Gallery:Step 4|Step 4: Time Pickers

Forecasting Stock Market Returns: Why We Don't Do It

Gallery:Step 4|Step 4: Time Pickers

A recent whitepaper from Vanguard1 analyzed the impact of sixteen different variables on subsequent equity returns. They included everything from the ten-year smoothed price-to-earnings ratio to consensus GDP growth. They even threw in annual rainfall which they referred to as a reality check because any variable that shows comparable explanatory power (6% of equity returns) can be safely ignored. Not surprisingly, eight of the sixteen variables showed equivalent or lower explanatory power than rainfall. These included trailing 10-year stock returns, GDP growth trend, earnings growth trend, and the 10-year Treasury yield. The variable that showed the highest explanatory power (43% of equity returns) was the ten-year smoothed price-to-earnings ratio which was first documented by Professor Robert Schiller of Yale University, the author of Irrational Exuberance. Nevertheless, 43% still leaves the majority of the returns unexplained, so investors should be extremely wary of relying on it as the basis for an asset allocation decision.

To Vanguard's credit, they do not state a single value as their forecasted equity return for the next ten years. Using their proprietary model, they display it as bell-shaped histogram that assigns probabilities to different ranges of returns. For example, they assess about a 24% probability that the nominal equity return will fall between 4% and 8%. A slightly lower probability is assigned to the 8-12% range.

IFA has consistently advocated a "no forecasting" approach to the market other than what we see in the historical data. The one data point that we have consistently seen is that the market does a good job of setting prices so that investors are rewarded for the risk they take. What goes into those prices is unknowable because it includes the information and forecasts of more than ten million investors per day.

 

 

Explanation of the Variables Tested in the Chart Above

  1. P/E 10 (Schiller CAPE) is the price-to-earnings ratio based on the prior ten years of earnings which adjusts for anomalies in the earnings cycle.
  2. P/E1 is the straightforward price-to-earnings based on trailing 1-year earnings.
  3. Government Debt/GDP is currently running around 73% (See this WSJ article), and is definitely on an increasing trend. Nevertheless, it is nowhere near its historical post World War 2 high when the market had a tremendous rally.
  4. The consensus building blocks model is based on the Gordon dividend growth model which uses the current dividend yield and the consensus estimate for earnings growth rates.
  5. The trend building blocks model is similar to the consensus building blocks model, but it uses the earnings growth rates of the prior ten years.
  6. Dividend yield is the trailing 1-year dividend divided by price.
  7. The Fed Model is a market-timing (or tactical allocation) technique based on the difference between the US stock earnings yield and the long-term government bond yield.
  8. Rainfall is the "reality check" variable.
  9. Trailing 10-year stock returns tests the hypothesis that after a strong bull run the market is "due" for a correction or a reversion to the mean.
  10. Trend GDP growth is the 10-year GDP growth rate.
  11. Trend earnings growth is 10-year corporate earnings growth rate.
  12. The 10-Year Treasury yield.
  13. Corporate profit margins are the percentage of GDP that is represented by corporate profits.
  14. Trailing 1-year stock returns shows that there is absolutely zero relationship between this year's return and last year's return.
  15. Consensus earnings growth is approximated by the trailing 3-year average growth rate.
  16. Consensus GDP growth is approximated by the trailing 3-year average growth rate.

 


1Davis, Joseph, Charles Thomas, and Roger Aliaga-Díaz, 2012. Forecasting Stock Returns: What Signals Matter, and What Do They Say Now? Valley Forge, Pa.: The Vanguard Group.