Published report (AskAmerica connector): http://127.0.0.1:61623/a/17552d7a567bae6e2ff7b3038548c456.html
(This is a loopback link generated by the local connector session — not fetchable from outside that session. The dashboard image is saved alongside this file as dashboard.png.)
Across 575 US research universities in FY2022 (NSF HERD funding-source data joined to NCES IPEDS endowment data on IPEDS UnitID — not just the handful of famous large-endowment schools that dominate press coverage), institutions fund a median of 25% of their own research from institutional funds (mean 28%), while the federal government funds a median of 51% (mean 52%). Both figures track NSF's own published national HERD totals (~25% institutional, ~55% federal for FY2022-23) closely, which cross-validates the sample.
Endowment SIZE is a very weak predictor of that institutional-funding SHARE. The raw-dollar correlation between endowment and % of research self-funded is essentially zero (r = -0.015). In log form it is only r = 0.11 (partial r = 0.10 after controlling for the size of the university's total research program, p = 0.015, n = 575). An OLS regression of institutional-share-of-research on log(endowment) and log(total R&D) has R² = 0.013 — endowment size explains about 1% of the cross-university variance in how self-funded a university's research is. Leave-one-state-out testing keeps the coefficient's direction stable (range 1.00–1.44 pp per log-endowment-unit, no sign flip across 54 states), but dropping New York's institutions alone pushes the p-value to 0.07 — so the direction is real but the statistical significance is not bulletproof.
The confound the political debate usually misses: endowment size correlates strongly with BOTH total research spending (r = 0.56) and federal research dollars (r = 0.56, nearly identical) — bigger, more prestigious universities have bigger endowments and bigger federal grant portfolios for the same underlying reason (institutional scale/prestige), not because one funds the other. A bigger endowment in raw dollars does not predict a smaller federal share.
Illustrative dispersion among the wealthiest schools (FY2022):
| University | Endowment ($B) | Total R&D ($M) | % Institutional | % Federal |
|---|---|---|---|---|
| Harvard | 50.9 | 1,308.5 | 31.8% | 48.0% |
| Yale | 41.4 | 1,191.2 | 28.9% | 57.9% |
| Stanford | 36.3 | 1,384.6 | 12.2% | 62.1% |
| Princeton | 35.1 | 461.2 | 22.3% | 47.4% |
| MIT | 24.6 | 989.2 | 11.8% | 53.7% |
| Penn | 20.7 | 1,791.3 | 31.4% | 48.8% |
| Notre Dame | 17.1 | 280.6 | 33.6% | 44.3% |
| Michigan | 17.1 | 1,770.7 | 34.6% | 54.8% |
| Columbia | 13.3 | 1,230.9 | 9.2% | 73.5% |
| Duke | 12.1 | 1,390.5 | 12.0% | 64.9% |
Columbia self-funds under 10% of research and depends on federal money for 74% of it despite a $13.3B endowment; Notre Dame and Michigan self-fund a third of research with smaller/comparable endowments. This matches independent reporting (Markov Processes International) that Columbia and Cornell are the most federal-grant-dependent large endowments, while Harvard and Princeton are the least.
research.nsf_herd_by_institution's own funding_source='Total' rollup silently excludes Federal dollars — it equals the sum of the five non-federal sources only. Verified nationally: SUM(funding_source=Total) for FY2022 = $43.70B, which exactly equals SUM(nonfederal components), while Federal alone = $53.97B; true total = $97.68B, matching NSF's published FY2022 HERD national total (~$97.8B) almost exactly. Taking the field at face value would understate total R&D by ~45% and badly overstate the institutional/federal funding split. This was filed via report_issue to the connector. All figures in this report use a corrected total (Federal + Institution funds + Business + State/local + Nonprofit + All other).research.nsf_herd_by_institution)edu.ipeds_financials)agent.md — this answerdashboard.png — the published report's dashboard (stat tiles, endowment-vs-share scatter for n=575, top-10-endowment bar chart)