The mechanics of statistical inference suggest that surprising novel empirical insights should not appreciably move priors about a phenomenon under investigation, even when statistically significant within the study. A few independent replications dramatically improve reliability, but incentives to replicate are rarely in place within the social sciences. This paper proposes a novel mechanism that promotes replications by leveraging mutually beneficial gains from trade between scholars, and uses experimental economics to highlight the approach. Our mechanism is simple: upon completion of a study finding surprising and significant results, the authors make the working paper available online, but commit never to submit it to a journal for publication. Instead, they offer co-authorship for a second, yet to be written, paper to scholars willing to independently replicate the study. The second paper references the original working paper, includes all pre- registered replications, and it is submitted to a peer- reviewed journal for publication. We apply our method to an investigation of the effects of Knightian uncertainty (ambiguity) on cooperation in allocation games, a pervasive and yet largely unexplored feature of most public goods. The original, voluntarily unpublished study (Butera and List 2017) unexpectedly found that ambiguity about the value of a public good facilitates cooperation. We report results from the original study and three independent replications, and show that while ambiguity has a positive effect in two replications for low-quality public goods, overall the original results do not pass a conservative replication test. We conclude that Knightian uncertainty likely has a limited impact on cooperation, corroborating the existing approach of focusing on strategic uncertainty.
Resum
Baixar pdf
Leave Comment