Experimentation and manipulation with preregistration

Abstract

Preregistration requires scientists to describe the planned research activities before their project begins. Preregistration improves transparency in empirical research and is an institutional response to scientific misconduct. This paper studies the impact of a preregistration requirement in a model in which a sender can generate information for a receiver by running private experiments. The sender can also engage in uninformative manipulation. This paper argues that a preregistration requirement can discourage p-hacking, but also result in even more detrimental faked studies.