January 30, 2012
By Justin Falk
This analysis used Current Population Survey data from 2005 through 2010 to compare the hourly wages of federal employees and workers in the private sector who have certain similar observable characteristics. In that comparison, we found that the arithmetic average of wages was about 21 percent higher for federal employees than for their private-sector counterparts among workers with no more than a high school education, was about the same in both sectors among workers with a bachelor’s degree, and was 23 percent lower in the federal sector among workers with a professional degree or Ph.D. Overall, federal wages were about 2 percent higher, on average, than wages of similar private-sector workers.
We found that the wages of federal employees were much less dispersed than those of employees with similar characteristics in the private sector—particularly among workers with more education. That aspect of the data causes semilog regressions to generate inconsistent estimates of percentage differences in arithmetic means. Consistent estimates of differences in arithmetic means—obtained using a quasi-maximum likelihood estimator that is robust to distributional misspecification—are substantially smaller than differences in geometric means estimated by semilog regressions. The differences in arithmetic means are more relevant for answering questions about how federal spending would change if federal workers were paid wages equal to those of measurably similar workers in the private sector.
The estimates do not show precisely what federal workers would earn if they were employed in the private sector. The difference between what federal employees earn and what they would earn in the private sector could be larger or smaller depending on characteristics that were not included in this analysis because such traits are not easy to measure. The results apply to the cost of employing full-time full-year workers. The analysis focused on those workers—who accounted for about 93 percent of the total hours worked by federal employees from 2005 through 2010—because higher-quality data were available for them than for other workers.