Michael E. Ezell

Desisting from Crime: Continuity and Change in Long-term Crime Patterns of Serious Chronic Offenders

1. Looking Forward 2. Crime Over the Life-Course: the Empirical Implications of Three Theories 3. Review and Hypotheses 4. The California Youth Authority 5. Data and Methods 6. Descriptive Summary of the Data 7. Age and Crime among Latent Classes of Serious Youthful Offenders 8. On the Relationship of Past to Subsequent Criminal Activity 9. Looking Backward and Forward Appendices

Modeling Multiple Failure Time Data: A Survey of Variance–Corrected Proportional Hazards Models with Empirical Applications to Arrest Data

Proportional hazards models are powerful methods for the analysis of dynamic social processes and are widely used in sociology to estimate the effects of covariates on event timing (e.g., time to arrest, birth, marriage). The proper statistical modeling of failure time data is an important analytical issue in sociology, but to date the field has largely neglected the application of these models to multiple failure time data in which the conventional assumption of the independence of failure times is not tenable. This paper critically describes a class of models known as variance-corrected proportional hazards models that have been developed by statisticians to take into account a lack of independence among failure times. The purpose is to provide an exposition and comparison for sociologists of several such models and associated methods for handling multiple failure time data that can be readily estimated in commonly available statistical software packages. We pay special attention to the data requirements necessary for estimation of the models. The paper concludes with an illustrative application of the models to analyze the arrest patterns of a sample of California Youth Authority parolees.

Examining the Role of Lifestyle and Criminal History Variables on the Risk of Homicide Victimization

This study uses a lifestyle and routine activities (LSRA) approach to examine the offender—victim overlap in the case of lethal victimization. Longitudinal arrest and mortality data from three samples of individuals released from the California Youth Authority (CYA) are used to examine the lifestyle and criminal history factors that influence the risk of homicide victimization. Results from counting process Cox proportional hazards models indicate that gang membership, the period after release from incarceration, violent arrest history, ethnicity and race, county of release, and family criminality are all significantly related to the bivariate and multivariate risk of homicide victimization. However, other lifestyle and criminality factors such as alcohol and drug abuse and total offense history fail to predict the risk of homicide victimization. The article concludes with a discussion of the results in terms of their implications for theory, public policy, and future research needs.

The Effect of Criminal History Variables on the Process of Desistance in Adulthood Among Serious Youthful Offenders

Given the theoretical disagreement between population heterogeneity and state dependence theories in regard to the relevance of specific criminal history variables, this study examines the effects of several criminal history variables on the risk of a new arrest in adulthood among a sample of serious youthful offenders. Findings show that even after controlling for prior differences in offending levels, the duration of time since the last arrest is a strong (and negative) predictor of the risk of a subsequent arrest. Results also suggest that criminal activity in early adulthood is more informative on risk prediction than elements of criminal past tied to the juvenile status.

Examining the Overall and Offense-Specific Criminal Career Lengths of a Sample of Serious Offenders

Building on several recent papers that have brought the study of criminal career lengths back to the attention of criminologists, this study used a sample of serious offenders to examine the length of their official criminal career. This study not only examined overall career length, but it also examined the career length for five categories of offenses (serious, violent, serious violent, property, and drug), the residual career length as a function of age, and the correlates of the average career length. The data used consist of a random sample of 1,957 offenders released from the California Youth Authority in 1981-1982 who were followed into their 30s. The results of the study are interpreted in terms of their implications for theory, public policy, and future research needs.