Kenyan Judicial Performance Improvement

The impact evaluation will measure three treatments: 1.) Simplified information on performance against peers; 2.) Top-down accountability feedback structure; 3.) Demand-side accountability feedback structure.

A randomized control trial methodology will be used to implement these identified interventions and to evaluate whether the feedback and increased engagement results in reduction of the case backlog, improvement in case processing times, and improved quality of judicial services. Treatment will be allocated randomly to all courts nationwide.

2016-02-01

2017-02-01

Court administrate data collected through a template called the Daily Court Return Template. This template will be used to primarily measure timeliness and efficiency of cases. Namely,time to disposition (the average and median time between filing of a case (or entry of plea) to conclusion);percentage of cases concluded within the time limit (360 days); case clearance rate (the number of resolved cases as a percentage of initiated cases for a specific period).

Court user satisfaction and employee engagement surveys will provide a means to decipher whether efforts to improve timeliness and efficiency has negative side-effects on quality. The survey will be conducted on annual basis and will ask court users and employees about their perceptions on service quality, facilities, perceptions of judicial officers and complaints handling.

The intervention implementation will use a fractional factorial design testing three binary treatments. There will be four treatment arms or clusters. To avoid that one of the treatments gets 2 or more of the courts located in the largest urban areas, we will stratify by city size. In particular, courts located within the largest urban areas will not be included in the same treatment group.

A randomized control trial methodology will be used to implement these identified interventions and to evaluate whether the feedback and increased engagement results in reduction of the case backlog, improvement in case processing times, and improved quality of judicial services.

4 clusters with 58 courts randomly assigned to each cluster

Yes

4 clusters

218 courts

58 courts control, 58 simple feedback form, 58 simple feedback from + top-down accountability, 58 simple feedback form + demand-side accountability

Indicators based on the Daily Court Returns Template: Power calculations have revealed that 58 courts per treatment arm are sufficient to achieve power to 80%. Intra-cluster correlation is not applicable here as we have 1 observation per court. We report 2 scenarios for the correlation between waves (which we expect to be quite high given the monthly nature of the data). We use as benchmark indicator the case clearance rate. Other inputs for the sample power calculation are: 5% significance level, power 80%, baseline probability: 60% (based on data from Sep 2015 DCRT), 6 baseline waves, and 12 follow up waves. Power calculations have confirmed that 58 courts per treatment arm should ensure acceptable level of statistical power. Indicators based on court user satisfaction and employee engagement survey: Power calculations have revealed that 58 courts per treatment arm are sufficient to achieve power to 80% under different levels of the intra-cluster correlation (rho). We use as benchmark indicator the probability of court users that report are satisfied with the court service. Other inputs for the sample power calculation are: 5% significance level, power 80%, baseline probability: 66% (based on data from the baseline), 1 baseline wave, and 2 follow up waves, correlation between waves: 0.6, and that 24 respondents will be interviewed in each court (as it was indeed the case for the baseline survey already conducted and will be for the follow ups).