The CPS survey is a national sample of the American population and contains about 20,000 white respondents and about 1,900 black respondents. Only the estimates for white workers are presented here.

In a multiple regression analysis, Tam uses occupational sex composition and a series of other independent variables (see Note 3 in Table 1) to predict hourly wages. This analysis is done separately for each gender.

Table 1. The effect of sex composition on hourly wages: unstandardized regression coefficients (standard errors in parentheses) White White Women Men Sex Composition -. 187 -.094 (.016) (.019) Control Variables INCLUDED INCLUDED R2 .470 .503 Notes: Dependent variable: the natural logarithm of hourly wages. 2. Sex Composition: the proportion of workers who are women in the respondent's occupation. 3. Other independent variables included are Geographical Location (South, West, Midwest), SMSA (residing in a standard metropolitan statistical area), veteran, union, schooling, work experience and its square, industry (22 dummy variables), and full-time. These estimates are not the focus of our interest, thus their estimates are omitted from the table