# Causality: Methods of Causal Inference in the Social Sciences

**Course Provider: **Prof Richard Breen

### Aims

The course introduces students to the "potential outcomes" or "counterfactual" model of causality and covers contemporary approaches to identifying and estimating causal relationships using observational data from the social sciences.

**Rubric**: Topics covered in the class include the potential outcomes model of causality, randomized control trials, matching, propensity score analysis, inverse probability treatment weighting, robustness and sensitivity tests, natural experiments and instrumental variables, control functions, regression discontinuity designs, fixed effects, and difference in difference models.

The course focuses on the identification of causal effects, the assumptions on which causal claims rest, and the estimation of causal relationship using statistical models. Basic knowledge of probability and of statistical methods such as OLS regression and logit and probit models is a pre-requisite. There are no practical classes in this course but students will be required to estimate models (in Stata or, preferably, R) and interpret the results.

Week 1: Review of Probability

Week 2: Review of Least Squares regression and properties of estimators

Week 3: The counterfactual model of causality, the fundamental problem of causality, randomized control trials.

Week 4: Matching estimators; regression and propensity scores

Week 5: Inverse probability of treatment weighting; robustness analysis

Week 6: Instrumental variables; natural experiments

Week 7: Control functions

Week 8: Regression discontinuity; fixed effects; difference in difference models

On successfully completing this course, students should have an understanding of the central role of causality in the social sciences and they should be able to cast a critical eye on the causal claims that social scientists make. Students should also have acquired a thorough knowledge of the potential outcomes approach to causality, the central role of assumptions in identifying causal effects, and they should be able to estimate a wide range of models for causal inference.

Weekly two-hour lectures.

Weekly problem sets. These include both theoretical and applied problems.

- Fox, J. (2008)
*A Mathematical Primer for Social Statistics*. QASS 159, Sage. - Morgan, S.L. and C. Winship. (2014)
*Counterfactuals and Causal Inference: Methods and Principles for Social Research*(2nd edition), Cambridge University Press. - Angrist, J. and J-S. Pischke. (2009)
*Mostly Harmless Econometrics*. Princeton University Press.