The Cross Sectional Relation between Dividend Yields and Stock Returns

If a firm's dividend policy is determined independently of its investment and operating decisions, the firm's future cash flows also are independent of its dividend policy. In this case, dividend policy can only affect the value of a firm by affecting the expected returns that investors use to discount those cash flows. For example, if dividends are taxed more heavily than capital gains, then, as noted earlier, investors must be compensated for this added tax by obtaining higher pretax returns on high-dividend yielding stocks. (They would not hold shares in such stocks and supply would not equal demand if this were not true.)

Stocks with high dividend yields do, in fact, have higher returns, on average, than stocks with low dividend yields. However, Blume (1980) documented that the relationship between returns and dividend yield is actually U-shaped. Stocks with zero dividend yields have substantially higher expected returns than stocks with low dividend yields, but for stocks that do pay dividends, expected returns increase with dividend yields. This finding is consistent with the idea that stocks with zero dividend yields are extremely risky, but for firms that pay dividends, higher dividends require higher expected returns because of their tax disadvantage.

8Studies by Eades, Hess and Kim (1984) and Grinblatt, Masulis, and Titman (1984) document positive returns on ex-dates for stock dividends and stock splits.

To test whether a return premium is associated with high-yield stocks, a number of studies estimate cross-sectional regressions of the following general form:9


Pj = the firm's beta Divyldj = the firm's expected dividend yield10 ej = the error term.

The hypothesis is that y2, which measures the effect of dividend yield on required returns, is positive to reflect the tax disadvantage of dividend payments, and that y1, the coefficient of beta, is positive to reflect the effect of systematic risk on returns. Most of these studies found that the coefficient of the expected dividend yield was positive, which they interpreted as evidence favoring a tax effect.

These interpretations assume that the beta estimates used as independent variables in the regression in equation (15.3) provide an adequate estimate of the stocks' risks. However, as discussed in Chapter 5, finance academics find weak support for the idea that market betas provide a good measure of the kind of risk that investors wish to avoid. Distinguishing between tax and risk effects is further compounded by the relation of the dividend yield to other firm characteristics that are likely to be related to risk and expected returns. For example, Keim (1985) showed that both firms paying no dividends and firms paying large dividends were primarily small firms. This suggests that the expected dividend yield may be acting as a proxy for firm size in the regression shown in equation (15.3). In addition to being related to firm size, dividend yield is correlated with a firm's expected future investment needs and its profitability—both attributes that are likely to affect the riskiness of a firm's stock.

Result 15.4 Stocks with high dividend yields are fundamentally different from stocks with low dividend yields in terms of their characteristics and their risk profiles. Therefore, it is nearly impossible to assess whether the relation between dividend yield and expected returns is due to taxes or risk.

Since it may be impossible to detect whether paying dividends increases a firm's required expected rate of return, one cannot be certain that a policy of substituting share repurchases for dividends will have a lasting positive effect on the firm's share price. Although some articles by finance academics claim that dividends increase a stock's required rate of return, these studies are open to interpretation.

9The first study to test this specification was Brennan (1970), who concluded that there was a return premium associated with stocks that have high dividend yields.

10The expected dividend yield rather than the actual dividend yield must be used in these regressions because of the information content of the dividend choice. For example, a firm that pays a high dividend in a given year is likely to have a high return in that year because of the favorable information conveyed by the dividend increase (which we will discuss in detail in Chapter 19). Miller and Scholes (1982) pointed out that this information effect was ignored in the early studies on this topic, and, as a result, the purported finding of a tax effect was spurious. However, Litzenberger and Ramaswamy (1982) measured an expected dividend yield, using information available prior to the time the returns were measured, and found the coefficient of the expected dividend yield to be positive and statistically significant, supporting the hypothesis of a tax-related preference for capital gains.

Citizens Utilities

Citizens Utilities provides an interesting case study for examining the effect of dividend yields on prices. From 1955 until 1989, Citizens Utilities had two classes of common stock that differed only in their dividend policy: Series A stock paid a stock dividend (which is not taxed) and series B stock paid a cash dividend (which generates personal income tax liabilities for shareholders). The company's charter required the stock dividend on series A stock to be at least of equal value to the cash dividend on series B stock. The stock dividends were, on average, about 10 percent higher than the cash dividends.

In the absence of taxes, the two stocks should trade at an average price ratio comparable to their dividend ratio. Taxable investors, however, would then prefer the series A stock which pays no taxable dividend. This suggests that the price of series A stock should exceed 1.1 times the price of the series B stock because the untaxed stock dividend paid by the series A stock is 10 percent higher than the taxed cash dividend paid by the series B stock. As shown by Long (1978), the price of series A stock before 1976 was somewhat less than 1.1 times that of the series B stock, but in the period examined by Poterba (1986), 1976-84, the ratio of the prices was about equal to 1.1. This evidence suggests that investors in Citizens Utilities stock prices were not influenced by their personal tax considerations. Furthermore, a study by Hubbard and Michaely (1997) showed that the relationship between the prices of the two classes of Citizens Utilities stock was largely unaffected by the Tax Reform Act of 1986 which substantially influenced the relative value of dividends and capital gains to taxable investors.11

While dividends may have had no effect on Citizens Utilities stock prices, we cannot generalize this finding to all firms. Since the two classes of Citizens' stock are essentially the same, any large difference in their prices presents an opportunity for arbitrage by tax-exempt investors. Indeed, the arbitrage argument described in Example 15.1 can be applied to show that in the absence of transaction costs the stock prices must be identical. This opportunity for arbitrage would not exist for other stocks, indicating that one might observe two closely related, but not identical, stocks with different dividends providing very different expected returns.

Stocks and Shares Retirement Rescue

Stocks and Shares Retirement Rescue

Get All The Support And Guidance You Need To Be A Success At Investing In Stocks And Shares. This Book Is One Of The Most Valuable Resources In The World When It Comes To

Get My Free Ebook

Post a comment