Monday, December 19, 2005
The Economic Death Penalty: Show me the Model!
Posner and Becker's blog posts on the economics of the death penalty provide an occasion for a few reflections.
Both Posner and Becker place a great deal of weight behind the empirical evidence that the death penalty deters crimes for which it is a punishment. (Becker explicitly disavows any other justification for the death penalty, Posner acknowledges that lots of other factors play into the calculus, but considers many of them a wash.) They don't go as far as the famous/infamous Vermeule/Sunstein line that capital punishment may be morally required, but the commitment to this sort of utilitarian line may lead them there.
Now, I happen to support the death penalty for certain serious crimes entirely independently of its deterrent effect. But if the argument for killing murderers is to rest entirely on the econometrics, I will have to be colored skeptical.
The problem is not, I think, with some platonic form of capital punishment. I do believe that even those with a criminal or murderous bent respond to some incentives, and that even if many murderers are addled, ignorant and very short-sighted, there might be enough of an effect on the margin for some death penalty regimes to deter serious crimes.
But of course our death penalty regime isn't that, and isn't even close. As Steven Levitt pointed out when I took his Economics of Crime , the death penalty is 1, very rare, even for murders, and 2, very slow.
So even if we posit a rational Beckerian potential murderer who sits down and weighs the expected costs and benefits, it seems unlikely that our death penalty regime should have much place in his calculus. In a state with no death penalty and life-without-parole, he faces, if caught and convicted, being locked in a box for the rest of his natural life. In a state with the death penalty and current safeguards, he faces the likelihood, if caught and convicted, of being locked in a box for the rest of his natural life, and a very small probability of being locked in a box for many years, very probably dying of natural causes, but possibly eventually being prematurely killed by the state.
Posner briefly adverts to this problem in his post, but suggests two counterarguments:
First, that the usual estimates for how rare the death penalty is miss the fact that many murders aren't eligible for the death penalty. He is responding to this argument by Steven Levitt and others, and his point seems fair, although he doesn't provide his own data, so it is hard to know how far it goes.
Second Posner suggests that option 2 (being locked in jail for ten years and then humanely killed) is so much worse than option 1 (being locked in jail for forty years and then dying on one's own or as a result of abuse or violence) that even a small probability of the former makes a big difference. (He says "most people would pay a substantial amount of money to eliminate" a .01 to .005 probability of death.)
This strikes me as rather improbable. First off, there are a number of life prisoners who would in fact prefer to die than to continue to live in what they consider to be unbearable and abusive captivity. That is why some prisoners have to be kept on suicide-watch. Second, even those who think that life in a box is better than no life at all should project this choice several years into the future, given the length of criminal trials, appeals, habeas, collateral habeas, and so on. Given that criminals tend to have disproportionately high discount rates, this further attenuates the possibility that a lethal injection several years down the line is going to make much difference. Third, owing to the current politics of the death penalty, there are quite a few very very smart lawyers and students willing to devote a lot of pro bono time to proving the innocence of anybody who looks likely to be executed, and there are a lot of extra protections that attach to a death penalty trial. Because the quality of one's representation and hearing depend at least somewhat on the proposed penalty, this is an additional reason that the threat of having to enter our Eighth Amendment death-is-different jurisprudence may not be the awfullest of awful fates.
Now, of course dying is bad, and most people most of the time would rather live than die. But given how rarely the death penalty is imposed, even for serious crimes, indeed given how rarely the death penalty is imposed, even for those on death row, it certainly doesn't seem common-sensical to think that our current death penalty system scares your marginal wrongdoer into doing anything differently at all. It is possible that some brilliant economist or behavior economist could design a rational-choice or boundedly-rational-choice model to explain given the current facts why we should think it is even conceivable that the death penalty deters crime, but until I see it, I am rather inclined to be skeptical of the very mixed empirical evidence. Has anybody seen a real-world model of how our death penalty might conceivably deter?
TrackBack URL for this entry:
Listed below are links to weblogs that reference The Economic Death Penalty: Show me the Model!:
» The deterrent effect of the death penalty from Statistical Modeling, Causal Inference, and Social Science
John Donohue sent me this paper reviewing evidence about the deterrent effect of the death penalty. I recommend the Donohue and Wolfers paper highly--at least on the technical side, it's the paper I would like to have written on this... [Read More]
Tracked on Feb 2, 2006 9:57:45 PM
Show me the model is indeed the key question. I've spent a fair amount of time pouring over the publicly available statistics on murder rates and the death penalty. The relationship is between the death penalty and murder rates is meager. To find one, you have to factor out a lot of factors and insert a lot of ad hoc considerations.
The murder rate of the District of Columbia, while not that far from those of other inner cities, is a grossly high outlier for any reasonably linear model. Alaska and Nevada turn out to have very high murder rates, while Montana, which one might naiively belive to have similar characteristics to Alaska, has very low murder rates. To the extent you can piece together a model, you can pick a wide variety of factors and get the results you want. The death penalty, liberal gun ownership laws, country music stations, a history of slavery, poor compensation for defense attorneys, Republican political dominance among white voters, Evangelical Christianity, and a host of other plausible and implausible factors cluster together -- choosing one over another is art and not science.
Posted by: ohwilleke | Dec 20, 2005 8:52:37 PM