Friday, August 11, 2006

The Expected Value of Governmental Quality

Tyler (or is it Tyrone?) links to my post on war as a case of a government failure, citing it as an example of the libertarian vice of “fixing the quality of government” before comparing it to the market.

Now, that may well be a libertarian vice (although see the response from Tyler’s co-blogger Alex). And I’ve probably been guilty of it from time to time. But I just don’t see how the accusation sticks here.

In that post, I readily admitted that the war in Iraq might have gone better – or not happened at all – had someone other than GWB held office. I agreed that some government officials have done especially poorly and deserve criticism; by implication, it’s possible for other government officials to beat the curve. In short, government can vary in quality. That seems a pretty obvious point, which few would deny. But the more important issue is that the incentives of government are systematically biased in an undesirable direction, especially in the absence of competition. This is the key point that Alex made in his original post and that I was affirming, and which non-libertarians repeatedly deny or ignore. Hence the willingness of neocons to believe we can achieve any foreign policy goal if we just have sufficient will, and the willingness of liberals to lay the blame for foreign policy failure entirely at the feet of the Bush administration.

But let me address Tyler’s post more analytically. The quality of both government and market outcomes display variance around the mean. But the systematically poor incentives of government strongly indicate that the mean will be lower for government than for the market. (I use the word “market” loosely here, to mean “whatever is the alternative to government.”) So we can regard the two spheres of action as having overlapping distributions, something like this:

This is, of course, an incredibly simplistic model whose only purpose is to address Tyler’s claim about libertarians “fixing the quality of government.” That phrase implies that libertarians regard all government as having uniformly low quality – a single point instead of a distribution. As the diagram shows, if both government and market outcomes vary in quality, then it’s possible for government to outperform the market! Fine. But is that what you should expect from expanding the general power of government, or from extending government regulation into a new area, or from inserting the U.S. military in another foreign war? From an ex ante perspective, what matters most is the mean.

If the libertarian vice is to assume that government quality will always be equal to its mean, the non-libertarian vice is to assume that we can simply choose whatever point we want on the curve. If we just put the right people in power, they apparently think, we’ll jump into that right tail where the government outperforms the market! This strikes me as a much more serious vice than making decisions based on expected value.

[BTW, if anyone knows how I can create a standard bell curve in PowerPoint, please let me know. I had to draw the two above using the curve tool.]


Jim Hu said...

Plot in Excel, then import

Vorn said...

First, let me compliment you on your graph. I think visual aids are very helpful to clarify the discussion.

That said, I do have some problems with your assumption that government must have poor incentives. There, are of course, problems with the means that we make government accountable to voters, just as there are problems with the ways we keep corporate executives accountable to shareholders. Certainly, the incentives of government officials are not always aligned with voters, just as the incentives for corporate executives are not always aligned with shareholders. Some amount of inefficiency is inevitable in both the government and the market. At the same time, surely actions can be done to improve incentives.

Overall, your diagram illustrates Tyler Cowen's criticism of you quite well. You are assuming that the ex ante distribution is fixed (even if the ex post results vary). You apparently believe it is impossible to create incentives in government so that the outcome curve for the government shifts to the right. (At least, you do not mention a shift as a possibility.)

Why can't we improve incentives for government actors, just as corporations will improve incentives for their employees?

A large corporation does not function much differently with respect to its employees than government. Both function mainly by command and usually individuals employees do not get any profit to incentivize them. Often corporations fail to consider employee incentives and that is at least in part (among many other reasons) why it might be smart to have McKinsey or another management consulting firm examine operations. It should be clear then, that you can change the incentive system for government employees, just as you can for corporate employees. In other words, you can shift your outcomes curve to the right.

Shifting the outcomes curve to the right was just what Madison and the others were doing when they designed seperation of powers, federalism, etc. into our system. While those are Constitutional changes, surely it is perfectly possible to make changes to the incentives of government actors to shift your outcomes curve to the right without constitutional changes.

I should further note that your outcomes curve arbitrary. Obviously something you made up rather than drew from empirical experience. May I suggest that one outcomes curve is not enough, but rather many. It doesn't make sense to have an aggregated outcomes curve, but rather, you need one for each possible government activity. May I suggest that the exact location of these curves will vary with function, and that sometimes the government curve will lie to the right of the market curve, while in other instances the reverse will be true.

Overall, your post vindicates Tyler Cowen's criticism. You have taken the quality of government as fixed. You did not illustrate the possibility that the outcomes curve for government could shift to the right. Furthermore, it illustrates my criticism, which is that libertarians have a tendency to simplistically assume that the market outcome curves lies to the right of the government outcomes curve in all circumstances, which is clearly false.

Jeff Brown said...

You could also import from Mathematica.

Henry said...

Vorn said, "Why can't we improve incentives for government actors, just as corporations will improve incentives for their employees?"

Maybe I'm not as economically literate as I should be, but isn't this the same as asking for the government to adopt market principles?

Glen Whitman said...

Vorn -- Yes, it's certainly possible to shift the curves one way or the other. I didn't say so specifically, but that was the point of Alex's post that I linked. Also, notice that most of the institutional devices for shifting the curve work by restricting the power or scope of government: federalism, separation of powers, enumeration of powers, bills of rights, and so on.

Thus, you can't make a simple leap from "we can improve the average quality of government" to "we can therefore trust government to do more things." Why not? Because we improve government quality in large part by constraining the government to do less.

And yes, the private sector has problems, too. Hence the existence of a left tail on the market curve. But at least market actors have a bottom-line incentive to solve their problems that government actors typically lack. The principal-agent problem that afflicts national government is far worse than the principal-agent problem that afflicts corporate governance. That is one among many reasons I think the center of the market curve lies to the right of the center of the government curve.

You are correct, of course, that a more sophisticated model would say the curves differ issue-by-issue. That's why many (most?) libertarians, myself included, admit a limited role for government in some areas.

Tim said...

I found Tyler's post baffling. He didn't explain how your post committed the "libertarian vice." The fact that even the best-case scenario of policy X is likely to be worse than not pursuing X doesn't in any way deny that policy X can be carried out in better and worse ways.