Did mandatory busing programs in the 1970s increase the school achievement of disadvantaged minority youth? Does obtaining a college degree increase an individual’s labor market earnings? Did the use of a butterfly ballot in some Florida counties in the 2000 presidential election cost Al Gore votes? Simple cause-and-effect questions such as these are the motivation for much empirical work in the social sciences. In this book, the counterfactual model of causality for observational data analysis is presented, and methods for causal effect estimation are demonstrated using examples from sociology, political science, and economics.
“This book is the first representative of a growing surge of interest among social scientists and economists to reclaim their professions from the tyrany of regression analysis and address cause-effect relationships squarely and formally. The book is unique in recognizing the equivalence between the counterfactual and graphical approaches to causal analysis and shows readers how to best utilize the distinct features of each. An indispensible reading for every forward-looking student of quantitative social science.” -Judea Pearl University of California, Los Angeles
“…Morgan and Winship have written an important, wide-ranging, careful, and original introduction to the modern literature on causal inference in nonexperimental social research.”
Canadian Journal of Sociology
This book is an excellent and relatively non-technical review of causal inference in the social sciences. The authors condense a huge literature that spans economics, statistics, sociology, philosophy, medical statistics, and computer science into manageable pieces appropriate for scholars and graduate students in the social sciences.
The authors’ primary contribution is linking the work on causal inference in diverse fields together, presenting a theoretically coherent view of causal inference that draws extensively on Judea Pearl’s work in philosophy and machine learning (see his book Causality: Models, Reasoning and Inference). The authors successfully illuminate the equations underlying the work of Paul Rosenbaum, Donald Rubin, Charles Manski, James Heckman, Joshua Angrist, Guido Imbens, James Robins, and Paul Holland (along with many others) by connecting them to Pearl’s fundamentally graphical view of causal thinking. The authors allow readers to grasp such a broad selection of research by presenting each element as a natural extension of an overarching theoretical perspective.
The book covers the strengths and weaknesses of many popular quasi-experimental approaches to causal inference, including conditioning (aka “controlling for other variables”), instrumental variables/natural experiments, case-to-case matching, propensity score matching, propensity score blocking, and propensity score weighting. It also presents a great overview of Charles Manski’s work on minimal identification approaches (i.e., “let’s see what the data can tell us if we invoke as few assumptions as possible”). Additionally, the book contains a chapter on causal inference and repeated observations/longitudinal data. The book leaves aside issues of variance estimation using these approaches, presumably because of its more technical nature and the large amount of research activity currently in progress.
This book is not a research “cookbook” in the sense that it will provide code snippets illustrating each technique (or any code snippets at all for that matter), so you will be disappointed if that’s what you are after. Its value is in providing a theoretically united and up-to-date review of causal inference in the social sciences (so you will actually know what you’re talking about as compared to simply pasting code into Stata/SAS/R/whatever).
This book should be on the shelf of any self-respecting quantitative social scientist, and it will provide serious intellectual fodder for anyone interested in causal inference more generally.
Product Details :
- Series: Analytical Methods for Social Research
- Paperback: 328 pages
- Publisher: Cambridge University Press; 1 edition (July 30, 2007)
- Language: English
- ISBN-10: 0521829984
- ISBN-13: 978-0521671934
- ASIN: 0521671930
- Product Dimensions: 9.1 x 6.1 x 0.9 inches