Who trades around the ex-dividend day? Evidence from NYSE audit file data.
Two types of traders have incentives to implement short-term, dividend-related trading strategies. The first type is securities dealers who have the same tax rates on both dividend and capital gains income, and very low transaction costs. Stock prices typically decline on the ex-dividend date, when the stock no longer trades with the right to a forthcoming dividend. If the expected capital loss differs from the dividend amount by more than their transaction costs, securities dealers will trade to profit from the difference. The second type is taxable corporations. Corporations were exempt from taxes on 70% of intercorporate dividends received during our sample period (subject to some additional constraints, discussed below). Because of the preferential tax treatment of dividend income relative to capital gains, corporations have a strong incentive to capture dividend income.
Previous research is consistent with the existence of short-term dividend trading. However, whether short-term dividend traders are taxable corporations or securities dealers remains an open question. Our paper addresses this question by examining audit file data from the TORQ database. In addition to price and volume data, the TORQ database includes information about the type of trader for both the buy and sell sides of each trade. These additional data allow more detailed tests of hypotheses on ex-dividend trading.
Specifically, if the expected capital loss is less than the dividend, securities dealers could buy the stock with the dividend and sell the stock without the dividend. Opposite predictions hold if the expected capital loss is greater than the dividend.(2) Dealers could (short) sell these securities before the dividend and purchase them afterward. As argued by Lakonishok and Vermaelen (1986) and Michaely and Vila (1996), short-term trading activity should decrease with transaction costs and increase with dividend yield.
Dividend capture by taxable corporations, which face lower taxes on dividend income than on capital gains, represents a special type of short-term trading. Corporate dividend-capture trading could also cause a net increase in cum-dividend buying and ex-dividend selling around ex-dates.(3) This trading should also be negatively related to transaction costs and positively related to yield.
In the TORQ database, trading by corporations is classified differently from trading by securities dealers. We are therefore able to distinguish between short-term trading by securities dealers and taxable corporations. This distinction is an important step in identifying the marginal ex-dividend trader, and therefore the determinants of ex-dividend returns and volume.
The paper is organized as follows. Section I discusses related research. Section II describes the data, and Section III defines theoretical predictions. Section IV discusses the statistical tests. Empirical results are reported in Section V, and Section VI concludes.
I. Related Research
Previous ex-dividend research has debated extensively whether the behavior of stock prices around ex-dividend days is determined by long-term tax-clientele investors (Miller and Modigliani, 1961, and Elton and Gruber, 1970) or short-term traders (Kalay, 1982, and Miller and Scholes, 1982). Recent research notes theoretical differences between short-term trading by securities dealers and short-term corporate dividend-capture trading (Karpoff and Walkling, 1990, and Michaely and Vila, 1995, 1996), but cannot distinguish empirically between the two.
Elton and Gruber (1970) describe tax-clientele traders as risk-neutral long-term investors who decide to buy or sell stock for reasons unrelated to the dividend, and who choose only the optimal time to trade during the ex-dividend period. In contrast, short-term traders trade specifically because of the dividend. In a dividend-capture trading strategy, traders buy the stock with the dividend and sell without the dividend, to "capture" the dividend income. The costs of this trading strategy are the capital loss and any associated transaction costs.
Many previous empirical tests of these dividend-trading theories examine ex-dividend price behavior. Eades, Hess, and Kim (1984, 1994), Karpoff and Walkling (1988, 1990), Grammatikos (1989), Fedenia and Grammatikos (1991), Michaely (1991), Robin (1991), Venkatesh (1991), Dubofsky (1992), Koski (1996), and Bali and Hite (1998), among others, analyze ex-dividend stock price behavior for US firms. Recent papers by Kato and Lowenstein (1995), Lasfer (1995), Michaely and Murgia (1995), and Frank and Jagannathan (1998) analyze ex-day returns for stocks in various foreign markets. Lakonishok and Vermaelen (1986) note that it is difficult to distinguish between ex-dividend-trading hypotheses using price data alone. For example, both the tax-clientele and short-term trading models predict a positive relation between the ex-dividend price decline as a percentage of the dividend, and dividend yield. Boyd and Jagannathan (1994) derive an equilibrium model in which this relation is complex and nonlinear.
Because inferences about ex-dividend trading using prices are limited, several papers analyze trading volume. Lakonishok and Vermaelen (1986) and Grundy (1985) examine both trading volume and prices around ex-dividend days to distinguish between tax-clientele effects and short-term trading. They find significant increases in trading volume. Michaely and Vila (1995, 1996) focus on short-term dividend trading, presenting and testing a model in which variations in investor tax rates are related to trading volume. They find that abnormal ex-dividend volume is positively related to tax heterogeneity and dividend yield, and negatively related to transaction costs and risk.
Overall, these papers document significant, short-term dividend-trading volume, but are unable to distinguish short-term trading by securities dealers from corporate dividend-capture trading. The TORQ database provides data that permit us to examine for the first time whether the short-term dividend-trading volume observed by Lakonishok and Vermaelen (1986) and Michaely and Vila (1996) is driven by taxable corporations or securities dealers.
II. Description of Data
Data for this paper come from the TORQ database distributed by the New York Stock Exchange (NYSE). The database includes transactions, quotations, order processing information, and audit trail data for 144 NYSE stocks for the period November 1990 through January 1991. These stocks are selected at random, subject to stratification by equity capitalization, so the sample includes a representative sample of firms by market capitalization. There are 63 trading dates during this period.
In our sample, we include as observations exdividend days of ordinary cash dividends for any of the 144 stocks during this period. The sample, summarized in Table 1, includes 70 ex-dividend observations (68 quarterly, one semi-annual, and one annual cash dividend payment). Dividend yields for these observations range from 0.24% to 4.21% and average 1.39%. There is also considerable cross-sectional variation in normal trading volume and transaction costs (as measured by bid-ask spreads) for the events included in the sample. Normal trading volume ranges from about 2,000 shares per day to over 1.6 million shares per day, and bid-ask spreads range from 0.16% to 7.80%.
For each ex-dividend observation, we extract from the TORQ database the entire audit file. This file includes the time, price, and volume for every trade, in addition to information about the account type of all parties to each trade.(4) As described in Hasbrouck, Sofianos, and Sosebee (1993), "The audit trail is a comprehensive trading record that supports NYSE surveillance operations and assists members in resolving trade disputes. The audit trail is a chronological reconstruction of trading in each stock, identifying the time and size of each trade and providing information on orders involved in each trade. The audit trail also indicates whether members participating in a trade acted as agents for customers or traded for their own account." We also extract bidand-ask quotations from the quote files to estimate the bid-ask spread, our measure of transaction costs.
An important advantage of this database over data used in previous research is that it makes available account-type information for both buyers and sellers. According to conversations with a representative of the NYSE, this account-type information is collected as follows: at the end of each trading day, any member who submits a trade is asked to provide further information about the trade, specifically, whether the trade is for the firm's own account, for a customer, or for another member (in which case the trade could be for either the other member's own account or for a customer of the other member). If the trade is for the member's own account, it is classified as Account Type P, a Proprietary Nonindividual trade. These Proprietary Nonindividual trades represent trades by dealers in securities. Note that Account Type P does not include trades by specialists. In the ex-dividend literature, taxneutral short-term traders are most often modeled as securities dealers, so their trades are classified in this group. Tax-neutral traders are traders who have the same tax rates on both dividends and capital gains. According to Kalay (1982), anyone can function as a tax-neutral short-term trader, given the tax treatment of short-term capital gains. However, securities dealers represent the investors with the lowest transaction costs, and are therefore the most likely tax-neutral short-term traders (see also Karpoff and Walkling, 1990).
If the trade is for the member's own customer, it is classified as Account Type I, an Individual trade. The traditional tax clienteles modeled by Elton and Gruber (1970) fall into this category. According to the NYSE, trading by customers who are nonmember taxable corporations are also included in the Individual trading category. Therefore, both traditional tax-clientele traders and corporate dividend-capture traders are classified as Account Type I. In principle, Account Type I could include trades by tax-neutral institutions such as pension funds. However, as discussed above, securities dealers are more likely to be the low-cost tax-neutral short-term traders.
Finally, trades for which the member acted as an agent for another member are classified as Account Type A, Agency Nonindividual trades. Because the other member could have traded either individually or for another customer, Agency Nonindividual trades could represent trades by tax clienteles, securities dealers, or taxable corporations.
The account-type variable also contains information on whether the trade is part of a program or index-arbitrage trading strategy. It is unlikely that these types of trades are executed for dividend-related purposes. Therefore, to remove the noise associated with these trading strategies from our measurement of dividend-related trading volume, we exclude these trades from our sample.
Table 2 summarizes our assumptions about the relevant investor account types and their relation to types of dividend traders. Throughout the remainder of the paper, we refer to the TORQ account types by their descriptions as related to dividend trading.
Table 2 also reports the proportion of normal buy and sell volume on the TORQ database attributable to the three account types on which we focus. Over 50% of volume is agency trades, dealers represent 6% of volume, and individuals and/or taxable corporations represent 13% of volume. The remaining volume includes program trades, index-arbitrage trades, and specialist trades.
[TABULAR DATA FOR TABLE 1 OMITTED]
[TABULAR DATA FOR TABLE 2 OMITTED]
III. Theoretical Predictions
This section outlines the alternative theories that explain dividend-related trading around the ex-date.
A. Trading by Securities Dealers
If the expected ex-dividend capital loss exactly equals the dividend, then there is no potential trading profit for tax-neutral traders. Therefore, we would not expect to see short-term dividend-related trading by securities dealers. However, the existence of clienteles (either individual or corporate) with tax-induced preferences for either dividend or capital gains income could cause the expected ex-dividend capital loss to differ from the dividend. If the clientele for a stock consists of individual investors with a preference for capital gains over dividends, we would expect the dividend to exceed the ex-dividend capital loss. Conversely, if the clientele for a stock consists of taxable corporations with a preference for dividend income, we would expect the capital loss to exceed the dividend.
Koski (1996) discusses corporate tax clienteles. Most empirical ex-dividend research examining sample periods after the 1986 Tax Reform Act finds that the ex-dividend capital loss is significantly less than the dividend for some low-yield stocks and significantly greater than the dividend for some high-yield stocks.(5) Both of these scenarios offer potentially profitable trading opportunities to securities dealers. If securities dealers expect that the ex-dividend capital loss will be less than the dividend, then they may buy the stock cum-dividend and sell it ex-dividend, provided that expected trading profits exceed round-trip transaction costs. (We refer to this strategy as "long-position dividend capture.") We should observe abnormal buy volume in the Dealers account type prior to an ex-date and abnormal sell-volume on or after the ex-date. If securities dealers expect that the ex-dividend capital loss will exceed the dividend by more than their transaction costs, they have incentives to sell the stock cum-dividend and buy it ex-dividend ("short-position dividend capture"). Since short-term traders are constrained by transaction costs, abnormal trading volume by securities dealers should be negatively related to transaction costs. Finally, as argued in Lakonishok and Vermaelen (1986) and Michaely and Vila (1996), the magnitude of the abnormal volume should increase with dividend yield.
B. Corporate Dividend-Capture Trading
Taxable corporations represent a distinct class of short-term traders. Corporations were exempt from taxes on 70% of intercorporate dividends received during 1990 and 1991 (subject to a 45-day holding period and some additional restrictions). Therefore, corporations had a strong incentive to capture dividend income. As mentioned above, corporations are included with Individuals in the TORQ account-type classification. Therefore, we have a unique opportunity to distinguish between securities dealers and corporate short-term traders. If corporations practice short-term dividend capture, then we should see an increase in purchases in the Individuals and/or Taxable Corporations account type cum-dividend, and a corresponding increase in sales ex-dividend.(6) This increase in abnormal volume should represent a cumulative net increase, and should not be offset by decreases in trading volume on other days. Abnormal volume due to corporate dividend-capture trading should be higher for stocks with higher yields, lower transaction costs, or both.
C. Tax-Clientele Trading
Tax-clientele traders, as modeled by Elton and Graber (1970), are risk-neutral long-term investors who decide to buy or sell stock for reasons unrelated to the dividend, and choose only the optimal time to trade during the ex-dividend period. As noted by Lakonishok and Vermaelen (1986), if the Elton and Gruber model holds, there should be no abnormal volume during the ex-dividend period. Lakonishok and Vermaelen further note that given either costly acceleration or delay of trading for tax reasons (see Green, 1980, and Grundy, 1985) or the presence of inframarginal traders,(7) there may be positive abnormal volume on the ex-date and the day before. For example, if a tax-clientele investor wants to accelerate the sale of a security to a date before the ex-day but it is costly to do so, then this investor will accelerate the trade no more than necessary and sell on the cum-date.
Similar arguments hold for a tax-clientele buyer who delays the purchase to the ex-date. Positive abnormal volume on the cum- and ex-dates will be offset' by negative abnormal volume on other dates. Because tax clienteles have already decided to trade for reasons other than the dividend, transaction costs are fixed with respect to the trade-timing decision (see Lakonishok and Vermaelen, 1986); any changes in trading volume due to tax-clientele traders should be unrelated to transaction costs. Although the primary focus of this paper is the distinction between short-term trading by securities dealers and taxable corporations, for the sake of completeness we include results concerning tax clienteles.
D. Summary of Theoretical Predictions
Table 3 summarizes the details of our theoretical predictions for securities dealers, corporate dividend-capture traders, and tax clienteles. Predictions for securities dealers include both long- and short-position dividend-capture strategies. Predictions for tax clienteles reflect the base case (the strict Elton and Gruber (1970) equilibrium model), and the costly acceleration model of Green (1980) or Grundy (1985), in which it is costly for traders to accelerate or delay trading for dividend-related purposes.
Our predictions are consistent with those defined in Lakonishok and Vermaelen (1986). However, Lakonishok and Vermaelen derive and test predictions about cumulative abnormal volume only in relation to dividend yield and transactions costs (columns 1 through 3 of Table 3). Notice that these predictions do not distinguish between dividend capture by securities dealers and taxable corporations; both trading strategies predict abnormal trading volume that is positively related to dividend yield and negatively related to transaction costs. The Lakonishok and Vermaelen predictions also do not distinguish between long- and short-position trading strategies by securities dealers. Our predictions are more specific (columns 4 through 6). TORQ information about the details of the trading strategy distinguishes between dividend-capture trading by securities dealers that should occur in Account Type P, and dividend capture by taxable corporations that should occur in Account Type I. We also distinguish between long- and short-position dividend-capture trading by securities dealers, using information in TORQ that differentiates cum- or ex-dividend purchases from sales.
We focus on the predicted abnormal trading volume specifically associated with the trading strategies under consideration. There may be additional abnormal trading volume, which would represent traders on the opposite side of the trade. In addition, trades by any of the dividend-trading groups discussed above (securities dealers, corporate dividend-capture traders, and tax clienteles) can be classified as Agency trades, if the trades are executed by a member firm acting as agent for another member. For completeness, we include trades in this category with our reported results.
IV. Statistical Tests
Following Lakonishok and Vermaelen (1986), we base our statistical inferences regarding abnormal trading activity on standardized abnormal volume (SAV). The TORQ database includes data for 63 trading days. We define the event window as the 11-trading-day period centered on the ex-dividend date (-5 [less than] t [less than] 5).(8) We define the normal trading period as the full sample period excluding the trading days within the event window.(9) We compute the mean [Mathematical Expression Omitted] and standard deviation ([Sigma]([V.sub.i,type])) of normal trading volume for each stock (i = 1 to 70) and type of trader (type = BUY I, BUY P, BUY A, SELL I, SELL P, and SELL A).
For each day within the event window, we compute abnormal volume for each stock by trader type, as
[Mathematical Expression Omitted] (1)
Since abnormal volume is difficult to interpret, we compute and report two descriptive statistics. We define percentage abnormal volume (% AV) as
[Mathematical Expression Omitted] (2)
and define SAV as
[Mathematical Expression Omitted] (3)
To determine the significance of abnormal trading around the ex-dividend date, we compute mean percentage abnormal volumes (per thousand [Mathematical Expression Omitted] and mean standardized abnormal volumes [Mathematical Expression Omitted] across stocks for each day in the event window by trader type. Mean standardized abnormal volume is
[Mathematical Expression Omitted] (4)
We compute [Mathematical Expression Omitted] in a similar fashion. We determine the statistical significance of abnormal trading by computing a t-statistic derived from mean SAV
[Mathematical Expression Omitted] (5)
[TABULAR DATA FOR TABLE 3 OMITTED]
Under the null hypothesis that trading activity in the event period is no different from normal trading activity (i.e., no short-term trading), we use the sample distribution of [SAV.sub.t,i,type] from the normal period to compute the t-statistic. By construction, the sample mean and standard deviation of [SAV.sub.t,i,type] are zero and one, respectively, for each type of trader in the normal trading period. If we further assume that [SAV.sub.t,i,type] is normally distributed, the t-statistic can be written
[Mathematical Expression Omitted] (6)
[SAV.sub.t,i,type] is actually right (or positively) skewed in our sample.(10) Johnson (1978) derives a modified $t$statistic that corrects for the effect of skewness on significance tests regarding sample means. The modified t-statistic is an adjustment to the conventional t-statistic
[Mathematical Expression Omitted] (7)
where S is the sample skewness. When skewness is positive, the conventional t-statistic is biased downward. We report both conventional and skewness-adjusted t-statistics. We are also interested in testing the hypothesis that abnormal trading volume around the exdividend day is positively related to dividend yield and negatively related to round-trip transactions costs, using the bid-ask spread as a proxy. We test this hypothesis by using ordinary least squares (OLS) to estimate
[SAV.sub.t,i,type] = [b.sub.0] + [b.sub.1][Yield.sub.i] + [b.sub.2] [Spread.sub.t,i] (8)
According to the tax-clientele hypotheses, the values of the coefficients [b.sub.1] and [b.sub.2] should be
[b.sub.1] = 0, [b.sub.2] = 0 (9)
Short-term trading by securities dealers implies that
[b.sub.1] [greater than] 0, [b.sub.2] [less than] 0 (10)
when we estimate this regression using data for Account Type P (securities dealers). When we estimate the regression using data for Account Type I (individuals and/or taxable corporations), corporate dividend-capture trading implies similar coefficients.
In these regressions, we compute dividend yield as
[Yield.sub.i] = [D.sub.i]/[P.sub.i] (11)
where [D.sub.i] is the cash dividend for Stock i and [P.sub.i] is the mean closing stock price for Stock i over Days -10 through -6 relative to ex-dividend Day 0. We obtain closing stock prices from the CRSP database. Our proxy for round-trip transaction costs is bid-ask spread.(11) The TORQ quotes file provides data on bid-and-ask quotes. For each stock in the sample, we extract every quote from the event period. For each quote, we compute percentage spread as
where A and B are ask and bid quotes, respectively. [Spread.sub.t,i] is the average percentage spread quoted for Stock i on event Day t.
Table 4 summarizes statistics that describe abnormal trading during the 11-day window surrounding the exdividend day, by trader account type for both buyers and sellers. Figures 1 and 2 plot cumulative abnormal volume by account type for purchases and sales, respectively. We also report results for aggregate volume (before partitioning by account type) for comparison with previous research. Statistics included are mean percentage abnormal volume, mean and median standardized abnormal volume, and two t-statistics (one conventional and one skewness-corrected) relating to mean standardized abnormal volume.
Positive skewness in the distribution of [SAV.sub.t,i,type] is evident in the fact that the mean exceeds the median on every date and for every type of trader. As discussed in Section IV, under the null hypothesis, the sample mean and standard deviation of [SAV.sub.t,i,type] are zero and one, respectively. If the null hypothesis is true and the distribution of [SAV.sub.t,i,type] is positively skewed, then we would expect the median [SAV.sub.t,type] to be negative, as found in Table 4.
Results for aggregate data (Panel A) show significant abnormal volume for several days around the ex-dividend day. These results are consistent with previous research. For example, in their overall sample, Lakonishok and Vermaelen (1986) find significant abnormal volume on Days t = -4 through t = +2 relative to the ex-dividend day. If we observe only this aggregate volume, we identify positive abnormal trading volume both cum- and ex-dividend that could be due to either securities dealers or corporate dividend-capture traders. This is consistent with Lakonishok and Vermaelen and Michaely and Vila (1996).
With the audit file data, we are able to distinguish between trading volume for these two strategies. Short-term trading by securities dealers should cause significant positive abnormal volume for the Dealers account type. In Table 4, Panel B, we see highly significant abnormal volume on several event days. These results suggest significant activity by securities dealers during ex-dividend periods. More specifically, if the expected capital loss is less than the dividend, we should see abnormal purchases cum-dividend and sales ex-dividend for the Dealers account type. Opposite predictions hold for stocks for which the expected capital loss exceeds the dividend. From Panel B, we see significant buying and selling both cum- and ex-dividend. One explanation is that Table 4 reflects both stocks for which the expected price decline exceeds the dividend (high-yield stocks) and those for which the expected price decline is less than the dividend (low-yield stocks).
To explore this hypothesis further, we divide the sample into three groups, stratified by dividend yield, and compute abnormal volume statistics for each group. Table 5 reports these statistics for event Days t = -1 and 0. Panel B of Table 5 reports dealer abnormal volume by dividend yield group. These results indicate that most of the abnormal purchases and sales by securities dealers are concentrated in high-yield stocks. Furthermore, mean percentage abnormal volumes for cure-dividend selling and ex-dividend purchasing of high-yield stocks are extremely high, 527% and 396% of normal volume, respectively.
We also examine cross-sectional variation in abnormal volume. Half of the 14 firms with SAV greater than two (abnormal volume more than two standard deviations above the mean) for Account Type P on the day prior to the ex-dividend day are gas or electric utilities.(12) Utilities are typically characterized by high, stable dividend yields and lower transaction costs, so it is not surprising that they are the target of dividend-related trading.
Our results are consistent with the hypothesis that for high-yield stocks, the expected capital loss exceeds the dividend, and that securities dealers execute a short-position dividend-capture strategy to profit. According to this explanation, if securities dealers trade to eliminate abnormal ex-dividend returns, then (subject to limitations imposed by their transaction costs) there should be no abnormal ex-dividend returns. Our estimates of mean ex-day returns are consistent with this explanation. For example, mean excess ex-day returns are 0.00184 and 0.00085 for the overall sample and high-yield subsample, respectively. Neither of these values differs significantly from zero.
No corresponding evidence of short-term trading by securities dealers exists for low-yield stocks. These results suggest that the ex-dividend capital loss for low-yield stocks should not deviate from the dividend sufficiently for short-term traders to profit over transaction costs. Panel C of Table 4 reports statistics for abnormal buy and sell volume for Account Type I, which includes individuals and/or taxable corporations. There is evidence of abnormal buying and selling by this group on several days in the event window. Abnormal [TABULAR DATA FOR TABLE 4 OMITTED] [TABULAR DATA FOR TABLE 5 OMITTED] buying and selling are strongest on the day prior to the ex-dividend date (t = -1), but are insignificant on the ex-dividend date itself. Figures 1 and 2 illustrate that there is no evidence that positive abnormal buy or sell volume on the cum- or ex-dividend date is offset by negative abnormal volume on other days. We interpret this as evidence against the tax-clientele hypothesis with costly acceleration or delay, which posits that there should be no net cumulative abnormal trading volume. 13 The strong abnormal volume prior to the ex-dividend date could be interpreted as evidence of dividend-capture trading by taxable corporations.
Panel C of Table 5 examines abnormal volume in each dividend-yield group by individuals and/or taxable corporations. Mean abnormal sales volume [TABULAR DATA FOR TABLE 6 OMITTED] (% AV) by individuals and/or taxable corporations in the high-dividend-yield group is 122.66% of normal and highly significant on Day t = -1.
Mean abnormal buy and sell volume is not statistically significant in the medium- and low-yield groups on Day t = -1. There is no statistically significant evidence of abnormal buy volume in high-yield stocks on the cum-dividend date. This is evidence against dividend capture by taxable corporations, at least on the cum- and ex-dividend dates.
Panels D in Tables 4 and 5 report results for Agency trades. Recall that this category can include tax clienteles, securities dealers, or corporate dividend-capture traders. Parties who wish to conceal their trading identity from other market participants (to implement particular dividend-trading strategies or for other reasons) can choose to trade through agents. Panel D shows highly significant abnormal purchases and sales on many event days. In Table 5, this volume spans stocks of all yields and is more pronounced cum-dividend than ex-dividend. These results collectively suggest that agency trades might include trades by more than one type of dividend trader.
To summarize, Tables 4 and 5 report significant evidence of dividend-related trading by securities dealers, and weaker evidence of dividend-related trading by taxable corporations. The fact that cumulative abnormal volume is positive, as shown in Figures 1 and 2, is evidence against the hypothesis that abnormal volume around the ex-day is caused by tax-clientele traders who are accelerating or delaying their trades due to the dividend. Short-term trading, either by securities dealers or by taxable corporations, implies that abnormal volume in a stock should be positively related to its dividend yield and negatively related to round-trip transaction costs. To test these additional predictions, we use OLS to estimate Equation (8).
We regress standardized abnormal volume (SAV) on dividend yield and transaction costs (using the bid-ask spread as proxy). Table 6 reports regression results for each type of trader, buy and sell, for Days t = -1 and t = 0. The most striking results appear on the cum-dividend day (Day t = -1). Aggregate volume, both buy and sell, is positively related to dividend yield and negatively related to transaction costs at conventional significance levels. This is consistent with results reported by Lakonishok and Vermaelen (1986). The regression results for account types P, I, and A are similar. In each case, the dividend yield coefficient is positive and the bid-ask spread coefficient is negative. Every coefficient is either significant or marginally significant.
Taken as a whole, these regression results provide additional support for the hypothesis that securities dealers and taxable corporations engage in dividend-related short-term trading on the cum-dividend day. Regression results for the ex-dividend date (Day t = 0) are much weaker. There is weak evidence that abnormal sell volume by individuals and/or taxable corporations on the ex-dividend date is related to dividend yield and transaction costs in the predicted manner. However, Tables 4 and 5 did not report significant abnormal volume for this group on this day.
This paper analyzes trading volume around ex-dividend days, using transaction data from the TORQ database. These data provide information about the identity of traders on the buy and sell sides of each transaction, and present a unique opportunity to partition aggregate trading volume and test more detailed hypotheses about ex-dividend-trading behavior.
We find strong evidence of significant abnormal volume by securities dealers around ex-dividend days. Abnormal volume is concentrated in high-yield stocks with low bid-ask spreads, as predicted. There is strong evidence that securities dealers sell these high-yield stocks cum-dividend and buy ex-dividend. We also find evidence of positive abnormal volume in high-dividend-yield stocks for the investor account type that includes both individual investors and taxable corporations. This abnormal volume is negatively related to bid-ask spreads, which is consistent with corporate dividend-capture trading.
Overall, we find strong evidence of dividend-capture trading by securities dealers, some evidence of corporate dividend-capture trading, but little evidence of tax-clientele trading.
We thank Erik Benrud, Minder Cheng, Peter Ciampi, David Dubofsky, Ed Rice, an anonymous referee, the Editors, and seminar participants at Washington University, the 1998 Eastern Finance Association meetings, and the 1998 Midwest Finance Association meetings for helpful comments.
1 The database is distributed by the New York Stock Exchange. See Hasbrouck (1992) and Section II for a description. The Petersen and Fialkowski (1994) and Knez and Ready (1996) empirical papers use this database.
2 Koski (1996) argues that these conditions are theoretically possible in equilibrium, particularly for high-yield stocks. Michaely (1991) provides empirical evidence that actual capital losses exceed the dividend for high-yield stocks. Lakonishok and Vermaelen (1986) discuss short-position dividend-capture strategies in detail.
3 Hereafter, we use the term "cum-dividend" to refer to trading with the dividend, i.e., just before the ex-dividend date.
4 In order to reflect the timing of ex-dividend trades accurately, we exclude all trades with nonstandard settlement. Koski (1993) and Angel (1998) discuss issues associated with nonstandard settlement. Angel (1998) also identifies an example of dividend-capture trading with nonstandard settlement in the TORQ database. If dividend-capture traders use nonstandard settlement, then excluding these trades could introduce a bias against finding evidence of abnormal dividend-related trading volume. However, we find that excluding nonstandard settlement trades has no appreciable effect on our empirical results.
5 See, for example, Fedenia and Grammatikos (1991); Michaely (1991); Robin (1991); Eades, Hess and Kim (1994); and Koski (1996).
6 Predictions for corporate dividend capture assume corporations satisfy the holding period and at-risk requirements required by the 1984 Tax Reform Act to qualify for the intercorporate dividend exclusion. Corporations may be able to satisfy these requirements and still trade cum- and exdividend. For example, according to Beatrice Garcia in The Wall Street Journal, May 19, 1986, a corporation with an ongoing investment in a stock and a desire to capture dividends could double its investment on the day prior to the ex-dividend day, and sell the initial holding on the ex-dividend day. Both positions satisfy holding period requirements.
7 Inframarginal traders are those with tax rates other than the tax rates of the marginal traders in particular clienteles.
8 We truncate the event windows for stocks with ex-dividend dates within three trading days of the beginning or end of the sample period. All stocks in the sample include, at a minimum, the five-trading-day period centered on the ex-dividend date.
9 Because the benchmark period includes days both before and after the event window, our research design minimizes potential biases due to time trends in volume.
10 Sample skewness of [SAV.sub.t,i,type] in the normal trading period ranges from a low of 2.506 for type = BUY 1 to a high of 3.127 for type = BUY P.
11 Previous ex-dividend research uses alternative proxies for transaction costs. However, most of these proxies are chosen because the variables have demonstrated significant correlations with bid-ask spreads (see Lakonishok and Vermaelen, 1986, and Karpoff and Walkling, 1988, for example). We follow more recent research and use bid-ask spreads directly (Karpoff and Walkling, 1990, and Michaely and Vila, 1996).
12 Our industry classifications are defined by two-digit SIC codes (codes 49xx in Table 1). Koski (1993) finds similar results.
13 Consistent with this result, Lasfer (1997) finds no evidence that scrip dividends for United Kingdom firms are motivated by tax savings for tax-paying individual shareholders.
Angel, J.J., 1998, "Nonstandard - Settlement Transactions," Financial Management (Spring), 31-46.
Bali, R. and G.L. Hite, 1998, "Ex-Dividend Day Stock Price Behavior: Discreteness or Tax-Induced Clienteles?" Journal of Financial Economics (February), 127-159.
Boyd, J.H. and R. Jagannathan, 1994, "Ex-Dividend Price Behavior of Common Stocks," Review of Financial Studies (Winter), 711-741.
Dubofsky, D.A., 1992, "A Market Microstructure Explanation of Ex-Day Abnormal Returns," Financial Management (Winter), 32-43.
Eades, K.M., P.J. Hess, and E.H. Kim, 1984, "On Interpreting Security Returns During the Ex-Dividend Period," Journal of Financial Economics (March), 3-34.
Eades, K M., P.H. Hess, and E.H. Kim, 1994, "Time-Series Variations in Dividend Pricing," Journal of Finance (December), 1617-1638.
Elton, E.J. and M.J. Gruber, 1970, "Marginal Stockholder Tax Rates and the Clientele Effect," Review of Economics and Statistics (February), 68-74.
Fedenia, M.A. and T. Grammatikos, 1991, "Portfolio Rebalancing and the Effective Taxation of Dividends and Capital Gains Following the Tax Reform Act of 1986," Journal of Banking and Finance (June), 501-519.
Frank, M. and R. Jagannathan, 1998, "Why Do Stock Prices Drop by Less than the Value of the Dividend? Evidence from a Country Without Taxes," Journal of Financial Economics (February), 161-188.
Garcia, B., 1986, "Trading to Nab Dividends Captures Investors' Fancy," The Wall Street Journal, May 19, Section B, 23.
Grammatikos, T., 1989, "Dividend Stripping, Risk Exposure, and the Effect of the 1984 Tax Reform Act on ExDividend Day Behavior," Journal of Business (April), 157-173.
Green, J., 1980, "Taxation and the Ex-Dividend Day Behavior of Common Stock Returns," NBER Working Paper 496.
Grundy, B., 1985, "Trading Volume and Stock Returns Around Ex--Dividend Dates," University of Chicago Working Paper.
Hasbrouck, J., 1992, "Using the TORQ Database," NYSE Working Paper (February).
Hasbrouck, J., G. Sofianos, and D. Sosebee, 1993, "New York Stock Exchange Systems and Trading Procedures," NYSE Working Paper 93-01 (April).
Johnson, N., 1978, "Modified T-Tests and Confidence Intervals for Asymmetric Population," Journal of the American Statistical Association (September), 536-544.
Kalay, A., 1982, "The Ex-Dividend Day Behavior of Stock Prices: A Reexamination of the Clientele Effect," Journal of Finance (September), 1059-1070.
Karpoff, J.M. and R.A. Walkling, 1988, "Short-Term Trading Around Ex-Dividend Days: Additional Evidence," Journal of Financial Economics (September), 291-298.
Karpoff, J.M. and R.A. Walkling, 1990, "Dividend Capture in NASDAQ Stocks," Journal of Financial Economics (December), 39-65.
Kato, K. and U. Loewenstein, 1995, "The Ex-Dividend-Day Behavior of Stock Prices: The Case of Japan," Review of Financial Studies (Fall), 817-847.
Knez, P. and M. Ready, 1996, "Estimating the Profits from Trading Strategies," Review of Financial Studies (Winter), 1121-1163.
Koski, J.L., 1993, "Nonstandard Settlement Trading on ExDividend Days," University of Washington Working Paper (September).
Koski, J. L., 1996, "A Microstructure Analysis of Ex-Dividend Stock Price Behavior Before and After the 1984 and 1986 Tax Reform Acts," Journal of Business (July), 313-338.
Lakonishok, J. and T. Vermaelen, 1986, "Tax-Induced Trading Around Ex-Dividend Days," Journal of Financial Economics (July), 287-319.
Lasfer, A., 1995, "Ex-Day Behavior: Tax or Short-Term Trading Effects," Journal of Finance (July), 875-897.
Lasfer, A., 1997, "On the Motivation for Paying Scrip Dividends," Financial Management (Spring), 62-80.
Michaely, R., 1991, "Ex-Dividend Day Stock Price Behavior: The Case of the 1986 Tax Reform Act," Journal of Finance (July), 845-859.
Michaely, R. and M. Murgia, 1995, "The Effect of Tax Heterogeneity on Prices and Volume Around the ExDividend Day: Evidence from the Milan Stock Exchange," Review of Financial Studies (Summer), 369399.
Michaely, R. and J.L. Vila, 1995, "Investors' Heterogeneity, Prices, and Volume Around the Ex-Dividend Day," Journal of Financial and Quantitative Analysis (June), 171-198.
Michaely, R. and J.L. Vila, 1996, "Trading Volume with Private Valuation: Evidence from the Ex-Dividend Day," Review of Financial Studies (Summer), 471-509.
Miller, M. and F. Modigliani, 1961, "Dividend Policy, Growth, and the Valuation of Shares," Journal of Business (October), 411-433.
Miller, M. and M. Scholes, 1982, "Dividends and Taxes: Some Empirical Evidence," Journal of Political Economy (December), 1118-1141.
Petersen, M. and D. Fialkowski, 1994, "Posted Versus Effective Spreads: Good Prices or Bad Quotes?" Journal of Financial Economics (June), 269-292.
Robin, A.J., 1991, "The Impact of the 1986 Tax Reform Act on Ex-Dividend Day Returns," Financial Management (Spring), 60-70.
Venkatesh, P. C., 1991, "Trading Costs and Ex-Day Behavior: An Examination of Primes and Stocks," Financial Management (Autumn), 84-95.
Jennifer Lynch Koski is an Assistant Professor of Finance at the University of Washington. John T. Scruggs is an Assistant Professor of Finance at Washington University in St. Louis.
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||New York Stock Exchange; Special Issue: Dividends|
|Author:||Koski, Jennifer Lynch; Scruggs, John T.|
|Date:||Sep 22, 1998|
|Previous Article:||Do firms use dividends to signal large future cash flow increases?|
|Next Article:||Dividend policy determinants: an investigation of the influences of stakeholder theory.|