Open academic book and magnifying glass examining overlapping university ranking trends and research data.

Reading University Rank Changes: What Recent Research Shows

Share

A move of a few places in a university ranking usually tells you less than it appears to. Five studies and reviews published between 2022 and 2026 point the same way: many institutions sit so close together that measurement error and design choices can reorder them; different ranking systems can disagree sharply about the same universities; methodology changes break year-to-year comparisons; and an institution's overall position is a weak guide to the quality of a specific program. The practical answer for students, families and institutions is to read ranks across several editions, several systems and the relevant subject, and to treat small differences as ties.

Why small rank differences are fragile

Measurement error makes close neighbors hard to separate

In a 2022 commentary in Significance, Adrian Barnett and Elizabeth Gadd argue that league tables present positions as precise even though every input is measured with error. To illustrate the point, they took citation data on 1,225 universities from the Leiden ranking for 2016 to 2019 and applied an assumed error of plus or minus 10 percent to the counts. Under that assumption, the median university overlapped with 76 other universities, and the overlap was larger for universities with smaller citation counts. Only two institutions at the very top stood clear of the rest.

The exercise does not show that any particular ranking is wrong. It shows that when the gaps between institutions are small relative to plausible measurement error, the order among neighbors is not well determined by the data. The authors also note that the weights used in composite tables appear to rest on expert opinion rather than evidence, and that changing established weights could produce dramatic changes in the tables.

Weights and normalization are design choices

A 2026 review in Frontiers in Research Metrics and Analytics by Konstantinos Latifis, Antonis Sidiropoulos and Georgios Evangelidis reconstructs the calculation pipelines of five widely used global rankings. It describes how each system selects indicators, normalizes and transforms them, and combines them with fixed weights into a single score. The authors argue that these design choices encode normative assumptions and drive discrepancies between systems, and that rankings should be read as algorithmic evaluative frameworks rather than neutral measurements. They also review alternatives that avoid a single fixed-weight score, such as dashboard reporting, tier and profile frameworks, and dominance-based methods, while noting that none of these is value-neutral either.

What ten years of national league tables show

Owen Tomlinson and James Butler, writing in BMC Medical Education in 2026, analyzed ten consecutive years (2016 to 2025) of medicine program rankings from three national league tables in the United Kingdom. They report three findings that matter for anyone reading a ranking:

  • Broad agreement between systems. Over the decade, the three tables correlated with each other at Spearman coefficients of 0.76 to 0.85, so programs ranked highly by one tended to be ranked highly by the others.
  • Disagreement at the top and movement within systems. The three programs with the highest ten-year mean positions were not the same across all systems, and some institutions showed high variance within a single table from year to year. Because that movement occurred under a single methodology, the authors suggest a "tight pack" of programs separated by small differences, and they conclude that small differences in quality are being overblown by a forced ordering.
  • A weak link between institution and program. An institution's overall rank was only weakly correlated with the rank of its medicine program (r = 0.36 to 0.38). A well-ranked university does not guarantee a well-ranked program, and the reverse also holds.

The authors advise applicants to weigh factors such as teaching approach, size and location, and not to rely on rankings alone.

When rankings disagree about the same universities

A study published in October 2026 by Fouad Ahmed Atallah in Frontiers in Research Metrics and Analytics examined how three of the most widely used global rankings represented universities in Saudi Arabia from 2022 to 2025. National policy there sets targets for the number of universities in the global top 100 and top 200 and refers to all three systems, so disagreement between them has direct consequences.

Agreement between the two most closely aligned systems was very high in 2022 (Spearman correlation 0.982) but fell to 0.726 in 2023 and 0.626 in 2024 before recovering to 0.932 in 2025. In 2025, agreement between one of those systems and the third was much weaker (0.437). Both of the closely aligned systems introduced substantial methodological revisions in their 2024 editions, and the author cautions that rank movements across that boundary should not be read automatically as changes in institutional performance.

The threshold question was the most striking. In 2025, one system placed three of the country's universities in its top 200, and the other two systems each placed one, and those two were different institutions. The author concludes that the systems are related but not interchangeable, and recommends that any report on threshold targets name the system used and show results from all three.

Does a higher rank mean better learning or research?

J. Paul Grayson and Kyle Grayson, in a 2026 article in the Canadian Journal of Higher Education, used United Kingdom data to test some assumptions behind league table indicators. The United Kingdom offers unusual data for this purpose: a national research assessment based on expert review, and a published "value added" measure that compares students' degree results with what their entry grades predicted.

  • Across 121 universities in the 2019 league tables, a regression of value added on institutional characteristics explained 20 percent of its variance. Entry grades had the largest effect (beta = 0.31). Student-to-staff ratio, spending on student services and research quality had small effects that were not statistically significant.
  • In a regression of research quality, total research grants and contracts had a large coefficient (beta = 0.55) that was not statistically significant. Grants are an input, the authors note, not a measure of output quality.
  • For departments of Politics and International Studies, the correlation between value added and the institution's overall league table position was zero. The correlation between departmental research quality scores and overall position was -0.42 (negative because rank 1 is best), which the authors describe as statistically significant but leaving most of the variance unexplained.

The authors caution that United Kingdom results cannot be generalized automatically. Their findings nonetheless suggest that inputs such as grant income or student awards are questionable proxies for quality, and that an overall rank can hide large differences between departments.

Five recent sources on how much a rank position can tell you
SourceSetting and dataMain findingWhat it means for readers
Barnett and Gadd (2022)Citation data for 1,225 universities, 2016 to 2019, with an assumed 10 percent errorMedian university overlapped with 76 othersNeighboring positions may not be meaningfully different
Latifis et al. (2026)Review of five global rankings and alternative methodsIndicator, normalization and weighting choices drive discrepanciesA rank reflects a design, not only performance
Tomlinson and Butler (2026)Three UK national tables for medicine, 2016 to 2025High agreement overall (0.76 to 0.85), but movement within tables and a weak institution-program link (0.36 to 0.38)Use multi-year averages and subject-level ranks
Atallah (2026)Three global rankings, Saudi universities, 2022 to 2025Agreement varied by year and system pair; top-200 counts depended on the systemName the system and edition; compare more than one
Grayson and Grayson (2026)121 UK universities, 2019 tables, plus one disciplineLeague position weakly related to research quality and unrelated to value added in one disciplineCheck what the indicators actually measure

Our interpretation

Read together, these sources suggest that a single year's movement of a few places is weak evidence of real change. Three separate mechanisms can produce movement without any change in the institution: measurement error among closely spaced institutions (Barnett and Gadd), methodological revisions between editions (Atallah), and the forced ordering of a tight pack (Tomlinson and Butler). Large, sustained differences that appear across several systems are more informative than small or isolated ones.

A second point is that the overall rank answers a different question from the one most applicants are asking. Two of the studies found only weak or no relationship between overall position and program-level measures. For a student choosing a program, the subject-level evidence and the indicators behind it usually matter more than the headline number.

These are interpretations of correlational and descriptive evidence. None of the studies shows that rankings cause particular outcomes, and none claims that rankings carry no information. Tomlinson and Butler found broad agreement between national tables over a decade, which suggests that rankings can capture stable, large differences even when they exaggerate small ones.

Practical steps for reading a rank change

For students and families

  1. Look at three or more editions. A program that holds a similar position for several years tells you more than one that moved once.
  2. Check for a methodology change. Before reading a jump or a drop, read the ranking's methodology notes for the edition. If the method changed, treat the two years as different measurements.
  3. Know how the ranking is published. UNIRANKS rankings, for example, are published by edition (currently UNIRANKS 2027) under a versioned methodology (V4.1.6.2), not updated continuously in real time; compare editions only after reading the methodology notes.
  4. Compare at least two rankings. If the systems disagree strongly, the disagreement itself is useful information about what each one rewards.
  5. Use the subject ranking where one exists. An institution's overall position can differ substantially from its position in a specific discipline.
  6. Treat near neighbors as similar. Institutions a few places apart, or in the same published band, may not differ in any way that matters to you.
  7. Ask what the indicators measure. Inputs such as income, grants or entry grades are not the same as learning outcomes or research quality.

For academics and institutional leaders

  1. Report the system and edition every time a rank is cited, and avoid comparing positions across a methodological break.
  2. Avoid announcing small movements as evidence of improvement or decline.
  3. Set targets on underlying performance, not on a single system's threshold, and if a threshold target is used, show results from every system it names, as Atallah recommends.
  4. Publish subject-level and indicator-level information so that applicants can look past the headline number.

Limitations

This review draws on five sources, all read beyond the abstract: four were read in full text, and one (Barnett and Gadd, 2022) was read from the article as displayed on its publisher's page. Barnett and Gadd's piece is a short commentary that uses an illustrative error assumption, not an estimate of actual measurement error. The review by Latifis and colleagues is a narrative, not a systematic, review. Two of the empirical studies use United Kingdom data and one covers a single country's universities, with small samples in some years (between 4 and 19 institutions ranked by each pair of systems). All the evidence is correlational or descriptive. Studies that could not be verified at source were not included.

Related reading on UNIRANKS

For more on rankings literacy, see the Local & International Rankings hub. Related articles discuss regional and global ranking systems and the shift toward transparency in rankings.

Editorial disclosure

UNIRANKS publishes university rankings, and its methodology includes a Local & International Rankings dimension. This article reviews independent research on how to interpret university rank changes. It does not describe how UNIRANKS scores institutions.

Research References

  1. Barnett, A., & Gadd, E. (2022). University League Tables have no Legs to Stand on. Significance, 19(4), 4-7. https://doi.org/10.1111/1740-9713.01663 (verified 20 November 2026)
  2. Latifis, K., Sidiropoulos, A., & Evangelidis, G. (2026). International university rankings: composite indicators and methodological approaches. Frontiers in Research Metrics and Analytics, 11, 1828850. https://doi.org/10.3389/frma.2026.1828850 (verified 20 November 2026)
  3. Tomlinson, O. W., & Butler, J. (2026). A ten-year retrospective analysis of medical school rankings in the United Kingdom. BMC Medical Education, 26(1), 278. https://doi.org/10.1186/s12909-025-08538-0 (verified 20 November 2026)
  4. Atallah, F. A. (2026). Beyond league tables: a multidimensional and rank-aggregation assessment of Saudi universities under vision 2030. Frontiers in Research Metrics and Analytics, 11, 1993617. https://doi.org/10.3389/frma.2026.1993617 (verified 20 November 2026)
  5. Grayson, J. P., & Grayson, K. (2026). Why the Rankings are Wrong: Lessons from Data Underlying U.K. University League Tables. Canadian Journal of Higher Education, 56(1), 35-54. https://doi.org/10.47678/cjhe.v1i1.190271 (verified 20 November 2026)

Sources

  1. UNIRANKS University Ranking Methodology V4.1.6.2. UNIRANKS. https://www.uniranks.com/methodology (accessed 20 November 2026)
Share this article