Factorial Designs, Model Selection, and (Incorrect) Inference in Randomized Experiments

We are grateful to Isaiah Andrews, Tim Armstrong, Prashant Bharadwaj, Arun Chandrasekhar, Clement de Chaisemartin, Gordon Dahl, Stefano DellaVigna, Esther Duflo, Graham Elliott, Andrew Gelman, Markus Goldstein, Macartan Humphreys, Guido Imbens, Hiroaki Kaido, Lawrence Katz, Michal Kolesar, Adam McCloskey, Craig McIntosh, Rachael Meager, Paul Niehaus, Ben Olken, Gautam Rao, Andres Santos, Jesse Shapiro, Diego Vera-Cossio, and many seminar participants for comments and suggestions. We are also grateful to the authors of the papers we reanalyze for answering our questions and fact checking that their papers are characterized correctly. Finally, we would like to thank Tim Armstrong, Adam McCloskey, Graham Elliott, Michal Kolesar, and Soonwoo Kwon, who graciously answered questions about the econometric methods they developed and how to implement them. Sameem Siddiqui provided excellent research assistance. All errors are our own. Financial support from the Asociación Mexicana de Cultura, A.C. is gratefully acknowledged by Romero.

A supplemental appendix is available online at https://doi.org/10.1162/rest_a_01317.