Matthew Tyler

My research develops new statistical methods for causal inference or to measure difficult-to-observe concepts. I use these new methods to answer important questions about campaigns & elections, political news, and partisan polarization. I am currently a Postdoctoral Scholar at the Democracy and Polarization Lab at Stanford University after having obtained my PhD in Political Science from Stanford in June 2021. My dissertation, Essays in Political Methodology, was advised by Justin Grimmer (chair), Jens Hainmueller, and Shanto Iyengar.

 

My job market paper shows how to estimate the causal effects of events without contemporaneous control units. This is a common problem; for example, researchers might try to gauge the effect of a national policy that affects every district, or try to examine the effect of department-wide police reform on individual police officers. In both cases, there may not be an obvious set of control cases, or we might not readily believe a control of convenience satisfies the "parallel trends" assumption in difference-in-differences. The core idea of my job market paper is to fit time series models on pre-treatment trends to forecast what would have happened for each unit had the treatment not been applied. These forecasted synthetic controls are then used to estimate the average or heterogeneous treatment effects of the event in question. My job market paper uses this new method to estimate the effect of FBI Director James Comey's letter (re: Hillary Clinton's handling of classified information) on political engagement in the 2016 US presidential campaign. Since every individual was potentially treated by the news coverage of Comey's letter on Oct. 28, 2016, there is no valid control group to measure the effect with conventional means. I use the forecasted synthetic control method on a panel of individual web browsing histories to show how Comey's letter had a large and enduring impact on the political engagement of undecided voters.