I had an old teacher send me all the research of John Russell Holmes that examined a century’s worth of investment data from 1871 to 1971. Isn’t that great? I love reading anything written in a textbook from the 1970s and earlier because the writing style is so clean–everything flows beautifully with the paragraphs neatly bound in a hypothesis, data, interpretation, conclusion format. The only downside is that there is less personality in the writing but this is offset by the elegance of logic and the chance to pick up on the anachronisms of the time. I had no idea the 1890s were called the “Gay Nineties” and the 1950s were called the “Fabulous Fifties” but then textbooks started phasing out the terms in the early 1970s to disambiguate from homosexuality.

Anyway, one of the first things that I learned is that the phrase “long-term investing” should really be used for periods of ten years, though ideally fifteen years or more. Five years is not adequate.

The attractiveness of using 1871 through 1971 as our laboratory is because of the nice quirk that the P/E ratio was the same in 1871 as 1971. The returns during this century, then, are entirely the product of business results–earnings per share growth plus dividend distributions as there were no valuation changes to act as a coefficient of distortion.

And here is what the Holmes study revealed: The average common stock return was 8.4%. You got 4.8% dividends on average, and 3.6% earnings per share growth over the century (note: this was the era before stock buybacks, so the lower dividend yields today encounter a counterbalance in the form of higher expected earnings per share growth resulting from a dwindling share count).

If you calculated the earnings growth plus dividend payments during each of the five two-decade stretches, the results were remarkably consistent. Businesses delivered 7.76% earnings growth plus dividends during the first twenty years, then 8.51% during the next twenty, then 8.25% during the next twenty, then 6.93% during the next twenty, and then 8.83% during the final twenty.

Despite this consistency–and again, I think remarkable is the right to describe it–in the operating results of the businesses during periods that included The Great Depression, WWII, and the post-World War II boom, the stock prices during this century did not mirror the consistency of the underlying businesses themselves.

Of the 96 possible five-year measuring periods during this century, exactly 69 of them delivered returns outside of the expected standard deviation of 5.4% to 11.4%. What does this mean? If you use five-year investment periods as your measuring stock, approximately 71.9% of the time you will either earn returns either excessively better or worse than what the actual performance of the businesses owned would merit.

When you stretch the holding period out to ten years, you get 91 measuring periods and only 50 of the measuring periods fall in the extremes, or 55.0% of the outcomes will be disproportionate (i.e. 45% of the time, you get what you deserved; 55% of the time, you will get much more or much less than you deserve). But really, fifteen years should be the definition of long-term investing. By then, there are 86 measuring periods, and only 35 fall in the extremes. In other words, you get what you deserve 59.3% of the time, and you get better or worse than you deserve 40.7%.

Now, of course, investors aren’t going to be too worried about the risk of getting better than they deserve (this would be a welcome risk, indeed!) but it is joined at the hip to getting worse than you deserve because periods of poor returns aren’t only the production of transitioning from fair value to undervalue but rather overvalue to undervalue.

The risk of getting worse than you deserve, defined as getting worse than 5.4% when a basket of stocks is delivering earnings growth plus dividend distributions of 8.5%, has a 37.5% chance of occurring during a five-year holding period, a 31.9% chance of occurring during a ten-year holding period, and a 22.1% chance of occurring during a fifteen-year holding period.

My lesson from the Holmes research? These people who sell stocks in response to price declines that occur over a period of month are playing a casino game. What surprised me is how much even a five-year holding is a game of chance for total returns–it’s basically a spin of the Roulette wheel. And the difference between five years and ten years as a holding period were not as drastic as I would have intuitively guessed. Although I suppose when I run the numbers of years like 1998-2008, 1999-2009, or 2000-2010, I shouldn’t be surprised by those odds of three years in a decade delivering worse performance than expected by measuring business results. It seems that you don’t really reach a substantial likelihood of having your investment returns match business performance results until you’ve held the stock for fifteen years, and even then, there is a still a one-in-five chance you will get less than expected.

There are also other ways you can minimize the ill effects of this study’s findings: (1) pick up the skill of identifying undervalued stocks; (2) find companies with a substantial likelihood of earnings per share growth over 8.5% so that even “disappointments” raise the floor of your returns; (3) own non-correlated businesses so that poor economic conditions for one industry benefit the other; and/or (4) own non-correlated assets like bonds, a strategy that I haven’t discussed because the bond terms available to U.S. investors have been terrible since 2008.

But the results of five years incorporate a fair amount of luck–or, put another, are the product of investor sentiment to such a degree that it overwhelms business performance. Seven times out of ten, you will either think you are a terrible stock picker or a brilliant stock picker when the actual performance of the stocks you own don’t actually suggest either. The temporal disconnect between business performance and deserved stock price changes in response to that business performance is one of the best arguments in favor of widespread portfolio diversification.