Article | Spring 2007
Capital Punishment and the Deterrence Debate
When economist Isaac Ehrlich published his 1975 article claiming that each execution averted an average of eight murders, he launched a 30-year war among legal scholars, social scientists, and advocates on both sides of the capital punishment debate. EhrlichÂ's work appeared in the midst of an unpopular moratorium on capital punishment imposed by the U.S. Supreme Court in its 1972 decision in Furman v. Georgia, and was extensively debated not only in the U.S. but across Western European countries considering abolition of capital punishment. Ehrlich analyzed murder and execution trends from 1933 to 1967, using what at that time were innovative regression techniques. Although Ehrlich placed many caveats and cautions on his estimates, the core claim of deterrence through execution had the impact of a sound bite, a bumper sticker, and an advertising slogan rolled into one. His work sparked heated controversy in professional journals, the popular press, and the polity. Several scholars raised serious doubts about the validity of EhrlichÂ's claims, including a devastating critique published in 1978 by a National Research Council panel on deterrence. Despite this criticism, EhrlichÂ's article was cited by the U.S. Solicitor General in Gregg v. Georgia, the 1976 U.S. Supreme Court decision that restored capital punishment after the brief four-year ban. The debate continued throughout the next two decades through recurring skirmishes in the social science literature, with neither side gaining either scientific or political traction.
The debate on deterrence was revived recently with the publication of more than a dozen new articles, primarily by economists, analyzing changes in execution and homicide rates since the death penaltyÂ's reinstatement. These studies leveraged new data archives on death sentences and executions to show that the selective implementation of capital punishment in some states and years led to dramatic decreases in homicides in the 38 states that permit capital punishment. These studies made very strong claims that each execution deterred between three and 32 murders. Some went even further, claiming that pardons and commutations had the opposite effect, increasing homicide rates.
Our recent study, however, examined homicide trends over the same time period and arrived at a sharply different conclusion. Published in the June 2006 Texas Law Review, the study examined fluctuating homicide rates with a particular focus on the subset of homicides directly targeted by executions—those that are punishable by death. We show that the use of total homicide rates, including those killings that would never be subject to the death penalty, is a type of aggregation error that can overstate causal effects on a portion of homicides that in fact are ineligible for executions. It can lead to the conclusion that there is deterrence when none exists. But looking at the smaller subset opens the possibility of detecting a marginal deterrent effect on that smaller subset. When considering this narrower and more statutorily precise target of capital punishment, we find no evidence of a deterrent effect.
The death penalty has long been reserved for only the most serious types of murders. The idea that Â"death is differentÂ" has guided states to craft death penalty statutes that limit executions to crimes considered particularly heinous, such as multiple homicides or the killing of children, or to those homicides reaching a predetermined grade of willfulness or premeditation. Examining aggregate homicide rates therefore misses the distinction of the particular crimes targeted by capital punishment and cannot isolate a deterrent effect. Some might say that the distinction is nominal, that killing someone in a bar fight is no different than the murder of a police officer or a murder in the course of a robbery, but we believe that the absence of intention in most homicides sets them apart from the small group of homicides that can be punished by death, and it is that constitutionally defined difference that matters when estimating deterrence. Our analysis therefore focuses on rates of capital-eligible homicides, and the Â"market shareÂ" of capital homicides as a fraction of total homicides.
To isolate that subset of homicides eligible for death, we examined data from the FBIÂ's Supplemental Homicide Reports (SHRs) between 1976 and 2003, which provide detailed, case-level information about the context and circumstances of each homicide known to the police, to identify death-eligible cases. We coded SHR homicides as capital- eligible if they included any of the seven elements commonly identified in capital punishment statutes across the states: (1) Â"felony homicides,Â" carried out during the commission of a robbery, burglary, rape or sexual assault, arson, or kidnapping; (2) killings of children below age six; (3) multiple-victim homicides; (4) Â"ganglandÂ" killings involving organized crime or street gangs; (5) Â"institutionÂ" killings where the offender was already confined in a government institution; (6) sniper killings; and (7) killings in the course of drug business.
The vast majority of homicides in the past thirty years were in fact not capital-eligible; just one in four homicides since 1976 were punishable by death. In reality, just a tiny fraction of even these death-eligible homicides have resulted in an execution. For the death penalty to be considered an effective deterrent, we concluded not only that there should be evidence a significant relationship between execution risk and homicide levels, but that this relationship should be stronger for death-eligible cases than it is for other types of homicide. In other words, the deterrent effect of capital punishment should cause the Â"market shareÂ" of death-eligible cases to fall in states and years with higher risk of execution. If, on the other hand, factors other than the death penalty are driving the deterrent effect apparent in aggregate-level analyses, then both death-eligible and death-ineligible homicide rates would fall, and the market share of death-eligible cases would stay fairly constant.
What we found, however, is that the rate of capital-eligible homicides has remained fairly stable over time, and has been insensitive to changes in the risk of execution. Also, the rate of non-capital homicides fell dramatically in the late 1990s both in death penalty states and non-death penalty states, causing capital homicides to comprise a rising share of the overall homicide rate. The rising market share in states and years with and without a legal death penalty statute shows that there is no evidence that capital homicide rates are lower when the risk of execution is greater. Because these rates are impervious to higher risks of punishment, we refer to this behavior as Â"the opposite of economicsÂ" and call into question prior claims of a deterrent effect.
Figure 1: Capital and Non-Capital Homicide Rate per 100,000 persons and Percent Capital, All States, 1976-2003
Our findings were put to a more standard econometric test, separating death-eligible from non-death-eligible homicides. We estimated a series of regression models to identify the effects of capital punishment on capital homicide after controlling for socioeconomic and crime conditions commonly associated with both homicide rates and punishment severity. We estimated both fixed effects and trajectory models to examine concurrent changes in the presence of a death penalty statute, changes in execution rates, and the rate of non-capital-eligible homicides at the state level. We also focused a part of our analysis exclusively on Texas, which has one of the highest homicide levels in the country, and by far the highest execution level. Our reasoning was that Texas is the most likely place where a deterrent effect of capital punishment would exist, and we therefore examined changes in county-level execution risk and homicide prevalence within each county.
Figure 2. Capital-Eligible Homicide Rates, Adjusted for Deterrence Components, Non-Capital Homicide Rates, Punishment Risk and Covariates by Executions Lagged 1 Year
At either the national level or looking solely at Texas, and using a variety of model specifications, we found no evidence of a deterrent effect of capital punishment on the rates of capital-eligible homicides. To the contrary, our models found that the market share of homicides that are capital-eligible grows over time, even as the risk of execution increases. We conclude that despite the recent drop in homicides as a whole, the threat of execution has had no deterrent effect on the very cases it was designed to prevent and that there has been no evidence of any differences between death penalty states and those that have rejected capital punishment. If the death penalty would deter any form of homicide, it should deter first and foremost those murders that might result in execution. This study found no evidence of deterrence in that small yet critical slice of homicides.
Amanda Geller is a doctoral student in the Columbia University School of Social Work. Her research focuses on crime, poverty, and statistical modeling in policy analysis.
Jeffrey Fagan is a Professor of Law and Public Health at Columbia University. He researches social contagion of violence, legal socialization of adolescents, social geography of domestic violence, jurisprudence of adolescent crime, drug control policy, capital punishment, racial profiling, and perceived legitimacy of the criminal law.
Franklin Zimring is the William G. Simon Professor of Law and Wolfen Distinguished Scholar at UC-Berkeley. He is best known for his studies of the determinants of the death rate from violent attacks; the impact of pretrial diversion from the criminal justice system; and criminal sanctions.
