A LATENT TRAIT APPROACH TO UNIFYING CRIMINAL CAREERS

We propose a latent trait model that simultaneously accounts for both participation in crime and the frequency of crimes, phenomena that the criminal career model attributes to different causal processes. The criminal career model is predicated on a categorical distinction between active offenders and nonoffenders, but the latent trait model assumes a continuous distribution of propensity to offend. Our specific statistical model relates a relatively stable and general latent propensity to engage in crime to the frequency of criminal behavior. The latent trait model successfully fit both the proportion of offenders (participation) and frequency of offending for several samples and several measures of offending. The model fit both samples of whites and nonwhites and both males and females. This shows that separate causal processes are not necessary to account for group differences in frequency and in participation, which disproves the major evidence in favor of the criminal career model. Finally, the latent trait model yielded evidence that disparate sex differences in rates of participation for different categories of offenses are consistent with a single difference on a latent trait. This demonstrates the latent trait model's potential for parsimoniously unifying knowledge about criminal careers.