Today, Modern Portfolio Theory is as accepted as air travel. But as with flying, some people question whether MPT is as safe as it seems.
The problem for many investment professionals is that MPT provides a model for economic behavior and is not always reliable.
"The models from the '60s and '70s are a really good starting point. And they suggest a way the investment world should behave under ideal conditions," said William Jacques, chief investment officer at Martingale Asset Management LP, Boston. "However, we know the world is not always rational and orderly, and one can profit from these deviations. Models are useful, like a pi?ata you can beat up on."
The issues with MPT fall into three main areas:
* Is the mean-variance optimizer the best way of calculating the optimal portfolio?
* Are markets efficient?
* Can one pick skilled money managers?
Criticizing model
Since the 1970s, numerous academics and investors have tried to poke holes in Harry Markowitz's mean-variance optimizer, the basic tool used for designing the most efficient portfolio, based on the relative risks and rewards among individual securities.
They have criticized the model for allowing huge errors to crop up from small changes in inputs, for looking at a single period, for not taking long-term wealth needs into account, for assuming static economic conditions, and for assuming rational behavior in an irrational world.
Peter Bernstein, head of the eponymous consulting firm, said in a recent speech that volatility, the proxy for risk used in the optimizer, often falls short of the mark. "Our definition of risk ... is in a sense a floating craps game," he explained. "So if we don't have a generally accepted proxy for risk, that stands up all the time, everything from performance measurement to asset allocation strategies are up for grabs," Mr. Bernstein said.
Dean Barr, chief investment officer of Deutsche Asset Management, New York, said the optimizer "is a great framework to start with, but it falls fairly short in solving some fairly important practical problems."
Yet even detractors acknowledge the mean-variance optimizer was a great contribution to investment practice. Its supporters are effusive: "Mean-variance optimization was just a profound innovation that just completely changed the mindset of investment professionals," said Mark Kritzman, managing partner of Windham Capital Management Boston LLC.
"There have been a lot of enhancements to it, to improve it or make it more flexible or more robust with respect to the quality of data or inputs, but nothing has really diminished its significance," Mr. Kritzman added.
The bottom line, however, is that the optimizer is a tool, and as such requires that reliable data be entered. "If you give garbage in, it will give you garbage out," Mr. Markowitz said. "So using portfolio theory no more guarantees wise investing than using double-entry bookkeeping guarantees that you are able to run a firm."
One common criticism, Mr. Kritzman said, is that changing assumptions in a small way can have a dramatic effect on the recommended portfolio.
But that result is not typical, he said, suggesting that some investors may be using the optimizer in a naive way, by relying solely on historical data. "A lot of people equate mean-variance optimization with extrapolation of historical returns or covariances," Mr. Kritzman said.
But the optimizer doesn't require use of historical prices; instead, one can obtain the implied volatility from options prices, he said.
A more significant criticism comes from Richard O. Michaud, president of New Frontier Advisors, Boston, who dubs the optimizer "an error maximizer." Instead, he has developed a resampling technique that basically takes the average of a series of efficient frontiers.
Other quantitative investment experts speak highly of Mr. Michaud's approach but say it tweaks the model and is not a radical overhaul.
Some critics, such as Mr. Barr, say the optimizer is weak when dealing with asymmetric return patterns - ones that don't fit onto a normal bell-shaped curve. Mr. Markowitz responded that the optimizer has proven robust for the vast majority of situations.
Critics also say the optimizer is flawed because it assumes a static economic environment. In real life, economic conditions are changing all the time.
The problem becomes even more intractable when trying to figure out how to provide enough wealth and income over the entire life of a pension fund, said John Y. Campbell, Otto Eckstein Professor of Applied Economics at Harvard University and a partner at Arrowstreet Capital LP, Cambridge, Mass. He noted that transaction costs don't fit neatly into most optimizers.
The optimizer also does not cope well with illiquid investments such as private equity because of problems in capturing the timing of cashflows, noted Grant Gardner, director of research for Frank Russell Co., Tacoma, Wash.
Amazingly, Mr. Markowitz outlined a possible approach to multiperiod analysis in his 1959 book, "Portfolio Selection." And, in a draft paper submitted recently to the Financial Analysts Journal, Mr. Markowitz and Erik L. van Dijk, a Dutch academic and managing director of Compendeon BV, Amsterdam, wrote that the traditional optimizer still does an extraordinarily good job of designing the optimal portfolio, based on a simplified model.
Efficient markets?
One of the biggest gripes about MPT is that many proponents - notably, not Mr. Markowitz - postulated that because markets are so efficient, money managers simply cannot beat them, at least with any measure of consistency. Yet active managers and behavioral finance experts have shown in example after example that investors often behave irrationally, as proof that smart investors can beat the market.
For example, investors often display overconfidence, extrapolating from recent experience. Or they seek to avoid losses because people are more likely to regret mistakes than remember correct decisions. People also tend to look for patterns, even where they don't exist, said Arnold S. Wood, Martingale's chief executive officer.
In a 1999 paper, Shlomo Benartzi, assistant professor at the University of California at Los Angeles' Anderson School, and Richard H. Thaler, the Robert P. Gwinn Professor of Behavioral Science and Economics at the University of Chicago's Graduate School of Business, demonstrated that defined contribution investors shied away from stocks when considering one-year returns - demonstrating "myopic loss aversion." Those same investors, however, would steer more of their assets into equities when shown long-term rates of return.
Elsewhere, the two academics have shown that 401(k) investors tend to divide their assets evenly over the number of investment options offered, regardless of their risk tolerance or retirement needs.
"The main message of behavioral finance is that individual investors haven't read as much Markowitz as they should," Mr. Thaler said in an interview. Even sophisticated institutional investors sometimes miss the boat, he said, noting the strong home-country bias that exists for virtually all pension funds around the world.
Yet, can you make money from behavioral finance?
"All the behavioral finance stuff describes a lot of things that probably do occur, but it's not a very practical platform for trying to make investment decisions," said Russell's Mr. Gardner.
Added Rob Arnott, managing partner of First Quadrant Corp., Pasadena, Calif.: "I've seen very little in behavioral finance that tells us what inefficiencies are going to be profitable."
The tech bubble
Perhaps the biggest recent challenge to efficient-market theory comes from the technology stock bubble of the late 1990s.
Initial public offerings that doubled in price on the first day, astronomically high tech stock valuations that totally ignored fundamentals, certainty over growth prospects for hot stocks - all are hallmarks of a market out of touch with reality.
These issues raise the question of whether Nobel Laureate William F. Sharpe was correct in saying the stock market is the largest single influence on returns, Mr. Bernstein said in his recent speech. "How, with a market, can there be a bubble? How does this fit into the notion of the rational market?"
Mr. Bernstein said the theoretical flaw is in positing that markets are in equilibrium, that correct prices for stocks exist. "This is a condition that can prevail only in the absence of uncertainty, only in a static rather than a dynamic environment, only when agents make decisions based on perfect foresight. Take away any of those conditions, and there is no such thing as equilibrium."
"In an efficient market ... everybody gets it right immediately," Mr. Bernstein continued. "In the real world, however, everybody gets it wrong. Prices are moving all the time. ... (T)here is no such thing as an equilibrium price."
Mr. Sharpe takes issue with both scores. He thinks "the market is almost efficient" and the challenge is to find where prices vary from efficiency. Stock prices reflect the consensus, and "those are the best estimates you can find," he said in an interview.
Measuring performance
Despite all the performance measurement tools that have evolved out of MPT, H. Russell Fogler, principal of Fogler Research & Management Inc., Gainesville, Fla., said: "The single biggest issue today (facing MPT) ... is how long does it take to measure a money manager?"
Statistically, it can take 80 years or better to know if that manager has demonstrated skill instead of luck, he said. "If you can't measure a manager with all these tools in a finite period because residual risk is so great relative to alpha, then you've got a strange situation," he said.
But statistics don't tell the whole story, said Andrew Lo, Harris & Harris Group Professor at the Massachusetts Institute of Technology and director of the MIT Laboratory for Financial Engineering, Cambridge, Mass.
A combination of "statistics and insight and modern financial analysis" helps one select managers, he said.
Russell's Mr. Gardner also looks to see if the market rewards a manager's particular approach.
Mr. Fogler added that the measurement problem is akin to Heisenberg's uncertainty principle, which states that in looking at subatomic particles, we can never know their precise conditions: "If you try to understand how the parts of an atom move, to understand that you have to insert a probe. As soon as you do that, it changes the behavior of the system.
"It's kind of the same thing in money management: If you really want to measure a manager, you have to get a benchmark. If you get a benchmark, you have now changed the behavior of the manager, because he can't go all over the place," Mr. Fogler explained.
Robert Kirby, senior partner of Capital Group Partners, Los Angeles, bemoaned that consultants have focused too heavily on managers' benchmark risk and not on performance: "If you believe you have picked a good manager, he should have as much benchmark risk as possible."
Focusing on tracking error from the benchmark is like "picking somebody to run a 100-yard dash and then giving them a bowling bowl in each hand before he starts, he said.