The influence of neighbourhood disadvantage on charge dismissal: the case of drunk driving

ABSTRACT

We use a sample of 78,160 cases involving adults who pleaded guilty to drunk driving in a New South Wales (NSW) court between 2014 and 2019 to assess the contribution of neighbourhood disadvantage to charge dismissal. Data are analysed using a multilevel random effects model with controls for magistrate identity and legal factors that must be considered by the magistrate when deciding whether to dismiss a PCA charge against an offender. Findings indicate that magistrates with a higher dismissal rate are more lenient towards individuals from advantaged or highly advantaged socio-economic neighbourhoods. Magistrates with a lower dismissal rate, are less lenient toward those from advantaged or highly advantaged socio-economic neighbourhoods. Neighbourhood disadvantage has a statistically significant effect on judicial willingness to dismiss a charge of drink driving and record no conviction. The effect is small but affects many people.

KEYWORDS:

• Drunk driving
• multi-level modelling
• sentencing
• socioeconomic disadvantage

Acknowledgements

The authors would like to thank the NSW Bureau of Crime Statistics and Research for access to the data required for this research. We would also like to thank the reviewers for their helpful feedback on the first draft of the article.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Access to data

Access to the ROD data employed in this study can be obtained on application to the NSW Bureau of Crime Statistics and Research at [email protected]. Copies of the data specification can be obtained from the first author.

Notes

1 Here and throughout the manuscript, the model is fitted using maximum likelihood estimation using the glmer procedure available in the R package lme4.

2 The argument control = glmerControl(optimizer = ‘bobyqa’) is specified within the glmer function. To ensure convergence of the algorithm, one may want to set a high number of function evaluation in the opitmizer by adding, eg, optCtrl = list(maxfun = 2e5).

3 Indeed, while we may have gained greater control over omitted variable bias in choosing an offence that had an objective measure of seriousness, the constraints in sentencing for an offence with this characteristic may have limited the scope for sentencing bias to emerge.

4 The 49 magistrates highlighted to have a higher dismissal rate dealt with 25,480 cases across a 6-year span, i.e., an average of 88 cases per year per magistrate. Of the cases dealt with, 11,009 concern individuals from advantaged areas and 14,831 from disadvantaged areas. According to the model, it is expected that 1,008 individuals from advantaged areas will be dismissed whereas 1,065 from disadvantaged areas will be dismissed. If the rate of dismissal was equal across defendants at 9.16%, there would be 1,359 individuals from disadvantaged areas who would be dismissed.