Panel:
Unlocking the Potential of Adminstrative Data
(Tools of Analysis: Methods, Data, Informatics and Research Design)
*Names in bold indicate Presenter
In the first paper, Vanessa Ríos-Salas from the University of Wisconsin will present on their efforts to integrate administrative records data on child welfare and educational outcomes to provide novel insights into how children placed in out-of-home foster care arrangements fare in school. The paper exploits the panel nature of the dataset they have constructed in order to reduce selection bias and provide important insights that have direct implications for child protection policy.
In the second paper, Brad Dudding (from the Center for Employment Opportunity) will discuss his research practitioner partnership with MDRC which involves application of machine learning techniques to predict which of CEO’s clients are most at-risk of dropping out of their nationally recognized program which uses peer supports and transitional employment to help reduce recidivism to the criminal justice system. This project has wide application to any programs which are concerned with the problem of client engagement/retention.
In the third paper, Rick Hendra from MDRC will discuss the latest results from the Subprime Lending Database Exploration Study. This study uses data provided by Clarity Services, the largest credit agency for subprime debt in the US. In this round of the study, MDRC is conducting a longitudinal analysis of the Clarity data to understand the movement into and out of debt and how policy and environmental factors affect those dynamics. The project is innovative due to the integration of administrative data with longitudinal ethnographic data.
In the fourth paper, Peter York from Community Science will present his work using child welfare data from the Broward County sheriff’s office. Using these administrative data, York and his colleagues built a predictive model which forecast with a high level of accuracy the likelihood that a case will return without another incident. This is very important and practical information for caseworkers because it can be used to accurately recommend if cases should go into either intensive out-of-home services or in-home community based programs. York and his colleagues find that the application of this model would lead to a 30 percent reduction in return cases.
Taken together, these four papers highlight the great promise from leveraging administrative analysis paired with advanced analytics. Overall the goal of this panel is to highlight the learning opportunities all around us provided by administrative data and the increasingly powerful tools to analyze them.