By Rob Bennett
The Stock-Return Predictor is a stock valuation calculator that performs a regression analysis of the historical stock-return data to reveal to investors the most likely annualized ten-year return for U.S. stocks starting from any of the various valuation levels. The calculator is rooted in a belief in Yale Professor Robert Shiller’s finding that valuations affect long-term returns and that, thus, long-term returns are to a significant extent predictable.
Investors can obtain higher returns at less risk by changing their stock allocations in response to dramatic changes in valuations, according to the calculator. For example, it shows that, at the P/E10 level (“P/E10” is the price of the S&P 500 over the average of its last ten years of earnings) that applied in 1982 (8), the most likely annualized 10-year return was 15 percent real. In contrast, the most likely annualized 10-year return in 2000 (when the P/E10 level was 44) was a negative 1 percent real.
There obviously is not one stock allocation that makes sense both when the likely long-term return is a positive 15 percent and when it is a negative 1 percent. The key to successful investing is getting your stock allocation right and it is not possible to do this following a Buy-and-Hold strategy if the risk/reward ratio for stocks is not a constant but an ever-changing phenomenon.
That said, the Stock-Return Predictor backs up claims that stocks are always the best asset class in the distant long-term. The highest valuation level on record is the 44 P/E10 level that applied in 2000. Even stocks purchased at those prices are likely to generate a solid 30-year return, according to the historical data; the Return Predictor puts the most likely return at the end of 30 years at 4.7 percent real. In 30 years, stocks have gone through a full secular bear/secular bull cycle and the effect of the starting-point valuation level is diminished.
The calculator does not report likely returns for time-periods of less than 10 years. The data shows a small amount of return predictability at five years, but the strength of the correlation between the starting point P/E10 level and the return obtained at Year Five is not sufficiently strong to provide meaningful guidance to investors. The statistical correlation is significant but not great at Year 10, stronger at Year 15 and surprisingly strong at Year 20.
The theory for why returns become predictable only after the passage of 10 years of time is that investor emotion is the dominant influence on prices in the short term but the economic realities assert themselves with the passage of time. The economic realities never come to complete dominance because investor emotion is present in each new short-term time-period. But the influence of any one short-term time-period diminishes as its effect is overwhelmed by the power of the economic realities to pull prices in the direction of fair value.
The 10-year predictions lack precision. The most likely annualized 10-year return starting from the fair-value P/E10 level of 15 is 5.6 percent real. There is an 80 percent chance of a return greater than 2.6 real in those circumstances and an 80 percent chance of a return less than 8.6 real. There is only a 5 percent chance of a return less than a negative 0.39 real and only a 5 percent chance of a return greater than 11.6 real.
The value of the calculator lies in the comparisons it makes possible for an investor seeking to make informed asset allocation decisions. Given the lack of precision in the predicted returns, it would be silly to place much focus on the difference between the results reported for a stock purchase made at a P/E10 level of 14 and those reported for a stock purchase made at a P/E10 level of 16. However, the differences in the results for a P/E10 level of 15 and a P/E10 level of 30 are stark. For an investor who buys into the premises in which the calculator is rooted to fail to make an asset allocation change in response to that dramatic a valuations swing would be foolhardy.
The Stock-Return Predictor challenges fundamental beliefs of the conventional model (the Buy-and-Hold Model) for understanding how stock investing works. The conventional thinking is that stocks pay higher returns than other asset classes because investors are being compensated for taking on more risk. The Predictor shows that, to the contrary, stocks offer the lowest returns when they are most risky and that stocks in some circumstances offer likely long-term returns significantly lower than the guaranteed returns offered by far safer asset classes.
Consider the return paid by Treasury Inflation-Protected Securities (TIPS) in January 2000, the height of the bull. Long-term TIPS were at the time paying a return in the neighborhood of 4 percent real. The most likely 10-year return for stocks was a negative 1 percent real. The investor choosing stocks because they are promoted as always being best for the long run was giving up five full percentage points of return for each of 10 years running, thereby accumulating over 10 years a shortfall of 50 percent of his starting-point portfolio value. Could it be that investors are compensated not for taking on real risk but for taking on perceived risk and that stocks were so popular in 2000 that the perceived risk of owning stocks was essentially zero (and that TIPS were widely perceived as a dangerous choice in comparison to stocks)?
The challenge to conventional thinking may go even deeper than that. Even the biggest stock cheerleaders acknowledge stocks to be a risky asset class. But is that so if long-term returns are highly predictable? If the calculator can be trusted to predict stock returns effectively, an investor who felt confident that he could stick with stocks so long as his 10-year annualized return never dropped into negative territory could invest in stocks heavily so long as the P/E10 level remained at 15 or lower. An investor able to accept a 20 percent chance of a negative 10-year return could invest heavily in stocks so long as the P/E10 level remained at 20 or below.
At P/E10 levels of 10 and below (we may be heading to these valuation levels within the next few years — we have seen them in the aftermath of every other trip to insanely high valuation levels), there are no bad outcomes. The investor who buys stocks when the P/E10 level is 10 has an 80 percent chance of seeing a 10-year annualized return of greater than 7.7 percent real and only a 5 percent chance of seeing an annualized 10-year return of less than 4.7 real.
Norbert Schlenker, a financial planner and co-owner of the Financial WebRing Forum, prepared a graphic comparing 10-year annualized returns for those following a Buy-and-Hold strategy with 10-year annualized returns for those following a Valuation-Informed Indexing strategy (shifting one’s stock allocation pursuant to the results of the Return Predictor). He concluded that: “The evidence is pretty incontrovertible. Valuation-Informed Indexing…is everywhere superior to Buy-and-Hold over 10-year periods.” The one exception identified by Schlenker, the period from the mid-1990s forward, follows the general rule if the effect of the 2008 price crash is considered (Schlenker’s graphic was prepared prior to the crash).
The most shocking implication suggested by the calculator is the possibility that publicization of its results might help lead us to realization of the long-promised but thus far highly elusive dream of an efficient market. If Shiller is right that the concept of overvaluation is a meaningful one, the nominal market price is not the proper “efficient” price. But could it be that the price that the market is trying to tell us is the true market price is the nominal market price as adjusted for the effect of overvaluation or undervaluation?
The market generates the P/E10 figure as much as it does the nominal market price. Perhaps all the confusion over whether the market is efficient or not, and if not, why not, stems from a failure to appreciate that the market is reporting the true market price in two steps.
If a large number of investors were to learn that the value proposition of stocks is a variable thing and that stocks were thus a less attractive asset class for them at some valuation levels than they were at others, each uptick in valuations would bring on stock sales. The stock sales would pull valuations back to fair value. If investors were taught to change their stock allocations in response to valuation shifts, the market price would be self-correcting!
Significant levels of overvaluation or undervaluation would become a logical impossibility and stocks would indeed always be priced right, as efficient market theorists argue they should be in a rational world. The suggestion here is that the only thing that has been keeping us from enjoying an efficient market is the insistence of the believers in the efficient market theory that market timing is not needed for the market to become efficient.
Markets in which investors do not change their stock allocations in response to big valuation shifts cannot achieve efficiency because valuations affect long-term returns and it is irrational for investors to ignore this reality when setting their stock allocations. The market is not automatically efficient. But investors can make it efficient if they are willing to change their stock allocations in response to valuation shifts.