Modern Portfolio Theory: Bruised, Broken, Misunderstood, Misapplied?
If the global financial crisis has left us with any enduring lessons, it’s that asset return distributions can be significantly skewed and asymmetrical, and that fat tails are the norm rather than the exception. So how can investment practitioners manage this reality? That question animated a recent presentation by Peng Chen, CEO Asia (ex-Japan) of Dimensional Fund Advisors, at the CFA Institute Thailand Investment Conference, which was cosponsored with CFA Society Thailand and the Securities Exchange Commission Thailand. (You can watch a replay of Chen’s 56-minute presentation on our blog.)
As Chen observed, the global crisis that began in 2008 has called into question many of the basic assumptions and central tenets of modern portfolio theory (MPT). But, in fact, the world of investing has long been observed to be “non-normal” insofar as asset returns historically have not followed a normal, or Gaussian distribution, and correlations, rather than being stable, have varied significantly across different market regimes. The result is that the benefits of portfolio diversification, the discipline of adhering to a strategic asset allocation policy, and other practical implications of MPT are being challenged.
After providing a brief overview of traditional MPT as it was first developed by Harry Markowitz in 1952 and enhanced in the following decades by numerous scholars and financial economists, including William Sharpe, Robert Merton, and Eugene Fama, Chen addressed two key questions:
- Did asset allocation and portfolio diversification fail?
- How should asset allocation and portfolio diversification be implemented in a non-normal world?
On this first question, Chen contended that diversification in fact did not fail at the security level and the basic asset class (stocks, bonds, cash) level. However, the increasing globalization of economic activity and capital markets has made stock markets around the world much more highly correlated, reducing the benefits of diversification across these markets, especially during periods of crisis.
In addition, diversification into alternative assets such as commodities, real estate, private equity, and hedge funds did not appear to provide much benefit. In a recent empirical study, Peng and his former colleagues at Ibbotson Associates found that many hedge fund strategies have significant systematic, or beta, risk embedded in them, so the diversification benefit of combining hedge funds with traditional asset classes like stocks and bonds is limited. Moreover, the average alpha of hedge funds is more than offset by high fees, which typically include a significant incentive component, and the alpha is in fact declining over time as the hedge fund industry grows and opportunities to exploit inefficiencies remain more or less the same.
Active managers, on average, do not add value because chasing alpha is a zero sum game, Chen argued, or even a negative sum game after trading costs and the high fees are accounted for. It is also very difficult to identify truly skillful managers who consistently outperform over long periods.
With regard to the second question above, Chen introduced as an improvement to the normal distribution the truncated Lévy flight (TLF) distribution, which accounts for “higher moments” (in a statistical sense) of skewness and kurtosis, in addition to the first moment (mean) and the second moment (variance). Asset returns, Chen argued, are better depicted by the TLF distribution, which provides a more effective tool to analyze downside risk. Instead of mean-variance optimization under MPT, an approach using mean-conditional value-at-risk optimization could be adopted to come up with an optimal asset allocation, he added.
Chen concluded with his asset class performance outlook. As he examined return on bonds versus stocks over the last 40 years, he found that most of the total return from bonds came from the coupon income. With bond yields at historic lows, he said he finds it hard to see bonds outperforming stocks over the next several years.
Watch a replay of Chen’s entire talk at the Thailand Investment Conference on 5 October 2012.
Please note that the content of this site should not be construed as investment advice, nor do the opinions expressed necessarily reflect the views of CFA Institute.
We have to give up these academic approaches to asset allocation. None of them work — it doesn’t matter whether you have two or four parameters — the parameters are not stable, and can’t be stable.
My previous comment was not approved. Perphaps I was a little too honest. So David when you say “we have to give up these academic approaches to asset allocation” please could you be so kind as to clarify what you mean by the phrase “academic approaches” and what exactly would you propose as a possible substitute?
Agree with Valeri. What does David Merkel mean by “academic approaches”? Quantitative Asset allocation models have been utilized in the industry since the 80s. What are your alternatives? Monday morning quarterbacking is way too easy
I think “academic” models are esential in understanding financial markets. There are many alternatives to MPT, that work better under market integration phenomenom, see the papers of G. Bekaert. Also, Black-Litterman model is a significant improvement in my opinion.
Merkel is under no obligation to provide a substitute or “better option.” If he is right that the current approach is not useful, that is enough. People prove things to be wrong all the time. They have no obligation to provide what is right or better.
Just joined this group and noticed this discussion. On the question posed by the article … Did asset allocation and portfolio diversification fail? … the answer given is no. The reason cited is that high correlation amongst traditional and alternative assets is to blame, not the theory itself.
However the starting premise of MPT is that there exist some stable lowly correlated assets that will help moderate volatility of portfolio returns and improve the risk/return tradeoff. If we now agree that correlations are not stable and that they cannot be relied upon in tail-events, then it seems to me that MPT is discredited… as tail events become ever more regular.
We need to turn our attention to finding ex-ante risk data reflecting our investment views of the likely prevailing regime(s). Therein lies the rub.
For too long the investment community has been inconsistent in assuming stable covariance for its risk analysis while at the same time disclaiming to clients that “past performance is no guarantee of future returns”. Some people might argue that “inconsistent” is not strong enough a word!