In interpreting the positive relationship between spousal education and one's earnings, economists have two major hypotheses: cross-productivity between couples and assortative mating. However, no prior empirical study has been able to separate the two effects. This paper empirically disentangles the two effects by using twins data that we collected from urban China. We have two major innovations: we use twins data to control for the unobserved mating effect in our estimations, and we estimate both current and wedding-time earnings equations. Arguably, the cross-productivity effect takes time to be realized and thus is relatively unimportant at the time of the wedding. Any effect of spousal education on wedding-time earnings should more likely be the mating effect. We find that both cross-productivity and mating are important in explaining the current earnings. Although the mating effect exists for both husbands and wives, the cross-productivity effect only runs from Chinese husbands to wives. We further show that the cross-productivity effect is realized by increasing the hourly wage rate rather than working hours.