Using consistently estimated occupational, wage, and hours equations, we calculate earnings differentials by gender, race, and ethnicity. For example, if the market treated women like men, the average women would have earned $133 more per week so that American women would have earned $338 billion more per year. We decompose the earnings differential into wage, hours, and occupational effects. Occupational segregation explains little of the earnings differential for women, but roughly a fifth of the differential for black and Hispanic men. For all groups, within-occupation wage discrimination is responsible for most of the earnings differential.