Statistics for Causal Inference in the Social Sciences
The goal of this course is to have students propose research questions that they are interested in, and to examine how (or whether) these questions could be answered. The main concern is with thinking clearly about answering empirical questions.
To this end, we will study the statistical foundations of two frameworks for drawing causal inference based on observational data. The first framework is based on approximating experimental design. I devote most attention to the potential outcome approach and matching. Instrumental variable methods are also discussed in this context. We will also consider for a small fraction of the course a second framework based on empirically testing the qualitative features of comparative statics of equilibrium relationships derived from a formal model.
As part of this course we will also pay particular attention to use of non-parametric tests for comparing empirical distributions.