Shiller’s Mistaken Understanding of Why Only Long-Term Returns Are Predictable


Valuation-Informed Indexing #188

by Rob Bennett

Yale Economics Professor and Nobel Prize Winner Robert Shiller changed our understanding of how stock investing works in a fundamental way with his finding that valuations affect long-term returns. His finding is so “revolutionary” (his word) that few of us today yet appreciate all of the many far-reaching implications of the breakthrough discovery. That includes Shiller!

A Look Back At Warren Buffett’s Legendary Washington Post Investment

Berkshire Hathaway Warren BuffettThroughout his long investment career, Warren Buffett has made thousands of investors and bought tens of businesses. However, a couple of examples really stand out. Q1 2020 hedge fund letters, conferences and more These companies stand out for a range of reasons. The Nebraska Furniture Mart, for example, stands out because it was such a Read More

Many people have a hard time understanding how the same 140 years of historical stock-return data could show both that short-term timing (changing your stock allocation with the expectation that you will see benefits from doing so within a year or so) NEVER works and that long-term timing (changing your stock allocation in response to a big valuation shift with the understanding that you may not see benefits from doing so for as long as 10 years) ALWAYS works. It’s a counter-intuitive finding. Why should stock returns that are so unpredictable in the short term become so predictable in the long term? Aren’t long-term results really just a string of short-term results?

Shiller has offered the following words in explanation: “How can it be that one-year returns are so apparently random and yet ten-year returns are mostly forecastable? In looking at one-year returns, one sees a lot of noise. But over longer time intervals the noise effectively averages out and is less important.”

I don’t find that a convincing explanation. There’s a measure of truth in it. But it is not a clear description of the dynamic that causes short-term predictions to be so much less effective than long-term ones.

Shiller’s statement is actually very much in tune with statements often made by Buy-and-Holders. Buy-and-Holders use the average long-term U.S. return of 6.5 percent real as their default assumption for the return that will apply on a going-forward basis. They then explain to investors that they may well not experience the average return in the first year in which they invest, or in the second, or in the third. The assure the investors that they will see that return if they stay the course, if they stick with their plan. They use the same term that Shiller uses — Noise — to describe derivations from the default return of 6.5 percent. The U.S. stock market delivers a long-term return of 6.5 percent real, the argument goes. You will see that return yourself if only you give the market time to cancel out noise coming from one direction (returns below 6.5 percent real) with noise coming from the other direction (returns above 6.5 percent real)

But Shiller cannot possibly be saying the same thing as the Buy-and-Holders are saying. All of his work points in the opposite direction. The core principle of Buy-and-Hold is that market timing is a bad idea. Shiller says that, when stock prices reach insanely dangerous levels, the market becomes a Ponzi scheme and investors must lower their stock allocations. His message is not that things average out over time. His message is that stocks are a far more risky asset class at times when prices are high than they are when prices are moderate or low.

The term “noise” suggests a meaningless factor, a factor that need not be given much consideration. The Buy-and-Holders are using the term properly, Shiller is not.

Buy-and-Holders believe that the returns of greater than 6.5 percent real and the returns of less than 6.5 percent real pop up randomly. Returns are determined by unforeseen economic and political events, in the Buy-and-Hold Model. Thus, deviations from the normal return cannot be predicted. The deviations are noise, distractions from the reality you should be keeping in mind at all times, the reality that the average long-term return is 6.5 percent real.

Under Shiller’s model, the deviations are NOT random and meaningless events. The reason why returns are predictable in the Shiller model is that the model posits that it is investor emotions that are the primary determinant of price changes. Emotional extremes in one direction beget in time emotional extremes in the opposite direction. High prices increase the probability of price drops and low prices increase the probability of price rises. There’s nothing “noisy” (or random, or meaningless) about this process. It is the essence of how stock investing works, according to the Shiller model.

Shiller predicted to the Federal Reserve in October 1996 that investors buying stocks at the insanely high prices that applied at the time would live to regret doing so within 10 years. He was off by two years. The crash came in September 2008.

His long-term prediction worked. Shiller showed that he possesses a far keener understanding than his critics of how the market works.

But is it right to call the wild price jumps we saw in 1997, 1998, 1998 and 1999 “noise.”

I say “no.” Those jumps were not noise. Those jumps were part of the process that caused Shiller’s long-term prediction to come true. Each insane price jump added to the long-term forces (the market’s desire to continue to function, which required that it eventually set prices properly) that pulled prices down so dramatically in 2008.

Short-term predictions don’t work because investor emotions are irrational. Investor emotions are the dominant influence on price changes and irrational factors cannot be effectively predicted. We can not identify precisely when price trends will change because there is no logic to the process by which they change.

Long-term predictions work because in the long-term the market must set prices effectively or collapse altogether. It is the very purpose of a market to set prices properly. There IS logic to the process by which prices are set properly. So that process IS predictable (just not precisely so).

I am not making a mere semantical distinction here. If investors are to gain confidence in their ability to predict long-term returns effectively, they must understand how one goes about doing it. We don’t gain the ability to predict long-term returns by tuning out “noise.” We gain the ability to predict long-term returns by accepting that it is investor emotions that determine short-term price changes and that rational forces always overpower such emotionalism in the long-term.

The series of short-term results really do determine the long-term result, just not in the manner in which we might first imagine them to do so.

Rob Bennett has recorded a podcast titled Bogle on Minsky: We Have Met the Market and It Is Us. His bio is here.