The effect of stock splits on liquidity - Evidence from the Dutch stock market

The effect of stock splits on liquidity

-

Evidence from the Dutch stock market

Boudewijn Anton Jan Leutscher1 Abstract

This paper studies the effects of stock splits on liquidity. Four different measures are used as a proxy for liquidity: the turnover ratio, the Amihud ratio, Zeros and the quoted bid-ask spread. The liquidity measures improve significantly around the announcement of a split. Moreover, after the announcement of a split most liquidity measures also remain above pre-split level up to one year after the execution of a split, suggesting the effect of a stock split on liquidity is long-lived. Furthermore, evidence is found that the change in Amihud ratio can explain the announcement effect; however it cannot be explained by the changes in the turnover ratio and Zeros. Overall, the results are mildly supportive for both the signalling hypothesis and the attention-grabbing hypothesis, as well as for the trading range/improved liquidity hypothesis.

Key words: Stock splits, Liquidity, Signalling, Attention-grabbing, Event study JEL classification: G14, G30

This thesis is submitted as a master’s thesis for the Master of Science in Finance (EBM866B20). Supervisor: MSc N. Heida, University of Groningen. I would like to thank MSc N. Heida for her useful feedback.

2

1. Introduction

"I don't want anybody buying Berkshire thinking they can make a lot of money fast. They're not going to do it, in the first place. And some of them will blame themselves, and some of them will blame me. They'll all be disappointed. I don't want disappointed people. The idea of giving people crazy expectations has terrified me from the moment I started selling stocks" - Warren Buffet

In January 2016, a Berkshire Hathaway A stock cost around $190,000 and was the highest priced share in the world. The CEO of Berkshire, Warren Buffet, is known for his investment talent and despite the extreme price of the stock he has always refused to split it. Keeping the price high is the way in which Buffet tries to withhold short-term investors from investing in the stock and creating excessive volatility. Even though Warren Buffet is not a proponent of stock splits, the management of other firms are, making stock splits still common around the world.

A stock split is a corporate event in which a publicly traded firm increases its number of shares outstanding, by dividing its existing shares into multiple shares. In this way, the proportion of equity owned by a shareholder remains the same. To express it differently: the size of the “pie” does not change, it will only be divided into smaller pieces (Schultz, 2000). Various big US firms such as Nike, Starbucks, Visa and Netflix thought 2015 was a good year to announce a stock split. The announced stock splits for these enterprises were in the range from two-for-one (Nike and Starbucks), four-for-one (Visa) and seven-for-four-for-one (Netflix). The announcement of a stock split by these four US firms was not a complete surprise, as they had relatively high and increasing stock prices over a long time horizon in common. Due to an elevation in stock price the probability a firm will split its stock increases, as a stock split is usually announced after an increase in stock price (Fama et al., 1969).

Assuming efficient markets (markets that adjust rapidly to new information), Grinblatt, Masulis, and Titman (1984) state that a stock split is just a cosmetic change which increases the number of stocks with a certain ratio and lowers the price of a stock with the same ratio. In other words, a stock split should not change the fundamental value of the firm itself. However, if markets are indeed efficient, the question arises why firms still announce stock splits when these stock splits do not directly influence the value of the firm and while costs are involved in undertaking a split.

3 other hand suggests that firms split their stock because such an event will grab the attention of investors. The limited attention of investors makes them prone to buy stocks that grab their attention. The trading range/improved liquidity hypothesis argues that stock splits are executed because firms prefer a certain price range for their stock (Copeland, 1979). If prices are high, lower stock prices will make stocks more affordable, especially to individual investors. All three hypotheses imply an improvement in liquidity for the splitting stocks, although these hypotheses have a different view on when this improvement in liquidity will arise and whether this improvement is a short- or long-term phenomenon. However, empirical research on the effect of stock splits on liquidity gives mixed results on the question whether stock splits improve liquidity.

Multiple studies suggest stock splits improve the liquidity of trading (Muscarella and Vetsuypens, 1996; Wulff, 2002; Huang, Liano, and Pan, 2015). Nevertheless, definite conclusions on the effect of stock splits on liquidity cannot be drawn, as other research suggests the existence of a negative relationship between stock splits and stock liquidity. Studies by Copeland (1979), Conroy, Harris, and Benet (1990) and Michayluk and Kofman (2001) find a decrease in liquidity after a split, measuring an increase in post-split bid-ask spreads. These discrepancies between the findings can be explained by the heterogeneity of the studied markets and periods and the different research approaches and variables that have been used.

Further research is needed to gather more evidence on the effect of stock splits. This paper therefore evaluates the effect of stock splits on liquidity and additionally provides an answer whether this effect can be explained by the signalling hypothesis, the attention-grabbing hypothesis or the trading range/improved liquidity hypothesis. Furthermore, the relationship between stock splits and abnormal returns is investigated. Finding shifts in liquidity around the announcement date or ex-split date of a stock split can be used to evaluate the explanations of abnormal returns around these dates. Besides, this paper contributes to existing research by focusing on the Dutch market, while most of the previous studies on stock splits focused on US markets.

4 returns in the period around the announcement of a split. The change in the Amihud ratio can explain the announcement effect, but the changes in the turnover ratio and Zeros cannot explain this effect. Overall, the results do not provide a conclusive answer why firms decide to split their stock. Some of the results on liquidity are in favour of the signalling hypothesis and attention-grabbing hypothesis, while other results suggest the trading range/improved liquidity hypothesis is better able to explain the decision to split a stock.

The remainder of this paper is organized as follows. In section 2, a literature overview is given. Section 3 describes the methodology used in this paper in more detail. Section 4 describes the data used and in section 5 the empirical results are presented. Conclusions and limitations of this research and suggestions for future research are given in section 6.

2. Literature review

In the following paragraphs, the theoretical background regarding stock splits is described. In section 2.1 the signalling hypothesis, the attention-grabbing hypothesis and the trading range/improved liquidity hypothesis are discussed to find an answer on the question why firms are splitting their stocks. In section 2.2 an overview is given of the effects of stock splits on liquidity and in section 2.3 abnormal returns due to a stock split and the relation with liquidity are discussed. In the final section, 2.4, hypotheses are developed.

2.1 Why firms are splitting their stocks

Earlier literature puts forth different hypotheses to explain why firms announce a stock split and these hypotheses expect different market reactions around the announcement date and execution date of a split. Of all these hypotheses, the following three received the most attention in the literature: the signalling hypothesis, the attention-grabbing hypothesis and the trading range/improved liquidity hypothesis. These three hypotheses all imply an increase in liquidity and are not mutually exclusive. However, they have different views on when this increase in liquidity will arise and whether this increase in liquidity is a short- or long-term phenomenon.

2.1.1 Signalling Hypothesis

5 enables managers to convey private information to the shareholders, reducing the existing information asymmetry between the two parties. In their paper, Grinblatt et al. (1984) argue that declaring a stock split can be seen as such an informative signal from managers to shareholders about the good future outlook of the firm. The signalling hypothesis therefore predicts that a stock split is mostly interpreted as a positive signal by investors, leading to an increase in investors’ demand and thus an increase in the stock price (Grinblatt et al., 1984). However, a signal will only be reliable when it is in the best interest of the managers to produce a truthful signal and producing a false signal will not be beneficial. Therefore, to prevent managers from sending out false signals, costs need to be associated with sending a signal (Crawford and Franz, 2001). Costs associated with a split may withhold firms without good future prospects to copy the signalling decision of firms with good future prospects (Lakonishok and Lev, 1987). Doran (1995), Brennan and Copeland (1988) and Yagüe, Gómez-Sala, and Poveda-Fuentes (2009) suggest stock splits are a credible signal, since the costs involved with the execution of a split exceed the benefits of sending out a false signal.

6 In short, the signalling hypothesis predicts an increase in demand for a stock in response to the positive signal of the stock split announcement. However, in an efficient market new information is rapidly incorporated in the price of a stock. It is therefore expected that the increase in liquidity will not last long after the announcement of a stock split.

2.1.2 Attention-grabbing Hypothesis

Barber and Odean (2008) state that attention is a scarce resource and investors are overloaded with choices. Their paper provides evidence that individual investors are net buyers of stocks that draw their attention. The announcement of a stock split is an example of an event that may attract the attention of investors, through which they become more likely to buy stocks of the splitting firm. Thus according to Barber and Odean (2008), more attention might lead to increased trading liquidity. This could especially be beneficial for small firms, as they are less often mentioned in financial media. Therefore, a stock split announcement may increase investors’ attention for small firms more than for large firms (Grinblatt et al., 1984). Similarly, Barber and Odean (2008) argue that small investors or individuals are more sensitive to the attention-grabbing news than large investors or institutions. As individuals have limited time and fewer resources to search for stocks, news announcements may have greater impact on their buying decision (Barber and Odean, 2008). For individual investors, Barber and Odean (2008) present evidence of attention driven buying, while they find no evidence for attention driven buying by institutional investors. Grinblatt el al. (1984) argue that attention for the firm possibly reduces the existing information asymmetry, as market analysts may reassess the value of these firms. Underpriced firms are interested in a revision of the value of the firm, while overpriced firms do not. Therefore, it is hypothesized that abnormal returns are present on the announcement day, reflecting the average underprizing of splitting firms. A weakness of the attention-grabbing hypothesis, mentioned by Grinblatt et al. (1984), is that it does not explain the beneficial effect of announcing stock splits above straightforward financial publications.

7 2.1.3 Trading range/Improved liquidity Hypothesis

The trading range/improved liquidity hypothesis holds that splitting a stock enables a firm to return its stock price to an optimal trading range and thus enhance the liquidity of the stock (Copeland, 1979). Especially small investors suffer from high stock prices, because such prices make it more difficult for them to buy or sell stocks in round lots (Yagüe et al., 2009).

Baker and Gallagher (1980) surveyed chief financial officers of firms listed on the New York Stock Exchange (NYSE) to investigate why firms split their stock. Their results show that the majority of the managers prefer to split a stock to bring the stock into a lower, more optimal, trading range. A lower stock price could result in a stock that comes within reach of more investors and result in a broader ownership base. Furthermore, respondents of the survey mention an improvement in trading liquidity as a reason to split a stock.

The motives for stock splits put forth in the survey by Baker and Gallagher (1980), are supported by empirical studies. Lakonishok and Lev (1987) studied the motivation for stock splits in a sample of American stock splits (n = 1015) between 1963 and 1982 and conclude that the main reason to announce a stock split is to restore the stock price to a normal trading range. This normal trading range is based on market and industry average stock prices as well as firm specific stock prices. Ikenberry, Rankine, and Stice (1996) support this view, providing evidence that stock splits generally occur when stock prices are high. Managements’ preference to restore the stock price to a lower trading range may stem from the wish to attract a specific group of investors or to achieve a certain dispersion in ownership (Easley, O’hara, and Saar, 2001). Lamoureux and Poon (1987) state that managers may prefer a diffused ownership base to protect themselves from takeover threats, arguing that a wider ownership base reduces the probability of a tender offer being accepted. Schultz (2000), and Easley et al. (2001) find support for this reasoning, showing an enlarged ownership base after a split. Dhar, Goetzmann, and Zhu (2004) and Kadapakkam, Krishnamurthy, and Tse (2005) find support for an increase in trading by small investors after a split. Small investors may be preferred by managers as they are on average less willing to interfere with the decisions managers make and to exercise control. Lakonishok and Lev (1987) and Dyl and Elliott (2006) find an increase in the number of shareholders after a stock split, arguing that firms may be splitting their stocks to increase the marketability of the stock.

8

2.2 Liquidity

The above paragraph discusses three main hypotheses to explain why firms are splitting stocks. All three hypotheses imply an improvement in liquidity for the split stocks. However, empirical results on the effect of stock splits on liquidity are somewhat ambiguous.

2.2.1 Liquidity effects of the announcement of a split

Not much research is done on the direct effects of stock splits on liquidity around the announcement date. Huang et al. (2015) study the effect of stock splits on liquidity in the US market between 1960 and 2010. They use five different liquidity measures: the turnover ratio, Amihud ratio, Zeros, Dollar Spread and the Relative Spread. Using these measures as a proxy for liquidity, they find a significant increase in liquidity for splitting firms in the five-day period around the announcement date. Moreover, they find that liquidity monotonically declines after the announcement date to the ex-split date, although it remains above the pre-split level. Joshipura (2008) studies the Indian Stock Market and also finds a significant improvement in liquidity following the announcement of a stock split.

2.2.2 Liquidity effects of the execution of a split

More research has been done on the effect of the execution of a split on liquidity, although with mixed results. Wulff (2002) evaluates the market reaction to stock splits for the German market and finds the percentage of days with trades and trading turnover significantly increase after the ex-split date, suggesting an improvement in liquidity. Muscarella and Vetsuypens (1996) study the effect of stock splits on liquidity in the US market between 1962 and 1993. They find that after a stock split, both the number of stocks traded as well as the total volume significantly increases. This is supported by Schultz (2000), who finds an increase in frequency and volume of small trades due to a stock split.

9 2.2.3 Abnormal returns and changes in liquidity

Next to liquidity, this paper studies abnormal returns because it may be possible that the change in liquidity can explain abnormal returns around the announcement date or execution date of a split. Empirical research finds significant abnormal returns around the announcement date. Investigating the US market, Fama et al. (1969) find abnormal returns around the announcement date using a sample of 940 stock split events. In consonance, Grinblatt et al. (1984) and Maloney and Mulherin (1992) also find significant cumulative abnormal return for US listed firms around the announcement of a split. Buijs and Gerritsen (2005) are one of the few investigating stock splits in the Dutch market. Their research focuses on stock splits in the period between 1983 and 2003. They find significant abnormal returns in the two days surrounding the announcement of a split, however there seems to be no effect of a stock split on the stock price in the two days surrounding the execution of a split. The existence of abnormal returns around the execution of a stock split is also studied for other markets. Grinblatt et al. (1984) for example do not only find abnormal returns around the announcement date, but they also find that post-announcement returns are often abnormally large around the ex-split date. In accordance, Wulff (2002) and Huang et al. (2015) also find positive and significant abnormal returns in the period around the execution of a split. Furthermore Huang et al. (2015) find a negative abnormal return one year after the execution of a split.

The existence of abnormal returns suggests increased buying pressure from investors. Some researchers argue that this increased buying pressure results from increased liquidity, although the empirical results on the relationship between abnormal returns and changes in liquidity are not unambiguous.

Wulff (2002) investigates the relationship between abnormal returns and liquidity. He evaluates the market reaction to stock splits for German firms and shows that changes in liquidity around the announcement date are not associated with abnormal returns in this period. More research is done by Huang et al. (2015) who examine whether changes in liquidity can explain the abnormal returns around the announcement date and the execution date. In contrast to Wulff (2002), they find that the change in liquidity can explain the split announcement effect while no relation is found between liquidity and abnormal returns in the period surrounding the execution of a split.

2.3 Hypotheses

10 Based on the three main hypotheses discussed and the expected effect of stock splits on liquidity, the following hypotheses are formed:

H1: The signalling hypothesis expects a significant increase in trading liquidity for a splitting stock around the announcement date, but not in any of the periods thereafter.

H2: The attention-grabbing hypothesis expects a significant increase in trading liquidity for a splitting stock around the announcement date, but not in any of the periods thereafter.

H3: The trading range/improved liquidity hypothesis expects a significant increase in trading liquidity for a splitting stock after the ex-split date and expects this effect is long lasting.

All three hypotheses imply an improvement in liquidity for the splitting stocks, although these hypotheses have a different view on the moment this improvement in liquidity will arise and whether this improvement is a short- or long-term phenomenon.

Therefore, this paper focuses on the stock liquidity effects of stock splits in the Dutch market and whether these liquidity effects can be explained by any of the three main hypotheses. Based on the findings by Huang et al. (2015) and in line with the signalling hypothesis and the attention-grabbing hypothesis, I expect to find a substantial increase in liquidity for the splitting stocks around the stock split announcement date, for firms listed on the Amsterdam Stock Exchange. Furthermore, I expect that after the increase in liquidity around the announcement date, liquidity will drop to the pre-split level in the long run. Moreover, I expect the improvement in liquidity does not last long.

This paper also investigates the relationship between stock splits and abnormal returns. In line with both the signalling hypothesis and the attention-grabbing hypothesis it is foreseen that abnormal returns are present around the announcement date but not at the execution date. Based on research by Huang et al. (2015), I expect that a positive association between liquidity and abnormal returns in the announcement period is present.

3. Methodology

11 The following sections elaborate on the methodology employed in this paper. First in section 3.1 the research periods are set out. Second, in section 3.2 the definition of liquidity is discussed and the liquidity measures are presented. Third, section 3.3 describes how changes in liquidity are investigated. Fourth, section 3.4 describes the methodology used to examine the abnormal returns. Finally, section 3.5 presents the regression analysis performed.

3.1 Research periods

Just like Huang et al. (2015) this paper investigates changes in stock liquidity in different periods around the split announcement date (AD) and ex-date (ED). Specifically, six periods over a two year period surrounding the split announcement date and the ex-date are examined (see Figure 1).

Figure 1: The research periods

This figure shows the six periods investigated to examine the effect of stock splits on liquidity.

The first period is the pre-announcement period (AD-252 to AD-3). This period starts 252 trading days before the split announcement and ends three trading days before the split announcement. The second period is the so-called announcement period (AD-2 to AD+2), which covers the trading days surrounding the announcement. MacKinlay (1997) states that including extra days around the announcement date makes it possible to examine the effect of investors acquiring information prior to the announcement. Sometimes there are rumours about a possible stock split announcement just before the official announcement of such a split, and therefore, to capture the real effect of the announcement, not only the announcement date but also two trading days before and two trading days after the announcement date are included in this period. The announcement-to-ex period (AD+3 to ED-1) is the third period investigated, and captures the period three trading days after the announcement until one trading day before the execution of the stock split. Thereafter, the ex-date period (ED0 to ED+4) captures the day of the execution of the split and the four trading days thereafter. The fifth and sixth period are the short-term ex period (ED+5 to ED+10) and the long-term post-ex period (ED+11 to ED+260) respectively. The long-term post-post-ex period is included to capture changes in liquidity in the long-term and comprises a period of a year (250 trading days), starting eleven trading days after the execution of a split.

3.2 Definition of liquidity and liquidity measures

12 paper, Amihud and Mendelson (1991) define liquidity as follows: "An asset is liquid if it can be bought or sold at the current market price quickly and at low cost". As liquidity in the stock market consists of different characteristics (Amihud, 2002) and the importance of these characteristics may change over time, it is not possible to capture all characteristics of liquidity in one measure. Therefore, in this paper, four different measures of liquidity are investigated: the turnover ratio, the Amihud ratio, Zeros and the quoted bid-ask spread. Using these four liquidity measures makes it possible to gauge different aspects of liquidity. For all four liquidity measures used, stock liquidity is measured on a daily basis, using time series data. In subparagraphs below the four liquidity measures are discussed.

3.2.1 Turnover ratio

This study uses trading turnover as a measure of liquidity. The turnover ratio is calculated as the average of the daily ratio of the number of shares traded divided by the total number of shares outstanding. The following formula is used:

𝑇𝑢𝑟𝑛𝑜𝑣𝑒𝑟 𝑟𝑎𝑡𝑖𝑜 =1 𝑛∑ 𝑉𝑂_𝑖,𝑡 𝑁𝑆𝑖,𝑡 𝑛 𝑡=1

(1)

𝑉𝑂𝑖,𝑡 is the number of shares traded of firm i on day t

𝑁𝑆𝑖,𝑡 is the number of shares outstanding of firm i on day t

𝑁 is the number of trading days for which data are available

Various studies (e.g. Lakonishok and Lev, 1987; Datar, Naik, and Radcliffe, 1998; Wulff, 2002; Huang et al., 2015) use stock turnover as a proxy for liquidity. This measure is preferred to the volume of traded shares as it standardizes volume into a statistic, making it possible to compare firms of different size. Datar et al. (1998) also mention to use turnover as a proxy for liquidity, as the turnover ratio is negatively correlated with the bid-ask spread. The higher stock turnover, the more liquid the share and the easier it is to trade a large number of shares without moving the market. If liquidity benefits are present for stocks as a result of a split, an increase in the turnover ratio is expected in one or all of the post-announcement periods.

3.2.2 Amihud Ratio

13 𝐴𝑚𝑖ℎ𝑢𝑑 = LN[1 +1 𝑛∑ |𝑅_𝑖,𝑡| 𝑃𝑖,𝑡𝑉𝑂𝑖,𝑡 𝑛 𝑡=1 ] (2) Where

𝑅𝑖,𝑡 is the daily stock return for firm i on day t, calculated as LN(RIt/ RIt−1)

𝑃𝑖,𝑡 is the stock price for firm i at the end of day t

𝑉𝑂𝑖,𝑡 is the trading volume for firm i on day t

𝑁 is the number of trading days for which data are available

To avoid taking the logarithm of zero when zero returns are made on a day, the number one is added as a constant to the formula before calculating the natural logarithmic value.

3.2.3 Zeros

The third liquidity measure investigated in this study is the measure Zeros, which is developed as a proxy for transaction costs. Lesmond, Ogden, and Trzcinka (1999) introduce the measure, which gauges the proportion of zero returns of a firm over a certain period. The following formula is used:

𝑍𝑒𝑟𝑜𝑠 =Number of trading days with zero return forfirm i in period t

Number of trading days for firm i in period t

(3)

The measure is based on the fact that stocks with lower liquidity are more likely to experience days with zero returns (Goyenko et al., 2009). Furthermore, Lesmond et al. (1999) argue that arbitrageurs will only trade if the information signal is sufficient to overcome the costs involved in performing the transaction. Thus, stocks with higher transactions costs are more likely to experience zero return days as it will take more time to overcome the transaction costs.

3.2.4 Quoted bid-ask spread

In line with Amihud and Mendelson (1986), this paper uses the quoted bid-ask spread as a proxy for liquidity. The quoted bid-ask spread is defined as the spread between the ask price and the bid price, divided by the midpoint of the ask and bid price. The formula to calculate the quoted bid-ask spread is the following:

𝑄𝑢𝑜𝑡𝑒𝑑 𝑏𝑖𝑑 − 𝑎𝑠𝑘 𝑠𝑝𝑟𝑒𝑎𝑑 =1_n∑ PAi,t−PBi,t

(PA_i,t+ PB_i,t)/2 n

t=1 (4)

Where

𝑃𝐴𝑖,𝑡 is the ask price of firm i at the end of day t

PBi,t is the bid price of firm i at the end of day t

𝑁 is the number of trading days for which data are available

14 per stock and the price per stock are related. The spread per stock tends to increase proportionally with the price of a stock, equalizing the transactions costs per euro exchanged. If spreads did not increase proportionally with the price per stock, it would become profitable for those who submit limit orders to narrow spreads on the stocks with a greater spread per euro exchanged. Ceteris paribus, the higher a stock is priced, the higher the spread. In this research, therefore, the quoted bid-ask spread is used, to be able to more precisely reflect the cost of trading. When the quoted bid-ask spread declines, the stock will become more liquid.

3.3 Changes in liquidity

To investigate whether liquidity improves for firms splitting their stock, the presence of significant differences in liquidity measures between the pre-announcement period and each of the five post-announcement periods is examined. To further evaluate changes in liquidity, a control sample consisting of splitting firms is used. This control sample allows controlling for the existence of non-event driven changes in liquidity, comparing liquidity measures of split firms with liquidity measures of control firms. Moreover, the control sample enables to investigate whether firms in the split sample become more liquid over time in comparison to firms in the control sample.

The parametric t-test and the non-parametric Wilcoxon signed-rank test, are used to investigate the effect of stock splits on liquidity. The Wilcoxon signed-rank test is used because using daily data may lead to non-normality in the data, as is stated by Brown and Warner (1985). The validity of a parametric test decreases if the data do not follow a normal distribution. Therefore, using both the parametric t-test and the non-parametric Wilcoxon signed-rank t-test strengthen the results if the same conclusions can be drawn from both tests. The t-tests and the Wilcoxon signed-rank tests are performed by means of software package SPPS version 23, testing for a zero mean and zero median respectively.

3.4 Abnormal returns

The methodology followed in this paper to calculate the abnormal returns is the “Event Study Methodology” exemplified by Brown and Warner (1985). An event study is mostly used to investigate whether unexpected events can influence the value of a firm. In case of efficient markets, the effects of such an unexpected event will immediately be reflected in the price of a stock. The findings of the event study therefore provide a general estimate about the influence of stock split on the value of a firm.

15 market’s return, the variance of the abnormal returns is reduced. As a result, the market model can lead to an increased ability to detect the effects of stock splits. Therefore, this paper employs the market model to compute the abnormal returns based on the recommendations of MacKinlay (1997). The first step in conducting an event study is to choose an estimation window and event window. An estimation window is used to estimate what the normal stock returns should be in the event window. The estimation window will be set at AD-252 to AD-3, which is the same period as the pre-announcement period presented in Figure 1. In this paper, multiple event windows are used, investigating abnormal returns in all five post-announcement periods. However, the abnormal returns in the announcement period (AD-2, AD+2) and the ex-date period (ED0, ED+4) are specifically of interest.

Before calculating the abnormal returns, the daily realized return of a stock is calculated using the following formula:

𝑅𝑖,𝑡= 𝑙𝑛( 𝑅𝐼_𝑖,𝑡

𝑅𝐼𝑖,𝑡−1) (5)

where 𝑅𝑖,𝑡 is the return for firm i on day t and RIi,t is the end of the day total return index for firm i on

day t respectively.

The following step is calculating the estimated returns using the market model methodology described by Brown and Warner (1985):

𝑅𝑖,𝑡= ∝𝑖+ 𝛽𝑖*𝑅𝑚,𝑡 +

i,t

where Ri,t and Rm,t are the return for firm i on day t and the return on the market portfolio m on day

t respectively. ∝iand βi are the parameters of the model and 𝑖,𝑡is the zero mean disturbance term.

In practice the market portfolio is usually proxied by a broad-based stock index to reflect the overall market fluctuations. This paper uses the Total Market Index of the Netherlands (Datastream item TOTMKNL-index), which includes every listed Dutch firm from 1973 onwards.

The next step is to measure the abnormal return (AR), which can be calculated as the difference between the actual return and the expected return. To calculate the abnormal return, the following formula is used, as defined by Brown and Warner (1985):

𝐴𝑅𝑖,𝑡= Ri,t− ∝̂i− βî *Rm,t (7)

where ARi,t is the abnormal return for firm i on day t, Ri,t denotes the return for firm i on day t, Rm,t

16 market model regression. The regression coefficients are estimated over a period of 250 days, from day -252 to day -3.

The abnormal returns must be aggregated in order to draw overall inferences for an event. The cumulative abnormal return is the sum of all abnormal returns and is calculated as formulated by MacKinlay (1997):

𝐶𝐴𝑅𝑖(𝜏1 ,𝜏2)= ∑ 𝐴𝑅𝑖,𝑡

τ₂

τ=τ1 (8)

where 𝐶𝐴𝑅𝑖(𝑡₁,𝑡₂) is the sample cumulative abnormal return over time interval (𝑡1 , 𝑡2).

In order to test whether abnormal returns are present, a parametric t-test is performed, testing whether abnormal returns differ significantly from zero. In addition, the Wilcoxon signed-rank test is used as a robustness check, testing if the median cumulative abnormal return is significant different from zero.

3.5 Regression models

To investigate if changes in liquidity can explain abnormal returns around the announcement and execution of a split, a regression analysis is employed. The influence of changes in liquidity is tested through ordinary least square regressions. The following regression models are estimated:

17 split will be announced when the price of the stock is relatively high compared to the market. If the primary motivation for a stock split is to be in a lower price range, a negative relation between PO_Price and the cumulative abnormal return is expected. This negative relation indicates that the reaction of the market is more positive for firms with a lower price after the split. STD is the change in the return standard deviation between the pre-announcement period and the long-term post ex-period and is used as a proxy for stock price elasticity of demand. Further, following Wulff (2002) and Huang, Liano, and Pan (2009) the variable Size is measured as the natural logarithm of the market value five trading days before the announcement of the split. As suggested by Dennis and Strickland (2003) size is included in the regression as a proxy for risk and asymmetry in information.

4. Data and descriptive statistics

This chapter outlines the data used in this paper and presents the descriptive statistics of the split sample. Section 4.1 presents the sample selection procedure. Section 4.2, outlines how control firms are selected to match with the split firms. Section 4.3 presents the data needed to construct the four liquidity measures. In the final section, section 4.4, descriptive statistics are given of the split sample.

4.1 Sample selection and requirements

This paper focuses on a sample of splitting firms listed on the Amsterdam Stock Exchange, today known as Euronext Amsterdam. The firms in this sample are listed on the Amsterdam Exchange Index (AEX), the Amsterdam Midcap Index (AMX), the Amsterdam Smallcap Index (AScX) or locally. Data are collected for firms splitting their stock in the period January 1995 to December 2014. The sample starts in 1995, as more daily data are available from that moment onwards. Only firms splitting their stock before the end of December 2014 are included, as data needs to be available 260 trading days after the ex-split date.

The main sources used to collect data are Datastream, the LexisNexis database and the website of Behr2_{, which contains a list of Dutch stock splits. First, firms listed on the Euronext Amsterdam which} split their stock in the past need to be found. Since Datastream does not provide the names of the splitting firms, nor the split factor or the ex-split dates of the splitting firms, the information on Behr.nl is used to compose such a list.

For a splitting firm to be included in the final sample, several requirements need to be met. First, the date a firm announced a split needs to be available in LexisNexis or another reliable source on the internet. Splitting firms with an unknown announcement date are not included in the sample. Second,

18 the stock split must have the characteristics of a forward stock split. Stocks splits which are actually stock dividends, claim emissions or reverse stock splits are eliminated. Various papers analysing stock splits (Park and Krishnamurti, 1995; Desai and Jain, 1997 ) state that stock splits with a split factor between 1 and 1.25 are actually stock dividends instead of stock splits. Therefore, to be incorporated in the final sample, the split factor must be at least 1.25. Stock splits with a split factor lower than one, are classified as reverse stock splits and these reverse stock splits bear other characteristics. Furthermore, stock splits announced due to a merger or acquisition are also not included in the sample. In addition, at least 60% of the data needs to be available for the following liquidity measures: the turnover ratio, the Amihud ratio, Zeros and the quoted bid-ask spread. Firms splitting their stock, with less than 60% data available in the period 252 trading days before the announcement date to the announcement date and from the execution date to 260 trading days after the execution date, do not appear in the final sample. To avoid dependence in overlapping data caused by firms splitting their stocks multiple times in a short time period, a split is only included in the analyses when the previous split was more than two years prior to the next split.

The announcement dates of the stock splits are hand-collected using the LexisNexis database. The date of the first publication of the announcement of a stock split in the press, is considered as the announcement date. Reliable public sources on the internet, such as firm websites, are used in case no press release of a stock split can be found in the LexisNexis database. Furthermore, including days on which the stock exchanges in the Netherlands are closed (e.g. holidays) may lead to a distortion in the data; these are therefore deleted manually.

The resulting sample eventually consists of 101 splits and 77 unique firms, as some firms announced a stock split more than once in the sample period. For the list of firms included in the sample, the corresponding announcement date, ex-split date and split factor, refer to Appendix A. Appendix C presents the sample of splitting firms for which data are available to construct the quoted bid-ask spread. In contrast to the other liquidity measures, fewer data are available on bid prices and ask prices especially before 1997, therefore a different sample is constructed to test for changes in the quoted bid-ask spread.

4.2 Sample selection control firms

19 (Amsterdam SE, EX, AEX & AMX) are added to this list. These dead stocks are included to make sure no survivor bias is present in this research. The TOTMKNL-index and the list LNLXSTCK consists of 117 active firms and 64 delisted firms respectively, making a total of 181 companies present in the candidate pool and available to match with the sample of splitting firms. To find the best match for every split in the sample separately, firms of the split sample are matched by hand with one of the 181 firms in the candidate pool.

The following criteria, based on the paper of Huang et al. (2015), are used to construct the control sample:

1) The price of the control firm is in a ten percent range of the price of the splitting firm three trading days before the announcement of the stock split. The stock price is used as a selection criteria as stock liquidity is correlated with the stock price of a stock as is shown by Gargett (1978).

2) The control firm did not execute a stock split in the period two years before the split announcement of the split firm.

3) From the control firms that meet the first and second requirement, the firm with the closest Amihud ratio to the splitting firm in the pre-announcement period (AD-252, AD-3) is selected. 4) At least 60% of the data need to be available for the following liquidity measures: the turnover

ratio, the Amihud ratio, Zeros and the quoted bid-ask spread.

5) When it was not possible to find a firm that meet all above the requirements, the first requirement, i.e. that the stock price of the control firm is in a ten percent range of the price of the splitting firm, is dropped. Instead, a firm is chosen which meet all the requirements except the first requirement.

4.3 Data liquidity measures

The four liquidity measures in this paper are constructed using various variables from Datastream. The subparagraphs below present the Datastream items used to construct each of the liquidity measures. 4.3.1 Turnover ratio

20 The Unadjusted Volume of Shares is used, since the Datastream item Adjusted Volume of Shares (VO) is adjusted for capital events (e.g. stock splits). Using the Adjusted volume of shares would bias the results especially after the ex-split date, as the number of shares outstanding will grow while the volume of traded shares is adjusted for this growth.

4.3.2 Amihud Ratio

The Amihud ratio is calculated using Datastream items: Return Index (RI), the adjusted stock price (P) and adjusted trading volume (VO). The RI measures the change in value of a share holding over a specific period in time. The RI is used instead of the Price Index (PI), as the RI takes into account dividend pay-outs by firms. The stock price and trading volume are adjusted for different capital events including stock splits.

4.3.3 Zeros

To construct liquidity measure Zeros, the return on a stock needs to be calculated. To calculate the daily return on a stock, Datastream item RI and formula 5 are used.

4.3.4 Quoted bid-ask spread

The quoted bid-ask spread is computed using Datastream items PA and PB which provide ask prices and bid prices at the closing of the market. As there are no reasons to expect zero values, zero values must be errors and are eliminated replacing these zeros with NA (Not Available).

4.4 Descriptive statistics

The final sample consists of 101 splits, executed by 77 firms in the period 1995 to 2014. Table 1 reports the number of splits per year for the period 1995 to 2014. The splits of the sample firms are not uniformly distributed over this period. Most of the splits are executed in the late 1990s, especially 1997 was a year in which many firms listed on Amsterdam Stock Exchange split their stock. In the period 2005-2008, the years before the credit crisis, the economy of the Netherlands grew with 2.8% 3 _while over the period 2009-2014 the economy shrunk with 0.2% (Notten, 2015). As firms split their stock after a rise in their stock price (Fama et al., 1969), this downturn of the economy can partly explain the low number of stock splits after 2007.

Table 1: Splits by year in the period 1995-2014

Year Number Percentage

21 2001 6 5.9% 2003 2 2.0% 2004 2 2.0% 2005 3 3.0% 2006 8 7.9% 2007 7 6.9% 2008 1 1.0% 2010 1 1.0% 2011 1 1.0% 2013 2 2.0% 2014 1 1.0% All 101 100.0%

This table presents the frequency of stock splits by year for the firms included in the sample over the period 1995-2014. The descriptive statistics concerning the split factors of the splitting firms are shown in Table 2. The split factor is defined as the number of stocks received for each stock owned before the split. The split factors of the sample vary between 1.25 and 30, while the average split factor is 3.80 with a median of 3. Similar to other research done on stock splits (e.g. Conroy et al., 1990; Ikenberry et al., 1996), the results show that certain split factors are more common than others. The most frequently used split factors in the split sample are respectively two and four.

Table 2: Splits by split factor

Split Factor Number of Splits (%)

10 4 (4.0%) 5 15 (14.9%) 4 27 (26.7%) 3 10 (9.9%) 2.5 9 (8.9%) 2 30 (29.7%) 1.25 3 (3.0%) Other 3 (3.0%) Total 101 (100.0%)

This table present the number of stock splits by split factor. The sample period is 1995-2014.

5. Empirical results

22

5.1 Change in stock liquidity

If firms derive liquidity benefits from a split, an increase in the turnover ratio and a decrease in the Amihud ratio, Zeros and the quoted bid-ask spread should be observed. To test for changes in liquidity, the mean and the median are examined for the four liquidity measures during the pre-announcement period (AD-252, AD-3), the announcement period (AD-2, AD+2), the announcement-to-ex period (AD+3, ED-1), the ex-date period (ED0, ED+4), short-term post-ex period (ED+5, ED+10) and the long-term post-ex period (ED+11, ED+260).

Panel A of Table 3 shows the mean and median turnover ratios. The mean turnover ratio is 0.42% in the pre-announcement period and increases to 0.71% in the announcement period. The turnover ratio drops to 0.46% in the announcement-to-ex period and remains 0.46% in the ex-date period. Thereafter the trading turnover increases to 0.52% in the short-term post-ex period and decreases again to 0.47% in the long-term ex period. Although the turnover ratio fluctuates in the different announcement periods, the ratio remains above the pre-announcement level in all post-announcement periods. In line with research of Huang et al. (2015), a peak in trading turnover is found in the announcement period. This peak could be the result of increased attention paid to the split firm around the announcement of the split, or increased optimism of investors about the future outlook of the firm.

Using the t-test to examine whether stock liquidity improves after a split announcement, the following conclusions can be drawn: the turnover ratio is significantly higher in the announcement period, the announcement-to-ex period, short-term post-ex period and the long-term post-ex period compared to the pre-announcement period. In contrast, no significant difference between the turnover ratio in the pre-announcement period and the ex-date period is found. The results of the

Table 3: Liquidity measures around stock splits

Intervals Mean Change Median Change

Panel A: Trading Turnover (%)

Pre-Announcement Period 0.416 0.347

Announcement Period 0.710 0.293*** 0.493 0.146***

Announcement-to-Ex Period 0.458 0.041** 0.375 0.027***

Ex-Date Period 0.456 0.040 0.339 -0.008

Short-Term Post-Ex Period 0.518 0.102*** 0.371 0.024***

Long-Term Post-Ex Period 0.471 0.055** 0.367 0.020*

Panel B: Amihud ratio (x 10-6

)

23

Announcement Period 24.795 -34.597*** 3.646 -3.333***

Announcement-to-Ex Period 25.318 -34.075*** 6.523 -0.456***

Ex-Date Period 34.121 -25.272*** 3.142 -3.837***

Short-Term Post-Ex Period 24.154 -35.239*** 3.140 -3.839*** Long-Term Post-Ex Period 29.885 -29.508*** 6.238 -0.741***

Panel C: Zeros (%)

Pre-Announcement Period 10.408 5.600

Announcement Period 6.931 -3.477*** 0.000 -5.600***

Announcement-to-Ex Period 6.804 -3.603*** 4.237 -1.363***

Ex-Date Period 10.099 -0.309 0.000 -5.600*

Short-Term Post-Ex Period 8.086 -2.322* 0.000 -5.600*** Long-Term Post-Ex Period 6.681 -3.727*** 4.400 -1.200***

Panel D: Quoted bid-ask spread (%)

Pre-Announcement Period 0.616 0.412

Announcement Period 0.509 -0.108 0.384 -0.028*

Announcement-to-Ex Period 0.453 -0.164** 0.334 -0.078***

Ex-Date Period 0.549 -0.067 0.390 -0.022

24 Wilcoxon signed-rank test are in line with the results of the t-test employed. In summary, trading turnover increases after the announcement of a split and remains above the pre-announcement level in all periods thereafter, although not significantly in the ex-date period.

Panel B of Table 3 reports the Amihud ratio, showing similar results as those of the turnover ratio. The Amihud ratio is lower in all post-announcement periods compared to the pre-announcement period, suggesting increased liquidity levels. However, the results show the mean Amihud ratio increases again in the ex-date period after the decrease in the announcement period and announcement-to-ex period and becomes closer to the pre-split level although staying significantly below this level. Both the results of the t-test and Wilcoxon signed-rank test show that the decrease in Amihud ratio is significant at 1% level for all periods. Thus, the results provide evidence for an improvement in liquidity around the announcement date and this improvement remains in the long run after the split.

Panel C of Table 3 contains information about the mean and median percentages of days with a zero return. The results show the mean of liquidity measure Zeros declines substantially from 10.41% in the pre-announcement period to 6.9% in the announcement period and declines even further to 6.8% in the announcement-to-ex period. However, in the ex-date period the number of days with zero return increase to 101% and become close to pre-split level again. This results in finding no significant differences in the number of days with zero return between the pre-announcement period and the ex-date period. In the periods after the ex-ex-date period the mean liquidity measure Zeros declines substantially and becomes significantly different from the pre-split level again. These results suggest improved liquidity in the announcement period and this improvement is long lived. However, just like the results found for the turnover ratio, liquidity measure Zeros does not find evidence liquidity significantly differs from the pre-split level in the ex-date period4_.

The mean and median quoted ask spread are reported in panel D of Table 3. The mean quoted bid-ask spread is lower in all post-announcement periods compared to the pre-split level. However, in line with the results for the liquidity measures turnover ratio and Zeros, the results are not significant in the ex-date period. Furthermore, there is no evidence for a significant difference in the quoted bid-ask spread between the pre-announcement period and the announcement period5_{. In the long-term} post-ex period the mean quoted bid-ask spread is below pre-split level, while the median quoted bid-ask spread is above pre-split level. Both results are significant, and the quoted bid-ask spread therefore gives an inconclusive answer to whether the liquidity improvement is sustained in the long run.

25 The results presented above suggest an improvement in liquidity around the announcement of a stock split. Most liquidity measures suggest this improvement in liquidity even last till one year after the split. This finding is not in line with research by Huang et al. (2015), who find the improvement in liquidity around the announcement of a split is short lived, and in the long run liquidity even falls below pre-split level. Furthermore, the results in Table 3 show that in the ex-date period the mean of the liquidity measures becomes closer to pre-split level, and only the Amihud ratio suggests liquidity in the ex-date period is significantly different from liquidity in the pre-announcement period. Finding an increase in the mean quoted bid-ask spread in the ex-date period compared to levels in the announcement period and announcement-to-ex period could be an explanation for this decline in liquidity, suggesting the costs of trading increase in the ex-date period. This is in line with the findings of Copeland (1979) and Conroy et al. (1990) who find an increase in the bid-ask spread after a split.

5.2 Liquidity difference between split sample and control sample

To further investigate if a stock split improves liquidity of a stock, liquidity measures of split firms are compared to those of control firms. Table 4 and 5 present the mean and median liquidity measures for the split sample and control sample respectively, as well as their differences. Furthermore, the adjusted changes in the four liquidity measures are reported, to examine whether changes in liquidity over time are different for firms in the split sample compared to firms in the control sample. To investigate if the results reported in Table 4 and 5 are significant, the t-test and the Wilcoxon signed-rank test are used, testing a zero mean and zero median respectively.

Panel A of Table 4 shows that split firms experience a significantly higher change in the turnover ratio than control firms in the announcement period, ex-date period and short-term post-ex period. In these three periods, firms in the split sample also experience a significantly higher trading turnover compared to firms in the control sample. Panel A of Table 5 shows almost the same results, although in the ex-date period no significant difference in the turnover ratio between split firms and control firms is found.

The results for the Amihud ratio reported in Panel B of Table 4 only show a significant, but positive, difference between split firms and control firms during the pre-announcement period. This finding suggests split firms are only less liquid than the control firms during this period. Furthermore, the change in Amihud ratio from the pre-announcement period to each of the post-announcement periods is significantly higher for split firms than for control firms, indicating the split firms become

Table 4: Mean liquidity measures for split firms and control firms

Intervals Splitting firms Control firms Difference

Adjusted Change

26

Pre-Announcement Period 0.416 0.414 0.002

Announcement Period 0.710 0.432 0.277*** 0.276***

Announcement-to-Ex Period 0.458 0.427 0.031 0.029

Ex-Date Period 0.456 0.360 0.096** 0.094**

Short-Term Post-Ex Period 0.518 0.401 0.118** 0.116***

Long-Term Post-Ex Period 0.471 0.435 0.036 0.034

Panel B: Amihud ratio (x 10-6

)

Pre-Announcement Period 59.393 29.980 29.412**

Announcement Period 24.795 30.699 -5.904 -35.316***

Announcement-to-Ex Period 25.318 21.169 4.149 -25.263***

Ex-Date Period 34.121 24.336 9.784 -19.628*

Short-Term Post-Ex Period 24.154 37.661 -13.507 -42.920***

Long-Term Post-Ex Period 29.885 42.923 -13.038 -42.451***

Panel C: Zeros (%)

Pre-Announcement Period 10.408 8.495 1.913**

Announcement Period 6.931 8.317 -1.386 -3.299*

Announcement-to-Ex Period 6.804 7.982 -1.177 -3.090***

Ex-Date Period 10.099 5.347 4.752** 2.840

Short-Term Post-Ex Period 8.086 3.960 4.125** 2.213

Long-Term Post-Ex Period 6.681 7.275 -0.594 -2.507***

Panel D: Quoted bid-ask spread (%)

Pre-Announcement Period 0.616 0.530 0.087

Announcement Period 0.509 0.474 0.035 -0.052

Announcement-to-Ex Period 0.453 0.450 0.003 -0.084

Ex-Date Period 0.549 0.406 0.143* 0.056

Short-Term Post-Ex Period 0.434 0.472 -0.039 -0.126

Long-Term Post-Ex Period 0.474 0.499 -0.026 -0.113

27 Table 5: Median liquidity measures for split firms and control firms

Intervals Splitting firms Control firms Difference

Adjusted Change

Panel A: Trading Turnover (%)

Pre-Announcement Period 0.347 0.346 -0.026

Announcement Period 0.493 0.309 0.189*** 0.215***

Announcement-to-Ex Period 0.375 0.367 0.026 0.052

Ex-Date Period 0.339 0.347 0.049 0.075*

Short-Term Post-Ex Period 0.371 0.302 0.060* 0.085***

Long-Term Post-Ex Period 0.367 0.353 -0.020 0.005

Panel B: Amihud Illiquidity (x 10-6

)

Pre-Announcement Period 6.979 6.072 0.965***

Announcement Period 3.646 2.894 -0.326 -1.291***

Announcement-to-Ex Period 6.523 3.490 -0.087 -1.052***

Ex-Date Period 3.142 2.458 0.256* -0.709

Short-Term Post-Ex Period 3.140 4.443 -0.121 -1.086***

Long-Term Post-Ex Period 6.238 4.969 0.025 -0.940***

Pancel C: Zeros (%)

Pre-Announcement Period 5.600 5.600 1.200*

Announcement Period 0.000 0.000 0.000 -1.200*

Announcement-to-Ex Period 4.237 4.815 0.000 -1.200***

Ex-Date Period 0.000 0.000 0.000** -1.200**

Short-Term Post-Ex Period 0.000 0.000 0.000** -1.200

Long-Term Post-Ex Period 4.400 5.200 0.000 -1.200***

Panel D: Quoted bid-ask spread (%)

Pre-Announcement Period 0.412 0.476 0.010

Announcement Period 0.384 0.352 0.050 0.041

Announcement-to-Ex Period 0.334 0.424 -0.002 -0.012

Ex-Date Period 0.390 0.258 0.021* 0.012

Short-Term Post-Ex Period 0.352 0.328 0.007 -0.003

Long-Term Post-Ex Period 0.463 0.419 0.038 0.029

28 more liquid than the controls firms in all post-announcement periods. Based on the findings of Huang et al. (2015), it is expected that the split firms become less liquid compared to the control firms in the period one-year after the split. However, the adjusted change of the Amihud ratio shows the opposite, indicating in the long-term post-ex period split firms become more liquid compared to control firms. Using the Wilcoxon signed-rank test, Panel B of Table 5 gives almost the same results although the adjusted change in the ex-date period is not significant.

In Panel C of Table 4 and 5 the results for liquidity measure Zeros are reported. The mean difference in the liquidity measure Zeros between the split sample and control sample is positive and significant in the pre-announcement period, the ex-date period and the short-term post-ex period6_{. This positive} difference implies that split firms experience a proportionally higher number of days with a zero return compared to the control firms in these periods. Significant negative adjusted changes are present in the announcement, announcement-to-ex and long-term post-ex period, suggesting firms in the split sample become more liquid in these periods compared to the firms in the control sample. Contrary to the results of the t-test, the Wilcoxon signed-rank test suggests firms in the split sample become more liquid around the ex-date compared to the firms in the control sample.

Panel D of Table 4 and 5 shows the results for the quoted bid-ask spread. The results suggest a higher quoted bid-ask spread in the ex-date period compared to the split sample, although only significant at a 10% level. Furthermore, none of the mean and median adjusted changes are significant; indicating there is no evidence for a different change in liquidity for split firms compared to control firms. However, finding no significant results for the quoted bid-ask spread could be caused by the smaller sample size. In comparison to the other liquidity measures, the sample for the quoted bid-ask spread is much smaller and only comprises 37 firms due to a lack of available data. This lower sample size decreases the power of the tests, making it harder to detect a meaningful difference between the groups.

In summary, comparing split firms and control firms and their change in liquidity measures over time give some mixed results. However, using the quoted bid-ask spread may not be suitable to detect significant differences between groups due to the low sample size. Therefore the primary focus will be on the other three liquidity measures when interpreting the results. The other three liquidity measures suggest that liquidity for split firms improve more than for control firms. The liquidity measures Amihud ratio and Zeros even suggest this improvement in liquidity is long-lived and still present one year after the execution of the split.

29

5.3 Cumulative abnormal returns

Table 6 describes the market performance of the firms in the split sample. The RUNUP rate is the return firms in the split sample made in the pre-announcement period. The mean (median) RUNUP rate is 40.69% (39.22%) in pre-announcement period. This result is in line with the findings of Fama et al. (1969) and Huang et al. (2015) who observe that firms announce a stock split after a rise in their stock price. The cumulative abnormal return is computed using formulas 5-8, calculating the difference between the return of the split firm and the expected return of the split firm based on the market model.

Table 6: Market performance

Mean Median

RUNUP (%) 40.686*** 39.222***

Cumulative Abnormal Return (%)

Announcement Period 2.649*** 2.330***

Announcement-to-Ex Period -3.122*** -0.311***

Ex-Date Period -0.111*** -0.437***

Short-Term Post-Ex Period 0.269*** -0.183*** Long-Term Post-Ex Period -27.106*** -20.93***

This table reports the market performance of the split sample. The RUNUP rate is the return made in the pre-announcement period (AD-252, AD-3). The cumulative abnormal return is the return difference between the return of the split firm and the expected return of the split firm. The expected return is calculated, using the market model, based on the stock performance in the pre-announcement period in relation to the performance of the TOTMKNL-index. The cumulative abnormal returns are calculated for five time periods: the announcement period (AD-2, AD+2), the announcement-to-ex period (AD+3, ED-1), the ex-date period (ED0, ED+4), short-term post-ex period (ED+5, ED+10) and the long-term post-ex period (ED+11, ED+260). The t-test and the Wilcoxon signed-rank test are used to test a zero mean and median respectively to examine whether the cumulative abnormal return differs significantly from zero. The symbols ***, ** and * denote statistical significance at the 1%, 5% and 10% levels, respectively.

Over the announcement period and the long-term post-ex period the returns for the split firms differ significantly from their expected returns based on the market model. Finding a positive cumulative abnormal return in the announcement period is in line with prior research by Grinblatt et al. (1984), Buijs and Gerritsen (2005) and Huang et al. (2015). The mean (median) cumulative abnormal return in the announcement period is positive with 2.65% (2.33%) and significant at a 1% level. In the long run the split sample firms are not able to meet their market model-based expected returns. Splitting firms experience on average 27.11% lower returns over the long-term post-ex period than expected. The underperformance in this period indicates that firms splitting their stock are not able to hold on to the return experienced in the year before the stock split announcement. This existence of a negative cumulative abnormal return one year after the execution of a split is consistent with prior research by Huang et al. (2015).

30 buying stocks after the announcement due to the inconvenience related with record dates and the delivery of the stocks of split firms. Therefore negative abnormal returns are expected as a result of a decrease in the price of the stock in this period. In contrast to other studies (see e.g. Grinblatt et al., 1984; Wulff, 2002; Buijs and Gerritsen, 2005; Huang et al., 2015) no evidence is found for abnormal returns over the ex-date period and the sign is also the opposite of what would be expected. Therefore, these findings do not suggest investors will wait till right after the execution of a stock to buy a stock that is in a lower trading range.

To summarise, significant abnormal returns only exist in the announcement period and long-term post-ex period. No significant evidence is found for the announcement-to-post-ex period, the post-ex-date period and the short-term post-ex period.

5.4 Abnormal returns and changes in liquidity

Now that the cumulative abnormal returns are calculated and analysed, the following step is to investigate whether the change in liquidity can explain the cumulative abnormal return observed in the announcement period. As defined by formulas 9-11 three regression analyse are employed to investigate this relationship. The quoted bid-ask spread is not used as an independent variable in the regression analysis, because the sample size was too small to draw a valid conclusion from the regression analysis.

31 Finding Amihud can explain the announcement effect but Turnover and Zeros cannot, may be related to the smaller improvement in liquidity based on the Turnover ratio and Zeros in the announcement period in comparison to the Amihud ratio. Furthermore, in none of the three models, the control variables post-split stock price (PO_Price), the size of the firm (Size) and the change in return standard deviation(STD) seem to have had any effect on the existence of the cumulative abnormal return in the announcement period.

Table 7: Announcement date abnormal return and the change in liquidity measures

Independent variables Model (1) Model (2) Model (3)

Intercept 0.049** 0.024 0.047* (2.035) (0.845) (1.888) Turnover 0.005 (0.390) Amihud (x10^2) -0.014* (-1.769) Zeros (x10^2) _(-0.488) -0.024 PO_PRICE (x 10^3) -0.008 0.159 0.054 (-0.012) (0.244) (0.082) STD 0.212 0.661 0.129 (0.254) (0.775) (0.150) Size -0.004 -0.001 -0.003 (-1.077) (-0.299) (-0.935) Adjusted R2 _{-2.6% 0.4% -2.6%} F-statistic 0.356 1.110 0.378

32

6. Conclusion, limitations and recommendations for future research

Conclusion

Although stock splits seem to be purely cosmetic events, prior research suggests that stock splits are associated with the existence of abnormal returns and changes in liquidity. Motives behind the decision to split a stock, are broadly discussed in literature. The most popular motives to split a stock all imply an improvement in liquidity although they have a different view on when this improvement will occur and whether this improvement will be long-lived. Both the signalling hypothesis and the attention-grabbing hypothesis expect an improvement in liquidity around the announcement date of the split and this improvement is expected to be short lived. In contrast, the trading range/improved liquidity hypothesis expects an improvement in liquidity right after the execution of the split and foresees this improvement to be long-lived. Therefore, this paper examines the effect of a stock split on liquidity by investigating a two year period surrounding the announcement date and execution date of the split. In addition, a regression analysis is performed, aiming to find an answer to whether the change in liquidity can explain the existence of abnormal returns in the announcement period. Research is done for stocks listed on the Amsterdam Stock Exchange in the period 1995-2014, using a final sample of 101 stock split events.

33 Thus, interpreting the results, no decisive answer can be given on which hypothesis best explains the motive behind stock splits. The results are mildly supportive for both the signalling hypothesis and the attention-grabbing hypothesis as well as the trading range/improved liquidity hypothesis.

Limitations and suggestions for future research

34

References

Amihud, Y., 2002. Illiquidity and stock returns: cross-section and time-series effects. Journal of Financial Markets 5, 31-56.

Amihud, Y., Mendelson, H., 1986. Asset pricing and the bid-ask spread. The Journal of Financial Economics 17, 223-249.

Amihud, Y., Mendelson, H., 1991. Liquidity, asset prices and financial policy. Financial Analysts Journal 47, 56-66.

Asquith, P., Healy, P., Palepu, K., 1989. Earnings and stock splits. Accounting Review 64, 387-403. Baker, H., Gallagher, P., 1980. Management's view of stock splits. Financial Management, 73-77. Barber, B., Odean, T., 2008. All that glitters: The effect of attention and news on the buying behavior of individual and institutional investors. Review of Financial Studies 21, 785-818.

Brennan, M., Copeland, T., 1988. Stock splits, stock prices, and transaction costs. Journal of Financial Economics 22, 83-101.

Brown, S., Warner, J., 1985. Using daily stock returns: The case of event studies. Journal of financial economics 14, 3-31.

Buijs, A., Gerritsen, D., 2005. Aandelensplitsingen in Nederland: een praktijkonderzoek. Vba journaal 2, 21-29.

Byun, J., Rozeff, M., 2003. Long‐run performance after stock splits: 1927 to 1996. The Journal of Finance 58, 1063-1085.

Connelly, B., Certo, S., Ireland, R., Reutzel, C., 2011. Signaling theory: a review and assessment. Journal of Management 37, 39-67.

Conroy, R., Harris, R., Benet, B., 1990. The effects of stock splits on bid‐ask spreads. The Journal of Finance 45, 1285-1295.

Copeland, T., 1979. Liquidity changes following stock splits. Journal of Finance 34, 115-141.

Crawford, D., Franz, D., 2001. Stock dividends and splits: anticipation, signaling, and market response. Journal of Accounting, Auditing & Finance 16, 141-166.

Datar, V., Naik, N., Radcliffe, R., 1998. Liquidity and stock returns: an alternative test. Journal of Financial Markets 1, 203-219.

Demsetz, H., 1968. The cost of transacting. The Quarterly Journal of Economics 82, 33-53.

Dennis, P., Strickland, D., 2003. The effect of stock splits on liquidity and excess returns: evidence from shareholder ownership composition. Journal of Financial Research 26, 355-370.