Why the Yale endowment model may still be fundamentally flawed

MPT is a great theory but completely flawed and unusable in creating a truly diversified portfolio. This becomes obvious in the way people treat the output; they add constraints to limit the allocation that can be made to any individual portfolio constituent and set floor levels as well. As soon as it becomes necessary to modify the output it becomes obvious that the method for creating that output is flawed.

In allocating to portfolio constituents, the question that should be asked is “will the past performance of this portfolio carry into the future?” In research my firm, Brandywine Asset Management, conducted in the late 1980s and early 1990s, what we discovered was that future performance is much more likely to match past performance if a portfolio is 'balanced’ across its constituents and the various return drivers underlying that portfolio’s trading strategies.

I discuss these topics throughout my book (which is an Amazon best-seller) and am happy to provide a link to one of the chapters where I briefly discuss the futility of allocating based on archaic “asset classes:” http://bit.ly/qG6oNE.

This is a well-constructed concise article illustrating the bounds imposed on MPT.

There is no question that MPT is a theoretical construct focused on standard deviation as a measure of portfolio risk. Also no question that MPT inputs are constrained, require assumptions which may not reflect future experience, etc. Also no question that most portfolio optimization software can lead to some very odd results, unless constrained by common sense. In a sense, perhaps we just take the concepts of MPT and expect them to deliver results they were never designed to provide.

The question I would ask as a read the article… what is the alternative? What other portfolio strategies would “work” better? And what proof exists as to such?

In posing such a question, it must be into the context of the goals which may exist for a particular investor’s portfolio. One can easily conclude, given the nearly unbounded irrationality of equity investors and the tremendous price fluctuations which can occur, that investors with shorter investment horizons (until the funds are withdrawn from the portfolio) should not be investing in equities. The risks are just too great.

At the same time, an investor with a 20-year time horizon, or greater, until the funds are likely to be utilized, is with a high degree of probability (but even then, not certain), likely to generate after-tax returns well in excess of those of fixed income assets (and certain other asset classes). One might surmise that, for such an investor – if discipline can be maintained – shorter-term market events should not be seen as troublesome.

Of course, the typical investor has a series of time horizons. And, for those whose other sources of funds are subject to risk (i.e., loss of income resulting from unemployment, etc.), any assumed time horizons are not set in stone. Therein lies just part of the complexity.

Another part of the complexity may also be self-induced by investment counselors, if they promote to their clients that they can manage the inherent volatility in the equities markets in the short term, such as through tactical asset allocation strategies (of which there are many, some of which have been dis-proven and many of which remain unproven) or other means, perhaps they are promising more than they can deliver. If, however, the adviser acknowledges to their clients that short-term price volatility in the overall equities markets is a price paid for the expected higher returns of equities (relative to fixed income and many other asset classes), then perhaps they have set the stage for a meaningful portfolio management discussion with the client.

MPT has not been “disproven” by the market events of 2008-9. What may have been disproven is the improper use of the MPT construct to achieve results for which MPT was never designed to achieve. MPT continues to have utility. Through a greater understanding of the many dimensions of risk and returns, the multiple and varying investment time horizons of individual clients, and adequate due diligence on investment strategies designed to minimize one or more types of risks, advisors can add value in the design and management of client portfolios.

Ron, as to your question “What other portfolio strategies would 'work’ better,” the answer is “one where the future performance most closely matches past performance.” Nothing else matters. For if a person can look at a back-tested portfolio performance and determine it is likely to recur, then they have the ability of creating portfolios to match their financial requirements. If there is no (or very little) predictability, then any effort put into creating a back-test is virtually worthless.

To accomplish this a person needs to create a balanced portfolio, one that is balanced across “return drivers,” not “asset classes.” (As I point out in my book, asset classes are simply long-only trading strategies that do not attempt to disaggregate their many separate return drivers). Once viewed in this fashion it is easy to create a truly diversified portfolio, rather than one constrained by the shackles of asset classes. There are literally hundreds of return drivers that can be developed into trading strategies that can then be incorporated into a balanced portfolio. Holding a long equity position is just one of them (and should receive a correspondingly equal allocation to the other trading strategies). I mention a number of these trading strategies in the book and provide more specifics of some of them in the book’s “Action Section” at www.JackassInvesting.com (a new interactive Action Section will be launched in July, providing even more strategy details).

We developed such a portfolio allocation model over twenty years ago and it remains as valid today as it was then.

I’m addressing the flaws I pointed out. For example, instead of using variance as the risk measure, I’m using maximum drawdown from peak as the risk variable. The main point of this article is to offer logic and evidence that the MPT mean-variance approach is inherently flawed.