Perhaps the most important takeaway from the decade that just came to a close is that U.S. equities’ return was nearly double their long-term average.

That’s the good news.

The bad news is that the decade to come is likely to be a lot less rosy.

Over the decade ending Dec. 31, the U.S. stock market on a dividend-adjusted and inflation-adjusted basis produced an 11.5% annualized return, according to data from Yale University professor Robert Shiller.

Prior to this past decade, the average back to 1871 was 6.6% annualized. In other words, stocks over the past decade beat their historical average by nearly five annualized percentage points.

No other major asset class deviated so much from its long-term average. Small-cap stocks over the past decade, for example, lagged their historical average (back to 1926) by just 1.5 annualized percentage points. Long-term Treasury bonds outperformed their historical average by just 0.3 annualized percentage points.

The reason stocks’ outperformance over the past decade bodes ill for their future is that equities exhibit a strong tendency to regress to their mean over approximately 10-year intervals.

Take the bottom of the financial crisis, for example: At that point stocks’ trailing 10-year return was minus 5.9% annualized on an inflation- and dividend-adjusted basis, according to the Shiller data.

From the perspective of regression to the mean, therefore, it shouldn’t have come as a big surprise that the following decade would be such a good one for equities.

One comeback to this argument is to question why I am choosing to focus on 10-year intervals over which regression to the mean supposedly takes place. Why not 20 years, for example? Wouldn’t we reach just the opposite conclusion if I focused on 20-year intervals?, since the top of the Internet bubble equities have performed well below average: 3.9% annualized after adjusting for dividends and inflation, nearly three annualized percentage points below their long-term average.

The answer to this comeback is that, at least historically, regression to the mean has been more pronounced over 10-year intervals than 20-year periods. To show this, I conducted a hypothetical test using Shiller’s data back to the late 1800s, tracking the performance of two strategies.

At the beginning of each decade, the first strategy became 100% equities if stocks’ return over the trailing decade was below average, and 100% in 10-year Treasuries (or equivalent) if stocks’ trailing 10-year return was above average. The second strategy was identical, except it only made transactions at the beginning of every other decade, based on stocks’ trailing 20-year return.

The accompanying chart reports the results. Since 1891, the 10-year regression strategy produced a 5.5% annualized return, versus 4.2% for the 20-year strategy. To be clear, I am not recommending that you mechanically follow this 10-year strategy. There no doubt are many ways of optimizing it, such as not waiting a full 10 years before betting that stocks will correct any extreme above- or below-average return. But the point of my comparing these two strategies’ return is to show that 10 years is a better regression interval to bet on than 20 years.

ts_chart_annualized-return (1)

Notice carefully that the subdued forecast for equities’ future returns is over the next 10 years. Even if it turns out to be correct, it tells us nothing about the path they take in getting there. It could be, for example, that stocks’ bull market continues for a while longer, before succumbing. Alternately, equities could soon enter into a severe bear market from which they only partially recovers over the next decade.

This crucial distinction between the short- and long-term trends brings to mind a metaphor used by Ben Inker, head of asset allocation at Boston-based money manager GMO. He analogizes the market to a leaf in a hurricane, saying that “you have no idea where the leaf will be a minute or an hour from now. But eventually gravity will win out and it will land on the ground.”