Changing Patterns of Assortative Mating for Educational, Substance Use, and Anthropometric Traits

Assortative mating (AM) shapes population structure and genetic architecture, yet its temporal dynamics remain incompletely understood. Using UK Biobank (UKB) spouse pairs ( N = 82,398), full siblings ( N = 16,273), and large unrelated samples from the UKB ( N = 333,491) and All of Us Research Program ( N = 228,875), we examined phenotypic and genetic assortment for eight behavioral, anthropometric, and socioeconomic traits. Spousal phenotypic correlations were substantial, particularly for educational attainment ( r spouse = 0.45), drinks per week ( r spouse = 0.42), smoking cessation ( r spouse = 0.33), and BMI ( r spouse = 0.26), and revealed two broad cross-trait clusters reflecting lifestyle and health/socioeconomic profiles. Spousal polygenic score (PGS) correlations were smaller but robust, with the largest effects for educational attainment ( r spousePGS = 0.16), height ( r spousePGS = 0.08), and smoking initiation ( r spousePGS = 0.06). Sibling PGS correlations exceeded the expected 0.50 for all traits except age at smoking initiation and cigarettes per day, indicating genetic assortment in the parental generation. Even-odd chromosome PGS correlations ( θ ), measuring within-person cross-chromosome covariance induced by assortative mating, showed consistent genome-wide signatures of AM across cohorts, with θ = 0.07 for educational attainment, θ = 0.04 for height and smoking initiation, and smaller cross-trait θ values that mirrored phenotypic clustering. Birth‐year analyses revealed cohort-specific changes: in the UKB, θ increased only for smoking cessation ( γ = 0.01), whereas in All of Us we observed increasing θ across birth years for smoking cessation ( γ = 0.014), smoking initiation ( γ = 0.017), and educational attainment ( γ = 0.02). Educational attainment θ remained stable (~ 0.07) across 1930–1970 UKB cohorts but rose from 0.04 in the 1930s to 0.09 by 2000 in All of Us. These findings indicate that genetic AM is pervasive, trait-specific, and, especially for education and some smoking behaviors, increasing across recent U.S. birth cohorts, with implications for how social and genetic variation becomes structured across generations.