More on the Economics of Capital Punishment

The recent execution by the State of California of the multiple murderer Stanley "Tookie" Williams has brought renewed controversy to the practice of capital punishment, which has been abolished in about a third of the states and in most of the nations that the United States considers its peers; the European Union will not admit to membership a nation that retains capital punishment.

From an economic standpoint, the principal considerations in evaluating the issue of retaining capital punishment are the incremental deterrent effect of executing murderers, the rate of false positives (that is, execution of the innocent), the cost of capital punishment relative to life imprisonment without parole (the usual alternative nowadays), the utility that retributivists and the friends and family members of the murderer's victim (or in Williams's case victims) derive from execution, and the disutility that fervent opponents of capital punishment, along with relatives and friends of the defendant, experience. The utility comparison seems a standoff, and I will ignore it, although the fact that almost two-thirds of the U.S. population supports the death penalty is some, albeit weak (because it does not measure intensity of preference), evidence bearing on the comparison.

Early empirical analysis by Isaac Ehrlich found a substantial incremental deterrent effect of capital punishment, a finding that coincides with the common sense of the situation: it is exceedingly rare for a defendant who has a choice to prefer being executed to being imprisoned for life. Ehrlich's work was criticized by some economists, but more recent work by economists Hashem Dezhbakhsh, Paul Rubin, and Joanna Shepherd provides strong support for Ehrlich's thesis; these authors found, in a careful econometric analysis, that one execution deters 18 murders. Although this ratio may seem implausible given that the probability of being executed for committing a murder is less than 1 percent (most executions are in southern states--50 of the 59 in 2004--which that year had a total of almost 7,000 murders), the probability is misleading because only a subset of murderers are eligible for execution. Moreover, even a 1 percent or one-half of 1 percent probability of death is hardly trivial; most people would pay a substantial amount of money to eliminate such a probability.

As for the risk of executing an innocent person, this is exceedingly slight, especially when a distinction is made between legal and factual innocence. Some murderers are executed by mistake in the sense that they might have a good legal defense to being sentenced to death, such as having been prevented from offering evidence in mitigation of their crime, such as evidence of having grown up in terrible circumstances that made it difficult for them to resist the temptations of a life of crime. But they are not innocent of murder. The number of people who are executed for a murder they did not commit appears to be vanishingly small.

It is so small, however, in part because of the enormous protraction of capital litigation. The average amount of time that a defendant spends on death row before being executed is about 10 years. If the defendant is innocent, the error is highly likely to be discovered within that period. It would be different if execution followed the appeal of the defendant's sentence by a week. But the delay in execution not only reduces the deterrent effect of execution (though probably only slightly) but also makes capital punishment quite costly, since there is a substantial imprisonment cost on top of the heavy litigation costs of capital cases, with their endless rounds of appellate and postconviction proceedings.

Although it may seem heartless to say so, the concern with mistaken execution seems exaggerated. The number of people executed in all of 2004 was, as I noted, only 59. (The annual number has not exceeded 98 since 1951.) Suppose that were it not for the enormous delays in execution, the number would have been 60, and the additional person executed would have been factually innocent. The number of Americans who die each year in accidents exceeds 100,000; many of these deaths are more painful than death by lethal injection, though they are not as humiliating and usually they are not anticipated, which adds a particular dread to execution. Moreover, for what appears to be a psychological reason (the "availability heuristic"), the death of a single, identified person tends to have greater salience than the death of a much larger number of anonymous persons. As Stalin is reported to have quipped, a single death is a tragedy, a million deaths is a statistic.

But that's psychology; there is an economic argument for speeding up the imposition of the death penalty on convicted murderers eligible for the penalty; the gain in deterrence and reduction in cost are likely to exceed the increase in the very slight probability of executing a factually innocent person. What is more, by allocating more resources to the litigation of capital cases, the error rate could be kept at its present very low level even though delay in execution was reduced.

However, even with the existing, excessive, delay, the recent evidence concerning the deterrent effect of capital punishment provides strong support for resisting the abolition movement.

A final consideration returns me to the case of "Tookie" Williams. The major argument made for clemency was that he had reformed in prison and, more important, had become an influential critic of the type of gang violence in which he had engaged. Should the argument have prevailed? On the one hand, if murderers know that by "reforming" on death row they will have a good shot at clemency, the deterrent effect of the death penalty will be reduced. On the other hand, the type of advocacy in which Williams engaged probably had some social value, and the more likely the advocacy is to earn clemency, the more such advocacy there will be; clemency is the currency in which such activities are compensated and therefore encouraged. Presumably grants of clemency on such a basis should be rare, since there probably are rapidly diminishing social returns to death-row advocacy, along with diminished deterrence as a result of fewer executions. For the more murderers under sentence of death there are who publicly denounce murder and other criminality, the less credibility the denunciations have.

Becker has presented in his post today a compelling restatement of the economic case for capital punishment. I have a few minor disagreements and qualifications, and I will first mention them and then respond to some of the very large number of comments that my last week's posting elicited.

I do not consider revenge an impermissible ground for capital punishment. Revenge has very deep roots in the human psyche. As I have long argued, basing the argument on work by evolutionary psychologists such as Robert Trivers, the threat of revenge must have played an essential role in maintaining order in the "ancestral environment." That is a term that evolutionary biologists use to describe the prehistoric era in which human beings evolved to approximately their current biological state. In that era there were no laws, police, etc., so the indignation that would incite a person to revenge himself upon an aggressor must have had substantial survival value and become "hard wired" in our brains. The wiring remains, and explains some of the indignation that people feel, especially but not only the friends and family members of murder victims, toward the murderer. It seems plausible to me (here modifying what I said in my original posting) that the net increment in utility that they derive from the execution (versus life imprisonment) of the murderer exceeds the net increment in disutility that the murderer derives from being executed rather than imprisoned for life. The strong support for capital punishment in public opinion polls provides limited support for this conjecture.

I do not favor public executions; nor dismemberment or other horrific modes of execution. The incremental deterrent effect might well be nontrivial, but would be outweighed by public revulsion. There is also the danger of brutalization. As Friedrich Nietzsche pointed out, making people squeamish is one of the projects of modernity, and may explain the banning of blood sports as well as the movement away from public and gruesome executions. The idea is that if people become unaccustomed to bloody sights they will be less likely to employ violence in their relations with other people. Still another objection to public and gruesome executions is that they offer murderers an opportunity to die as heroes by showing fortitude.

I agree that marginal deterrence is important and that it generally argues for reserving the heaviest sentences for the most serious crimes. But there are two important qualifications. First, a very heavy sentence may be necessary to deter a minor crime because the likelihood of apprehension is very low. The expected punishment cost of crime is, as a first approximation (ignoring attitude toward risk), the punishment if imposed multiplied by the probability of imposition, so if the probability is very low a compensating increase in punishment is indicated. This does not impair marginal deterrence as long as the crimes are not close substitutes: a heavy fine for litterers will not increase the robbery rate, whereas capital punishment for robbers would increase the murder rate (of robbers' victims)--were it not for my second qualification. Even if murder and robbery were both capital crimes, there would be marginal deterrence because the police would search much harder for a robber who murdered his victim; the more extensive search would compensate, in part anyway, for the loss of the information that the victim could have given the police to identify the robber. Moreover, capital punishment is merely a ceiling; even if robbery were a capital crime, judges and juries would be much less likely to impose the death sentence on a robber who had not killed his victim than on one who had.

Marginal-deterrence theory provides, however, a compelling reason to execute prisoners sentenced to life without parole who murder in prison; the threat of a sentence of imprisonment can have no deterrent effect on them.

Becker mentions the possibility of racial discrimination in execution. Studies done some years ago--I do not know whether they would be descriptive of current practice--revealed the following pattern: murderers of black people were less likely to be executed than murderers of white people. Since blacks were more likely to murder other blacks than to murder whites, this meant that blacks were less rather than more likely to be executed than whites, relative to the respective murder rates of the two races. (Blacks commit murders at a much higher rate than whites.) The explanation offered was that judges and juries tended to set a lower value on black victims of murder than on white ones. From this some observers inferred that capital punishment discriminates against blacks. The inference is incorrect. The proper inference is that murderers of blacks are underpunished.

I turn now to the comments on my posting. A long comment by "ohwilleke" makes a number of interesting points, but they do not support his opposition to capital punishment. He notes first of all that many factors influence the murder rate besides the probability of execution. That is true, but it does not, as he suggests, make it "insanely difficult to make any econometric estimate" that is not "meaningless." Econometrics, which is to say the set of statistical methods used by economists to try to tease out causal factors, enables the particular factor of interest, in this case the probability of execution, to be isolated. The methods are not entirely reliable, which is why neither Becker nor I claim that economists have proved that capital punishment deters; we merely claim that there is significant evidence that it does. I note how many commenters remark correctly that murder rates are higher in the South, even though that is where most executions occur, than in other regions. But that is not an argument that executions do not deter. The higher the background rate of murder, the more severe one expects punishments to be. A high murder rate implies a high expected benefit from murder and so the expected cost of punsihment has to be jacked up to offset that greater benefit.

Ohwilleke's comment claims that the "error rate" in capital punishment is 10 percent. This is incorrect. Not a single person among the 119 that he contends were erroneously sentenced to death was executed. That is a zero error rate.

One commenter asks whether capital punishment "really deter[s] the type of person who actually does the murdering?" Of course if someone in a state that has capital punishment commits the kind of murder that puts him at risk of such punishment, the threat of capital punishment has not deterred. This is the usual situation with criminal punishment. People who commit crimes are people for whom the expected cost of punishment, combined with the other costs of crime, is less than the expected benefit of the crime. The purpose of punishing these people is not to deter them--by committing the crime in the face of threatened punishment they have shown themselves to be undeterrable--but to deter people who, were it not for the expected punishment cost, would commit the crime because its other costs were lower than its expected benefit.

Finally, several comments usefully point out that capital punishment has a secondary deterrent effect: it induces murderers to plead guilty and receive a life sentence.