From The Horse’s Mouth: Yale's Endowment Officer Makes Financial Sense

If you don't know who David Swensen is, you should. He is the Chief Investment Officer of Yale's endowment and is widely considered one of the top institutional managers in the world. He has also written two well-received books on portfolio management and investing. An article about him from Summer of 2007 is very enlightening.

Yale has done very well with Mr. Swensen at the helm:

On his 22-year watch, Yale's endowment has grown tenfold and has spun off enough cash to fund about a third of the university's annual operating budget. Swensen has cranked out an average annual return of 16.1 percent — the best performance of any college endowment in the land; by comparison, the S&P 500 posted an average annual gain of 12.3 percent during that period.

(from interview linked above)

Mr. Swensen is well known partly for his strategy of taking on asset classes that achieve maximum diversification benefits relative to equities--like timber. This concept is at the core of portfolio theory, but many investors do not really understand what it means at a deep level. Assets with low correlations to equities (like commodities and other asset classes) can increase the return of a portfolio without increasing risk. The remarkable performance in Yale’s endowment is undoubtedly substantially due to the expert exploitation of diversification benefits between asset classes. When I read this interview, what really jumped out at me was the following statement:

Oddly enough, Swensen actually strives to produce equity market-like results for the Yale endowment. The expected return of his target portfolio is presently 10.1 percent per year with a standard deviation of 11.8 percent. While Swensen aims for the broad market's return, he diversifies the portfolio's other assets to provide insulation from the equity market's potential for outsized losses in any given year.

(from interview linked above)

Make sure that you understand this statement. Mr. Swensen believes that the most return that he can achieve for this risk level (annualized standard deviation in return of 11.8%) is 10.1% per year. We must conclude that even with all of the range of assets available to him, Mr. Swensen believes that this is the best that he can plan on.

This projected return level at this level of risk will be familiar to those who have read my articles. I have consistently found (using our own forward-looking portfolio analysis tool, QPP) that the best that investors in well-diversified portfolios can reasonably plan for is a 1-to-1 relation between expected return and standard deviation in return (i.e. 10.1% in return and 11.8% in standard deviation)--and here we have David Swensen at Yale telling us that this is about what he is aiming for!

What is even more interesting is that a range of other well-respected institutional investors and analysts have concluded the same thing.

The Retail Investor’s Portfolio

In the interview with Mr. Swensen, he discusses what he thinks individual investors should be doing. Notably, he does not suggest that retail investors take on non-equity assets like timber that have made Yale's performance so strong. Instead, he proposes a plain vanilla portfolio that did not look all that great to my eye (below).

Swensen’s model retail investor portfolio

Mr. Swensen's suggested retail portfolio looks like just about every other pie chart portfolio you will ever see. This portfolio has stocks, bonds, and real estate and the stocks include global indices. When I ran this portfolio through our forward-looking portfolio management software, Quantext Portfolio Planner [QPP], the results were essentially what I expected. QPP projects that this portfolio will match the expected return the S&P500 on a going forward basis (8.2% per year), with less risk (the projected SD is 11.8% vs. 15% for the S&P500). This is okay, but far from spectacular—and, notably, far below the 1-to-1 ratio between expected return and standard deviation that Mr. Swensen is planning for with the Yale endowment.

Mr. Swensen is famous for seeking out asset classes with low correlation to broad equity indices, such as a range of commodities, timberland and other real assets. Where are these in the retail portfolio? Mr. Swensen only has 12% of Yale’s portfolio in domestic equities but he is proposing that retail investors put 30% of their assets in domestic equities. While it is true that Mr. Swensen has access to private equity and other assets that the average retail investor cannot easily include in his/her portfolio, it is certainly possible to get a lot closer to the Yale model.

To broadly replicate the kind of performance that Mr. Swensen has engineered for Yale in a retail portfolio, I replaced all of the Vanguard funds with ETFs and then added commodities (via (DJP)), a timber REIT (PCL), a very large electrical utility (EXC) and a large oil company (COP). The idea here is to provide a significant exposure to commodities and real assets. I have also ditched the short-term bond fund. Our new model portfolio looks like this:

Modified retail investor portfolio

I used the commodity index (^DJC) which DJP tracks for the analysis because DJP has a very short history. The modified retail portfolio contains only 5% in very short term bonds, and moves most of the bond allocation to TIPS. This portfolio also has only 10% in the broad REIT index, but has another 5% in the timber REIT, PCL. ICF and PCL have very little correlation to one another (28%), however, because of the very different nature of their businesses. The addition of COP and EXC gives the portfolio more exposure to real assets.

Quantext Portfolio Planner projects that this portfolio has an expected return of 10.1% per year, with a standard deviation of 11.6% per year---almost exactly what Mr. Swensen says that he has targeted for the Yale portfolio.

There is no magic in these results. Commodities and real assets provide powerful diversification benefits that will boost portfolio returns without increasing risk—which is exactly what we have shown here. The use of forward-looking models to come up with a portfolio which effectively exploits diversification benefits in this way is discussed in depth in this article (the same one referenced earlier).

This conceptual model of building a core out of index funds or ETFs and then spiking the portfolio with individual stocks is explained in less technical terms in Yes, You Can Supercharge Your Portfolio by Ben Stein and Phil DeMuth, and they use QPP to do it!

Bringing It All Together

If you ask most retail investors how much risk they have in their portfolios and how much return they expect to receive for bearing this risk, they will simply have no idea. If you asked an advisor the expected return he/she is targeting for a specific client portfolio and the expected risk level, you would often not get much more information—at least not quantitative information. By contrast, David Swensen has a very firm idea of the expected return (i.e. going forward) in Yale’s portfolio, the expected volatility (i.e. the standard deviation), and the most return he can realistically plan for at a given level of risk. These are crucially important pieces of information.

When David Swensen says that he is targeting about 10% in return with a standard deviation of about 12%, this should get your attention. There is actually remarkable consistency among forward-looking estimates of the best that anyone can do among institutional investors and analysis, and Mr. Swensen’s targeted risk and return is right in line with these estimates. I have analyzed a wide range of portfolios and have consistently found that a rough 1-to-1 ratio between annual expected return and annualized standard deviation in return is the best that can be planned for on a forward-looking basis.

Forward-looking models (like Quantext Portfolio Planner) are the standard of practice in institutional money management, and the technology is gradually making its presence known among retail investors and their advisors (as shown by the Stein-DeMuth book). What would your portfolio look like in one of these forward-looking models?

Note: all projections in this analysis used default settings in QPP and used three years of historical data (through 12/31/2007) as inputs.

Comments

[TraderPig]

Dr. Considine,

I've followed your ariticles for some time and would like to thank you for those very insightful thoughts.

I noticed that you'd like to include some individual stocks into the portfolio with some broard market index products (index funds or ETFs). My question is: is the QPP tool suitable to analyse the risks/returns of these individual stocks? I'm actually not comfortable with the rusults of these stocks when the analysis only involves with 3-5 year data. There're too many non-systematic risks for the individual stocks.

Thank you.

TraderPig

[GeoffConsidine]

Hi TraderPig:

It is always good to be skeptical. QPP/QRP have been extensively tested out-of-sample in varying market conditions and the portfolio projections of risk and return for both stocks and funds turn out well.

I tend to run the model using three years of data, but the model the combines these data with long-term capital markets data on risk/return relationships--so its not as though the model only knows about the last three years.

For an example of this kind of testing, see this article:

seekingalpha.com/artic...

Regards,

Geoff

[Jim Kerns]

Geoff,
Aren't there supposed to be some charts with this article? JK

[GeoffConsidine]

Yes-there are supposed to be figures. I have asked SA to re-post this article. It is available in complete form here:

www.quantext.com/David...

[whidbey]

This is an excellent approach if only we can learn to set the course and await the results. I management "otherpeoplesmone... and they are often beyond appreciation of rationality (or Yale): "just the numbers please, and is that really your fee for these results? " Very enjoyable and well presented.

[User 155471]

Excellent article. It is the first time I`ve seen a practical mathematical relationship between risk and reward. My question regards the use of standard deviation. In your example, are you saying that the QPP model predicts a maximum deviation of 11.6% return above and below zero return or around 10.1% return? If around 10.1%, the return could vary from -1.5% to 21.7%. Do I have this correct?

[GeoffConsidine]

Standard deviation is relative to the expected return--so you are looking at 10.1% per year plus/minus 11.8%. That said, the plus/minus 1 standard deviation is not the extreme case. In general, only 2/3 of outcomes will fall within the average +/- 1 standard deviation. 95% of outcomes will fall withing 2 standard deviations.

[AAI]

Great piece, Geoff. Thanks for the headsup on QPP as well.

[pizon]

What is the affect of taxes using your strategy of funds (spdrs, ishares, etc.) versus owning baskets of diversified stocks in each equity asset class?

[GeoffConsidine]

Pizon:

Good question on taxes. In taxable accounts (obviously) it is generally better to hold individual securities and to rebalance only when the risk-return balance of the portfolio shifts outside of your desired range or when the diversification benefits shrink because of some over-concentration. This is the best approach. Many investors and advisors feel uncomfortable comparing the risk/return benefits of ETF's / index funds to a basket of individual securities because they lack the tools to really capture equivalence and risks of default in individual stocks--but that can be handled.

The average retail investor will capture the vast majority of the benefits of this kind of strategy with index funds, however, and he/she is typically not willing or unable to manage the individual securities to capture these benefits.

Geoff

[mydoghatescats]

I always look forward to your stuff. I read this one a while ago but I'm working on my CFP and ran across something similar.

In studying for my CFP we are told to use CV (Coefficient of Variability) or Standard Deviation divided by Return (SDev/Ret) as a comparison tool in picking portfolios. I say your measure, Return divided by Standard Deviation, the inverse of CV, is more intuitive.

It directly, proportionally measures, as opposed to inversely measures, the thing we are looking for: stability, gain with the least pain, the biggest return-bang for the smallest risk-buck. It also directly measures the degree of foolishness we are exhibiting by believing it can be much much greater than 1 (or what the broad market action of portfolios, passive and active, tell us what is currently feasible).

Thanks for your great articles. I guess I will always be looking in the rear view mirror when investing, but your work helps me turn around and look out the front.

I used to be a deck officer in the merchant marine. When I was starting out, I can remember that on more than one occasion a captain would come up on the bridge and catch me with my head glued to the radar on a perfectly sunny, high visibility day. Then they would point to the window and shake their head at my ninnyness for not appreciating that radars, or any tools we create to measure things are all fine and good, but sometimes you need to look out the window at reality.

If the ratio you describe is unclaimed you should name it: CDS, or Considine's Coefficient of Stability. It's basically just showing the Efficient Frontier, and they already gave out the Nobel Prize for that piece of work, but they name ships after people for far less, so lay claim. And it's a hell of lot more lucid and plainspeaking way to describe the Efficient Frontier than the way it is described in my CFP textbook.

Hope you sell lots of QPP software. Maybe when I get my CFP and figure out how to use this HP10BII Business Calculator I'll be needing it.

[User 199831]

VFISX is Vanguard Short Term Bond fund, not an Ultra Short. Thought you may want to correct this typo. Enjoyed your article and found it very useful. Thanks.

[Ranger Ric]

Geoff:

I have always supported the concept on diversification and your excellent work has put some important refinements on this concept. I have three questions.

Retail investors have some ability to invest in "hedge-like" funds through long-short mutual funds and private equity through individual companies like Blackstone and Fortress and the Powershares ETF PSP. You have not discussed these alternatives in your work. Do these alternatives not adequately address the asset areas that Swenson is using or do you believe that these asset classes do not add any diversification benefits?

Why do you think that Swenson not have utilities as an asset class as you recommend?

Finally, it appears that diversification benefits are not static. You have pointed out that foreign stocks do not provide the diversification benefits today that one might expect. I believe this is due to increasing globalization. While your models make sense to me, do you see them changing over time?

[GeoffConsidine]

Ranger Ric:

First "hedge like" asset classes require considerable care. I have not written about these but they are certainly worth examining. I prefer to get as much as I can get from basic asset allocation, etc. Public private equity funds are a different animal that private equity of course...I have some doubts about the viability of a public private equity stock: a lot of the argument for private equity is that a private firm can be run more efficiently than a public one--so what happens when they go public? BX and FIG do not exactly inspire confidence...

As to why Swensen does not look at utilities as an asset class: I don't know. I think that utilities look like the kind of thing he would like. This is an interesting question. Obviously I see utilities as a rather special "class."

The idea that the right portfolio changes over time is correct. In fact, the old idea among institutional investors of a permanent "policy portfolio" is quite outdated. Inst. investors increasingly revisit asset allocations and shift their portfolios as the world changes.