Anyway - to all those caveats let me add another one. Bryan Caplan discussed this recently and reported:
"The literature on wage differentials between public and private sector employees spans roughly four decades, originating with Smith's [1976a, 1976b, 1977] seminal series of papers. The core of her analysis is the estimation of conventional human capital earnings functions. For example, in Smith [1976b] she uses 1973 Current Population Survey (CPS) data to estimate for each gender a regression of the logarithm of the wage on various worker characteristics such as years of schooling and race, including a series of dichotomous variables indicating whether each individual worked in the federal, state, or local government sectors (the private sector is the omitted category). For males, she finds wage differentials relative to the private sector of 19 percent in federal government and -4.9 percent in local government. The coefficient on the state government variable is statistically insignificant. The differentials for female workers are 31 percent in federal government, 12 percent in state government, and 3.6 percent in local government."So the exercise and the assessment of differentials is fine (although he goes on to talk about more refinements). We ought to ask whether the research is really finding comparable work in the private sector or not (in a lot of cases, likely not), but what I want to question is how we want to interpret this.
It's 1973, after all.
If we find a 31 percent wage differential in the federal government does that mean:
1. Government overpays workers because it isn't exposed to market discipline, or
2. Private sector underpays women
I'm guessing it's a mix of both and it's hard to know what to do with this without knowing how it splits out between the two. We often use market prices as benchmarks. There can be good reason for doing that, but I'm not sure I would for an analysis of women in 1973.