Court appearance rates displayed on the RA form and other public-facing materials are intended to reflect recent, stable, and empirically observed patterns of court appearance among cases with similar RA scores. This brief describes the methodology and procedure CJA follows when updating the RA scores.

This project aims to update the RA form to reflect current court appearance patterns and evaluate how these updates affect judges’ release decisions and subsequent reappearance rates.

CJA maintains external interactive, user-friendly tools that present up-to-date data on arrests and release decisions throughout New York City. The process of updating these dashboards has historically been manual and time-consuming. To streamline this process, DART has developed a Python-based automation tool that updates these dashboards more efficiently.

A list of links and resources for CJA data infrastructure, datahub, DART infra, dashboards, and Power BI.

Human subjects research conducted at CJA is reviewed by the Center for Justice Innovation’s Institutional Review Board (IRB). This guide outlines the IRB submission and review process for CJA DART staff.

The IB CASU Pilot integrates learnings from the IB interviews into a new call script for CJA’s Court Attendance Support Unit (CASU), operated within the Outreach Department. CASU provides live call, SMS, email, and postal mail notifications to individuals who have been issued a bench warrant or stayed warrant following a missed court date.

Following bail reform, the number of individuals with high needs who are mandated to supervised release has increased. In response, the City allocated $13.1 million to develop and launch a new model of supervised release called Intensive Case Management (ICM). The goal of ICM is to better support participants with higher needs through community-based case management and to improve key pretrial outcomes (e.g., via court appearance and avoiding new arrests).

At a high level, this project aims to causally estimate the impact of Supervision Release (SR) versus Release on Recognizance (ROR) by exploiting quasi-random variation in judge behavior. Policy releveance of this work stems from the fact that SR is resource-intensive, and it is unclear whether it improves outcomes relative to ROR for marginal defendants.

MDRC provided a memo giving a high-level explanation of their analysis methods for the ICM study. This memo contains the following sections: 1. Covariates and Missing Data, 2. Weighting, 3. Clustering by Date of Random Assignment, 4. Models, and 5. Appendix Tables. Claudia Mastrogiacomo from NYCJA recreated their weighting scheme.

Project scoping docs help scope actionable data science projects. A pre-analysis plan (PAP) describes how researchers plan to analyze the data from a randomized or quasi-randomized evaluation.