EmpiricalRequires evidence

Pooled across the three large replication projects of top psychology, economics and Nature/Science experiments, papers that failed to replicate were cited more than papers that replicated, by about 16 more citations a year (the authors write "cited almost 16 times more per year"; we read it as an absolute difference), and the gap did not close after the failure was published: only 12% of later citations mentioned it. Experts in prediction markets had foreseen which would fail, so the citations went to findings reviewers had reason to doubt.

Institutional design