Age, Criminal Careers, and Population Heterogeneity: Specification and Estimation of a Nonparametric, Mixed Poisson Model*

This article addresses three issues that are central to the criminal career debate. First, is the life course of individual offending patterns marked by distinctive periods of quiescence? Second, at the level of the individual, do offending rates vary systematically with age? In particular, is the age‐crime curve single peaked or flat? Third, are chronic offenders different from less active offenders? Do offenders themselves differ in systematic ways? Using a new approach to the analysis of individual criminal careers—based on nested, mixed Poisson models in which the mixing distribution is estimated nonparametrically—we analyze a panel data set that tracks a sample of males for more than 20 years. Our results provide empirical evidence in support of some features of criminal propensity theory and some in support of conventional criminal careers theory. In support of latent‐trait criminal propensity theory, the individual‐level average offense rate (per unit of time) varies as a function of observable individual‐level characteristics and unobservable heterogeneity among individuals, and the age trajectory of the offense rate is generally single peaked rather than flat. On the other hand, in support of conventional criminal careers theory, models that incorporate a parameter that permits periods of active as well as inactive offending across age have greater explanatory power than those that do not. In addition, the nonparametric, discrete approximation to the population distribution of unobservable heterogeneity in the individual‐level mean offense rate facilitates identification of four classes of offenders—nonoffenders as well as individual‐level characteristics that are unique to each group. Problems of theoretical explanation and empirical generalizability of these results are described.