Unlike data we obtain from experiments, observational data rarely come nicely pre-packaged for hypothesis testing, and especially less so, for causal inference. Yet some of the most central political processes and outcomes, such as terrorism, repression, and other forms of political violence, are often not compatible with experimental research. For the study of many such processes, observational data are all we will ever have access to. Students will learn the research tools and methods for causal inference with observational data, the assumptions behind these tools, and what common pitfalls to watch out for when working with non-experimental data.
The course covers the principles of panel and temporal design, difference-in-difference design, regression disountinuity design, and the instrumental variable approach, as well as network analysis, web-scraping, and text analysis. Applications draw from topics in the study of international relations, economics, political behavior, and statistics to offer a diverse set of tools for processing and analyzing different types of data. Applications include war and conflict, terrorism, international trade, social media, elections, and representation.