Research Data Sharing Statistics 2026: What Actually Happens When Someone Requests Your Data

Most published papers now carry a data availability statement. Many promise data “available on reasonable request.” But when researchers actually send those requests, the results tell a very different story. Across multiple independent audits conducted between 2006 and 2024, the gap between what data availability statements claim and what researchers actually deliver is one of the most consistent findings in open-science research. This article brings together the key research data sharing statistics — covering request success rates, compliance audits, data rot, and field-by-field differences — so you know what the evidence actually says in 2026.

The Promise vs. Practice Gap

Data availability statements (DAS) have become standard. According to a 2024 audit of Springer Nature publications, 64% of articles published in 2023 indicated that data was freely accessible via some direct route — up from 50% the year before. But “indicated freely available” and “actually available” are not the same thing. A cross-sectional study examining 3,191 papers in the BMJ, JAMA, NEJM, and The Lancet found that only 38.1% of data availability statements described genuinely open data. When journals mandated data sharing, that share rose to 61.4%; when they merely encouraged it, the rate fell to just 22.3%.

In other words, the policy environment matters more than the statement itself. A DAS is not proof of data access — it is a claim, and claims require verification. Understanding what the research data sharing statistics actually show is essential for any researcher designing a methodology, writing up their data management plan, or attempting to build on prior work.

For students working on dissertations, writing a compliant data availability statement is increasingly a formal requirement — but the compliance landscape it sits within is messier than most guides acknowledge.

Data Request Success Rates: What Audits Found

Several independent research teams have systematically requested data from authors whose papers claimed it was available, then measured response and delivery rates. The results are remarkably consistent.

Psychology: A Decades-Long Record

The landmark study in this area comes from Wicherts et al. (2006), who contacted authors of 249 articles published in APA journals and found that only 25.7% of teams shared their data upon request. A decade later, Vanpaemel and colleagues requested data from 394 papers published across four APA journals in 2012; the rate had risen, but only to 38% — meaning nearly two-thirds of requests still went unfulfilled.

A 2023 replication published in PLOS ONE confirmed that this pattern persists even after years of open-science advocacy. The study, a direct replication of Wicherts, Bakker and Molenaar (2011), found that data sharing upon request remained far below the rates suggested by stated intentions.

Biomedical Research: The Reasonable Request Fiction

A 2022 audit published in ScienceDirect focused specifically on authors who had stated that datasets were “available on reasonable request” — the most common DAS phrasing (used in 42% of papers sampled). The findings were stark:

• Only 14% of contacted authors responded at all
• Only 6.8% actually shared their data
• The gap between stated intent and delivery was therefore over 93 percentage points

A parallel analysis of Max Planck Society publications found that data could be obtained from approximately 20% of publications claiming availability on request — better, but still a long way from what the statements imply.

The Miyakawa Natural Experiment

Perhaps the most striking evidence comes from a 2020 editorial in Molecular Brain. Editor-in-chief Tsuyoshi Miyakawa requested raw data for 41 submitted manuscripts as a condition of peer review. The result: 97% of authors did not provide it. Twenty-one manuscripts were withdrawn without providing data. Of the 14 papers that subsequently appeared in other journals — 12 in journals with data-on-request policies — Miyakawa requested the data again. He received no response in 10 cases, a refusal in 1 case, and an incomplete dataset in 1 case.

Data Rot: How Availability Declines Over Time

Even when data exists at the time of publication, it may not survive. Vines et al. (2014), published in Current Biology, requested datasets from 516 ecology articles published between 2 and 22 years prior. They found that the odds of data being extant decreased by 17% for every year since publication. Email addresses expired, hard drives died, and researchers moved institutions — all eroding the practical availability of data that may once have been accessible.

A 2022 follow-up published in PMC on data rot at a single US university confirmed this decay pattern holds in modern settings. The longer a paper sits, the less likely any data associated with it can actually be retrieved.

This has direct implications for replication science. If you are designing a study that builds on prior empirical work and intend to request the original data — plan for the possibility that it no longer exists, regardless of what the data availability statement says.

Data Sharing by Academic Field (2024 Data)

Research data sharing statistics vary substantially by discipline. The State of Open Data 2024 report, produced by Digital Science, Figshare, and Springer Nature, combined Dimensions citation data, Springer Nature DAS analysis, and the Make Data Count corpus to produce the most comprehensive field-level comparison available.

The spread between raw sharing rates and FAIR-criteria compliance is particularly revealing. Physics shares data at a high rate, but most of it is not deposited in a way that meets Findable, Accessible, Interoperable, and Reusable standards — meaning it may be shareable in practice today but fragile over time. Environmental science leads on both dimensions, likely reflecting the long data-archiving culture in ecology and geoscience journals.

For dissertation students designing empirical studies, understanding the FAIR data principles is increasingly important — not just for journal submission, but because funders such as Horizon Europe and UKRI now assess data management plans against FAIR criteria as part of grant review.

Across all fields, the State of Open Data 2024 report noted that public data sharing broadly increased from 41% to 47% among PLOS-comparable journals between 2019 and 2022, and repository use among researchers rose from 23% to 28% in the same period. Progress is real but incremental.

DAS Compliance Audits in Medical Journals

Medical and clinical research carry their own compliance audit trail. A meta-research study examining 3,191 papers across four leading medical journals — the BMJ, JAMA, NEJM, and The Lancet — from the introduction of DAS policies through December 2022, found significant variation in what data availability statements actually promised.

In papers with applicable DAS:

• 38.1% described openly available data overall
• 61.4% described open data when the journal mandated sharing
• 22.3% described open data when the journal only encouraged sharing

The implication is clear: encouragement alone produces roughly the same rate as having no policy. Mandates — with enforcement — produce significantly more genuine openness. The challenge is enforcement: journals that have begun post-publication checks on DAS compliance frequently encounter authors who have moved institutions or whose datasets are no longer accessible.

A separate JMIR study from 2025 examined data-sharing statements specifically in public, environmental, and occupational health journals, confirming that even in high-stakes YMYL research, actual compliance with stated data-sharing intentions remains a persistent gap. The full study is available at JMIR.

Do Mandates Actually Work?

The research data sharing statistics consistently point in one direction: mandates with teeth work better than voluntary norms. A classic demonstration comes from ecology. A 2013 arxiv preprint (later published) examined mandated data archiving in ecology journals and found that archive-mandating journals produced substantially higher rates of actual data deposit than those relying on author discretion.

The NIH’s updated Data Management and Sharing (DMS) policy, effective January 25, 2023, requires all NIH-funded researchers to submit a DMS Plan with their grant application and to report on compliance in annual Research Performance Progress Reports from October 2024 onward. Early reports from implementing institutions suggest that the policy has substantially increased the proportion of funded projects with formal data management plans — though aggregate compliance statistics for 2024–2025 are still being consolidated at institutional level.

For researchers designing empirical studies, these trends have concrete implications for the methodology chapter. Funders increasingly review data management plans as part of the grant process, and examiners in some fields are beginning to expect explicit data availability commitments in thesis methodology sections. If you intend to use secondary datasets, it is also worth reviewing whether your reuse is exempt from full ethics review — the rules differ by data type and jurisdiction, as covered in detail in the guide to IRB approval for secondary data.

Students working with datasets — whether original or secondary — should also understand how to preregister their study on OSF or AsPredicted, since preregistration and data sharing together provide the strongest signal of reproducible research practice.

Tools like Tesify can help students draft compliant data management sections and data availability statements that accurately reflect their actual plans — reducing the gap between stated policy and research practice from the very start of a project.

Frequently Asked Questions