In direct contrast to conventional wisdom and most economic models of gender differences in age of marriage, we present robust evidence that men and women who are married to differently-aged spouses are negatively selected. Earnings analysis of married couples in the 1970, 1980, 1990 and 2000 Decennial Censuses finds that male earnings decrease with within-couple age difference, regardless of whether the man is older or younger than his wife. In contrast, female earnings increase with within-couple age difference. We argue and present evidence that women in differently-aged couples have higher earnings not because of positive selection, but because their hours of work increase in response to partnering with a lower earning man. We test for negative selection into differently-aged couples using three measures: average earnings per hour in occupation using Census data, cognitive skills assessments from the National Longitudinal Survey of Youth 1979 cohort (NLSY79), and measures of physical appearance from the National Longitudinal Study of Adolescent Health (Add Health). The point estimates indicate negative selection on all of these characteristics, although statistical significance varies by outcome and sample.