3.0

credits

Average Course Rating

This course introduces strategies for estimating causal effects from a counterfactual perspective when conditioning techniques, such as matching and regression, do not identify the parameter of interest. After a review of scenarios when such conditioning will fail, the course then presents intervention designs, explaining randomization from both a potential outcome and causal graph perspective. The challenges to implementation of these designs are then discussed, with a special focus on large-scale randomized trials in education research. The course then considers the most prominent designs for causal inference in observational research in the presence of troubling unobservables: instrumental variable estimators, pre-post longitudinal designs, regression discontinuity, and estimation via exhaustive mechanisms. The course concludes with a consideration of credible avenues for investigation when point identification cannot be achieved, including an analysis of bounds and the estimation of a provisional estimate followed by a sensitivity analysis.