"The committee concludes that research to date on the effect of capital punishment on homicide is not informative about whether capital punishment decreases, increases, or has no effect on homicide rates. Therefore, the committee recommends that these studies not be used to inform deliberations requiring judgments about the effect of the death penalty on homicide. Consequently, claims that research demonstrates that capital punishment decreases or increases the homicide rate by a specified amount or has no effect on the homicide rate should not influence policy judgments about capital punishment."
Why has the research found it so hard to sort out cause and effect in this situation? The NRC report emphasizes two reasons, but discusses a number of others:
1) One problem is that deciding on the effect of capital punishment requires saying, "Compared to what?" But existing studies don't pay enough attention to what the alternative punishment might have been. Here the NRC report: "Properly understood, the relevant question about the deterrent effect of
capital punishment is the differential or marginal deterrent effect of execution over the deterrent effect of other available or commonly used penalties, specifically, a lengthy prison sentence or one of life without the possibility of parole. One major deficiency in all the existing studies is that none specify the noncapital sanction components of the sanction regime for the punishment of homicide."
2) Another problem is that studies of the deterrent effect of capital punishment need to make some assumptions about how potential murderers perceive that penalty. Are they aware of how often it is imposed, and under what circumstance, and their actual chances of receiving the penalty? In many studies, these underlying assumptions are not spelled out clearly, or even at all.
3) There's a classic cause-and-effect problem in studying the deterrent effects of any criminal penalty, whether fines or imprisonment or capital punishment. Say that there is one jurisdiction where lots of crimes occur and another where not many crimes occur, for whatever reason. The first jurisdiction thus imposes lots of criminal penalties, and the other jurisdiction doesn't. In this situation, a naive statistical test will observe that high level of crime are correlated with high levels of penalties, and low levels of crime are correlated with low levels of penalties. But it would be foolish to argue that the levels of penalties are causing the levels of crime. Instead, one needs to figure out how increases or decreases in the level of penalties would affect levels of crime, which is a much harder question, especially because changes in penalties often occur as a reaction to levels of crime.
4) The statistical problem here is a difficult one. To help illustrate why, here's a graph of the number of death sentences and executions in the U.S. since 1974. Back in 1976, U.S. Supreme Court decisions had made nearly impossible for states to execute anyone, and the number of death penalty sentences was somewhat lower as well. Then the number of death penalty sentences rises, and a few years later the number of executions rises. In the last decade or so, the number of death sentences had dropped substantially, but the number of executions has dropped by less, surely because of that large number of death sentences originally given back on the 1980s and 1990. Referring back to the earlier point, an obvious question here is the extent to which potential murderers are aware of these patterns in sentencing and executions.
Next, here's a graph of homicide rates since 1974. Notice that it starts off at about 10 per 100,000 population, and then starting around 1990 drops off to about half that level. In other words, about a decade after capital punishment sentences rise, and at about the same time as the execution rate starts to rise, the homicide rate drops off. A naive statistical comparison between these patterns using national data might well suggest that higher levels of executions preceded a drop-off in murders.
But of course the deal penalty is not equally likely across states. New York state sentenced only 10 people to death from 1973-2009, and executed none during that time. California and Texas both sentences large numbers of people to death, but California actually carried out only 13 people from 1976-2009, while Texas executed 447 people. However, these differences in death sentences and actual executions seem to have very little effect on the murder rate in these three states, which essentially follows the national pattern in all three cases.
Of course, researchers in this area are fully aware of these difficulties in looking at the data over time and across states, and have applied a wide array of methods to this data. Given these issues, it is perhaps no surprise that the NRC report lists studies in the last decade which find that each execution deters five other murders, or 18 other murders, as well as studies that find that capital punishment deters no murders at all, and studies that find that the conclusions one draws from the data are quite fragile, depending on small differences in the statistical tests that are run.
Of course, the issue of whether capital punishment deters is not the only issue in making policy choices about capital punishment. The NRC report is careful to point out that it is not considering the moral arguments for or against capital punishment, nor is it looking at the arguments over whether the penalty is administered in a consistent or nondiscriminatory fashion. Personally, my own moral sense would not rule out the concept of the death penalty for the most extreme and egregious cases of murder. I do worry that its application seems to vary so greatly across jurisdictions, and across racial groups, and by the quality of the lawyers involved in the case. I also worry that in a world without capital punishment, those who have already committed crimes that could land them in prison for life (say, kidnapping) have an incentive to kill potential witnesses, because in a situation without capital punishment, there is no additional penalty for doing so.
The NRC report is careful to point out several times that a lack of solid empirical support for whether capital punishment deters doesn't prove that such an deterrent effect does not in fact exist. As the old saying among empirical researchers goes: "Absence of evidence is not evidence of absence."
On the particular issue of the uncertainty of whether or how much capital punishment deters future murders, I struggle with a conundrum that was put to me many years ago. Take as a starting point that we aren't sure whether capital punishment deters or not, and we must make a choice whether to execute certain murderers or not. Then consider four possibilities: 1) we execute some murderers and it does deter others, so we save innocent lives; 2) we execute some murderers and it doesn't deter others; 3) we don't execute any murderers and it wouldn't have deterred anyone if we did; and 4) we don't execute any murderers but it would have deterred some future murderers if we had done so. Those who worry about executing those who don't really deserve such a grave punishment, or even who may be innocent, have a point. But if it's possible that capital punishment may deter, and we genuinely aren't sure, then we also need to take into account in our policy calculations the possibility that executing the most egregious murderers might save innocent lives.