Version 2.1, published Oct 21, 2026: 5 criteria, 3 sources, 1 of the 5 measures backed by peer-reviewed research.
David Swensen's philosophy, applied to ETFs
In the tradition of David Swensen, this seat reads 5 criteria from ETFIQ data, version 2.1 of Oct 21, 2026. How many real bets do you own?
Bio
David F. Swensen was born in 1954. He graduated from the University of Wisconsin-River Falls and came to Yale as a graduate student in 1975. He earned his PhD at Yale in 1980, advised by the Nobel laureate James Tobin and by William Brainard.
Before returning to Yale he worked on Wall Street, at Salomon Brothers and Lehman Brothers, where Yale credits him with pioneering work on interest rate swaps. In 1985 Yale hired him to manage its endowment, which then stood at $1.3 billion. He ran it until his death; by 2021 it exceeded $31 billion.
Yale reported that over his 35 years, measured through June 30, 2020, the endowment earned 13.1% a year. That was 3.4 percentage points a year better than the mean of the endowments tracked by Cambridge Associates, and 4.3 points a year better than a traditional portfolio of 60% stocks and 40% bonds. Yale credits him with $45.6 billion of gains in dollar terms, including $36.0 billion of value added against the Cambridge Associates mean. Yale Investments describes the Yale Model as a strategy pioneered by Swensen and Dean Takahashi.
He taught throughout his career, co-teaching the seminar "Investment Analysis" with Takahashi for more than two decades, in Yale College and the School of Management. In 2014 Yale added the words "and teacher" to the citation of his honorary degree. He wrote two books: *Pioneering Portfolio Management* (2000) for institutions and *Unconventional Success* (2005) for individuals.
He died in New Haven on May 5, 2021, aged 67, after a long illness.
Philosophy
Swensen's investing rests on two principles: lean toward equity-like assets, and diversify. In *Pioneering Portfolio Management* he described diversification as the investor's free lunch, because spreading money across assets that behave differently lowers risk without giving up expected return.
For Yale he applied that principle through what Yale Investments calls the Yale Model: a long-term approach built on partnerships with outside investment managers, producing a portfolio that is equity-oriented and diversified across asset classes. Those partnerships, with outside managers, are not something an individual saver can buy.
His second book is the one that matters most for ETF investors, because it is written for individuals and sets out a simpler path. *Unconventional Success* argues that individuals face a for-profit fund industry whose interests conflict with theirs, and that they should therefore build simple portfolios of low-cost, passively managed funds.
The book sets out six core asset classes: domestic stocks, foreign developed-market stocks, emerging-market stocks, real estate, US Treasury bonds and Treasury inflation-protected securities. To count as core, an asset class has to pass three tests: it adds a basic, valuable and distinct quality to a portfolio, such as high expected return or low correlation with other holdings; its returns come from the market rather than from a manager's skill; and it trades in broad, deep, investable markets. His model portfolio puts 30% in domestic stocks, 15% in foreign developed stocks, 5% in emerging-market stocks, 20% in real estate, 15% in Treasury bonds and 15% in inflation-protected Treasuries. Asset classes that fail at least one of the tests are non-core and left out of the model.
He paired that with discipline. Part two of the book deals with market timing, in two chapters, one titled "Chasing Performance" and the other "Rebalancing": the discipline of bringing a portfolio back to its targets rather than following what has recently done well.
On funds he was blunt. Part three's chapters document the performance deficit of actively managed mutual funds and the reasons for it. Chapter 11 is titled "The Exchange-Traded Fund Alternative," and the book's afterword is titled "Failure of For-Profit Mutual Funds." Not-for-profit fund managers, in his view, served investors best because no profit motive stood between the manager and the investor.
The two books apply the same principles to different investors. An endowment with a long horizon, a skilled staff and access to outside managers can build the Yale Model; the saver he wrote for in 2005 gets six low-cost core funds instead.
Key works
- David F. Swensen, *Pioneering Portfolio Management: An Unconventional Approach to Institutional Investment* (Free Press, 2000; revised edition 2009). - David F. Swensen, *Unconventional Success: A Fundamental Approach to Personal Investment* (Free Press, 2005).
What we cannot know
Yale's results reflect private funds and outside managers that are not public, so the part of the 13.1% that came from asset allocation rather than from manager selection and access cannot be measured from Yale's public figures.
Source quality
Strong. Both of his books, with the contents and the model portfolio confirmed, and Yale's own published account of his record; the one gap is that two quote pages were not pinned.
In his words
“The act of diversification provides a free lunch of enhanced returns and reduced risk…”
The Swensen-style Diversification seat
Swensen-style Diversification is one seat of the IQ Investment Committee. It applies David Swensen's own books to ETFIQ data.
What the investor published
In ETFIQ’s words, with the source of each. Sourced to David Swensen's own books.
Diversification lowers risk without giving up expected return, which makes it one of the few free lunches in investing. (David F. Swensen, Pioneering Portfolio Management: An Unconventional Approach to Institutional Investment (Free Press, first edition, 2000), p. 67)
A portfolio rests on six core asset classes: US stocks, foreign developed stocks, emerging market stocks, real estate, US Treasury bonds and TIPS. (David F. Swensen, Unconventional Success: A Fundamental Approach to Personal Investment (Free Press, first edition, 2005), Introduction, p. 34 (Table I.1))
Each asset class needs enough weight to matter, at least 5 to 10 percent, and none should be large enough to do damage, at most 25 to 30 percent. His model mix runs from 5% to 30%. (David F. Swensen, Unconventional Success: A Fundamental Approach to Personal Investment (Free Press, first edition, 2005), chapter 3, p. 83; Table I.1, p. 34, repeated p. 84)
Narrow niche products, single countries and sectors among them, expose investors to risks they should not take; low-cost broad index funds are the answer. (David F. Swensen, Unconventional Success: A Fundamental Approach to Personal Investment (Free Press, first edition, 2005), chapter 11, pp. 314-315)
Research behind the measures
Peer-reviewed research on how each measure is taken. It never adds a criterion, and where it weakens one the criterion says so.
Itzhak Ben-David, Francesco Franzoni, Byungwook Kim and Rabih Moussawi (2023), Competition for Attention in the ETF Space, The Review of Financial Studies 36(3), 987-1042 (doi:10.1093/rfs/hhac048). Specialised ETFs, which track niche portfolios and charge higher fees, lost about 30% risk-adjusted over their first five years, driven by the overvaluation of their stocks at launch, unlike the early broad-index ETFs.
Harry Markowitz (1952), Portfolio Selection, The Journal of Finance 7(1), 77-91 (doi:10.1111/j.1540-6261.1952.tb01525.x). Choosing only for the highest expected return never favours a diversified portfolio. Treating the variance of the whole portfolio's return as a cost, alongside expected return as a benefit, makes diversification rational.
Gary P. Brinson, L. Randolph Hood and Gilbert L. Beebower (1986), Determinants of Portfolio Performance, Financial Analysts Journal 42(4), 39-44 (doi:10.2469/faj.v42.n4.39). Across 91 large US pension plans from 1974 to 1983, the policy mix of asset classes explained on average 93.6% of the variation of each plan's quarterly returns over time; that is the share of ups and downs, not of the level of returns or of the differences between plans.
Compare and model modes
In compare mode funds are ordered by the seat’s score, and ties break on the effective number of holdings, highest first.
In model mode: Fill the six core asset classes within Swensen's bounds, 5% to 30% each, with the broadest fund in each.
How the seat is tested
The cases this seat must pass before it ships, with the result each must give.
A global total-market fund: Supports. Thousands of companies, none above 5%.
A broad fund with one 6% position: Supports (The effective number of holdings: Supports, One of Swensen's six core asset classes, owned whole: Supports). A core asset class held whole, with the weight spread like 120 equal positions: one name a little above 5% does not make it concentrated.
A fund with 40% of its weight in 10 names: Supports (The weight in the ten largest positions: Challenges, One of Swensen's six core asset classes, owned whole: Supports). Ten names carry 40%, which the seat states, but it is a core asset class spread like 35 equal positions: an ETFIQ threshold never decides the seat alone.
A fund over 25% in one company: Challenges (The effective number of holdings: Challenges, One of Swensen's six core asset classes, owned whole: Challenges). A narrow slice with a quarter in one company: few real bets.
A broad sector fund: Challenges (Not a narrow niche product: Challenges). Spread over many companies, but one sector: the narrow niche product Swensen warned against (pp. 314-315).
Swensen's own model mix as a portfolio: Supports (Swensen's bounds on each asset class: Supports). 30% US stocks, 15% foreign developed, 5% emerging, 20% real estate, 15% Treasuries, 15% TIPS: inside his own bounds.
A two-fund portfolio, 70% US stocks and 30% bonds: Challenges (Swensen's bounds on each asset class: Challenges). Many holdings, but 70% in one asset class, more than twice Swensen's ceiling.
Three asset classes at 30%, 30% and 40%: Challenges (Swensen's bounds on each asset class: Challenges). Three of the six core classes missing, and one above 30%.
VT: Supports. About 10,000 companies worldwide, top ten 22.5%. Figures read on Oct 22, 2026.
SMH: Challenges. 25 semiconductor companies, top ten 68.1%. Figures read on Oct 22, 2026.
AOR: any result but Not enough data (Swensen's bounds on each asset class: Challenges). A fund of seven funds: read through, it is not seven bets, but its US stocks sleeve (38.57% of the book) is above Swensen's 30% ceiling. Figures read on Oct 22, 2026.
BND: any result but Supports (One of Swensen's six core asset classes, owned whole: Challenges). A total bond fund holding corporate and mortgage bonds beside Treasuries: outside Swensen's core classes, which take Treasuries and TIPS only.
Changes to this seat
Version 1.0, Oct 21, 2026: First version.
Version 2.0, Oct 21, 2026: The 1940 Act and tax-code tests moved to the ETFIQ Skeptic, since Swensen did not frame diversification through them; a required criterion reads whether a fund owns a core asset class whole or a narrow niche (chapter 11); the effective-holdings lines are 30 and 15, after SSGA's own figures, so a total-market fund can Support; the asset-class test needs all six classes at 5% or more.
Version 2.1, Oct 21, 2026: Owning a core class whole is read from the book, never the segment label; the classes Swensen sets outside his six (corporate and other non-core bonds, chapter 4) read against the core criterion; the niche line is its own gate.
Where to next
Scope
Every ETF with a filed book, and every portfolio. A fund of funds is read through to what its funds own, so its issuers are the holder's exposure.
Source strength: Strong: the investor’s own books, letters or memos. Swensen's own two books. His asset-class bounds are used as thresholds for a portfolio; inside one fund, where he gave no number, the seat uses ETFIQ's figures, each labelled. The two tests US law sets for a diversified fund are not Swensen's, so they moved to the ETFIQ Skeptic.
Lenses apply publicly known investing principles to ETFIQ data. No investor named is affiliated with or endorses ETFIQ. Not investment advice.
Sources
The published work each criterion is sourced to
| Source | What it provides | Read | Spec |
|---|---|---|---|
| David F. Swensen, Unconventional Success: A Fundamental Approach to Personal Investment | Diversification as a core principle, the six core asset classes and their weights (Table I.1, p. 34), the model portfolio and the bounds on each asset class (chapter 3, Table 3.1), and the warning against narrow niche ETFs (chapter 11) | Free Press, first edition, 2005 | document ↗ |
| David F. Swensen, Pioneering Portfolio Management: An Unconventional Approach to Institutional Investment | Diversification as the free lunch of investing: risk falls without giving up expected return | Free Press, first edition, 2000 | document ↗ |
| State Street Global Advisors (Haselgard and Crasta), Concentration is up; diversification isn't gone | Context for the measure, not Swensen: the effective number of holdings, 1 over the sum of squared weights, about 50 for the S&P 500 today against about 120 before the pandemic | State Street Global Advisors, Sept 17, 2026, 2026 | document ↗ |
Definitions and method
Does the fund own one of Swensen's six core asset classes whole, or a class he sets outside them? Measure: core or niche. core where the fund owns one of Swensen's six core asset classes whole, read from its book: for US stocks, market weights (the weighting measure's test, from the fund's own N-PORT) over companies making up at least 80% of the US market's value, or, where those are not read, holdings carrying at least 80% of VTI's weight in proportion to it; for other stocks, holdings carrying at least 80% of the broad cap-weighted fund of the fund's own house for its region in proportion to it (VEA, VWO, VT or VNQ where the house has none), so a momentum, quality, minimum-volatility or price-weighted fund is a slice; for Treasuries and TIPS, at least 90% of the book in US Treasury paper. partial for a slice or a tilt of a core class. noncore for the classes Swensen sets outside his six (corporate, high-yield, tax-exempt, foreign and asset-backed bonds, and bond funds that mix them in). niche for a sector, a single country or narrow region, or any fund on the themes, leverage, crypto, buffer, income or alternatives desks. other for the rest. Null where the book cannot be compared, or the reference prints its weights to two decimals. Supports where core or niche equal to core. Challenges where core or niche one of noncore, niche. Mixed between. Threshold: ETFIQ’s figure, because the investor published none. The six classes are Swensen's (Table I.1, p. 34), and so are the classes he sets outside them: corporate, high-yield, tax-exempt, foreign and asset-backed bonds have "unattractive characteristics" (chapter 4, "Non-Core Asset Classes", p. 92; corporate bonds p. 103). Owning a class whole is read from the book, never from the fund's label: for stocks, the book must carry at least 80% of a broad cap-weighted reference's weight in proportion to it. The reference funds, 80% and the proportionality test are ETFIQ's; a slice or a tilt reads Mixed. Weight: 35%; required. Applies to: what is read (a fund or a portfolio) equal to fund and the category not allocation. Sourced to: David F. Swensen, Unconventional Success: A Fundamental Approach to Personal Investment (Free Press, first edition, 2005), Introduction, p. 34 (Table I.1); David F. Swensen, Unconventional Success: A Fundamental Approach to Personal Investment (Free Press, first edition, 2005), chapter 4, pp. 92 and 103. Research: none cited; the criterion rests on the source alone.
Is the fund a single sector, a single country or another narrow product? Measure: core or niche. core where the fund owns one of Swensen's six core asset classes whole, read from its book: for US stocks, market weights (the weighting measure's test, from the fund's own N-PORT) over companies making up at least 80% of the US market's value, or, where those are not read, holdings carrying at least 80% of VTI's weight in proportion to it; for other stocks, holdings carrying at least 80% of the broad cap-weighted fund of the fund's own house for its region in proportion to it (VEA, VWO, VT or VNQ where the house has none), so a momentum, quality, minimum-volatility or price-weighted fund is a slice; for Treasuries and TIPS, at least 90% of the book in US Treasury paper. partial for a slice or a tilt of a core class. noncore for the classes Swensen sets outside his six (corporate, high-yield, tax-exempt, foreign and asset-backed bonds, and bond funds that mix them in). niche for a sector, a single country or narrow region, or any fund on the themes, leverage, crypto, buffer, income or alternatives desks. other for the rest. Null where the book cannot be compared, or the reference prints its weights to two decimals. Supports where core or niche one of core, partial, noncore, other. Challenges where core or niche equal to niche. Mixed between. Threshold: ETFIQ’s figure, because the investor published none. Swensen's own warning: narrow niche products, individual country and sector funds among them, clutter the ETF landscape (pp. 314-315), and sensible investors avoid the vast majority of ETF offerings (p. 329). Which segments are niches is ETFIQ's mapping. It decides the seat alone, and weighs little otherwise because its force is that veto. Weight: 5%; a Challenge here makes the seat Challenge. Applies to: what is read (a fund or a portfolio) equal to fund and the category not allocation. Sourced to: David F. Swensen, Unconventional Success: A Fundamental Approach to Personal Investment (Free Press, first edition, 2005), chapter 11, pp. 314-315 and 329. Research, supports the measure: Itzhak Ben-David, Francesco Franzoni, Byungwook Kim and Rabih Moussawi (2023), Competition for Attention in the ETF Space, The Review of Financial Studies 36(3), 987-1042. Specialised ETFs, the narrow niche products Swensen warned against, lost about 30% risk-adjusted over their first five years.
How much of the fund rides on ten names? Measure: the ten largest issuers. Weight of the ten largest issuers in the look-through book, share classes of one issuer combined, percent of net assets. Cash rows, and US government and agency debt (never a foreign government's), are left out, as the 1940 Act's diversification test leaves government securities out. A fund of funds is read through: each underlying fund's book replaces its row, scaled by its weight. Null where fewer than ten issuers remain, where government debt is more than 75% of the book, or where the book's positive weights sum to more than 105% (a book of swaps and options). Supports where the ten largest issuers below 25%. Challenges where the ten largest issuers at or above 40%. Mixed between. Threshold: ETFIQ’s figure, because the investor published none. ETFIQ threshold. Swensen sets bounds for asset classes, not for positions, so under 25% and 40% or more in ten names are ETFIQ's, weighted like the seat's other criteria and never deciding it alone. The S&P 500's ten largest companies were about 40% of it in October 2026 (40.65% of VOO by company, Sep 30 book): a finding the seat states, not a verdict. Weight: 15%. Sourced to: David F. Swensen, Unconventional Success: A Fundamental Approach to Personal Investment (Free Press, first edition, 2005), chapter 3, p. 83; David F. Swensen, Pioneering Portfolio Management: An Unconventional Approach to Institutional Investment (Free Press, first edition, 2000), p. 67. Research, context, not a test of the measure: Harry Markowitz (1952), Portfolio Selection, The Journal of Finance 7(1), 77-91. Diversification is judged by the variance of the whole portfolio; a weight count like this one is a proxy that ignores how positions move together.
How many equal-sized bets is the fund really making? Measure: the effective number of holdings. 1 divided by the sum of squared weights over the same issuers as the ten-largest measure, the weights rescaled to sum to 1 (the inverse Herfindahl index). Null on the same terms as the ten-largest measure. Supports where the effective number of holdings at or above 30. Challenges where the effective number of holdings at or below 15. Mixed between. Threshold: ETFIQ’s figure, because the investor published none. Swensen gives no count. 30 and 15 are ETFIQ's, set against SSGA's rule of thumb that 25 to 30 stocks remove most of the volatility diversification can, and its figure of about 50 effective holdings for the S&P 500 in 2026 (about 120 before the pandemic). Version 1.0's 100 left even VTI reading Mixed. The S&P 500 reads about 41 here against SSGA's "about 50", partly because ETFIQ combines share classes (GOOGL with GOOG) into one issuer; the rest of the gap is not measured. Weight: 25%. Sourced to: David F. Swensen, Pioneering Portfolio Management: An Unconventional Approach to Institutional Investment (Free Press, first edition, 2000), p. 67. Research, context, not a test of the measure: Harry Markowitz (1952), Portfolio Selection, The Journal of Finance 7(1), 77-91. The inverse of the sum of squared weights is a concentration count, not the portfolio variance Markowitz uses, so ten stocks that move together still count as ten.
For a portfolio or an allocation fund: does any asset class carry more than Swensen allowed? Measure: the largest core asset class. The largest share of the look-through book in any one of Swensen's six core asset classes: US stocks, foreign developed stocks, emerging market stocks, real estate (REITs), US Treasury bonds and Treasury inflation-protected bonds (TIPS). Weight outside the six (corporate credit, commodities, cash) is in no class. Supports where the largest core asset class at or below 30% and the smallest core asset class at or above 5%. Challenges where the largest core asset class above 30% or the smallest core asset class below 5%. Mixed between. Threshold: the investor’s own figure (David F. Swensen, Unconventional Success: A Fundamental Approach to Personal Investment (Free Press, first edition, 2005), chapter 3, p. 83). The ends of Swensen's own ranges (each class at least 5 to 10 percent and at most 25 to 30 percent, pp. 33, 34 and 83), read at their loose ends: all six classes held, none above 30% and none below 5%. Version 1.0 counted only the classes held at all, so three classes at 30% each could Support. Weight: 20%; a Challenge here makes the seat Challenge. Applies to: what is read (a fund or a portfolio) equal to portfolio or the category equal to allocation. Sourced to: David F. Swensen, Unconventional Success: A Fundamental Approach to Personal Investment (Free Press, first edition, 2005), chapter 3, p. 83; Table I.1, p. 34, repeated p. 84. Research, context, not a test of the measure: Gary P. Brinson, L. Randolph Hood and Gilbert L. Beebower (1986), Determinants of Portfolio Performance, Financial Analysts Journal 42(4), 39-44. The policy mix of asset classes explained most of how pension plans' returns moved over time, which is why the seat reads asset classes; the paper sets no bounds, and its 93.6% is never quoted as support for one.
Each criterion reads one ETFIQ measure against two thresholds. Each criterion reads one measure against the seat's two thresholds: Supports where the supports condition holds, Challenges where the challenges condition holds, Mixed between them, Not enough data where the measure is missing. A criterion whose scope does not cover the subject is left out, and a seat whose scope does not cover it does not apply. The seat's score is the weighted mean of its readable criteria at +1, 0 and -1. A criterion marked veto that Challenges makes the seat Challenge, whatever the rest reads or could not read. Otherwise the seat reads Not enough data when the criteria it could read carry less than half of its weight, or when a criterion marked required cannot be read; it Supports at 0.5 or more and Challenges at -0.5 or less, and is Mixed between. A criterion that caps the seat keeps it from Supporting when it Challenges. A criterion the seat needs keeps it from Supporting unless it reads Supports itself. A seat that reads a fund's holdings states its own rule on its page. The weights. The core class carries the most (35%) because it is Swensen's first rule; effective holdings 25%; the asset-class bounds 20% for a portfolio; the top ten 15%, an ETFIQ line he did not draw; the niche line 5%, because its force is its veto. The weights are ETFIQ's and were not set to land any fund.
What this cannot tell you
How to check any of this
This seat’s criteria, thresholds, weights and cases are in the committee file, the same file the committee reads.
Cite this pageCitation, APA and Markdown
ETFIQ, David Swensen's philosophy, applied to ETFs, data as of Oct 21, 2026. https://etfiq.com/investment-committee/david-swensen/
ETFIQ. (Oct 21, 2026). David Swensen's philosophy, applied to ETFs. Retrieved from https://etfiq.com/investment-committee/david-swensen/
[David Swensen's philosophy, applied to ETFs (ETFIQ, Oct 21, 2026)](https://etfiq.com/investment-committee/david-swensen/)
Free to use with attribution for figures ETFIQ computes; issuer data stays under its owner's terms. The underlying files are at Open data.