Do Price Multiples Predict Market Returns?
A large body of work demonstrates that price multiples, such as the dividend-to-price ratio, predict stock returns. As a result, modern asset pricing theory increasingly incorporates time-varying expected returns. The majority of the empirical work underpinning these findings uses U.S. stock market data going back to 1926.
Benjamin Golez and Peter Koudijs contribute to the literature on return predictability with their January 2017 study, “Four Centuries of Return Predictability.” They examined whether dividend yields predict returns in a sample that covers four centuries of data, going back to the stock market’s earliest years in the 17th century.
Thus, they provide an out-of-sample test on whether results that hold in the recent U.S. period are generalizable to other times and places. Their sample covers annual stock market data for the most important equity markets of the last four centuries: the Netherlands/U.K. (1629‒1812), U.K. (1813‒1870) and U.S. (1871‒2015).
They analyzed the data for each subperiod individually and for the sample as a whole, asking whether the dividend-to-price ratio forecasts returns over the succeeding one-, three- and five-year periods.
Summary Of Finds
- Annual nominal returns are 8% on average, and vary between 6% in the early data and 12% in the recent U.S. period.
- While inflation is much higher after 1945, average real returns across periods are more similar, varying between 6% and 8%.
- Returns are more volatile after 1870; the standard deviation of real returns is roughly 50% higher compared with the earlier periods.
- Estimated risk premiums were relatively low in the early part of the data, 2 to 3%, in comparison with 6 to 8% in the U.S. after 1870.
- Annual real dividend growth rates are around 2% on average. In the 17th and 18th centuries, they are below 1%, picking up in the 19th century, with an average growth rate of 5% before falling to about 2% in the later part of the data.
- For the entire period, most of the real return to investors has come from dividend yields, with 37% coming from price appreciation. However, this has changed in the most recent period, where price appreciation accounts for 57% of real returns.
- With 384 annual observations, the authors were able to reject the null hypothesis of no return predictability over both short and long horizons—dividend yields do predict returns.
- Expected returns are time-varying and related to the business cycle, with expected returns increasing in recessions.
- The dividend-to-price ratio predicts dividend growth rates, but only for the period before 1945. The dramatic fall in the propensity of firms to pay dividends may explain this change. During the 17th and 18th centuries, firms paid out close to 100% of their earnings to shareholders. In 1945, this number was still around 80%. By 1982, however, the dividend-to earnings ratio had fallen to approximately 45%.
The authors concluded that their results “indicate that return predictability from dividend yields has been a robust characteristic of financial markets over the last four centuries. Our findings are robust to a number of statistical tests proposed in the literature, including Monte Carlo simulations. … The implied variation in expected returns lines up well with the business cycle, with on average high returns following downturns. This is true for both early and more recent data.”
Wesley Gray and Jack Vogel, authors of the 2012 paper, “Enhancing the Investment Performance of Yield-Based Strategies,” contributed to the research on the ability of dividends to predict returns by examining different yield metrics.
The Power Of Dividend Yield In Predicting Returns
Gray and Vogel began by noting that the percentage of firms paying a dividend has declined from 63% in 1972 to 36% in 2011. They also cited the 2001 study, “Disappearing Dividends: Changing Firm Characteristics or Lower Propensity to Pay?” by Gene Fama and Ken French, which found that even after controlling for firm characteristics, firms have become less likely to pay dividends. One reason for the disappearing dividends is that firms have increasingly substituted share buybacks for dividends.
And Gustavo Grullon and Roni Michaely, authors of the 2002 paper, “Dividends, Share Repurchases, and the Substitution Hypothesis,” found that: “Repurchases have not only become an important form of payout for U.S. corporations, but also that firms finance their share repurchases with funds that otherwise would have been used to increase dividends.” They concluded: “Firms have gradually substituted repurchases for dividends.”
To address the problem of the disappearing dividends, Gray and Vogel examined four metrics to see if they had predictive value:
- Dividends — DIV
- Dividends plus repurchases — PAY1
- Dividends plus net repurchases (repurchases minus equity issuance) — PAY2
- Dividends plus net repurchases plus net debt paydown — SH/YD
Their data covered the period 1971 through 2011 for the largest 2,000 stocks. Following is a summary of their findings:
- Controlling for exposures to the market, size, value and momentum factors, DIV and PAY1 strategies have no alpha (excess return) after controlling for either the three- or four-factor models—they don’t generate statistically reliable excess risk-adjusted returns.
- Regardless of the yield metric chosen, the predictive power of separating stocks into high- and low-yield portfolios has lost considerable power in the last 20 years. In the latter half of the sample, from 1992 through 2011, DIV loses any forecasting ability it might have had in the previous time period.
- Splitting a yield category by payout percentage doesn’t improve risk-adjusted performance—firms with higher payout ratios don’t earn higher risk-adjusted returns.
Poor-Risk Adjusted Bet
Interestingly, while the authors didn’t find that buying high-yielding stocks is a good “bet,” they did find that buying low-yielding securities is a poor risk-adjusted “bet.” They also found that there was information contained in the other metrics: PAY2 and SH/YD provided economically and statistically significant alphas. Using these metrics, the lowest-yielding quintile stocks produced statistically significant negative alphas, and the highest-yielding ones produced statistically significant positive alphas.
The dividends plus net repurchases plus net debt paydown metric—SH/YD—produced the most robust results over the full period. The authors concluded: “Our evidence corroborates what previous authors have concluded: dividend yield is no longer an effective metric to predict future returns. We also find evidence that more holistic metrics of yield have stood the test of time, but even their predictive ability has fallen.”
Gray and Vogel’s findings contribute to the literature that demonstrates that dividend strategies are neither good substitutes for safe income nor a reliable way to generate market-beating returns. However, they also show that the dividend yield (with the appropriate adjustments) does have explanatory power in terms of future returns.
This commentary originally appeared April 3 on ETF.com
By clicking on any of the links above, you acknowledge that they are solely for your convenience, and do not necessarily imply any affiliations, sponsorships, endorsements or representations whatsoever by us regarding third-party Web sites. We are not responsible for the content, availability or privacy policies of these sites, and shall not be responsible or liable for any information, opinions, advice, products or services available on or through them.
The opinions expressed by featured authors are their own and may not accurately reflect those of the BAM ALLIANCE. This article is for general information only and is not intended to serve as specific financial, accounting or tax advice.
© 2017, The BAM ALLIANCE