Acciai 2023 — 86% non-sharing, 1,634 PNAS and Nature-portfolio papers from 2017-2021

Acciai, Schneider, and Nielsen ran an audit experiment in 2022 on 1,634 papers published 2017-2021 in PNAS and Nature-portfolio journals. The papers' data-availability statements promised data on request. 86% of the requests went unfulfilled.

The corpus is high-impact, recent, and explicitly committed in writing. The non-sharing rate sits squarely in the C-0002 range and demonstrates that the pattern persisted into the 2020s in the journals research universities most often cite as exemplary. The replication of Vines 2014 (E-0001) and Gabelica 2022 (E-0002) seventeen years later, in a different field, with the same direct-contact methodology, is what establishes the C-0002 baseline as steady-state rather than artifactual.

The four cohorts (E-0001, E-0002, E-0003, E-0004) span two decades, four fields, and four funder regimes. The convergence on 73-93% non-delivery is the strongest available evidence that the baseline is structural rather than cohort-specific.