The Sept-Oct 2008 issue of Journal of Criminal Justice has an article by Bijou Yang and David Lester titled "The deterrent effect of executions: A meta-analysis thirty years after Ehrlich." Here is the abstract:
In 1975, Ehrlich published a seminal paper in American Economic Review which argued that executions prevent murders in America. Subsequent empirical studies varied in their methodology and the time-period/region/country covered, and therefore it is difficult to draw a clear conclusion about the deterrent effect of executions. This article applies a meta-analysis to combine the results from refereed studies in order to summarize objectively the findings. The overall results of the meta-analysis supported the deterrent effect of executions, but the evidence for a deterrent effect depended on the type of study carried out (time-series and panel data versus cross-sectional data and the effects of publicity).
The discussion section contains this passage:
The authors left out a variable that I would like to see included in future studies. That is whether the data came from the pre-Furman (1972) era or the post-Gregg (1976) era. The law of capital punishment in the United States changed drastically in that interval, in a way that may very well be relevant to the subject. If capital punishment is narrowed to a smaller class of offenses but used relatively more often within that class, as Furman and Gregg require, one would expect a greater deterrent effect. Studies with post-Gregg data might be finding deterrence more often than those with pre-Furman data for that reason.
This issue is related to another problem with deterrence studies. The dependent variable is usually, if not always, all homicides or all murders and voluntary manslaughters. But voluntary manslaughter and second degree murder are not capital offenses, and we would not expect the death penalty for first-degree murder (premeditated or in the course of another felony, usually) to deter them. The relevant dependent variable, first-degree murders, is diluted in the much larger class of homicides. The entire group of studies may underestimate the actual deterrent effect for this reason.
The article has a feature that seems to be standard in articles finding deterrence these days. In the last four paragraphs, the authors feel compelled to say that they aren't necessarily endorsing the death penalty, reciting a list of issues on which they have no apparent expertise and cite very dubious figures. Twenty-one death row inmates have been exonerated by DNA in a three-year period? I don't think so. Even the DPIC's notoriously loose list only claims 17 as those in which DNA "played a role" over the entire 35 years. If these disclaimers are de rigueur for researchers, it would be nice if they at least noted that these propositions are disputed and cited the contrary authority.
Notwithstanding that quibble, the article is a significant contribution to the literature. We have added it to our list of abstracts on the subject.
The presence of a deterrent effect in this meta-analysis depended upon the type of study. The statistically significant deterrent effect was found most clearly for the time-series studies and for the panel studies. In contrast, the cross-sectional studies, the studies of the effect of single executions, and the studies of newspaper and television publicity gave mixed results, and deterrent or brutalization effects failed to reach statistical significance.It's not too hard to imagine why studies of the effect of a single execution or of execution publicity are less likely to find a significant effect. These studies only measure one component of the deterrent effect. If a person is less likely to commit murder because of his awareness that (a) the death penalty is generally enforced in the state, plus (b) an execution was carried out last month, a study that only measures (b) is only measuring one component, probably the weaker of the two by a substantial margin.
It is intriguing to ask why the methodologies incorporating the dimension of time (time-series and panel data) gave evidence for a deterrent effect, while the cross-sectional (ecological studies over regions) did not. This difference in the results of time-series and cross-sectional studies is found in other areas (such as the socioeconomic predictors of aggregate suicide rates) (see Lester & Yang, 1995). Perhaps individuals are more likely to make decisions based on past conditions in their own community rather than by comparing their conditions with that of others in different ecological areas. The reasons for these differences in the outcome of the meta-analysis, however, are far from clear. It is hoped that the present meta-analysis will stimulate thought and research on this issue.
The authors left out a variable that I would like to see included in future studies. That is whether the data came from the pre-Furman (1972) era or the post-Gregg (1976) era. The law of capital punishment in the United States changed drastically in that interval, in a way that may very well be relevant to the subject. If capital punishment is narrowed to a smaller class of offenses but used relatively more often within that class, as Furman and Gregg require, one would expect a greater deterrent effect. Studies with post-Gregg data might be finding deterrence more often than those with pre-Furman data for that reason.
This issue is related to another problem with deterrence studies. The dependent variable is usually, if not always, all homicides or all murders and voluntary manslaughters. But voluntary manslaughter and second degree murder are not capital offenses, and we would not expect the death penalty for first-degree murder (premeditated or in the course of another felony, usually) to deter them. The relevant dependent variable, first-degree murders, is diluted in the much larger class of homicides. The entire group of studies may underestimate the actual deterrent effect for this reason.
The article has a feature that seems to be standard in articles finding deterrence these days. In the last four paragraphs, the authors feel compelled to say that they aren't necessarily endorsing the death penalty, reciting a list of issues on which they have no apparent expertise and cite very dubious figures. Twenty-one death row inmates have been exonerated by DNA in a three-year period? I don't think so. Even the DPIC's notoriously loose list only claims 17 as those in which DNA "played a role" over the entire 35 years. If these disclaimers are de rigueur for researchers, it would be nice if they at least noted that these propositions are disputed and cited the contrary authority.
Notwithstanding that quibble, the article is a significant contribution to the literature. We have added it to our list of abstracts on the subject.
Leave a comment