Abstract
Political science has witnessed an explosion in the use of experiments. In response, authors, reviewers, and journal editors have sought to devise best practices for experimental work by promoting (i) a bias against the use of convenience samples, (ii) the inclusion of multiple experiments per study, and (iii) the requirement of pre-registration. However well-intended, these efforts only increase the cost of experimental research. Without any clear scientific basis, journals are requiring pre-registration exclusively for experimental research; researchers aggrandize the use of nationally representative and other non-convenience samples, even when there is no inferential target; and editors and reviewers often expect multiple experiments per study, when only one is necessary for causal inference. We argue these norms are theoretically bankrupt and exacerbate institutional inequalities and elitism in the discipline. We highlight the problems of these norms and provide a set of recommendations to head off these bad practices.