Why the Yale endowment model may still be fundamentally flawed

Brooke’s note: Two years ago, we published a compelling account of how Yale’s famously successful endowment went off the rails and how the “Yale Model” cost people and institutions billions of dollars. See: The Yale Endowment Model of Investing Is Not Dead. The white paper, which first appeared in IMCA’s Investments & Wealth Monitor, became the most-read investment strategy piece RIABiz has ever posted. Financial advisor and Harvard economics graduate Timothy Keating detailed how and why the seemingly invincible model took a catastrophic nosedive the wake of the 2008 crash. But Keating argued persuasively, in that piece and in follow-up article a year later (How the Harvard and Yale endowment models changed to avoid a repeat of 2009), that despite the giant paper losses sustained during the depths of 2009, the endowment model — based on modern portfolio theory — was, in fact, sound. The culprit, he argued, was a lack of liquidity and a failure to model for truly extreme events. Now, we are pleased to present an impassioned counterargument from another erudite financial advisor — and Harvard alum — Robert Boslego. Reports of MPT’s death, he says, most assuredly have not been exaggerated — as the next, inevitable market crash will prove all too well, in Boslego’s view.

In his article, “The Yale Endowment Model of Investing Is Not Dead”(1) Timothy J. Keating concluded “that the problem is neither with modern portfolio theory nor asset allocation.” I disagree. While modern portfolio theory is a beautiful conceptual framework for thinking about investment strategies, it is flawed and unreliable when it comes to actually using it.

Having evaluated the endowment model and results, I find:

• The resulting asset allocations are not diversified when it comes to risk, they actually have a high correlation to stock market returns;

• Their “diversification” failed to protect them during the 2008/09 crisis because equity markets collapse together during financial market disasters;

• Endowments haven’t made major changes since before the 2008/09 crisis to their asset allocations, which are based on modern portfolio theory, a process that is inherently flawed because it requires inputs that are unpredictable, rendering its solutions unreliable; and so

• Endowments remain exposed to future financial disasters.

I speak to each of these points below.

The endowment model

Yale chief investment officer David F. Swensen has received much credit for “Pioneering Portfolio Management.”(2) When he began in 1985, the Yale Endowment was just under $1 billion, and it has grown to more than $17 billion, as of June 2011.

Mr. Swensen, like many portfolio managers, bases his investment approach on the statistically beautiful modern portfolio theory. But he found that applying MPT is much more problematic than non-practitioners may be aware. The idea that there is an “engineering-like” solution called the “efficient frontier” to be derived through “optimization” totally mischaracterizes the reality of the task, which is highly judgmental. In fact, it was because of such serious limitations of MPT that Mr. Swensen pioneered portfolio management and created The Yale Model.

The Yale Model has been emulated by many of the largest university endowments. Today, the “endowment model” is often used to describe a theory and practice of investing. “The model is characterized by highly diversified, long-term portfolios that differ from a traditional stock/bond mix in that they include allocations to less traditional and less liquid asset categories, such as private equity and real estate, as well as absolute return strategies.”(3)

'Diversification failed to protect asset values’

The endowment model is purportedly characterized by “highly diversified, long-term portfolios.” Not true from a risk standpoint. For example, the Stanford University endowment had an 89% annual correlation to stock market returns for fiscal years 2001-11 (see Exhibit 1). If you look at the individual asset classes, they are highly correlated (see Exhibit 2). Though this 11-year period is relatively short, a high correlation makes sense over the long term: the Yale Model is 95% equity-market-based.

1.

Over the short term, when there is a financial crisis, correlations among equity asset classes approach 100% because they all collapse together. From a risk standpoint, these portfolios lack diversification when it is most needed.

Following the 2008-09 financial crises, the major endowments reported:

Harvard University: “With a few notable exceptions, nearly every asset class did poorly … Our real estate portfolio, for example, suffered a loss of over 50% during the year … While diversification has been a mainstay and a driver of the portfolio’s return over the long term, the benefits of diversification did not bear out through the rapidly evolving and widespread events that unfolded in FY 2009.”(4)

2.

Yale University: “During this period, equity exposure hurt results, diversification failed to protect asset values and illiquidity further detracted from performance.(5)

Stanford University: “In the period from September 2008 through March 2009, diversification failed to protect the portfolio as correlations rose dramatically across all asset classes.”(6)

Modern Portfolio Theory: 'Predicting the unpredictable’

The theoretical framework for MPT relies on mean-variance optimization, an approach developed by 1990 Nobel laureate in economic sciences Harry Markowitz when he invented portfolio theory in one afternoon in 1950.(7) He had no stock market experience (8), which may explain the limitations that practitioners have confronted when they have tried to apply his theory over the years. He was a 23-year-old grad student in search of a thesis topic.(9)

3.

To conduct a mean-variance optimization, there are three inputs for each asset class: expected return (mean), expected volatility (variance, usually described by the standard deviation) and expected correlation of each asset class to each other (co-variance). The output is the “efficient frontier” of portfolios on a risk-return map. For each point along the efficient frontier, there is no portfolio that has a better return for the amount of risk taken. This sounds precise and “optimal,” but that impression could not be further from the truth.

There are a lot of practical problems when it comes to actually applying mean-variance optimization, many of which are pointed out by Mr. Swensen himself:

Unpredictable, unstable inputs to M-V. For starters, the inputs must be predicted. The first question is, for what time frame? M-V is a single-period solution.

Future returns, variances and co-variances are all highly unstable and can’t be predicted with much confidence over any time frame. When a model is based on a set of unpredictable and unstable inputs, the outputs are unreliable.

4.

In a recent interview, Dr. Markowitz said, “perhaps the actual computation of efficient frontiers may be of practical use. It will be so under two conditions: First, people want to use mean and variance; and second, They can come up with reasonable estimates [of the model’s inputs], which I believed would involve a combination of statistics and security analyst views.” (10)

Equity returns are unpredictable because markets do not behave rationally. Mr. Swensen’s colleague at Yale, economist Robert J. Shiller, came to the conclusion that prices are more volatile than the fundamentals they are supposed to reflect. “After all the efforts to defend the efficient-markets theory, there is still every reason to think that, while markets are not totally crazy, they contain quite substantial noise, so substantial that it dominates the movements of the aggregate market.” (11)

“In the international economy, politics matters at least as much as economics … they assume that political crises will be rare. But such crises — and their catastrophic effects on businesses —happen much more frequently than we imagine. On the curve that charts both the frequency of the events and the power of their impact, the 'tail’ of extreme political instability is not reassuringly thin, but dangerously fat.” (12)

5.

Simply using assumptions of long-term growth in equity markets can be wrong for a long time. “Consider an investor in Japanese equities in 1989. A portfolio invested in the Nikkei at the end of 1989 suffered a decline of 73% over the subsequent two decades,”(13) for an annual average return of -6%.

Invalid probability distributions. To generate the M-V output, the “efficient frontier” assumptions about how the returns are distributed are required. The conventional approach is to use a “normal distribution,” which has been disproved as a “true” representation of the dispersion and likelihood of financial returns by analysts.

A case in point: the 1987 stock market crash. The domestic equity market fell by 22.6% in a single day, a 25-standard-deviation event, according to “normal distributions.” A 22-standard deviation is one in a google (which is the digit 1 followed by 100 zeros):
.00000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000001.

According to reliable radiometric dating techniques used by geologists, the Earth is at least 4.5 billion years old, so suffice it to say this drop in price had no detectable chance of happening if returns are described by a normal distribution, even if the stock market had opened on Day 1 of the Earth’s creation!

6.

Little useful guidance. “Unfortunately, the mean-variance optimization approach provides little useful guidance in choosing a portfolio,”(14) according to Mr. Swensen himself. “Unconstrained mean-variance runs usually provide solutions unrecognizable as reasonable portfolios … because the process involves material simplifying assumptions, adopting the unconstrained asset allocation point estimates produced by mean-variance optimization makes little sense.”(15)

Incomplete output “Even in the unlikely event that the academic (M-V) approach proves helpful, decision makers wonder how the chosen approach will interact with the institutions spending policy.” (16)

The Yale Model is an extended and improved version of MPT, but many of the same fundamental problems remain. It adds simulations and probability analysis but requires the same unpredictable inputs and flawed “normal distribution” assumption.

The big risks: Maximum drawdown and behavioral risk in a disaster

Equity asset allocations haven’t changed much from before the crisis (see Exhibits 3, 4 and 5). As a result of this undiversified equity market exposure, endowments are exposed to a much larger cumulative loss (or drawdown from peak) than occurred in 2008/09.

Without an effective risk management strategy in place, endowments are exposed to what I call the “behavioral risk” of a financial disaster, which is the true risk, not “variance.”

“What matters is not how often you are right, but how large your cumulative errors are.”(17) Maximum drawdown from peak, or peak-to-valley (“P2V”) is the percentage of how much the portfolio has dropped from its highwater mark to its lowest point before establishing a new highwater mark. For example, I calculated the P2V for each asset class in Stanford’s portfolio from 2001-2011 and found some very large drawdowns: the P2V for private equity was a staggering 63%, while the P2V for real estate was 46% (see Exhibit 6). A drawdown of 30% should sound alarm bells.

If the endowment is experiencing a very large drawdown, say 50% or more, at some point in the process convictions about the correctness of the investment strategy will change because reaction to risk is emotional. “Both risk detection and risk avoidance are not mediated in the 'thinking’ part of the brain but largely in the emotional one.”(18) At some point in the process, decision-makers will decide to preserve what is left rather than risk the survival of the university as they know it, on “diversification that failed to protect the portfolio” or assumed future long-term returns.

Conclusions

Mr. Swensen has been pioneering its development and The Yale Model is an extended and improved version, but many of the same fundamental problems remain.

The large university endowments shifted their allocations to equities and equity-like markets over the years, and despite the diversification of asset classes, they are not diversified when it comes to risk. So when the financial markets collapsed in 2008-09, their portfolios were hit hard, even more than they expected because they thought they were diversified.

Furthermore, the endowments remain exposed to the risk of blowup during future market panics or crashes, which are virtually certain to happen. Finally, during financial disasters that cause very large losses, endowments are exposed to a behavioral risk that will shut down the strategy, opting to preserve what is left for survival.

See: How an ex-University of Chicago endowment chief is teaming up with a $4-billion AUM RIA to go after $500-million institutional accounts.

Notes:

1. Originally published in Investment Management Consultants Association’s March/April 2010 issue of Investments & Wealth Monitor.
2. Pioneering Portfolio Management: An Unconventional Approach to Institutional Investment, David F. Swensen, The Free Press, Copyright © 2000 by David F. Swensen.
3. Harvard Management Co. Endowment Report, October 2010.
4. Harvard Management Co. Endowment Report, August 2009. https://www.hmc.harvard.edu/docs/2009%20HMC%20Endowment%20Report.pdf
5. Yale Endowment Report, 2009
6. Report from the Stanford Management Co., August 2009. https://www.smc.stanford.edu/sites/default/files/site_files/Report%20from%20SMC%202009.pdf
7. Markowitz Interview, www.afajof.org/afa/all/Harry%20Markowitz%20Transcript.doc., Page 3.
8. Ibid. 7, page 4.
9. JOIM Conference Series: Harry M. Markowitz interview with Richard O. Michaud, San Diego, March 6, 2011, Journal of Investment Management, Vol. 9, No. 4, (2011), p. 2, © JOIM. https://www.joim.com/summaries.pdf
10. Ibid. 9, page 4.
11. “From Efficient Market Theory to Behavioral Finance,” Robert J. Shiller, Cowles Foundation Paper No. 1055, Cowles Foundation for Research in Economics, Yale University, page 90. https://www.econ.yale.edu/~shiller/pubs/p1055.pdf
12. The Fat Tail, Ian Bremmer and Preston Keat, Oxford University Press, © 2009 by Oxford University Press, Inc.
13. Ibid. 5, page 14.
14. Ibid. 2, page 123.
15. Ibid. 2, pages 105, 107
16. Ibid. 2, page 130.
17. Ibid. 2, page 149.
18.“Fooled by Randomness,” Nassim Nicholas aleb, Random House, Copyright © 2004 by Nassim Nicholas Taleb, page 38.

Robert Boslego is managing director of Boslego Risk Services, a consulting firm in Santa Barbara, Calif. He earned an AB cum laude in economics from Harvard College and an MBA from the Stanford University Graduate School of Business. Contact him at Boslego@Boslego.com.