Do negative campaign ads help politicians win elections? Are there learning benefits to regular class attendance at university? Is maternal employment good for children’s development and, if so, why? If you are interested in answering such or similar questions, you are in the business of causal inference; your aim is to use empirical data to learn about causal relations between specific variables.

This class provides an introduction to the state-of-the-art theory and practice for causal inference. It goes beyond the notion that statistical association doesn’t necessarily imply causation and specifies precisely under which conditions we can endow correlations with a causal interpretation. In addition to conventional regression approaches the class introduces inverse probability weighting as a useful statistical technique for causal inference. The main topics of the class are the construction and application of graphical causal models, the precise definition of causal effects in terms of interventions and counterfactuals, and the conditions and techniques to learn causal effects from data.

The focus is never on the mathematical derivation of statistical methods, but on an intuition for the conditions under which these methods allow valid causal inference, and for the scenarios under which they break down. Providing step-by-step practical guidance for (1) specifying prior knowledge and assumptions regarding the research question, for (2) defining the effect of interest, for (3) assessing whether this effect can be learned from available data, and for (4) statistical estimation (in Stata) and (5) interpretation of the results, this class will strengthen your ability to evaluate existing research, aid you in the formulation of precise and novel research questions, and provide tools for answering them. To facilitate learning, lectures are complemented with assignments and student collaboration.