A Note on the Effect of Personal Taxes and Dividends on Capital Asset Prices
Murphy, J. Austin, Akron Business and Economic Review
A Note On The Effect Of Personal Taxes And Dividends on Capital Asset Prices(*)
In spite of the 1986 change in the tax laws that equate the personal tax rate on both dividend and capital gains income, the effect of differential taxation of capital gains and dividends continues to be an issue of primary concern to both policy makers and corporate financial officers. Brennan has hypothesized that a higher personal tax rate on dividends than on capital gains might result inhigher required returns on stocks that pay high dividends. Even with equal tax rateson capital gains and dividend income, the option investors have to defer capitalgains income on stocks might still lead to higher returns being required on stocks having higher current dividend yields. On the other hand, Black and Scholes have hypothesized that, because of the preferential tax treatment of dividend income received by corporations (70-80% of dividend income is excludable under current tax law), corporate investors might require a lower return on stocks that pay high dividends,(1) whereas many tax-exempt entities such as pension funds might be indifferent to the dividend yield. In addition, Miller and Scholes have developed a model that shows that individual investors can shield dividend income from personal taxation and thus also be indifferent to stock dividend yields.
Litzenberger and Ramaswamy have stated that the existence of a higher or lower required return on stocks with high dividend yields is an empirical issue. In an empirical test, the authors discovered evidence of a higher risk-adjusted return on stocks with high dividend yields over the interval 1936-77.(2) These results are disturbing, since they imply that corporate and tax-exempt investors can earn abnormally high returns just by investing in high dividend-yielding stocks and that corporate managers can lower the required return on their stock (and thus raise their stock price) just by reducing their dividend.(3)
However, Litzenberger and Ramaswamy observed that their results might have been biased by misspecification of the market proxy used in adjusting for risk. Blume and Hess have further stated that the use of historical estimates of risk might have distorted the results. The purpose of this research is to utilize a different measure of risk in order to reexamine the Litzenberger and Ramaswamy model of the effect of dividend yield on required returns.
In the Litzenberger and Ramaswamy model, the required expected return on an asset is a function of its systematic risk or beta B and its dividend yield, i.e., (1) [Mathematical Expression Omitted] where r, d, and B denote the total return, the dividend yield, and the beta of the subscripted asset, respectively; subscripts j and f denote the risky asset j and the risk-free asset f, respectively; | denotes a random variable; E is the expected value operator; and the g terms are parameters to be estimated. The beta measures the contribution of the asset to the variance risk of a diversified portfolio and is computed as (2) [Mathematical Expression Omitted] where m denotes the market portfolio of all assets.
The [g.sub.1] parameter represents the premium expected return for beta risk and should be positive. On the other hand, the value of the [g.sub.2] parameter depends on whether investors require a premium return for differential dividend yields. Obviously, if [g.sub.2] is greater (less) than 0, then a higher return is required on assets with high (low) dividend yields. To test for the sign of [g.sub.2], the cross-sectional average returns above the risk-free rate on a large number of risky assets can be regressed on their estimated betas and average dividend yields above the risk-free rate. To validly conduct such a regression, several econometric problems must be addressed.
To begin, Litzenberger and Ramaswamy have noted that the choice of the proxy for the market portfolio can lead to systematic biases in the measure of betas. …