Assortative mating is the tendency for individuals to marry partners with similar traits. Positive assortative mating means similar individuals pair (the highly educated marry the highly educated); negative assortative mating means dissimilar individuals pair (high earners marry low earners, consistent with pure comparative-advantage specialization). The direction and strength of assortative mating on a given trait depends on whether that trait is a complement or substitute in household production — a structural prediction of Becker's (1973) theory of marriage and Weiss and Willis's (1997) empirical investigation.
In Becker's model, marriage is a matching market. Individuals on each side of the market rank potential partners by the household surplus they can generate together. A stable matching (no blocking pair) requires that no two individuals outside the assigned match could both benefit by pairing with each other instead. The key result: if traits are complements in the household production function, the unique stable matching is the positive assortative one — the highest-trait man matches the highest-trait woman, and so on down the distribution. If traits are substitutes, the stable matching is negative assortative.
Two approaches:
Correlations at time of marriage (Weiss and Willis 1997): Direct measurement of spousal trait similarity using retrospective survey data at the date of marriage. Avoids endogeneity from post-marriage investment (e.g., education completed after marriage). Disadvantage: limited to variables measured at marriage.
Cross-sectional household correlations (Mare 1991 and successors): Measure correlations in education or income across currently married couples in census or survey data. Subject to selection into marriage and divorce — couples who divorced are not observed, and if divorce is correlated with educational mismatch, the observed correlation in intact marriages overstates initial assortative mating.
A growing literature (following Schwartz and Mare 2005; Greenwood et al. 2014) argues that rising educational assortative mating since the 1960s has mechanically contributed to rising household income inequality, because:
This mechanism operates independently of changes in individual-level inequality — even if the individual earnings distribution held constant, stronger assortative mating would increase household income variance.
Educational homogamy is one of the strongest predictors of marital stability in U.S. data. The same-education diagonal in Table 2 of Weiss and Willis (1997) shows 5-year divorce probabilities falling from 23% (High School [HS]/HS) to 9% (College/College). This operates through two channels: (a) similar-education couples face fewer preference conflicts over child investment, leisure, and consumption; (b) educational similarity signals initial match quality, buffering the marriage against subsequent earning-capacity shocks.
The near-zero sorting on earning capacity (0.2 correlation) has an important implication: in a world of rising female labor force participation and rising female relative earnings, there is no counterbalancing tendency to sort into marriages that accommodate this (because earning capacity is a substitute, not a complement). Rising female earning capacity therefore mechanically raises divorce risk through the outside-option channel identified in Match Quality and Marital Dissolution, independently of preferences.
Because household income inequality is partly a product of assortative mating, policies aimed at reducing individual earnings inequality may have smaller effects on household welfare inequality than expected if assortative mating is intensifying simultaneously. Conversely, anything that reduces educational stratification in the marriage market (e.g., residential integration, mixed-socioeconomic status (SES) schools) would dampen the mating-driven component of household income divergence.