Marital Sorting versus Stochastic Sorting

Inequality in household earnings has increased enormously in the last decades. The variance of joint male and female earnings has gone up eightfold since 1960. This has often been attributed to marital sorting, the fact that married partners tend to have more similar levels of education. In this paper we investigate how this contrasts with stochastic sorting, the fact that earnings have become more volatile. We characterize conditions for stochastic sorting and propose a measure for marital sorting (or mismatch). Contrary to popular belief, we find that marital sorting has not changed once we account for the changing distributions of education: the distribution of educational attainment has shifted to the right (and more so for the females than for the males). The lion share of the increase in the variance in household earnings stems from the fact that over time: 1. the variance of earnings has increased; and 2. that now there are many more highly educated earners whose variance is substantially higher the that of the low educated. Marital sorting on education contributes nearly nothing. There is a modest increase in the correlation of earnings (from slightly negative to 11%), which could be attributed to marital sorting based on (unobserved) characteristics other than education. But that does not take away from the fact that 80% of the increase in the variance of household income is due to stochastic sorting.