My third book, “Rational Investing in Irrational Times,” was published in 2002. The book presented and analyzed 52 common mistakes investors make. By the time the updated version of the book was published in 2011, as “Investment Mistakes Even Smart Investors Make and How to Avoid Them,” the list of mistakes had grown to 77 (a full catalog of those errors is provided at the end of this article). Unfortunately, if I were to again update the book, the list would be even longer.

One of the additions I would make involves a mistake that, while expected among some individual investors, is surprisingly common among many professional advisors. Specifically, it’s the error of using historical stock and bond returns as the best estimator of their future returns. The rationale for the behavior is typically that forecasts other than those using historical returns are nothing more than opinion. Historical returns are then employed as inputs for planning tools (such as Monte Carlo simulations) used to develop strategies for retirement, including the amount of savings required, a safe withdrawal rate for determining spending and the appropriate asset allocation.

To explain why using historical returns is an error, I’ll review the academic literature on the subject. Investment decisions should be based not on opinions, but on solid, peer-reviewed academic research. Before discussing the literature, the following example addresses the issue of whether using historical returns is even a logical approach. Unfortunately, as you’ll see, the use of historical returns in retirement planning is highly likely to lead to failure (meaning investors outlive their assets) because, for both stocks and bonds, it leads to a forecast that’s extremely unlikely to occur. The reason is that stock valuations are much higher than they have been in the past, and bond yields are much lower.

From 1926 through 1948, the S&P 500 produced an annualized nominal return of 6.3% and an annualized real (inflation-adjusted) return of 5%. Relying on the historical record would lead you to forecast the same returns going forward. From 1949 through 1999, the S&P 500 returned 13.7% in nominal terms and 9.8% in real terms. The return over the full period, from 1926 through 1999, rose to 11.3% in nominal terms and 8.1% in real terms. Let’s see if it makes any intuitive sense to increase your forecast from 6.3% to 11.3% in nominal terms and from 5% to 8.1% in real terms because nominal returns were 13.7% and real returns were 9.8% over the prior 51-year period.

On January 1, 1949, the price-to-earnings ratio (P/E) of the S&P 500 was 6.6. That produces an earnings yield (E/P) of 15.2%. The Shiller CAPE 10 stood at 10.1, producing an E/P of 9.9%. On January 1, 2000, the P/E ratio stood at 29.0, producing an E/P of 3.4%, and the Shiller CAPE 10 stood at an even higher 34.4, producing an E/P of 2.9%. Does it make sense to forecast a nominal return of 6.3% and a real return of 5% when the E/P is 15.2%, and also to forecast a nominal return of 11.3% and a real return of 8.1% when the E/P is 3.4% and the Shiller E/P is just 2.9%? The answer is an obvious: no. Yet, that’s what you do if you rely on historical returns.

Common sense, which is all-too-uncommon, should lead you to conclude that this isn’t logical. Higher prices paid for the same dollar of earnings should lead one to forecast lower, not higher, returns. And the academic literature demonstrates current valuations do in fact provide us with the best estimate of future returns. In other words, high valuations don’t forecast high future growth in earnings (which would be required to generate a high return). Instead, the research shows high valuations predict low future returns. And from 2000 through 2015, the S&P 500 went on to return just 4.1% in nominal terms and 1.9% in real terms. It’s easy to see the damage that could be done if one built a retirement plan that relied on a nominal return forecast of 11.3% and a real return forecast of 8.1%.

The following is an even simpler and clearer explanation of why using historical returns is illogical. From 1926 through 2015, the return on long-term (20-year) Treasury bonds was 5.6%. As I write this, the yield on the 20-year Treasury is just 1.7%. Which is the best estimate of future bond returns, the historical return of 5.6% or the current yield of 1.7%? Obviously, the answer is that the current yield is the perfect predictor of the return over the next 20 years, at least in nominal terms.