This research project received funding through the 2023 Institute for Diversity Science Seed Grant Program
Principal Investigator: Felix Elwert, Sociology and Biostatistics and Medical Informatics, UW–Madison
Co-Investigator: Ang Yu, Department of Sociology, UW–Madison
Abstract: We develop a new framework for explaining the causal contribution of a (social, psychological, or economic) factor on group-based outcome inequalities. Our framework innovates beyond existing methods in two crucial respects. First, we identify a previously overlooked mechanism, by which the differential selection of the factor to individuals within groups can lead to outcome inequalities across groups. Second, we develop a suite of highly flexible machine-learning estimators with several desirable statistical properties. In contrast to existing estimators, which rely on generalized linear models, our estimators do not require any functional-form or distributional assumptions, are root-n consistent, asymptotically normal, semiparametrically efficient, and doubly robust. This is an advanced project that generalizes our ongoing work from binary to categorical and continuous group definitions and causal factors, and from continuous to binary and categorical outcomes. We provide an easy-to-use open-source R package to facilitate the implementation of our framework by empirical diversity scientists.