Investigating bias towards Aboriginal people in police bail decisions.

AIM: To investigate whether police custody managers exhibit bias towards Aboriginal defendants in their bail decisions, once accounting for the risk of reoffending and absconding when on bail. METHOD: We compare the rates of different measures of bail misconduct for Aboriginal and non-Aboriginal defendants at the margins of release on bail, using an outcome test design. If marginal Aboriginal defendants have lower rates of bail misconduct than marginal non-Aboriginal defendants, then we conclude that custody managers apply a stricter threshold of release and are biased against Aboriginal defendants. We calculate outcomes for marginal defendants using the quasi-random allocation of defendants to police custody managers with differing propensities to grant bail. By using an instrumental variable strategy, we approximate the 'marginal defendant' as the defendant who would have been denied bail if they were allocated to a harsher custody manager but would have been granted bail if they were allocated to a more lenient custody manager. RESULTS: Rates of bail misconduct are similar between Aboriginal and non-Aboriginal defendants among our measures of misconduct. For example, Aboriginal defendants on the margins of release are 20.2 p.p. more likely to reoffend or abscond on bail than Aboriginal defendants refused bail, while the corresponding number for non-Aboriginal defendants is 19.8 p.p. CONCLUSION: We find no evidence to suggest that there is bias, in the form of a stricter threshold of release by police custody managers in their bail decisions, between Aboriginal and non-Aboriginal defendants. This suggests that strategies to reduce disparities in the likelihood of receiving bail should focus either on other criminal justice decisions (such as the decision to charge or arrest accused persons) or research should focus on structural factors of the criminal justice system (such as the potential presence of institutional bias). [ABSTRACT FROM AUTHOR]