CAPITAL GAINS TAXES: Empirical Analysis 2

The return is computed over the five-day period, Tuesday, April 29, 1997 through Monday, May 5, 1997, during which most of the uncertainty surrounding the budget agreement was resolved. One advantage of the budget agreement for this analysis is that few specific provisions in the agreement (other than the capital gains change) clearly benefit one group of firms over another and, hence, potentially confound the analysis. Nevertheless, a series of control variables is added to mitigate the possibility that omitted correlated variables spuriously affect the study’s inferences.

The dividend variable captures the presence of (or extent of) dividend payments. To the extent the change in expected capital gains tax rates affects valuation, the effect should be most pronounced for firms that are less likely to pay dividends, all else equal. Thus, p1 is the coefficient of primary interest and is expected to be negative.

The analysis is based on the 2,000 largest U.S. corporations as reported by Datastream. Of the 1,975 sample firms with complete data, 728 (37 percent) pay dividends. For those firms, the mean dividend yield is 2.8 percent. Table 1 presents additional descriptive statistics.

Dividend Status Regressions

The initial regression investigates the stock price change of dividend-paying and nondividend-paying stocks, as follows:

Return = a + P1 Dividend Dummy + e (13)

Because dividend initiations and omissions are rare, past dividend status is assumed to predict future status and, hence, the impact of capital gains tax rate changes on investors.

The simple regression results in Table 2, Column A suggest dividend status is an important explanatory variable for stock price reactions to the budget accord. As predicted, the coefficient on the dividend variable is negative (-6.24 with a t-statistic of -17.1), indicating nondividend-paying companies outperformed dividend-paying companies during the event week. The mean return over the week in question is 6 percent for dividend-paying stocks and 12 percent for the non-dividend-paying firms. The return differential is consistent with the capital gains tax reduction being particularly good news for non-dividend-paying stocks. website

To assess the uniqueness of this week’s differential stock response, equation (13) was reestimated for every week between January 1995 and the event week. Of those 129 weeks, the coefficient on the dummy variable in the event week was the largest in absolute terms, suggesting that, whatever occurred during the week was highly unusual. However, one potential remaining concern is that non-dividend-paying stocks always react more strongly to factors that move the market. If so, the large difference between dividend and non-dividend-paying stocks may be solely attributable to the increase in stock prices during the week of the accord.