Which Government Employees?

by Reihan Salam

My friend Mike Konczal has written a very interesting post, but I worry that it might be confusing for casual readers:

Something that is strongly implied in the Bowles-Simpson chairman’s mark is that government employees are overpaid (see here).

The see here link takes you to Mike’s interview with Greg Anrig. I don’t think it tells us much about what Bowles-Simpson implies, strongly or otherwise.

The Bowles-Simpson proposal concerns itself with federal employees, not state and local employees. One assumes that we should be fine-grained, e.g., are we talking about state, local, or federal employees? This makes a big and significant difference. For example, federal employees have had defined contribution pensions since the Reagan era, and the job functions at the federal level tend to be quite different from the state and local level.

Despite the defined contribution difference — which certainly helps matters — federal employees are overcompensated by a considerable margin when we use the same kind of adjustments for workforce characteristics seen in the EPI study. As Andrew Biggs and Jason Richwine noted in September:


A Congressional Budget Office study concluded that the average federal worker resides two-thirds of one pay grade above a similar private sector employee. A more recent academic study found a larger gap of three-quarters of a grade.

Official pay statistics are badly skewed as a result. If federal employees are over-promoted, federal positions can be underpaid, while federal workers remain overpaid. …


We estimate that federal employees are overpaid by roughly $40 billion annually, on top of job security and other perks that make beleaguered private sector workers envious. Not every federal worker is overpaid and the most talented people often receive less working for government, but average federal pay is substantially higher than in the private sector.

It should not be surprising that federal employees quit their jobs at only one-third the rate of private sector workers.

Given that Bowles-Simpson centers on the federal government, this seems salient.

The rest of Mike’s post is actually about state and local employees. And the story really is different for state and local employees, who appear to be less well compensated in cash terms but who generally have defined benefit pensions and fairly expensive retiree health benefits, which complicates the picture. 

Reihan Salam brought up a corollary argument, Andrew Bigg’s argument that you have to include unfunded benefit liabilities in these comparisons.

Yes, I’d say this is true when it comes to state and local employees and not federal employees, in light of the defined contribution difference for the feds. I’m partly to blame here — my post pivoted to state and local numbers, not to the federal numbers. (Here is Biggs on state and local.) 

The first is an apples-to-apples comparison of workers among education levels. The government’s workforce is more educated than the private workforce. For instance, the government’s “college plus” level is 54%, while all private workforce is 35%. “Some college” is 14% of government workers, 19% of the private workforce. So this is important to control for.

Mike provides an EPI chart on state and local government workers. I am pleased to see that he endorses this method, which he and Biggs and Richwine can agree on. This suggests that we’re within striking distance of a consensus on the whether or not federal employees are overcompensated. If you embrace the EPI method, Bowles-Simpson starts to make a lot of sense.

Mike then suggests the state and local pay premium for non-college workers derives from the collapse in the earnings floor in the private sector, which is an interesting formulation. The statutory minimum wage is one way of thinking about the earnings floor. Any other way of thinking about the earnings floor is question-raising: what should be the basis of compensation if not thinking through productivity, the reservation wage, the desire to attract reliable personnel for a given function, etc.?

But this shouldn’t imply that Mike is wrong. This is ultimately a normative question, and it could be that Mike believes that (a) everyone should be paid a lot more and (b) if we can’t impose a higher wage floor (e.g., a higher statutory minimum wage), perhaps for fear of the consequences for employment levels among ex-offenders, people with limited English proficiency, teenagers, etc., we can at least pay non-college state and local employees more than non-college private sector employees at taxpayer expense, despite the fact that at least some of the taxpayers in question face straitened circumstances of their own. I have a different vision of how the public sector should work: I want to deliver high-quality public services in a cost-effective manner.

At the end of Mike’s post, he includes a lengthy excerpt from a National Institute of Retirement post. I’m hoping that Andrew Biggs will respond to this directly. The authors of the study make a number of claims that I think you should judge for yourself, so please read the excerpt Mike provides, if not the paper itself, very carefully. My main question is this: what happens when we factor in the cost of retiree health benefits and the value of guaranteed annual returns? The former cost is not factored into BLS’s compensation measure, and the implies a very significant, and expensive, taxpayer backstop.

Here’s my grand proposal: using the EPI-AEI method of adjusting for worker characteristics, let’s sharply lower federal salaries adjust state and local salaries up and down as necessary to reach parity with private salaries, and let’s match public retiree health benefits and guaranteed annual returns to the levels that are common in the private sector. I’m guessing that this will dramatically reduce the wage bill at all levels of government. 

The Agenda

NRO’s domestic-policy blog, by Reihan Salam.