This material is excerpted from The Ivy Portfolio: How to Invest Like the Top Endowments and Avoid Bear Markets, Copyright © 2009 by Mebane T. Faber and Eric W. Richardson. All rights reserved. By permission of John Wiley & Sons.
David Swensen graduated from the University of Wisconsin in 1975 with a degree in economics. He then attended Yale and received his Ph.D. under the legendary Nobel Laureate James Tobin (his doctoral dissertation focused on the eponymous “Tobin’s Q”). Upon graduation, he worked at Salomon Brothers for a three-year stint, where he structured the first financial swap transaction in history between IBM and the World Bank.
After a brief time at Lehman Brothers, Tobin offered Swensen the position of managing the Yale Investment Office (YIO), to which Swensen famously replied, “I don’t know anything about Portfolio management.” Tobin countered, “That doesn’t matter. We always thought you were a smart guy and Yale needs you” (Capital Ideas Evolving).
Swensen agreed to an 80% cut in pay, and while he now makes a little over $1 million a year, he could be earning many times that amount managing a hedge fund or a fund of funds in the private sector.
Swensen is clearly motivated by factors more meaningful than a large Wall Street paycheck. “I had a great time on Wall Street, but it didn’t satisfy my soul,” he says. “And I’ve always loved educational institutions. My father was a university professor, my grandfather was a university professor. So there must be something in the genes.” (NPR, All Things Considered).
How did David Swensen go about constructing this portfolio that was so far removed from commonly accepted allocations of the day?
The Yale portfolio is constructed based on academic theory— namely a framework known as mean-variance analysis. The technique was originally developed by Harry Markowitz in concert with Swensen’s mentor Tobin, and eventually earned Markowitz a Nobel Prize in 1990. It really boils down to “don’t put all your eggs in one basket,” or in other words, diversification works. You can put together a bunch of risky assets (stocks, real estate, commodities) and as long as they don’t all move together in a correlated fashion, the combined portfolio is less risky than the individual parts. Roger Gibson has some great examples of multi asset-class investing in his academic papers and his book, Asset Allocation.
Mean-variance analysis uses the expected returns of various asset classes, the expected risk (volatility, or how much an asset bounces around), and the expected correlation to find the portfolios with the highest return for a given level of risk (or lowest risk for a given level of return). A correlation of +1.0 means that two assets have a perfect positive relationship (they move together), and 1.0 means that they have a perfect negative relationship (they move opposite one another).
Figure 2.2 is a basic chart showing numerous asset classes and the curved line known as the efficient frontier.1 Portfolios situated on the curved line have the highest return for a given level of risk (and vice versa).
This framework often leads to overweighting assets with low (or negative) correlation to the existing portfolio, such as real assets and hedge funds.2 One of the problems with mean variance analysis is that it is very dependent on the inputs. It is very simple to use historical market returns as they are a fact. The mean variance optimization will tell you what the best allocation is for the past. We care about the best allocation in the future.
For example, there is a very interesting piece on market history from the Global Financial Data Guide to Total Returns. Looking at historical data from capital asset returns in the 1800s, Dr. Bryan Taylor finds that:
- Most people invested in bonds, not stocks.
- Virtually all of an equity investor’s returns came in the form of dividends, not capital gains.
- There was little difference in the returns to stocks and bonds.
- Since the government did not issue treasury bills and deposits were not federally insured, there was no “risk free” investment available to investors.
- Bond and dividend yields declined over the course of the century as the risk to investors and inflation declined.
- Although prices rose and fell in any given year, from 1815 to 1914, there was no overall inflation in the US and in most countries on the Gold Standard.
Taylor states, “What is interesting about these points, which would have been taken as given before 1914, is that during the twentieth century none of these assumptions proved to be true. By the end of the twentieth century, most investors were investing in stocks, not bonds, depended on capital gains, not dividends, received a large premium on stocks over bonds, had risk-free investment alternatives, saw interest rates rise during most of the twentieth century, and suffered from the worst inflation in human history.”
A mean variance optimization performed at the beginning of the twentieth century would have resulted in far different results than one performed at the beginning of the twenty-first century. Assumptions that you may be using today as fact (stocks outperform bonds, small caps outperform large caps, etc), could prove to be unreliable in the future.
Will stocks return 10% going forward? How volatile are commodities going to be over the next 50 years? Has private equity seen its day in the sun? Will foreign stocks become more correlated to domestic stocks? As Niels Bohr said, “Prediction is very difficult, especially about the future.”
While expected returns and expected volatilities are difficult to forecast, correlations are even harder. Combining assets with correlations that are all over the map does little to help your portfolio. The best way to go about the process is to combine assets with consistently low correlations to reduce risk.3 Table 2.2 shows that commodities have the least correlation with traditional asset classes over the long term.
While this table is useful, the process is made more difficult because correlations change depending on the market environment.
Ray Dalio, founder of the $150 billion hedge fund group Bridgewater Associates, expresses his opinion in the book 2020 Vision: “We don’t assume stable correlations, we look at a range of past correlations to stress-test our portfolio based on different correlation assumptions and we structure our portfolios to have no unintended consequences that would favor one type of economic environment over another.”
As an example that correlations can (and do) change, Figure 2.3 charts the three-year rolling correlations of stocks and bonds since 1903.
(This measures the correlation between stocks and bonds, updated every month using only the past three years of monthly returns.) While on average stocks and bonds are not very correlated with each other at .15, the correlation has varied from –.61 to .62 (and the one-year rolling correlation from –.85 to .86).
Table 2.3 shows the one-year rolling correlations between the main asset classes and U.S. stocks. On average U.S. stocks are only partially correlated to foreign stocks and Real Estate Investment Trusts (REITs), and less so to U.S. government bonds and commodities. However, over any one 12-month period the returns can be nearly identical. The takeaway is that even though some asset classes are on average negatively correlated with one another, in the short run they can move together almost identically. This matters little over the very long run, but for individuals looking to project their financial situation over the next 1 to 10 years, there are significant implications. Over the short term anything can hap- pen as evidenced by nearly every asset class declining in 2008.This is one reason why risk management is so important.
Correlations should be viewed as a tendency, not an absolute, and they are certainly something you cannot count on. A famous Wall Street saying is that in times of economic shocks, all correlations go to 1! (Meaning when it hits the fan, everything goes down together.)
Yale also uses a technique called Monte Carlo simulation, which stress tests the portfolio over thousands of different scenarios to come up with a range of likely outcomes. This simulation can give insight into the most likely outcome, as well as the chances of best- and worst-case occurring scenarios. (The moniker Monte Carlo was coined by U.S. physics researchers in the 1940s as a reference to the repetitive nature and randomness of possible outcomes in a gambling casino.)
Another way to stress test these correlations is to look at the returns when equities had their worst months. Table 2.4 shows the 10 worst months in stocks since 1972 and the returns of the other asset classes in those months.
The performance of foreign stocks and REITs did very little to dampen the losses from U.S. stocks. Only bonds did a good job of per- forming strongly in the months when stocks did poorly and did so nearly all of the time.
These mathematical techniques are useful to gain perspective, but they have other problems such as accounting for structural changes in markets and modeling liquidity and rare low-probability events (as evidenced by the 2007 herding/liquidity-based quantitative equity hedge fund meltdown). The 2007 Yale endowment report reads, “Investment management involves art as much as science.”
These methods, as well as a little common sense and experience, lead Yale to the current Policy Portfolio for 2007. The Policy Portfolio is simply their target portfolio for the year. See Table 2.5. (A more detailed Policy Portfolio follows later in this chapter; we included this simplified version to compare Yale’s endowment to the average endowment.)
Yale does not break out its investments in real assets. We have assumed an even split between real estate and commodities, but in reality the university has a diverse allocation of timber, oil and gas partnerships, real estate investments, and other real assets.
Compared with the average endowment, Yale has:
- Less stock exposure.
- Much less bond exposure.
- Much more real estate, commodities, private equity, and hedge funds.
The portfolio reaches for high returns, thus the endowment is biased towards equity, and equity-like asset classes, which total 96% of the endowment. Bonds, due to their vulnerability to inflation, are in the portfolio only as a hedge against deflation. Yale has a large allocation to real assets and nontraditional asset classes due to their return potential and diversifying power. The long-term time horizon (forever) is suited to exploiting illiquid and less efficient markets such as venture capital, leveraged buyouts, oil and gas, timber, and real estate.
This target mix is expected to produce returns after inflation of6.3% with risk (volatility) of 12.4%. The Yale 2007 endowment report states that their measure of inflation is “based on a basket of goods and services specific to higher education that tends to exceed the Consumer Price Index (CPI) by approximately one percentage point.” The Commonfund, a nonprofit devoted to management of college and university endowments, tracks an index called the Higher Education Price Index (HEPI). It ended the 2007 year at 3.4%, roughly 1% above the CPI figure of 2.6%.
When you add the HEPI inflation numbers back into the above numbers it results in an expected return target of 9.7%, with volatility of 12.4%. See Table 2.6.
While Yale uses this table as its bogey return to beat, it attempts to outperform this allocation using active management.
Footnotes:
1. If you really want to get deep on this subject Google “risk parity” for the papers by Bridgewater and PanAgora (which will also be linked on the web site www.theivyportfolio.com).
2. It does not make sense to label real assets as alternative asset classes. Humans have been buying and selling land and commodities far longer than they have traded shares of IBM. Hedge funds are not a separate asset class, but rather funds that trade existing asset classes.
3. Want more information on this subject? Read Mastering the Art of Asset Allocation, which has 174 pages on the correlations of asset classes. William Coaker also has two papers in a series titled “The Volatility of Correlation” and “Emphasizing Low Correlated Assets” that have great tables of lots of asset classes and the distributions of correlations. The links to these and other papers can be found at www.theivyportfolio.com.
Read more articles by Mebane Faber
This material is excerpted from The Ivy Portfolio: How to Invest Like the Top Endowments and Avoid Bear Markets, Copyright © 2009 by Mebane T. Faber and Eric W. Richardson. All rights reserved. By permission of John Wiley & Sons.
