Associations between gambling, substance misuse and recidivism among Canadian offenders: a multifaceted exploration of poor impulse control traits and behaviours

Abstract

Moderate correlations among gambling, substance use and crime suggest these three behaviours may each be indicators of a single underlying problem trait, such as poor impulse control. We tested whether self-reported traits prospectively predicted future criminal recidivism, when accounting for incarcerated adult offenders' past criminal behaviour, substance use and gambling behaviour. We took a multi-construct approach to measuring components of impulse control, utilizing three questionnaires that assess sensation seeking, poor premeditation, and reward sensitivity. Male participants incarcerated in two federal institutions were invited to complete self-report booklets; participation was self-selected and fully voluntary. Results indicated moderate correlations between measures of trait impulse control, self-reported drug use, and official criminal history (N = 140). Gambling problems prospectively predicted post-release criminal recidivism (n = 83), even when accounting for past criminal behaviour, impulse control traits and substance misuse. Although they represent a small percentage of this sample, results suggest offenders with high gambling problems represent a particularly at-risk group for future crime, who were more than four times likely to reoffend compared to offenders without problem gambling. These results suggest there may be mutual impact of problem behaviours on life outcomes above the contribution of poor impulse control.

Keywords:

• problem gambling
• crime recidivism
• impulsivity
• substance misuse
• sensation seeking
• Cox regression survival analysis

Funding

This research was supported by a grant from the Ontario Problem Gambling Research Centre.

Notes

Caleb D. Lloyd is now at the Department of Psychology, University of Texas El Paso.

^1. To determine if dichotomizing measures weakened or strengthened prediction, we followed Walters' (1997) strategy, and used Cox regression to predict recidivism using the dichotomized versions of the measures, while simultaneously controlling for the continuous version of the measure. For alcohol use (ADS), neither version of the measure predicted recidivism alone, and thus neither version was significant when entered together. For drug use (DAST), both versions predicted recidivism alone, but the continuous version (p < .15) of the measure performed better when both were entered together. For gambling (PGSI), the dichotomized version performed better (p < .15) when entered with the continuous version. These results suggest that the continuous version of the DAST explains more variance in recidivism outcome compared to the dichotomized version, whereas the dichotomized version of the PGSI explains more variance in recidivism outcome, compared to the continuous version.

^2. We conducted our prediction models by alternately including or excluding individuals with no reported substance or gambling behaviour; results remained substantively the same in both approaches, but the main result displayed in Table 4 only approached significance when excluding non-gamblers, compared to reaching significance when including non-gamblers (see Table 4 note). This is likely due to reduced power, as limiting our sample size to n = 56 for a regression model with 5 predictors does not meet the minimal rule-of-thumb criterion which suggests including 15 participants per predictor.

^3. Examination of the variance inflation factor suggested that multicollinearity among the two impulse control measures was not problematic, but shared variance and small sample size likely contributed to the non-significant finding. When entered alone in Block 2, BIS-11 Motor was not significant, whereas ImpSS scores were significant. In the remaining models, impulse control traits were not significant whether entered alone or together.

^4. We also tested whether BIS-11 Motor or ImpSS scores moderated the gambling-recidivism relationship, and found non-significant results. These results can be obtained from the first author, upon request.

Additional information

Notes on contributors

Caleb D. Lloyd

Caleb Lloyd, MA, is a PhD Candidate in Forensic Psychology at Carleton University, Ottawa, Ontario, Canada. He has published research projects on the relationship between offenders' perceptions of desistance and recidivism, the impact of co-participants on treatment group outcomes, and judges' consideration of psychopathy when delivering indeterminate sentences in Canada.

Nick Chadwick

Nick Chadwick, BA, is an MA Candidate in Forensic Psychology at Carleton University, Ottawa, Canada. He has contributed to research projects on the use of evidence-based practices in community supervision as well as the use of dynamic risk assessment in the prediction of recidivism.

Ralph C. Serin

Ralph C. Serin, PhD, is Associate Professor in the Department of Psychology at Carleton University. He is a clinical psychologist who worked in federal corrections from 1975 to 2003 in various capacities. He has consulted with the National Institute of Corrections and the Centre for Effective Public Policy and is an advisor to the National Parole Resource Centre (US). He is currently engaged in research collaboration with the Departments of Corrections in numerous US states and internationally regarding violent offenders, dynamic risk assessment, evidence-based practice, parole decision making, and offender change.