For years, financial advisers have used modern portfolio theory to build diversified portfolios. The industry was happy with MPT until the bottom fell out of the financial markets in 2008.
The problem, however, isn’t with MPT but with the way it has been implemented.
Many advisers don’t use a sufficient number of asset classes in their portfolios. Adding alternative asset classes, such as absolute return, commodities, real estate and managed futures, makes a positive difference.
“Thin-slicing” traditional asset classes helps, as well.
Most advisers haven’t developed a solid process for developing expectations about the future returns, risk and correlations of the asset classes that they use. They get poor results using MPT because they plug flawed inputs into their optimizers.
The first problem is the use of long-term historical averages to derive these inputs. Long-term averages mask reality.
For example, in his book “Stocks for the Long Run: The Definitive Guide to Financial Markets Returns and Long Term Investment Strategies” (McGraw-Hill, 2007), Jeremy Siegel said that the long-term return of the U.S. stock market is 7%. In fact, it is almost never 7%, even over fairly long time periods, though it does pass through 7% occasionally on its way up and down.
Risk changes over time, too. Using static long-term standard deviations to assess risk doesn’t allow you to take this into account.
In fact, when measuring risk, using standard deviations is a problem because it treats upside volatility the same as downside volatility. Taking only downside risk into account in the portfolio construction process produces better results.
Correlations also change over time. Day to day, markets appear far less correlated than they are when they are falling.
This fact is masked by long-term historical averages.
Looking ahead works much better. Results can be improved by using future-looking return, risk and correlation expectations for each asset class.
This means that allocations will change as expectations change, resulting in a more dynamic allocation process.
Let’s talk about implementation of the asset allocation strategy.
Skill exists. This is not a controversial idea outside the world of investment management.
Yet some maintain that skill doesn’t exist among investment managers — or if it does, we have no reliable way of finding it. This view is simply wrong. The problem is in how advisers attempt to identify skill.
Many of the venerable practices that they use to try to find skilled investment managers are seriously flawed. These flaws make it difficult for advisers to generate solid returns for their clients.
One of the biggest problems is the widespread use of style boxes.
Advisers look for managers who fit neatly into a box and then punish them if they commit the sin of “style drift.” The rationale for this approach is that drifting managers mess up our perfect asset allocation pies.
As it turns out, our asset allocation pies aren’t so perfect after all. For every portfolio that sits on the efficient frontier, others, with a different asset class composition, sit essentially on the same point.
An optimized portfolio is only as good as the inputs we use to create it. Because our inputs are never exactly on target, demanding slavish adherence to our asset allocation strategies is an exercise in form over substance.
As it turns out, most good managers don’t pay much attention to style drift anyway. At least that was the finding of Russ Wermers, a professor at the University of Maryland, when he studied the behavior of active mutual fund managers.
Mr. Wermers also found that manager skill isn’t limited to one style box.
Managers who exhibited the greatest style drift actually produced the best performance. A skilled manager can find good stocks in any style box.
Another benefit of allowing managers to drift a bit is that collectively, they provide a level of asset allocation thinking in addition to whatever you may come up with on your own. You might think you are pretty good at asset allocation, but you aren’t perfect.
Everyone can benefit from the added intellectual horsepower that comes from a stable of skilled managers.
Another practice that leads to poor manager selection is the use of a single benchmark to measure manager skill. If managers don’t pay any attention to style boxes and the best managers are associated with high levels of style drift, why do we think that comparing managers to a single index will tell us very much about how skilled they are?
Fielding Best Team
Now let’s talk about combining managers in portfolios. The legendary football coach Knute Rockne said: “As a coach, I play not my 11 best but my best 11.”
This is how we should build portfolios. But most advisers build portfolios by bringing together a collection of star performers to represent each asset class.
This isn’t a good idea.
If you combine a group of “best of breed” managers without considering how they will interact together in a portfolio, they are likely to have similar performance characteristics.
For example, if you choose them during a period when small-cap stocks have been doing better than large-cap stocks, and growth has been outperforming value, you are likely to get a portfolio full of managers who are “growthier” and more heavily invested in small-cap stocks. When large-cap and value come back into favor, all your managers will tend to underperform at the same time.
The MPT tool isn’t broken but has been used ineptly by many practitioners. There is no way to avoid declines in portfolio value entirely when markets become difficult.
But the good news is that you can still do very nicely for your clients, even during difficult times, if you build diversified portfolios and manage them in an intelligent and disciplined fashion.
Scott A. MacKillop is president of Frontier Asset Management LLC.