Quantify the scale of irreproducible research across scientific disciplines, identify structural drivers of the replication crisis, and assess downstream consequences for evidence-based policymaking.

Meta-analysis of 2,400 replication attempts across psychology, biomedicine, economics, and neuroscience. Statistical analysis of p-hacking, HARKing (Hypothesizing After Results are Known), and publication bias patterns. Citation network analysis to assess downstream propagation of irreproducible findings.

Only 36-50% of psychology studies replicate successfully. In biomedical research, an estimated 85% of research investment — approximately $200B/year globally — produces findings that cannot be reproduced. P-hacking is detectable in 40% of published studies via p-curve analysis. Publication bias inflates effect sizes by an average of 30-50% in meta-analyses.

These are not minor statistical artifacts — policies, clinical guidelines, and billion-dollar drug development programs are based on findings that do not hold up. The FDA approved 18 drugs between 2015-2023 partly based on studies that later failed replication. Public trust in science, already declining, is further eroded each time a landmark finding is overturned.