Online Program Home
  My Program Register Now!

Abstract Details

Activity Number: 119 - SPEED: Government and Health Policy
Type: Contributed
Date/Time: Monday, July 30, 2018 : 8:30 AM to 10:20 AM
Sponsor: Health Policy Statistics Section
Abstract #328670
Title: Nonparametric Machine Learning with Variable Selection for Synthetic Controls
Author(s): Christoph Kurz* and Laura Hatfield and Sherri Rose
Companies: Helmholtz Zentrum Muenchen and Harvard Medical School and Harvard Medical School
Keywords: causal inference ; machine learning ; longitudinal data ; diabetes ; synthetic control

The selection of a suitable control group in the evaluation of policy interventions is a challenging problem. While synthetic control methodology is a growing area of research in this space, it has largely focused on parametric approaches to creating a weighted control group based on aggregate-level outcome data over a long pre-intervention time period. Typical settings in health policy have pre-intervention periods that span only months or a few years. Our work focuses on flexible nonparametric machine learning techniques to create appropriate synthetic controls using pre-intervention aggregate-level outcomes and covariates across a small number of time points. The methodology includes an automated variable selection procedure to identify the minimal needed set of covariates to form the optimal synthetic control group. We present our new framework in an application estimating the effect of Medicaid expansion on the prevalence of diabetes.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2018 program