How the Journal Impact Factor Is Calculated: Citable Items, Front Matter, and Self-Citation

Short answer: the Journal Impact Factor (JIF) is citations in year Y to a journal’s citable items from years Y-1 and Y-2, divided by the count of those citable items. Almost every explainer stops there. The mechanics that actually decide whether a journal’s number moves are what counts as a “citable item” in Clarivate’s own document-type coding, why the numerator and denominator draw from different populations of content, and what Clarivate does — and does not do — about self-citation. This page works through each of those, with Clarivate’s own definitions and the peer-reviewed and editorial literature that has examined them.

The exact formula, and the two populations it draws from

The Journal Impact Factor for a given year’s Journal Citation Reports (JCR) release is:

Citations recorded in year Y to a journal’s content published in years Y-1 and Y-2, divided by the number of “citable items” that journal published in those same two years.

The 2026 JCR release, for example, uses 2025 citation data against content published in 2023-2024. The two halves of that fraction are not drawn from the same set of content, and that asymmetry is the whole story:

• The numerator counts citations to anything. If an editorial, a letter, a news item, or a correction published in the window gets cited, that citation counts toward the journal’s total — Clarivate does not filter the numerator by document type.
• The denominator counts only “citable items.” Per Clarivate’s own account of the calculation, “review” articles carrying more than 100 references are coded distinctly, and letters and other front-matter content are excluded from the citable-item count entirely (source: Clarivate, “The Clarivate Impact Factor” essay, clarivate.com/academia-government/essays/impact-factor/).

A journal can therefore raise its Impact Factor without publishing a single additional citable article, purely by having more of its non-citable content get cited — or by having less of its citable-eligible content reclassified out of the denominator. Whether that happens in practice, and how much it actually moves the number, is the part worth being precise about.

What Clarivate actually counts as a “citable item”

Clarivate’s Web of Science indexing assigns every piece of content a document type, and only three types are treated as citable for JCR purposes: Article, Review, and Proceedings Paper. Everything else — editorial material, letters, news items, meeting abstracts, corrections, front matter — is indexed but excluded from the denominator.

The most detailed public account of how that classification is actually made comes from two Thomson Reuters (Clarivate’s predecessor) staff scientometricians, writing in JAMA: Marie E. McVeigh and Stephen J. Mann, “The Journal Impact Factor Denominator: Defining Citable (Counted) Items” (JAMA, 2009;302(10):1107-1109). They describe classification as a judgment call made at the level of a journal section, not an individual article, based on seven characteristics of the section’s typical content:

• A descriptive, substantive title (not a generic label like “Editorial”)
• Named author(s) with institutional affiliation
• An abstract or structured summary
• Length exceeding roughly half a printed page
• Original data presented in figures or tables
• A list of cited references
• A citation density (references per page) consistent with research literature rather than commentary

McVeigh and Mann also reported the actual scale of the citable/non-citable split across the indexed literature: of roughly 1.1 million items indexed from publication year 2000, 793,395 (72.8%) were classified as articles or reviews — and that 72.8% accounted for 97.6% of all citations those items accrued over the following eight years. The 2005 cohort showed almost the same split (about 71.9% citable). In other words, for the literature as a whole, the citable/non-citable line tracks the citing/non-cited line closely — most of the divergence that matters is concentrated in a minority of edge cases, not a systemic mismatch.

The front-matter loophole: how it works, and what’s actually been shown about it

The mechanism itself is simple and undisputed: reclassifying a section from “Article” to “Editorial Material” removes its items from the denominator but does nothing to the numerator — any citations that content already earned, or goes on to earn, still count. Because Clarivate’s classification is made at the section level and, per McVeigh and Mann, generally stays fixed in Clarivate’s authority files unless a publisher requests reconsideration, a journal that successfully argues a research-adjacent section (a “Perspective,” a “Commentary,” a “Hindsight” retrospective) belongs outside the citable count gets a denominator reduction that persists year over year.

The clearest public illustration of the sensitivity is a set of counterfactual recalculations by Phil Davis in the Scholarly Kitchen, “Citable Items: The Contested Impact Factor Denominator” (February 10, 2016) — these are Davis’s own re-runs of the arithmetic under an alternate classification, not documented cases of Clarivate actually changing a published figure:

• Reclassifying the Journal of Clinical Investigation‘s “Hindsight” section as citable moved its Impact Factor from 13.262 to 13.583.
• Reclassifying Science Translational Medicine‘s “Commentary” and “Perspective” sections as citable moved its figure from 15.843 to 18.598.
• Reclassifying PLOS Medicine‘s “Policy Forum,” “Essay,” and “Health in Action” sections as citable — Clarivate’s default treats them as Editorial Material — dropped its figure from 14.429 to 8.447, the largest of the three because those sections make up a bigger share of the journal’s output.

Those numbers show the mechanism can move a journal’s figure by a meaningful margin when a large content share sits near the classification boundary. But set against McVeigh and Mann’s finding — using real data from BMJ, JAMA, The Lancet, and NEJM — that citable/non-citable reclassification did not materially move those journals’ actual Impact Factors, the honest summary is: the mechanism is real and mathematically clean, but the magnitude of its effect in practice depends entirely on how much of a specific journal’s content sits in the classification’s gray zone. No independently verified case of Clarivate reclassifying a section and a journal’s published Impact Factor changing as a direct result was found for this page — if you’re evaluating a specific journal’s number, treat “look at how much of its output sits in ambiguous front-matter sections” as the actual due-diligence question, not a general assumption of gaming.

The 2-year vs. the 5-year Impact Factor

Every JCR release reports two versions of the same calculation, differing only in the citation window:

• The standard (2-year) JIF — citations in year Y to citable items from Y-1 and Y-2, as above. This is the figure almost everyone means by “impact factor.”
• The JIF 5-Year — the identical calculation, extended to citations in year Y against citable items from Y-1 through Y-5.

The two-year window was chosen for currency, not statistical robustness — Eugene Garfield designed it to surface journals whose content was being cited quickly. That makes it volatile for fields with slow citation uptake (much of mathematics, the humanities, some engineering disciplines), where a paper’s real citation trajectory hasn’t developed yet at the two-year mark. The 5-year window smooths that volatility by giving slower-citing content three additional years to accrue citations before it drops out of the count, at the cost of being less current. Clarivate itself frames the 5-year figure as an alternative some users will find more informative, not a replacement for the standard figure (source: Clarivate, “The Clarivate Impact Factor” essay).

Self-citation: what Clarivate actually does about it

This is the point most often stated wrong: Clarivate does not strip self-citations out of the standard, published Journal Impact Factor. A journal’s own self-citations to its own recent content count in the numerator exactly like any other citation. What Clarivate does instead is report the raw citing-journal data separately, alongside the standard figure, and monitor for abuse:

• The data is available, not the correction. Per Clarivate’s own explanation, a journal’s self-citations run at roughly 13% of its total received citations on average, and JCR’s Citing Journal Listing gives users the components to calculate an analogous, self-citation-excluded Impact Factor themselves — Clarivate does not publish that adjusted figure as the headline JIF.
• Self-citation rate is reported as a separate figure alongside JIF in every journal’s JCR profile, precisely so a reader can see how much of a given figure is self-referential without Clarivate having to adjudicate a “correct” adjusted number.
• Suppression is the actual enforcement mechanism, not subtraction. Clarivate has periodically suppressed journals from JCR entirely — omitting them from that year’s rankings — when self-citation, or citation stacking (reciprocal over-citation with a small group of partner journals), accounts for a disproportionate share of the citations behind a journal’s Impact Factor. Clarivate does not publish the exact threshold that triggers a review, by design, to make the criterion harder to reverse-engineer and game. Clarivate’s own analysis of the 2014 JCR Science Edition found 85% of indexed journals sat at or below a 15% self-citation rate — useful context for judging whether a specific figure looks unusual.

Net effect: a high self-citation rate on its own is not evidence of manipulation and does not get quietly netted out of the number you see — it is disclosed, and it is the trigger for a manual suppression review only when it (or reciprocal stacking) becomes extreme.

What this means when you’re reading a specific journal’s Impact Factor

Put together, the mechanics above translate into three concrete checks, beyond the headline number:

• Check the journal’s self-citation rate in its JCR profile alongside the JIF — both are reported together for exactly this reason.
• If a journal’s output leans heavily on named sections (perspectives, commentaries, hindsight pieces, policy essays) rather than standard research articles, its published figure may be more sensitive to Clarivate’s document-type classification than a journal publishing mostly conventional articles and reviews.
• A 2-year figure and a 5-year figure that diverge substantially is itself informative — it usually means the field’s citation uptake is slow relative to the two-year window, not that anything is wrong with the number.

None of this substitutes for the field’s own responsible-metrics guidance: DORA and the Leiden Manifesto both recommend against using a journal-level average — however precisely it’s calculated — as a stand-in for the quality of an individual article or author. See Impact Factor vs. h-index for why a journal-level metric and an author-level metric answer different questions entirely, and Essential Science Indicators for how Clarivate’s separate ESI product uses citation thresholds for a different purpose (identifying highly cited papers and fields, not ranking journals). For the practical task of pulling a specific journal’s current number, see How to Check a Journal’s Impact Factor.

Frequently asked questions

Do editorials and letters count toward a journal’s Impact Factor?

Citations to them count in the numerator if anyone cites them. They do not count in the denominator — Clarivate’s document-type coding excludes editorial material, letters, news items, and meeting abstracts from the citable-item count, even though citations to that content still add to the numerator.

Can a journal legitimately dispute how its content is classified?

Yes — per McVeigh and Mann’s account, classification is made at the section level and a publisher can request Clarivate reconsider it; absent that request, the existing classification persists in Clarivate’s authority files year over year.

Does a high self-citation rate mean a journal is manipulating its Impact Factor?

Not on its own. Clarivate discloses each journal’s self-citation rate alongside its JIF rather than adjusting the headline figure for it. It becomes a manipulation question only when self-citation or reciprocal citation stacking reaches a level extreme enough to trigger Clarivate’s (undisclosed-threshold) JCR suppression review.

Is the 5-year Impact Factor “more accurate” than the standard 2-year figure?

Not inherently — it answers a different question. The 5-year figure trades currency for stability, which matters more for slow-citing fields; the 2-year figure stays closer to what’s being cited right now, which is what it was originally designed to surface.

Sources

• Clarivate, “The Clarivate Impact Factor” (clarivate.com/academia-government/essays/impact-factor/) — Clarivate’s own explanation of the calculation, citable-item treatment, the 5-year alternative, and self-citation data availability.
• McVeigh ME, Mann SJ. “The Journal Impact Factor Denominator: Defining Citable (Counted) Items.” JAMA. 2009;302(10):1107-1109. (jamanetwork.com/journals/jama/fullarticle/184527)
• Davis P. “Citable Items: The Contested Impact Factor Denominator.” The Scholarly Kitchen, February 10, 2016. (scholarlykitchen.sspnet.org/2016/02/10/citable-items-the-contested-impact-factor-denominator/)
• Clarivate, “A Refresher Course on JCR Journal Suppression Policies” (Clarivate blog) — self-citation and citation-stacking suppression policy; exact thresholds deliberately unpublished.

jcr.clarivate.com itself (the live subscription platform) could not be fetched directly in this drafting session; the calculation mechanics above are sourced to Clarivate’s own publicly accessible essay and to the peer-reviewed/editorial literature examining the methodology, not to a paraphrase of a paywalled page.