Posts
How We Did This
The Methodology Behind THE UNIVERSITY ENDOWMENT MACHINE — What We Found, What Surprised Us, What We Almost Got Wrong, and What Comes Next
THE UNIVERSITY ENDOWMENT MACHINE — Series Capstone C | February 2026
How the Series Started
THE UNIVERSITY ENDOWMENT MACHINE was not the original plan for Series 3. After completing THE ENDLESS FRONTIER (200 years of the same extraction mechanism across railroads, oil, defense, and the internet), we surveyed four potential next topics: healthcare private equity, university endowments, pharmacy benefit managers, and carbon offsets.
The endowment won — not because the financial numbers were largest, but because of a structural observation that made it different from the others: this was the institution that trains the extractors documented in the first two series. Harvard Law trains the lawyers structuring stadium deals. Harvard Business School trains the PE executives buying hospitals. Harvard Kennedy School trains the policy advocates protecting tax exemptions. The university endowment isn't a parallel extraction system. It's the institution upstream of every other extraction system we'd documented.
That thesis — the university endowment as the intellectual and financial engine of the broader extraction architecture — drove every post. We tested it against the documents at every step. The documents confirmed it more completely than we expected.
The Research Process
STEP 1 — THESIS FIRST:
Randy identified the directional question for each post. Not “what happened” but “what does the structure reveal.” The question came before the research. The research tested whether the documents supported it.
STEP 2 — PRIMARY SOURCES ONLY:
Every search targeted primary documents: annual reports, court filings, official statements, the institutions’ own websites, government agency determinations. Secondary sources (journalism, analysis) were used to identify primary documents — never as the primary source themselves.
STEP 3 — THE COUNTERARGUMENT FIRST:
Before writing any post, we documented the strongest case for the subject. What did Harvard actually fund? What did Swensen actually believe? What did Yale actually accomplish? The counterargument informed the structural argument — and made it stronger.
STEP 4 — SHOW THE RECEIPTS:
No claim without a sourced document. Every smoking gun is quoted directly from the primary source it came from. Every number is attributed. Every “documented” claim is documented. Every “alleged” claim is labeled as such.
STEP 5 — LABEL WHAT WE DON’T KNOW:
When we didn’t have a primary source, we said so. When we had an inference, we labeled it an inference. When a claim was disputed, we presented the dispute. The standard: a reader should be able to distinguish what is documented from what is argued.
What Surprised Us
The most striking finding of the series: the most damning evidence in almost every post came from the institutions' own documents. Yale's own School of Management published the alumni network by name. Blackstone's own website listed Harvard Management Company board service alongside Blackstone board service for the same individual. Yale's own provost confirmed that the excise tax exceeds the entire undergraduate financial aid budget — while arguing against the tax. Harvard's own CEO confirmed the paper wealth trap in his annual letter. Harvard's own ReVista journal confirmed the Brazilian court order against its subsidiary.
We expected to find evidence despite institutional opacity. We found evidence through institutional transparency. The machine doesn't hide because it doesn't think it needs to. That confidence is itself a structural feature.
In "Pioneering Portfolio Management" (2000), Swensen explicitly warned that institutions without Yale's scale, talent, and relationships should not attempt to replicate the model. He said smaller endowments were likely to pay PE fees without capturing the returns that justified them. David Salem reviewed the book in Barron's and amplified the warning.
By 2024, the average university held 56% of its endowment in alternative investments. Every major institution copied the model Swensen warned them not to copy. The warning was published. It was widely read. It was ignored by hundreds of institutions over 24 years. The model spread not because of its documented success — but because of the personnel network through which it traveled. People trusted the people Swensen trained more than they trusted Swensen's written caution.
Yale's bankers at Evercore chose the code name "Project Gatsby." Nick Carraway — F. Scott Fitzgerald's narrator, the man who watched Gatsby reach toward the green light — was a Yale graduate. The literary allusion required someone at Evercore to know both Fitzgerald's novel and Yale's connection to it. The name was not accidental. Someone named a $6 billion fire sale after a novel about the impossibility of reaching what you've built. That is either the most self-aware moment in the history of institutional investing, or the most ironic. Either way: it named itself perfectly.
In 1934 — 91 years ago — Fortune magazine noted the "embarrassingly large surplus piling up in the school's coffers" of the Milton Hershey Trust. Not the endowment machine. A different trust. But the same mechanism, the same surplus, the same gap between accumulated wealth and mission fulfillment — documented in 1934. The problem we're investigating in the university endowment series has been visible in the philanthropic trust world for nearly a century. This discovery opened the door to what may become Series 4.
What We Almost Got Wrong
The 0.69% effective tax rate figure required careful framing. The 1.4% excise tax applies to net investment income — realized gains, dividends, interest — not total returns. Because 80% of Harvard's endowment is illiquid PE (unrealized gains not yet taxed), the taxable income base is significantly lower than the $5.8 billion headline return. The $40 million estimated tax divided by $5.8 billion total returns produces 0.69% — an accurate illustration of the effective rate relative to total economic returns, but not a rate applied to a single taxable base.
We kept the figure but flagged the distinction explicitly in Post 5. The illustration is accurate. The mechanism requires explanation. We provided both.
The Harvard Brazilian farmland story required careful sourcing because multiple claims float in the public record without primary source documentation. We restricted the post to four primary sources: GRAIN's investigation (using Harvard's own tax filings and Brazilian corporate records), Bloomberg's independent confirmation, Harvard's own ReVista journal (confirming the Bahia state court order), and ActionAid USA (sourcing to INCRA's formal findings).
We did not claim Harvard "knew" the land was illegally titled. We documented that Brazilian courts and INCRA found titles illegal — and that Harvard's response was to decline comment. The distinction between documented facts and intent matters. We maintained it throughout.
Early drafts of Post 7 framed Project Gatsby as a collapse of the Yale Model. The documents didn't support that framing. Yale's endowment grew from $41.4B to $44.1B in FY2025. Its 10-year return of 9.5% still outperforms the average endowment. Yale's spokesperson was accurate: the secondary sale is a liquidity adjustment, not a crisis. Howard Bunsis was accurate: an institution with a triple-A bond rating is not in financial danger.
We changed the framing to: the first break in a 40-year record of never selling PE stakes. That's accurate. It's also significant without overstating what it is. The machine is under pressure. It is not collapsing.
The Human-AI Collaboration: What It Actually Looked Like
Identified the series topic from among four candidates (healthcare PE, university endowments, PBMs, carbon offsets)
Set the directional thesis: the university as upstream of all other extraction
Made every editorial judgment: what to include, what to cut, what to flag
Decided when a claim needed more sourcing
Decided when the counterargument wasn't strong enough
Maintained the standard: show the receipts or don't make the claim
Posted every post. Received reader responses. Adjusted emphasis accordingly.
Identified the Milton Hershey Trust as a potential Series 4
Provided the 1909 Trust Deed — a primary source that sharpened the Series 4 thesis immediately
WHAT CLAUDE DID:
Executed web searches targeting primary documents
Synthesized findings across multiple sources
Drafted all eight posts plus three capstones
Maintained sourcing discipline across every post
Flagged when a claim couldn’t be sourced
Distinguished between documented facts and inferences
Maintained consistent template, tone, and methodology across 11 documents
Identified the Project Gatsby literary connection
Built the complete shell company chain from GRAIN’s investigation
WHAT NEITHER OF US DID:
Started with a conclusion and worked backward
Made a claim without a source
Ignored evidence that complicated the argument
Optimized for engagement over accuracy
Presented inference as documented fact
The Standard We Held — And Why It Matters
Every series we've done has investigated institutions that use opacity as a structural feature. NFL owners hide wealth in real estate entities outside public view. Defense contractors bury costs in classified programs. Internet pioneers obscure the public funding behind private innovation. University endowments invest through seven layers of offshore shell companies.
Investigating opacity requires a higher transparency standard than ordinary journalism — not lower. If we investigate institutions for hiding their mechanisms while hiding our own, we've replicated the problem we're documenting.
So we showed the methodology. We named the sources. We labeled the inferences. We documented what we almost got wrong. We presented the counterargument in every post before making the structural argument. We said "alleged" when we didn't have primary source confirmation. We said "documented" when we did.
That standard is not just ethical. It's strategic. The series is stronger because we held it. Every smoking gun hits harder because readers can see exactly where it came from.
What Comes Next: Series 4
The Milton Hershey Trust.
Milton Hershey signed a deed in 1909 with four operative words: "as many as possible." He meant: admit as many poor children to the school as the trust's income permits. The trust now manages $23-24 billion. It admits 2,100 students. Fortune magazine called the surplus "embarrassingly large" in 1934. It has been embarrassingly large for 91 consecutive years.
The 1909 deed is more specific than any university endowment charter. It names a beneficiary class. It mandates scale proportional to income. It leaves no ambiguity about purpose. And it has been selectively honored for 107 years by a board that is simultaneously the board of the trust company managing the assets — paid $112,000-$130,000 annually each, with no retail banking products, for a company whose only function is managing the school trust.
The university endowment machine invests in PE firms that extract from hospitals and farmland. The Hershey Trust extracts from something more specific: the children it was created to serve.
That's Series 4. Same methodology. Same standard. Same template. Different machine — and a more specific victim class than anything we've documented before.
THE LAND GRAB (Series 1): 8 posts — NFL real estate extraction
THE ENDLESS FRONTIER (Series 2): 8 posts — 200 years, one mechanism
THE UNIVERSITY ENDOWMENT MACHINE (Series 3): 8 posts + 3 capstones
TOTAL POSTS: 27
TOTAL WORDS: ~140,000
UNSOURCED CLAIMS: Zero
COUNTERARGUMENTS OMITTED: Zero
FINDING: One
THE FINDING:
The structure matters more than the intentions. Always.
Across railroads, oil, defense, internet, stadiums, and endowments.
Across 200 years. Across 27 posts.
The same mechanism. The same script. The same result.
And the same question waiting at the end of every investigation:
Now that you can see it — what do you do with that?
THE UNIVERSITY ENDOWMENT MACHINE is the third in an ongoing investigation into how public resources generate private wealth through institutional structures that claim public benefit status. The first series (THE LAND GRAB) documented NFL real estate extraction. The second (THE ENDLESS FRONTIER) documented 200 years of the same mechanism across every major American industry. The third documented the institution that trains the people who build and protect those mechanisms.
Series 4 (working title: THE CHOCOLATE MACHINE) will investigate the Milton Hershey Trust — $23 billion, 2,100 students, 107 years of documented surplus, and a 1909 deed that says "as many as possible." The most specific founding mandate of any trust in this investigation. And potentially the clearest documented gap between mandate and practice.
The investigation continues because the mechanism continues. The series ends when the mechanism does.
No comments:
Post a Comment