Research > Research Themes > Innovations in Methods > Statistics and econometrics

Rosenbaum’s work is highly influential in population studies, since the primary focus of his career has been the problem of causal inference in observational data. In his text Observational Studies and high-profile publications in Journal of the American Statistical Association and elsewhere he has been an innovator of propensity scores, the statistical model for causal inference, non-parametric matching, and sensitivity analysis for unobservables. Berk, who was at the CCPR at UCLA, has been dedicated to the practical application of a variety of statistical methods to real-world policy problems. He won the Lazarsfeld Award from the American Sociological Association for a series of methodological insights summarized in Regression Analysis: A Constructive Critique, which is a skeptical, forward-looking treatment of statistical practice in policy and social science. He is currently working on ensemble methods, or “data mining.” Allison is also a Lazarsfeld Award winner; he has made innovative contributions in areas from event history analysis to change score models to imputation of missing data to fixed effects models. He is very influential; in addition to high profile scientific papers, he has “vertically integrated” his expertise through texts, two published through SAS (the software company) and short courses given in Philadelphia and around the world. Smith has several publications in Sociological Methodology and elsewhere, introducing or giving new applications of the analysis of association, Rasch models for paired population data, the Rubin model and causal inference; and is currently working on problems of inference for “double samples” of nonrespondents. This is a group on a rough axis statistics-sociology-demography. Most all of the methods have counterparts, in the isomorphic, arithmetic sense, within econometrics—specifically microeconometrics—but econometrics is usually differentiated since its methods derive from economic theory and problems.

The PSC’s signature strength in microeconometrics is organized on a continuum from nonparametric to fully parametric, highly structured approaches to inference, with Todd (who is well-known for work on matching estimators) at one end and Wolpin, who has been an innovator in developing dynamic micro level structural econometric models with multiple future states of the world that permit investigation of counterfactual policies at the population level. De Paula and F. Cunha, who is about to join the PSC, “sit” somewhere in the middle, although everyone works together, and the continuum is both fluid and embedded in the core problem of ex ante policy evaluation: How do you evaluate policies that have not been tried? The statistics and econometrics group can be cognitively restructured along other dimensions, e.g., expertise on propensity score matching (Rosenbaum/Todd/Smith) or innovative thinking about experiments in population settings, which would include Rosenbaum, Berk, Smith, Todd, Wolpin, Behrman, and Mutz, who is PI of an NSF project that embeds experimental methods in general population surveys.