is a major driver of household-level income inequality.
How big? Jeremy Greenwood, Nezih Guner, Georgi Kocharkov, and Cezar Santos report in a new paper that if in 2005 the matching of husbands and wives had been random, the Gini coefficient—the most common summary measure of income inequality—would have been 0.34 rather than its real-world 0.43, a difference of almost 25 percent.