93% of Researchers Who Promise to Share Data in Their Publications Never Actually Do When Asked

Among 1,792 manuscripts where authors stated they were willing to share their research data upon request, 1,669 (93%) either did not respond or declined when actually asked. Even journals with mandatory data sharing policies see compliance rates no better than journals without such mandates. Why it matters: other researchers cannot verify or build upon published findings, so entire fields accumulate unreplicated claims that may be false, so funding agencies waste billions on research programs built on unverifiable foundations, so clinical and policy decisions get made based on data no one else has examined, so public trust in science erodes as high-profile failures (like the Alzheimer's amyloid-beta fabrication scandal) reveal that no one checked the underlying data. The structural root cause is that data sharing imposes real costs on researchers (averaging $29,800 per compliance effort) with no career reward, while journals and funders impose mandates they never enforce, creating a system where the statement 'data available upon request' functions as a performative ritual rather than a genuine commitment.