UniRank
QS THE ARWU Methodology

Which University Ranking Is Most Reliable: A Side-by-Side Test of QS, THE, and ARWU

“Most reliable” is not an overall property that the published evidence establishes for any one ranking. The defensible decision rule is purpose-specific: consult the Academic Ranking of World Universities 2025 (ARWU 2025) for research volume and Nobel or Fields Medal output; the Times Higher Education World University Rankings 2026 (THE WUR 2026) for teaching and the learning environment; the QS World University Rankings 2026 for employer perception and employability; and the QS or THE edition that matches the particular meaning of international outlook being investigated.

Those are different uses of “reliable”. One ranking may be more relevant to a question without being more accurate overall. No publisher in this comparison publishes an overall reliability figure, so unirank.io does not convert the weights into a composite reliability score.

Research output, citations and Nobel-level performance

For research volume and Nobel-level output, begin with ARWU 2025. Its published criteria are:

  • Alumni: alumni winning Nobel Prizes or Fields Medals, weighted at 10%.
  • Award: staff winning Nobel Prizes or Fields Medals, weighted at 20%.
  • HiCi: Highly Cited Researchers selected by Clarivate, weighted at 20%.
  • N&S: papers published in Nature and Science, weighted at 20%.
  • PUB: papers indexed in SCIE and SSCI in the Web of Science, weighted at 20%.
  • PCP: per-capita academic performance, weighted at 10%.

ARWU 2025 states that its N&S measure covers research articles published from 2020 to 2024, while its PUB measure uses papers indexed in 2024. It also uses the Highly Cited Researchers list issued in November 2024. This makes ARWU 2025 the most directly applicable of these tables to award attainment and research-paper volume, but it does not turn the table into a measure of teaching quality, employability or internationalisation.

For faculty-normalised citation intensity, consult QS 2026. QS assigns 50% to its Research and Discovery lens, divided between Academic Reputation at 30% and Citations per Faculty at 20%. QS explains that Citations per Faculty divides citation counts by faculty headcount to account for institutional size. Academic Reputation is not a direct publication-volume measure; it is based on academic experts’ nominations.

THE WUR 2026 separates research into Research environment: 29% and Research quality: 30%. Research environment comprises Research reputation at 18%, Research income at 5.5% and Research productivity at 5.5%. Research quality comprises Citation impact at 15%, Research strength at 5%, Research excellence at 5% and Research influence at 5%. THE therefore offers a broader research profile, while ARWU 2025 is the clearer starting point for the specific question of research volume and Nobel or Fields Medal output.

Teaching and the learning environment

For teaching and the learning environment, THE WUR 2026 has the most directly labelled teaching pillar. THE assigns 29.5% to Teaching, distributed across:

  • Teaching reputation: 15%
  • Staff-to-student ratio: 4.5%
  • Doctorate-to-bachelor’s ratio: 2%
  • Doctorates-awarded-to-academic-staff ratio: 5.5%
  • Institutional income: 2.5%

These are published weights, not a guarantee that the resulting rank answers every question about teaching. Institutional income, for example, is structurally different from a classroom observation.

QS 2026 assigns 10% to its Learning Experience lens, consisting of Faculty Student Ratio at 10%. That is a student-facing scale measure, but the supplied QS methodology does not present it as a complete teaching assessment comparable to THE’s named teaching pillar.

ARWU 2025 publishes no teaching or student-facing indicator and therefore supplies no teaching weight for this purpose.

Employer perception and employability

For employer perception and employability, consult QS 2026. Its Employability and Outcomes lens carries 20%, divided between Employer Reputation at 15% and Employment Outcomes at 5%.

QS states that Employer Reputation is collected through its Employer Survey and measures the reputation of institutions and their programmes among employers. The supplied QS methodology names Employment Outcomes but does not reproduce its detailed construction, so it cannot support a more specific interpretation of that indicator from the available evidence.

THE WUR 2026 lists Industry: 4%, comprising Industry income at 2% and Patents at 2%. Neither is an employer-reputation survey indicator. THE’s patent measure concerns patents citing university research, which is different from asking employers how they perceive an institution.

ARWU 2025 contains no employer-reputation or employability indicator. A zero should not be inserted where a publisher has not assigned a weight.

International outlook

The appropriate internationalisation ranking depends on what “outlook” means.

For international faculty, student composition, diversity or research networks, consult QS 2026. Its Global Engagement lens carries 15%, divided into:

  • International Faculty Ratio: 5%
  • International Research Network: 5%
  • International Student Diversity: 0%
  • International Student Ratio: 5%

For THE’s broader international-outlook construct, consult THE WUR 2026, which assigns 7.5% to International outlook. THE also lists a study abroad metric at 0 per cent. Because that metric currently has no weight, it cannot discriminate between institutions in the published THE WUR 2026 total score. The supplied methodology extract does not provide the names or individual percentages of the other metrics within the International outlook pillar, so a complete component comparison cannot be reconstructed from the available evidence.

ARWU 2025 publishes no internationalisation indicator or weight.

Surveys, indicators and score construction

QS 2026 is survey-based only in part. Academic Reputation uses academic experts’ nominations, while Employer Reputation uses an employer survey. Its other indicators include citations, faculty ratios, international measures, employment outcomes and sustainability; QS assigns Sustainability at 5%.

THE WUR 2026 uses 18 indicators, but it is not survey-free. The methodology says its Teaching reputation at 15% and Research reputation at 18% are underpinned by an Academic Reputation Survey. THE reports that the survey ran from November 2024 to January 2025 and that, combined with its 2024 results, supplied more than 108,000 responses. Any description of THE as having no reputation survey would conflict with the publisher’s own WUR 2026 methodology. What THE’s listed indicators do not include is a dedicated employer-reputation measure.

THE standardises each indicator and combines them in the published proportions. It uses a cumulative probability function, generally a version of Z-scoring, with an exponential component for Teaching reputation, Research reputation, Research excellence, Research influence and patents.

ARWU 2025 uses six academic or research-performance indicators and no reputation survey. For each indicator, the highest-scoring institution receives a score of 100, while other institutions are calculated as a percentage of that top score. This describes relative scoring within ARWU; it is not an overall reliability measure.

Coverage makes cross-ranking comparisons conditional

The QS World University Rankings 2026 includes over 1,500 institutions and represents over 100 locations. QS also warns that corrections to university performance may be made after publication.

THE WUR 2026 describes its population as research-intensive universities, but the supplied methodology does not state the number of ranked institutions or countries in the published table. Its inclusion rules nevertheless show that it is a selected population: universities can be excluded if they do not teach undergraduates, if research output falls below 1,000 relevant publications between 2020 and 2024, with a minimum of 100 a year, or if 80 per cent or more of output is concentrated in one of 11 subject areas.

ARWU 2025 says that more than 2,500 universities are actually ranked, although only the best 1,000 are published. Its candidate pool includes institutions meeting specified award, Highly Cited Researcher, Nature or Science, and Web of Science publication conditions. The available results evidence does not provide a complete country count. Its supplied country extract covers ranks 1–30, and the country for rank 30 is blank, so it cannot establish ARWU’s full geographic coverage.

Consequently, comparing the same universities across these tables is not necessarily like-for-like: the tables can use different eligibility rules, indicator definitions and published populations.

Reading ranks without inventing certainty

ARWU 2025 prints entries marked “14 (tied)” and “26 (tied)”. Those are published tied positions and should not be converted into sequential places. Its footnote says institutions within the same rank range are listed alphabetically.

The available THE WUR 2026 evidence is methodological rather than a complete results table, so it cannot establish which institutions received tied ranks or published rank bands. No cross-publisher median or reliability conclusion can be calculated from the supplied rank evidence.

The title’s question therefore has no defensible overall winner. Reliability depends on whether the edition’s population, indicators and weights match the reader’s question. The publisher that appears strongest on one axis may provide no usable evidence on another, and unirank.io does not fill that gap with an invented score.