Study 1
Rahman and Weatherburn (2020) conducted a quasi-experimental design to assess the effect of imprisoning individuals convicted of a driving-under-the-influence-of-alcohol (DUI) offense on their DUI recidivism within 6 months, 24 months, and 5 years? time.
The sample included individuals who appeared in New South Wales criminal courts for DUI offenses from January 1, 2000, to June 30, 2018. The data source was the New South Wales Bureau of Crime Statistics and Research Reoffending Database. Cases were selected if they involved an adult who had been convicted with a principal offense of ?exceeding the prescribed content of alcohol? (DUI) where their blood?alcohol concentration was medium or high; the principal penalty was imprisonment or a suspended sentence; and the individual was not charged with any indictable offense. This resulted in a dataset of 9,384 individuals appearing in New South Wales criminal courts from January 1, 2000, to June 30, 2018. There were 3,946 individuals who received a sentence of imprisonment (the treatment group) and 7,336 individuals who received a suspended sentence of imprisonment (the comparison group).
The treatment group mostly were male (94 percent), were 35 years or older (66.3 percent), and were non-Indigenous (57.9 percent). Some resided in major cities (41 percent) and were from highly socioeconomically disadvantaged areas (31.3 percent). Of the treatment group, 91.5 percent had legal representation, 35.6 percent had three or more concurrent offenses, 67.7 percent had a high DUI range (blood?alcohol concentration) at index, 57.5 percent had four or more previous court appearances, and 60.1 percent did not have a previous imprisonment. The comparison group mostly were male (85.0 percent), were 35 years or older (64.2 percent), were non-Indigenous (57.2 percent), resided in major cities (50.6 percent), and some were from socioeconomically disadvantaged areas (30.4 percent). Of the comparison group, 87.9 percent had legal representation, 56.9 percent had one concurrent offense, 69.2 percent had a high DUI range at index, 34.5 percent had four or more previous court appearances, and 86.6 percent did not have a previous imprisonment. There were statistically significant differences between the groups on age, gender, race, remoteness of residence, socioeconomic status, legal representation, number of concurrent offenses, DUI blood?alcohol range, number of prior court appearances, and prior penalties (e.g., fines). These variables were used as controls in the analysis.
Reoffending was measured in consideration of both elapsed time and free time. Elapsed time was the time between case finalization and of the date of death, reconviction, or the end of the observation period (June 30, 2018) and did not exclude time spent in custody. To conduct the elapsed-time analysis, three dummy variables were created capturing any DUI reoffending from case finalization within 6 months, within 24 months, and within 5 years. Free time reoffending was examined using the custody data from the Reoffending Database and was defined as the number of days from disposal to the earlier of either the individual?s date of death or the end of the observation period of the data (June 30, 2018), excluding all further custodial episodes. Indicator variables were constructed to carry out a free-time analysis for whether a person reoffended with a DUI offense within 6 months, 24 months, and 5 years of free time.
The random allocation of cases to judicial officers was exploited in an instrumental variable analysis to account for observed and unobserved differences between individuals with a DUI offense who were given a suspended prison sentence and individuals with a DUI offense who were imprisoned. Unresidualized and residualized measures of judge severity were measured by their propensity to imprison individuals with a DUI offense. The analysis was undertaken using a two-stage?least-squares linear probability approach to examine differences between the groups in reoffending within 6 months, 24 months, and 5 years. Subgroup analysis was conducted with individuals with their first DUI offense, and with individuals with a prior DUI conviction.