U.S. flag

An official website of the United States government, Department of Justice.

Methodological Issues in Survey Research on the Inhibition of Crime

NCJ Number
Journal of Criminal Law and Criminology Volume: 72 Issue: 3 Dated: (Fall 1981) Pages: 1094-1101
Date Published
8 pages
This reanalysis of Grasmick and Green's data makes clear that their conclusions about the inhibition of illegal behavior are not warranted; the original analysis studied the relationships between threat of legal punishment (L), threat of social disapproval (S), moral commitment to legal norms (M), and other factors.
The other factors are self-reported past involvement in illegal behavior (Ip) and estimated future involvement in illegal behavior (If). The authors interpret their data on the basis of an assumed underlying causal model in which L, S, and M each influence Ip and If. Their findings are consistent with the proposition that each of the independent variables inhibits participation in crime. However, Grasmick and Green make no attempt to rule out other interpretations. Excluding correlated measurement errors, there are two rival interpretations to consider: (1) that L, S, and M are not causes of involvement in illegal behavior, but rather are consequences of such involvement and (2) that the relationships are spurious, with nonvanishing correlations being produced by unmeasured exogenous causes of the observed variables. Each of these possibilities is examined. The reexamination of the data shows that they do not establish that the threats of legal punishment or of social disapproval inhibit self-reported illegal behavior, nor do they demonstrate that moral commitment to legal norms discourages illegality. The data also were consistent with the assumption that an omitted exogenous variable influences these inhibition variables and past criminality, with no causal influences among those measured variables. The comment concludes that what is needed in survey research on the subjective aspects of the social control of illegality is the formulation of causal models that incorporate all relevant variables, not just control variables, as well as a data collection design that will allow these models to be estimated without specification bias. Six footnotes and three figures are provided. (Author summary modified)

Date Published: January 1, 1981