We study intergenerational mobility and assortative mating in the US, using an extended-kin design and longitudinally linked records from the 1900-1940 Census. By comparing many different kinship moments we can address the influence of measurement error and study the “anatomy” of intergenerational transmission in terms of assortative and other margins. Our findings suggest that mobility was lower than conventional estimates suggest. Particularly striking is the degree of assortative matching, with the spousal correlation in latent advantages exceeding 0.8 (or 40% larger than the spousal correlation in schooling). Second, we compare the level and anatomy of educational mobility across regions. Despite greater spatial segregation, assortative matching was only marginally stronger in the South than in other regions; instead, high kinship correlations in the South can be attributed to strong intergenerational transmission and schooling being a better indicator for latent status. Finally, we show that early 20th century US had only slightly less mobility, and moderately more sorting, than modern-day Sweden. The considerably larger kinship correlations observed in the US are primarily driven by the fact that schooling better reflects latent socio-economic status in the US compared to Sweden.