“On Theory and Identification: When and Why We Need Theory for Causal Identification,” Tara Slough, Columbia University
QUANTITATIVE RESEARCH METHODS WORKSHOP
What is the role for theory in identification-driven research designs? I argue that not only is theory important for the interpretation of causal findings, but in many research designs, theory is necessary for the identification of standard causal estimands. In particular, I show that when empiricists study a sequence of post-treatment behavioral outcomes, post-treatment selection can prevent the identification of standard causal estimands, even when standard (empirical) identification assumptions hold. In these cases, articulation of a theory, or model of the world, that defines the post-treatment selection processes is necessary to define a set of identified estimands. Using a stylized example of crime, reporting, and recording, I illustrate how variants of a theory imply the identification of different estimands, holding constant the research design. I then present this result more generally by considering the conditions under which the invocation of a theory is necessary for causal identification. I consider the implications for different research designs. This paper illustrates the need for theory in identification-driven research and provides guidance for research design.
Tara Slough is a Ph.D. candidate in Political Science at Columbia University and a Predoctoral Fellow at the University of California, Berkeley. In 2020, she will join New York University as Assistant Professor of Politics. Her research interests include the political economy of institutions and development as well as applied methodology. Tara is a recipient of the National Science Foundation Graduate Research Fellowship and Doctoral Dissertation Improvement Grant. Tara earned undergraduate degrees in violin performance and political science from Rice University and was a professional violinist before graduate school.
The Quantitative Research Methods Workshop series is being sponsored by the ISPS Center for the Study of American Politics and The Whitney and Betty MacMillan Center for International and Area Studies at Yale with support from the Edward J. and Dorothy Clarke Kempf Fund.