Performance marketing for AI-era growth – backed by models, not gut feeling.
TL;DR
  • 162 universities have been in the ARWU top 100 across 24 editions. 58 of them were there in all 24.
  • The United States: 58 of 101 in 2003, 37 of 100 in 2026. In places 1 to 20 it held 15 and holds 14. The loss is in places 21 to 100: 43 then, 23 now.
  • Mainland China: none through 2015, 16 of 100 in 2026. Between 2003 and 2026 no European country gained or lost more than two universities.
  • Turnover grew: 3.45 entries per edition from 2004 to 2014, 4.92 from 2015 to 2026. For a shortlist I read a university's line across editions before its place in one.
How I counted
  • 24 editions of the Academic Ranking of World Universities (ARWU), 2003 to 2026, places 1 to 100, read from ShanghaiRanking's published lists.
  • In 2003, 2006, 2007, 2011 a tie at the bottom puts 101 universities into the top 100. There the denominator is 101.
  • A merger or a rename is a new university in my count.
  • An entry is a university that is in this edition and was not in the previous one: a first appearance or a return.
  • Mainland China and Hong Kong are counted apart, as ARWU lists them.

24 editions in one chart

Chart: ARWU Country Dynamics, 2003 to 2026, on its own page.

Show the data: universities per country, 2003, 2016 and 2026
Country or region200320162026
United States585037
Mainland China0216
United Kingdom988
Switzerland345
Australia265
Germany534
France234
Canada443
Japan542
Netherlands332
Sweden332
Belgium122
Denmark122
Israel122
Hong Kong002
Singapore012
Norway111
South Korea001
Austria100
Finland110
Italy100
Russia010
All101100100

Table 1. Universities in the top 100 by country or region, 2003, 2016 and 2026. Countries with none in all three editions are left out; the full grid is in the dataset.

Harvard never moved. The US lost 21 places.

Harvard has held place 1 in all 24 editions. Over those 24 editions the United States went from 58 of 101 universities in the top 100 to 37 of 100.

The top 20 barely moved. The reshuffle is in places 21 to 100: the US lost 20 there, mainland China gained 15.

Country or regionPlaces 1 to 2021 to 5051 to 100
United States15 → 1420 → 1323 → 10
Mainland China0 → 10 → 50 → 10
United Kingdom4 → 31 → 24 → 3
All other countries1 → 29 → 1024 → 27
All20 → 2030 → 3051 → 50

Table 2. Universities by band of places, 2003 → 2026. In 2003 a tie at place 100 puts 51 universities in the last band.

What the lines show

Zoom out

PeriodEditionsEntriesPer editionFirst appearancesReturnsMergers and renames
2004 to 201411383.4525130
2015 to 202612594.9236236

Table 3. Entries into the top 100 by period. The periods have 11 and 12 editions, so they are compared per edition.

Two objections to that number, and what is left after them:

Largest moves, 2003 to 2026

Among the 64 universities in the top 100 in both 2003 and 2026:

University20032026Places
The University of Melbourne9237+55
The University of Manchester8951+38
University of Copenhagen6534+31
University of Bristol5591−36
Uppsala University5994−35
University of California, Santa Barbara2661−35

Table 4. The three largest rises and the three largest falls, 2003 to 2026, among the 64 universities in both editions.

Before a university goes on the shortlist

Data

Also in this series
The Problem Statement The four-constraint optimisation problem and why every obvious answer fell apart.
How I Built It Why ARWU over QS and THE, why 100 universities, and what 1,258 normalised pairs made possible.
What the World Actually Offers CS at 95 universities. AI at 12. France above Imperial. What the coverage matrix actually showed.
The Unfashionable Four Four A-level subjects. Mathematics named by 22 of the 23 UK-taught programmes; Chemistry as the fourth opens all 23.
The Floor Keeps Rising The scarce skill in data and computing has shifted seven times since the 1880s. The foundation underneath it has never moved.
About the author

Maria Atamanova is a Senior Growth & Performance Marketing Manager and applied mathematician. Twelve years of growth in apps, e-commerce, EdTech and AI data platforms.

The method here – define the criteria, build the dataset, test every obvious answer – is the same one I use for growth teams; see the case studies.