The endowment model's 20-year winning streak had consultants and asset managers imitating its high-risk strategies -- and then came the crash
Brooke’s Note: With all due respect to Warren Buffett and Bill Gross, the most gilded name in investing is David Swensen for what he has done in taking Yale’s struggling endowment and turning it into a mighty one. Being possessed of this knowledge is a good feeling because we all like to believe that there are paragons of accomplishment out there that are, if not infallible, worthy of reverence. But reverence, whether it is for cyclists who win the Tour de France seven times or for founders of Internet companies with can’t-miss stocks, is a dangerous game. But does that even apply to Swensen? Robert Boslego shows in this article that it may especially apply to the endowment manager — according to Yale’s own reports.
In The Yale Endowment 2010 report (1), there is a section titled Monte Carlo Simulations, which has not received a lot of attention, probably due to its “nerd-only” appeal (I read it). Buried within this discussion, however there are two assessments worthy of much attention by RIAs:
• Yale says it runs a 17% — nearly 1-in-5 — risk of losing half of its endowment (in real terms) over the next 50 years, and a 28% risk of a “spending disruption,” defined as a 10% (or greater) decline in support for the university’s operating budget, over the next five years, by employing the Yale Model of Investing.
• Yale says that the “average endowment” runs an even larger risk—28%— of losing half of its assets (in real terms) over the next 50 years based on their asset allocations, and a 35% risk of a “spending disruption” over the next five years. (If Yale is correct, a large portion of endowment gifts may just end up in the “bonus pools” of Wall Street traders, hardly their intended destination.)
The report concludes by saying: “Yale’s simulations show relatively significant probabilities of circumstances that would be traumatic for educational institutions, highlighting the tenuous balance between protecting endowment purchasing power and maintaining a steady and substantial stream of spending.” See: What we all feared: 'Better’ disclosure yields worse results, according to Yale professor’s study.
These assessments raise two important questions:
• Why are the risks so high?
• Is there a better alternative to the Endowment Model of Investing?
As to the first of these issues, why the risks are so high, I believe that:
• The Yale model is 95% invested in equity and equity-like assets — primarily illiquid investments, such as hedge funds, private equity, and real estate.
• Equity markets exhibit excess volatility, behave irrationally, and are prone to periodic financial crises involving manias, panics and crashes; investments in illiquid asset classes are even riskier than public equities.
• The Yale model, and more broadly, the endowment model, are based on Modern Portfolio Theory, which is a “single period solution,” meaning they make allocations to asset classes that they do not intend to change no matter what happens; they do not actively manage their risks.
The case for MPT
As Yale chief investment officer David F. Swensen explains in a lecture that he gave to Yale MBAs in 2008 (which you can watch on YouTube)(2), the Yale model rests on two core tenets:
• “If you’re investing with a long time horizon, having an equity bias makes sense.” 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. and
• Diversification. “Because [Harry] Markowitz (inventor of MPT) tells us that diversification is a free lunch … for any level of return, you can reduce the risk, and for any given level of risk, you can increase the return.” See: Why diversification is still the go-to risk killer.
How the ’08 crash changed everything
The question institutional investors are asking now is whether the events of the past few years require a reappraisal of principles underpinning the Yale model. In 2008, worldwide equity markets collapsed and many assets that conventional investment wisdom —until then — regarded as effective equity diversifiers, such as commodities, also experienced dramatic falls, explains the EDHEC-Risk Institute in Paris in a recent report.(3) These events dashed the exaggerated hopes placed in traditional forms of diversification and led investors to pay increased attention to the volatility and downside risk of equity holdings, if not to question their allocations altogether.
Swensen’s golden rules of asset management — stocks for the long run and diversification — seem to be going out of fashion. Pension-fund managers that may have years to ride out losses on their portfolios until they turn into gains are increasingly throwing in the towel in favor of less volatile, lower-returning bonds. For the first time in more than a decade, more of the $1.246 trillion assets represented by the 100 largest U.S. corporate pension funds are now in bonds instead of equities.(4) Corporations are weary of stock market volatility; they want less risk in their portfolios. See: The 4 biggest investment performance myths — and how they can torpedo advisor-client trust.
Financial crises are often associated with the peaks of business cycles. Not every upswing in business excess leads inevitably to mania and panic, but the pattern occurs often enough.(5)
What happens, basically, is that some event changes the economic outlook. New opportunities for profits are seized, and overdone, in ways so closely resembling irrationality as to constitute mania. Once the excessive character of the upswing is realized, the financial system experiences a sort of 'distress,’ in the course of which the rush to reverse the expansion process may become so precipitous as to resemble panic. In the manic phase, wealthy and creditworthy investors switch out of money or borrow real or illiquid financial assets. In panic, the reverse movement takes place, from real or financial assets to money, or repayment of debt, with a crash in the prices of commodities, houses, building, land, stocks, bonds—in short, in whatever has been the subject of the mania. (6) Because the financial system is so intertwined, when credit tightens, that adversely impacts virtually all asset classes. See: 12 for 2012: A dozen issues you and your clients should really be focusing on this year.
This type of market behavior is inconsistent with the theory of rational expectations, in which prices reflect future fundamentals, which is at the core of the efficient markets theory. In 1981, Yale economist Robert J. Shiller (7) argued that the volatility of the equity markets was so large that it dominates the movements of the aggregate market. By the late 1980s, following that controversial finding, no study could produce a predictive link between stock market movements and subsequent fundamentals. (8) Ironically, in 1990, Markowitz was given the Nobel Prize for MPT, which rests on the unproven or disproved efficient-markets theory. See: The Yale endowment model of investing is not dead.
Consequently, economic research shifted to finding psychological behavioral models to explain market behavior. In 2002, Daniel Kahneman became the Nobel laureate in economics, despite being a research psychologist and never having taken a single economics course. Behavioral economics had been recognized as the replacement to a failed theory.
In Swensen’s book (9), there is so little discussion of “risk management” that those two words don’t even appear together in the index. He covers it briefly under “The Role of Market Timing,” and states, “If market timing involves betting against the stock market by reducing equity holdings and increasing cash positions, long-run expected portfolio returns decline as the market timer’s position decreases risk levels.”
Swensen goes on to say, “Because such activity lowers anticipated returns, market timers must succeed substantially more than 50% of the time to post a winning record. The wind is in the . The wind is in the speculator’s face…” (my emphasis) (10).
There seems to be some confusion on this important point. Speculating is taking an exposure (long or short) in an attempt to benefit from a price change, and hedging is a risk management technique for reducing an exposure to avoid a loss. As fiduciaries, RIAs may consider a “winning record” to be one which avoids heavy losses during financial crises and earns reasonable returns during the “good times.”
Extraordinary popular delusions
The inputs to the Endowment Model are predictions of future market returns, volatility and the covariances among asset classes. As I’ve previously explained, there are many problems with such a model. See: “Why the Yale endowment model may still be fundamentally flawed”: http://www.riabiz.com/a/13639069/why-the-yale-endowment-model-may-still-be-fundamentally-flawed. Feedback models are not attempting to predict prices, but rather to respond effectively to conditions that often set off price shocks. See: Can the behavior of RIA clients be changed? Yes … but.
One of the oldest theories of financial markets is price-to-price feedback. (11) Charles MacKay, in his influential book, Memoirs of Extraordinary Popular Delusions (1841), described the famous tulip-mania episode in Holland in the 1630s, a speculative bubble in tulip bulbs, with words that suggest feedback and the ultimate results of feedback.
Feedback models are based on human nature, the observation that people react gradually to price changes over months or years, not to just yesterday’s price change. The models are built to recognize that there are other shocks, besides feedback, influencing price.
The Endowment Model is simply not built to recognize or manage these risks. More portfolio risk management—and less speculation—is needed to produce a more stable performance both in the short and the long run.
The success of the Endowment Model pre-2008 over a 20-year period when asset prices rose (1986-2006) strongly encouraged smaller endowments, pension funds, foundations, investment consultants and asset managers to imitate these high-risk strategies.
Very few investors have the risk appetite of the large endowments. Moreover, Yale estimates that the “average endowment” takes more risk than it does. So what’s the alternative investment strategy, does it have a solid basis, and where is the proof that it works?
An investment strategy involving risky assets is essentially a risk management strategy because all that an investor can control is the risk exposures that are taken. Returns are determined by the market and are beyond the control of any single investor.
Portfolio risk results from what is in the portfolio and how large it is. I call the risk that derives from what is in a portfolio the “horizontal risk,” and the risk that derives from how large a portfolio is the “vertical risk.”
The orientation of Vertical Risk Management is to manage risk by decreasing exposures to prevent large losses of principal. Its other goal is to take risk to earn returns when risk levels are at more tolerable levels.
VRM involves combining two approaches: stop-loss, and price feedback. A stop-loss, a predetermined exit point, is a protection from a “Black Swan,” says the author of “The Black Swan.”(12) It is also a solution to the “disposition effect,” which includes the tendency of investors to hold losing investments too long. (13)
The typical stop-loss approach does not generally provide guidance as to when to increase exposure. VRM addresses that issue by providing an algorithm for decreasing and increasing position sizes based on a price-feedback model, having its foundation in Behavioral Economics, not MPT or the efficient markets theory.
Back-testing over the long run
VRM was back-tested utilizing daily values of the Dow Jones Transportation Index (DJT), going back to its origin in 1928 and proceeding daily through 1999, a period of 71 years, including 17,889 days, encompassing the period of the Great Depression and its aftermath. Although the Index could not be traded as such, this simulation was used to test the strategy and to compare its results with the DJT risk and returns. See: 10 investment ideas that STILL don’t work.
Each day using the latest closing price, the algorithm “decides” how long the position would be, and VRM makes the adjustment at the close of the following trading day. This procedure was used to prove that all of the information that was needed could have been known.
The position size can range from 0% (no positions) to 130%, where 100% is the dollar size of the portfolio. The liquidity rule was that 25% of the position could be added or decreased in a given day. While an endowment as large as Yale’s could not reasonably change its equity exposures as fast as contemplated in the discussion to follow, the exposures of RIAs’ more typical portfolios could.
VRM was designed to avoid Black Swan portfolio impacts. The 1929 stock market crash is conventionally said to have occurred on Oct. 24 and Oct. 29, but the bulk of the downturn occurred over the next three years (see Exhibit 1). VRM began reducing positions on Oct. 18, 1929 and had only a 5% exposure on Oct. 24 and no exposure on Oct. 29. VRM remained largely out of the market in the 1930s because of the high risk it detected.
Over the 10 year period 1929-1938, the DJT lost 78% whereas the VRM strategy gained 38% (see Exhibit 2). The largest peak-to-valley (P2V) drawdown (computed annually) for DJT was 83%, whereas the largest P2V for VRM was only 8%.
For the entire period, 1929-1999, VRM had a total compounded return of 20,371% compared with 1,442% for the DJT. The maximum drawdown calculated annually for VRM was 21% compared to 84% for the DJT (see Exhibit 3).
The primary purpose of a “Walk-forward test” is to prove the validity and robustness of an investment strategy. It is one of the very best methods available. (14)
After the VRM strategy was tested over the historical “in-sample” period of 1929-1999 (as above), it was then simulated over a new sample period of different market data. All of the specifications of the strategy were exactly the same as those used for the DJT simulation. This process is also known as “out-of-sample” testing and double-blind testing.
The VRM strategy was simulated daily from the end of June 2000 through the end of June 2011, using a simple, constant portfolio of 10 top holdings, equally weighted, rebalanced weekly, of SPY, the exchange-traded fund, which is designed to provide investment results that, before expenses, generally correspond to the price and yield performance of the S&P 500 Index.
This period included years of both bull markets, FY 2004, 2007, and 2011, when SPY gained 19%, 20% and 30%, respectively, as well as the financial crisis, FY 2008, 2009, 2010, when SPY lost 13%, 26%, and 16%, respectively. For the whole period, FY 2001-2011, SPY lost 19% and experienced a 46% P2V drawdown. See: How the Harvard and Yale endowment models changed to avoid a repeat of 2009.
By contrast, VRMS (i.e., VRM for SPY), produced a total return of 144% (including transaction costs) over the period. The VRMS total return was better than the Harvard and Stanford endowment returns but was a little less than Yale’s (see Exhibits 4 and 5).
However, the P2V for VRMS was superior at just 14%. Yale, Stanford and Harvard had P2Vs of 25%, 26% and 27%, respectively (see Exhibit 6).
It is important to recognize that The Endowment Model is an intentionally and inherently risky investment strategy, and that the Yale Investment Office acknowledges that. Since most investors do not fit into their risk pool, RIAs should not be swayed into following The Endowment Model just because the elite universities use it. It’s also important to consider that there is no guarantee that investors will stick with an RIA or investment strategy if they are down 50% (or more), and there are many reasons to suspect they would not.
Unlike The Endowment Model, VRM is a risk management strategy that reduces risk to avoid large losses and uses modestly leveraged (at times), liquid, publicly traded ETFs or stocks to try to earn higher returns during periods when large risks are not detected. As a systematic strategy, VRM can be objectively tested over past periods and applied in real-time. It is based on an effective risk management approach, the stop-loss, combined with a long-standing theory of why and how markets experience manias, panics and crashes, coupled with the objective of “sitting out” panics and crashes. See: How ETFs have been oversold when it comes to flexibility, lower costs and tax efficiency.
The thinking behind VRM is that it is prudent for investors to be on the sidelines at times to avoid heavy losses, and that fast or sizable losses trigger predefined exit points. Investors who don’t have multibillion-dollar positions can be more nimble than the ultra-rich endowments and use their smaller size to their advantage. Ultimately, confidence that VRM has effective controls in place to avoid major losses is critically important.
(1) See page 23: http://www.yale.edu/investments/Yale_Endowment_10.pdf
(2) Open Yale Course Financial Markets 09. Guest Lecture by David Swensen, http://www.youtube.com/watch?v=fWAP4hsa9_8
(3) “The Benefits of Volatility Derivatives in Equity Portfolio Management,” Lionel Martellini and Renata Guobuzaite, Copyright EDHEC 2012, Forward by Federic Ducoulombier, Director, EDHEC Risk Institute-Asia.
(4) “Whither the Yale Model?” http://blogs.reuters.com/unstructuredfinance/2012/04/04/whither-the-yale-model/
(5) Charles P. Kindleberger, Manias, Panics and Crashes: A History of Financial Crises, John Wiley &Sons, Inc., Copyright © by Charles P. Kindleberger, Fourth Edition, 2000, pages 1-3.
(6) Ibid. 5.
(7) Robert J. Shiller, “From Efficient Market Theory to Behavioral Finance,” Cowles Foundation Paper No. 1055, Cowles Foundation for Research in Economics, Yale University, 2003, page 90. http://www.econ.yale.edu/~shiller/pubs/p1055.pdf
(8) Ibid. 7.
(9) David F. Swensen, Pioneering Portfolio Management: An Unconventional Approach to Institutional Investment, The Free Press, Copyright © David F. Swensen.
(10) Ibid. 9, page 55.
(11) Ibid. 7, page 91.
(12) Nassim Nicholas Taleb, “Fooled by Randomness,” Random House, Copyright © 2004 by Nassim Nicholas Taleb, page 131.
(13) Shefrin and Statman, “The Disposition to Sell Winners Too Early and Ride Losers Too Long: Theory and Evidence,” (1985).
(14) Robert Pardo, The Evaluation and Optimization of Trading Strategies, Copyright© 2008 by Robert Pardo, pages 247, 252.
Robert Boslego is managing director of Boslego Risk Services, a consulting firm in Santa Barbara, California. 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.