Stock Price Behavior on Ex-Dividend Dates. Hui-Ju Tsai * This Draft: 2/5/2018

Transcription

1 Stock Price Behavior on Ex-Dividend Dates Hui-Ju Tsai * This Draft: 2/5/2018 We examine stock price behavior on ex-dividend dates with the consideration of dynamic adjustment of the bid-ask spread. For both NYSE- and NASDAQ-listed stocks, we show that the bid-ask spread is increased on ex-dividend dates. More importantly, the adjustment of bid and asked prices on ex-dividend dates is asymmetric so that more of the carrying cost is borne by buyers than by sellers, who are more likely to conduct dividend capturing activities. We show that the asymmetry is more pronounced for NASDAQ-listed than for NYSE-listed firms, especially when the abnormal trading volume surrounding ex-dividend dates is high. JEL Classification: G14, G35 Keywords: Ex-dividend date; NASDAQ; Bid-ask spread; Market microstructure * Tsai is the corresponding author at the Washington College, 300 Washington Avenue, Chestertown, MD 21620, htsai2@washcoll.edu.

2 I. Introduction The topic of abnormal stock returns around ex-dividend dates has been of considerable interest to financial economists as well as practitioners since it was first documented in Campbell and Beranek (1955). Several hypotheses have been provided in the literature to explain stock price behavior around ex-dividend dates. The well-known tax hypothesis by Elton and Gruber (1970) indicates that the price-drop-to-dividend ratio being less than one on ex-dividend dates is due to differential tax treatments on dividend income and capital gains. Their work is followed by a number of empirical studies that examine the relation between tax rules and abnormal stock returns around ex-dividend dates (see, e.g., Booth and Johnston, 1984; Eades, Hess, and Kim, 1984; Poterba and Summers, 1984; Barclay, 1987; Michaely, 1991; Robin, 1991; Lasfer, 1995; Michaely and Vila, 1995; Green and Rydqvist, 1999; Bell and Jenkinson, 2002; and Kadapakkam and Martinez, 2008). Kalay (1982) argues that the discrepancy between price drop and the dividend amount on ex-dividend dates gives tax-neutral or tax-advantaged investors an arbitrage opportunity and thus the price-drop-to-dividend ratio should reflect the transaction costs faced by marginal arbitragers. Some studies show that the abnormal stock returns around ex-dividend dates are related to market microstructure. Dubofsky (1992) and Dubofsky (1997) indicate that the abnormal returns on ex-dividend dates are caused by NYSE Rule 118 and AMEX Rule 132, by which open limit buy orders are reduced on ex-dividend dates to reflect dividend payments, whereas open limit sell orders are not adjusted. Bali and Hite (1998) believe that price discreteness is the main factor that causes the price-drop-to-dividend ratio to be less than one. Their argument, however, is not supported by later evidence found by Graham, Michaely, and Roberts (2003), who examine the price-drop-to-dividend ratio after the tick size goes from one-eighth to one-sixteenth and then to 1

3 decimals. Meanwhile, Frank and Jagannathan (1998) propose the bid-ask bounce theory, arguing that the market makers tend to buy shares on the cum-dividend date and sell on the ex-dividend date. However, Koski and Michaely (2000) and Graham, Michaely, and Roberts (2003) do not find evidence supporting the bid-ask bounce theory. Several studies find that investors conduct dividend capture activities by buying shares before ex-dividend dates and selling shares on or after ex-dividend dates. Lakonishok and Vermaelen (1986), for instance, find that there is abnormally high trading volume around exdividend dates, especially for high-yield stocks (See, also, Karpoff and Walking, 1988; Rantapuska, 2008; Henry and Koski, 2017). Investors dividend capturing activities imply that market makers have to carry more inventory on the ex-dividend dates. In order to compensate for the additional carrying cost, market makers may increase the bid-ask spread on ex-dividend dates. Current studies that model stock price behavior surrounding ex-dividend dates, however, either do not explicitly consider bid-ask spread or assume a constant bid-ask spread on the cum- and exdividend dates. Our study relaxes this constraint by allowing market makers to dynamically adjust the bid-ask spread on ex-dividend dates. We find that the bid-ask spread is increased on exdividend dates for both NYSE- and NASDAQ-listed firms. More importantly, we find an asymmetric adjustment of bid and asked prices on the ex-dividend date. The magnitude of increase in the asked price is larger than the magnitude of decrease in the bid price on the ex-dividend date. The asymmetric adjustment is more pronounced for NASDAQ-listed than for NYSE-listed firms especially when the abnormal trading volume around the ex-dividend date is high. The evidence suggests that the dynamic change of the bid and asked prices is an important factor that affects stock price behavior on the ex-dividend date especially for NASDAQ-listed stocks. Moreover, complementary to the findings in Mortal, Paudel, and Silveri (2017) that the average price-drop- 2

4 to-dividend ratio of NASDAQ-listed firms converges with that of NYSE-listed firms after NASDAQ s introduction of SuperMontage at the end of 2002, the asymmetric adjustment of bid and asked prices of NASDAQ-listed stocks is not significantly different from that of NYSE-listed firms after Our evidence agrees with the argument that market makers increase the bid-ask spread on the ex-dividend date to compensate for higher liquidity costs, and they dynamically adjust the bid and asked prices so that more of the liquidity cost is borne by less-elastic retail investors. Based on current tax regimes, compared to institutional investors, retail investors face higher tax rates on dividend income than on capital gains. Thus, tax-neutral or tax-advantaged institutional investors are more likely to conduct dividend capture activities (Kalay, 1982). When institutional investors conduct dividend capturing activities by buying shares before or on the cum-dividend date and selling shares on or after the ex-dividend date, market makers face a higher probability of carrying a larger inventory on the ex-dividend date. Our finding suggests that market makers increase the bid-ask spread on the ex-dividend date in order to compensate for higher liquidity costs. Furthermore, since retail investors are less elastic than institutional investors, market makers tend to adjust the bid and asked prices so that more of the carrying cost is borne by retail investors than by institutional investors. The asymmetric adjustment is more pronounced for NASDAQ-listed firms before 2003, which might be attributed to its higher liquidity costs and lower price efficiency before the introduction of SuperMontage. The remainder of the paper is organized as follows. A summary of studies on stock price behavior surrounding ex-dividend dates is provided in section II. We present hypothesis development and model construction in Section III. Data selection process and descriptive 3

5 statistics are described in Section IV. Results of empirical analysis are presented in section V. In section VI, we provide evidence after Section VII concludes the paper. II. Literature Review The earliest study that documented positive abnormal returns on the ex-dividend date can be traced back to Campbell and Beranek (1955). This study shows that the average price drop on the ex-dividend date is less than the amount of the dividend. Their work is followed by a large body of literature that proposed alternative explanations to justify the abnormal returns around the ex-dividend date. Elton and Gruber (1970) show that the abnormal returns on the ex-dividend date are mainly driven by tax heterogeneity. They argue that since investors consider after-tax returns, the price-drop-to-dividend ratio should reflect the marginal rate of substitution between dividend income and capital gains. When the tax rate on dividend income faced by marginal investors is higher than that on capital gains, the price-drop-to-dividend ratio is less than one on the exdividend date. Further, Elton and Grube (1970) find evidence supporting the tax-clientele hypothesis. According to the hypothesis, higher dividend payments are more attractive to investors with lower tax brackets, and thus the price-drop-to-dividend ratios increase with dividend payments. Elton and Gruber s (1970) tax-related argument is supported by several empirical studies that examine ex-dividend day stock price behavior surrounding tax reforms in the U.S. Whitworth and Rao (2010), for instance, examine stock prices around ex-dividend dates in the U.S from 1926 to 2005 and find evidence consistent with the tax hypothesis (see, also, Barclay, 1987; Michaely, 1991; Robin, 1991; and Michaely and Vila, 1995). Moreover, Whitworth and Rao (2010) show that the positive relation between dividend yield and price-drop-to-dividend ratio is 4

6 strengthened during periods when the tax differential between dividend income and capital gains increases. Elton and Gruber s (1970) framework is also tested in countries other than the U.S. Bell and Jenkinson (2002) examine ex-dividend day stock price behavior before and after a major change in dividend taxation in the U.K. in 1997 and find evidence supporting the tax clientele effect (see, also, Poterba and Summers, 1984; and Lasfer, 1995). Studying Swedish lottery bonds that offer tax advantages to coupon payments relative to capital gains, Green and Rydqvist (1999) observe that the behavior of bond prices around ex-dividend dates is consistent with the tax hypothesis. Meanwhile, Booth and Johnston (1984) study the ex-dividend day price-drop-todividend ratio in Canada but fail to find much evidence supporting the tax clientele effect. Evidence inconsistent with Elton and Gruber s (1970) tax hypothesis is also found in Kadapakkam and Martinez s (2008) study on the ex-dividend day abnormal returns in Mexico that offer dividend tax credit to individual investors. Their finding agrees with Eades, Hess, and Kim s (1984) observation that the ex-dividend day returns of several taxable and non-taxable distributions in the U.S. are inconsistent with the implication of the taxation hypothesis. Not all investors face unfavorable tax treatment on dividend income relative to capital gains. Corporations, for instance, may prefer dividends because 70% of their dividend income is tax free. Institutional investors and tax-exempt investors are indifferent to the difference between dividend income and capital gains according to current tax regime. Kalay (1982) argues that the discrepancy between price drop and dividend amount gives tax-neutral or tax-advantaged investors an arbitrage opportunity. These arbitragers trade to capture dividends by buying on or before the cum-dividend date and selling on or after the ex-dividend date. Kalay (1982) argues that the pricedrop-to-dividend ratio should reflect the transaction cost faced by marginal arbitragers. Kalay s 5

7 (1982) dividend capture hypothesis is supported by several empirical studies. Lakonishok and Vermaelen (1986), for instance, find that there is abnormally high trading volume around exdividend dates, especially for high-yield stocks (See, also, Karpoff and Walking, 1988). Several studies find that the abnormal trading volume on the ex-dividend date is affected by several factors. Dhaliwal and Li (2006), for instance, show that ownership heterogeneity increases with the abnormal trading volume on ex-dividend dates. Examining the Finnish stock market, Rantapuska (2008) finds that idiosyncratic risk, transaction costs, and dividend yields are important factors that affect the short-term trading behavior on ex-dividend dates. Similarly, Henry and Koski (2017) examine transactions from institutional investors and find that a significant abnormal trading volume is positively related to dividend yield, but negatively to idiosyncratic risk. Boyd and Jagannathan (1994) propose a model in which the price-drop-to-dividend ratio in equilibrium is determined by the interaction among taxable individuals, tax-advantaged dividend capturers, and tax-neutral arbitragers. Similarly, Michaely and Vila (1996) propose a dynamic trading clientele theory, showing that the price-drop-to-dividend ratio is the result of the interaction of investors who face different tax charges on dividend income relative to capital gains. Consistent with Michaely and Vila (1996), Michaely and Murgia (1995) examine stocks listed on the Milan Stock Exchange and find that the abnormal trading volume on the ex-dividend date is positively related to taxation heterogeneity. Another strand of research claims that the positive abnormal return on the ex-dividend date is related to market microstructure. Dubofsky (1992) and Dubofsky (1997) show that due to NYSE Rule 118 and AMEX Rule 132, open limit orders to buy stock must be reduced by the cash dividend amount or the next lower eighth on the ex-dividend date if the dividend amount is not a multiple of eighth, whereas the limit orders to sell are not changed. Dubofsky (1992) and Dubofsky 6

8 (1997) demonstrate that these rules lead to positive abnormal ex-dividend day returns (see also Akhmedov and Jakob, 2010). Bali and Hite (1998) argue that due to price discreteness, price drop on the ex-dividend date is less than the dividend amount, but within one tick of the dividend amount. Their argument, however, is not supported by later evidence that arises when the tick size goes from one-eighth to one-sixteenth, and later to decimals (see, for instance, Graham, Michaely, and Roberts, 2003). Meanwhile, Frank and Jagannathan (1998) propose the bid-ask bounce theory, which states that the market makers tend to buy shares on the cum-dividend date and sell shares on the ex-dividend date and thus cause positive abnormal returns on the ex-dividend date. However, Kadapakkam (2000) and Graham, Michaely, and Roberts (2003) do not find evidence supporting the bid-ask bounce theory. III. Hypothesis Development and Model Construction Several studies show that investors conduct dividend capture activities by buying shares before ex-dividend dates and selling shares on or after ex-dividend dates. This implies that market makers have to carry more inventory on the ex-dividend dates. In order to compensate for the additional carrying cost, market makers may increase the bid-ask spread on ex-dividend dates. Current studies that model stock price behavior surrounding ex-dividend dates, however, either do not explicitly consider bid-ask spread or assume a constant bid-ask spread on the cum- and exdividend dates. Our framework relaxes this constraint by allowing market makers to dynamically adjust the bid-ask spread on the ex-dividend date. Specifically, we consider the following framework: P ex, ask Pcum, ask D Sask, and (1) 7

9 P ex, bid Pcum, bid D Sbid, (2) where P ex, bid and ex ask P, denote the bid and asked prices on the ex-dividend date, and P cum, bid and P cum, ask represent the bid and asked prices on the cum-dividend date, respectively. S bid and S ask respectively measure the adjustment of bid and asked prices on the ex-dividend date. If market makers do increase the bid-ask spread on the ex-dividend dates, ( S bid + S ask ) should be positive. Figure 1 illustrates the adjustment of bid and asked prices on the ex-dividend date. The bid and asked prices on the ex-dividend date is determined by the bid and asked prices on the cumdividend date, dividend payment, and the adjustment of bid and asked prices. This model is based on Jakob and Ma s (2003) framework, but here we allow the market makers to adjust the bid-ask spread on the ex-dividend date to compensate for additional carrying costs. More importantly, our model allows the market makers to asymmetrically adjust the bid and asked prices. According to Conrad and Conroy (1994), in order to compensate for additional carrying costs, the market makers adjust the bid and asked prices such that investors with the most inelastic demand bear more of the cost. When the dividend capturing behavior is conducted, mostly by institutional investors, we anticipate the market makers to receive more sell orders from institutional investors relative to retail investors and more buy orders from retail investors relative to institutional investors on the ex-dividend date. Since institutional investors have more bargaining power than retail investors, we hypothesize that the market makers impose more carrying costs on the retail investors than on the institutional investors. That is, the market makers adjust the bid and asked prices so that the increment of asked price ( S ask ) is higher than the decrease in bid price ( S bid ). 8

10 According to Jakob and Ma (2003), the change in security price on the ex-dividend date can be written as 1 cum Pcum, bid expex, ask ex Pex bid, (3) P cumpcum, ask 1, where ΔP denotes the change of share price from the cum-dividend to the ex-dividend dates, and cum and ex respectively represent the probability of a buy order on the cum- and ex-dividend dates. To measure the probability of a buy order, we follow Keim (1989) and Conrad and Conroy (1994) by defining P P cum cum, bid cum, and (4) Pcum, ask Pcum, bid P P ex ex, bid ex, (5) Pex, ask Pex, bid where Pcum and Pex are closing transaction prices on the cum- and ex-dividend dates, respectively. A higher value of indicates a tendency to close at the asked price and thus implies a higher probability of a buy order. Substituting equations (4) and (5) into equation (3) gives S S P P D P Sbid ex ask bid ex cum cum, ask cum, bid. (6) Thus, the difference between dividends and price change on the ex-dividend date is determined by the probability of buying on the cum- and ex-dividend dates, the bid-ask spread, and the dynamic change in the bid and asked prices on the ex-dividend date. 9

11 IV. Data and Descriptive Statistics We consider all dividend payments with code 12N2 made by domestic companies in CRSP during the period from January 1, 1983 through December 31, We require share price on the cum-dividend date to be at least $1 and the dividend amount at least $0.01 (Whitworth and Rao, 2010). To be included in the sample, firms must have the same number of shares outstanding on the cum- and ex-dividend dates, the trading volume on both dates must be positive, and the dividend yield must be no more than 0.1 (Kadapakkam, 2000; Mortal, Paudel, and Silveri, 2017). Announcements with more than 365 days or less than 4 days between ex-dividend dates are excluded from the sample. P We follow the literature by computing price-drop-to-dividend ratio as PDR= P D cum E ex, where D denotes the dividend amount, and P cum and E P ex respectively represent the closing stock price on the cum-dividend date and expected stock price on the ex-dividend date. E P ex is obtained by discounting the actual ex-dividend day closing price by the expected stock return on the same day. To determine the expected stock return, we apply the market model on returns during days (- 50, -6) and (6, 20) relative to the ex-dividend date and require that there must be at least 30 days of returns available. We are left with a sample of 193,523 dividend announcements after these screenings. To remove outliers, we sort the announcements according to their price-drop-todividend ratios and remove events ranked within the top or bottom one percent from our sample. The removal of outliers leaves us with a sample of 109,955 and 68,470 dividend announcements made by 2,683 and 3,303 NYSE-listed and DASDAQ-listed firms, respectively. 1 We require the companies to have share code 10 or 11 in order to be included in the analysis. 10

12 Table 1 summarizes the descriptive characteristics of the sample. Not surprisingly, the average size of firms listed on the NYSE is much larger than that of firms listed on the NASDAQ, with an average market value equal to $6.80 billion and $1.25 billion, respectively. On the exdividend date, the trading volume of NYSE-listed firms is significantly higher than that of NASDAQ-listed firms. The difference in the liquidity between the two groups of firms is also reflected in the bid-ask spread. On the ex-dividend date, the average bid-ask spread of NYSElisted and NASDAQ-listed firms are 0.23 and 0.35, respectively, which are equal to 0.86 percent and 1.90 percent of stock price. The average amount of dividends paid by NYSE-listed firms is $0.23, which is higher than the $0.15 paid by NASDAQ-listed firms. However, due to the fact that NASDAQ-listed firms on average have lower stock prices, the dividend yields of these two groups of firms are not significantly different. The average dividend yields of NYSE-listed and NASDAQ-listed firms are equal to 0.71 percent and 0.72 percent, respectively. On average, there are around days between declaration dates and ex-dividend dates and days between ex-dividend dates and payment dates. The number of days from declaration dates to ex-dividend dates and from ex-dividend dates to payment dates are slightly higher for NYSE-listed than for NASDAQ-listed firms. The average return on the ex-dividend date for NASDAQ-listed firms is 0.47 percent, which is higher than the average return of 0.16 percent for NYSE-listed firms. The average and median price-drop-to-dividend ratio of NYSE-listed firms are 0.91 and 0.93, whereas the corresponding ratios of NASDAQ-listed firms are 0.51 and Consistent with Mortal, Paudel, and Silveri (2017), the price-drop-to-dividend ratio of NASDAQ-listed firms is significantly lower than that of NYSE-listed firms. They argue that the differences in price-drop-to-dividend ratios between NYSE- and NASDAQ-listed firms are mainly due to market structures, especially in that 11

13 the NASDAQ has higher execution costs and lower price efficiency than the NYSE. Moreover, they show that due to the recent structural changes in the NASDAQ, the price-drop-to-dividend ratio of NASDAQ-listed firms converges with and is not significantly different from that of NYSElisted firms. V. Empirical Analysis A. Summary Statistics To implement regression analysis, we further require that the data from CRSP has positive bid and asked prices on the cum- and ex-dividend dates with the asked price being higher than the bid price. This screening leaves us with a sample of 69,950 and 67,013 dividend announcements made by NYSE-listed and DASDAQ-listed firms, respectively. 2 Panels A and B of Table 2 present the summary statistics of bid and ask returns, probabilities of buying, and the bid-ask spread for NYSE- and NASDAQ-listed firms, respectively. The bid (ask) return is determined by P ex,bid - P cum,bid + D P cum,bid ( P ex,ask - P cum,ask + D P cum,ask ). The bid (ask) return of NYSE-listed firms is 0.15 (0.15) percent, whereas the bid (ask) return of NASDAQ-listed firms is 0.49 (0.52) percent. In untabulated tests, we find that the bid and ask returns of NASDAQ-listed firms are higher than that of NYSE-listed firms at a significance level of 0.01, which is consistent with previous findings that the price-drop-to-dividend ratio of NASDAQ-listed firms is significantly lower than that of NYSE-listed firms. The bid-ask spread of NYSE-listed firms on the cum-dividend date is , which is significantly different from the spread of on the ex-dividend date at a significance 2 Before 2003, there are no bid and asked prices available for stocks listed on the NYSE. We conducted a robustness test by using low bid and high asked prices in place of bid and asked prices in our analysis and obtained similar results. The results are available from the authors upon request. 12

14 level of For NASDAQ-listed firms, the bid-ask spread on the cum-dividend date is , which is significantly lower than the bid-ask spread of on the ex-dividend date at a significance level of Moreover, our test shows that the bid-ask spread of NASDAQ-listed firms is significantly higher than that of NYSE-listed firms on the cum-dividend and ex-dividend dates. The probabilities of buying NYSE-listed firms on the cum- and ex-dividend dates are equal to 0.54 and 0.53 percent, respectively, both of which are higher than 0.5 at a significance level of That is, on the cum- and ex-dividend dates, investors are more likely to submit a buy order. Additionally, the probability of submitting a buy order is higher on the cum-dividend date than on the ex-dividend date. For NASDAQ-listed firms, the probabilities of submitting a buy order on cum- and ex-dividend dates are equal to 0.51 and 0.49, respectively. Our statistical analysis shows that the probability of submitting a buy order on the cum-dividend date is significantly higher than 0.5, whereas the probability of buying on the ex-dividend date is significantly lower than 0.5. That is, while it is more likely for investors to submit a buy order on the cum-dividend date, investors tend to submit a sell order on the ex-dividend date. The evidence is consistent with the dividend capturing theory, which shows that investors buy on or before the cum-dividend date to receive dividends and sell shares after the stock goes ex-dividend. B. Regression Result Based on equation (6) in Section III, we conduct the following regression analysis separately on NYSE- and NADSAQ-listed firms: D -DP = b 0 + b 1 l ex + b 2 ( l ex - l cum ). 13

15 To examine if a significant difference between the two groups of firms exists, we also conduct the following regression: D - DP = b 0 + b 1 l ex + b 2 ( l ex - l cum ) + éë b 3 + b 4 l ex + b 5 ( l ex - l cum ) ù û * I, where I is an indicator that equals 1 for NASDAQ-listed firms and 0 otherwise. The regression results are presented in Table 3. The coefficient of (λex-λcum) estimates the bid-ask spread on the cum-dividend date (i.e., P cum ask Pcum, bid, ). Since NASDAQ-listed stocks are usually smaller issues and are less frequently traded, we expect to see the bid-ask spread of NASDAQ-listed firms being larger than that of NYSE-listed firms. Thus, if the model is well specified, both the coefficients of (λex- λcum) and of (λex- λcum)*i should be significantly positive. Consistent with the prediction, the estimated bid-ask spreads on the cum-dividend date for both NYSE- and NASDAQ-listed firms are significantly positive, with the latter being significantly larger than the former. The coefficient of λex reflects the total adjustment of bid and ask spread, which is significantly positive for NASDAQ- and NYSE-listed firms, respectively. The positive adjustment of bid and ask spread agrees with recent findings in Ainsworth and Lee (2014) in relation to the Australia stock market, where the effective bid-ask spread increases on the exdividend date. The increase in the bid-ask spread on the ex-dividend date agrees with our argument that on the ex-dividend date, market makers have to carry a larger inventory and thus they increase the bid-ask spread to compensate for higher carrying costs. Thus, contrary to the theoretical assumption made by other studies in the current literature that the bid-ask spread on the cum- and ex-dividend dates remains the same, our result finds an increase in the bid-ask spread on the exdividend date for both NYSE- and NASDAQ-listed stocks. 14

16 Since the intercept is an estimate of negative bid spread (-Sbid) and the coefficient of λex reflects the total adjustment of bid and ask spread, if the adjustment of bid and asked prices on exdividend dates is symmetric, the coefficient of λex should be about twice as much as the negative of intercept estimate. Our test shows that that the coefficient of λex is significantly higher than the double of the negative intercept, suggesting that the increment in asked price being higher than the adjustment of bid price. This result agrees with our argument that the market makers adjust bid and asked prices on the ex-dividend date so that the inventory carrying costs are borne more by inelastic retail investors than by institutional investors. Institutional investors do not face higher tax rates on dividend income relative to capital gains, and thus are more likely to conduct dividend capturing activities and become sellers on the ex-dividend date. The marker makers would adjust bid and asked prices so that retail investors have to pay a bit more at the asked price to buy shares on the ex-dividend date, while institutional investors can sell their shares at relatively better bid prices. Furthermore, we find the asymmetry is more pronounced for NASDAQ stocks. For NASDAQ-listed firms, the estimated intercept and coefficient of λex is positive and different from zero at a significance level of This suggests that after the adjustment of dividend payment, the bid and asked prices on ex-dividend dates are higher than that on cumulative dates. Compared to the adjustment of bid and asked prices of NYSE-listed stocks, the more pronounced asymmetric adjustment of bid and asked prices of NASDAQ-listed firms on the ex-dividend date seems to suggest that market makers of NASDAQ-listed stocks tend to favor institutional investors who conduct dividend capturing activities, while retail investors bear most of the carrying costs on the ex-dividend date. 15

17 C. Ex-dividend Day Excess Trading Volume Lakonishok and Vermaelen (1986) show that there is abnormally high trading volume around ex-dividend dates and that the phenomenon is more prominent for high-yield stocks. To estimate excess trading volume on ex-dividend dates, we compare trading volume during period [-3, 3] (days -3 to 3 relative to the ex-dividend date) to trading volume during period [-45, -4] [4, 45]. Specifically, we compute the average trading volume during days -45 to -4 and 4 to 45 relative to the ex-dividend date and subtract it from trading volume from days -3 to 3 to obtain excess daily trading volume during the event period. The excess daily trading volume is then standardized by the average trading volume during the non-event period. In addition, we estimate the average excess trading volume during the event period [-3, 3] by subtracting the average trading volume during the non-event period from the average trading volume during the period [-3, 3] and then normalize it by the average trading volume during the non-event period. Table 4 Panels A, B, and C present summary statistics of the standardized excess trading volume during the event period for the entire sample, firms listed on the NYSE, and firms listed on the NASDAQ, respectively. Additionally, we conduct t-test to check if the excess trading volume is significantly different from zero. Table 4 Panel A shows that the excess trading volume surrounding the ex-dividend date is significantly positive. For the entire sample, the average excess trading volume on the cum- and ex-dividend dates respectively are 16.8% and 12% of the average trading volume during the nonevent period. The excess trading volume on days -3 and 3 are still significantly positive, but the magnitude is smaller than the abnormal volume on the cum- and ex-dividend dates. The average excess trading volume during event period [-3, 3] is 9.06%, which is significantly positive at a significance level of We find that the excess trading volume surrounding the ex-dividend 16

18 date is higher for firms listed on the NYSE than for firms listed on the NASDAQ. For instance, for NYSE-listed firms, the respective average excess trading volume on the cum- and ex-dividend dates are 20.30% and 17.16% of the average trading volume during the non-event period, which are higher than the values of 12.88% and 6.65% for firms listed on the NASDAQ. Similarly, the average excess trading volume of NYSE-listed firms during the event period [-3, 3] is 10.54%, which is higher than the average value of 7.52% of NASDAQ-listed firms. To further examine if dividends distribution is related to excess trading activities and to identify factors that affect excess trading volume surrounding the ex-dividend date, we conduct the following regression model: EV = β 0 + β 1 DivYield + β 2 MV + β 3 ( 1 P ) + β 4Risk + β 5 Inst + β 6 Inst 2 + β 7 DivYield Inst + β 8 DivYield Inst 2 + ε. The dependent variable is the average excess trading volume during the event period [-3, 3]. DivYield denotes dividend yield and is calculated by dividing dividend amount by stock price on the cum-dividend date. MV represents the market value of the company which is calculated by multiplying the security price by the number of shares outstanding on the cum-dividend date. Following Naranjo, Nimalendran, and Ryngaert (2000) and Dhaliwal and Li (2006), we use 1/P as a proxy for transaction costs, where P is the security price on the cum-dividend date. To control for risk, we use the standard deviation of security returns during days (-50, -6) and (6, 20) relative to the ex-dividend date as an estimate of security risk. Dhaliwal and Li (2006) show that the abnormal trading volume surrounding ex-dividend dates is related to institutional ownership, so we include institutional ownership (Inst) in the regression model and consider its interaction with dividend yield. 17

19 The regression results for the entire sample are presented in Table 5. The abnormal trading volume surrounding ex-dividend dates is positively related to dividend yield at a significance level of The positive relationship between dividend yield and excess trading volume agrees with the dividend capturing theory. We find that the abnormal trading volume is negatively related to risk of securities, with the coefficient of risk being at a significance level of The abnormal trading volume is negatively related to the market value of security and security price. Consistent with the finding of Dhaliwal and Li (2006), abnormal trading volume is affected by institutional ownership and the relationship is nonlinear. The abnormal trading volume surrounding the ex-dividend date increases with ownership heterogeneity. Table 5 also presents the results separately for NYSE- and NASDAQ-listed firms. Overall, for both groups of firms, the excess trading volume is positively related to dividend yield and negatively to security risk and security market value. The evidence suggests that investors engage in dividend capturing behavior especially when dividend yield is high, but they are discouraged by the risk of securities. The relationship between institutional ownership and trading volume, however, is not as clear as when the entire sample is used. We find some evidence of a nonlinear relationship between excess trading volume and institutional ownership, but the evidence is weak. D. Trading Volume and Price Adjustments To further examine whether the asymmetric adjustment of bid and asked prices on the exdividend date is related to dividend capturing activities, we add a dummy variable of high excess trading volume surrounding ex-dividend dates to the regression model: D P = β 0 + β 1 λ ex + β 2 (λ ex λ cum ) + β 3 H + β 4 Hλ ex + β 5 H(λ ex λ cum ), 18

20 where H is a dummy variable that equals one if the abnormal trading volume surrounding exdividend dates is high and zero otherwise. To identify events with high excess trading volume, we rank sample events according to their abnormal trading volume during the event period and identify those ranked within the top one-third as the events with high excess trading volume. Based on the above model, we conduct regression analysis on the NYSE- and NADSAQ-listed firms separately. Additionally, we combine both groups of firms and run the following regression model on the entire sample: D P = β 0 + β 1 λ ex + β 2 (λ ex λ cum ) + β 3 H + β 4 Hλ ex + β 5 H(λ ex λ cum ) + [β 6 + β 7 λ ex + β 8 (λ ex λ cum ) + β 9 H + β 10 Hλ ex + β 11 H(λ ex λ cum )] I where I is an indicator that equals one for NASDAQ-listed firms and zero otherwise, and H is a dummy variable that equals one if the abnormal trading volume surrounding ex-dividend dates is high and zero otherwise. The results are presented in Table 6. For firms listed on the NYSE, we do not find significant effects of excess trading volume on the asymmetric adjustment of bid and asked prices on the ex-dividend date. However, for the NASDAQ sample, we find a more asymmetric adjustment of bid and asked prices on the exdividend date for stocks experiencing high trading volume surrounding the event period. Specifically, the bid price on the ex-dividend date is higher than that on the cum-dividend date by a larger amount for firms with high excess trading volume than for firms without high excess trading volume. Firms with high excess trading volume have a higher adjustment of asked prices on the ex-dividend date than firms without excess trading volume. We also conduct the regression analysis on the entire sample and obtain similar results. The asymmetric adjustments of bid and asked prices are more pronounced for NASDAQ-listed firms, especially when the trading volume is high. Our study provides evidence that the inventory costs surrounding ex-dividend dates is not 19

21 being fairly shared among market participants; market makers adjust bid and asked prices so that more of the costs is borne by buyers of securities on the ex-dividend date than by sellers, who are more likely to conduct dividend capturing activities. Furthermore, the asymmetric adjustment is more pronounced in the NASDAQ than in the NYSE, complementing the conclusion made by Mortal, Paudel, and Silveri (2017) that the NASDAQ has higher execution costs and lower price efficiency than the NYSE before its recent structural change in VI. Evidence After 2003 Since the introduction of SuperMontage at the end of 2002, it is shown that there is improvement in the efficiency of the NASDAQ market (see, e.g., Chung and Chuwonganant, 2009). More importantly, Mortal, Paudel, and Silveri (2017) find that the price-drop-to-dividend ratio of NASDAQ-listed firms converges with that of NYSE-listed firms after To examine if the differences in the asymmetric adjustment of bid and asked prices on the ex-dividend date between the two markets are still significant after the recent structural changes in the NASDAQ, we conduct the regression analysis again while only considering events with ex-dividend dates occurring in 2003 or after. The results are presented in Table 7. We find that the adjustment of bid and asked prices of NASDAQ-listed firms is not significantly different from that of NYSE-listed companies after Consistent with the conclusion made in Mortal, Paudel, and Silveri (2017), our evidence suggests that recent structural changes not only improve the efficiency of the NASDAQ market, but also cause the NASDAQ market to become more similar to the exchange market. However, the coefficient of I* (λ ex λ cum ) is 0.056, which is different from zero at a significance level of Thus, the bid-ask spread 20

22 of NASDAQ-listed companies is still significantly higher than that of NYSE-listed stocks, even after recent changes in the NASDAQ. Table 8 presents the results when trading volume is considered. For the NASDAQ sample, we find that high trading volume is somewhat associated with asymmetric adjustment of bid prices on ex-dividend dates but to a lesser degree when compared to the results from when the entire sample period is used. The result from a combined sample of NYSE- and NASDAQ-listed firms suggests that, for firms with excess trading volume, sellers of NASDAQ-listed stocks on the exdividend date receive more favorable prices than sellers of NYSE-listed stocks. As to asked price, contrary to the result from when the entire sample period is used, we do not see an association between trading volume and adjustment of asked price after Overall, we find the adjustment of bid and asked prices for NASDAQ-listed firms converges with that of NYSE-listed firms, and trading volume has less of an effect on the asymmetric adjustment of bid and asked prices after VII. Conclusions We examine stock price behavior on ex-dividend dates with the consideration of dynamic adjustment of the bid-ask spread. We find that the bid-ask spread is increased on the ex-dividend date for both NYSE- and NASDAQ-listed firms. We show that the market makers tend to adjust the bid and asked prices on the ex-dividend date so that more of the carrying cost is borne by buyers on the ex-dividend date than by sellers who are more likely to conduct dividend capturing activities. Furthermore, the asymmetric adjustment of bid and asked prices on the ex-dividend date is more pronounced for NASDAQ-listed than for NYSE-listed firms, especially when the abnormal trading volume surrounding the ex-dividend date is high. Consistent with previous 21

23 studies, we find positive abnormal trading volume surrounding ex-dividend dates and the trading volume being positively related to dividend yield, but negatively to risk of securities. Our study demonstrates the importance of considering the dynamic adjustment of bid and asked prices while studying the stock price surrounding ex-dividend dates. Moreover, by considering events with exdividend dates occurring in 2003 or after, we find that the asymmetric adjustment of bid and asked prices of NASDAQ-listed firms is not significantly different from that of NYSE-listed companies. The evidence suggests that the recent structural changes in the NASDAQ market, such as the introduction of SuperMontage, not only improve its price efficiency but also cause the NASDAQ market to become more similar to the NYSE market. 22

24 References Ainsworth, Andrew, and Adrian Lee, 2014, Waiting costs and limit order book liquidity: Evidence from the ex-dividend deadline in Australia, Journal of Financial Markets 20, Akhmedov, Umid, and Keith Jakob, 2010, The ex-dividend day: Action on and off the Danish exchange, The Financial Review 45, Bali, Rakesh, and Gailen L. Hite, 1998, Ex-dividend day stock price behavior: Discreteness or taxinduced clienteles, Journal of Financial Economics 47, Barclay, Michael, 1987, Dividend, taxes, and common stock prices: The ex-dividend day behavior of common stock prices before the income tax, Journal of Financial Economics 19, Bell, Leonie, and Tim Jenkinson, 2002, New evidence of the impact of dividend taxation and on the identity of the marginal investor, Journal of Finance 57, Booth, L.D. and D.J. Johnston, 1984, The ex-dividend day behavior of Canadian stock prices: Tax changes and clientele effects, Journal of Finance 39, Boyd, John H., and Ravi Jagannathan, 1994, Ex-dividend price behavior of common stocks, Review of Financial Studies 7, Campbell, James, and William Beranek, 1955, Stock price behavior on ex-dividend dates, Journal of Finance 10, Conrad, Jennifer, and Robert Conroy, 1994, Market microstructure and the ex-date return, Journal of Finance 49, Chung, Kee H., and Chairat Chuwonganant, 2009, Transparency and market quality: Evidence from SuperMontage, Journal of Financial Intermediation 18, Dhaliwal, Dan, and Oliver Li, 2006, Investor tax heterogeneity and ex-dividend day trading volume, Journal of Finance 61, Dubofsky, David A., 1992, A market microstructure explanation of ex-day abnormal returns, Financial Management 21, Dubofsky, David, A., 1997, Limit orders and ex-dividend day return distributions, Journal of Empirical Finance 4, Eades, Kenneth, Patrick Hess, and E. Han Kim, 1984, On interpreting security returns during the ex-dividend period, Journal of Financial Economics 13,

25 Elton, Edwin J., and Martin J. Gruber, 1970, Marginal stockholder tax rates and the clientele effect, Review of Economics and Statistics 52, Frank, Murray, and Ravi 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 47, Graham, John R., Roni Michaely, and Michael R. Roberts, 2003, Do price discreteness and transactions costs affect stock returns? Comparing ex-dividend pricing before and after decimalization, Journal of Finance 58, Green, Richard C., and Kristian Rydqvist, 1999, Ex-day behavior with dividend preference and limitations to short-term arbitrage: The case of Swedish lottery bonds, Journal of Financial Economics 53(2), Henry, Tyler R., and Jennifer L. Koski, 2017, Ex-dividend profitability and institutional trading skill, Journal of Finance 72, Jakob, Keith, and Tongshu Ma, 2003, Order imbalance on ex-dividend days, Journal of Financial Research, Kadapakkam, Palani-Rajan, 2000, Reductions of constraints on arbitrage trading and market efficiency: An examination of ex-day returns in Hong Kong after introduction of electronic settlement, Journal of Finance 55, Kadapakkam, Palani-Rajan, and Valeria Martinez, 2008, Ex-dividend returns: The Mexican puzzle, Journal of Banking and Finance 32, Kalay, Avner, 1982, The ex-dividend day behavior of stock prices: A re-examination of the clientele effect, Journal of Finance 37, Karpoff, Jonathan M., and Ralph A. Walkling, 1988, Short-term trading around ex-dividend days: Additional evidence, Journal of Financial Economics 21, Keim, Donald B., 1989, Trading patterns, bid-ask spreads, and estimated security returns: The case of common stocks at calendar turning points, Journal of Financial Economics 25, Koski, Jennifer, and Roni Michaely, 2000, Prices, liquidity, and the information content of trades, Review of Financial Studies 13, Lakonishok, Josef, and Theo Vermaelen, 1986, Tax induced trading around ex-dividend days, Journal of Financial Economics 16, Lasfer, M. Ameziane, 1995, Ex-day behavior: Tax or short-term trading effects, Journal of Finance 50,

26 Michaely, Roni, 1991, Ex-dividend day stock price behavior: The case of the 1986 Tax Reform Act, Journal of Finance 46, Michaely, Roni, and Maurizio Murgia, 1995, The effect of tax heterogeneity on prices and volume around the ex-dividend day: Evidence from the Milan stock exchange, Review of Financial Studies 8, Michaely, Roni and Jean-Luc Vila, 1995, Investors heterogeneity, prices, and volume around the ex-dividend day, Journal of Financial and Quantitative Analysis 30, Michaely, Roni, and Jean-Luc Vila, 1996, Trading volume with private valuation: Evidence from the exdividend day, Review of Financial Studies 9, Mortal, Sandra, Shishir Paudel, and Sabatino Silveri, 2017, The impact of market structure on exdividend day stock price behavior, Financial Management 46, Naranjo, Andy, M. Nimalendran, and Mike Ryngaert, 2000, Time variation of ex-dividend day stock returns and corporate dividend capture: A reexamination, Journal of Finance 60, Rantapuska, Elias, 2008, Ex-dividend day trading: Who, how, and why?: Evidence from the Finnish market, Journal of Financial Economics 88, Poterba, James, and Lawrence H. Summers, 1984, New evidence that taxes affect the valuation of dividends, Journal of Finance 39, Robin, Ashok, 1991, The impact of the 1986 Tax Reform Act on ex-dividend day returns, Financial Management 20, Whitworth, Jeff, and Ramesh P. Rao, 2010, Do tax law changes influence ex-dividend behavior? Evidence from 1926 to 2005, Financial Management 39,

27 Table 1 Descriptive Statistics This table presents average, median, and standard deviation (S.D.) of the following variables: 1) DivYield: dividend yield, 2) Return: stock return on ex-dividend dates, 3) PDR: price-drop-to-dividend ratio, 4) DivAmt: dividend amount, 5) Price: stock price on exdividend dates, 6) Spread: bid-ask spread on ex-dividend dates, 7) ShareOut: number of shares outstanding, 8) MarketVal: Market value, 9) Vol: trading volume, 10) Ex_Ann: number of days from dividend announcement dates to ex-dividend dates, 11) Pay_Ex: number of days from ex-dividend dates to payment dates. ***, **, and * denote significance at the 1%, 5%, and 10% level, respectively. NYSE NASDAQ Average Median S.D. Average Median S.D. DivYield * Return *** PDR 0.91*** DivAmt 0.23*** Price 35.75*** Spread 0.23*** ShareOut 160,408*** 46, ,269 42,574 10, ,624 MarketVal 6,808,055*** 1,417,029 21,669,123 1,250, ,185 10,774,379 Vol 934,887*** 154,000 3,876, ,490 16,589 2,763,443 Ex_Ann 23*** Pay_Ex 23***

28 Table 2 Summary Statistics of Stock Price Behavior on Ex-dividend Dates This table shows average, median, and standard deviation (S.D.) of stock returns, bid-ask spread, and the probability of buying on exdividend and cum-dividend dates. Bidreturn (Askreturn) is the return on ex-dividend dates based on closing bid (asked) price. Spread_ex and spread_cus represent the bid-ask spread on ex- and cum-dividend dates, respectively. Location_ex and location_cum measure the probability of buying on ex-dividend and cum-dividend dates, respectively. NYSE NASDAQ Average Median S.D. Average Median S.D. Bidreturn Askreturn Spread_ex Location_ex Spread_cum Location_cum

29 Table 3 Adjustment of Bid and Asked Prices on Ex-dividend Dates This table presents regression results of the following model: D -DP = b 0 + b 1 l ex + b 2 ( l ex - l cum ), where D represents dividend amount, ΔP denotes the change of share price from cum-dividend to ex-dividend dates, and cum and ex respectively represent the probability of a buy order on the cum- and ex-dividend dates. The regression is conducted on NYSE-and NASDAQ-listed firms separately and on the entire sample. To control for heterogeneity and autocorrelation, the Newey-West (1987) correction is applied to all regressions. ***, **, and * denote significance at the 1%, 5%, and 10% level, respectively. NASDAQ NYSE ALL Coefficient Std. Error t-statistic Coefficient Std. Error t-statistic Coefficient Std. Error t-statistic Intercept *** ** ** λ ex *** *** *** λ ex λ cum *** *** *** I *** I*λ ex * I* (λ ex λ cum ) *** R-squared Adjusted R-squared F-statistic Prob(F-statistic)