Assortative Mating

Assortative mating (or marital sorting, marital homophily) is the tendency of people to form partnerships with others similar to themselves along traits such as education, income, class background, race, and religion. Gary Becker’s 1973 marriage-market model launched the economics literature: matching is a market, and positive assortative matching emerges from complementarities in household production. The concept matters beyond demography because who marries whom shapes the intergenerational transmission of advantage — marriage is one of the channels through which inequality replicates itself. ¹

Measurement is harder than it looks

Comparing sorting across cohorts or countries runs into a mechanical problem: when the marginal distribution of a trait changes (say, more people attend college), raw intermarriage rates change even if underlying preferences don’t. Chiappori, Costa Dias, Meghir & Zhang (NBER WP 33354, 2025) show that different summary measures — odds ratio, normalized trace, likelihood ratio — can yield opposite conclusions about whether sorting by education rose or fell between the 1950s and 1970s US birth cohorts. They give these measures axiomatic characterizations and a structural economic interpretation of the odds ratio, then apply them to show how US marital sorting by education actually changed while both educational attainment and its labor-market returns were shifting. The methodological lesson generalizes: any claim that “homophily is increasing” needs a measure that separates changes in the trait distribution from changes in matching propensity. ²

Exposure versus preference

Why is intergroup marriage so rare? Two families of explanation: people prefer own-group partners, or people rarely get exposure to other groups. Distinguishing them matters for policy — if exposure binds, integration policies (inclusionary zoning, school redistricting, housing vouchers) move marriage patterns as a side effect; if preferences bind, they mostly don’t.

Goldman, Gracie & Porter (NBER WP 35140, 2026) provide the cleanest causal decomposition to date, using linked US tax and Census records covering roughly the full population of 1982–1989 birth cohorts:

• The raw gaps are large. Only 3.1% of people from bottom-quartile parental income families have a top-quartile spouse; only 2.1% of Black Americans born in the 1980s have a White spouse. Observed intermarriage rates are 2–4× below no-homophily benchmarks. Marriage is also intensely local: 67.6% of married couples lived within 50 census tracts of each other five years before marriage, and the probability of marrying someone from one’s own tract is ~20× that of marrying someone 50 tracts away. ³
• The instrument. They exploit random sex-ratio variation in childhood neighborhoods — whether other race/class groups happened to have more opposite-sex children of similar age — as an instrument for opposite-sex “market tightness” in young-adult marriage markets. The instrument’s distribution matches Bernoulli sex assignment almost exactly, an unusually checkable identification assumption. ⁴
• Class and race diverge. A 10pp increase in cross-class market tightness raises interclass marriage by 1.3pp (42%) — exposure matters for class. The equivalent estimate for White-Black marriage is 0.006pp, with a 95% CI ruling out effects above 0.36pp — exposure does not detectably move White-Black intermarriage. (Exposure does raise interracial marriage for other racial pairings, with magnitudes resembling the class result — the White-Black boundary is distinct.) ⁵
• Scaling up. Embedding the IV estimates in a Choo–Siow transferable-utility matching model with geography, they simulate a no-segregation counterfactual: interclass marriage rises from 3.1% to 4.6% (closing 35% of the homophily gap), while White-Black intermarriage rises only from 2.1% to 3.0% (closing 4% of its gap). Residential segregation explains more than a third of marital sorting by class but less than 5% by race. ⁶

Stakes

Marital sorting compounds other inequalities: it concentrates (or disperses) household resources, social capital, and the environments children grow up in. The Goldman–Gracie–Porter results imply that desegregation policy has meaningful downstream effects on class intermarriage — a channel for mobility policy that operates through the marriage market rather than the labor market — while the persistence of the White-Black gap despite exposure points to mechanisms (preferences, norms, contact conditional on exposure) that residential integration alone does not reach.

Cross-domain connections

• stable-matching — the same formal object from the other side: Gale–Shapley and Choo–Siow model the mechanics of two-sided matching; assortative mating is the empirical pattern those markets produce, and the GGP model literally builds on Choo–Siow
• filter-bubble — the exposure-versus-preference decomposition has the same shape as Bakshy et al.’s algorithm-versus-individual-choice decomposition of ideological exposure; in both cases the preference term dominates the naive intuition
• early-1970s-structural-break — marital sorting is a transmission mechanism in the inequality literature’s account of widening dispersion since the 1970s (assortative mating by education rose as returns to education rose)
• norms — the GGP race result (exposure doesn’t move White-Black intermarriage) is evidence for a norm-level rather than opportunity-level constraint on the matching market

Open questions

• What drives the White-Black distinction? The GGP design rules out neighborhood exposure as the binding constraint, leaving preferences, contact conditional on exposure, or non-residential frictions — the authors note Ciscato (2024) attributes high same-race marriage largely to search frictions, and the two results are not formally inconsistent.
• Does the class effect generalize beyond parental-income quartiles to wealth, education, or occupational class?
• How do online dating platforms — now the modal way couples meet — reshape the distance gradient that makes neighborhoods matter (35% of online daters restrict search to their own neighborhood)?

Sources

• Benjamin Goldman, Jamie Gracie, and Sonya Porter 2026 — Who Marries Whom? The Role of Segregation by Race and Class — the exposure-vs-preference causal decomposition (NBER WP 35140, April 2026)
• Pierre-André Chiappori, Monica Costa Dias, Costas Meghir, and Hanzhe Zhang 2025 — Changes in Marital Sorting: Theory and Evidence from the US — measurement theory for marital sorting; education-sorting trends (NBER WP 33354, January 2025)

Footnotes

1. Pierre-André Chiappori, Monica Costa Dias, Costas Meghir, and Hanzhe Zhang 2025 — Changes in Marital Sorting: Theory and Evidence from the US ↩

2. Pierre-André Chiappori, Monica Costa Dias, Costas Meghir, and Hanzhe Zhang 2025 — Changes in Marital Sorting: Theory and Evidence from the US ↩

3. Benjamin Goldman, Jamie Gracie, and Sonya Porter 2026 — Who Marries Whom? The Role of Segregation by Race and Class ↩

4. Benjamin Goldman, Jamie Gracie, and Sonya Porter 2026 — Who Marries Whom? The Role of Segregation by Race and Class ↩

5. Benjamin Goldman, Jamie Gracie, and Sonya Porter 2026 — Who Marries Whom? The Role of Segregation by Race and Class ↩

6. Benjamin Goldman, Jamie Gracie, and Sonya Porter 2026 — Who Marries Whom? The Role of Segregation by Race and Class ↩