Much economic research seeks to combine economic and statistical modeling to distill
policy-relevant conclusions from data. The transparency of this process—whether the intended
audience can understand the evidence behind the conclusions as well as the conclusions
themselves—is crucial to its success. In this project, we develop a framework for understanding
the communication of scientific findings to an audience, apply the framework to important
problems in empirical economics, and study how economists can reach more robust
policy-relevant conclusions from data.
The project will produce three academic articles. The first, "A Model of Scientific
Communication" (Andrews, Shapiro), will use economic theory and statistical decision theory to
propose and develop a model of scientific communication and contrast its implications with
those of a classical model of statistics. The second, "Transparency in Structural Estimation"
(Andrews, Gentzkow, Shapiro), will define a working notion of transparency for structural
estimation and discuss approaches to achieving it. The third, "Instrument Selection and
Sensitivity to Misspecification in Structural Models" (Andrews, Barahona, Gentzkow,
Rambachan, and Shapiro), will develop and illustrate the importance of instrument choice in
determining the sensitivity of estimates of structural models to misspecification.