June 4, 2010

Groundbreaking study of sex offender life courses

Challenge to actuarials: 4 distinct trajectories ID'd

Actuarial tools to predict offenders' future risk are all the rage. They provide a veneer of science in that on average their simple formulas work somewhat better than the flip of a coin. But a bit of sleight of hand is involved. They work only by lumping everyone together, making the vast differences among individuals with similar risk scores magically disappear. Thus, they say little about the risk of the specific offender standing in court awaiting judgment.

In addition to masking differences between individuals, actuarial risk assessment tools such as the Static-99 and the MnSOST-R ignore changes within an individual over time. As offenders age, they tend to rack up more arrests, which are scored as historical risk factors that elevate risk. But paradoxically, as men reach their 40s their days of crime are numbered. Many actuarially minded evaluators show a remarkable ignorance of the robust criminological literature on desistance, viewing sex offenders through an insular and mechanistic lens of history as destiny.

In the first study to directly challenge these actuarial fallacies by examining the offending trajectories of adult sex offenders from early adolescence to adulthood, a group of Canadian criminologists has identified four distinct offending trajectories, and in the process found a couple of surprises.

The four trajectories, identified by Patrick Lussier of Simon Fraser University and his colleagues through a longitudinal, retrospective study of 250 convicted sex offenders in a federal prison, were -- in order from most to least prevalent:
  • Very-low rate (56%): The most common trajectory involves a very low rate of offending over the time period examined, from ages 12 to 35. Most of these men were child molesters. Their offending appeared to be transitory and limited.
  • Low-rate desistors (26%): This group followed the age-crime curve identified by criminologists such as Sampson and Laub and Moffitt for offenders in general. This trajectory takes off gradually in adolescence, peaks in young adulthood, and gradually declines in the mid-30s. Offenders begin with general criminal activity and escalate over time to more serious crimes, including sex offending. Paradoxically, sex offending begins just as their overall criminal activity is slowing down.
  • Late bloomers (10%): This group is largely neglected in scientific literature about sex offending, according to the researchers. Rather than following the typical age-crime curve, late bloomers start their offending in adulthood, and gradually increase into their mid-30s. Like the low-rate desistors, this group progresses from nonsexual, nonviolent crimes to sex crimes. Many of these offenders sexually assault adolescent females.
  • High-rate chronics (8%): This group somewhat matches that known to criminologists as the "life-course persistent" group. The smallest of the four groups, it is also the most criminally active. These offenders start out as juvenile delinquents and offend frequently as adults, with sex offenses as just one component of general criminality. Most of the sex offenders in this group raped adult women.
These findings have a notable implication for risk assessment with juveniles. Despite their highly divergent rates of crime and desistance, three out of the four groups are not distinguishable during adolescence. As Frank DiCataldo explains in his new book, Perversion of Youth, the smart gambler will place bets that any random juvenile sex offender will NOT go on to become a chronic sex offender as an adult.

Lussier and his colleagues are critical of the actuarial tools for failing to capture the desistance process:
Some individuals might be considered high-risk offenders when their criminal activity is actually in the desistance process. This might be particularly true for the low-rate desistors and the high-rate chronics. Others, such as the late-bloomers, might be underestimated by actuarial tools considering that their criminal involvement started later and did not accumulate the risk factors included in the actuarial tools, in spite of the fact that their offending is accelerating…. We are left wondering how current actuarial risk assessment tools can account for the diversity of offending trajectories of sex offenders and the dynamic aspect of their offending over [the] life course.
The researchers report that they will study the predictive validity of their model in a future study. They also recommend further studies to extend the age range past the mid-30s, to better understand the various trajectories over the entire life course.

The study is: "Criminal trajectories of adult sex offenders and the age effect: Examining the dynamic aspect of offending in adulthood," by Patrick Lussier, Stacy Tzoumakis, Jesse Cale, and Joanne Amirault, in the current issue of the International Criminal Justice Review.

Other newly published articles on the actuarial controversy:

Campbell, T.W., & DeClue, G. (2010). Flying Blind with Naked Factors: Problems and Pitfalls in Adjusted-Actuarial Sex-Offender Risk Assessment. Open Access Journal of Forensic Psychology. Available ONLINE.
ABSTRACT: Actuarial instruments are typically the centerpieces of evaluations pursuant to civil commitment statutes for sex offenders. Almost as frequently as they rely on actuarial instruments, evaluators adjust actuarial data via weighing additional variables that are (presumably) correlated with recidivism. Typically, however, such variables are only weakly related to reoffending. This article reviews many problems and pitfalls undermining Adjusted Actuarial Assessment (AAA) and reports data demonstrating how ill advised this procedure is. Publicly available data do not support a claim in a recent meta-analysis (Hanson and Morton-Bourgon, 2009, p. 7), "For all three measures, for all types of raters, and for all outcomes, the adjusted scores showed lower predictive accuracy than did the unadjusted actuarial scores." Based on available data, at its best, AAA neither increases nor decreases the accuracy of actuarial classification. At its worst, AAA dilutes actuarial accuracy.
Craig, L.A., & Beech, A.R. (2010). Towards a guide to best practice in conducting actuarial risk assessments with sex offenders. Aggression and Violent Behavior, 15, 278-293.
ABSTRACT: Assessing the risk of further offending behavior by adult sexual perpetrators is highly relevant and important to professionals involved in public protection. Although recent progress in assessing risk in sexual offenders has established validity of actuarial measures, there continues to be some debate about application of these instruments. Increasingly forensic practitioners are being requested to give expert witness evidence in formal settings where actuarial risk estimates are being examined. This is true in the Sexually Violent Predator (SVP) hearings in the United States and the Parole Board Hearings in the United Kingdom. It is important therefore for practitioners using actuarial scales in adversarial settings to have a thorough understanding of methodological limitations of the technology and possible errors and inaccuracies of reporting actuarial risk estimates in individual cases. The aim of this paper is to summarize strengths and weaknesses of actuarial risk data, and to contribute to developing guidance on best practice when using actuarial measures in adversarial settings. This paper is organized into six areas: (1) Actuarial scales in practice; (2) Understanding risk prediction concepts; (3) Factors known to affect actuarial estimates; (4) Can we use group data to assess risk in individual cases; (5) Choosing which actuarial risk measure to use; and (6) Reporting actuarial risk estimates. It is hoped this paper goes some way to establishing guidance on the best practice of actuarial scales and associated limitations.