Abstract Predominantly female occupations pay less than “male” occupations, even after adjusting for skill demands. The devaluation perspective says that this is because gender bias influences employers’ decisions about the relative pay of occupations. The longitudinal implication of the devaluation perspective is that when occupations feminize, their relative pay should go down. In the queueing or relative-attractiveness view, occupations’ reward levels affect their sex composition, with less attractive occupations going to women because employers prefer men and can get them in occupations that pay well. To test these views about change over time, we use a fixed-effects model with lagged independent variables and data from the 1983–2001 Current Population Surveys. There is only slight evidence that the feminization of occupations lowers their wages, and no evidence that a fall in occupations’ relative wages leads to feminization. But percent female is associated with lower wages in every year. We speculate that in the early history of new occupations and organizations, there was a causal relationship between gender composition and wages, and it has been frozen in by institutional inertia in relative wage structures.