Association for Public Policy Analysis & Management

Panel: Uses of Predictive Modeling in Three Policy Areas

(Tools of Analysis: Methods, Data, Informatics and Research Design)

Across policy domains, researchers are benefiting from increasing access to more granular and frequent data and increased computing power to work with larger, richer datasets. There is growing interest in using such data to address research questions about patterns of behavior. In particular, predictive modeling—which has long been used in business and marketing research—is gaining currency as a way to identify individuals who may benefit from targeted intervention. Predictive modeling uses the experiences of individuals whose outcomes are known to model and predict outcomes of individuals whose outcomes are not yet known. While this approach offers benefits to agencies engaged in continuous improvement efforts and looking to allocate resources more efficiently, interpretation of and inferences from results are limited. In keeping with the conference theme, this panel presents three studies that discuss the benefits and limitations of predictive analytics in different policy arenas while also detailing ways to produce results that are practically meaningful and actionable to policymakers and practitioners. Each of these studies stems from partnerships or close working relationships with service providers. The first paper is a case study from K-12 education using rich student-level data to predict key school performance outcomes and milestones toward high school graduation. The second paper is a case study from supportive housing, using individual-level data from multiple government agencies to predict high-cost usage of housing services. The third paper is a policy brief designed to guide health care policymakers and practitioners in how to use predictive modeling results. The panel offers perspectives from a diversity of institutions, disciplines and policy arenas to catalyze an informed discussion about uses of a research method and model growing in popularity. The two discussants will synthesize findings and discuss implications across policy arenas.