Measuring and testing for gender discrimination in physician pay: English family doctors

In 2008 the income of female GPs was 70%, and their wages (income per hour) were 89%, of those of male GPs. We estimate Oaxaca decompositions using OLS models of wages (income/hours) and 2SLS models of income. The elasticity of income with respect to hours is 0.91 for female GPs and 0.29 for male GPs, so that log wage models are misspecified. The conventional discrimination measure (the unexplained difference in mean log income) is sensitive to the counterfactual (30% using male returns vs 11% using female returns), to the use of OLS vs. 2SLS (19% vs 11%, female counterfactual), but not to dropping insignificant female interactions. The unexplained pro-male difference arises because the pro male difference in regression constants offsets the pro-female difference in the effect of hours on income. We propose a set of new direct tests for within workplace gender discrimination based on a comparison of the differences in income of female and male GPs in practices with varying proportions of female GPs and with female or male senior partners. The direct tests produce mixed results. An indirect test, comparing GPs actual income with the income they report as an acceptable reward for their job, shows that female GPs are not more likely than male GPs to report that their actual income is less than acceptable income, whereas GPs from ethnic minorities and overseas qualified GPs are significantly more likely to do so.