Deterrence: Evidence and Inference

Because of space limitations and the short time I have been given to prepare my response to the two critiques of my work published in this issue of the Journal, I shall confine myself principally to the critique by Bowers and Pierce. 1 I do that not because the paper by Baldus and Cole 2 does not warrant a detailed reply but because an elaborate response to the central issues they raise is contained in a study of mine now in progress which deals critically with published research by Sellin and others. 3I choose to focus on the Bowers and Pierce piece also because, as I hope to show, their work largely misinterprets and misapplies the framework I have developed for testing the deterrence hypothesis.Addressing their work critically provides the opportunity to elaborate upon some pertinent aspects of my research, hopefully for the benefit of interested scholars.The conclusions of my time series study of murder basically are two: (1) that previous research never adequately tested a set of direct and specific implications suggested by a general theory of deterrence and (2) that my empirical findings, while tentative and inconclusive by the very nature of observational statistics, are not, inconsistent with rather sharp implications emanating from this theory, including the hypothesized deterrent effect of the conditional risk of execution.Bowers and Pierce tacitly accept the first conclusion and seek to evade the second, evidently on faulty grounds.In their efforts to obscure the empirical findings, they have selectively deleted observations, utilized an inferior regression specification, considered irrelevant variables and correlations, and revealed in the process misunderstanding of elementary statistical concepts, as I discuss in points III, IV, and V below.They do not provide evidence based on a systematic statistical analysis showing that capital punishment, or punishment in general, does not deter crime.Essentially, they make only the point that the observed deterrent effect of the risk of execution can be confounded