Cash dividend payouts by firms have been observed to have some effect on the prices of shares, and, therefore, the returns to investors holding the shares. However, there is no unanimous agreement as to why the price changes take place. Some relatively non-contestible evidence from different markets in different countries has led to a general acceptance that a change in dividend leads to an information effect, which is consistent with the earlier valuation theories and the later capital asset pricing theories. A drop in the share price by an amount less than that of the dividend payout at the ex-dividend day has been observed: this systematic price drop has been interpreted by some researchers as suggesting a tax effect arising from the tax consideration of the investors trading at the ex-day.
What has not received general agreement is the issue of whether or not a dividend per se effect (after controlling for the information effects from dividend changes) is compatible with the valuation theories. There is a widespread belief that a dividend per se effect is not compatible with the valuation theory which suggests that the cash flows from the investments in the underlying firm determine the price of a security. Some researchers have claimed that a dividend per se effect is compatible with the after-tax capital asset pricing theory in the face of market imperfections. One study with an empirical control for the tax factor produced evidence anomalous to the current body of theories.
This study uses two classes of share securities in the Singapore equity market with different tax treatments of the dividend streams. The cash dividends received by the shareholders from one class of securities are exempted from personal income tax while the dividends of other securities are taxed in the hands of the receiver. Any difference in the behaviour of the two classes, therefore, can be attributed to the discriminating tax factor under certain conditions.
The research design for this study is the classical test of unequal samples differing with respect to a control factor, namely the tax factor. The strategy followed is to examine first the securities differing with respect to the control factor and testing for differences in other-than-the-tax factor. The first set of results suggests that the two classes of securities are similar with respect to a number of market-relevant variables and that the tax treatment of dividend streams is the only difference. Next, the investigation proceeds to find out if the assumptions of the models used and the statistical tests are generally tenable for the data set to be used in this research. Support for the validity of the assumptions is also obtained.
Next, two strategic hypotheses are tested about the tax-induced per se effect suggested by rational investor behaviour in an efficient market using the after-tax capital asset pricing models with tax costs as a market imperfection. The null hypotheses tested are:
i. The average return of the dividend taxed securities is not different from that of the dividend not-taxed securities, and
ii. Given that the after-tax asset pricing model is a correct description of the equilibrium pricing of share securities, the dividend taxed and the dividend not-taxed securities have no significant difference. The null test uses the model to test for no per se effect.
Solutions to a number of problems are derived for (i) measuring the systematic risk of securities accurately in the presence of serious thinness in trading, (ii) contemporaneous covariance in the residual returns and (iii) errors in variables, so that the resulting estimates of the coefficients from the after-tax capital asset pricing model are unbiased, and the test results efficient.
The null hypotheses are not supported by the findings of this research. A systematic higher average return is observed for the dividend taxed securities. The performances of the two sets of securities, as estimated by the after-tax models, are not similar. The coefficients for the tax impact measured by the dividend variables are statistically significant especially with the unbiased estimates from generalised least squares regression.
Given the institutional arrangements for one class of securities with taxable dividend streams in the market under study, it is argued that investors form rational expectations about the impact of the tax factor on their estimates of the end-of- period cash flows from their investments in the dividend taxed securities. Through such a market clearance process, the tax-induced effect emerges as the dividend per se effect observed in this study.