which is greater than the 6.5%($65/$1000)they would obtain from holding medium-payout stocks,or the 5%($50/$1000)they would obtain from holding low-payout stocks.What about institutions?They'will not be prepared to hold low-payout stocks,since the return on them is $100/$1050 =9.52%.This return is less than the 10%($100/$1000)they can get on the other two stocks and the opportunity cost they obtain from holding foreign assets,so they will try to sell.Again,there is downward pressure on the price of low-payout stock.Therefore,the price must fall from $1050 to $1000 for equilibrium to be restored.A similar argument explains why the prices of other stocks are also $1000.Thus,in equilibrium,the price is independent of payout policy and dividend policy is irrelevant,as in the original Miller and Modigliani theorys Several studies have attempted to distinguish between the case of the static model in which everybody is taxed the same,and the static clientele model in which investors are taxed differently.Perhaps the easiest way to make the distinction is to investigate the relation between the marginal tax rates of stockholders and the amount of dividends paid. Blume,Crockett,and Friend(1974)found some evidence from survey data that there is a modest (inverse)relation between investors'tax brackets and the dividend yield of the stocks they hold.Lewellen,Stanley,Lease and Schlarbaum (1978),using individual investor data supplied by a brokerage firm,found very little evidence of this type of effect.Both studies indicate that investors in high tax brackets hold substantial amounts of dividend-paying stock Table 2 corroborates these findings for the last 30 years.It is evident that individuals in high tax brackets hold substantial amounts of dividend-paying stocks.There is no evidence that their dividend income relative to capital gains income is lower than that of investors in low tax 5 The equilibrium here is conceptually different from the one in Miller(1977).Miller presents an equilibrium in which there is a strict clientele.In the equilibrium here,potential arbitrage by institutions ensures one price for all stocks,regardless of their dividend policy.The existence of a strict tax-clientele is inconsistent with no-arbitrage. See also Blume(1980). 26
which is greater than the 6.5% ($65/$1000) they would obtain from holding medium-payout stocks, or the 5% ($50/$1000) they would obtain from holding low-payout stocks. What about institutions? They’ will not be prepared to hold low-payout stocks, since the return on them is $100/$1050 = 9.52%. This return is less than the 10% ($100/$1000) they can get on the other two stocks and the opportunity cost they obtain from holding foreign assets, so they will try to sell. Again, there is downward pressure on the price of low-payout stock. Therefore, the price must fall from $1050 to $1000 for equilibrium to be restored. A similar argument explains why the prices of other stocks are also $1000. Thus, in equilibrium, the price is independent of payout policy and dividend policy is irrelevant, as in the original Miller and Modigliani theory.5 Several studies have attempted to distinguish between the case of the static model in which everybody is taxed the same, and the static clientele model in which investors are taxed differently. Perhaps the easiest way to make the distinction is to investigate the relation between the marginal tax rates of stockholders and the amount of dividends paid. Blume, Crockett, and Friend (1974) found some evidence from survey data that there is a modest (inverse) relation between investors’ tax brackets and the dividend yield of the stocks they hold. Lewellen, Stanley, Lease and Schlarbaum (1978), using individual investor data supplied by a brokerage firm, found very little evidence of this type of effect. Both studies indicate that investors in high tax brackets hold substantial amounts of dividend-paying stock. Table 2 corroborates these findings for the last 30 years. It is evident that individuals in high tax brackets hold substantial amounts of dividend-paying stocks. There is no evidence that their dividend income relative to capital gains income is lower than that of investors in low tax 5 The equilibrium here is conceptually different from the one in Miller (1977). Miller presents an equilibrium in which there is a strict clientele. In the equilibrium here, potential arbitrage by institutions ensures one price for all stocks, regardless of their dividend policy. The existence of a strict tax-clientele is inconsistent with no-arbitrage. See also Blume (1980). 26
brackets.According to the clientele theory,this phenomenon should not occur.For example, firms should be able to increase their value by switching from a policy of paying dividends to repurchasing shares. Elton and Gruber(1970)sought to identify the relation between marginal tax rates and dividend yield by using ex-dividend date price data.They argued that when investors were about to sell a stock around its ex-dividend date,they would calculate whether they were better off selling just before it goes ex-dividend,or just after.If they sold before the stock went ex- dividend,they got a higher price.Their marginal tax liability was on the capital gain,represented by the difference between the two prices.If they sold just after,the price would have fallen because the dividend had been paid.They would receive the dividend plus this low price,and their marginal tax liability would be their personal tax rate times the dividend.In this setting,we can make a direct comparison between the market valuation of after-tax dividend dollars and after-tax capital gains dollars.In equilibrium,stocks must be priced so that individuals'marginal tax liabilities are the same for both strategies. Assuming investors are risk neutral and there are no transaction costs,it is necessary that: P。tP。-P)=pt-P+Dl-t) (9) where PB stock price cum-dividend (the last day the stock is traded with the dividend) PA expected stock price on the ex-dividend day (the first day the stock is traded without the dividend) Po stock price at initial purchase D dividend amount 27
brackets. According to the clientele theory, this phenomenon should not occur. For example, firms should be able to increase their value by switching from a policy of paying dividends to repurchasing shares. Elton and Gruber (1970) sought to identify the relation between marginal tax rates and dividend yield by using ex-dividend date price data. They argued that when investors were about to sell a stock around its ex-dividend date, they would calculate whether they were better off selling just before it goes ex-dividend, or just after. If they sold before the stock went exdividend, they got a higher price. Their marginal tax liability was on the capital gain, represented by the difference between the two prices. If they sold just after, the price would have fallen because the dividend had been paid. They would receive the dividend plus this low price, and their marginal tax liability would be their personal tax rate times the dividend. In this setting, we can make a direct comparison between the market valuation of after-tax dividend dollars and after-tax capital gains dollars. In equilibrium, stocks must be priced so that individuals’ marginal tax liabilities are the same for both strategies. Assuming investors are risk neutral and there are no transaction costs, it is necessary that: P - t (P - P ) = P - t (P - P )+ D(1- t ) B g B 0 A g A 0 d (9) where PB = stock price cum-dividend (the last day the stock is traded with the dividend) PA = expected stock price on the ex-dividend day (the first day the stock is traded without the dividend) P0 = stock price at initial purchase D = dividend amount 27
tg personal tax rate on capital gains ta personal tax rate on dividends. The left-hand side of (9)represents the after-tax receipts the seller would receive if he sold the stock cum-dividend and had bought it originally for Po.The right-hand side represents the expected net receipts from sale on the ex-dividend day.Rearranging, P。卫=1L D 1-t, (10) If there are clienteles with different tax brackets,the tax rates implied by the ratio of the price change to the dividend will differ for stocks with different levels of dividends.The implied tax rate will be greater the higher the dividend yield,and,hence,the lower the tax bracket of investors.Elton and Gruber find strong evidence of a clientele effect that is consistent with this relation. 5.1.I The role ofrisk In the simplest versions of the theories presented above,risk has been ignored.In practice,because risk is likely to be of primary importance,it must be explicitly incorporated in the analysis. As Long (1977)pointed out,there is an implicit assumption in the argument of a tax clientele that when there is risk,there are redundant securities in the market.An investor can achieve the desired portfolio allocation in risk characteristics without regard to dividend yield. In other words,investors can create several identical portfolios in all aspects but dividend yield. Keim (1985)presented evidence that stocks with different yields also have different risk characteristics.Zero-dividend-yield stocks and stocks with low-dividend -yields have 28
tg = personal tax rate on capital gains td = personal tax rate on dividends. The left-hand side of (9) represents the after-tax receipts the seller would receive if he sold the stock cum-dividend and had bought it originally for P0. The right-hand side represents the expected net receipts from sale on the ex-dividend day. Rearranging, . 1- t 1- t = D P - P g B A d (10) If there are clienteles with different tax brackets, the tax rates implied by the ratio of the price change to the dividend will differ for stocks with different levels of dividends. The implied tax rate will be greater the higher the dividend yield, and, hence, the lower the tax bracket of investors. Elton and Gruber find strong evidence of a clientele effect that is consistent with this relation. 5.1.1 The role of risk In the simplest versions of the theories presented above, risk has been ignored. In practice, because risk is likely to be of primary importance, it must be explicitly incorporated in the analysis. As Long (1977) pointed out, there is an implicit assumption in the argument of a tax clientele that when there is risk, there are redundant securities in the market. An investor can achieve the desired portfolio allocation in risk characteristics without regard to dividend yield. In other words, investors can create several identical portfolios in all aspects but dividend yield. Keim (1985) presented evidence that stocks with different yields also have different risk characteristics. Zero-dividend-yield stocks and stocks with low –dividend –yields have 28
significantly higher betas than do high-yield stocks.This finding implies that it may be a nontrivial task to choose the optimal risk-return tradeoff while ignoring dividend yield. Depending on the precise assumptions made,some models that incorporate risk are similar to the simple static model,in that there is a tax effect and dividend policy affects value. On the other hand,other models are similar to the static clientele model in that there is no tax effect and dividend policy does not affect value.Therefore,most of the literature has focused on the issue of whether or not there is a tax effect. Brennan(1970)was the first to develop an after-tax version of the CAPM.Litzenberger and Ramaswamy (1979,1980)extend his model to incorporate borrowing and short-selling constraints.In both cases,the basic result is that for a given level of risk,the compensation for a higher dividend yield is positively related to the differential taxes between dividends and capital gains: E(Rit-Ra)=ar+a2B+as(dat-Ra) (11) Equation(11)describes the equilibrium relation between a security's expected return E(Rit),its expected dividend yield(dit),and its systematic risk (Bit).Finding a significantly positive a3 is interpreted as evidence of a tax effect.That is,two stocks with the same risk exposure (same beta)will have the same expected return only if they have the same dividend yield.Otherwise, the stock with the higher dividend yield will have a higher expected return to compensate for the higher tax burden associated with the dividend. Several researchers have tested such a relation,including Black and Scholes (1974), Blume (1980),Morgan (1982),Poterba and Summers (1984),Keim (1985),Rosenberg and Marathe (1979),Miller and Scholes(1982),Chen,Grundy,and Stambaugh(1990),and Kalay 29
significantly higher betas than do high-yield stocks. This finding implies that it may be a nontrivial task to choose the optimal risk-return tradeoff while ignoring dividend yield. Depending on the precise assumptions made, some models that incorporate risk are similar to the simple static model, in that there is a tax effect and dividend policy affects value. On the other hand, other models are similar to the static clientele model in that there is no tax effect and dividend policy does not affect value. Therefore, most of the literature has focused on the issue of whether or not there is a tax effect. Brennan (1970) was the first to develop an after-tax version of the CAPM. Litzenberger and Ramaswamy (1979, 1980) extend his model to incorporate borrowing and short-selling constraints. In both cases, the basic result is that for a given level of risk, the compensation for a higher dividend yield is positively related to the differential taxes between dividends and capital gains: E( ) R - R = a + a + a (d - Rft) 3 it 1 2 it it ft β (11) Equation (11) describes the equilibrium relation between a security’s expected return E(Rit), its expected dividend yield (dit), and its systematic risk (βit). Finding a significantly positive a3 is interpreted as evidence of a tax effect. That is, two stocks with the same risk exposure (same beta) will have the same expected return only if they have the same dividend yield. Otherwise, the stock with the higher dividend yield will have a higher expected return to compensate for the higher tax burden associated with the dividend. Several researchers have tested such a relation, including Black and Scholes (1974), Blume (1980), Morgan (1982), Poterba and Summers (1984), Keim (1985), Rosenberg and Marathe (1979), Miller and Scholes (1982), Chen, Grundy, and Stambaugh (1990), and Kalay 29
and Michaely (2000).The empirical results are mixed.Several of these studies find a positive yield coefficient,which they attribute to differential taxes. Black and Scholes (1974)performed one of the earliest (and one of the most influential) tests.Using annual data,and a slightly different version of equation (11),they tested the tax effect hypothesis: Ri=Yo+Rm-YoB,+y(di-dm)dm+i=1....N (12) where R:=the rate of return on the ith portfolio Yo=an intercept term that should be equal to the risk-free rate,R based on the CAPM R=the rate of return on the market portfolio B=the systematic risk of the ith portfolio Y=the dividend impact coefficient d=the dividend yield on the ih portfolio,which is measured as the sum of dividends paid during the previous year divided by the end-of-year stock price dm=the dividend yield on the market portfolio measured over the prior 12 months i=the error term To test the tax effect,Black and Scholes formed portfolios of stocks and used a long-run estimate of dividend yield (the sum of prior-year dividends divided by year-end price).Their null hypothesis was that the dividend-yield coefficient is not significantly different from zero.This hypothesis cannot be rejected for the entire time period(1936 through 1966)or for any of the ten-year subperiods.Black and Scholes concluded that "..it is not possible to demonstrate that 30
and Michaely (2000). The empirical results are mixed. Several of these studies find a positive yield coefficient, which they attribute to differential taxes. Black and Scholes (1974) performed one of the earliest (and one of the most influential) tests. Using annual data, and a slightly different version of equation (11), they tested the tax effect hypothesis: [ ] R - + (d - d )/ d + , i =1,..., N ~ R = + ~ 1 i m m i i γ0 m γ0 βi γ ε (12) where R ~i = the rate of return on the ith portfolio γ0 = an intercept term that should be equal to the risk-free rate, Rf, based on the CAPM R ~ m = the rate of return on the market portfolio βi = the systematic risk of the ith portfolio γ1 = the dividend impact coefficient di = the dividend yield on the ith portfolio, which is measured as the sum of dividends paid during the previous year divided by the end-of-year stock price dm = the dividend yield on the market portfolio measured over the prior 12 months εi = the error term To test the tax effect, Black and Scholes formed portfolios of stocks and used a long-run estimate of dividend yield (the sum of prior-year dividends divided by year-end price). Their null hypothesis was that the dividend-yield coefficient is not significantly different from zero. This hypothesis cannot be rejected for the entire time period (1936 through 1966) or for any of the ten-year subperiods. Black and Scholes concluded that “... it is not possible to demonstrate that 30