Comparable worth is designed to raise the earnings of women assumed to be penalized for working in female-dominated occupations. Comparable worth advocates assume that the relation between earnings and percentage female in an occupation is due to crowding or other forms of discrimination. An alternative explanation is that the relation stems from women freely choosing different occupations. In other words, preferences are an omitted variable. In our study, we first replicate previous research that has used cross-sectional data to find a negative relation between earnings and percentage female (in an occupation) for both men and women. However, using longitudinal data to control for time-invariant omitted variables, we find that while men's estimated penalty is not reduced, the percentage female penalty falls substantially for women and is not statistically significant. These results imply that estimates of the percentage female effect based on cross-sectional data may be inflated for women. An exception to this general finding is that women with intermittent labor force participation do experience a sizeable penalty for working in female-dominated occupations. Hence, this pattern of results suggests that a comparable worth policy would most likely benefit women with discontinuous employment--perhaps an unintended outcome.