Most research "rankings" are substantially reputation surveys: people are asked which universities they think do good research, and the answers correlate with which universities they have heard of. There is a more direct approach, which is to count the research.
Rankings measure reputation
Reputation is not worthless. It is just slow, self-reinforcing, and close to impossible to move with anything an institution does in a given decade. It also systematically underrates two kinds of school: the specialist that is excellent in one field, and the institution that got good recently.
Counted output has the opposite properties. It is current, it is checkable, and it does not care whether the person filling in a survey has heard of you.
Three sources that measure output
- An open index of scholarly work gives lifetime works, citations, and an institutional h-index for 2,393 institutions. The h-index is the honest one of the three: an institution with an h-index of 47 has 47 works each cited at least 47 times, so neither one famous paper nor ten thousand ignored ones will move it.
- NIH grant records cover 381 institutions, median $2.84M. Narrow, deep, and the single best read on biomedical strength.
- The NSF research-spending survey covers 543 institutions, median $13.99M. This is money spent rather than work produced, which makes it the useful denominator: output per dollar is a far more interesting question than either number alone.
The distribution is brutal
The median institution has an h-index of 47. The top of the table is above 2,700. That is not a gap you close, and no marketing budget has ever meaningfully moved it.
Research spending concentrates the same way: the top of the table spends $3.80B a year against a sector median of $13.99M.
The median institution's h-index is 47. The top of the table is above 2,700. Competing on 'research excellence' as a general claim is competing in someone else's category.
What this is good for
- Stop making the unqualified claim. "A leading research university" is either verifiable or it is noise, and for most institutions the counted data says noise. The specific claim survives contact with the data. The general one does not.
- Find the field where you are actually top-decile. Institutional totals are dominated by size. Field-level strength is where a mid-sized school genuinely beats famous ones, and it is the only research claim worth putting in front of a prospective graduate student.
- Use spend as the denominator. Output per research dollar is a question the rankings never ask, and it is the one where a well-run institution can win outright.
- Know that AI answer engines read the index, not the brochure. When a model answers "who does good research in X", it is drawing on the indexed record and the public web, not your homepage adjectives.
A note on method: this is an open bibliometric index, not a citation-industry product, and it inherits the usual problems. Author affiliations are noisy, medical-school output can attach to a hospital rather than a parent university, and lifetime totals reward age. Use it to find where you are strong and to sanity-check a claim. Do not use it to build a league table, because the world has enough of those.
The bottom line
You cannot out-market a reputation survey. You can find the field where the counted evidence is on your side, and say that instead.