RAND is proposing to build on a prior effort where we developed a risk prediction model for job applicants with a criminal history. RAND is proposing to build prototype empirical tools that could be used by employers to more accurately and fairly predict the recidivism risk posed by a job applicant.

The specialized expertise of CJARS staff along with their one-of-a-kind curated data collection on justice involved individuals will be useful for this effort

To accomplish this, CJARS will do the following:

1. Procure, install, and configure a research server that will allow non-CJARS researchers (external to CJARS) to conduct research using CJAR’S anonymized justice data. This server should be robust enough to allow several researchers to run computationally demanding algorithms simultaneously.
2. Work with RAND staff to set up a research environment so that it will be capable of running our previously developed prediction model.
3. Work with RAND staff to obtain appropriate Institutional Review Board clearances (at RAND and University of Michigan) for conducting this research.
4. Provide RAND staff with an initial orientation and start up advice on the most effective and reliable approaches to transform the relevant CJARS data elements (variables) into the format needed for our prediction model.
5. Work with RAND to obtain the appropriate clearances to extract appropriately anonymized statistical outputs from the research server.