This three-day workshop provided an introduction to the potential outcomes approach to causal inference. This approach, also known as the counter factual framework or Rubin model, is increasingly becoming the dominant framework for causal inference in the social and biomedical sciences. It has many natural applications in policy and program evaluation.
This workshop introduced participants to the foundational insights of the potential outcomes approach to causality and demonstrated its utility using a range of hands-on examples from education, program evaluation, and social demography. A thorough and substantively grounded conceptual understanding was privileged over abstract mathematical proofs and out-of-context software recipes.
- Conceptual basics
- Potential outcomes
- The “Fundamental Problem of Causal Inference”
- Randomized experiments
- Experimental analogy for observational data
- Propensity score estimation
- Relationship to standard methods
- Change score estimation
- Instrumental variables
- Types of causal effects
- Point vs time-varying treatments
- Direct and indirect causal effects
- Dynamic vs static treatment regimes
- Graphical rules of identification (Pearl’s DAGs)
- The importance of prior knowledge for causal estimation
- Inverse probability of treatment weighting (IPTW)
- Marginal structural models for longitudinal data
Participants were encouraged to read pages 659-69, 671-78 in Winship, Christopher and Stephen L. Morgan. 1999. “The Estimation of Causal Effects from Observational Data.” Annual Review of Sociology 1999.
Hamilton, Lawrence C. 2006. Statistics with Stata. Brooks/Cole Publishers, particularly chapters 1, 2, 4, 6, 10, 11, provides a good introduction to the Stata software package.
Christopher Winship is the Diker-Tishman Professor of Sociology and also a member of the faculty at the John F. Kennedy School of Government at Harvard University. He is editor of Sociological Methods and Research. His research on problems of causation and selection has appeared in The American Sociological Review, The American Journal of Sociology, Sociological Methodology, The Journal of Mathematical Sociology, and elsewhere. He is currently writing a book on causal inference. A noted teacher of statistics for social scientists, he frequently lectures on causality in the U.S. and abroad. His current applied research deals with the effect of schooling on mental ability, the cooperation between the Boston Police Department and the Ten Point Coalition of black inner-city ministers, and institutional differences in the college graduation rates of minorities. Before coming to Harvard in 1992 he was a professor of Sociology and Economics at Northwestern University, where he also was a founding member of the Department of Statistics.
Felix Elwert is a graduate student in the Department of Sociology and the Department of Statistics at Harvard University. He works on the application of causal inference in sociology and demography. Current applied work concerns the estimation of interspousal health effects and the causes and consequences of death and divorce for American families. His work on race differences in the widowhood effect is forthcoming in the American Sociological Review. Elwert has taught causal inference in various academic and professional settings from Harvard to Uzbekistan. He is a Graduate Associate of the Institute for Quantitative Social Science at Harvard.