American Community Survey · 2014 to 2023 microdata

Who marries whom in America

Computed from Census microdata:   married couples, linked person by person. Every number is a real weighted estimate, filtered live. Split by men and women, by immigrant vs U.S.-born, by age, and by same-sex vs opposite-sex marriage.

Source: U.S. Census Bureau American Community Survey Public Use Microdata Sample, 2014-2018 and 2019-2023 5-year files joined. "Newlywed" means married in the 12 months before the survey. Spouses are linked inside households (reference person plus spouse), the only linkage PUMS permits, so couples living in someone else's household are not observed. Error bars use the ACS 80 replicate weights to measure each group's design effect rather than assuming a simple random sample.


Years
Born
Couple
Age

What the data actually shows

Each card pins its own filters, shown on its chip, so the headline stays true no matter what the controls above are set to. "Explore this finding" opens the full write-up.

Who lives here, first

Marriage rates mean little without knowing how many people are in each group and how many are immigrants. This is the population everything else draws from.

Over time

Ten years is still a short window and single-year estimates for small groups bounce around, so treat the level as solid and the slope as suggestive. Points resting on fewer than 25 respondents are dropped rather than drawn.

The full who-marries-whom matrix

Share of spouses: marries own group faded row = small sample

The gender gap

How differently women and men of the same group marry. The difference view plots one bar per group instead of two, so the gap itself is what you read.

Measure
Shape

Combine groups into blocs

Groups collapsed into blocs, pooled by actual population rather than a flat average of rows, so East Asian by South Asian reads directly.

Share of spouses: marries own bloc

Every pairing, ranked

One row per group-and-spouse pairing under the current filters. Click a column to sort. n is the unweighted respondent count behind the estimate.