CJB235
Estimating the impact of audio-visual link on being granted bail
Download this publication
Summary
Background
The aim of this study is to estimate the causal impact of appearing via audio-visual link (AVL) on the likelihood of being granted bail.
Audio-visual link describes the video conferencing equipment to facilitate court appearances without the defendant being physically present.
To estimate the impact of appearing via AVL on bail outcomes, we compare individuals who have their first court bail hearing via AVL at two NSW Correctional Centres, Amber Laurel and Surry Hills, between Jan 2018 and Feb 2020 with similar individuals over the same period.
The credibility of the estimates hinge on two factors:
1) The extent to which we have observed and modelled the factors that influence the bail decision of the magistrate, and
2) The extent to which the allocation of AVL is ‘as good as random’ after controlling for all observed factors.
Three statistical approaches (logistic regression, Mixed effects regression and a generalised random forest) are used to adjust for the observed differences between these two groups and estimate the causal impact of appearing via AVL.
Key findings
Figure 1. Average impact of AVL - full sample
We find no evidence for a meaningful disadvantage to appearing via AVL on whether the defendant is granted bail. When estimated using a logistic regression or a generalised random forest, we estimate a slight increase in likelihood, an increase of approximately 2.5%. When estimated using a mixed effects approach, we estimate a small reduction in the likelihood of being granted bail of 1.5%.
Our estimates suggest that if any impact exists, it is a relatively small effect. The 95% confidence intervals of our estimate ranges between defendants being 3.7% less likely to be granted bail if they appear via AVL (with other factors held constant) and 4.2% more likely to be granted bail.
We also find no evidence for the impact of AVL on bail outcomes for specific subgroups. Estimates by gender, Aboriginality, SEIFA quartile, remoteness of area and offence type are all largely consistent with the estimates of the average impact across all bail hearings.
Conclusion
Overall, we find no evidence that appearing via AVL causes defendants to be less likely to be granted bail. This study should not be taken as an endorsement of the use of AVL for court appearances, or used in lieu of a more holistic evaluation of the use of AVL, without consideration of the relevant costs, benefits and implementation issues.