10 May 2013

Rearden Hammers

EconLog | Bryan Caplan | Public vs. Private Sector Compensation: A Case of Curious Controversy

Bewerunge and Rosen's working paper on public- versus private-sector compensation begins with a discussion of recent controversy:
... Exploring [compensation parity] is challenging for two reasons. First, the human capital of public and private sector workers may differ. If, for example, public sector workers have more education than private sector workers, then it is neither surprising nor objectionable that they earn higher wages. This is precisely the argument made by former White House budget director Peter Orszag
On the one hand, yes. On the other, no.

The yes is obvious, so why the no? Because adjusting for education is a measure of inputs, and what we're really concerned with is outputs.

What if every hammer the government owned cost $1000, when private sector hammers cost $50? That's outrageous, right? The Government is drastically overpaying for hammers! Taxpayers are being cheated!

Then along comes the White House to explain that no, no, no, it's perfectly reasonable for federal hammers to cost more because they're made of mithril. Mithril is more expensive than steel, so the gov't is totally justified in paying more. Once you do the proper regressions and clean the data up just so... presto bango! the government is paying a reasonable price for its hammers.

But do they need a mithril hammer? Is the mithril hammer doing a better job of driving nails? Are they getting things built any faster with the mithril hammers? Are their buildings more robust? Are they hammering adamantine nails into cavorite boards? Or are they just banging things with top-of-the-line, A1 grade, gold star, best-in-class, Hephaestus-approved, super-premium, special reserve edition Mjölnir-class mallets for shits and giggles?

If thisis getting the job done fine, you don't need one of these. Which, by the way, is $200 and made of solid titanium.

To a first approximation you should definitely account for things like education when doing these comparisons. But keep in mind that's just an approximation. It's assuming an implicit conclusion: that the government needs all the highly credentialed people it hires.

For example, the number of K-12 teachers with advanced degrees has ballooned to 48%. Are we getting any better education with all those MAs we're paying for? When even the Center for American Progress says no I think the answer is pretty clear. (HuffPo and NPR are also in agreement, so this is clearly no radical capitalist thing.) Statistically "correcting" for the education level of half of the teachers in the country seems reasonable until you stop to ask whether those extra years of schooling are actually providing any extra benefit in addition to their cost.

PS I expect everyone in this debate who's putting such high stock in accounting for human capital to do the same when the topic of conversation turns to, for example, pay disparities by gender & race, or comparisons between CEO and average employee compensation.

PPS It's important to note the actual thesis of Caplan's post: despite Orszag's assertive claims, the literature completely disagrees. This just goes to show that if you speak in a confident enough voice and start your sentences with some version of "Studies have shown..." you can get away with making almost any claim.

Of course, it helps immensely to have a sufficiently sycophantic or credulous audience.


  1. First of all, one should note that the cost of the benefits for Feds is just about constant no matter what the base salary is. This is because there is no difference in the amount that the govt spends on health care among govt workers. This is probably the biggest benefit. They are all in the same system, getting the same level of subsidy. They get the option to pick which plan they want each year, so the final cost to the individual can vary, but the govt share is the same for nearly every plan (the Post Office is a special case, and their financial balance sheet demonstrates it). Retirement benefits depend on final High-3 salary, so higher earners can cost more in retirement, but there are fewer of them, so this sort of binning does not reflect that cost. Also the new pension system requires the govt to kick in up to 5% of everyone's salary, so that is just a fixed percentage of overall payroll.

    The govt gives out relatively few bonuses, mostly to high earners, mostly to convince them to stick around and not defect to the private sector.

    The biggest confounder in all of this is the fact that govt workers do not really do much of that lower level work any more. It is generally contracted out to the rivate sector. Even in the STEM areas, much is done by contractors. More and more govt workers sit in offices running contracts and thinking about policy and doing the HR stuff that every organization has to do. Those functions require higher educational levels, because they require people who can read and write and think at a higher level than the average college dropout. We don't all agree with their conclusions, and some of them are down-right Machevellian or just plain evil, but Machievelli (sp?) and the devil both have very advanced degrees. These modern bureaucrats have to deal with politicians who are the devil on steriods, who speak with triple-forked tongues, who never forget a slight or a name. And congressional staff members who can be out on the street the day after the next election if their local supporters decide to rebel.

    I have to admit that I used to BE one of those overpaid Feds, and I now reap the rewards in retirement. I tried to hire people to work with me in the DC area, and the ones who were young and just slightly experienced (see the example in my comment below), and wanted to move to the area for family reasons, could be hired at a reasonable price. On the other hand, an expert in reactor fuel design from a local utility who only had to sell his house 75 miles away and buy one closer, required a salary at the top of the scale. As did a senior thermal-hydraulicist from a national laboratory where he was making more than he ever would working for the govt.

    These sorts of people are difficult to find. As I said in another comment, you could just assume that the jobs could be filled by any engineer, but do you really want to trust the safety of your family when they get on an airplane to someone who doesn't understand how airplanes stay in the air? There seem to be a lot of people like this in policy positions, and I, for one, don't like it.

    Also, for jobs in the DC area, the price of housing is a real killer. People who move there from Pittsburgh or NC (my main areas of recruitment) were shell-shocked by the prices, and it can be difficult to convince someone with a spouse that they will have to give up their house and 5 acres for an apartment for the move. The Feds who work in Iowa don't get as much, and their pay scales are generally pretty low. DC (and a few metro areas) is the place where the highest compensation rates exist, and the density is highest in DC. There has been NO recession in the DC area, I have been told.

    1. Agreed pretty much across the board. My mother shifted from the private to public sector in the DC area about 20 years ago. Although in her defense, the projects she manages (or did, until she was promoted out of them about two years ago) were as close to private-sector as you can get (i.e. they were self-funding and consistently in the black).

      I don't want to get into a giant discussion about the overall compensation levels of government employees. We'd need to account for benefits, job security, cost of living, sectoral composition, etc. I just wanted to bring up the one point that adjusting for educational credentials is a red herring.

  2. One last comment - neat hammer. I used to lust after a titanium anchor for my boat, but could never justify the cost. Could not justify this hammer, either, but maybe I will do some searching for a socket or wrench set...

  3. I am now starting to read the original paper, and I find all sorts of problems with it. As an example, it uses data from studies going back 30 years, and data from modern studies done my UMich in 2006 of people over 50. People under 50 represent 7% of the total.

    Unfortunately, they don't realize that pension plans change over time. Feds my age(63) are/were mostly in the old system, which was very generous. This generosity caused concern in teh Congress, so in the mid-80s they created a new plan that is a combined defined contribution/defined benefit/IRA/SS system. This study does not talk about this factor at all, but it is certain to have a strong effect on the results, because the pension benefits are quite different. And they come out of different pots of money in the govt (i.e., SS or the general treasury). I don't have any experience with state or local pensions, but I seem to think that in a LOT of states (Ca and Il come to mind), the pols have quite over-promised their pension schemes in recent years, and this sort of stuff doesn't seem to be captured.

    Coyote has a recent post about "analyses" like this - he wonders whether they are all junk. I think this is another example to add to the pile. They don't really understand the actual phenomena that they are trying to analyse, but instead just take stacks of data that were derived in different time periods, sift out the pieces they don't understand or are too hard to deal with, apply some voodo assumptions or homemade statistical tricks, and VOILA, we get a result. It is like you tried to do an analysis of the physics of your own current universe based on sets of observations done in different parallel universes where F=MA was not consistent, but varied by say a factor of 10 or so, or maybe with an exponent thrown in somewhere.

    Lord, please forgive the creator of Lotus 1-2-3, he did not know what sorts of mischief he was enabling!

  4. you can get away with making almost any claim

    ... as has been shown by numerous studies.

    1. Too true. Sometimes I wish I had low enough levels of shame to publish some of these. My publication count would be tripple what it is now!

  5. The government has to measure by inputs. Government output is negative.

    1. Ha! Yes!

      But the labor economists who are doing these sorts of studies are not so bound.