Replication data for: Contributions to Causal Inference for Political Science

This thesis presents five independent essays that advance causal inference in political science. It is divided into a methodological and an empirical part. The methodological part presents a suite of statistical techniques, called synthetic inference methods, that allows researchers to construct control groups that more accurately resemble the treated units than is possible by commonly used methods of covariate adjustment. Synthetic inference is based on the idea of using weighted averages of control units to create so-called synthetic comparison units or comparison groups that resemble either a single treated unit or a group of treated units in all relevant characteristics that may confound a comparison. The thesis develops two variants of synthetic inference for empirical settings that are typically encountered in political science: comparative case studies (essay 1, with Alberto Abadie and Alexis Diamond) and cross-sectional studies (essay 2). Both methodological essays demonstrate the methods in real world applications and provide companion software for implementation . The empirical part of the thesis presents three original empirical studies that contribute answers to previously unanswered causal questions about (a) the financial rewards to serving in Parliament (essay 3, with Andy Eggers), (b) the impact of foreign free media on the stability of authoritarian regimes (essay 4, with Holger Kern), and (c) the impact of economic concerns on public attitudes toward immigration essay 5). The empirical essays advance the debates in these substanti ve fields by combining newly collected data and a design-based approach to causal inference. Design-based inference provides an effective strategy to identify valid control groups in settings where statistical control is insufficient since units potentially differ on more characteristics than can be measured and controlled for in a statistical model.