Judicial Decisions as Public Goods

We propose two interventions to improve case disposition and outcomes: courtroom monitoring and case clerking.

Courtroom Monitoring: Simply watching a public proceeding may improve the quality of that proceeding. This intervention will assign one courtroom monitor to attend all portions of a randomly selected court case. At the first meeting for the case, the monitor will introduce herself to the magistrate and the advocates present. During the trial, the monitor will follow the proceeding, compiling two kinds of information. First, she will fill out an app-based questionnaire on procedural aspects of the trial like witnesses called, cross-examination, and sentencing mitigation statements. Second, she will record outcome information like length of trial, verdict, and sentence/fine.

Magistrates' case clerks: Clerks will be randomly assigned to magistrates for several months, working with the magistrate to produce judgments for cases that magistrate is working on. Case clerks reduce the administrative and research costs related to the composition of a legal judgment by providing support on the recording of evidence, legal research, and drafting of judgment language. In this implementation, clerks will be hired from the class of graduating law students and trained on judgment writing, legal research, and sentencing guidelines prior to their first case assignment.

2019-01-21

2020-08-31

Monitoring: Time to judgment; verdict; sentence/fine; number of witnesses; length of judgment (in words); quality of judgment (algorithm scored and textual complexity).

Clerks: Length of judgment (in words); quality of judgment (algorithm scored and textual complexity); appeals rate; time between last hearing and judgment delivery; verdict; sentence/fine.

The experimental design features two parallel experiments, one for each intervention. The case observation intervention will focus on four case types, effectively blocking on case type: “running down” (personal injury), sexual offenses, grievous bodily harm, and succession. Cases will be assigned to treatment sequentially, as the case enters the court registry system, via an app-based randomization tool, into one of three groups: out of sample, observed case, and unobserved case. If "out of sample", no followup of the case or data collection will be performed. If "observed", the case will be followed across its lifecycle by an observer who will collect relevant procedural and outcome data. If "unobserved", no observer will be present in the courtroom, but procedural and outcome data will be culled from administrative records about the case. The clerk intervention will randomly assign clerks to magistrates for a period of 3 months (one quarter). During that period, the clerk will assist the magistrate on case summary and judgment writing.

Both interventions will be implemented via a computerized random number generator upon entry of case (or magistrate) into the data collection app. For the case observation intervention, we may implement a biased coin design after an initial startup phase if sub-group balance is problematic. This approach will retain randomization probabilities to enable proper analysis.

For case observation: unit of randomization is the case. For clerks: unit of randomization is the magistrate-time period. Time period will likely be defined as 3 months (one quarter), depending on total number of magistrates available for the intervention at rollout.

No

This is not a clustered intervention.

For case observation: approximately 400 cases per case type. This will vary depending on actual case flow through court stations, and may be higher for some kinds of high frequency case types (like traffic offenses). For clerk intervention:

For case observation: 200 cases for each of treatment and control groups for four case types, for a total of approximately 1600 observations. For clerk intervention: Approximately seventy magistrates will be included across 2 years (8 quarters). In a given quarter, approximately 20 magistrates will be provided with a clerk, leading to 160 treated units and 400 control units.