Causal Inference with Cross-Sectional Data

Description: This course covers the potential outcomes approach to identification and estimation of causal (or treatment) effects in several situations that arise and various empirical fields. The settings include unconfounded treatment assignment (with randomized assignment as a special case), confounded assignment with instrumental variables, and regression discontinuity designs. We will cover doubly robust estimators assuming unconfoundedness and discuss covariate balancing estimators of propensity scores. Local average treatment effects, and some recent results on including covariates in LATE estimation, also will be treated. Regression discontinuity methods, both sharp and fuzzy designs, and with control variables, round out the course.

Participants should have good working knowledge of ordinary least squares estimation and basic nonlinear models such as logit, probit, and exponential conditional means. Sufficient background is provided by my introductory econometrics book, Introductory Econometrics: A Modern Approach, 7e, Cengage, 2020. My book Econometric Analysis of Cross Section and Panel Data, 2e, MIT Press, 2010, covers some material at a higher level. I will provide readings for some of the more advanced material. While the focus here is on cross-sectional data, many of the methods have been applied to panel data settings, particularly to difference-in-differences designs. Course material, including slides and Stata files, will be made available via Dropbox.

The lectures and the Q&A sessions will be recorded and the recordings will be made available to participants.