Table of Contents

Part I. Introduction: 1. The basic framework: potential outcomes, stability, and the assignment mechanism; 2. A brief history of the potential-outcome approach to causal inference; 3. A taxonomy of assignment mechanisms; Part II. Classical Randomized Experiments: 4. A taxonomy of classical randomized experiments; 5. Fisher's exact P-values for completely randomized experiments; 6. Neyman's repeated sampling approach to completely randomized experiments; 7. Regression methods for completely randomized experiments; 8. Model-based inference in completely randomized experiments; 9. Stratified randomized experiments; 10. Paired randomized experiments; 11. Case study: an experimental evaluation of a labor-market program; Part III. Regular Assignment Mechanisms: Design: 12. Unconfounded treatment assignment; 13. Estimating the propensity score; 14. Assessing overlap in covariate distributions; 15. Design in observational studies: matching to ensure balance in covariate distributions; 16. Design in observational studies: trimming to ensure balance in covariate distributions; Part IV. Regular Assignment Mechanisms: Analysis: 17. Subclassification on the propensity score; 18. Matching estimators (Card-Krueger data); 19. Estimating the variance of estimators under unconfoundedness; 20. Alternative estimands; Part V. Regular Assignment Mechanisms: Supplementary Analyses: 21. Assessing the unconfoundedness assumption; 22. Sensitivity analysis and bounds; Part VI. Regular Assignment Mechanisms with Noncompliance: Analysis: 23. Instrumental-variables analysis of randomized experiments with one-sided noncompliance; 24. Instrumental-variables analysis of randomized experiments with two-sided noncompliance; 25. Model-based analyses with instrumental variables; Part VII. Conclusion: 26. Conclusions and extensions.

Promotional Information

This text presents statistical methods for studying causal effects and discusses how readers can assess such effects in simple randomized experiments.

About the Author

Guido W. Imbens is Professor of Economics at the Graduate School of Business, Stanford University. He has held tenured faculty positions at Harvard University, the University of California, Los Angeles, the University of California, Berkeley, and Stanford University. He is a fellow of the Econometric Society and the American Academy of Arts and Sciences. Imbens has published widely in economics and statistics journals, including Econometrica, The American Economic Review, the Annals of Statistics, the Journal of the American Statistical Association, Biometrika, and the Journal of the Royal Statistical Society. Donald B. Rubin is John L. Loeb Professor of Statistics at Harvard University, where he has been professor since 1983 and department chair for thirteen of those years. He has authored or coauthored nearly four hundred publications (including ten books), has four joint patents, and has made important contributions to statistical theory and methodology, particularly in causal inference, design and analysis of experiments and sample surveys, treatment of missing data, and Bayesian data analysis. Rubin has received the Samuel S. Wilks Medal from the American Statistical Association, the Parzen Prize for Statistical Innovation, the Fisher Lectureship, and the George W. Snedecor Award from the Committee of Presidents of Statistical Societies. He was named Statistician of the Year by the American Statistical Association, Boston and Chicago chapters. He is one of the most highly cited authors in mathematics and economics with nearly 150,000 citations to date.

Reviews

'This book offers a definitive treatment of causality using the potential outcomes approach. Both theoreticians and applied researchers will find this an indispensable volume for guidance and reference.' Hal Varian, Chief Economist, Google, and Emeritus Professor, University of California, Berkeley

'By putting the potential outcome framework at the center of our understanding of causality, Imbens and Rubin have ushered in a fundamental transformation of empirical work in economics. This book, at once transparent and deep, will be both a fantastic introduction to fundamental principles and a practical resource for students and practitioners. It will be required readings for any class I teach.' Esther Duflo, Massachusetts Institute of Technology

'Causal Inference sets a high new standard for discussions of the theoretical and practical issues in the design of studies for assessing the effects of causes - from an array of methods for using covariates in real studies to dealing with many subtle aspects of non-compliance with assigned treatments. The book includes many examples using real data that arose from the authors' extensive research portfolios. These examples help to clarify and explain many important concepts and practical issues. It is a book that both methodologists and practitioners from many fields will find both illuminating and suggestive of further research. It is a professional tour de force, and a welcomed addition to the growing (and often confusing) literature on causation in artificial intelligence, philosophy, mathematics and statistics.' Paul W. Holland, Emeritus, Educational Testing Service

'A comprehensive and remarkably clear overview of randomized experiments and observational designs with as-good-as-random assignment that is sure to become the standard reference in the field.' David Card, Class of 1950 Professor of Economics, University of California, Berkeley

'This book will be the 'Bible' for anyone interested in the statistical approach to causal inference associated with Donald Rubin and his colleagues, including Guido Imbens. Together, they have systematized the early insights of Fisher and Neyman and have then vastly developed and transformed them. In the process they have created a theory of practical experimentation whose internal consistency is mind-boggling, as is its sensitivity to assumptions and its elaboration of the key 'potential outcomes' framework. The authors' exposition of random assignment experiments has breadth and clarity of coverage, as do their chapters on observational studies that can be readily conceptualized within an experimental framework. Never have experimental principles been better warranted intellectually or better translated into statistical practice. The book is a 'must read' for anyone claiming methodological competence in all sciences that rely on experimentation.' Thomas D. Cook, Joan and Sarepta Harrison Chair of Ethics and Justice, Northwestern University, Illinois

'In this wonderful and important book, Imbens and Rubin give a lucid account of the potential outcomes perspective on causality. This perspective sensibly treats all causal questions as questions about a hidden variable, indeed the ultimate hidden variable, 'What would have happened if things were different?' They make this perspective mathematically precise, show when and to what degree it succeeds, and discuss how to apply it to both experimental and observational data. This book is a must-read for natural scientists, social scientists and all other practitioners who seek new hypotheses and new truths in their complex data.' David Blei, Columbia University, New York

'This thorough and comprehensive book uses the 'potential outcomes' approach to connect the breadth of theory of causal inference to the real-world analyses that are the foundation of evidence-based decision making in medicine, public policy and many other fields. Imbens and Rubin provide unprecedented guidance for designing research on causal relationships, and for interpreting the results of that research appropriately.' Mark McClellan, Director of the Health Care Innovation and Value Initiative, Brookings Institution, Washington DC

'This book will revolutionize how applied statistics is taught in statistics and the social and biomedical sciences. The authors present a unified vision of causal inference that covers both experimental and observational data. They do a masterful job of communicating some of the deepest, and oldest, issues in statistics to readers with disparate backgrounds. They closely connect theoretical concepts with applied concerns, and they honestly and clearly discuss the identifying assumptions of the methods presented. Too many books on statistical methods present a menagerie of disconnected methods and pay little attention to the scientific plausibility of the assumptions that are made for mathematical convenience, instead of for verisimilitude. This book is different. It will be widely read, and it will change the way statistics is practiced.' Jasjeet S. Sekhon, Robson Professor of Political Science and Statistics, University of California, Berkeley

'Clarity of thinking about causality is of central importance in financial decision making. Imbens and Rubin provide a rigorous foundation allowing practitioners to learn from the pioneers in the field.' Stephen Blyth, Managing Director, Head of Public Markets, Harvard Management Company

'A masterful account of the potential outcomes approach to causal inference from observational studies that Rubin has been developing since he pioneered it fourty years ago.' Adrian Raftery, Blumstein-Jordan Professor of Statistics and Sociology, University of Washington

'Correctly drawing causal inferences is critical in many important applications. Congratulations to Professors Imbens and Rubin, who have drawn on their decades of research in this area, along with the work of several others, to produce this impressive book covering concepts, theory, methods and applications. I especially appreciate their clear exposition on conceptual issues, which are important to understand in the context of either a designed experiment or an observational study, and their use of real applications to motivate the methods described.' Nathaniel Schenker, Statistician

'The book is well-written with a very comprehensive coverage of many issues associated with causal inference. As can be seen from its table of contents, the book uses multiple perspectives to discuss these issues including theoretical underpinnings, experimental design, randomization techniques and examples using real-world data.' Carol Joyce Blumberg, International Statistical Review

'Guido Imbens and Don Rubin present an insightful discussion of the potential outcomes framework for causal inference … this book presents a unified framework to causal inference based on the potential outcomes framework, focusing on the classical analysis of experiments, unconfoundedness, and noncompliance. The book has become an instant classic in the causal inference literature, broadly defined, and will certainly guide future research in this area. All researchers will benefit from carefully studying this book, no matter what their specific views are on the subject matter.' Matias D. Cattaneo, Journal of the American Statistical Association

'Guido Imbens and Donald Rubin have written an authoritative textbook on causal inference that is expected to have a lasting impact on social and biomedical scientists as well as statisticians. Researchers have been waiting for the publication of this book, which is a welcome addition to the growing list of textbooks and monographs on causality … the authors should be congratulated for the publication of this impressive volume. The hook provides a unified introduction to the potential outcomes approach with the focus on the basic causal inference problems that arise in randomized experiments and observational studies.' Alicia A. Lloro, Journal of the American Statistical Association