The method the brokerage firm used to generate projected returns for my hypothetical portfolio—extrapolating past performance—is common. It is logical, and, in some cases, defensible. It is also, unfortunately, dangerous.
As the fine print suggests, past performance doesn't guarantee future results. What it usually does guarantee is that we will expect more of the same going forward. The human brain being what it is, most of us develop expectations about the future by observing the past—by, as Warren Buffett puts it, looking in the rearview mirror. Unfortunately, in the market realm, most of what we can easily observe—periods of three years, five years, 10 years—are often too limited to be useful as a forecasting tool. Why? Because market cycles usually last 30-40 years, not three to 10 years. The images in the rearview mirror, therefore, even over periods that feel like eternity, often bear little resemblance to what lies ahead.
In the financial markets, the "long term" is long. Over the past 200 years, U.S. stocks have, on average, returned approximately 10 percent a year (about 7 percent, after adjusting for inflation). For many of those 200 years, however, stocks have returned nothing—or worse. The fallow periods, moreover, have not just lasted months or years. They have lasted decades. In a 2001 Fortune article (the wisest, pithiest snapshot of market history I've ever read), Buffett observed that the 20th century encompassed three major bull markets in which the Dow jumped more than 11,000 points and three major stagnant markets in which the Dow lost 292 points. The three bull markets, in aggregate, lasted 44 years; the three bear markets 56 years. For more than half of the century, in other words, stock performance stank.
The most recent market cycle spanned 34 years, from 1966-2000. The bull phase, the one we all remember, lasted 18 years (1982-2000), and it took the Dow from just over 800 to just under 12,000. The bear phase—the one almost no one remembers—lasted 16 years (1966-1982— 16 years!), and it took the Dow down nearly 20 percent. Lest this tempt you to rush out and buy bonds, average bond returns from 1966-1981 were worse than those on stocks (bonds can be dangerous when inflation is rising, a fact worth remembering now).
Because market cycles are so long, assumptions based on 10-year averages can lead to unfortunate expectations. If one had applied the 10-year analysis to the S&P 500 index from 1974-1983, for example, one might have concluded that the index would deliver low single-digit returns over the next 10 years—an outlook paltry enough to cause one to consider reducing exposure to stocks. In fact, over the next 10 years, the index delivered low double-digit returns (the late '70s and early '80s was a great time to buy, in part because stock performance had been so crappy for so long). Applying the same 10-year historical analysis to the S&P 500 today suggests that, over the next decade, the index will deliver low double-digit returns (thus the 12 percent assumption used by the brokerage firm). Broader history, however, suggests that actual performance will be worse—as it must be, if the 10-year average return is, eventually, to regress back down to the 200-year 10 percent mean.
The problem with most projections based on past performance is that they don't fully account for the length of market cycles and the tendency toward mean-regression. With this in mind, let's look again at the brokerage firm's assumptions
|Asset Class||Est. Annual Return||Est. Long-Term Return *|
Most of the assumptions suffer from the market-cycle and mean-regression problem. If, for example, after two decades of dropping, interest rates either remain where they are or, more likely, begin to regress upward to the mean, bond returns will probably sag (when rates rise, bond prices drop). It is, therefore, hard to imagine that municipal bonds will deliver the same returns over the next 10 years as they did over the last 10, when yields were initially higher and capital appreciation from falling rates juiced up returns. Similarly, after two decades of above-trend stock performance, we're probably due for a decade or two below trend (international stocks, interestingly, are below the mean, so, here, the firm's assumption may be low). It is always possible, of course, that, as the firm implicitly suggests, stock and bond performance for the next 10 years will equal or exceed performance for the last 10. More likely, however, we are in the early years of a 10-year to 20-year "regression" in which the returns on financial assets will disappoint.
So what might alternative assumptions look like? After consulting the work of a few of the gurus of mean-regression—Jeremy Grantham of Grantham, Mayo, Van Otterloo (7-year asset class projections), Robert Shiller of Yale University ( Irrational Exuberance and numerous papers), and Andrew Smithers of Smithers & Co. (article on stock market valuation)—
|Asset Class||Firm Assumptions||My Assumptions|
|T-bills||4.5 percent||4.0 percent|
|Municipal Bonds||7.2 percent||4.5 percent|
|International Stocks||4.5 percent||6.0 percent|
|U.S. Stocks||11.8 percent||2.0 percent|
|Funds of Funds||9.8 percent||5.0 percent|
Do these changes have an impact on the median projected return of my hypothetical portfolio? Yes.
Median annual return using the brokerage firm's assumptions: 9 percent.
Median annual return using my assumptions: 4 percent.
Well, 4 percent a year is a far cry from 9 percent, but the brokerage firm is still going to make me money, right? Yes, probably, but in all likelihood less than that suggested by either of the median estimates. Once again, the wisdom lies in the fine print: "Returns are shown before the deduction of fees … and assume reinvestment of income and no transaction costs or taxes" (my italics). The returns are also shown before the impact of inflation.
Tomorrow, I'll talk about what these absences mean.