I recently discussed in this column some changes in our profession's portfolio management activities, and I've started a longer discussion with my newsletter audience about whether it's time to move beyond the whole Modern Portfolio Theory/Mean-Variance Optimization methodologies toward something, for lack of a better term, I call the New Paradigm (NP). I have no idea what NP will look like, but I outlined what it should accomplish and passed on some ideas that have been floated recently.

A tsunami of emails in response to this heresy suggested that if this were a different place and time, I should be burned at the stake or fed to lions. The emailers accused me of being (gasp!) a market timer: If you're trying to adjust client portfolios based on current valuations, you're "predicting the future." If you invest with managers who aren't precisely confined to style boxes, your clients will fall victims of the dreaded "style drift." Many suggested that we didn't need this discussion, since diversifying along the efficient frontier is the only optimal way to deploy a client's assets--anything else is market timing.

FESSING UP

Interestingly, I received nearly as many responses from advisers timidly fessing up that they've been making small but constant tactical decisions in client portfolios or investing outside the boxes, but were afraid to say so openly for fear of being ostracized by the community. Some reported consistent above-market results, simply by avoiding or seeking out asset classes that have strayed far from their normal value parameters. But the most common claim was that they've been delivering below-market volatility. Based on an informal survey, I would guess that more than half of all advisers are making at least small deviations from a strict MPT/MVO strategy. So the profession is already beyond the current paradigm.

Why stray? As most of you know, Modern Portfolio Theory was first articulated by Harry Markowitz in the 1950s, who suggested going beyond the idea that if you held many different assets, you would experience more stable returns. Markowitz tried to quantify which assets tended to move up or down with or against which other assets--what he called "correlation coefficients."

Thus, government bonds are weakly correlated with stocks--when stocks rise in value, there will often be an associated rise in bond prices--but bonds and stocks don't move precisely in unison. Stocks and cash are hardly correlated at all. Real estate, commodities and the various kinds of domestic and international stocks and bonds all have positive or negative correlations with one another.

Markowitz found that you could combine the expected returns, the expected standard deviations and the correlation coefficients to create portfolio mixes that would theoretically give you more return per unit of volatility than other mixes of the same asset classes. In MVO terms, the portfolio is "optimized" and on the "efficient frontier"--a graph of different asset mixes that produces a curve like a horizontal fishhook with the barb pointing down.

Unfortunately, the problems with this elegant mathematics are becoming more obvious each year. The expected returns and volatility don't ever precisely match the returns and volatility that you actually experience with the portfolio going forward, so the inputs to this MVO process can never be defined in advance.

In addition, correlation coefficients can change over time. David Reilly, chief economist at Rydex, has been offering presentations that include a slide of the efficient frontiers, in retrospect, for different decades, and they're all over the map: Some look like the standard fishhook curve; others resemble a nearly straight line or an obverted fishhook, scattered about different parts of the risk/return spectrum.

Meanwhile, evidence suggests that correlations suddenly shoot up exactly when you need diversification most. Whenever markets tank, domestic and international stocks, bonds, REITs and all other assets tend to get caught up in the same panic selling. The old joke is that the only thing that goes up in a down market is correlations.

BEHIND THE BELL CURVE

Finally, the whole mean-variance calculation rests on an assumption that market returns will be distributed according to the classic bell curve. But a graph of the actual returns of many asset classes shows that a number of them tend to be scattered in the thin tails of their various curves in much greater numbers than the mean and standard deviation alone can account for, looking more like a cluster of low hills than one symmetrically shaped one.

As Bryce James of SmartPortfolios in Seattle notes, based on the traditional standard deviation calculations around the mean return, an event like the 7.18% one-day loss on October 27, 1997 should be a one in 50 billion trading days event. The August 27, 1998 loss of 4.1% represented a one in 20 million event, or one in 100,000 years worth of trading days. This was followed, four days later (August 31, 1998) by a one-day drop of 6.4%. The odds of the two happening in the same month--never mind the same week--based on standard deviation calculations around the mean is approximately one in 100 billion years of trading days. Consider that the universe is estimated to be less than 15 billion years old, and you get some idea of the extent of the "fat tail" problem.