William C. Horrace (University of Arizona,USA)
On the Ranking Uncertainty of Labor Market Wage Gaps
This paper uses multiple comparison methods to perform inference on labor market wage gap estimates from a regression model of wage determination. The regression decomposes a sample of workers' wages into a human capital component and a gender specific component; the gender component is called the gender differential or wage gap and is sometimes interpreted as a measure of sexual discrimination. Using data on fourteen industry classifications (e.g. retail sales, agriculture), a new relative estimator of the wage gap is calculated for each industry. The industries are then ranked based on the magnitude of these estimators, and inference experiments are performed using "multiple comparisons with the best" and "multiple comparisons with a control". The inference indicates that differences in gender discrimination across industry classifications is statistically insignificant at the 95% confidence level and that previous studies which have failed to perform inference on gender wage gap order statistics may be misleading.