"Generalized Synthetic Control Method: Causal Inference with Interactive Fixed Effects Models” Political Analysis 25/1
Selection committee: Jennifer Pan (Stanford), Pablo Barberá (LSE), and Jonathan Katz (CalTech)
On behalf of this year's Miller Prize Committee (Jennifer Pan, Pablo Barberá, and Jonathan Katz, any myself (recused)), I am pleased to announce that the 2018 Miller Prize for the best work appearing in Political Analysis (in 2017) was awarded to Yiqing Xu for his article entitled "Generalized Synthetic Control Method: Causal Inference with Interactive Fixed Effects Models” (https://doi.org/10.1017/pan.2016.2). Please join me in congratulating Xu on this honor. The award citation is pasted below. Thank you to the committee for their work on this.
Time-series cross-sectional (TSCS) data is widely used in the social sciences to examine the impacts of policies and events on aggregate units such as states, schools, or firms. Commonly researchers employ a difference-in-differences framework that simply assumes away the presence of time-varying confounds by imposing a parallel trends assumption. However, since units are heterogeneous and often follow their own trends this parallel trends assumption is unrealistic in many social science settings, leading to invalid inferences. Xu’s Political Analysis article “Generalized Synthetic Control Method: Causal Inference with Interactive Fixed Effects Models” makes an important contribution that advances the methodology for causal inference with TSCS data. It provides researchers with a flexible estimator that constructs more plausible counterfactual trends by generalizing the synthetic control method and combining it with interactive fixed effects. Xu’s method allows for time-varying confounds and contains the often used two-way fixed effects model as a special case. Xu's estimator is widely applicable and provides a unique contribution that highlights how political methodology can contribute to exporting methodological advances to other disciplines.