Citation

Yunker JA. Crime Delinq. 1982; 28(1): 96-124.

Copyright

(Copyright © 1982, SAGE Publishing)

DOI

10.1177/001112878202800107

PMID

unavailable

Abstract

The great majority of the many recent multiple regression studies of homicide rate determination fail to find executions to have a statistically significant negative effect on homicides. A notable excep tion is the work of Isaac Ehrlich. This literature was recently reviewed in this journal by Richard McGahey. McGahey's review included a nontechnical description of the multiple regression method, but only very sketchy treatment of its potential problems. This paper supple ments McGahey's article by reviewing the several potential problems that often invalidate statistical tests of social theories. Particular em phasis is put on the identification problem, or the difficulty in disen tangling one particular relationship from a system of interacting social relationships sufficiently to allow valid estimation by statistical meth ods such as multiple regression. It is shown that there may be very serious identification problems in the estimation of the relationship between executions and homicides, and that serious questions may be raised regarding the validity of any particular estimation of this rela tionship. In light of these problems, the investigator cannot judge a given hypothesis solely on the grounds of indications from a given statistical test. Rather, the inherent plausibility of the hypothesis itself must be taken into account. The paper concludes with a brief argu ment that the hypotheses relied upon by the defenders of capital pun ishment possess greater inherent plausibility than those relied upon by its opponents.