The Alumni Factory

Universities describe themselves in mission statements. Their alumni describe them better. This article takes the 250 educational institutions with at least 2,000 notable alumni in Wikidata — a panel discovered from the data, not curated, which is why Eton College, West Point and the Paris Conservatoire sit beside Harvard — and reads each one as a factory with an output mix: so many politicians, so many physicists, so many footballers per thousand famous graduates.

One anomaly, kept on purpose. The Imperial University of Dorpat (today's Tartu, Estonia) ranks 8th with 15,269 "notable" alumni — because Estonian editors mass-imported historical matriculation registers into Wikidata. It is a reminder of what this data is: not a ranking of eminence, but a map of who got recorded. We flag it wherever it appears.

1 The league table, as the archive sees it

Harvard leads with 25,368 recorded notable alumni; Berkeley, Stanford, Washington and Tokyo follow. The list is American-heavy — partly real dominance of the notability economy, partly English-language Wikipedia's home-field advantage, a bias this whole article inherits.

The 30 institutions with the most notable alumni

People in Wikidata with "educated at" → the institution

Data table (all 250)

Figure 1. Collegiate universities are split across entities (Harvard Law/Business/Medical School and Harvard College all appear separately further down the list; Oxbridge colleges likewise), so true university-level totals are higher than any single bar. Dorpat's bar is the import artifact flagged above.

2 Fingerprints

Divide each school's alumni among thirteen career fields and the mission statements become unnecessary. The Beaux-Arts de Paris is 88% visual artists; the Paris Conservatoire, 78% musicians; Moscow's Gerasimov film institute, 68% film and stage; West Point, 53% military (plus, delightfully, 18% athletes — the Academy's sports programs produce Olympians). ENA — France's presidential nursery — runs 51% politics and government. The great research universities, by contrast, are all the same fingerprint: half science-and-academia, a spread of everything else. Elite generalism is itself a type.

What each factory makes

Share of the institution's notable alumni in each field · darker = larger share · specialists above, generalists below

Data table

Figure 2. Rows sum to 100% across the thirteen fields (people are deduplicated within a field but can appear in several). "Science & academia" absorbs anyone recorded as a researcher or university teacher — one reason research universities look science-heavy: their alumni's academic careers are the best-databased occupation on Wikidata. Hover any cell.

3 You can read coeducation in the cohorts

Plot the share of each school's notable alumni who are women, by birth decade, and admissions history reappears as demography. Princeton's line sits at 1% for people born in the 1930s — who would have applied around 1950, two decades before coeducation — jumps with the 1940s cohort (the first to arrive after 1969), and reaches 41% for the 1990s-born. Yale moves in lockstep. Oxford, which had women's colleges from the 1870s, never sits near zero. Tokyo, whose line reaches only 26% for the 1980s cohort, shows how slowly elite pipelines can open even without formal exclusion. And Eton's line is zero by construction — a reminder that some pipelines never opened at all.

Women among notable alumni, by birth cohort

Nine institutions · birth decades 1870–1990 · dotted marks approximate first coeducational admissions

Data table

Figure 3. Birth decade ≈ enrollment two decades later. Thin segments mark cohorts with fewer than 100 recorded people. Vertical ticks: Oxford grants women degrees 1920; Cambridge 1948; Tokyo admits women 1946; Yale & Princeton coeducate 1969 (shown at the birth cohort ~19 years prior). The gap between the tick and the line's rise is the pipeline's lag — notability takes a career.

4 Explore any of the 250

The original version of this analysis was a dashboard comparing professions and gender across elite universities. Here it is, rebuilt on the full discovered panel: pick any institution — a Soviet film school, a Japanese private university, a Texas land-grant — and read its output mix, its gender history, and its most-linked alumni.

Institution explorer

Institution

Data table (occupations)

Figure 4. Left: the institution's ten most common recorded occupations. Right: women's share of notable alumni by birth cohort (line thins below 100 people). Chips: the institution's most-interlinked alumni on Wikidata — a rough celebrity ranking. All caveats from the methods section apply per institution.

What the factory metaphor finally yields is not a ranking but a taxonomy. There are schools that make one thing superbly (a conservatory, an academy, a grande école); schools that make everything (the research universities, interchangeable at this altitude); and schools whose product changed mid-century — the moment half of humanity was finally allowed onto the floor. The archive records all three, along with its own fingerprints on the lens.

Data & methods

Panel. All Wikidata items pointed to by ≥2,000 humans' "educated at" (P69), discovered by aggregate query 2026-07-09 — 250 institutions including secondary schools, academies, conservatories, and sub-entities of collegiate universities (Harvard's schools, Oxbridge colleges are separate items; we do not merge them).
People. 1,203,511 person-institution records; 1,053,018 distinct humans with gender (P21), birth year (P569) and sitelink counts. A person educated at several panel institutions appears under each.
Fields. 13 buckets over English occupation labels (P106), keyword rules in scripts/wikidata_aggregates.py; a person counts once per bucket per institution. Occupations with <25 uses panel-wide are unlabeled and excluded (~6% of pairs).
What "notable" means. Having a Wikidata entry — a joint product of achievement, documentation, and editor attention. Mass imports (Dorpat's matriculation registers, academic-paper author databases) inflate some institutions and the science bucket generally. English-language and Western coverage is densest.
Gender shares use recorded male/female; other recorded genders (~0.2%) are in totals, and people missing P21 or P569 are excluded from cohort charts — exclusions that fall hardest on older, less-documented alumni.
Birth cohort ≈ enrollment + ~19 years. Coeducation ticks in Figure 3 are placed at (admission year − 19), an approximation.
Post-1985 cohorts are incomplete: notability accrues with career age, so recent cohorts over-represent early-fame fields (athletes, performers).