Long-term Trends and Short-term Variations in Valuation Ratios

Does decomposition of widely used valuation ratios into components that reflect long-term trend and shorter-term variation from trend reveal predictability? In their November 2009 paper entitled “Do Decomposed Financial Ratios Predict Stock Returns and Fundamentals Better?”, Xiaoquan Jiang and Bong-Soo Lee explore decomposition of the dividend-price, earnings-price and book-to-market ratios into stochastic trend and cyclical components. The stochastic trend component measures variations in longer-term trend (fundamental structural changes), while the cyclical component measures shorter-term deviations from this trend. The study employs rolling four-quarter sums of dividends and earnings, with the latter smoothed via a ten-year moving average, and accounting data to model older book values. Using quarterly S&P 500 Index returns and valuation metrics for the S&P 500 over the period 1926-2008, they find that:

• The stochastic trend components for all three valuation ratios slope downward over the sample period, suggesting structural change. However, the cyclical component reverts to a mean of approximately zero.
• Without decomposition, the three valuation ratios relate positively to future stock market returns, significantly (weakly) at a long (short) horizon. Predictive power during 1926-1951 is much stronger than that during 1952-2008.
• With decomposition, in sample (potentially incorporating look-ahead bias):
  • The stochastic trend components relate negatively to future stock market returns at long horizons, perhaps reflecting a longer-term persistence of returns.
  • The cyclical components of all three valuation ratios relate positively to future stock market returns, perhaps reflecting shorter-term (business-cycle) mean reversion.
  • The opposing relationships help explain why valuation ratios without decomposition exhibit little or no predictive power at short horizons and some predictive power at long horizons.
  • Used distinctively but in combination, the power of the components to predict stock market returns peaks at an horizon of about 12 quarters, with explanatory power over the entire sample period of 45%, 72%, and 77%, respectively, for the dividend-price, earnings-price and book-to-market ratios. At a one-quarter horizon, the respective explanatory powers are 5%, 13% and 13%.
• With decomposition, out of sample:
  • Some of the cyclical components of valuation ratios display statistically significant predictive power for stock market returns, but the stochastic trend components do not.
  • The cyclical component of the earnings-price ratio is the best predictor of stock market returns.
• The cyclical component of the dividend-price ratio predicts the dividend growth rate, especially at short horizons. The cyclical (stochastic trend) component of the earning-price ratio predicts the earnings growth rate at short (long) horizons. The stochastic trend component of the book-to-market ratio predicts accounting returns at long horizons.

In summary, evidence indicates that decomposition of valuation ratios into long-term trends and shorter-term (business cycle) variations may substantially enhance their abilities to predict stock market returns.

You May Also Enjoy...

• The Accuracies of Different Valuation Multiples (Ratios)
• Don Hays on Long-term Cycles and Shorter-term Trends
• Volatility and Valuation Ratios
• Stock Market Valuation Ratio Trends
• Long-Term Outperformance from Trends Defined by Moving Averages