Please note: This course will be taught in hybrid mode. Hybrid delivery of courses will include synchronous live sessions during which on campus and online students will be taught simultaneously.
Ryan T. Moore is Associate Professor of Government at American University, Senior Social Scientist at The Lab @ DC, and Fellow at the US Office of Evaluation Sciences. Ryan’s research interests centre around statistical political methodology, with applications in American social policy. He develops and implements methods for political experiments, ecological data, missing data, causal inference, and geolocated data. Among his publications are sixteen peer-reviewed journal articles, a book chapter, and several software packages and applications. His work has appeared in Political Analysis, The Lancet, Nature Human Behaviour, the Journal of Public Policy, and the Journal of Policy Analysis and Management, among other outlets. Ryan received his Ph.D. in government and social policy, along with his A.M. in statistics, from Harvard University.
Course Content
While experiments provide the soundest evidence about causality, in the social sciences, many important causal questions are not subject to experimentation. Methods for causal inference from observational data attempt to bridge the gap between the experimental ideal and the realities of data and evidence as we find them in the empirical world. This course lays foundations in the potential outcomes model, then introduces a variety of methods for causal inference from observational data, each tailored toward particular data environment. Throughout, we will develop enough formal intuition to understand the methods and the settings in which they are appropriate. With that understanding, the course focuses on applications and substantive interpretation of causal results. Specific methods include subclassification, matching, regression adjustment, regression discontinuity designs, difference-in-differences, synthetic controls, and instrumental variables.
Course Objectives
Participants will gain understanding of the potential outcomes model, how experiments represent a model for causal inference, and how we seek to approximate this ideal in a variety of observational data settings. Throughout, participants will learn application in the R statistical language. This course is suitable for participants at a variety of levels, including exceptional undergraduates, master’s degree and Ph.D. students, and those with a Ph.D. This course will serve well as a first introduction to observational causal inference, or as a refresher in the most important applied methods.
Course prerequisites
Students should have encountered conventional topics in introductory statistics, such as null hypothesis significance tests, confidence intervals, and linear regression. We will reintroduce such topics as needed. Students should have some familiarity processing data with R or Stata, or be willing to learn.
Representative Background Reading
Hernán Miguel A. The C-Word: Scientific Euphemisms Do Not Improve Causal Inference From Observational Data. American Journal of Public Health. 2018 May 108(5):616-619. doi: 10.2105/AJPH.2018.304337
Barret, Malcolm, Lucy D’Agostino McGowan, and Travis Gerke. Causal Inference in R. 2025. Chapter 1: https://www.r-causal.org/chapters/01-casual-to-causal
Required Text: Cunningham, Scott. Causal Inference: The Mixtape (this will be provided by ESS).
Background knowledge required
Maths
Linear Regression = elementary
Statistics
OLS = elementary
Computer Background
R or Stata = elementary
Students should have some familiarity processing data with R or Stata, or be willing to learn.
Please note: Recordings will only be available to online students for the length of the course’s duration for the purpose of catching up on missed content.
