[From the working paper, “Which findings should be published?” by Alexander Frankel and Maximilian Kasy]

“There have been calls for reforms in the direction of non-selective publication. One proposal is to promote statistical practices that de-emphasize statistical significance … Another proposal is for journals to adopt Registered Reports, in which pre-registered analysis plans are reviewed and accepted prior to data collection … Registered Reports guarantee that publication will not select at all on findings…”

“…In this paper we seek the optimal rule for determining whether a study should be published … In this framework, we will show that non-selective publication is not in fact optimal. Some findings are more valuable to publish than others. Put differently, we will find a trade-off between policy relevance and statistical credibility.”

“… The optimal publication rule defined in this manner selects on a study’s findings.”

To read the article, click here.