Prior research indicated that dividend yield variation is a function of expected returns and not of the expected growth in dividends. The authors’ research supports the conclusion that, although true for the stock market in the aggregate, the result of previous research is not true for portfolios consisting of small and value stocks, in which future dividend changes have a greater impact on dividend yields.
What’s Inside?
The authors’ objective is to challenge the generally held conviction that variations in dividend yield are a function of only expected returns. The authors find that this view does not hold for portfolios consisting of small and value stocks. The authors base their conclusions on their calculations of dividend yield decompositions over time periods ranging from 1 to 20 years. Because dividend yield is often used as a proxy for expected stock returns, practitioners conducting asset pricing studies may find this research useful because it suggests that the use of dividend yield in this manner may not be valid for all classes of stocks.
How Is This Research Useful to Practitioners?
The authors note that this research is important for both the stock return predictability literature and the asset pricing literature. The most obvious benefit to practitioners would be in the latter area. To the extent that practitioners are engaged in the search for mispriced stocks, this research presents the opportunity to refine model specifications, particularly for certain classes of stocks.
How Did the Authors Conduct This Research?
The authors base their analysis on annual data for the time period 1928–2010. Return data (with and without dividends) for the value-weighted stock index are obtained from CRSP. Dividend growth rates are determined based on the combination of the total return and return without dividends series. Portfolios sorted on size and book-to-market are also needed for the empirical analysis. For both size and book-to-market portfolios, only portfolios consisting of the top (highest) 30% of stocks and bottom (lowest) 30% of stocks are used in order to maintain continuity of dividend data.
The authors perform long-horizon regressions to estimate future returns, dividend growth, and dividend-to-price ratios based on the current dividend-to-price ratio. The results of these regressions can then be combined to create predictive coefficients for return, dividend growth, and dividend yield, which can then explain changes in the variance of the current dividend yield.
For portfolios grouped by size, the t-statistic for the dividend growth variable for small stocks is significant at the 0.05 level for time horizons greater than 10 years. In contrast, the results are insignificant for all time horizons for large stocks. When examining the results for the return variable, however, the t-statistic is significant for the large-stock portfolio for all time horizons. The return statistic for the small-stock portfolios demonstrates a smaller degree of significance beyond three years. For both large-stock and small-stock portfolios, the dividend yield variable is significant for time periods up to 10 years but is insignificant for longer horizons.
Book-to-market portfolios are used to differentiate value stocks from growth stocks. The dividend growth variable for growth stocks is significant for horizons greater than four years, and for value stocks, the dividend growth variable is highly significant beyond the short-term time horizon. For growth stocks, the return variable also plays an important role in explaining the variation in dividend yield.
Following prior research, the authors also use an alternative variance decomposition based on first-order vector autoregression (VAR). The conclusions reached are qualitatively similar using first-order VAR.
Abstractor’s Viewpoint
For practitioners, this research creates an interesting opportunity to refine models used to detect mispriced stocks. But what is left unsaid is the extent to which this research may generate exploitable differences between price and value for the practitioner.