ECON 7240 - Causal Inference and Machine Learning

Fall or Spring. 3 credits. Student option grading.

Prerequisite: ECON 6190.

C. Qiu.

This course introduces econometric and machine learning methods that are useful for causal inference. Modern empirical research often encounters datasets with many covariates or observations. We start by evaluating the quality of standard estimators in the presence of large datasets, and then study when and how machine learning methods can be used or modified to improve the measurement of causal effects and the inference on estimated effects. The aim of the course is not to exhaust all machine learning methods, but to introduce a theoretic framework and related statistical tools that help research students develop independent research in econometric theory or applied econometrics. Topics include: (1) potential outcome model and treatment effect, (2) nonparametric regression with series estimator, (3) probability foundations for high dimensional data (concentration and maximal inequalities, uniform convergence), (4) estimation of high dimensional linear models with lasso and related methods, (5) estimation of high dimensional generalised linear models with L1 regularisation, (6) introduction to other machine learning methods such as neural networks, regression trees and random forests, (7) inference on semiparametric models with high dimensional components, orthogonalisation, de-biased machine learning, (8) other related topics, such as balancing methods, treatment choice problems, etc. Class slides will be circulated and students are expected to read theoretic and empirical research papers that involve machine learning methods.