This study investigates the effects of unobserved earnings-related endowments on marital outcomes, focusing on potential gender differences. Using a less stringent, more plausible assumption that monozygotic twins share more similar endowments relative to dizygotic twins, we develop a novel identification strategy from the twins experiment. Using unique Chinese twins survey data, we find that men with genetic endowments related to higher earnings marry earlier and have younger, taller wives; such women marry later and have older husbands with higher education and income. Results cast doubt on the conventional assumption of independence between unobservables and observable marital attributes.