Abstract:
Recidivism is an issue of great complexity as so many factors come into play. Other research has identified offending being affected by many factors including personal history and upbringing, necessary support, and connections to the community. Many studies reference stable housing as being a crucial platform to reducing recidivism and successful reintegration into the community. Other research has primarily focused on the overall effect of wrap around services combined with stable housing or whether stable housing is a significant predictor of recidivism, rather than focusing on the estimated effect of stable housing. This research explores the possibility of causal effects of stable housing after release from prison on recidivism.
Eligible people from six New Zealand prisons were approached to participate in the study. Participants were interviewed pre-release and approximately 6 and 12 months post-release. Four different analytic techniques were employed, all with covariate adjustment: i) logistic regression; ii) model averaging; iii) non-response weighted logistic regression; and iv) Inverse Probability Treatment Weighting.
Bivariate analysis showed participants who reported living in unstable housing approximately six months after release were between 1.5 to 4.6 times more likely to reoffend. The strongest bivariate relationship was between housing type and re-imprisonment (Chi-squared test p-value = 0.020). However, when key confounding variables identified were adjusted for, the estimates of the effects across each analytic technique showed agreement of no practically significant effect across models; almost all confidence intervals of estimated effects of stable housing spanned zero.
Despite some evidence of significant associations between stable housing measures and recidivism in bivariate analyses, there was no evidence the effects of stable housing were significant when other suspected confounding variables were incorporated in the analyses. This suggests either the sample size was too small to identify the significance of causal effects of stable housing, or these effects were explained by other suspected confounding variables (Skelly et al., 2012).
Keywords: recidivism; reoffending; stable housing; cohort study; observational study; logistic regression; model averaging; non-response; Inverse Probability Treatment Weighting (IPTW)