The measurement of specialization and escalation in the criminal career: An alternative modeling strategy

Recent research using indices of specialization and escalation, such as the forward specialization coefficient and the escalation coefficient, have generally shown that repeat offenders tend to commit the same type or a more serious type of crime on successive arrests. Unfortunately, there are two important limitations to the use of specialization and escalation indices: (1) the meaning and interpretation of the coefficients is often unclear, and (2) the coefficients cannot be tested for statistical significance across groups. In an attempt to account for these limitations and to extend prior research in this area, this paper applies a class of log-linear models developed for studying social mobility tables with matched categories (for one or more groups) to crime-type switching tables. The benefits of using these models, in comparison with prior specialization and escalation research, are that the parameter estimates can be interpreted directly as tests of specialization and escalation in a meaningful way and the parameter estimates can be tested for equality across groups, such as age, race, and gender. The application and interpretation of these models are illustrated with arrest data from a sample of felony offenders in Michigan.