The data availability statement is the last thing most authors write and the first thing an editor checks. It's usually a paragraph, sometimes a sentence, and it answers one question: where are the data behind this paper, and how can a reader get them? Most journals now require one. Many funders expect the published version to match what you promised in your data management plan. And a surprising number of manuscripts get sent back not because the science is wrong but because the statement is vague, missing, or contradicts what's in the methods.
This guide covers what the statement needs to say, the handful of situations that cover almost every paper, wording you can adapt, and the mistakes that get flagged. It's written for researchers, but if you support researchers (in a library, a research office, or an IRB), it doubles as a review checklist.
What the statement is for
Journals ask for it so that readers can find, reuse, and check the data. Funders care because public money paid for the data and their policies (NIH's Data Management and Sharing Policy, NSF's data sharing requirements, and the federal public-access push behind them) expect the data to be findable at publication. Reviewers use it to judge whether the results could be verified. And future you will use it, because two years on you'll be asked for the data and won't remember which drive it lived on.
A good statement does three things: it says what data exist, where they are (with a persistent identifier if possible), and under what conditions someone can get them. Everything else is detail.
The five situations that cover almost every paper
Nearly every manuscript falls into one of these. Find yours, then adapt the wording.
1. The data are in a public repository
This is the best case and the one editors like most. Name the repository, give the DOI or accession number, and state the license if the repository asked you to pick one.
The data that support the findings of this study are openly available in Zenodo at https://doi.org/[DOI], under a CC BY 4.0 license. Analysis code is available at https://github.com/[user]/[repo] and archived at https://doi.org/[DOI].
If you haven't deposited yet, do it before submission. Most repositories issue a reserved DOI you can cite while the record stays private until the paper is accepted. The repository guide covers how to pick one.
2. The data are about people and can't be shared openly
Human-subjects data are the most common reason for a restriction, and the mistake here is stopping at "the data are not publicly available due to privacy." That's a reason, not a plan. Say what's restricted, why, and how a qualified researcher could get access.
The survey data contain information that could compromise participant privacy and are not publicly available. De-identified data are available from [the ICPSR restricted-use system / the corresponding author / the university's data steward at data@example.edu] to researchers who meet the criteria for access to confidential data and sign a data use agreement. The survey instrument, codebook, and analysis code are openly available at https://doi.org/[DOI].
Notice the second half. Even when the raw data must stay closed, the instrument, the codebook, and the code usually can be shared, and sharing them shows good faith. Check your consent form before you promise anything: if participants were told the data would never leave the research team, a restricted-access repository may still be off the table, and the statement should say so plainly.
3. The data come from a third party you don't control
Licensed datasets, administrative records from a state agency, proprietary data from a company, institutional records covered by FERPA. You can't post them, but you can tell the reader exactly how you got them.
Student record data were obtained from [institution's office of institutional research] under a data sharing agreement and cannot be redistributed by the authors. Researchers may request access from [office and contact]. The code used to clean and analyze the data is available at https://doi.org/[DOI].
4. The data are public and you only used them
If your paper reanalyzes IPEDS, the Census, a NASA archive, or someone else's deposited dataset, you still need a statement. The reader needs to reproduce your extract, not just find the source.
This study used publicly available data from the Integrated Postsecondary Education Data System (IPEDS), National Center for Education Statistics, provisional release for 2022-23, accessed [date] at https://nces.ed.gov/ipeds. The specific survey components, variables, and the code used to construct the analysis file are available at https://doi.org/[DOI].
Versions and access dates matter more than people expect. Public datasets get revised. Someone who pulls IPEDS next year may not get the numbers you got. The IPEDS guide has more on why.
5. No new data were created
Theoretical papers, reviews, and some methods papers produce no dataset. Say so, rather than leaving the section blank, because a blank section looks like an oversight.
No new data were created or analyzed in this study. Data sharing is not applicable to this article.
What "available upon request" really means
For years the default statement was "data are available from the corresponding author upon reasonable request." Many journals still accept it. A growing number discourage it, and some have stopped accepting it without a stated reason. The problem is that it rarely works: several studies that emailed authors of papers with that statement found most requests were ignored or refused, and the fraction that produced usable data was small.
Use it only when there's a real reason the data can't go in a repository, and if so, say the reason. "Available upon request" as a way to avoid the deposit step is the wording most likely to draw an editor's query. If you're worried about being scooped, most repositories offer an embargo. If you're worried about the effort, deposit takes an afternoon and the data organization guide gets the files ready.
Mistakes that get a manuscript sent back
- The statement contradicts the methods. The methods section says you collected interviews; the statement says no data were created. Or the DMP submitted with the grant promised public deposit and the statement says "upon request." Editors and program officers do compare.
- A link instead of an identifier. A lab website or a Google Drive link is not a persistent location. Use a DOI or an accession number from a repository that commits to keeping the record.
- A repository name with no record. "Data are deposited in Dryad" with no DOI means the reader can't find it. If the record is embargoed, cite the reserved DOI and say when it opens.
- Silence about the code. Many journals now ask about code as well as data. Even if they don't, the analysis scripts are often what a reader needs most. The reproducibility guide covers what to include.
- Restriction without a route. "Not available due to ethical restrictions" with no contact, no process, and no mention of what can be shared reads as a closed door. Give the route, even if it's narrow.
- Placeholder text left in. [DOI] and [repository] survive into published papers more often than you'd think. Search the final PDF for square brackets.
Before you submit
Read the journal's author guidelines for the exact heading (some say "Data Availability," others "Data Accessibility" or "Availability of data and materials") and where it goes, usually after the conclusion with the funding and conflict-of-interest declarations. Check whether the journal has a data policy tier that requires deposit rather than a statement. Confirm the DOI resolves. Make sure every co-author agrees with what's promised, especially the one who holds the data. And if the work was grant-funded, reread your data management plan and make the statement match it, or explain why it changed.
If you'd rather not do this alone
Reviewing data availability statements, matching them to funder plans, and getting datasets ready for deposit is a large part of research data services work, and it's something I help individual researchers and small research offices with. If your institution wants a standing process for this rather than a scramble at each submission, the guide to starting research data services describes how that looks, or get in touch.
Common questions
Is "data available upon request" still acceptable? Often, but less and less. Follow-up studies found most such requests go unanswered, so journals are moving away from it. If there's no legal or ethical reason to withhold the data, deposit them and cite the DOI.
What do I write if my data are about people and can't be shared? Say what's restricted, why (consent terms, IRB conditions, a data use agreement), and how a qualified researcher could get access. Share the instrument, codebook, and code if you can.
Do I need a statement if I only used public data? Yes. Name each source, cite it (the data citation guide shows how), give the version or access date and the variables you used, and link to the code that builds your analysis file.
Where does it go in the manuscript? Usually a titled section after the conclusion, next to funding and conflict-of-interest declarations. Many journals also ask for the same text in a submission form. Check the author guidelines.