Methods and limitations · OpenAlex data as of November 17, 2026
How the numbers are made
Every figure on this site comes from OpenAlex, a free and open index of research. Other sources are used only to check it. This page explains the rules, and where the numbers are likely to be wrong.
What is counted
Source. OpenAlex. The data was downloaded on November 17, 2026, and every figure on the site reflects OpenAlex as of that date. OpenAlex is run by the nonprofit OurResearch and released into the public domain (CC0). It is the only source for the numbers on this site.
Institutions. 17 Connecticut colleges and universities: Central Connecticut State University, Connecticut College, Eastern Connecticut State University, Fairfield University, Quinnipiac University, Sacred Heart University, Southern Connecticut State University, Trinity College, United States Coast Guard Academy, University of Bridgeport, University of Connecticut (incl. UConn Health), University of Hartford, University of New Haven, University of Saint Joseph, Wesleyan University, Western Connecticut State University, Yale University. UConn Health (the medical school and hospital in Farmington) is listed separately in OpenAlex and counted under UConn here, each article once. Community colleges and other organizations are out of scope.
An article counts for an institution if any author lists it as an affiliation, including its departments and units (OpenAlex’s institution “lineage”). An article with authors at two Connecticut institutions counts for both. Statewide totals count each article once. Without Yale means articles with at least one author at one of the other 16, each once. Yale is the only Connecticut institution on 64% of the articles, so it shapes the statewide figure.
Work types. Journal articles and reviews that are not retracted. Books and book chapters are counted separately on each profile and are not in any other figure. Editorials, letters, preprints, datasets and conference papers are not counted.
Publishers and subjects. Publishers are grouped under their parent company, as OpenAlex records them. The site shows the familiar name where OpenAlex uses the owner’s: Taylor & Francis appears in OpenAlex as Informa, and Cambridge and Oxford University Presses as their universities. The data files keep OpenAlex’s names. Fields and topics are OpenAlex’s subject labels, one main topic per article.
Years. 2015 to 2025, the last complete year. 2026 is shown separately and labeled incomplete, because new articles take months to be indexed and some publishers open articles a year or more after publication.
What is left out
Three kinds of records are excluded from every figure. They stay in the data files, flagged, so they can be checked.
| Reason | Works |
|---|---|
| Repository-only records with no DOI (mostly student work, course material and catalog stubs that OpenAlex types as articles) | 1,342 |
| Meeting abstracts printed in journal supplements (Neurology, The FASEB Journal, Journal of Clinical Oncology, American Heart Association journals, Blood and a few others), found by their DOI pattern | 1,255 |
| Known affiliation errors: authors at Washington College in Maryland matched to Trinity College (founded as Washington College), where no Connecticut author remained | 14 |
The list of excluded works is on the Downloads page. Meeting abstracts without one of the known DOI patterns are still counted, so a few remain.
How open access is classified
Each article gets one category, from OpenAlex’s open access status. OpenAlex looks at where free copies exist and under what license, and largely decides by the journal: whether it is fully open access, and whether it is listed in the Directory of Open Access Journals (DOAJ).
- Diamond
- Free to read in a journal that charges authors nothing (fee confirmed as $0). Statewide, 0.9% of articles.
- Open journal, fee not confirmed
- Free to read in an open access journal, but the data does not confirm whether authors pay a fee. Statewide, 4.8% of articles.
- Gold
- Free to read in a fully open access journal that usually charges authors a fee. Statewide, 19.9% of articles.
- Hybrid
- Free to read in a subscription journal because the author paid a fee for this article. Statewide, 9.8% of articles.
- Bronze
- Free to read on the publisher’s site without an open license. The publisher could take it back. Statewide, 14.4% of articles.
- Green
- The journal version is paywalled, but a free copy is in a repository (an online archive). Statewide, 17.6% of articles.
- Closed
- Behind a paywall, and no free copy could be found. Statewide, 32.7% of articles.
Free to read means any category except closed. Bronze counts as free to read, but bronze articles have no open license and the publisher can withdraw them.
Diamond is split in two. OpenAlex calls an article “diamond” when the journal is open access and it has no fee on record. That is not the same as a confirmed fee of zero. This site reports “diamond” only where the listed fee is $0, and “open journal, fee not confirmed” for the rest (8,329 articles statewide). Both are free to read. Neither is counted in the fee estimates.
Article fee estimates Estimate
Many journals charge an article processing charge (APC) to publish open access. OpenAlex records each journal’slisted price, not what anyone paid. The estimates add up listed prices, in US dollars, for gold and hybrid articles, 2015 to 2025.
- Corresponding author estimate (the more realistic figure): articles where a corresponding author (the author who usually arranges payment) is at the institution. Only 82.5% of articles statewide have a corresponding author’s institution recorded, so this can understate. When authors at two Connecticut institutions are both corresponding, each institution’s own figure includes the full fee, and the statewide figure counts it once.
- Upper bound: articles with any author at the institution, even when the fee was likely paid elsewhere.
- Articles with no listed price are left out, as are articles in the “fee not confirmed” category. Statewide, 2,595 articles in the corresponding author estimate have no listed price.
Listed prices compared with fees actually paid
No Connecticut institution reports its payments to OpenAPC, the public database of fees universities have paid. But other universities, mostly in Europe, have reported paying for 393 articles with Connecticut co-authors. For those same articles, the fees paid came to 77% of today’s listed price for gold articles and 71% for hybrid (compared in euros, including VAT). For gold articles published 2015–2019 it was 68%, and for 2020–2025 87%, which fits listed prices rising over time. Using the typical fee paid per journal instead gives a similar result (73% for gold, covering 37% of Connecticut gold articles).
So the listed-price estimates on this site are likely too high, especially for earlier years. Payments can also be lower through discounts, waivers and library agreements, which OpenAPC does not fully capture. This comparison was run on the OpenAlex data of November 15, 2026.
A special case: 55 articles that charge authors nothing (mostly in physics journals funded by the SCOAP3 library consortium) appear in OpenAPC because a library reports its share of the consortium cost per article. They stay in the no-fee categories here.
Repository candidates (green proxy) Proxy
Whether an author may put a paywalled article in a repository depends on the publisher’s policy. The main policy database, Open Policy Finder (formerly Sherpa Romeo), is licensed CC BY-NC-ND, which conflicts with releasing this site’s data into the public domain. So the site uses a proxy built only from OpenAlex.
- A journal shows deposit evidence if at least 3 of its articles worldwide (2015–2025) are free only as an accepted manuscript in a repository. Deposits in PubMed Central and Europe PMC are not counted, because most follow funder rules (such as NIH’s) that publishers accept only for funded articles.
- A Connecticut article is a candidate if it is closed or bronze, has no free copy anywhere except the publisher’s page (no free preprint either), and is in a journal that shows deposit evidence.
Statewide that gives 43,141 candidates (24.6% of articles): 37,594 closed and 5,547 bronze. The threshold matters: with 1 deposit as evidence the count is 55,803, and with 10 it is 25,656. This is a starting point for outreach. Embargoes and each publisher’s current policy still need checking.
Checks against other sources
A sample of 510 articles (30 per institution, spread over 2015–2025) was looked up in other sources. Because every institution has the same number, these are sample rates, not statewide estimates. None of these sources changes a figure on the site. The checks were run on the OpenAlex data of November 15, 2026 and are rerun by hand from time to time, so their counts can differ slightly from the figures elsewhere on this page, which use data as of November 17, 2026.
- Unpaywall agrees with OpenAlex on 501 of 501 articles it found. This is expected: Unpaywall is made by the same nonprofit and now runs on the OpenAlex database, so it is not an independent check.
- OpenAIRE, a European research graph built separately, is the main independent check. Where it gives an answer, it agrees on whether an article is free to read for 358 of 407 articles (88.0%). It has no access information for another 97. Every disagreement was examined by hand. Where OpenAlex says free and OpenAIRE says closed (26 articles), OpenAlex looked right in most cases, usually because OpenAIRE had not linked a repository or publisher copy. Where OpenAlex says closed and OpenAIRE says open (23), the picture was mixed: some came from old or misread data at OpenAIRE, and some revealed free copies in Connecticut repositories that OpenAlex misses (see Limitations).
- DOAJ: OpenAlex’s “in DOAJ” flag matches DOAJ for 99.9% of 3,329 journals checked. For the 8,344 “fee not confirmed” articles (2015–2025), DOAJ lists the journal as charging no fee for 2,706, charging a fee for 858, and no article fee but other charges for 83. 4,594 are in journals DOAJ does not list, and 103 were published before the journal became open access. This is reported here only. The category figures stay as OpenAlex has them.
- CORE, a repository aggregator, has not been used yet while its license terms for this kind of site are confirmed.
Full cross-check report (on GitHub)
Limitations
- OpenAlex misses some free copies in Connecticut’s own repositories. The cross-check of November 15, 2026 found records in Digital Commons repositories and in Yale’s EliScholar that offer a public download but are marked not free in OpenAlex. Up to 553 articles called closed have such a record (an upper bound, because some records hold no file). The Journal of Financial Crises, published by Yale on EliScholar, appears free to read but has 143 articles labeled closed. Free-to-read shares are therefore slightly too low for some institutions, notably Fairfield and the University of New Haven. Source: the cross-check report.
- Affiliation errors. OpenAlex matches authors to institutions automatically. Spot checks of the raw affiliation text confirmed 97 to 100 percent of each institution’s articles. One systematic error (Washington College in Maryland matched to Trinity) is corrected. Others may remain.
- Classification depends on the journal. Gold, diamond and hybrid depend on journal-level data (DOAJ listing, listed fees). A journal that changed its model, or added a fee, may be classified by its current state.
- Recent years are still moving. The share of articles free to read peaked in 2023 and is lower for 2024 and 2025, mostly because of fewer bronze articles. Some publishers open articles a year or more after publication, so this is likely partly a timing effect. Later snapshots will show.
- OpenAlex changes every day. Records are added, merged, corrected and reclassified continually, so counting again on another day gives slightly different numbers, including for older years. Every figure on this site reflects OpenAlex as of the download date shown at the top of each page, and the monthly refresh brings them up to date.
- Fees are listed prices, not payments. See the comparison with OpenAPC.
- Books and the humanities are covered less well. OpenAlex is strongest for journal articles in the sciences and medicine. Humanities scholarship, which appears more in books, is undercounted, so institutions with strong humanities programs are described less completely.
- Sources disagree. OpenAIRE agrees with OpenAlex on free-to-read status for about nine in ten sampled articles. Neither is complete.
Reproducing the figures
The pipeline is open source on GitHub. A Python script downloads the data from OpenAlex, saves every response, and logs every query, so each number on this site traces back to a saved query. One command rebuilds every table from the saved responses. Tests check the metric definitions on small examples with known answers.
A separate program, written without seeing the pipeline code, recomputes the published figures from the raw data, and a sample of counts was checked by hand against the OpenAlex website. The data is refreshed monthly. Each refresh downloads everything again, reruns the tests and the separate recomputation, and compares the results with the previous month. It publishes only if every check passes, and each refresh is listed on the Changelog.
All the tables behind the charts can be downloaded on the Downloads page.
Sources, licenses and how to cite
- OpenAlex (CC0). Priem, J., Piwowar, H., and Orr, R. (2022). OpenAlex: A fully-open index of scholarly works, authors, venues, institutions, and concepts. arXiv:2205.01833.
- OpenAIRE Graph (CC BY 4.0), used for the cross-check. OpenAIRE Research Graph, doi:10.5281/zenodo.3516917.
- DOAJ journal metadata (CC0), used for the cross-check.
- OpenAPC: contains information from OpenAPC, made available under the Open Database License (ODbL) 1.0. Used for the fee comparison.
- Unpaywall, used for the cross-check.
This site’s own data is released into the public domain (CC0), except the cross-check files built from OpenAIRE and OpenAPC, which keep those sources’ licenses. The code is MIT licensed.
How to cite this site (APA 7th edition)
Misciagna, C. (2026). Connecticut Open Access Observatory [Website and data set]. https://ct-oa-observatory.vercel.app/
To cite one page, name the page and give its address, for example:
Misciagna, C. (2026). Methods and limitations. Connecticut Open Access Observatory. https://ct-oa-observatory.vercel.app/methods/
The figures change when the data is refreshed. If you quote a specific number, also say that it uses OpenAlex data as of November 17, 2026, so readers can match it to the right version on the Changelog.