Ehrlich's first point is that if one is searching for deterrence it is the law in action (i.e., the actual incidence of executions) rather than the law on the books (i.e., the presence or absence of the death penalty) which is crucial. His second point is that in order to spot deterrent effects other factors which might affect homicide rates, such as conviction rates and unemployment rates, must be held constant. Many of those who believe that Ehrlich's work is fundamentally flawed nevertheless accept these criticisms. This article follows Sellin's approach but takes account of Ehrlich's criticisms. Instead of comparing states on the basis of whether or not they have capital punishment statutes, it compares states on the basis of the number of murderers executed. It does this by correlating differences in executions with differences in homicide rates. Focusing on differences in this way does not separate out causal factors other than executions for specific control, but it is arguably a reasonably good control for the variety of often unmeasurable factors that are historically specific to given states and likely to affect homicide rates. The results of this analysis are consistent with the basic finding of Sellin and others who have followed his procedures. The data provide no reason to believe that executions deter homicide. At the same time nothing about the data suggests that states that do not execute murderers enjoy lower homicide rates on this account. The results of the study may be limited because only states Sellin compared are examined and a number of arbitrary decisions had to be made. All such decisions were made a priori on theoretical grounds and are specifically noted in the paper. However, for these reasons it might not be unfair to treat the study as a pilot for an as yet unborn larger study that would look at data from the forty-eight contiguous states.
Lempert, Richard O. "The Impact of Executions on Homicides: A New Look in an Old Light." Crime and Delinquency 29 (1983): 88-115.