Online Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 563 - Statistical Methods for State Health Policy Evaluation
Type: Invited
Date/Time: Thursday, August 6, 2020 : 3:00 PM to 4:50 PM
Sponsor: Health Policy Statistics Section
Abstract #309445
Title: Partial Identification in Difference-In-Difference Designs When the Parallel Trends Assumption Fails
Author(s): Raiden Hasegawa* and Dylan Small and Luke Keele and Daniel Webster
Companies: Google and University of Pennsylvania and University of Pennsylvania and Bloomberg School of Public Health, Johns Hopkins University
Keywords: Difference-in-Difference; Parallel Trends; Gun Violence; Voter ID laws; Partial Identification; State Policy Evaluation

In the comparative interrupted time series design, the change in outcome in a group exposed to treatment in the periods before and after the exposure is compared to the change in outcome in a control group not exposed to treatment in either period. The standard difference-in-difference estimator for a comparative interrupted time series design is biased when the assumption that the counterfactual trends under control for the comparison groups are parallel fails. Under a weaker set of assumptions than parallel trends, we provide a partial identification strategy that requires constructing two candidate control groups that systematically vary on confounding factors that interact with history. Next, we develop a method to assess the sensitivity of these estimates to departures from the assumptions required for partial identification. Finally, we present a graphical placebo test for the interval estimate produced by our bracketing method. We illustrate this collection of methods with two state-level policy analyses: a study of the repeal of a handgun purchasing law in Missouri and one of the enactment of stricter voter ID laws in Georgia and Indiana.

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

Back to the full JSM 2020 program