In Part I of my discussion of When Markets Collide, by Mohamed El-Erian, I looked at a range of relatively abstract conceptual themes that the book raises. In this follow-up, I will explore the model portfolio allocation that Mr. El-Erian provides.
The Model Portfolio
Mr. El-Erian proposes an model portfolio that he feels will be well-positioned to deal with the markets of the future. This portfolio, along with its projected performance (from the book), are shown below.
Mr. El-Erian posits that this portfolio reflects the secular market shifts that he sees coming, and accounts for the correlations between these asset classes. His statements imply that a forward-looking portfolio analysis was used to derive this portfolio, presumably conditioned by his specific expectations (via something like Black-Litterman).
When Mr. El-Erian calculates the expected annual return and standard deviation for this portfolio, where do these numbers come from? I infer that he is using a quantitative portfolio model that combines his expectations for the various asset classes, and calculates the net risk in the portfolio—which is a function of how correlated these assets are. While he does not say so explicitly, I believe that what he has done is to locate a forward-looking efficient frontier and tried to locate this portfolio on the frontier, as in the linked analysis [PDF file].
This general approach is a standard of practice among institutional investors, but is entirely foreign to the vast majority of individual investors. Portfolio theory tends support diminishing allocations to broad market-weighted indices and increasing allocations to asset classes like real estate, commodities, and a range of ‘hard asset’ sectors—even without the kind of long-term trends that Mr. El-Erian sees coming.
While private equity is listed as one of the asset classes in this portfolio, Mr. El Erian has the following to say:
“Having a general exposure to private equity is neither a necessary nor sufficient condition for obtaining superior investment results.”
The discussion of hedge funds (which fit into the Special opportunities category) is similarly cautionary.
In a range of institutional research, and in my own research, I have noted that diverse forward-looking portfolio analyses suggest that about the best we can plan on for a portfolio in this risk range is a 1-to-1 ratio between average annual return and standard deviation. David Swensen, the Chief Investment Office at Yale’s endowment has stated that he is targeting an average annual return of 10.1% with a standard deviation of 11.8%, again consistent with this sort of 1-to-1 ratio.
I was, therefore, intrigued that Mr. El-Erian projects a model portfolio that is consistent with this 1-to-1 ratio. While it is easy to find portfolios that have beaten this 1-to-1 ratio over extended periods of history, this benchmark is a hard one to plan on beating going forward. The models that produce these 1-to-1 profiles, implicitly account for the uncertainty in our estimates of future returns, risks, and correlations among assets.
Perhaps the most unusual feature of this portfolio is the specific focus on infrastructure as a distinct asset class. Mr. El-Erian explains this as being due to low correlation between this class of assets and broad stock and bond indices---portfolio theory, once again. I have previously written about the particular attractiveness of utilities, transport, and other “infrastructure” asset classes on this basis.
There is increasing interest in infrastructure due to aging infrastructure in the U.S. and the fact that emerging economies are making (and must continue to make) massive investments in infrastructure (source: NY Times). This, combined with the positive portfolio effects of these assets and the companies that build them, creates an interesting narrative.
Commodities have a substantial presence in this portfolio, and the continued attractiveness of commodities is one of the secular trends discussed earlier. Even if you have a fairly dim view of the long-term expected returns from commodities, the low correlation between commodities and equities make this an attractive asset class.
The high weight given to “real assets” in this portfolio (27% of the portfolio vs. 15% for U.S. equities) is notable. While this may seem somewhat radical to many investors, forward-looking portfolio models strongly support this approach—which is why these asset classes show up strongly in the model portfolios that I write about (as in the All-ETF portfolio discussed above). Mr. El-Erian notes that institutional investors have moved broadly towards higher allocations in real assets.
It is interesting that projected volatility for this portfolio is for standard deviation of 8%-12%. This is very moderate for a portfolio with only 19% bonds. While we cannot see into the guts of the model inputs that drove these outputs, the projections do not suggest a future that is more risky than outlooks from a range of standard models. I think that Mr. El-Erian sees the potential for market dislocations in the potential for ‘fat tail’ events.
When looking at the specific allocations to different classes of equities, Mr. El-Erian cautions about the way that one invests in these categories—specifically that passive equity indices “reflect the world of yesterday rather than the world of tomorrow.” What he means here is that the common standard of market capitalization weighted indices is too biased towards what has done well in the past—an argument that is increasingly well accepted. He also cautions that so-called fundamental indices (which are most often touted as the best alternative to market cap weights) are subject to the potential for “construction biases.” This means that theses indices, specifically motivated by their historical performance, may simply be over-fitting history and have less forward-going value. I agree on both of these points. There is no reason to believe that a market cap weighted index or a fundamental index of an entire market (like the S&P500 or the RAFI 1000) is the best way to invest in that market.
With these caveats about thinking about how to represent an asset class as broad as U.S. equities or emerging economies, we can still look at a few broad strokes. This portfolio has almost equal allocations between domestic equities, foreign developed equities, and emerging markets equities. This is not as radical as it may appear, given that there are high correlations between broad U.S. equity indices like the S&P500 (IVV) or Russell 2000, the EAFE index (EFA) and a broad emerging markets index (EEM):
Correlation Matrix (three years of monthly returns, through July 2008)
If we construct an equity portfolio that is 35.7% in EFA, 35.7% in IVV, and 28.6% in EEM (to match Mr. El Erian’s proportions), we end up with a portfolio that has risk and return characteristics quite like that of the Russell 2000 (IWM). In other words, the high relative allocation to emerging markets does not make the equity portion of this portfolio especially risky—though it does have Beta of 120%. In the near term, the question would be whether the emerging market allocation is due for some correction, given the very high returns from emerging markets over recent years.
A portfolio-centric approach to analyzing the asset allocation is, essentially, fairly agnostic with regard to the country of origin of the underlying stocks. This results in the diminished U.S. orientation of the model portfolio.
The Map Is Not the Territory
One of Mr. El-Erian’s key themes is that the models that we have used in the past may not work in the future—and of course he is correct. The world is not static and our conceptual models are especially challenged in times of great change. Did our models of financial markets really break in the recent market conditions? Perhaps. Mr. El-Erian cites investor response to prolonged low levels of VIX, the volatility index, as an example of recent dislocations. Broad market volatility ran at very low levels for a number of years—and investor response was exactly what Minsky’s financial instability hypothesis would suggest: investors leveraged up and took riskier bets. The fact that many investors discounted the probability of a return to a much higher volatility environment was not a result of broken models, however. Options prices and Monte Carlo models were projecting a return to a more volatile market back in 2006.
The fact that many investors ignored the models and levered up their portfolios in the hopeful belief that market volatility had been forever tamed does not mean that the models failed. Edward Altman has come to similar conclusions with regard to the ratings agencies and the collapse of Enron and WorldCom. It is often the case that the models are correct, but people ignore them—and that, in my opinion, explains a great deal of the current financial meltdown.
Frankly, I believe that portfolio models have held up quite well in the recent market meltdown---when people paid attention to the models. It is often the case that the potential of large short-term gains on the trading floor leads to decisions that short-change risk management, and I lay much of the blame for the current losses on this effect. By contrast, witness the solid performance of Yale and Harvard in recent years as examples of how portfolio management models (applied properly) can protect a portfolio from these massive dislocations. The Yale [PDF file] and Harvard portfolios show trends in asset allocation consistent with Mr. El-Erian’s model portfolio.
The model portfolio that Mr. El-Erian proposes represents a radical departure from the norms that most retail investors are used to, but is generally consistent with trends that we are seeing in institutional research. The emerging paradigm tends to lower allocations to broad domestic equity indices—indeed to all market-cap-weighted equity indices, and to increase allocations to some specific sectors. For the individual investor, the message is to use a portfolio-centric approach to asset allocation and to break away from the arbitrary ‘pie chart’ allocations that remain the basis of most retail portfolios.