What the Impact Factor, CiteScore, and h-Index Really Measure

Three numbers dominate the conversation about where research is published and whose contributions are taken seriously: a journal's impact factor, its CiteScore, and a researcher's h-index. Each is quoted as though it settled a question of standing. Yet each measures something narrow and specific, and each is routinely asked to answer questions it was never built to answer.

Research metrics exist because influence is difficult to observe directly and citations offer a countable trace of it. When one paper cites another, it leaves a record that the earlier paper was used, and counting those records produces a number that stands, imperfectly, for influence. Every metric discussed here is built on that single act of counting. What separates them is whose citations they count, over what period, and what the resulting figure is taken to represent.

The citation as the common unit

A citation is not a measure of quality. It records that one author referred to another's paper, for reasons that range from building on a finding to disputing it or noting it in passing. Counted in bulk, citations reveal how widely a body of research has been used, and that is a real and useful signal. But the signal is coarse. It cannot distinguish approval from criticism, a central influence from a passing mention, or a genuine intellectual debt from a courtesy. Every number derived from citations inherits this coarseness, however sophisticated the formula applied to it.

The level each number describes

The three metrics do not describe the same object, and confusing their levels is the most common error made with them. The impact factor and CiteScore are properties of a journal: they say something about the average article in a title over a period, and nothing about any particular article within it. The h-index is a property of an individual researcher, or of any defined body of research. Treating a journal metric as though it certified a single paper, or an author metric as though it ranked a field, applies a number to a question it cannot answer.

The impact factor is the oldest and most widely quoted of the three, and the most widely misused.

What the Impact Factor measures

The Journal Impact Factor, published annually by Clarivate in the Journal Citation Reports and drawn from the Web of Science database, measures the average number of citations received in a given year by the articles a journal published in the two preceding years. In outline, it divides the citations counted in that year by the number of citable items published across the two-year window. The result is a single figure describing how often, on average, a recent article in that journal was cited.

What the Impact Factor hides

The figure is an average, and averages conceal their distribution. Citations are distributed very unevenly: in most journals a small number of heavily cited articles account for a large share of the total, while the majority of articles are cited far less often than the impact factor implies. The number therefore describes the journal, not the article, and a paper published in a high-impact-factor journal may itself attract few citations or none. Two further limits matter. The two-year window suits fast-moving fields and penalizes those, such as mathematics and much of law, in which citations accumulate slowly over many years. And because the items counted in the numerator and those counted in the denominator are defined differently, the figure is sensitive to editorial choices unrelated to research quality. Above all, the impact factor is specific to its field: a value that marks an outstanding journal in one discipline is unremarkable in another, and comparison across fields carries no meaning.

CiteScore is a more recent journal metric, built on a different database and a wider base of sources.

What CiteScore measures

Published by Elsevier and calculated from the Scopus database, CiteScore measures citations over a four-year window to the documents a journal published in those same four years, divided by the number of those documents. Its window is longer than the impact factor's, and it counts a broader range of document types, including conference papers and book chapters, in both the citations counted and the items divided by. Because Scopus indexes more titles than the Web of Science core collection, CiteScore is available for a far larger number of journals.

How CiteScore differs from the Impact Factor

The longer window gives citations more time to accumulate, which narrows the disadvantage that slow-citing fields face under the impact factor, though it does not remove it. The wider source base and the consistent treatment of documents in numerator and denominator make the figure harder to distort. But CiteScore remains a journal metric with the journal metric's central limitation: it describes the average article in a title, not any single article, and it cannot be read as a verdict on a particular paper. Nor are its values interchangeable with impact factors, since the two are computed from different databases over different windows. A journal's CiteScore and its impact factor are not the same measurement on the same scale, and setting one against the other compares the systems as much as the journals.

The h-index differs from the other two in what it describes: not a journal, but a researcher or a defined body of research.

What the h-Index measures

Proposed in 2005 by the physicist Jorge Hirsch, the h-index is defined by a single rule: a researcher has an h-index of h if h of their publications have each been cited at least h times. An author with an h-index of twenty has twenty papers, each cited at least twenty times. The measure deliberately combines two things a plain citation count does not: productivity, in the number of papers, and sustained impact, in the citations each attracts. A single highly cited paper cannot by itself produce a high h-index, and a large number of rarely cited papers cannot either.

What the h-Index misses

The construction that gives the h-index its balance also fixes its limits. It rises with the length of a career, since more papers and more accumulated citations become possible with time, so it compares a senior and an early-career researcher on unequal terms. It is insensitive to the citations a paper attracts beyond the threshold: an article cited thousands of times counts no more toward the index than one cited exactly h times, so a landmark paper leaves no special mark on the figure. It is strongly field-dependent, because citation rates differ from one discipline to the next. It cannot fall, and it says nothing about a researcher's role in the papers it counts, treating sole and marginal authorship alike. And its value depends on the database used to compute it, so the same researcher carries different h-indices in Web of Science, Scopus, and Google Scholar.

A figure that carries weight in decisions becomes a target, and each of these metrics can be moved by effort directed at the number rather than at the research.

The figures can be inflated

Because the impact factor and CiteScore count citations to a journal, they can be raised by practices that add citations without adding value. A journal that presses authors to cite its own recent articles, or an arrangement in which journals cite one another reciprocally, lifts the figure with no change in the quality of what is published. The h-index is open to a milder form of the same effect, since an author's own citations to earlier papers count toward it. None of this makes the metrics worthless, but it means a figure cannot be read in innocence of how it was produced, and an unusually high value invites the question of how it was reached.

The databases do not cover everything

Every one of these numbers is computed from a particular database, and each database indexes a particular slice of the literature. Journals in English and in well-funded fields are covered more completely than regional titles, non-English research, and emerging areas, so a researcher or a journal working outside the well-indexed core will show lower figures for reasons that have nothing to do with the merit of the research. The number reflects what the database can see, and what it cannot see does not count.

Beyond the specific limits of each metric lies a set of questions that none of the three can answer, and that they are nonetheless asked to answer daily.

The quality of a single article

No citation metric assesses whether a particular study is well designed, whether its conclusions follow from its evidence, or whether it will prove to be correct. A journal's impact factor is an average over its contents and certifies nothing about the article in front of the reader; the h-index describes a career and certifies nothing about a single paper within it. The only assessment of an individual paper's soundness is the one that reading it, and the review it has passed through, can provide. A number attached to the journal or the author is not a substitute for that reading. The recognition that these figures are misapplied in exactly this way has itself become formal: research-assessment reform efforts, among them the San Francisco Declaration on Research Assessment, advise against using journal-based metrics as a proxy for the quality of an individual article or as a factor in decisions on hiring, promotion, and funding.

The comparisons the numbers cannot bear

Because every one of these figures is shaped by the citation culture of a field and by the database that produced it, the comparisons most often made with them are the ones they can least support. A high figure in a fast-citing discipline set beside a lower figure in a slow-citing one measures the disciplines, not the research. An author metric compared across career stages measures seniority as much as achievement. A journal figure from one database set against a figure from another compares the databases. Used within a single field, at the level each metric describes, the numbers carry real information; carried across those boundaries, they measure the boundary rather than the thing.

The most recent research scores lowest

Citations take time to appear. A paper published this year has had almost no opportunity to be cited, and a researcher's newest research contributes least to any of these figures precisely when it may matter most. The impact factor's two-year window and CiteScore's four-year window both measure a period that has already passed, and the h-index reflects citations accrued over years. As measures of established influence this is appropriate, but it means none of the three can register the value of research whose influence has not yet had time to form. The frontier of a field is always underrepresented in the numbers that describe the field.

Read each number as what it is: a count of citations, at a stated level, over a stated period, from a stated database. Ask first what the metric describes, a journal or a person, and refuse to let a journal figure stand in for the worth of a single article or a single author. When a metric is quoted, note the field it belongs to and compare only within that field, since a value that is high in one discipline is ordinary in another. Look past the headline figure to the distribution beneath it, remembering that a journal's average is not the fate of the average paper in it. And treat any single number as one input among several, read alongside the research itself, rather than as the conclusion. Where the question concerns a single paper, the citation count of that paper, and above all its content, answer it far better than any figure attached to the journal that carries it.

Understood this way, the impact factor, CiteScore, and h-index are neither meaningless nor authoritative. Each measures something narrow and real, and each is silent on the questions of quality and importance that matter most. The figures are a starting point for judgment, not a replacement for it, and the reader who knows precisely what each one counts is the reader least likely to be misled by it.