Does Living Near People From Different Backgrounds Increase Intermarriage?
Goldman, Gracie, and Porter (2026) examine whether residential segregation helps explain why Americans rarely marry outside their racial or economic class groups. They use linked federal tax records, U.S. Census data, and neighborhood information covering millions of Americans. They find that greater exposure to people from different economic backgrounds substantially increases interclass marriage, while greater interracial exposure has little causal effect on interracial marriage. Their model estimates that eliminating residential segregation would raise interclass marriage from 3.1% to 4.6%, eliminating 35% of class-based marital sorting. White-Black marriage would rise from 2.1% to 3.0%, eliminating only 4% of racial sorting.
Why This Article Was Selected for The Policy Scientist
Marriage patterns matter well beyond household formation because they can reinforce or weaken the transmission of economic advantage across generations. This makes the study especially timely as residential segregation and declining economic mobility remain prominent concerns in the United States. The article makes an important contribution by separating the effects of geographic exposure from preferences in explaining marriage across class and racial lines, extending the large literature on assortative mating and inequality. The population-scale Census and tax data are exceptionally strong, and the instrumental-variable design provides substantially more credible causal evidence than conventional multivariate regression. The findings should generalize broadly within the United States, although less confidently to countries with different residential and marriage patterns.
Full Citation and Link to Article
Goldman, B., Gracie, J., & Porter, S. R. (2026). Who marries whom? The role of segregation by race and class. American Economic Review, 116(10), 3623–3673.
Central Research Question
The study asks why Americans so rarely marry across racial and economic class lines and, more specifically, how much residential segregation contributes to this pattern. The authors distinguish between two explanations for marital sorting. People may prefer partners who share their racial or socioeconomic backgrounds, or they may be willing to marry across groups but have relatively few opportunities to meet potential partners from those groups. The analysis therefore seeks to isolate the causal effect of exposure from preferences and other forms of selection. The central comparison is between class, measured using parental income, and race, with the principal racial analysis focusing on White-Black marriages. The broader question is whether reducing residential segregation would meaningfully increase marriage across these boundaries.
Previous Literature
The study builds on Becker’s (1973) foundational work on assortative mating, which provides the economic framework for understanding how individuals sort into marriages. Subsequent research has documented how marital sorting can affect inequality within and across generations, including Chadwick and Solon (2002), Schwartz (2010), Eika, Mogstad, and Zafar (2019), Buckles et al. (2023), and Olivetti et al. (2024). The authors shift attention from the consequences of marital sorting to one of its possible causes: whether people have sufficient opportunities to encounter potential partners from other groups.
The study also connects to research on segregation, social interaction, and economic opportunity. Chetty et al. (2022) show that cross-class social connections are strongly associated with economic mobility, while Cutler and Glaeser (1997), Ananat (2011), and Derenoncourt (2022) examine broader consequences of segregation. Research on marriage specifically has produced seemingly conflicting evidence. Corno, La Ferrara, and Burns (2022), Gordon and Reber (2018), and Merlino, Steinhardt, and Wren-Lewis (2019) find evidence that increased interracial exposure can affect marriage patterns. Chyn, Collinson, and Sandler (2023), using the Gautreaux housing program, similarly find effects from changes in neighborhood exposure. Other research, including Fisman et al. (2008) and Hitsch, Hortaçsu, and Ariely (2010), documents substantial preferences for same-race partners. The present study attempts to reconcile these findings by distinguishing exposure from preferences and examining both small and large changes in segregation.
Data
The analysis combines unusually comprehensive administrative data from three principal sources: the 2000 and 2010 decennial censuses, federal income tax records covering numerous years between 1979 and 2019, and the American Community Survey from 2005 through 2019. Individual records are linked across sources using Census Bureau identifiers. The primary analysis contains approximately 31.1 million people and focuses on individuals born from 1982 through 1989, with marriage generally measured at age 30.
Economic class is defined using average parental household income when children were ages 13 through 17. Parents are ranked nationally within their child’s birth cohort, and the principal class comparison concerns children from the bottom and top parental-income quartiles. Race is obtained from Census and ACS records, with the principal racial comparison involving non-Hispanic White and non-Hispanic Black Americans.
The geographic information is particularly important. The researchers reconstruct childhood and young-adult residential locations and measure potential marriage markets using nearby census tracts and similarly aged individuals. This permits them to determine not simply whether neighborhoods are segregated but whether individuals were actually exposed to plausible potential partners from other groups. Geographic proximity proves highly consequential: 67.6% of married couples had lived within 50 census tracts of one another at some point at least five years before marriage, despite such neighborhoods containing only a tiny fraction of the national population.
Methods
The principal methodological challenge is selection. A simple correlation between diverse neighborhoods and intergroup marriage cannot establish that exposure causes intermarriage because people who are more willing to marry outside their group may also choose more diverse neighborhoods. The authors address this problem with an instrumental-variable design based on random variation in the sex composition of childhood neighborhoods.
The instrument exploits whether nearby families from another racial or class group happened to have relatively more opposite-sex than same-sex children of similar ages. Because sex at birth is effectively random, these demographic differences generate plausibly exogenous variation in the availability of potential partners. The authors call this “market tightness.” Because most people remain reasonably close to where they grew up, childhood variation in potential partners predicts the composition of their young-adult marriage markets.
The authors conduct extensive validity and robustness tests, including census-tract and family fixed effects. They then incorporate their causal estimates into a spatial marriage-market model based on Choo and Siow (2006). The model accounts for geographic distance, race, class, and unobserved characteristics associated with residential sorting. It is estimated using simulated method of moments and allows the researchers to move beyond localized changes in potential partners to simulate how national marriage patterns might change under alternative levels of residential segregation.
Findings/Size Effects
Intergroup marriage is uncommon. Only 3.1% of people raised in bottom-quartile-income families marry someone raised in a top-quartile family. Among Black Americans, only 2.1% marry a White spouse. If spouses were selected without regard to class from the existing married population, the corresponding interclass rate would be 6.3%. For Black Americans, random selection from the existing married population would produce an 8.6% White-spouse rate.
Descriptively, greater neighborhood integration strongly predicts intergroup marriage. Among people from low-income families, a 10-percentage-point increase in high-income neighbors is associated with a 2.2-percentage-point, or 71%, increase in the probability of marrying someone from a high-income family. Among Black Americans, a 10-point increase in White neighbors is associated with a 0.9-point, or 42.9%, increase in the probability of marrying a White spouse. These associations, however, combine causal exposure effects with residential selection.
The instrumental-variable estimates reveal a major difference between class and race. A 10-percentage-point increase in cross-class market tightness increases interclass marriage by approximately 1.3 percentage points. Given an observed interclass marriage rate of only 3.1%, this is a substantial effect. By contrast, a comparable 10-point increase in cross-race market tightness produces essentially no detectable increase in White-Black marriage. The estimates therefore reject a racial exposure effect comparable in magnitude to the class effect.
The spatial model adds an important qualification. Small, targeted reductions in segregation can increase both forms of intermarriage. A 10-point increase in cross-class exposure is predicted to raise interclass marriage by 3.0 percentage points. A comparable increase in cross-race exposure in highly segregated settings raises White-Black marriage by about 1.0 percentage point, closely matching the 1.1-point quasi-experimental estimate from the Gautreaux Project.
The contrast becomes especially clear in the national no-segregation counterfactual. Eliminating residential segregation by class increases interclass marriage from 3.1% to 4.6%, a 1.5-percentage-point increase. This eliminates 35.2% of the gap between actual interclass marriage and the benchmark in which marriage occurs without class-based sorting. Residential segregation therefore accounts for more than one-third of class-based marital homophily.
Eliminating racial residential segregation increases White-Black marriage from 2.1% to 3.0%. Although that 0.9-point increase is substantial relative to the low baseline, it closes only 4.3% of the much larger gap between observed interracial marriage and the corresponding no-homophily benchmark. Thus, residential segregation explains only a small fraction of White-Black marital sorting.
Conclusion
The study concludes that residential exposure has fundamentally different explanatory power for class and racial sorting in marriage. Limited contact with people from different economic backgrounds substantially constrains interclass marriage. If Americans were geographically integrated by parental income, interclass marriage would increase considerably, although most class-based sorting would remain.
For White-Black marriage, residential geography provides a much less complete explanation. Targeted reductions in extreme segregation can increase interracial marriage, but eliminating residential segregation altogether would close only about 4% of the overall racial marriage gap. The strong descriptive relationship between racial integration and interracial marriage therefore largely reflects factors beyond the causal effect of residential proximity itself. More broadly, the results demonstrate that observed relationships between segregation and social outcomes can substantially overstate the causal importance of exposure when people select neighborhoods based on characteristics that also influence their relationships.
Comments