A shortcoming of these models is that they neglect the nonpecuniary aspects of crime. The empirical model uses an underlying geometric probability process and explicitly accounts for the moral social compromise involved in becoming a criminal. The quantification of the criminal's moral and social sensitivity is done using data based on a consolidated file of police records and a cohort survey of criminals and noncriminals. It is found, on the basis of this unique data set, that the included individual criminals are risk averse and that gang membership reduces social sensitivity. Figures, tables, 22 references. (Author abstract modified)