Null Results Are 40 Percentage Points Less Likely to Be Published, Creating a Systematically Distorted Scientific Literature

Studies with statistically significant results are three times more likely to be published than those with null results, and strong results are 60 percentage points more likely to even be written up by their authors. In social science, only 10 out of 48 null results were published compared to 56 out of 91 strongly significant results. Why it matters: the published literature systematically overstates effect sizes and treatment efficacy, so meta-analyses that guide policy produce inflated estimates, so governments and organizations implement interventions that appear effective in the literature but fail in practice, so billions in public funds are wasted on programs that looked good on paper but were built on a biased evidence base, so the public experiences a pattern of promised breakthroughs that never materialize, eroding trust in science and expert recommendations. The structural root cause is that journal editors and reviewers use statistical significance as a proxy for 'interestingness,' while researchers' careers depend on publishing in high-impact journals that prefer novel positive findings, creating a self-reinforcing cycle where null results are neither submitted, reviewed, nor published despite being equally scientifically valuable.