TANF Data Collaborative Pilot: Analytical Tools for Education and Employment Services in Virginia

The project seeks to create unbiased analytical tools for Virginia Department of Social Services (VDSS) to customize education and employment services for TANF participants, improving their labor market success upon program exit. Demographic characteristics, household compositions, public benefit receipt, and past education/employment activities are analyzed for predictive purposes. 

A cohort of TANF case leads who exited the program in 2019 was identified, and their labor market outcomes within a year of exit were predicted. Data from various sources, including TANF enrollment, cross-program benefit usage, demographics, geographic location, household composition, and education/employment activities, were utilized. The pilot team used MDRC's predictive analytics tool and conducted variable importance analysis to design a tool for case workers to recommend suitable services to clients. Multiple sets of predictor variables of increasing complexity were tested, incorporating logistic regression and machine learning algorithms. The team evaluated the models based on simplicity, performance, and bias avoidance. Additionally, variable importance analysis was conducted to identify TANF program services most associated with labor market success for case leads. The project aimed to provide case workers with effective tools for personalized service recommendations. 

The project in Virginia developed accurate predictive models for TANF participants' labor market outcomes. Bias in predicting outcomes for certain subgroups is being addressed. Plans include refining variable importance analysis, building a new predictive model, incorporating external data, and gathering feedback for tool development and deployment. 

This project serves as a replicable model for staff from other agencies looking for examples of efforts that support the use of administrative data for learning and improvement. Policymakers, researchers, and organizations seeking to expand the use of data in state TANF agencies may find interest in this project.