The Spectacular Failure of the Endowment Model

Ennis furthermore says, “Moreover, they have underperformed with remarkable consistency. Their poor performance is no accident. Rather, it is structural in nature.”

The public pension funds – an asset pool of $4.5 trillion – underperformed by an average of 1.52% per year. Had they not underperformed and instead been invested passively in an index fund, those funds would have accumulated almost 20% more money than they have now.

The endowment funds – an asset pool of about $600 billion – underperformed worse, by an average of 1.87% a year.

In both cases, the underperformance was equal to their fees. This equivalence is unsurprising when one considers the nature of the funds’ performance.

But the performance was worse than what was reported. There is a convention among pensions and endowments to report performance gross of internal management fees. For example, in the case of Yale, the costs incurred by Swensen and his team of analysts and researchers are not deducted from its reported performance.¹ Based on its most recent IRS form 990, such expenses for Yale were approximately $90 million, which works out to about 34 basis points annually on its AUM.

They are running closet index funds

The performance of these funds over time is statistically indistinguishable from that of a combination of two or three market index funds – a U.S. stock market index fund, a bond index fund, and an international equity index fund – less the fees paid.

Mr. Ennis discovered that fact by performing a “style analysis,” of the type that has been applied widely since it was proposed by William F. Sharpe in 1988.

Style analysis fits the performance over time of an allocation to investment “style” categories to the actual performance of a fund over time (in this case, the “style” categories are U.S. equities, bonds, and international equities). It thus infers the allocation of the fund to those styles.

Ennis’s result is that if the allocation had been made to the three index funds with allocations matching the style analysis, the performance would have precisely matched the performance of the pension funds (gross of fees). The fit had an r-squared of 0.991 (i.e., 99% of the variation in the performance of the pension funds was explained by the index fund). The pension fund beta relative to the benchmark was 0.999.

In short, the funds could have realized the same performance for a fee, not of 1.52% or 1.87%, but of about one basis point a year (for the large funds; smaller funds would have to pay slightly more). That cost of one basis point is less than one-one-hundred-and-fiftieth of what they actually did pay.

But Yale does it

Mr. Ennis says, referring to endowment funds, “…it is important to acknowledge that there exists a long-standing ethos within the endowment management community in the U.S. to the effect that these institutions are, or should be, savvy investors. I won’t begin to attempt to describe this ethos further; nor will I attempt to justify or critique it. Suffice it to say the phenomenon is real: Trustees of private universities with large endowments invariably believe that they can and should do better than average. In their eyes, this requires an active approach to investing.”

I saw this ethos up close 20 years ago when I chaired the investment committee of a nonprofit organization and was charged with advising on the investment of its small quasi-endowment.

I had successfully advocated index fund investing when three other board members who were power players – an attorney with a top Washington DC law firm, the CEO of a large oil company, and a major donor – started to press for investing in hedge funds. When I asked why, the attorney simply said, “Yale does it.”

Yes, Yale’s endowment fund, managed by Swensen, had seen market-beating growth for many years, mostly because of its allocation to alternative investments like timberland, hedge funds, and private equity.

But that didn’t mean that merely investing in hedge funds – any hedge funds (which is what the power players advocated) – was likely to produce good results going forward for everyone, or even for David Swensen himself (it didn’t).

Nevertheless, the drive to imitate Swensen was overwhelming, even if it became senseless in its particulars.

The “endowment model”

Major institutional investment consultants turned the fascination and envy over Swensen’s results into what they thought or pretended was an “expert system”: the “endowment model.” An “expert system” is an attempt to embody in an artificial intelligence (AI) program whatever it is that an expert does. This is done by observing what the expert does or questioning her about it, then modeling it as a kind of algorithm, so that it can be applied by those who don’t have access to the expert.

My knowledge of this algorithm comes largely from reading a 2010 book about it, The Endowment Model of Investing: Return, Risk, and Diversification, by Martin L. Leibowitz, Anthony Bova (both of Morgan Stanley) and P. Brett Hammond (of TIAA-CREF). I wrote a scathing review of the book that July. (See also a related article here.) I won’t go into detail in this article about what I wrote, but suffice it to quote one sentence: “The book is a lengthy and repetitious exposition of a worthless, fraudulent pseudoscience.”

The point of the book was to pretend to use a Markowitz-type asset allocation model to force a high (but not too high) allocation to “alternatives” such as hedge funds and private equity. The book said that the model had to be “tortured” to make it do this.

This was a deeply flawed “expert system.” Its whole point was to mimic the allocations to alternatives as “asset classes” that Swensen had used, without any analysis or consideration of what exactly was in those asset classes. You can be sure this is not what Swensen did.

How has the endowment model been applied?

Ennis describes what has happened because of the purveyance of this endowment model. Allocations by endowments and pensions to hedge funds, private equity and real estate increased enormously, from almost nothing in the 1990s and early 2000s to as much as half the portfolio or more.

Those allocations may be prescribed by a mean-variance model. The model is forced to set the allocations high by feeding high expected returns for those asset classes into the model. The model is also forced not to set them too high by the imposition of maximum percentage constraints on their allocations – otherwise, because of their high return expectations, the model would allocate 100% of the portfolio to them. The result is that the model drives the allocations to alternative investments to their preset maximum allocation constraints, making use of the model gratuitous.

Once these allocations to those “asset class” silos are set, the administrators of the fund are required to hire managers for each of these asset classes to fill out the quota. It doesn’t matter if, for example, no good hedge funds or private equity funds can be found at a reasonable cost, or not enough of them. They must be hired to fill out the quota anyway.

A result has been the explosion of hedge funds and private equity funds. Ennis points out that “Hedge fund assets under management in 1997 totaled approximately $118 billion. The figure grew 27-fold to $3.2 trillion in 2018,” and, “The number of private equity firms active grew more than tenfold between 1995 and 2018, from fewer than 1,000 to roughly 10,000.”

And yet even now, hedge fund assets comprise only $3.2 trillion and private equity $3.4 trillion, compared to a total value of global stocks and bonds of $180 trillion.

How can such an overallocation to a small niche market not affect its pricing, and move it back toward market efficiency – or beyond?

Some practitioners of the endowment model do not use mean-variance optimization or other mathematical models to set a target asset allocation. They may rely on a consultant or an internal committee to set those allocations. But there is no evidence that they are any better at selecting investments that will outperform, and are burdened by a market that has become vastly more efficient over time.

Results

The results experienced by these funds represent the inevitable triumph of the efficient market model over the endowment model.

The purveyors of the endowment model ignored two almost universally valid maxims: (1) “Past performance is not a predictor of future performance”; and (2) the Schwert Rule (enunciated by University of Rochester Distinguished University Professor G. William Schwert in 2003), “After they are documented and analyzed in the academic literature, anomalies often seem to disappear, reverse, or attenuate.”

Trying to replicate on a massive and automated scale the performance of Yale’s endowment fund under David Swensen and a very few others in what Ennis calls the “Golden Age of Alternative Investments” from 1994 to 2008, who went boldly into a small niche market and selected their investments very carefully, was always a fool’s errand. But it has made a lot of money for hedge fund and equity fund managers and the consultants who recommend them. Yale’s performance was outstanding during the Golden Age. However, since that time it has underperformed the market, as this article shows.

I don’t see the blind allegiance to the endowment model ending any time soon, in spite of its well-documented failure.

Economist and mathematician Michael Edesess is adjunct associate professor and visiting faculty at the Hong Kong University of Science and Technology, chief investment strategist of Compendium Finance, adviser to mobile financial planning software company Plynty, managing partner and special advisor at M1K LLC, and a research associate of the Edhec-Risk Institute. In 2007, he authored a book about the investment services industry titled The Big Investment Lie, published by Berrett-Koehler. His new book, The Three Simple Rules of Investing, co-authored with Kwok L. Tsui, Carol Fabbri and George Peacock, was published by Berrett-Koehler in June 2014.

¹ For those who want to be wonky, the reported performance is not GIPS compliant, since it does not represent the net returns to investors.

Read more articles by Michael Edesess