400010 SE Advanced research design and causal inference (2023S)

This course provides an overview of different advanced quantitative methods that are used in the social sciences to draw inferences about causal relationships from large-N observational data. We first introduce Neyman and Rubin's "potential outcomes framework" of causality that forms a theoretical basis for the course and then survey five different classes of popular methods for causal inference:

1) Matching methods
2) Instrumental variables
3) Difference-in-differences
4) Synthetic control
5) Regression discontinuity designs

These methods claim to advance on standard regression models by adjusting for selection bias on observables and unobservables. With regard to each method covered, we will address its theoretical foundations and assumptions, practical considerations and challenges, critical discussions of applications, implementation in software as well as interpretation of results.

Each session will consist of 1) an interactive lecture element, in which you participate through live polls and mini tasks; 2) a computer lab, in which you practice the implementation of the methods in R (or STATA); 3) a discussion of an application of the method in a published research paper.

While this is a course in advanced research design using quantitative methods, no prior knowledge of causal inference methods is expected. A basic understanding of quantitative methods (e.g. multiple regression analysis) is desired, but students with strong motivation may also acquire this knowledge in parallel to the course. The first session of the course will provide a quick review of multiple regression analysis. A solid understanding of research design in the social sciences is assumed. Some prior familiarity with R is an asset; some familiarity with STATA is helpful.