Offending and Early Death in the Cambridge Study in Delinquent Development

Abstract

Criminal offending has many negative outcomes. Researchers have recently begun to consider the impact offending has on offenders' health, with a few studies focusing on differential mortality risk. Yet, prior research has been limited due to selective samples, restricted range of follow-up time period, limited set of explanatory variables, and lack of theoretical guidance. This paper examines the risk of early death among 411 South London males in the Cambridge Study in Delinquent Development followed into their late 50s. Attention is paid not only to differential risk of death between nonoffenders and offenders, but also to the risk within the population of offenders and through consideration of theoretical frameworks and associated predictor variables. Results show that high-rate chronic offenders evince the highest risk of death, an effect that continues even after controlling for childhood individual and environmental risk factors as well as participation in a range of analogous behaviors.

Keywords:

• early death
• offending
• health
• trajectories
• longitudinal

Notes

1. Of course, within specific, serious offender samples, it may be expected that the risk of death by homicide may be heightened (Lattimore et al., 1997; Piquero et al., 2005).

2. It is important to briefly review the results of two recent investigations of the Cambridge data that have attended to issues related to health generally, and hospitalization, disability, and death in particular. Shepherd et al. (2009) found that childhood (impulsivity) and parental predictors of offending, self-reported delinquency at age 32, and conviction history predicted death and disability by age 48. Piquero et al. (2011) found that, by age 48, offender trajectories differed significantly from one another on being registered disabled and hospitalization, with high-rate chronic offenders evincing the highest risk for both outcomes. Our study extends these efforts in two specific ways. First, we include a wide range of theoretically relevant predictors of mortality including: childhood individual and environmental predictors (the former which include several measures that assess self-control), unique offender trajectories, and involvement in deviant/imprudent/health-engendering behaviors (drug use, excessive smoking, and heavy drinking). Inclusion of this wide range of predictors affords the opportunity to assess several theoretical perspectives in order to predict early death in one of the most comprehensive long-term investigations of this issue. Second, we also update earlier studies of mortality in the Cambridge data by extending the follow-up to the end of 2010 (age 57 on average). As a reflection of the importance of additional follow-up years, consider deaths in the Cambridge data by age 40 (n = 10 deaths), age 48 (n = 17 deaths), and age 50 (n = 25 deaths). For the current study’s record search, the number of deaths totaled 31 out of the 411 males. A long-term follow-up study is important in order to examine the full range of outcomes and their correlates.

3. Since (at least until the 1970s) conviction in London was very likely after arrest, the requirement of conviction in London does not represent a major distinction from the arrest data which is the more commonly available official record data in the USA (Blumstein et al., 1985, p. 194). The minimum age of criminal responsibility in England is 10.

4. A semiparametric mixed Poisson model was used to detect distinct offending trajectories to age 40. This approach recognizes and assesses the possibility of variation underlying longitudinal crime patterns and is open to the possibility that there are meaningful subgroups within a population that follow distinct developmental trajectories that are not identifiable ex ante (Nagin, 2005). The model makes no parametric assumptions about the distribution of these developmental trajectories in the population. Instead, the distribution is approximated by a finite number of groups. Each category within the multinomial mixture can be viewed as a point of support, or grouping, for the distribution of individual-level trajectories. The model, then, estimates a separate point of support (or grouping) for as many distinct groups that are identified in the data. This approach permits detection and examination of individuals following distinctive developmental (offending) trajectories in time- and age-based data. A five-trajectory solution provided the best fit to the data: (1) nonoffenders (62.3%), (2) low-adolescence-peaked (18.6%), (3) very low-rate chronics (11.3%), (4) high-adolescence-peaked (5.4%), and (5) high-rate chronics (2.5%). Common to criminal careers research (Piquero, Farrington, & Blumstein, 2003), a very small number of individuals commit a large proportion of antisocial behavior in the CSDD.

5. The attrition rate has been unusually low for such a long-term survey, and since we use conviction records, attrition is essentially nonexistent. The only types of “missing data” concern males who are not at risk of conviction because they were abroad, incapacitated, or dead. All males contribute data when they are alive, i.e. they are censored after the death age. According to Piquero, Farrington, & Blumstein (2007), 10 of the 411 men died by age 40, while only 6.8% (28 of the 411 men) were incarcerated at some point before age 40. The mean time served was 1.5 years. Incarceration did not alter patterns of offending because of the infrequent use of incarceration and the small amount of time off the street among those who were incarcerated.

6. And of course, early death reduces the number of convictions that a man would have received so in a way these relationships are underestimated.

7. The estimate for the high-adolescence-peak group (OR = 3.45) approached significance (p < .06).

8. As noted above, we also estimated a logistic regression predicting mortality from a variable summarizing the subject’s conviction history and the results demonstrated a positive and significant effect. We also estimated this model with the two childhood risk factors and again observed that conviction history was positively and significantly related to mortality but that neither of the risk indexes was significant. Our decision to estimate these set of models was based on Gottfredson and Hirschi’s argument that there is an unnecessary complication associated with offender trajectories and that a simple dichotomy of offender/nonoffender or simple control for frequency (to include the nonoffenders) is sufficient, especially since, in their view, the highest rate offenders (who have the lowest self-control) are found at the tail of a continuous frequency distribution. While these results showed that conviction history was a significant predictor, it is important to bear in mind that the size and substance of this effect (as measured by the OR) is not very large. We believe that the insight gained from unpacking the offenders (as in the text) is useful because we can capture the size and substance of the effect at different ranges of the offending distribution and the risk of mortality is better visualized in that context. More generally, past research also supports studying variation in mortality within offenders. Stattin and Romelsjö (1995) found that the proportion of their sample who died by age 33 was 3% among all those who were convicted, but the proportion increased to 4.7% for those who had been convicted twice or more and 7.2% for those convicted four or more times. Instead of parceling the offender sample into groups based on some arbitrary designation, the adoption of the trajectory methodology is more objective (Nagin, 2005).

9. There is a significant association between both drug use measures and the trajectory classification, with high-rate chronic offenders reporting the highest prevalence of self-reported drug use and nonoffenders reporting the lowest prevalence of self-reported drug use.

10. As is true for drug use, there is a significant association between excessive drinking and the trajectory classification, with high-rate chronic offenders reporting the highest prevalence of excessive drinking and nonoffenders reporting the lowest prevalence of self-reported drug use. More generally, this suggests that future research should assess the correlates of mortality within trajectories (see Shepherd, Farrington, & Potts, 2002; Vaillant, 1987). Data limitations with respect to small sample size, especially among the high-rate chronics in the CSDD, precluded such an exploration.

11. Alcohol consumption has been linked to fewer illnesses and fewer infections (Shepherd, Farrington, & Potts, 2004).

Additional information

Notes on contributors

Alex R. Piquero

Alex R. Piquero is Ashbel Smith professor of Criminology, EPPS, University of Texas at Dallas, adjunct professor Key Centre for Ethics, Law, Justice, and Governance, Griffith University Australia, and co-editor, Journal of Quantitative Criminology. His research interests include criminal careers, criminological theory, and quantitative research methods. He has received several research, teaching, and service awards and is fellow of both the American Society of Criminology and the Academy of Criminal Justice Sciences.

David P. Farrington

David P. Farrington, O.B.E., is a professor of Psychological Criminology at the Institute of Criminology, Cambridge University, and adjunct professor of Psychiatry at Western Psychiatric Institute and Clinic, University of Pittsburgh. His major research interest is in developmental criminology, and he is the director of the Cambridge Study in Delinquent Development, which is a prospective longitudinal survey of over 400 London males from age 8 to 48. In addition to 550 published journal articles and book chapters on criminological and psychological topics, he has published over 80 books, monographs, and government publications.

Jonathan P. Shepherd

Jonathan P. Shepherd is a professor of Oral and Maxillofacial Surgery and director of the Violence and Society Research Group at Cardiff University. He won the 2008 Stockholm Criminology Prize for his research and its application to violence prevention. His research group won a 2009 Queen's Prize in Higher Education for Cardiff University. Since the mid-1990s and utilizing longitudinal data from the Cambridge Study in Delinquent Development, he has led a series of investigations of links between offending, victimization, and health outcomes. He has also led a series of randomized evaluations of cognitive behavioral interventions designed to reduce alcohol misuse and posttraumatic stress disorder. He is a fellow of the Academy of Medical Sciences and a member of Council of the Royal College of Surgeons.

Katherine Auty

Katherine Auty is studying for a PhD in the Forensic Psychiatry Research Unit, Barts and The London School of Medicine and Dentistry, Queen Mary, University of London. She is using data from the Cambridge Study in Delinquent Development to examine the intergenerational transmission of psychopathy and antisocial behavior. Prior to this, she gained an MSc in Public Policy at Queen Mary, University of London.