Misreporting Statistical Non-significance Levels to Support Hypothesized Relationships

Abstract

Recent research has documented a range of problematic behaviors by academic researchers including p-hacking, hypothesizing after results are known, and even fraud. This study tests whether the “publish or perish” aphorism might explain the occurrence of one such action: misreporting statistical significance levels to support hypothesized relationships. Drawing from an integrative theoretical model, we hypothesize that articles with senior academic lead authors are more likely to have misreported statistical significance findings to support hypothesized relationships than those led by junior authors. Findings from a study of 55 high impact articles appearing in top empirical management journals confirm these expectations. Overall, nearly 20 percent of the articles report support for hypotheses which are actually non-statistically significant, this occurrence is related to senior ranked lead authors, and the relationship is robust relative to the research rankings of the lead author’s institution, their PhD program, and the number of coauthors on the article.