Ex-Dividend Day Bid-Ask Spread Effects in a Limit Order Book Market Setting

Ex-dividend Day Bid-Ask Spread Effects in a Limit Order Book Market Setting

Andrew B. Ainsworth♣

Adrian D. Lee♦

Last updated: November 4, 2010

Abstract

Liquidity and transaction costs are key considerations in short-term trading strategies. Literature on ex-dividend day trading however seldom considers bid-ask spread magnitudes. We find the effective bid-ask spread increases substantially on the ex-dividend day for stocks in the limit order book setting of the Australian Stock Exchange. We attribute the higher spread to: i) a reduction in the cost of delaying execution, ii) an increase in the proportion of impatient traders and iii) uncertainty about the ex-dividend price drop. These effects result in less aggressive limit orders and thereby higher spreads on the ex-dividend day, consistent with predictions of dynamic models of the limit order book. Dividend yield and franking credit level are also positively related to the ex- dividend spread increase as a result of differences in trader valuations. Our results indicate that the actual profitability of traders utilising a dividend-induced trading are likely to be lower after considering the higher spread.

Keywords: bid-ask spread, liquidity, dividend capture, dividend clientele, ex-dividend day.

♣ Corresponding author. Finance Discipline, University of Sydney Business School, University of Sydney, NSW 2006 Australia Ph: +61 2 9036 7992 Email: [email protected] ♦ School of Finance and Economics, University of Technology Sydney, Sydney NSW 2000 Australia.

Ex-dividend Day Bid-Ask Spread Effects in a Limit Order Book Market Setting

Abstract

Keywords: bid-ask spread, liquidity, dividend capture, dividend clientele, ex-dividend day.

1. Introduction

Short term-term trading around the ex-dividend day remains a prevalent feature of equity markets worldwide1, despite Heath and Jarrow (1988) showing that an uncertain price drop between the cum- and ex-dividend days will remove any risk-free arbitrage opportunities. Consideration for liquidity and transaction costs thereby determines the presence and profitability of short-term trading opportunities around the ex-dividend day.2 However, the literature utilising the costly arbitrage models of Boyd and Jagannathan (1994) and McDonald (2001) implicitly assumes that the bid-ask spread is constant around the ex-dividend day. This assumption may not be warranted given the recent evidence of Zhang, Russell and Tsay (2008) that bid-ask spreads are time-varying.

Our study examines the bid-ask spread around the ex-dividend day in the limit order book setting of the Australian Stock Exchange (ASX). The ASX provides a rich environment allowing us to examine whether the bid-ask spread is affected by the limit order book structure and trading on dividends with franking (imputation) credits. For example, Ainsworth, Fong, Gallagher and

Partington (2008) document that buying and selling pressure present around the ex-dividend day in

Australia and that these imbalances significantly impact returns. There is also much evidence that the imputation tax credits paid in Australia do have some value, and market participants alter their trading as a result (see, for example, Walker and Partington (1999), Cannavan, Finn and Gray

(2004), and Ainsworth, Fong, Gallagher and Partington (2010)). Importantly, dividends in

Australia have been increasing (Pattenden and Twite (2008)), rather than decreasing as in the US

(Fama and French (2001)) suggesting a growing importance in trading around the ex-dividend day.

The majority of ex-dividend day studies focus on abnormal returns and abnormal volume and do not directly examine the behaviour of liquidity and the bid-ask spread around the ex-

1 See, for example, Kalay (1982), Eades, Hess and Kim (1984), Lakonishok and Vermaelen (1986), Michaely (1991),

Michaely and Vila (1996), Rantapuska (2008), and Ainsworth, Fong, Gallagher and Partington (2010).

2 See, for example, Kalay (1982), Boyd and Jagannathan (1994), Michaely and Vila (1996), and McDonald (2001).

dividend day, despite its influential role in the ex-dividend equilibrium. An exception is Graham,

Michaely and Roberts (2003) who report the average effective spread on cum-dividend and ex- dividend days for NYSE stocks. Their focus is on the impact of changes in the minimum tick size on the ex-dividend premium and they do not directly test whether the ex-dividend and cum- dividend bid-ask spread are statistically different. A cursory look at Table 1 of Graham, Michaely and Roberts (2003) suggests that both the quoted spread and the effective spread are similar between the ex-dividend day and the cum-dividend day. Koski and Michaely (2000) examine liquidity and the information content of trades around different events, including the ex-dividend day. They find that spreads are wider during announcement periods compared to ex-dividend periods, consistent with information asymmetry increasing the adverse selection component of the spread. Again, they do not directly consider differences in bid-ask spreads on ex-dividend days and ordinary trading days.

There is however, indirect evidence that bid-asks spreads would differ between ex-dividend days and other trading days. For example, in the US, Koski (1996) finds that there is purchasing pressure in the cum-dividend period and selling pressure in the ex-dividend period for high yield stocks in 1983 and 1988. Koski notes that this dividend-driven buying and selling pressure could impact the size of the bid-ask spread. It is hypothesized that the ex-dividend day attracts both uninformed dividend capture and avoidance traders, as well as market makers and other short-term traders, who are indifferent between capital gains and dividends, and trade against the tax-motivated traders (Kalay (1982)). Market makers, in particular, are primarily concerned with providing liquidity and earning the bid-ask spread by buying at the bid price and selling at the higher ask price. Utilising the dynamic models of the limit order book presented by Foucault (1999), Foucault,

Kadan and Kandel (2005), and Roşu (2009), we are able to hypothesise how these different trader groups will choose between market and limit orders in Australia’s pure order-driven market. These models of symmetrically informed agents show that the cost of delaying execution is likely to affect the spread around the ex-dividend day. With the market being closed by a call auction, there is 2

likely to be little change in the proportion of impatient traders before the ex-dividend day.

However, the cost of delaying execution is likely to increase as the ex-dividend deadline approaches. This cum-dividend increase in waiting cost will lead to more aggressively priced limit orders, a more resilient market and a lower bid-ask spread according to these models. After the ex- dividend deadline has passed, there should be a decline in waiting costs for traders in the market, on average. If this is the case, then limit orders will not be priced as aggressively, the market less resilient, and spreads will be higher. Uncertainty about the ex-dividend price drop is also a factor that could lead to a higher spread on the ex-dividend day.

Using executed trades data for stocks listed on the Australian Stock Exchange (ASX) from

1990 to 2008, we find that the ex-dividend effective half-spread is 0.36 cents or 27% higher than the effective half-spread on the cum-dividend day. It is also 0.26 cents or 18% higher than the average daily effective half-spread in a benchmark period between t-50 to t-6 relative to the ex- dividend day. The effective spread peaks on the ex-dividend day and returns nearly immediately to the average spread level. This finding is robust across dividend yields, franking level and a non- event benchmark spread. We investigate several possible explanations for the higher spread. We find that the abnormal volume of on-market trading remains positive on the ex-dividend day suggesting there is no substantial liquidity drain on on-market trading by dividend capture traders going off-market, despite the increase in off-market trading. We find that depth at the best ask quote increases on the cum-dividend day and decreases significantly at both the best bid and ask on the ex-dividend day. This is consistent with more aggressively priced limit orders on the cum- dividend day, and less aggressively priced limit orders on the ex-dividend day. Increased information asymmetry also does not appear to be present as the realized spreads do not differ between the cum- and ex-dividend days, although a decomposition of the spread supports the notion that adverse selection increases on the ex-dividend day. The higher ex-dividend spread is concentrated in the first hour of trading, consistent with uncertainty surrounding the determination of the ex-dividend price and an increase in the proportion of impatient trades that were unable to 3

unwind dividend-induced trading strategies in the opening call auction. However, the spread is persistently higher throughout the ex-day, relative to the cum-dividend day and a non-event benchmark period, suggesting that patient traders have lower expected costs of delaying execution after the ex-dividend deadline has passed. We also find that franking and dividend yield are positively related to the ex-dividend spread after controlling for common determinants. This suggests that valuation differentials and waiting costs are important drivers of the ex-dividend spread.

The remainder of the paper is organised as follows. Section 2 discusses institutional details of the Australian Stock Exchange and the imputation tax system. Section 3 outlines the hypotheses based on the literature regarding dynamic models of the limit order book. Section 4 discusses the data and section 5 presents the results on how the bid-ask spread changes around the ex-dividend day, with section 6 concluding.

2. Institutional Details

The ASX is the only domestic stock exchange in Australia. Stocks are traded using the electronic limit order book system called the Stock Exchange Automated Trading System (SEATS).

This system commenced operation on October 19, 1987 and it fully replaced the trading floor system on September 4, 1990 (see Aitken, Brown and Walter (1996)). There are no designated market makers on the ASX, but brokers are free to trade as principal, or pseudo market makers (see

Aitken, Garvey and Swan (1995)). SEATS opens for trading from 10:00 am and it operates in a continuous open limit order book mode until 4:00 pm. The ASX also operates an opening call auction and in 1997 it introduced a closing call auctions (Comerton-Forde (1999)). Up until

November 25, 2005, SEATS terminals displayed in real time the full limit order book and the broker IDs of the underlying bid and ask orders.

Explicit transactions in effect during the sample period are stamp duty and brokerage commissions. Prior to July 1, 1995, stamp duty was charged at 0.30% of the transaction amount.

After July 1995 and prior to July 1, 2001, stamp duty was halved to 0.15%. From July 1, 2001, stamp duty was abolished.3 Brokerage is typically charged as a percentage of the transaction amount. Since brokerage commissions are deregulated during the sample period, the rates brokers charge varies across time and broker.

Australia introduced an imputation tax system in July 1987.4 There have been four notable changes in the imputation system that would affect ex-dividend trading. We separate our sample into four tax regimes to assist in our hypothesis testing. Tax regime 1 is from the beginning of the sample in February 1990 until June 30, 1997. After June 30, 1997, a 45-day holding period rule was announced whereby investors were required to hold a stock for at least 45 days to be eligible for the tax credits (tax regime 2). However, the holding period rule was only enacted on July 15,

1999 (tax regime 3). A 50% discount on capital gains was also introduced on September 16, 1999 and is included in tax regime 3. On June 30, 2000 the Australian government introduced cash refunds for unutilised franking credits (tax regime 4). Certain investors including individuals, complying pension funds, and income tax-exempt charities were able to fully benefit from the tax credits, irrespective of their tax liabilities. Changes also occur to the corporate tax rate during our sample period. The corporate tax rate was lowered from 36% to 34% on July 1, 2000 and further reduced to 30% on July 1, 2001. The reduction in the tax rate reduces the value of franking credits, though this effect is likely to be minimal.

3. Theory and Hypothesis Development

Our hypotheses are developed from the dynamic models of the limit order book presented by Foucault (1999), Foucault, Kadan and Kandel (2005), and Roşu (2009). These models do not include informed traders, with trading motivated by liquidity needs. The ex-dividend day provides a natural setting to examine the role of liquidity trading in financial markets as information

3 ASX Media Release ‘No stamp duty on securities’: http://www.asx.com.au/about/pdf/stampduty010701.pdf

4 See Ainsworth, Fong, Gallagher and Partington (2008) for a more detailed description of the imputation system.

asymmetry is likely to be lower following earnings and dividend announcements that precede the ex-dividend day. These dynamic models of the limit order book involve traders with subjective valuations above and below fundamental value. In the ex-dividend setting, the subjective valuation could reflect differences between tax rates on capital gains and dividend income, and the ability to utilise tax credits. Dividend capture traders, such as individual investors, will place a high value on the dividend and tax credit and will buy on the cum-dividend date and/or sell on the ex-dividend date. Foreign investors are unable to utilise tax credits and would therefore avoid the dividend payment. Depending on the size of the subjective valuation, certain investors could be indifferent to the dividend.

Foucault (1999) focuses on the risk of being picked off and the risk of non-execution. The risk of being picked off increases on the ex-dividend day, given the uncertainty surrounding the ex- dividend price. Foucault shows that the spread has a reservation component related to adverse selection and an execution risk component that is related to non-competitive behaviour. The bid- ask spread will increase when the execution risk of limit orders increases as limit order traders capture a larger share of the differences in the subjective valuations (e.g. tax credits). Foucault notes that impatience will increase if there is an increase in the relative number of traders who place a higher subjective value on the stock. In the ex-dividend setting, this equates to an increase in the proportion of traders who place a higher value on the dividend and tax credit. Traders will need to increase their bid prices to offset the decline in execution probability because of more competition on the buy side of the order book. Traders who do not value the dividend and tax credit as highly will switch to using sell limit orders to capture a greater share of the bid-ask spread, with improved execution probabilities.

Foucault, Kadan and Kandel (2005) and Roşu (2009) formally include trader impatience in their models of limit order book dynamics. Impatient traders have a larger waiting cost per unit of time and the expected total waiting cost is determined by the delay between order submission and execution. Foucault, Kadan, and Kandel show that the market will be more resilient, and the bid- 6

ask spread subsequently smaller, when the population of traders is dominated by patient traders, or if waiting costs are relatively higher. Patient traders will submit aggressive limit orders to reduce their time to execution if there is more competition from other patient traders. Similarly, if the cost of delaying execution increases, then there is an incentive to submit more aggressive limit orders to reduce this cost.

Waiting costs and the risk of non-execution are likely to increase as the ex-dividend day approaches for dividend-induced short-term traders or long-term traders accelerating their pre- determined purchase or sale decisions to the cum-dividend period. If waiting costs were to increase by a large enough amount for a subset of patient traders, then the proportion of impatient traders may also increase as these formerly patient traders switch to submitting market orders. However,

Pagano and Schwartz (2003) show that spreads are actually smaller in the last half-hour of trade when the market closes using a call auction. They posit that the closing call auction will lead to the placement of more limit orders instead of market orders. This finding is consistent with the predictions of Foucault, Kadan, and Kandel, as the proportion of impatient traders will not increase.

As the ASX operates a closing call auction we anticipate that the proportion of impatient traders would not increase substantially as there exists an opportunity to trade before the close on the cum- dividend day, albeit at an uncertain price. After a stock begins trading ex-dividend, it is likely that waiting costs will be relatively lower in the absence of a trading deadline. Any prediction about the proportion of impatient traders in the ex-dividend period hinges upon whether they are able to fill their orders in the opening auction. Therefore, we are uncertain as to whether the composition of patient and impatient traders will change on the ex-dividend day.

In summary, the model of Foucault, Kadan, and Kandel implies that the bid-ask spread in the cum-dividend period should be lower, and the market more resilient, as patient traders post more aggressively priced limit orders to reduce the time to execution, given higher waiting costs.

On the ex-dividend day, limit orders will not be priced as aggressively, spreads will be higher, and the market less resilient, as a result of lower waiting costs. We also anticipate an increase in 7

adverse selection on the ex-dividend day due to the uncertainty regarding the ex-dividend price, and therefore an increase in the bid-ask spread. We expect that differences in the bid-ask spread between the ex-dividend day and the cum-dividend day will be larger in those stocks paying full franked dividends at a high yield, as this is where differences in subjective valuation are hypothesised to be largest. These are also the stocks where long-term traders accelerate trades to the cum-dividend period, and will therefore have reduced trading on the ex-dividend day.

Furthermore, we hypothesise that valuation differentials are likely to be greatest earlier in our sample (tax regime 1), before tax credit eligibility restrictions were introduced, and later in our sample (tax regime 4), when certain investors, including individuals, charities, and pension funds, became eligible for cash refunds of previously unutilised franking credits.

4. Data

We obtain intraday trading data from the ASX SEATS system supplied by the Securities

Industry Research Centre of Asia-Pacific (SIRCA) for the period February 19, 1990 to December

31, 2008. This data captures all trades that took place on the ASX SEATS system where all trading by member brokers are reported. Each transaction in the dataset consists of the timestamp, ticker, price, bid and ask quotes just prior to each transaction, trade flags and buyer and seller broker identifiers (IDs). Trade flags indicate whether the trade was buyer or seller initiated, an opening or closing auction trade, technical crossing, an off-market trade, or an odd-lot trade. We also source closing day prices from SIRCA. Dividends, ex-dividend dates, capitalisation adjustments, and month-end share market capitalisation data are sourced from the Australian School of Business’

Centre for Research in Finance Share Price and Price Relative Database (CRIF SPPR). Similarly to

Bell and Jenkinson (2002) we limit the sample to the largest 250 stocks by market capitalisation at the end of the prior month in order to remove thin trading stocks. Furthermore, we remove dividend events where the cum-dividend day stock price is below $1 and exclude foreign stocks such as US depositary receipts.

Following Huang and Stoll (1996), the effective half-spread is measured as the absolute value of the traded price less the midpoint of the bid and ask:

 at + bt  Effective Spread = pt −   , (1)  2 

where pt is the traded price, at and bt are the ask and bid prices immediately prior to the trade occurring respectively. We also calculate the realized half-spread following Huang and Stoll

(1996). The realized half-spread measures the price movement after a trade is executed from the perspective of the limit order and is calculated as:

Realized half − spread = ( pt+τ − pt )δ , (2) where pt is the traded price, pt+τ is the first trade price after τ minutes have passed and δ equals +1 if the trade is a buy limit order (traded at the bid price), or equals -1 if the trade is a ask limit order

(traded at the ask price). We set τ to between five minutes and no later than ten minutes. If no trade occurs after ten minutes on the day, we do not calculate a realized spread.

Following Michaely and Vila (1996) we calculate abnormal volume as the ratio of daily volume divided by the total shares outstanding (Vi) over the expected volume (EVi). The expected volume is measured as the average daily Vi between t-45 to t-6 and t+6 to t+45 from the ex-dividend date as:

t−6 t+45 ∑Vi,t + ∑Vi,t t−45 t+6 EV = , (3) i T where T is the number of days the stock traded in the 80-day estimation window. The abnormal volume (AVi,t) for each day between t-5 and t+5 is calculated as:

Vi,t AVi,t = −1. (4) EVi

5. Results

5.1 Descriptive Statistics

Table 1 reports the descriptive statistics for the dividend events. As we are focusing on the dollar spread, it is important to note that the mean dividend is 14 cents and median dividend is substantially smaller at eight cents. Any changes in the effective spread will be eroding an amount of these dividend values. The dividend yield is calculated as the cash dividend divided by the closing cum-day price, and has a mean of 2.27%. The median franking level of 100% shows that most companies pass on tax credits to investors, although the average is around 64%. The high proportion of fully franked dividends is also evident in Panel B where more than half of the 5781 dividend events in the entire sample are fully franked. The distribution of dividend events across tax regimes is also worth highlighting, with over 80% of the dividend events occur in tax regimes 1 and 4, and only 11% and 6% in tax regimes 2 and 3, respectively.

[INSERT TABLE 1]

5.2 Effective Half-Spread around the Ex-Dividend Day

Table 2 reports our baseline results showing that the effective half-spread is statistically higher on ex-dividend days than on cum-days or days prior to the ex-dividend day. For comparison to the effective half-spread on the ex-dividend day, we use the spread on the cum-dividend day and a benchmark that is the average daily spread from t-50 to t-6 relative to the ex-dividend day. As shown in Panel A for the entire sample, the ex-dividend day half-spread of 1.695 cents is 0.26 cents higher than the average spread from t-50 to t-6 and 0.36 cents higher than the cum-dividend day spread. Both differences are statistically significant at the one percent level. This represents about an 18 to 26% increase in the average cost of trading using market orders. To compare the magnitude of the ex-dividend day spread to other days, we examine the daily average spread between t-45 to t+45 of the ex-dividend day in Figure 1. As can be seen, the spread on the ex-

dividend day is the highest over the 91 days, and ranges from about 1.3 cents to 1.45 cents across the other days.

[INSERT TABLE 2]

[INSERT FIGURE 1]

The results in Table 2 show that the spread premium is positive and mostly statistically significant when sorting by stock characteristics and time. Table 2 Panel B sorts dividend events into three equally sized groups by dividend yield. As seen in the ‘Ex-Benchmark’ column, the spread premium increases across all three groups by around 0.35 cents, though in percentage terms the average difference between the cum- and ex-dividend days is monotonically increasing from

19% in the low dividend yield group to 35% in the high dividend yield group. We find similar results for the franking level (Panel C) and three equally sized groups sorted by the benchmark effective half-spread (Panel D). The spread premium increases from high spread stocks (based on the t-50 to t-6 daily average spreads) of 1.06 cents to 0.04 cents in the low spread stocks, statistically significant at the one percent level. Consistent results are found for difference between the cum- and ex-dividend days.

While our results clearly show that the ex-dividend effective spread is higher than on other days, we find mixed evidence supporting our hypotheses. We find that that percentage increase in the spread between the cum- and ex-dividend days is higher for fully franked stocks and high yield stocks. These observations are consistent with our predictions regarding changes in waiting costs from theoretical models of an order-driven market, as the spread varies inversely with waiting costs.

If we assume that waiting costs do not change then our results indicate that there is actually an increase in the proportion of impatient traders on the ex-dividend day relative to the cum-dividend day. This could be due to traders unable to unwind short-term trading positions in the opening auction and a desire to avoid price risk by using market orders to reduce time to execution.

However, we do not find support for the changes across tax regimes, with the differences in the spread actually lower in tax regime 1 and 4 (Panel E). 11

5.3 Abnormal Volume

In this section we test whether the trading volume and on-market liquidity can explain the spread difference between the ex-dividend day and other days. The higher spread on the ex- dividend day may be due to traders opting to trade off-market in order to buy or sell a large quantity of shares quickly without price impact. Thus, uninformed dividend avoidance or capture trading off-market may increase on-market information asymmetry and thereby increase the bid-ask spread

(Seppi (1990) and Grossman (1992)). Given, the documented increase in the bid-ask spread, we expect the ex-dividend day to have negative abnormal on-market volume and positive abnormal off- market volume if this switch takes place. We would also expect a stronger effect on volume for stocks where the difference in valuation amongst traders is greater – namely fully franked and high yield stocks.

Table 3 reports average abnormal volume for all, off-market only and on-market only trades between t-5 to t+5 of the ex-dividend day. We find that abnormal volume on the ex-dividend day is positive and statistically significant for both on-market and off-market trades. The higher volume traded in the off-market on the ex-dividend day does not appear to significantly remove volume from the on-market. For the full sample in Table 3 Panel A, off-market volume on the ex-day is over 240% higher than expected volume, statistically significant at the one percent level. This is also higher than on the cum-dividend day (82%), and the highest of all the surrounding days. On- market abnormal volume is positive and statistically significant in the five days prior to the ex- dividend day, and on the ex-dividend day itself. This suggests that the higher overall trading volume on the ex-dividend day is primarily due to the substantial increase in off-market trading.

However, while there is a greater concentration in off-market trading, on-market trading is not abnormally lower and therefore, is inconsistent with uninformed dividend capturers abnormally reducing liquidity on the ex-dividend day. Sorts on dividend yield (Panel B), franking level (Panel

C) and average effective spread from t-50 to t-6 (Panel D) show results consistent to the full sample.

The exception is for on-market trading in the high dividend yield and fully franked groups on the 12

ex-dividend day with statistically significant abnormal volume of 9.2% and 7.7%, respectively.

Medium spread stocks also have abnormally higher on-market volume of 6.2%. In sub sampling by tax regimes in Panel E, we find negative and statistically significant abnormal volume of -9% during the ex-dividend day in tax regime 3 consistent with the 45-day rule being enacted and reducing the profitability of short-term trading strategies. While this suggests that the lower volume creates higher spreads, it cannot fully explain why we find the spread premium in all our tax regimes.

[INSERT TABLE 3]

5.4 Information Asymmetry

While the abnormal volume suggests that there is no reduction in trading volume and therefore a fall in information asymmetry, it does not directly measure information asymmetry. To directly measure information asymmetry, we examine the realized spread on the ex-dividend day and compare it to the cum-dividend day and the t-50 to t-6 average. Table 4 reports the realized half-spreads. We find that realized spreads do not differ greatly from the ex-dividend day compared with the spread on the cum-day. However, the realized spread does differ significantly from the average from t-50 to t-6 for those stocks paying high yield, fully franked dividends with a low non- event bid-ask spread. Therefore, there is only weak evidence that the increase in the spread on the ex-dividend day is in response to higher information asymmetry. However, the results could indicate that traders’ order placement strategies respond differently to information asymmetry on the cum-dividend day than on the ex-dividend day.

[INSERT TABLE 4]

5.5 Abnormal Market Depth

This section looks at whether reduced market depth could explain the difference between the spread on the ex-dividend day and other trading days. If waiting costs increase then limit order traders are likely to improve upon the best quotes to ensure quick execution, so we anticipate more 13

aggressively priced limit orders. For stocks already trading at the minimum spread this could lead to an increase in depth at the best bid and ask during the cum-dividend period. On the ex-dividend day, a decline in waiting costs will reduce the aggressiveness of orders and a decline in the depth is expected given the reduced need for the majority of traders to have quick execution.

We calculate an abnormal market depth measure for the first level of bid and ask depth separately in a similar manner to the calculation of abnormal volume. We report our results in

Table 5. Our results show that on the cum-dividend day the ask depth is positive and statistically significant for the entire sample, high dividend yield, fully franked and low spread stocks where the gains from trade are greatest. In tax regimes 3 and 4 both the bid and ask depth are abnormally larger. The increase in depth is expected for tax regime 4 after tax credits are refunded, however, we do not expect the increase to occur in tax regime 3. On the ex-dividend day the abnormal bid and ask depth is negative and statistically significant for the entire sample (Panel A). The abnormal depth is also negative and statistically significant at either the bid or ask across dividend yields

(Panel B), franking levels (Panel C) and tax regimes (Panel D). Although the lower depth is not consistent across the different sorts on stock characteristics, the results show that patient liquidity suppliers do become less aggressive. Therefore, the changes in the depth are consistent with our predictions regarding change in the waiting costs faced by patient traders.

[INSERT TABLE 5]

5.6 Intraday effective spreads

To further understand the potential causes of the ex-dividend spread increase, we examine the hourly effective spreads throughout the trading day. The hourly spreads will allow us to determine whether the opening auction and uncertainty plays any role in the ex-dividend spread premium. Table 6 shows that the greatest difference in the spread is concentrated in the first hour of trade after the market opens on the ex-dividend day. However, the ex-dividend spread is consistently higher than both the cum-dividend spread and the benchmark spread throughout the

day, albeit declining during the middle of the day. This pattern of spread behaviour is pervasive across all stock characteristics and tax regimes, and suggests there are potentially two factors driving the ex-dividend spread premium early in the trading day. Firstly, uncertainty surrounding the ex-dividend price drop after the opening auction could lead to a widening of spreads that is resolved as the trading progresses. Secondly, impatient traders who do not execute their orders in the opening call auction are likely to submit market orders. As a result, there could potentially be an increase in the proportion of impatient traders early in the trading day that declines after they unwind their dividend trading strategies. The larger bid-ask spreads early in the trading day could arise as a result of a temporary increase in trader impatience coupled with the decline in waiting costs for patient traders after the ex-dividend day.

[INSERT TABLE 6]

5.7 Spread Decomposition

To determine whether adverse selection is an important factor in the spread increase on the ex-dividend day, we undertake a decomposition of the spread following Lin, Sanger and Booth

(1995). We exclude opening and closing auction trades, and off-market trades in the decomposition. The results in Table 7 show that the adverse selection component of the spread does increase between the cum- and ex-dividend days. Quote revisions as a proportion of the effective spread increase from 49% to 53%. This increase is likely to reflect uncertainty regarding the fair value of the ex-dividend security. Given that there is less depth at the best quotes on the ex- dividend day, it is not surprising that the adverse selection component of the spread increases as quote revisions by a given trader are more likely to affect the quoted spread. The 3.4% increase in the adverse selection component for the full sample between the cum- and ex-dividend days equates to 0.234 cents or 65% of the average difference between the spreads on the cum- and ex-days presented in Table 2. Adverse selection makes up 74% of the spread difference between the ex- dividend day and the benchmark period. The order processing component declines as a percentage

on the ex-day, but still increases by 20% in dollar terms. The order persistence component accounts for 15% of the dollar increase in the effective spread.5 The results for the adverse selection component are generally similar in percentage terms across the different stock characteristics, except for the low benchmark spread group, where the increase in adverse selection accounts for the entire increase in the spread. These are the stocks where short-term traders are most active in implementing their strategies and monitoring of limit orders may be higher.

[INSERT TABLE 7]

5.8 Regression evidence

Although the decomposition results provide an insight into how the components of the spread change between the cum- and ex-dividend days, it is important to ascertain what observable factors are driving the change in spreads. The determinants of the spread are estimated using common factors contained in the literature (Stoll (2000) and Comerton-Forde and Tang (2009)) as well as dividend related factors:

Ex Ex Ex ESi,t = α 0 + α1Di,t + α 2Volumei,t + α 3Volumei,t Di,t + α 4σ i,t + α 5σ i,t Di,t + α 6Ticki,t . (5) Ex Ex Ex +α 7Ticki,t Di,t + α 8 Franki,t + α 9 Franki,t Di,t + α10 DYi,t + α11DYi,t Di,t + ε i,t

Ex ESi,t is the effective spread, Di,t is a dummy variable equal to one for ex-dividend days and zero

for cum-dividend days, Volumei,t is the natural logarithm of average daily trading volume between

t-50 and t-6, relative to the ex-dividend day, σ i,t is the stock return variance over the same time

interval, Ticki,t is the minimum tick size for each stock based on its ex-dividend price, Franki,t is

the percentage to which the dividend carries tax credits, and DYi,t is the dividend as a percent of the cum-dividend day closing price. The ex-day dummy variable is interacted with explanatory variables to assess whether there exert different influences on the spread once a stock beings trading

5 For brevity, we omit the results for the order processing and order persistence components. These are available on request from the authors.

ex-dividend. We report t-statistics based on standard errors clustered by stock and ex-dividend date.

Table 8 contains the regression results. Column 1 contains the spread determinants, excluding dividend yield and the percentage level of franking. The volume, volatility and tick size all have the expected signs and are statistically significant on the cum-dividend day. On the ex- dividend day, the effect of volume on the spread decreases significantly from -0.782 to -1.028 per unit of benchmark period log volume. Interestingly, these cross-sectional variables are unable to explain the difference in spread, with the ex-dividend dummy positive and significant. These coefficients are broadly similar when we include franking, dividend yield and their interaction with the ex-dividend day dummy variable (column 2). Fully franked and higher dividend yield stocks have lower spreads on the cum-dividend day, consistent with differences in valuation affecting trading. Fully franked stocks experience a 0.2 cent increase on the ex-dividend day, with higher dividend yield stocks also having higher spreads once ex-dividend trading begins. However, the ex- dividend dummy variable remains significant. The inclusion of fixed effects (columns 3 and 4) does not have a substantial impact on the magnitude of the coefficients. The tick size becomes insignificant as there will be less stock-level variation in tick size. Fully franked stocks no longer have lower spreads on the cum-dividend day. We are still able to detect higher spreads for fully franked and higher yield stocks at the 10% significance level on the ex-dividend day. The ex- dividend day dummy variable is still significant and positive when fixed effects are included.

[INSERT TABLE 8]

Increases in waiting costs are likely to be more important in stocks where valuation differences amongst traders are greatest. If this is the case then we would expect differences in spreads to be larger for fully franked, high yield stocks and for increases to be smaller for stocks paying unfranked and low yield dividends. Furthermore, the particular functional form in which dividend yield and franking enter into Equation 5 is unknown. To address these issues, we estimate

Equation 5, omitting franking and dividend yield, for each intersection of the dividend yield groups 17

and the franking level groups. The results are presented in Table 9. The fully franked and higher yield regression results in columns 3 and 5 to 8 mirror those contained in Table 8. Where significant, all coefficients have the expected sign.6 The key result is that for stocks with a lower yield and lower franking levels, there is no significant difference between the cum- and ex-dividend day spreads, as the ex-dividend day dummy is insignificant (columns 1, 2 and 4). These are the stocks where we do not expect substantial differences in valuation, and therefore substantial gains from trade. This is consistent with predictions from limit order book models, as waiting costs for these stocks are unlikely to change around the ex-dividend day, and therefore, we do not observe a spread increase.

[INSERT TABLE 9]

6. Conclusion

We document a substantial increase in the bid-ask spread on the ex-dividend for stocks listed on the ASX between 1990 and 2008. The ex-dividend day effective half-spread is around 18 to 27% higher than both the cum-dividend day and a non-event benchmark period. The effective spread spikes upwards on the ex-dividend day before returning to its average level. This pattern of spread behaviour is pervasive across dividend yields, franking level, a non-event benchmark spread, as well as changes in Australia’s tax code. Theoretical models of the limit order book (Foucault

(1999), Foucault, Kadan and Kandel (2005), and Roşu (2009)) show that adverse selection, trader impatience and the cost of delaying execution are likely to influence the size of the spread in a pure order-driven market. We hypothesise that the cost of delaying execution will increase in the cum- dividend period before declining after the stock begins trading ex-dividend. As a result, patient traders will post more aggressive limit orders in the cum-dividend period leading to lower spreads.

6 The results are similar if stock fixed effects are included.

After the ex-dividend deadline has passed, patient traders will submit less aggressive limit orders and spreads will be wider.

Consistent with these predictions, we find that depth at the best ask quote increases on the cum-dividend day and decreases significantly at both the best bid and ask quote on the ex-dividend day. In addition, the ex-dividend spread increase is concentrated in the first hour of trading, consistent with both uncertainty surrounding the determination of the ex-dividend price and an increase in the proportion of impatient trades that were potentially unable to unwind dividend- induced trading strategies in the opening call auction. However, the spread is persistently higher throughout the ex-dividend day consistent with a decline in waiting costs for patient traders. At the daily level, we do not observe an increase in information asymmetry as realized spreads do not increase. However, a spread decomposition shows that the adverse selection component of the spread accounts for about two-thirds of the spread increase, indicating that that uncertainty regarding the ex-dividend price is important. Analysis on determinants of the spread indicate that franking level and dividend yield are positively associated with the ex-dividend spread, consistent with taxation induced valuation differentials. The results in this paper have important implications for the presence of profitable short-term trading opportunities, indicating that expected profits are likely to be higher than actual profits for traders utilising a dividend-induced trading strategy.

References

Ainsworth, A. B., K. Y. Fong, D. R. Gallagher, and G. Partington, (2008). Taxes, price pressure and order imbalance around the ex-dividend day, University of Sydney. Ainsworth, A. B., K. Y. L. Fong, D. R. Gallagher, and G. Partington, (2010). Institutional trading around the ex-dividend day, University of Sydney. Aitken, M., P. Brown, and T. Walter, (1996). Infrequent trading and firm size effects as explanations for intraday patterns in returns on seats, SIRCA. Aitken, M. J., G. T. Garvey, and P. L. Swan, (1995). How brokers facilitate trade for long-term clients in competitive securities markets, Journal of Business 68, 1-33. Bell, L., and T. Jenkinson, (2002). New evidence of the impact of dividend taxation and on the identity of the marginal investor, Journal of Finance 1321-1346. Boyd, J. H., and R. Jagannathan, (1994). Ex-dividend price behavior of common stocks, Review of Financial Studies 7, 711-741. Cannavan, D., F. Finn, and S. Gray, (2004). The value of dividend imputation tax credits in Australia, Journal of Financial Economics 73, 167-197. Comerton-Forde, C., (1999). Do trading rules impact on market efficiency? A comparison of opening procedures on the Australian and Jakarta stock exchanges, Pacific-Basin Finance Journal 7, 495-521. Comerton-Forde, C., and K. M. Tang, (2009). Anonymity, liquidity and fragmentation, Journal of Financial Markets 12, 337-367. Eades, K. M., P. J. Hess, and E. H. Kim, (1984). On interpreting security returns during the ex- dividend period, Journal of Financial Economics 13, 3-34. Fama, E. F., and K. R. French, (2001). Disappearing dividends: Changing firm characteristics or lower propensity to pay?, Journal of Financial Economics 60, 3-43. Foucault, T., (1999). Order flow composition and trading costs in a dynamic limit order market, Journal of Financial Markets 2, 99-134. Foucault, T., O. Kadan, and E. Kandel, (2005). Limit order book as a market for liquidity, Review of Financial Studies 18, 1171-1217. Graham, J. R., R. Michaely, and M. R. Roberts, (2003). Do price discreteness and transactions costs affect stock returns? Comparing ex-dividend pricing before and after decimalization, Journal of Finance 58, 2611-2636. Grossman, S., (1992). The informational role of upstairs and downstairs trading, Journal of Business 65, 509-528. Heath, D. C., and R. A. Jarrow, (1988). Ex-dividend stock price behavior and arbitrage opportunities, The Journal of Business 61, 95-108. Huang, R., and H. Stoll, (1996). Dealer versus auction markets: A paired comparison of execution costs on NASDAQ and the NYSE, Journal of Financial Economics 41, 313-357. Kalay, A., (1982). The ex-dividend day behavior of stock prices: A re-examination of the clientele effect, Journal of Finance 37, 1059-1070. 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 69, 313-338. Koski, J. L., and R. Michaely, (2000). Prices, liquidity, and the information content of trades, Review of Financial Studies 13, 659-696. Lakonishok, J., and T. Vermaelen, (1986). Tax-induced trading around ex-dividend days, Journal of Financial Economics 16, 287-319. Lin, J.-C., G. C. Sanger, and G. G. Booth, (1995). Trade size and components of the bid-ask spread, Review of Financial Studies 8, 1153-1183. McDonald, R. L., (2001). Cross-border investing with tax arbitrage: The case of German dividend tax credits, Review of Financial Studies 14, 617-657.

Michaely, R., (1991). Ex-dividend day stock price behavior: The case of the 1986 Tax Reform Act, Journal of Finance 46, 845-859. Michaely, R., and J. L. Vila, (1996). Trading volume with private valuation: Evidence from the ex- dividend day, Review of Financial Studies 9, 471-509. Pagano, M. S., and R. A. Schwartz, (2003). A closing call's impact on market quality at euronext paris, Journal of Financial Economics 68, 439-484. Pattenden, K., and G. Twite, (2008). Taxes and dividend policy under alternative tax regimes, Journal of Corporate Finance 14, 1-16. Rantapuska, E., (2008). Ex-dividend day trading: Who, how, and why?: Evidence from the Finnish market, Journal of Financial Economics 88, 355-374. Roşu, I., (2009). A dynamic model of the limit order book, Review of Financial Studies 22, 4601- 4641. Seppi, D., (1990). Equilibrium block trading and asymmetric information, Journal of Finance 73- 94. Stoll, H. R., (2000). Friction, The Journal of Finance 55, 1479-1514. Walker, S., and G. Partington, (1999). The value of dividends: Evidence from cum-dividend trading in the ex-dividend period, Accounting and Finance 39, 275-296. Zhang, M. Y., J. R. Russell, and R. S. Tsay, (2008). Determinants of bid and ask quotes and implications for the cost of trading, Journal of Empirical Finance 15, 656-678.

Table 1 Descriptive Statistics

The sample is for dividends paid between February 1990 and December 2008. Panel A reports summary statistics on the dividend events. The cash dividend is the dividend payment in cents. The franking level is the percentage tax credit attached to the cash dividend. The dividend yield is measured as the cash dividend divided by the closing cum-day price. Panel B reports the distribution of dividend events across the various tax regimes and franking levels across the sample period.

Panel A. Summary Statistics Mean Median Q1 Q3 Std. Dev. Min Max Cash Dividend (c) 14.05 8.00 4.50 15.00 21.99 0.10 280.00 Franking Level (%) 64.24 100.00 0.00 100.00 45.53 0.00 100.00 Dividend Yield (%) 2.27 2.08 1.53 2.76 1.66 0.06 79.37 Franking Credit Yield (%) 0.34 0.33 0.00 0.55 0.34 0.00 8.40 Panel B. Distribution of Ex-Dividend Day Events by Tax Regime and Franking Level Fully Franked Partially Franked Unfranked Total Tax Regime 1 924 148 509 1581 Tax Regime 2 348 67 232 647 Tax Regime 3 171 38 119 328 Tax Regime 4 1828 383 1014 3225 Total 3271 636 1874 5781

Table 2 Effective Half-Spreads around the Ex-Dividend Day

This table reports the equally weighted average effective half-spread in cents measures across dividend events. Benchmark is the average effective half-spread from t-50 to t-6 days prior to an ex-dividend date for a given stock. T- stats using clustered standard errors on the ex-dividend date are reported for the paired differences. ‘Ex- Benchmark’ and ‘Ex-Cum’. **, * denote significance at the 1 and 5% level.

Panel A. Entire Sample Ex- Benchmark Cum Ex T Ex-Cum T N Benchmark 1.439 1.338 1.695 0.256** (7.83) 0.357** (9.40) 5781 Panel B. Dividend Yield 0(Low) 2.006 1.899 2.262 0.256** (3.27) 0.363** (3.90) 1927 1 1.224 1.142 1.517 0.292** (5.75) 0.374** (7.05) 1927 2(High) 1.087 0.973 1.306 0.219** (7.30) 0.333** (10.15) 1927 Panel C. Franking Level Zero 1.472 1.446 1.645 0.174** (2.60) 0.199* (2.36) 1874 Partly 1.130 1.084 1.415 0.285** (3.90) 0.331** (4.38) 636 Fully 1.481 1.326 1.778 0.297** (7.82) 0.452** (11.19) 3271 Panel D. Benchmark Effective Half-Spread 0(Low) 0.529 0.529 0.571 0.042** (9.96) 0.041** (9.54) 1927 1 0.831 0.803 0.973 0.143** (12.79) 0.170** (13.77) 1927 2(High) 2.959 2.682 3.541 0.582** (6.06) 0.859** (7.72) 1927 Panel E. Tax Regimes Regime 1 1.617 1.492 1.910 0.293** (4.97) 0.418** (5.48) 1581 Regime 2 1.647 1.473 2.052 0.405** (2.89) 0.579** (4.55) 647 Regime 3 1.595 1.434 2.081 0.485** (3.65) 0.647** (3.62) 328 Regime 4 1.295 1.226 1.479 0.184** (4.64) 0.253** (5.44) 3225

Table 3 Abnormal Volume around the Ex-Dividend Day

This table reports the equally weighted average daily abnormal volume, between t-5 and t+5 of the ex-dividend day for all trades, off-market trades and on-market trades. Abnormal volume is calculated as the actual daily volume divided by the number of shares outstanding over the expected volume divided by the number of shares outstanding. The expected volume for a dividend event is the average volume between t-45 and t-6 and t+6 to t+45. The sample period is from Feb 1990 to Dec 2008. **, * denote significance at the 1 and 5% level using t-stats with clustered standard errors on the ex-dividend date.

Panel A. Full Sample Type t-5 t-4 t-3 t-2 t-1 Ex-day t+1 t+2 t+3 t+4 t+5 N All 0.073** 0.128** 0.141** 0.174** 0.286** 0.191** 0.028* 0.020 -0.009 0.044* -0.002 5781 Off -0.349** -0.123 -0.192** -0.035 0.823** 2.427** 0.088 -0.232** -0.350** -0.026 -0.273** 5781 On 0.075** 0.131** 0.152** 0.177** 0.261** 0.044** 0.017 0.020 -0.002 0.018 0.002 5781 Panel B. Dividend Yield 0(Low) All 0.038 0.095** 0.110** 0.138** 0.148** 0.173** 0.035 0.062* -0.007 0.037 0.012 1927 1 0.092** 0.116** 0.098** 0.121** 0.243** 0.195** 0.002 0.010 -0.006 0.048 0.000 1927 2(High) 0.090** 0.173** 0.217** 0.263** 0.467** 0.205** 0.047 -0.013 -0.013 0.049 -0.019 1927 0 Off -0.425** -0.364** -0.406** -0.268* 0.470** 2.629** 0.118 -0.309** -0.306** -0.116 -0.295** 1927 1 -0.236** -0.109 -0.108 0.031 0.686** 2.620** -0.029 -0.212* -0.430** 0.016 -0.264* 1927 2 -0.386** 0.103 -0.062 0.132 1.315** 2.031** 0.176 -0.173 -0.314** 0.021 -0.261* 1927 0 On 0.042 0.098** 0.125** 0.147** 0.127** 0.008 0.018 0.064** 0.003 0.022 0.014 1927 1 0.085** 0.121** 0.107** 0.123** 0.229** 0.030 -0.005 0.007 -0.002 0.009 0.000 1927 2 0.097** 0.173** 0.223** 0.262** 0.426** 0.092** 0.038 -0.012 -0.006 0.023 -0.007 1927 Panel C. Franking Level Zero All 0.031 0.063* 0.038 0.124** 0.167** 0.022 -0.049 -0.040 -0.020 0.103* -0.041 1874 Partly 0.119* 0.120* 0.084** 0.108** 0.261** 0.300** 0.030 0.014 -0.007 0.030 0.006 636 Fully 0.089** 0.167** 0.212** 0.216** 0.359** 0.267** 0.072** 0.055** -0.002 0.014 0.018 3271 Zero Off -0.321** -0.290** -0.295** 0.024 0.460* 0.610** -0.175 -0.308** -0.285** 0.215 -0.148 1874 Partly -0.175 -0.337* -0.314* 0.010 0.884** 5.256** 0.424* -0.446** -0.459** 0.006 -0.127 636 Fully -0.399** 0.014 -0.109 -0.077 1.020** 2.917** 0.173 -0.146 -0.366** -0.169 -0.373** 3271 Zero On 0.025 0.060* 0.043 0.124** 0.154** -0.010 -0.047 -0.039 -0.015 0.034 -0.036 1874 Partly 0.094* 0.133* 0.100** 0.106** 0.229** 0.029 0.016 0.021 0.000 0.009 0.005 636 Fully 0.100** 0.171** 0.224** 0.221** 0.328** 0.077** 0.054** 0.053** 0.005 0.011 0.024 3271

Panel D. Benchmark Effective Half-Spread Table 3 continued Type t-5 t-4 t-3 t-2 t-1 Ex-day t+1 t+2 t+3 t+4 t+5 N 0 (Low) All 0.080** 0.111** 0.143** 0.138** 0.249** 0.182** 0.023 -0.024 -0.011 0.053 -0.055* 1927 1 0.067** 0.133** 0.138** 0.172** 0.279** 0.280** 0.031 -0.010 -0.013 0.026 0.011 1927 2 (High) 0.074** 0.140** 0.143** 0.214** 0.331** 0.111** 0.031 0.095** -0.002 0.055 0.039 1927 0 Off -0.312** -0.127 -0.152 -0.019 1.009** 3.458** 0.332* -0.243** -0.332** -0.016 -0.427** 1927 1 -0.297** -0.155 -0.179 -0.002 0.768** 2.862** -0.027 -0.301** -0.346** -0.103 -0.200 1927 2 -0.440** -0.085 -0.247* -0.084 0.694** 0.960** -0.043 -0.149 -0.373** 0.043 -0.192 1927 0 On 0.070** 0.109** 0.144** 0.137** 0.214** 0.032 0.007 -0.020 -0.010 0.001 -0.044 1927 1 0.072** 0.141** 0.147** 0.179** 0.260** 0.062** 0.019 -0.010 -0.004 0.017 0.014 1927 2 0.083** 0.143** 0.164** 0.217** 0.308** 0.037 0.025 0.091** 0.009 0.036 0.039 1927 Panel E. Tax Regimes 1 All 0.039 0.083** 0.075* 0.090** 0.334** 0.044 0.031 0.026 -0.001 0.036 0.084** 1581 2 0.057 0.120 0.088 0.137** 0.311** 0.139** 0.002 0.001 -0.016 0.024 -0.001 647 3 0.042 -0.011 0.009 0.044 0.210** 0.083 0.014 -0.084 -0.113 0.143 -0.018 328 4 0.096** 0.166** 0.198** 0.235** 0.265** 0.285** 0.033 0.031 0.000 0.043 -0.042* 3225 1 Off -0.142 -0.025 -0.028 0.111 1.268** 0.379** 0.091 0.042 -0.230** 0.244 0.078 1581 2 -0.238 0.132 -0.267 0.190 0.968** 2.558** 0.251 -0.167 -0.181 0.230 -0.138 647 3 -0.194 -0.276 0.221 -0.001 0.464 2.507** 0.335 -0.617** -0.053 0.833 0.259 328 4 -0.487** -0.206 -0.299** -0.153 0.613** 3.396** 0.030 -0.336** -0.472** -0.294** -0.524** 3225 1 On 0.049 0.087** 0.081* 0.093** 0.274** 0.025 0.028 0.017 0.008 0.033 0.085** 1581 2 0.059 0.118 0.104* 0.141** 0.299** -0.030 -0.005 0.003 -0.013 0.020 0.009 647 3 0.039 0.009 0.015 0.064 0.216** -0.090* -0.005 -0.065 -0.102 0.101 -0.006 328 4 0.094** 0.167** 0.209** 0.236** 0.251** 0.081** 0.018 0.033 0.006 0.002 -0.038 3225

Table 4 Realized Half-Spreads around the Ex-Dividend Day

This table reports the equally weighted realized spreads in cents across dividend events. ‘Benchmark’ measures the average realized spread between t-50 and t-6 days from the ex-dividend day. The realized half-spreads are in cents. The sample period is from Feb 1990 to Dec 2008. t-stats with ex-dividend date clustered standard errors are reported for the paired differences ‘Ex- Benchmark’ and ‘Ex-Cum’. **, * denote significance at the 1 and 5% level.

Panel A. Full Sample Ex- Benchmark Cum Ex Benchmark T Ex-Cum T N 0.341 0.369 0.376 0.034** (2.85) 0.006 (0.43) 5781 Panel B. Dividend Yield 0(Low) 0.445 0.487 0.480 0.034 (0.91) -0.008 (-0.19) 1927 1 0.295 0.326 0.321 0.026* (2.56) -0.005 (-0.34) 1927 2(High) 0.283 0.295 0.326 0.043** (2.77) 0.032 (1.89) 1927 Panel C. Franking Level Zero 0.202 0.244 0.232 0.030 (0.86) -0.012 (-0.34) 1874 Partly 0.445 0.441 0.439 -0.007 (-0.23) -0.002 (-0.07) 636 Fully 0.401 0.427 0.446 0.045** (3.48) 0.018 (1.08) 3271 Panel D. Benchmark Effective Half-Spread 0(Low) 0.081 0.137 0.130 0.049** (4.32) -0.006 (-0.45) 1927 1 0.267 0.287 0.298 0.031** (3.22) 0.011 (0.92) 1927 2(High) 0.675 0.684 0.698 0.023 (0.73) 0.014 (0.37) 1927 Panel E. Tax Regimes 1 0.191 0.212 0.212 0.021 (1.56) 0.000 (-0.01) 1581 2 0.284 0.309 0.291 0.007 (0.30) -0.018 (-0.46) 647 3 0.386 0.412 0.375 -0.011 (-0.45) -0.036 (-0.97) 328 4 0.422 0.454 0.473 0.051* (2.57) 0.019 (0.83) 3225

Table 5 Average Abnormal Time-weighted Bid and Ask Market Depth around the Ex-Dividend Day

This table reports the equally weighted average daily bid and ask market depth, between t-5 and t+5 of the ex-dividend day. Abnormal time-weighted bid (ask) market depth is calculated as the daily number of shares at the best bid (ask) price divided by the number of shares outstanding over the expected time weighted daily number of shares at the best bid (ask) price divided by the number of shares outstanding. The expected time weighted bid or ask depth for a dividend event is the average daily time-weighted number of shares at the best bid or ask price between t-45 and t-6 and t+6 to t+45. The sample period is from Feb 1990 to Dec 2008. **, * denote significance at the 1 and 5% level using t-stats with clustered standard errors on the ex-dividend date.

Panel A. Full Sample Depth t-5 t-4 t-3 t-2 t-1 Ex-day t+1 t+2 t+3 t+4 t+5 N Bid -0.071** 0.028 -0.033 0.001 0.023 -0.118** -0.052 -0.024 -0.016 0.033 -0.027 5781 Ask -0.011 0.056 0.000 0.071 0.111** -0.079* -0.023 0.000 -0.062** -0.022 -0.070** 5781 Panel B. Dividend Yield 0(Low) Bid -0.103** -0.041 0.016 -0.040 0.025 -0.105** -0.018 0.017 -0.059* 0.013 0.045 1927 1 -0.060 0.086 -0.001 0.001 0.042 -0.113* -0.028 -0.044 -0.044 0.004 -0.044 1927 2(High) -0.051 0.039 -0.114* 0.044 0.001 -0.136* -0.109* -0.045 0.054 0.083 -0.083 1927 0 Ask -0.003 -0.010 -0.016 0.054 0.017 -0.109** -0.014 0.055 -0.108** -0.021 -0.015 1927 1 -0.050 0.028 -0.012 0.078 0.062 -0.035 -0.102 -0.024 -0.016 0.013 -0.079 1927 2 0.021 0.151* 0.027 0.081 0.253** -0.092 0.048 -0.030 -0.061 -0.059 -0.116** 1927 Panel C. Franking Level Zero Bid -0.054 -0.016 -0.082 -0.020 0.004 -0.150** -0.200** -0.039 -0.103* 0.022 -0.049 1874 Partly -0.121** 0.266 0.022 0.198 -0.014 0.014 -0.099* -0.016 0.086 0.022 -0.140** 636 Fully -0.071** 0.007 -0.015 -0.024 0.041 -0.126** 0.043 -0.017 0.013 0.042 0.007 3271 Zero Ask -0.024 -0.068* -0.043 0.063 0.094 -0.081 -0.073 -0.082* -0.089* -0.090** -0.101* 1874 Partly 0.025 0.222 0.085 0.087 0.131 -0.103* -0.049 0.126 0.035 -0.027 -0.087 636 Fully -0.010 0.096 0.008 0.072 0.116** -0.073 0.011 0.023 -0.065* 0.017 -0.050 3271 Panel D. Benchmark Effective Half-Spread 0(Low) Bid -0.042 -0.011 -0.070** 0.023 0.043 -0.133* -0.097* -0.122** -0.093** 0.008 -0.080* 1927 1 -0.046 0.119 -0.046 0.006 0.053 -0.132** 0.001 -0.068 0.088 0.051 0.066 1927 2(High) -0.127** -0.025 0.017 -0.025 -0.027 -0.089 -0.059 0.120 -0.043 0.042 -0.067 1927 0(Low) Ask -0.044* 0.135* 0.035 0.122* 0.059* -0.150** -0.062 -0.052 -0.056* -0.046 -0.115** 1927 1 -0.004 0.004 0.057 -0.028 0.160* 0.063 0.038 -0.022 -0.064 -0.046 -0.100* 1927 2(High) 0.016 0.030 -0.093 0.120 0.113 -0.149** -0.044 0.076 -0.065 0.025 0.006 1927

Panel E. Tax Regimes Table 5 continued t-5 t-4 t-3 t-2 t-1 Ex-day t+1 t+2 t+3 t+4 t+5 N 1 Bid -0.200** 0.108 -0.184 -0.017 -0.105 -0.186* -0.084 -0.063 -0.087 0.049 0.030 1581 2 0.028 0.036 -0.076** -0.001 0.164 -0.042 -0.045 0.027 -0.036 -0.012 0.014 647 3 -0.040 -0.023 0.080 0.005 0.137* -0.149** -0.082* -0.129** -0.107** -0.021 -0.029 328 4 -0.033* -0.007 0.036 0.011 0.045** -0.097** -0.034 -0.005 0.030 0.041 -0.062** 3225 1 Ask -0.110 0.128 -0.173 0.067 0.101 -0.038 0.000 0.067 -0.171* -0.102 -0.134 1581 2 0.022 0.066* 0.045 0.040 0.044 -0.107** -0.060* -0.039 -0.050 -0.028 -0.004 647 3 0.020 0.003 -0.009 -0.002 0.097* -0.117** -0.104** -0.051 -0.003 0.017 0.019 328 4 0.027* 0.025 0.074** 0.087** 0.130** -0.089** -0.018 -0.018 -0.018 0.013 -0.062** 3225

Table 6 Hourly Effective Half-Spreads

This table reports the equally weighted average effective half-spread in cents measures across dividend events. Benchmark is the average effective half-spread from t-50 to t-6 days prior to an ex-dividend date for a given stock. t- stats using clustered standard errors on the ex-dividend date are reported for the paired differences ‘Ex-Benchmark’ and ‘Ex-Cum’. **, * denote significance at the 1 and 5% level.

Panel A. Entire Sample Ex- Hour of trade Benchmark Cum Ex T Ex-Cum T N Benchmark 10 to 11(first) 1.261 1.174 1.812 0.551** (11.02) 0.638** (12.40) 4930 11 to 12 1.101 0.986 1.339 0.238** (7.25) 0.353** (10.05) 4933 12 to 1 1.028 0.933 1.115 0.087** (3.05) 0.182** (6.44) 4641 1 to 2 0.955 0.919 1.005 0.050* (2.08) 0.086 (1.50) 4096 2 to 3 1.026 0.961 1.134 0.108** (3.08) 0.172** (4.64) 4884 3 to 4(last) 1.157 1.111 1.259 0.102** (4.43) 0.148** (4.84) 5222 Last - first -0.209 -0.173 -0.633 -0.424** (-11.78) -0.460** (-11.13) 4663 Panel B. Dividend Yield 0(Low) 10 to 11(first) 1.621 1.529 2.145 0.524** (7.71) 0.615** (8.40) 1637 11 to 12 1.355 1.266 1.669 0.314** (3.81) 0.403** (4.81) 1633 12 to 1 1.287 1.188 1.401 0.114 (1.67) 0.213** (3.43) 1543 1 to 2 1.191 1.270 1.278 0.087 (1.45) 0.008 (0.05) 1390 2 to 3 1.279 1.239 1.426 0.147 (1.63) 0.187 (1.90) 1617 3 to 4(last) 1.429 1.455 1.601 0.172** (3.44) 0.146* (2.13) 1718 Last - first -0.335 -0.283 -0.738 -0.403** (-5.98) -0.455** (-5.40) 1543 1 10 to 11(first) 1.131 1.060 1.777 0.645** (5.40) 0.717** (6.04) 1641 11 to 12 1.036 0.899 1.226 0.190** (4.93) 0.327** (6.97) 1646 12 to 1 0.945 0.834 1.038 0.093* (2.03) 0.204** (4.05) 1531 1 to 2 0.851 0.760 0.875 0.024 (0.99) 0.114** (3.98) 1316 2 to 3 0.958 0.885 1.074 0.115* (2.51) 0.189** (4.15) 1623 3 to 4(last) 1.068 0.976 1.120 0.052 (1.34) 0.144** (3.01) 1746 Last - first -0.161 -0.132 -0.606 -0.445** (-7.20) -0.475** (-6.69) 1557 2(High) 10 to 11(first) 1.033 0.935 1.518 0.485** (9.90) 0.583** (11.07) 1652 11 to 12 0.915 0.797 1.126 0.210** (5.68) 0.329** (8.10) 1654 12 to 1 0.855 0.777 0.909 0.054* (2.53) 0.132** (5.04) 1567 1 to 2 0.818 0.718 0.855 0.037 (1.42) 0.137** (5.12) 1390 2 to 3 0.843 0.763 0.905 0.062* (2.50) 0.142** (5.17) 1644 3 to 4(last) 0.981 0.910 1.064 0.083* (2.47) 0.154** (4.60) 1758 Last - first -0.131 -0.107 -0.556 -0.425** (-9.09) -0.450** (-9.47) 1563

Panel C. Franking Level Table 6 continued Ex- Hour of trade Benchmark Cum Ex T Ex-Cum T N Benchmark Zero 10 to 11(first) 0.966 0.957 1.492 0.526** (4.23) 0.535** (4.36) 1526 11 to 12 0.872 0.794 1.074 0.202** (3.86) 0.280** (4.47) 1500 12 to 1 0.860 0.792 0.904 0.044 (0.79) 0.112* (2.46) 1391 1 to 2 0.773 0.879 0.821 0.048 (1.13) -0.058 (-0.32) 1167 2 to 3 0.847 0.799 0.885 0.037 (0.88) 0.086* (2.27) 1473 3 to 4(last) 0.970 0.989 1.033 0.063 (1.84) 0.044 (1.02) 1637 Last - first -0.111 -0.086 -0.424 -0.313** (-6.52) -0.338** (-5.62) 1412 Partly 10 to 11(first) 1.270 1.164 1.694 0.424** (3.88) 0.530** (4.83) 592 11 to 12 1.041 0.931 1.143 0.102 (1.46) 0.212** (2.74) 589 12 to 1 0.943 0.889 0.963 0.020 (0.49) 0.075 (1.75) 571 1 to 2 0.898 0.846 0.901 0.003 (0.08) 0.055 (1.23) 535 2 to 3 0.927 0.902 1.023 0.096* (2.03) 0.121 (1.89) 589 3 to 4(last) 1.015 0.980 1.205 0.189* (2.50) 0.224** (2.83) 604 Last - first -0.288 -0.247 -0.550 -0.262* (-2.51) -0.303** (-2.82) 575 Fully 10 to 11(first) 1.419 1.294 2.011 0.592** (12.19) 0.717** (13.64) 2812 11 to 12 1.235 1.099 1.520 0.285** (6.11) 0.421** (8.85) 2844 12 to 1 1.134 1.015 1.257 0.123** (3.20) 0.242** (5.81) 2679 1 to 2 1.057 0.955 1.118 0.061 (1.79) 0.163** (4.33) 2394 2 to 3 1.139 1.058 1.287 0.147** (2.65) 0.229** (3.81) 2822 3 to 4(last) 1.289 1.205 1.395 0.105** (3.25) 0.189** (4.26) 2981 Last - first -0.243 -0.203 -0.762 -0.518** (-10.35) -0.558** (-9.60) 2676 Panel D. Benchmark Effective Half-Spread 0(Low) 10 to 11(first) 0.559 0.555 0.685 0.126** (11.46) 0.130** (12.26) 1862 11 to 12 0.521 0.519 0.555 0.034** (6.39) 0.037** (6.28) 1859 12 to 1 0.516 0.515 0.532 0.016** (4.61) 0.017** (4.22) 1819 1 to 2 0.514 0.511 0.528 0.014** (3.43) 0.017** (3.47) 1697 2 to 3 0.511 0.506 0.524 0.013** (3.16) 0.018** (4.29) 1833 3 to 4(last) 0.522 0.527 0.539 0.017** (4.41) 0.012** (2.62) 1871 Last - first -0.039 -0.030 -0.151 -0.111** (-10.31) -0.120** (-10.81) 1813 1 10 to 11(first) 0.959 0.911 1.321 0.361** (13.67) 0.410** (14.82) 1729 11 to 12 0.815 0.776 0.940 0.125** (8.34) 0.164** (9.81) 1724 12 to 1 0.789 0.761 0.849 0.060** (4.70) 0.088** (5.77) 1636 1 to 2 0.780 0.749 0.849 0.069** (3.78) 0.100** (4.86) 1464 2 to 3 0.757 0.736 0.799 0.042** (4.02) 0.063** (5.02) 1715 3 to 4(last) 0.805 0.785 0.859 0.053** (4.28) 0.073** (4.65) 1819 Last - first -0.161 -0.138 -0.465 -0.304** (-11.24) -0.327** (-10.48) 1662 2(High) 10 to 11(first) 2.627 2.375 4.015 1.388** (7.88) 1.640** (9.19) 1339 11 to 12 2.266 1.898 2.928 0.662** (5.70) 1.030** (8.42) 1350 12 to 1 2.145 1.810 2.376 0.231* (2.12) 0.566** (5.29) 1186 1 to 2 2.031 1.926 2.115 0.084 (0.84) 0.189 (0.76) 935 2 to 3 2.077 1.876 2.400 0.323* (2.56) 0.524** (3.94) 1336 3 to 4(last) 2.352 2.213 2.615 0.263** (3.47) 0.402** (4.02) 1532 Last - first -0.535 -0.441 -1.605 -1.070** (-8.13) -1.164** (-7.70) 1188

Panel E. Tax Regimes Table 6 continued Ex- Hour of trade Benchmark Cum Ex T Ex-Cum T N Benchmark 1 10 to 11(first) 1.300 1.213 1.858 0.558** (9.25) 0.645** (9.99) 1125 11 to 12 1.194 1.068 1.496 0.302** (4.60) 0.429** (6.00) 1126 12 to 1 1.068 1.018 1.225 0.157** (3.40) 0.207** (3.88) 975 1 to 2 0.950 0.860 0.972 0.022 (0.60) 0.112* (2.57) 660 2 to 3 1.123 1.076 1.323 0.200* (2.39) 0.246** (2.76) 1100 3 to 4(last) 1.316 1.256 1.475 0.159** (2.98) 0.219** (3.20) 1267 Last - first -0.082 -0.076 -0.504 -0.423** (-5.83) -0.428** (-5.04) 995 2 10 to 11(first) 1.507 1.365 2.213 0.706** (4.30) 0.848** (5.09) 511 11 to 12 1.397 1.241 1.717 0.319** (3.39) 0.476** (5.17) 513 12 to 1 1.248 1.120 1.231 -0.016 (-0.30) 0.111 (1.55) 459 1 to 2 1.114 1.010 1.141 0.027 (0.40) 0.130* (2.01) 359 2 to 3 1.290 1.136 1.346 0.056 (0.55) 0.210* (2.12) 519 3 to 4(last) 1.417 1.394 1.631 0.214* (2.47) 0.237* (2.50) 584 Last - first -0.265 -0.291 -0.885 -0.621** (-3.86) -0.595** (-3.61) 481 3 10 to 11(first) 1.712 1.364 2.362 0.651** (5.83) 0.998** (7.49) 268 11 to 12 1.412 1.260 1.842 0.430** (3.17) 0.582** (4.50) 269 12 to 1 1.315 1.131 1.659 0.343 (1.89) 0.528** (2.64) 260 1 to 2 1.198 1.042 1.556 0.358 (1.68) 0.514* (2.21) 224 2 to 3 1.174 1.090 1.921 0.747* (1.99) 0.831* (2.25) 273 3 to 4(last) 1.434 1.601 1.807 0.373** (2.68) 0.206 (0.85) 306 Last - first -0.399 0.022 -0.846 -0.446** (-3.33) -0.868** (-5.35) 255 4 10 to 11(first) 1.165 1.111 1.679 0.514** (7.10) 0.568** (7.73) 3026 11 to 12 0.989 0.889 1.172 0.183** (4.28) 0.283** (6.14) 3025 12 to 1 0.956 0.858 1.013 0.057 (1.50) 0.155** (4.44) 2947 1 to 2 0.918 0.912 0.952 0.035 (1.28) 0.041 (0.51) 2853 2 to 3 0.931 0.877 0.956 0.025 (0.91) 0.079* (2.41) 2992 3 to 4(last) 1.015 0.949 1.045 0.030 (1.29) 0.096** (3.07) 3065 Last - first -0.226 -0.204 -0.617 -0.391** (-9.15) -0.413** (-8.23) 2932

Table 7 Adverse Selection Component of the Spread

This table presents the adverse selection component of the spread following a spread decomposition using the method of Lin, Sanger and Booth (1995). The percentage of the spread due to adverse selection is equally weighted across dividend events. ‘Benchmark’ measures the average adverse selection component between t-50 and t-6 days from the ex-dividend day. The sample period is from Feb 1990 to Dec 2008. t-stats with ex-dividend date clustered standard errors are reported for the paired differences ‘Ex-Benchmark’ and ‘Ex-Cum’. **, * denote significance at the 1 and 5% level.

Panel A. Full Sample Ex- Benchmark Cum Ex Benchmark T Ex-Cum T N 0.489 0.493 0.527 0.038** (8.89) 0.034** (6.49) 3949 Panel B. Dividend Yield 0(Low) 0.538 0.545 0.569 0.032** (5.17) 0.024** (3.08) 1355 1 0.467 0.472 0.505 0.038** (5.64) 0.033** (3.89) 1256 2(High) 0.459 0.459 0.503 0.044** (5.91) 0.045** (4.67) 1338 Panel C. Franking Level Zero 0.412 0.424 0.473 0.062** (6.34) 0.049** (4.10) 1078 Partly 0.510 0.508 0.537 0.027** (3.01) 0.028* (2.56) 539 Fully 0.519 0.520 0.549 0.030** (6.18) 0.028** (4.76) 2332 Panel D. Benchmark Effective Half-Spread 0(Low) 0.376 0.375 0.423 0.048** (7.01) 0.048** (6.22) 1690 1 0.555 0.559 0.581 0.026** (4.22) 0.022** (2.77) 1428 2(High) 0.604 0.617 0.644 0.040** (4.23) 0.027* (2.23) 831 Panel E. Tax Regimes 1 0.551 0.554 0.596 0.045** (4.77) 0.042** (3.38) 662 2 0.585 0.599 0.620 0.035** (2.85) 0.021 (1.38) 323 3 0.536 0.539 0.579 0.043* (2.41) 0.040 (1.94) 191 4 0.459 0.462 0.496 0.036** (6.78) 0.033** (5.14) 2773

Table 8 Determinants of the Effective Spread

This table reports results from the estimation of the following regression: ES = α +α DEx +α Volume +α Volume DEx +α σ +α σ DEx +α Tick +α Tick DEx + i,t 0 1 i,t 2 i,t 3 i,t i,t 4 i,t 5 i,t i,t 6 i,t 7 i,t i,t Ex Ex α8Franki,t +α9Franki,t Di,t +α10DYi,t +α11DYi,t Di,t + εi,t where ES is the effective spread, DEx is a dummy variable equal to one for ex-dividend days and zero for cum-dividend days, Volume is the natural logarithm of average daily trading volume between t-50 and t-6, relative to the ex-dividend day, σ is the stock return variance over the same time interval, Tick is the minimum tick size for each stock based on its ex-dividend price, Frank is the percentage to which the dividend carries tax credits, and DY is the dividend as a percent of the cum-dividend day closing price. The ex-day dummy variable is interacted with explanatory variables to assess whether there exert different influences on the spread once a stock beings trading ex-dividend. Regression in column 3 and 4 include fixed effects. We report the absolute value of t-statistics based on standard errors clustered by stock and ex-dividend date in parentheses . ***, **, * denote significance at the 1, 5 and 10% level.

(1) (2) (3) (4) Constant 9.917*** 10.888*** 4.925*** 5.244*** (5.10) (5.14) (5.69) (5.90) ExDum 3.594*** 3.330*** 3.321*** 3.052*** (3.86) (3.31) (3.68) (3.09) Volume -0.782*** -0.793*** -0.394*** -0.400*** (5.56) (5.59) (6.73) (6.82) Volume * ExDum -0.246*** -0.242*** -0.244*** -0.239*** (3.87) (3.76) (3.76) (3.66) Volatility 17.725** 18.255** 13.222*** 14.477*** (2.33) (2.34) (2.74) (2.96) Volatility * ExDum -3.061 -3.589 -2.897 -3.360 (0.65) (0.73) (0.60) (0.67) Tick 0.883*** 0.889*** 0.538 0.540 (2.83) (3.27) (1.62) (1.63) Tick * ExDum -0.107 -0.158 0.134 0.092 (0.33) (0.49) (0.54) (0.37) Franking -0.437* -0.066 (1.71) (0.74) Franking * ExDum 0.204** 0.191* (1.98) (1.83) DivYield -25.617*** -12.826*** (3.98) (4.27) DivYield * ExDum 6.244* 6.500* (1.77) (1.82) Stock fixed effects Yes Yes Obs. 11562 11562 11562 11562 Adj. R2 0.215 0.225 0.549 0.550

Table 9 Determinants of Effective Spreads: Dividend Yield and Franking Partitions

Ex Ex Ex Ex This table reports results from the estimation of the following regression: ESi,t = α0 + α1Di,t + α2Volumei,t + α3Volumei,t Di,t + α4σ i,t + α5σ i,t Di,t + α6Ticki,t + α7Ticki,t Di,t + εi,t . ES is the effective spread, DEx is a dummy variable equal to one for ex-dividend days and zero for cum-dividend days, Volume is the natural logarithm of average daily trading volume between t-50 and t-6, relative to the ex-dividend day, σ is the stock return variance over the same time interval, and Tick is the minimum tick size for each stock based on its ex- dividend price. The ex-day dummy variable is interacted with explanatory variables to assess whether there exert different influences on the spread once a stock beings trading ex- dividend. Regressions are estimated for nine groups based on the intersection of their dividend yield tercile and franking percentage. We report the absolute value of t-statistics based on standard errors clustered by stock and ex-dividend date in parentheses . ***, **, * denote significance at the 1, 5 and 10% level.

Unfranked Partially Franked Fully Franked Low DY Mid DY High DY Low DY Mid DY High DY Low DY Mid DY High DY (1) (2) (3) (4) (5) (6) (7) (8) (9) Constant 28.874*** 15.606** 4.456*** 6.478*** 5.464*** 5.970*** 9.001*** 5.243*** 3.376*** (3.89) (2.49) (4.83) (2.67) (3.11) (3.58) (7.31) (6.43) (4.78) ExDum -1.814 4.803 1.964*** 4.917 3.636** 5.489** 5.364*** 4.693*** 4.805*** (0.44) (1.59) (2.76) (0.90) (2.37) (2.50) (3.50) (2.82) (5.43) Volume -2.023*** -1.024** -0.319*** -0.527*** -0.390*** -0.456*** -0.750*** -0.417*** -0.308*** (4.20) (2.54) (4.82) (3.91) (3.47) (3.94) (6.88) (7.25) (7.78) Volume * ExDum 0.077 -0.330* -0.128** -0.477 -0.237** -0.366** -0.350*** -0.350*** -0.296*** (0.30) (1.69) (2.50) (1.64) (2.39) (2.31) (4.06) (3.71) (5.98) Volatility 55.898** -17.472 11.936 10.273 9.672 25.263*** -0.295 2.155 11.328*** (1.99) (0.98) (1.58) (0.96) (1.32) (3.52) (0.04) (0.53) (2.71) Volatility * ExDum 2.369 2.790 -4.994 -27.005 -3.596 -2.719 -9.560 -6.106 5.088 (0.10) (0.21) (1.33) (0.94) (0.53) (0.23) (1.01) (0.64) (0.83) Tick -2.254 -1.579 0.272* 1.496*** 0.468* 0.556** 1.878*** 1.057*** 1.275** (1.46) (1.42) (1.89) (2.70) (1.85) (2.29) (2.74) (4.66) (2.19) Tick * ExDum 0.947 -0.465 -0.083 2.223 -0.236 -0.323 -0.390 0.189 -0.823 (0.93) (1.04) (0.94) (1.30) (1.16) (1.59) (0.43) (0.21) (1.34) Obs. 1108 1346 1294 440 430 402 2306 2078 2158 Adj. R2 0.341 0.246 0.273 0.202 0.338 0.304 0.258 0.208 0.228

Figure 1 Average Daily Effective Spread around the Ex-Dividend Day

The sample is for dividends paid between February 1990 and December 2008. This figure presents the equally weighted daily average effective half-spread in cents across dividend events from t-45 to t+45 relative to the ex-dividend date.