The use of synthetic control and other covariate adjustment strategies for policy evaluation (306613)
Keywords: opioid policy, synthetic control, difference-in-differences, interrupted time series
The broad class of comparative interrupted time series methods (sometimes termed “difference-in-differences”) have become a useful tool for studying the effects of programs or policies that come into effect at a particular point in time, when measures of the outcome of interest are available before and after the policy/program change, and for locations that experienced the policy shift and some comparison locations that did not. A benefit of these approaches is that they can be used even when only aggregate data is available at each location. Use of these methods, and methods development around them, is skyrocketing, but many statistical questions remain, including such as 1) when individual level data is available, what are the right strategies to use, 2) how to adjust for differences between intervention and comparison locations, and 3) how to articulate and assess the underlying assumptions. This talk will prevent an overview of current research in this area, including synthetic control and newer augmented synthetic control approaches, and present motivating examples from gun policy and opioid policy.