Performance marketing for AI-era growth – backed by models, not gut feeling.
TL;DR
  • 162 universities have been in the ARWU top 100 across the 24 editions from 2003 to 2026. 58 of them are in all 24; the other 104 rotate through the remaining places.
  • The United States had 58 universities in 2003 and 37 in 2026. Mainland China had none before 2016 and 16 in 2026.
  • Entries per edition: 3.45 from 2004 to 2014 (11 editions), 4.92 from 2015 to 2026 (12 editions). 6 of the later 59 entries are mergers or renames; without them it is 4.42.
  • Three tables of my own counts, CC BY 4.0. The ranking lists themselves are published by ShanghaiRanking, one per edition.
Download

arwu_top100_dynamics_universities_2026-10.csv. 162 rows, one per university, 16 columns.

arwu_top100_dynamics_country_year_2026-10.csv. 552 rows, 23 countries and regions × 24 editions, 10 columns.

arwu_top100_dynamics_exits_2026-10.csv. 98 rows, one per exit, 4 columns.

Entries and exits, edition by edition

EditionUniversitiesEntriesFirst appearancesReturnsExits
200410099010
20051003213
20061013122
20071011011
20081003214
20091002022
20101001011
20111013302
20121004315
20131002022
20141007527
20151003123
20161009639
20171005235
20181005325
20191004224
20201007617
20211003123
20221008538
20231003213
20241006426
20251002112
20261004314

Table 1. An entry is a university in the edition that was missing from the previous one: a first appearance or a return. In 2003, 2006, 2007, 2011 a tie at place 100 puts 101 universities in the list.

Two periods

PeriodEditionsEntriesPer editionFirst appearancesReturnsMergers and renames
2004 to 201411383.4525130
2015 to 202612594.9236236

Table 2. The periods have 11 and 12 editions, so they are compared per edition. 2003 has no previous edition and opens neither period.

How long universities stay

Editions in the top 100Universities
2458
20 to 2310
10 to 1939
5 to 923
2 to 419
113

Table 3. 162 universities by the number of editions, out of 24, in which they are in the top 100.

Mergers and renames

UniversityEventFirst editionRank then
Sorbonne Universitymerger 2018201836
Paris-Saclay Universitymerger 2020202014
PSL Universitymerger 2019202036
University of Parismerger 2019202065
Université Grenoble Alpesmerger 2020202099
Université Paris Citérename 2022202278

Table 4. A merger or a rename gets a new id, so it counts as one exit and one entry. Each event was checked on the university's own page.

Codebook

arwu_top100_dynamics_universities_2026-10.csv

ColumnWhat it holds
university_idStable slug of the university. A merger or a rename gets a new id.
name_latestName as ARWU printed it in the last edition with the university in the top 100.
country_codeISO 3166-1 alpha-2. Mainland China is CN and Hong Kong is HK.
country_arwuCountry or region as printed by ARWU.
first_editionFirst edition in the top 100 (from 2003).
last_editionLast edition in the top 100 (up to 2026).
editions_in_top100Editions out of 24 in the top 100.
best_rankBest world rank across those editions.
best_rank_first_yearFirst edition with that best rank.
first_rankWorld rank in the first edition.
worst_rankLowest world rank while in the top 100.
rank_2003World rank in 2003; empty if not in the top 100.
rank_2026World rank in 2026; empty if not in the top 100.
exitsTimes the university left the top 100.
re_entriesTimes it came back after at least one edition out.
lineageMerger or rename that created this id, with the year; empty otherwise.

arwu_top100_dynamics_country_year_2026-10.csv

ColumnWhat it holds
yearEdition.
country_codeISO 3166-1 alpha-2.
country_arwuCountry or region as printed by ARWU.
n_top100Universities of the country in the top 100 of the edition. Zero is a count.
n_top10Of them, in places 1 to 10.
n_top50Of them, in places 1 to 50.
best_rankBest world rank of the country; empty if none.
best_university_idUniversity at that rank; two ids with a semicolon on a tied place.
n_entriesUniversities of the country that entered in this edition (first appearance or return); empty for 2003.
n_exitsUniversities of the country that were in the previous edition and are missing from this one; empty for 2003.

arwu_top100_dynamics_exits_2026-10.csv

ColumnWhat it holds
year_exitFirst edition in which the university is missing after being in the previous one.
university_idUniversity id.
last_rankWorld rank in the last edition before the exit.
next_statusreturned_YEAR: the first later edition with the university back; not_returned: not back up to 2026.

Method and limitations

Where these numbers are used
The Top 100 Is a Moving Target (Article 6) The article that reads these tables, with the interactive chart.
ARWU Top 100 by Country – dataset Universities per country and the best-ranked university of each, 2003 to 2026.

Cite this
Atamanova, M. (2026). ARWU Top 100 Dynamics, 2003–2026 [Data set]. mariascales.com. https://mariascales.com/research/long-bet/data/arwu-top100-dynamics/

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.