Abstract
The traditional way to verify the adequacy of a theory to the data has been to perform a single experiment in which the assumptions underlying the theory are tested. While one experiment in itself is not sufficient to guarantee the acceptance or the rejection of a theory, attempts to replicate an already performed experiment have often been met with skepticism. Big team science has the potential to overcome this, making it possible to perform several instances of the same experiment in different countries and contexts and to considerably increase the sample size.
This makes it possible to reliably test several related theories at the same time. As recently expressed, ‘initiatives to assess the robustness of findings […] should aim to simultaneously test competing ideas operating in the same theoretical space’ (Tierney et al., 2020, p. 291). This approach has been likened to the gales of creative destruction that were proposed by Schumpeter in innovation economics.
The availability of statistical methods for the comparison of a collection of models incorporating these competing ideas is a prerequisite for this process of simultaneous verification of several theories. The aim of this presentation is to discuss these methods and make a case for information criteria as a general method to perform multi-criteria model selection in a creative destruction setting.
References
Warren Tierney, Jay H. Hardy III, Charles R. Ebersole, Keith Leavitt, Domenico Viganola, Elena Giulia Clemente, Michael Gordon, Anna Dreber, Magnus Johannesson, Thomas Pfeiffer, Hiring Decisions Forecasting Collaboration, Eric Luis Uhlmann (2020), Creative destruction in science, Organizational Behavior and Human Decision Processes, 161, 291-309.
Raffaello Seri
Professor of Econometrics
My research interests include statistics, numerical analysis, operations research, psychology, economics and management.