Gender Differences in Pay

As an example, consider a case where a pay differential, estimated across all members of an alleged class of employees, masked considerable variation between employees in different jobs.

An opposing expert estimated a pay regression for hourly employees working at a service providing company and identified an aggregate pay differential to the detriment of women. Among other problems with the regression, this estimated pay difference is at best an average measure. It cannot inform whether common patterns in outcomes are present throughout the proposed class.

We looked beneath the aggregate result and found considerable variation between jobs in the estimated pay differentials between women and men. The pay differentials among employees in the same job varied in size, from very close to zero to differences of approximately 3% in either direction. Of the twenty- three positions analyzed, just eight showed differentials to the disadvantage of women.

In the remaining positions women were actually advantaged relative to men.
The absence of a uniform pattern – illustrated by the size and direction of the estimated differentials varying between jobs – was confirmed through a statistical test that rejected the equality of these differentials across jobs. This demonstration of variability undermined the notion of a common class of employees.