We use census and household survey data to document China's educational assortative marriage and its evolution between 1990 and 2009. Empirical results suggest that men are increasingly likely to marry with women with similar education levels in China since the early 1990s, which is also true for urban areas and for different provinces. We then calculate the counterfactual Gini coefficients that would prevail if marriage matching was random in terms of education. For China in 2005, the inequality of per capita household income would drop from 0.508 to 0.476 if marriage was random. For urban areas in 2009, assortative marriage in education also increased the Gini coefficients by around 2 percentage points (from 0.316 to 0.337). The decomposition exercise shows that the increase in the return to education is the major contributor to the increase in urban household income inequality between 1990 and 2009, and the change in the assortative marriage pattern plays a minor role.