THE IMPACT OF THE 1986 TAX REFORM ON EX-DIVIDEND DAY VOLUME AND PRICE BEHAVIOR CHUNCHI WU * & JUNMING HSU **

Transcription

1 EX-DIVIDEND DAY VOLUME AND PRICE BEHAVIOR THE IMPACT OF THE 986 TAX REFORM ON EX-DIVIDEND DAY VOLUME AND PRICE BEHAVIOR CHUNCHI WU * & JUNMING HSU ** Abstract - This paper examines the impact of the 986 Tax Reform Act (TRA) on trading volume and stock price behavior around the ex-dividend day. The paper shows that changes in capital gain taxes exert a differential effect on the trading volume and ex-day premiums of stocks with different yields and transaction costs. Results show that the trading volume around the ex-dividend day declines after the implementation of the tax law. There is evidence that short-term traders continue to focus on high-yield stocks after the 986 tax reform and the magnitude of exdividend day premiums is directly related to their short-term trading activity. Ex-dividend day premiums and abnormal trading volume around the ex-dividend day are significantly affected by transaction costs, and these effects appear to be stronger after 986. * School of Management, Syracuse University, Syracuse, NY -0. ** Department of Business Administration, Tunghai University, Taichung, Taiwan. INTRODUCTION In September 986, the most sweeping revision of the U.S. tax code in nearly half a century became law. The new tax law eliminated the preferential tax treatment of long-term capital gains. Under the old law, capital gains on assets held at least six months were taxed at only about 0 percent of the rate applied to regular income. Beginning in January 987, the differential taxation between dividends and capital gains substantially narrowed. This change represents the most significant capital gains tax increase in recent years. Whether this tax reform has influenced stock price and trading volume is an important issue, because it bears on federal tax revenue consequences and corporate financing strategies. The objective of this paper is to examine the impact of the 986 Tax Reform Act (TRA) on trading volume and stock price behavior around the ex-dividend day. Capital gains and dividend taxes may affect market valuation of dividends and corporate dividend policy. According to the traditional view, if dividends are taxed more heavily than capital gains, 77

2 NATIONAL TAX JOURNAL VOL. XLIX NO. investors should demand a higher before-tax premium from dividend income. Investors can receive this premium if the drop in stock price on the ex-dividend day is less than the amount of the dividend. However, short-term traders could easily eliminate this premium through dividend stripping. The 986 tax reform provides a unique opportunity for examining the effect of income tax rates on the exdividend stock price and trading. Similar to that of Michaely (99), this paper analyzes the behavior of exdividend day premiums before and after the implementation of the 986 TRA. However, whereas Michaely s analysis focused on stock price behavior, this paper examines the impact of the 986 TRA on trading volume around the exdividend day and explores the linkage between ex-dividend day price and short-term traders behavior. Michaely found that () the average premiums were similar before and after the 986 TRA and were indistinguishable from one; () the tax reform did not change the incentives of short-term traders and had a very small impact on corporate traders incentive to trade around the ex-dividend day; and () the reduction in transaction costs and structural changes in the U.S. financial markets may have resulted in the dominance of institutional traders in the trading around exday, and, consequently, a change in the individual investor s tax rates has no significant effect on the ex-day prices. Like Michaely, our paper finds that the average ex-day premiums after 986 are generally similar to the average premium in 986. Also, our results are consistent with Michaely s conclusion that transaction costs play an important role in the short-term trading after the tax reform. However, in contrast to Michaely, our results suggest that the 986 TRA has significantly affected short-term traders incentives to trade around the ex-dividend day. It is found that the trading volume around the exdividend day declines after the implementation of the 986 TRA. Furthermore, there is evidence that short-term traders have continued to focus on high-yield stocks since the 986 tax reform, and the magnitude of exdividend day premiums is directly related to their trading activity. However, shortterm trading activity and its influence on the ex-dividend price apparently have declined in the post-reform period. Our results suggest that taxes of dividends relative to capital gains have important effects on valuation of dividends and the trading activity around the ex-dividend day. Our findings thus support the old view of dividend taxation, as in Poterba and Summers (98, 98), suggesting that taxes are important determinants of dividend policy. Our results cast doubt on the new view of Auerbach (979a, 979b), Gordon and Bradford (980), and Bradford (98), based on the notion of tax capitalization, that dividend decisions are unaffected by the dividend and capital gain taxes. Bernheim and Wantz (995) find that there is a strong positive relationship between dividend tax rates and the share price response per dollar of dividends. Their result supports the dividend signaling hypothesis that an increase in dividend taxation should increase the share price response to dividend changes (or the dividend signaling effect). Although we do not directly address the issue of dividend signaling, our findings suggest that the 986 tax reform should decrease the share price response per dollar of dividends as a result of a reduction in the relative tax burden of dividends versus capital gains. 78

3 EX-DIVIDEND DAY VOLUME AND PRICE BEHAVIOR The remainder of this paper is organized as follows. The first section discusses the major implications of the 986 TRA for ex-dividend price and volume behavior. The second section describes the data and the empirical methodology. The third section reports the empirical results. Finally, the last section summarizes the main findings of this paper. THEORETICAL FRAMEWORK Elton and Gruber (970) examine the issue of tax-induced dividend clienteles. Under the assumptions that there are no transaction costs and short-term traders, no restrictions on short sales, and investors are risk-neutral, they show that the marginal tax rates of the marginal investor can be inferred from the exdividend day price drop relative to the dividend amount. Their results imply that the differential tax rates on dividends and capital gains lead to a tax clientele equilibrium. Recently, Michaely and Vila (995a) suggested that, in the presence of a perfect clientele, there should be no tax-related gains from trade among the investors around the ex-dividend day, because investors are confronted with the same relative after-tax valuation of dividends and capital gains. In the absence of disagreement on the relative valuation, there should be no profit opportunity and, consequently, no abnormal volume around the exdividend day. Thus, if a perfect clientele exists, we should not observe a significant change in trading volume around the ex-day after the 986 tax reform. The model suggested by Elton and Gruber (970) does not take risk into account. In reality, trading around the ex-dividend day is often subject to substantial price risk for both individual and institutional traders. This trading risk is even more apparent for corporate traders who are subject to the dividend exclusion restriction. Heath and Jarrow (988) showed that there may be no arbitrage opportunity on the exdividend day, even if the ex-day premium is not equal to one, because of the risk involved in the dividend-related trading. Thus, the argument of shortterm arbitrage alone cannot explain the price formation around the ex-dividend day. To account for the risk involved in dividend capture, Michaely and Vila (995a) proposed an equilibrium model to explain trading volume and stock price behavior around the ex-day. They showed that the ex-day price drop relative to the dividend amount is a function of the average relative tax rates of dividends versus capital gains across traders and the total risk in the economy. They also showed that the trading volume around the ex-day is positively related to the heterogeneity of investors income taxes and the dividend size and is negatively related to the variance of stock price. The model developed by Michaely and Vila (995a) offers specific testable implications for both ex-day premiums and trading volume. Therefore, it is an especially suitable framework for empirical investigation of the stock price behavior and trading volume around the ex-day. According to Michaely and Vila, the ex-day price drop can be expressed as P B P A = αd ε e Xσ / K e 79

4 NATIONAL TAX JOURNAL VOL. XLIX NO. where P B = the price per share cum-dividend, P A = the price per share ex-dividend, α = the average preference for dividend income relative to capital gains weighted by tax-adjusted risk tolerance D = the dividend per share announced by the firm ε e = the ex-day stock price disturbance X = the aggregate demand for the stock on the ex-day σ e and = the ex-day variance of stock price K = the aggregate risk tolerance in the economy. In notations, K = and N Σi= K i equation divided by the dividend amount D: E[(P B P A )/D] = α Xσ e /KD. The first part of equation is the weighted average tax rate of dividends relative to capital gains, where the weight is inversely proportional to each agent s risk aversion. Taxes thus affect the ex-day premium, and the less risk-averse agents have more weight in affecting this premium. The second part is the premium for the risk involved in the ex-day trading. Note that the expected ex-day premium will not be equal to one, even if α is equal to one, because σ is positive. The higher the e risk involved in dividend-related trading, the lower the expected ex-day premium. Michaely and Vila (995a) also showed that the trading volume around the exdividend day is α = N Σi= K i α i / N Σ Ki i= where K i = /[ρ i ( τ i)], g αi = ( τ i )/( d τ i ), g ρi is the constant risk aversion coefficient, and τ i and τ i are the d g dividend and capital gain tax rates for investor i. By construction, K i is the taxadjusted risk tolerance, and α is the average of N agents tax rates of dividend income relative to capital gains, α i, weighted by their tax-adjusted risk tolerance. The expected ex-day premium can be derived by taking the expectation of 5 { V e = N DΣ (α i α) K i /σ e i= The trading volume is positively related to the difference between an agent s tax rate of dividends relative to capital gains α i and the weighted mean α. That is, the more heterogeneous the tax structure is, the higher the trading volume around the ex-dividend day. Also, as shown in equation 5, volume contains additional information about investors tax preferences. Based on the above equilibrium relationships, we can establish several empirical }. 80

5 EX-DIVIDEND DAY VOLUME AND PRICE BEHAVIOR The new tax law also reduced the maximum income tax rate of corporations from 6 to 0 percent in 987 and to percent in 988. At the same time, the dividend tax exemption rate dropped from 85 percent in 986 to 70 percent in 988. The maximum dividend tax rate of taxable corporations thus rose from 6.9 to 0. percent, approximately a 7 percent increase. This increase in the effective dividend tax rate should reduce the incentive for dividend stripping. Previous studies (Lakonishok and Vermaelen, 986; Karpoff and Walkling, 990) have indicated that stocks with high dividend yields are the main targets of incorporated traders. The new tax law makes incorporated traders dividend capture less profitable and, therefore, discourages their trading around the exhypotheses with respect to ex-day premiums and trading volume before and after the 986 tax reform. The 986 TRA increases the capital gain tax rate. This increase in the capital gain tax directly affects the magnitude of the expected equilibrium ex-day premium in equation. The capital gain tax rate τ i g affects the average preference for dividend income versus capital gains, α, the aggregate risk tolerance, K, and the aggregate demand for the stock, X, on the ex-dividend day. Since the effects of τ g on K and X are of the same magnitude (see Michaely and Vila, 995a), the net effect of the increase in the capital gain tax on the ex-day premium mainly depends on its impact on α. It can be shown that α / τ g is positive. The new tax law increases each individual s capital gain tax rate τ i and the taxinduced preference for dividend income g versus capital gains α i, which is equal to ( τ i )/( τ i ). Since α is a weighted d g average of α i, an increase in α i for each individual leads to an increase in α. Other things being equal, the expected ex-day premiums in equation should increase after the elimination of the preferential tax treatment of capital gains. This argument leads to our first hypothesis. HYPOTHESIS : The average (expected) ex-day premium (P B P A )/D will be larger in the post-reform period than in the pre-reform period. Similar to Michaely (99), the hypothesis above holds only in expectations. The effect of the capital gain tax increase will be smaller if trading for stocks is dominated by short-term traders or by more risk-averse agents. The equilibrium relationship in equation 5 provides important empirical implications for the trading volume around the ex-dividend day. After the 986 tax reform, there were fewer tax brackets (rates); i.e., there were tax brackets before the tax reform and only after. The reduction in the number of tax brackets and rates decreases the tax heterogeneity. According to equation 5, the more homogeneous the tax structure is, the lower the trading volume around the ex-dividend day. In the extreme case with only one tax bracket, all investors agree upon what should be the consequences of dividend distribution. In the lack of disagreement on the relative value of dividend income versus capital gains, no dividend-related trading will occur. Conversely, when the tax structure is more heterogeneous, there will be greater disagreement on the relative value of dividends, which creates more profitable opportunities for trading and, consequently, higher trading volume. The above argument leads to the following hypothesis. HYPOTHESIS : The 986 TRA will have a negative impact on the abnormal trading volume around the ex-dividend day. 8

6 NATIONAL TAX JOURNAL VOL. XLIX NO. dividend day. This reduction in trading volume can also be explained using equation 5. Let α c and α L be the tax rates of dividends relative to capital gains for incorporated traders and longterm individual investors, respectively. The 986 TRA increases the effective dividend tax rate for incorporated traders and decreases α c. The mean α also decreases but does not fall as much as α c, since the weight of incorporated traders is less than one and α L increases after the 986 TRA. The increase in α L will be small if high-yield stocks are held by investors with low income tax rates. The term α c α (>0) decreases and the trading volume drops, which generally depends on the proportion of trading by corporations. Since trading of highdividend yield stocks is dominated by incorporated traders, the volume for high-dividend yield stocks is likely to drop more around the ex-dividend day. 5 The following hypothesis summarizes this argument. HYPOTHESIS : The 986 TRA will have a more negative impact on the trading volume of high-dividend yield stocks around the ex-dividend day. Another important factor affecting a trader s short-term trading is transaction costs. Michaely and Vila (99) showed that trading volume is lowered by the presence of transaction costs. Higher transaction costs not only reduce an investor s net returns but also make tax arbitrage riskier. Transaction costs differ among stocks. In general, the new tax law discourages short-term trading of stocks with higher transaction costs more than that of stocks with lower transaction costs. The new tax law increases the effective tax rate of dividends earned by incorporated traders and therefore, reduces the net profits of dividend capture and makes trading of stocks with higher transaction costs unprofitable. Also, the exday premium will be smaller for stocks with higher transaction costs because dividend capture trading for these stocks is discouraged by the new tax law. The ex-day premium will decrease as the dividend capture activity decreases (see also Michaely and Vila, 995a). Thus, the ex-day premiums will appear to be negatively associated with the transaction costs of trading. We summarize this argument below. HYPOTHESIS : In the post-reform period, short-term trading volume around the ex-day will be lower and the ex-day premium will be smaller for stocks with higher transaction costs. Transaction costs are usually negatively related to stock price and firm size (Karpoff and Walkling, 990). Therefore, we should observe a positive relation between ex-day premiums and firm size or stock price. This positive relation is expected to be stronger in the post-reform period. Given a higher effective tax rate and lower profits for dividend capture, short-term trading tends to be more sensitive to transaction costs. In the following section, we discuss the data and the empirical methodology. DATA AND EMPIRICAL METHODOLOGY Data were obtained from the Center for Research in Security Prices (CRSP) Daily NYSE/AMEX File. The data contain regular cash dividend distributions, daily price, and firm capitalization for the period from September 98 December 990. Daily trading volume for the entire year is available on the 99 CRSP tape starting in 986. The sample includes a total of, companies and,599 trading days. The 98 data were mainly used to compute the residual variance of the market 8

7 EX-DIVIDEND DAY VOLUME AND PRICE BEHAVIOR model for firms with ex-dividend days in the early part of the first quarter in 985. The data sample excludes exchange offers, tender offers, stock splits, liquidating dividends, subscription rights, and new share offerings. In addition, the following rules were adopted for further data screening. First, dividends paid by foreign currency were excluded. Second, to avoid distortions due to dividend announcements, all observations from which the announcement and ex-dividend dates were six or fewer trading days apart were dropped from the sample. Third, cash dividends lower than $0.5 were omitted to reduce biases caused by the minimum allowable price fluctuation. The final sample includes,6 events in 985,,87 events in 986,,787 events in 987,,8 events in 988,, events in 989, and 77 events in 990. To avoid a potential bias induced by high price volatility of the October 987 crash, the 987 data contain only the records for the first nine months. For the purpose of comparability, we compute statistics using the first nine months of 986. We also compute the results for the entire years of 986 and 987 and report them when they are materially different from the nine-month results. The empirical investigation involves estimation of both abnormal trading activity and premiums around the ex-dividend day. We discuss next the estimation and test procedures for abnormal volume and ex-dividend day premiums. To estimate abnormal transaction volume, we first calculate normal dollar trading volume based on the daily trading volume over a period of 5 to 5 days before the ex-dividend date. The selection of this estimation period avoids the potential problem of overlapping with the ex-dividend period of the previous quarterly dividend and the dividend announcement date. 6 For each ex-dividend event, daily abnormal trading volume in both dollar and percentage terms is calculated for an -day period beginning five days before the ex-dividend date and ending five days after the ex-dividend date. The abnormal volume is taken from the difference between the actual and normal trading volumes. Then, the average daily abnormal trading volume in both dollars and percentage is obtained based on all ex-dividend cases for the period of five days before the exdividend date to five days after. The t-statistics for average abnormal volume are computed as follows: 6 N Σ AV / N t,i i= t = σ(av) where AV t,i is the abnormal volume on day t for each event case i =,..., N and σ(av ) is the standard deviation of the abnormal volume on day t. 7 All t- statistics are calculated based on the event time method. To estimate the premium around the exdividend day, we estimate a weighted mean premium using a generalized least-squares method suggested by Michaely (99): 7 p = where N Σi= N Σ i= (d i /σ i ) p i (d i /σ i ) p i = the difference between the cumdividend price (P B,i ) and ex-dividend price 8

8 NATIONAL TAX JOURNAL VOL. XLIX NO. (P A,i ) divided by dividend per share (D i ), d i = the dividend per share divided by the cum-dividend price per share, σ i = the residual variance obtained from the ordinary least-squares regression of the market model, and i =,,..., N is the subscript representing each ex-dividend event case. The above formula corrects two sources of heteroskedasticity: the security s own conditional variance σ and the dividend i yield (d i ) effect. Following Michaely (99), we estimate the market model by least squares using return data in days 5 to and + to +5, where day zero is the ex-dividend date.the residuals obtained from the regression are used to estimate the variance term. EMPIRICAL RESULTS Table A reports the average premiums by years before and after the tax reform. Mean ex-dividend day premiums for the years and 990 are all close to one. The Z-statistics show that the average premiums for these years are not significantly different from one. However, the mean ex-day premiums are significantly different from one for 98, 985, and 989. The results for 98 and 985 are consistent with Grammatikos s (989) findings that exday premiums are less than one for the period The last column in Table A shows the results of the Fisher sign test, whereas column 5 reports the percentage of exday premiums that are greater than one. The results show that the number of exday premiums below one is not statistically different from the number of premiums above one only for the years 987 and 990. For the remaining years, the number of premiums below one is significantly higher than the number of premiums above one. In Table B, we report average exdividend day premiums by removing the restriction that cash dividends have to be larger than $ The results are now closer to those reported by Michaely (99). All ex-day premiums in the post-reform period and in 986 are now insignificantly different from one. Also, consistent with Michaely, the Fisher sign test shows that the number of ex-day premiums below one is not statistically different from that above one for Therefore, the difference between our results in Table A and Michaely s results is mainly attributed to the restriction on the magnitude of cash dividends in our sample. However, the Fisher sign test still indicates that the number of exdividend day premiums below one is greater than the number of premiums above one for the years 985, 988, and 989. A comparison of the ex-day premiums in the pre- and post-reform periods shows some variations. In Table A, the ex-day premium in 986 is not significantly lower (at the five percent level) than that in any of the years from A similar pattern is also found in Table B. Based on the results for the year 986 and after, we could not support Hypothesis that the 986 TRA increased the ex-day premium. However, the results show that the ex-day premium significantly increased after 985. The 986 TRA reduced the heterogeneity of the tax structure. The theory predicts that the abnormal trading 8

9 EX-DIVIDEND DAY VOLUME AND PRICE BEHAVIOR TABLE IA EX-DAY PREMIUMS BEFORE AND AFTER THE 986 TAX REFORM (DIVIDEND PER SHARE > 0.5) Standard Percentage Fisher Year Mean Deviation Z-Value >.0 Z-Value Note: Cash dividends lower than $0.5 are excluded in the calculation of ex-day premiums. TABLE IB EX-DIVIDEND PREMIUMS BEFORE AND AFTER THE 986 TAX REFORM Standard Percentage Fisher Year Mean Deviation Z-Value >.0 Z-Value Note: Cash dividends lower than $0.5 are included in the calculation of ex-day premiums. volume on the ex-dividend day should decline as a result of the tax law change. Table shows the average cumulative abnormal trading volume (from t = to 0) for the years before and after the tax reform. Note that complete daily trading volume records for the entire year were available on the 99 CRSP tape starting in 986. Comparing the trading activity of 986 with those of the remaining years, the abnormal volume (in thousands) around the ex-dividend day generally declined after the tax reform. For instance, the cumulative abnormal trading volume declined from,5 in 986 to 6,78 in 990, in dollar terms (thousands), and from.90 in 986 to.85 in 990, in percentage terms. The last column of Table reports the average abnormal volume over the period The cumulative abnormal volume dropped to 9,578 in dollar terms and.57 in percentage terms during this period. These results support Hypothesis that the 986 TRA adversely affected the abnormal trading volume before the ex-dividend day. The abnormal trading around the exdividend day began to have a larger decline in 988. The relatively large volume in 987 may be due to trading by Japanese life insurance companies. These companies are known to have been very active around the ex-dividend day in 987 (Koski, 99). Also, although the 986 TRA increased the effective dividend tax rate for taxable corporations, the dividend exemption rate did not drop to 70 percent until 988. In 987, the rate was 80 percent 85

10 NATIONAL TAX JOURNAL VOL. XLIX NO. TABLE CUMULATIVE ABNORMAL TRADING VOLUME AROUND THE EX-DIVIDEND DAY Years ,5,95 8,90 7,56 6,78 9,578 (.89) (9.9) (5.) (.9) (9.) (9.).90%.0% 5.75% 5.8%.85%.57% Note: Reported are cumulative abnormal dollar volume (in thousands) from t = to 0, t-statistics (in parentheses), and abnormal percentage volume. and the exemption rate increased if the corporate ownership was 0 percent or more. Thus, the full effect of the tax increase on short-term trading was more likely to take place since 988. As indicated earlier, the 986 TRA may have a greater impact on the trading volume of high-yield stocks. To evaluate the impact of the tax reform on the volume of stocks with different dividend yields, the sample is divided into ten dividend yield deciles for each year. Table reports the mean ex-day premiums and abnormal volume for each dividend yield group for the periods before and after the tax reform. As shown in the table, short-term trading centers on high-yield stocks. As in Lakonishok and Vermaelen (986), the abnormal volume reported in Table is the average daily abnormal volume using data for days to. Average abnormal volume is generally very high for the three highest dividend yield deciles. Average ex-dividend day premiums in Table generally increase with dividend yields. Average premiums are much larger for the highest three dividend yield deciles, especially for the period While average ex-day premiums are often significantly larger than one for the high-yield deciles, the average premiums for low-and-middle yield deciles are frequently significantly smaller than one. As suggested by Michaely and Vila (995a), the ex-day premiums can be less than one because of the risk involved in the dividendrelated trading. The ex-day premiums can also be less than one, because the effective capital gain tax rate may still be less than the ordinary income tax rate. An individual may defer realization of accrued capital gains to reduce the effect of tax rate on such gains. This lock in effect implies that the real capital gain rate may be less than the ordinary income tax rate. The abnormal dollar volume for stocks in deciles 8 and 0 declines after 986. The abnormal volume declines even more sharply after 988. These results support Hypothesis that predicts a negative impact of the 986 TRA on the abnormal trading volume of high-yield stocks. This negative impact on volume is much greater after 988. The short-term trading volume is negatively related to transaction costs. Since transaction costs are negatively correlated with firm size (see Lakonishok and Vermaelen, 986), firm size can be used as a proxy for the transaction cost variable. To examine the effect of transaction costs, we divide the sample into different size groups. We first rank stocks according to the dividend yield and then the market capitalization of the firm. Four dividend yield groups 86

11 EX-DIVIDEND DAY VOLUME AND PRICE BEHAVIOR Decile Yield (%) TABLE AVERAGE EX-DAY PREMIUMS AND ABNORMAL VOLUME BY YIELD DECILES Mean z-value Abnormal Volume ($) Abnormal Volume (%) ,99,59 77, 9,6 5,06 6, 0,067,96,6,080,0,76 6, 6,00 9, Notes: Average ex-day premiums and abnormal trading volume are reported for ten dividend deciles for the periods and , respectively. Column reports the cutoff points for the dividend yield groups. The abnormal volume is the average daily excess volume from t to t. are formed, and, within each dividend yield group, four subgroups are formed according to the market capitalization. In total, the sample is divided into 6 groups. Because the number of observations in 98 is not sufficiently large, we focus on the period in this analysis. Table presents the average ex-dividend day premiums and abnormal volume by dividend yields and firm size. As indicated, the abnormal trading volume increases with dividend yields and firm size. These results are generally consistent with Lakonishok and Vermaelen s (986) findings. Within each dividend yield group, abnormal trading volume is positively correlated with firm size. A comparison of the results for the preand post-reform periods indicates that the 986 TRA tends to have a more negative impact on the trading volume of high-yield stocks of smaller firms. To examine the effects of the tax reform, we regress ex-day premiums against dividend yields and the transaction cost variables for the periods before and after the tax reform. Table 5A reports the results. In the first regression, we incorporate dividend yield and the inverse of stock price as explanatory variables. Following Karpoff and Walkling (988), we use stock price inverse here as a proxy for the transaction cost variable. The results show that the coefficients of dividend yields are significantly positive in both subperiods and The coefficient of the price inverse is negative and significant (at the one percent level) only in the post-reform period. The higher the stock price inverse is, the higher the transaction cost. The results show that higher ex-day premiums are associated with lower transaction cost stocks. The results also show that the marginal 87

12 NATIONAL TAX JOURNAL VOL. XLIX NO. Yield Group TABLE AVERAGE EX-DAY PREMIUMS AND ABNORMAL VOLUME BY YIELDS AND FIRM SIZE Size Group Mean z-value Abnormal Volume ($) Abnormal Volume (%) ,798, ,57,966, ,5 6,0 0, , ,09, ,06 9, ,766 8, Notes: Firm size is measured by the market capitalization of the firm. The abnormal volume is the average daily excess volume from t to t. Reported are average figures for the periods and , respectively. effect of transaction costs on ex-day premiums in the post-reform period is greater than that in the pre-reform period ( 5.99 versus.9). 5 In the second regression, we replace stock price inverse with firm size (market capitalization) as a proxy for the transaction cost variable. The results are similar to the first regression except that the firm size variable has a less significant effect on ex-dividend day premiums. The coefficient of firm size is positive and significant at the ten percent level in the post-reform period. Consistent with the finding of Eades, Hess, and Kim (99), our results show that dividend yields and transaction costs affect the ex-day premium. In addition, we find that transaction costs have a larger effect on the exday premium in the post-reform period. 6 88

13 EX-DIVIDEND DAY VOLUME AND PRICE BEHAVIOR Year TABLE 5A CROSS-SECTIONAL REGRESSION RESULTS: EX-DAY PREMIUMS AS THE DEPENDENT VARIABLE Constant (5.95) 0. (6.05) Dividend Yield 7.7 (.7) 6.5 (.) /Price.9 ( 0.75) Size 0.00 (0.06) (8.95).7 (5.97) 5.99 (.90) 0.88 (7.7).76 (5.56) 0.00 (.8) Note: t-statistics are reported in parentheses. Year TABLE 5B CROSS-SECTIONAL REGRESSION RESULTS: ABNORMAL VOLUME ($) AS THE DEPENDENT VARIABLE Constant.0 (.0) 0.50 ( 0.0) Dividend Yield /Price (.075),676.9 (8.5) Note: t-statistics are reported in parentheses. 0.9 ( 7.05).658 ( 0.06) Table 5B reports the results of regressing abnormal volume against dividend yields and the transaction cost variable. As shown in the table, the coefficients of these two variables are highly significant and of the correct sign. Similar to Michaely and Vila (99), the abnormal trading volume is positively related to dividend yield and negatively related to transaction costs. However, we find that the negative effect of transaction costs on abnormal trading volume is much larger in the postreform period. The difference between the coefficients of the transaction cost variable (price inverse) in the two periods is significant (t =.099). The results are consistent with the contention in Hypothesis that the new tax law induced a more negative impact on short-term trading of stocks with higher transaction costs. Summary and Conclusions This paper examines the impact of the 986 tax reform on the trading volume and stock price behavior around the exdividend day. The paper finds that the trading volume around the ex-dividend day declined after the implementation of the 986 TRA. Thus, the change in the tax structure appears to affect adversely the incentives of short-term traders to trade around the ex-dividend day. Also, there is evidence that shortterm traders have continued to focus on high-yield stocks since the 986 tax reform, and the magnitude of exdividend day premiums is directly related to their short-term trading activity. However, short-term trading activity and its influence on the ex-dividend price apparently have declined in the postreform period. Furthermore, the paper finds that abnormal trading volume around the ex-dividend day is significantly affected by transaction costs of dividend stripping, and these effects appear to have become much stronger since 986. Our findings suggest the importance of taxes in corporate dividend decisions. Both ordinary and incorporated investors seem to recognize that taxes affect their investment returns. If taxes are not relevant, then the 986 TRA should not 89

14 NATIONAL TAX JOURNAL VOL. XLIX NO. affect the short-term trading activity around the ex-day. The short-term trading volume around the ex-day reflects investors tax preferences. The differences in investors income tax rates result in a disagreement on the relative value of dividends, which creates profitable opportunities for trading and, consequently, higher volume around the ex-day. The 986 TRA reduces the number of tax brackets and the heterogeneity of tax structure. The empirical results are consistent with the contention that this decrease in the tax heterogeneity results in less disagreement on the relative value of dividends versus capital gains and, hence, lowers the shortterm trading volume around the ex-day. Our results therefore support the old view (Poterba and Summers, 98, 98) that taxes affect the relative valuation of dividends and corporate dividend policy. Our results cast doubt on the new view (Auerbach, 979a, 979b; Bradford, 98) that dividends are unaffected by dividend and capital gain taxes. The new view is based on the notion that the stock market capitalizes tax payments associated with dividends and that this capitalization leaves investors indifferent between dividends and capital gains. Recently, Bernheim and Wantz (995) found that there is a strong positive relationship between dividend tax rates and the dividend signaling effect. Our results imply that the 986 TRA should decrease the dividend signaling effect, measured by the share price response per dollar of dividends, because the relative cost of dividends versus capital gains is reduced. Our findings also suggest that any studies addressing the issue of the effectiveness of dividend signaling (e.g., Kao and Wu, 99a, 99b) should explicitly account for the tax effect to better explain corporate dividend behavior. ENDNOTES This study was funded by the George E. Bennett Center for Tax Research, School of Management, Syracuse University. We are grateful to Maurice Harris, William Lane, and especially to the Editor, Joel Slemrod, and two anonymous referees for extremely helpful comments. An earlier version of this paper was presented at the 99 Financial Management Association meetings. Although the preferential tax treatment of capital gains was greatly reduced, there has been a maximum tax rate on capital gains. For instance, in 987, the maximum tax rate on capital gains was 8 percent while it was percent for dividend income. Since the enactment of the Tax Reform Act of 986, many studies have focused on its impact on securities trading (see Henderson, 990; Badrinath and Lewellen, 99; Bolster, Lindsey, and Matrusi, 989), ex-dividend day premiums (Michaely, 99; Michaely and Vila, 995a), real estate investments (Copley and Garris, 989), tax revenue and effective capital gain tax rates (Poterba, 987; Coyne, Fabozzi, and Yaari, 989), and other related issues. We thank a referee for pointing this out. See also proposition in Michaely and Vila (995b) for a similar argument. 5 This result can be explained by a numerical example. Let the weights of the agents be π i = K i / K, i = c, L, and a, for incorporated traders, longterm investors, and short-term arbitragers, respectively. Let, π c + π L + π a =, α = π c α c + π L α L + π a α a, and M = π c α c α + π L α L α + π a α a α. The trading volume (in percentage) can be expressed as V e /X = / (MKD/Xσ ) = / M(D/P)/ e ν, where ν = Xσ /KP. Following Michaely and Vila e (995a), we set the dividend yield (D/P) equal to one percent and the risk premium ν = 0.00, and assume that long-term investors have a marginal tax rate of 50 percent on dividend income and 0 percent on capital gains. Also, let π c = 50 percent, π L = 0 percent, and π a = 0 percent. The corporate tax rates are as indicated in text. Then, before the enactment of the 986 TRA, α c =.7, α L = 0.65, α a =, α =.5 and V e /X = percent. After the tax reform, α c =.60, α L =, α a =, α =.8, and V e /X = 9 percent. Thus, the abnormal volume drops. In general, we can show that, after the tax reform, M = ( π c )π c (α c α L ), since α L = α a. For π c 0.5, the higher π c is, the lower M and V e /X. 6 Lakonishok and Vermaelen (986) use the period from 65 to 5 days before the ex-dividend date for estimation of the normal volume. Grammatikos (989) has indicated that their estimation period could overlap with the ex-dividend day of the previous quarterly dividend. 90

15 EX-DIVIDEND DAY VOLUME AND PRICE BEHAVIOR 7 The t-statistic for average abnormal volume could be drawn from a nonstandard distribution, which is skewed or leptokurtic. 8 The average (raw) ex-day premiums reported by Grammatikos (989) are and for the periods before and after the 986 tax reform, respectively. 9 Michaely (99) does not impose the minimum cash dividend restriction. We impose this restriction to reduce the potential bias for the measures of exday premiums of very low dividend paying stocks due to the minimum price fluctuation set at $/8. 0 Michaely (99) reports the results of the Fisher sign test for the years of , 986, and 987. We also examined the abnormal turnover, which is the abnormal dollar volume relative to the outstanding share value. The results for this alternative measure are not materially different from our reports. Similar results also occur for the shorter period. For example, the differences between the level of abnormal volume in 986 and 987 are significant with t-test statistics equal to.670 and.7, for deciles 8 and 0, respectively. The annual results are available upon request. Our grouping is different from that of Lakonishok and Vermaelen (986). They used average daily trading volume as a proxy for firm size, while we employed market capitalization reported in the CRSP tape as the proxy. The regression includes dividend yield and stock price inverse, which may not be independent of the past values of the disturbance term. This might cause a stochastic regressor bias. 5 However, the difference is insignificant due to high standard errors. 6 The study by Eades, Hess, and Kim (99) assigns a dummy variable with a value of one for the period between January 987 October 989, in the return regression, to examine the effect of the 986 tax reform. The coefficient of this dummy variable is not significant, which suggests that the 986 tax law change does not significantly affect the ex-day return. REFERENCES Auerbach, Alan J. Share Valuation and Corporate Equity Policy. Journal of Public Economics No. (June, 979a): Auerbach, Alan J. Wealth Maximization and the Cost of Capital. Quarterly Journal of Economics 9 No. (August, 979b): 6. Badrinath, S. G., and Wilber G. Lewellen. Evidence on Tax-Motivated Securities Trading Behavior. Journal of Finance 6 No. (March, 99): Bernheim, B. Douglas, and Adam Wantz. A Tax-Based Test of the Dividend Signaling Hypothesis. American Economic Review 85 No. (June, 995): 5 5. Bolster, Paul J., Lawrence B. Lindsey, and Andrew Matrusi. Tax-Induced Trading: The Effect of the 986 Tax Reform on Stock Market Activity. Journal of Finance No. (June, 989): 7. Bradford, David F. The Incidence and Allocation Effects of a Tax on Corporate Distributions. Journal of Public Economics 5 No. (February, 98):. Copley, R. E., and J. M. Garris. The Effect of the Tax Reform Act of 986 on Real Estate Returns. Appraisal Journal 57 No. (April, 989): 0. Coyne, Christopher, Frank J. Fabozzi, and Uzi Yaari. Taxation of Capital Gains with Deferred Realization. National Tax Journal No. (December, 989): Eades, Kenneth, Patrick Hess, and E. Han Kim. Time Series Variation in Dividend Pricing: Tax Effects and Dividend Capturing. Journal of Finance 9 No. 5 (December, 99): Elton, Edwin, and Martin Gruber. Marginal Stockholder Tax Rates and the Clientele Effect. Review of Economics and Statistics 5 No. (February, 970): Gordon, Roger H., and David F. Bradford. Taxation and the Stock Market Valuation of Capital Gains and Dividends: Theory and Empirical Results. Journal of Public Economics No. (October, 980): Grammatikos, Theoharry. Dividend Stripping, Risk Exposure, and the Effect of the 98 Tax Reform Act on the Ex-Dividend Day Behavior. Journal of Business 6 No. (April, 989): Heath, David C., and Robert A. Jarrow. Ex- Dividend Stock Price Behavior and Arbitrage Opportunities. Journal of Business 6 No. (January, 988): Henderson, Yolanda K. Capital Gains Tax Rates and Stock Market Volume. National Tax Journal No. (December, 990):. Kao, Chihwa, and Chunchi Wu. Rational Expectations, Information Signaling and Dividend Adjustment to Permanent Earnings. Review of Economics and Statistics 76 No. (August, 99a): Kao, Chihwa, and Chunchi Wu. Tests of Dividend Signaling Using the Marsh-Merton Model: A Generalized Friction Approach. Journal of Business 67 No. (January, 99b):

16 NATIONAL TAX JOURNAL VOL. XLIX NO. Karpoff, Jonathan M., and Ralph A. Walkling. Dividend Capture in NASDAQ Stocks. Journal of Financial Economics 8 Nos. (November December, 990): Koski, Jennifer L. Nonstandard Settlement Trading on Ex-Dividend Days. University of Washington Working Paper. Seattle: University of Washington, 99. Lakonishok, Joseph, and Theo Vermaelen. Tax Induced Trading Around Ex-Dividend Days. Journal of Financial Economics 6 No. (July, 986): Michaely, Roni. Ex-Dividend Day Stock Price Behavior: The Case of the 986 Tax Reform Act. Journal of Finance 6 No. (July, 99): Michaely, Roni, and Jean-Luc Vila. Trading Volume with Private Valuations: Theory and Evidence From the Ex-Dividend Day. Review of Financial Studies (forthcoming, 99). Michaely, Roni, and Jean-Luc Vila. Investor s Heterogeneity, Prices, and Volume Around the Ex-Dividend Day. Journal of Financial and Quantitative Analysis 0 No. (June, 995a): Michaely, Roni, and Jean-Luc Vila. A Theory of Trading Volume with Transaction Costs: Application to Tax Motivated Trading. Cornell University Working Paper. Ithaca: Cornell University, 995b. Poterba, James M. How Burdensome Are Capital Gains Taxes? Journal of Public Economics No. (July, 987): Poterba, James M., and Lawrence H. Summers. Dividend Taxes, Corporate Investment, and Q. Journal of Public Economics No. (November, 98): Poterba, James M., and Lawrence H. Summers. New Evidence that Taxes Affect the Valuation of Dividends. Journal of Finance 9 No. 5 (December, 98):