I’m fairly certain that almost everyone is tired of reading about comparisons of public sector and private sector pay, but a new paper in the Journal of Economic Perspectives is worth reading if you are at all interested in the issue. They lay things out very clearly, and bring some new evidence to bear using micro level data from an employer based survey that includes non-wage compensation. Their main result is that, compared to the private sector, state public sector workers are paid 3-10% more, and local public sector workers are paid 10-19% more.

The paper includes a useful discussion of the issues that divide many papers on this topic and which explain why different researchers often find different answers to these questions. It boils down to what should and shouldn’t be controlled for in the regression analysis. While it’s not always clear which is which, there are broadly speaking to kinds of factors:

….skill-related factors that an individual can transfer from job to job and a second set of variables that are descriptive of the job or sector and possibly indicative of noncompetitive pay differentials such as  rent-sharing.

Studies will differ in whether they consider employer size and union wage premiums as reflecting skill-related factors or non-competitive pay differentials. The authors agree with that I have argued in past blog posts, that neither should be controlled for in the regression:

We treat union status and organizational size as not reflecting worker skills. Controlling for union coverage seems inappropriate, because union wage premia probably do not refl ect ability differences, and those in the public workforce would not likely take their public sector unionization rates with them if they were to move to the private sector…  Troske (1999) tests several explanations of the employer size-wage effect and a significant unexplained premium remains. This and other evidence leaves the door open for the possibility that rent-sharing may be involved. Absent evidence that larger public sector organizational size reflects unobserved ability, we do not control for employer size.

That union status should not be included seems pretty clear, but as the authors acknowledge the firm size issue is not so certain. As a recent CBO report points out, the size premium could reflect a higher degree of specialization at larger firms. However, even if this is the case, the higher specialization may not be transferrable between jobs. In the end, I would agree with the authors of the JEP paper that the best approach is to not control for firm size.

I don’t think regression analysis like this can answer the public sector compensation premium question with certainty, but it is informative. And to the extent that it is useful, I believe these authors take the correct approach and provide the best estimates that the available data can give us. If you’re going to cite empirical analysis on the public sector compensation premium, this is the study you should cite, not other studies that make questionable assumptions or lack sufficient data.