Identification and Causal Inference

Abstract

Most policy relevant research questions in the social sciences face the same challenge: How can we identify a causal impact of one variable on another when we cannot use a controlled experiment? This course will teach program evaluation methods for causal analysis based on non-experimental (i.e. observational) data, derive the underlying theory and discuss recent applications.

Objective

The main objective of this course is to make PhD students familiar with program evaluation methods to estimate causal effects. The course covers Difference-in-Differences/Event Study estimations, Instrumental Variables Estimators, Regression Discontinuity Designs, Matching Methods, and Post-Double Machine Learning. The course will cover the underlying theory, illustrate the connection to classical regression analysis and randomized control trials (field experiments), show how these different methods relate to each other and how they differ in terms of the required identifying assumptions as well as data needs. Recent research papers will be discussed to illustrate their use. The course has an applied focus. The goal is to place students in the position to have a broad toolkit of quasi-experimental methods and to apply these methods in their empirical research.

Resources