Liberal bloggers have cited the work of John Schmitt at CEPR and Bender and Heywood as showing that, despite conservative complaints, state and local public sector workers are actually underpaid compared to private sector workers. I have disputed those results elsewhere, but I’ll let that fight go for today. What I want to look at is whether federal employees are paid more than private sector workers.

Using the same data from those studies that Krugman and Cohn cite favorably (CPS ORG), and the filters and variables from Bender and Heywood*, I’ve taken a look a public sector pay for federal employees. As a first pass I found that federal workers are paid 18.2% more than private sector workers. This includes controls for age, gender, marital status, education, state residence, race, and union coverage. Using a more conservative regression that includes occupational controls and hours worked it comes to 17.8%.

So there’s probably some systematic differences between federal workers and private sector workers that we’d want to account for. One important difference is that a lot of federal employees live in Washington D.C., which I assume has a higher than average standard of living than average. To control for this I re-run the regressions dropping out all workers who live in D.C., and then I tried it again dropping everyone in Virginia as well. Still around 17%.

Maybe federal employees are more concentrated in cities, again with higher cost of living. To control for this I dropped all workers living in a Combined Metropolitan Statistical Area.  Federal employee pay premium goes up to 24.2%. Dropping D.C., V.A., and MSAs gives you around the same.

Just for kicks, if you run it without controlling for union coverage, as I’ve argued you should, the wage premium is 31.8% if you use the geographic restrictions above, and 24.3% if you don’t. In contrast to state and local public sector wage premiums, which I found were concentrated primarily in low education workers with wage discounts for more educated workers, excluding these regressions to workers with college and post college only still has a wage premium of 16.2%.

A note for skeptics: If there’s other controls you would like to see included I’m happy to throw them in and re-run the regressions. There’s no need to call me a hack because I didn’t control for bananas per capita or something, just convince me in the comments that I should and we’ll see what happens. It may be that controlling for some factor I haven’t considered will make these results go away. I’ll grant that it’s certainly possible.

But until then… No matter how I cut it, using the same data and methodology that Paul Krugman and others have stood behind as proving local and state public sector workers were underpaid by around 5% shows that federal workers are overpaid by over 15%.

*I would have loved to do the analysis using Schmitt’s filters and variable definitions as well, but he hasn’t yet responded to my emails