Why Firms split their stocks in the Netherlands

UNIVERSITEIT VAN AMSTERDAM

Why Firms split their stocks in the

Netherlands

This study investigates why firms split their stock. I analyse a sample of stock splits on the Dutch stock market over the period 1987 through 2012. My analysis of the cross-sectional distribution of the split factor provides support for the “optimal price range” hypothesis in the stock split sample. This study finds that the coefficients of the relative price of the pre-split price to the market average price are positive.

I.

Introduction

This paper studies stock splits in the Netherlands between 1987 and 2012. The reason I wrote this paper is the recent stock split by the Coca Cola Company on the 27th of July in 2012. The stock split of 2:1 means the total shares of Coca Cola doubled from 5.6 billion shares to 11.2 billion shares; that is one of the historical effects of a stock split. It is to be noted that since 1919 when Coca Cola began trading, the recent stock split makes it the tenth time that the company would have its stock split and the first in the last 16 years. On the surface, Coca Cola's recent stock split does not affect the current holdings of existing shareholders. Simply put, it only doubles their holdings while the price of each stock drops to half of its price at the time of the split.

In this paper stock splits in The Netherlands are studied. This paper used the data from 1987 till 2012 because that was all the available data. In this analysis I focus on how firms choose the split factor to split their stocks. A lot of research has been done on this topic, but most studies used the U.S. stock market to research the split factor. This study extends this to the Netherlands. In this study we try to answer the following research question: What has been the price to split shares in The Netherlands between 1987 and 2012?

Baker and Powell (1993) show in various survey studies that companies use a stock split to increase the marketability of the shares. There is an optimal price range for the stock, where investors with less capital can easily purchase the stock and transactions costs are not too high, because of the fixed per-share transaction cost component, for investors with more capital e.g. institutional investors. Therefore, the argument goes, there exists an optimal price range that equilibrates the preferences of these classes of investors (Schrama &

Eijgenhuijsen, 2006).

In their study Wu and Chan (1997) found that their hypothesis about the optimal price range theory in the stock split sample is significant. According to their study there is a

positive effect of the relative price of the pre-split price to the market average price.

Lakonishok and Lev (1978) found for their data, from 1963 to 1982, that stock splits are used to restore stock prices, which increased considerably during an unusual growth period, to a normal range, defined in terms of market and industry-wide price and firm-specific prices.

Therefore, I expect in this paper that the split factor is explained by the market average prices. The market average price is the price of an index divided by the number of companies that are listed in that index1. This hypothesis is based on prior empirical findings.

This paper exists of three parts. First, this paper is going to highlight the existing literature about stock splits, the focus is on two theories why firm managers decide to split stocks and the hypothesis development. The second part shows us the research design; the used data and methodology. Last, the results of the regression are highlighted.

1

Suppose the AEX is noted 350, then the market average price of the AEX is 14 (350 / 25 = 14).

II.

Literature review

A stock split is used to reduce the nominal value of a stock by a certain factor, while there are more stocks issued increased by the same factor. The nominal stock capital is unchanged2. Also the control and any other rights of the shareholder do not change, nor the cash flows of the enterprise (Schrama & Eijgenhuijsen, 2006).

In theory the announcement of a share split should not bring any significant change in the share prices. For this reason the value of the share should decrease with the same factor which the number of shares increases. From empirical researches in the United States (Fama et al., 1969) becomes clear however that the announcement of a share split goes paired with a significant share increase of average two to five percent, where has been corrected for the market return.

Theories

There are numerous explanations for the existence of stock splits and the reaction of their prices. These explanations can be classified into two groups: signalling and price range.

Signalling

Financial decisions are used by managers, given asymmetric information between managers and investors, to signal information to the investors. This is called signaling (Lakonishok & Lev, 1978).

In particular, researchers believe that companies use a stock split to transfer

information to the investors about the good prospects of the undertaking, an increase of future dividends or profits. The reason of using a stock split for the information transmission is that analysts assume that the manager expects that the share will rise even further after a stock split. A pessimistic manager will not split a share, because bad results can make a splitted share go even under the acceptable lower limit, what has often due to that the share will drop even further. Moreover, it is stated that many indirect costs will have to be made if it appears that a company has given incorrect information afterwards. In the case of false signals the reputation of a company is at stake (Schrama & Eijgenhuijsen, 2006).

2_{Suppose you have a share with a nominal value of 100 euro. After a 5: 1 split, you have five} shares each with a nominal value of 20 euro.

However, there are no direct costs associated with sending false signals. Direct costs should be associated with sending false signals for a signaling device to be valid. Firms with low expected performance should not be able to copy the signaling decisions of firms with high expected performance (Lakonishok & Lev, 1978).

Price range

The second theory is seen by the companies as the main reason for a stock split. Baker and Powell (1993) show in various survey studies that companies use a stock split to increase the marketability of the shares. There is an optimal price range for the stock, where investors with less capital can easily purchase the stock and transactions costs are not too high, because of the fixed per-share transaction cost component, for investors with more capital e.g.

institutional investors. Therefore, the argument goes, there exists an optimal price range that equilibrates the preferences of these classes of investors (Schrama & Eijgenhuijsen, 2006). Managers adjust their stock prices to an optimum by splitting their stocks or

distributing stock dividends for the purpose of a broader and heterogeneous base of stockholders or a wider marketability of their stock (Lakonishok & Lev, 1978).

In a survey about managers’ motives for splitting a firm is shown that 98.4 percent of the respondents agreed that splits make the purchase of stocks easier for investors with small capital and 93.7 percent believed that splits keep a firm's stock price in an optimal range and increase the number of stockholders (Lakonishok & Lev, 1978).

In their research called On Existence of An “Optimal Stock Price”: Evidence from Stock Splits and Reverse Stock Splits in Hong Kong Wu and Chan (1997) researched the market effect of a stock split or a reversed stock split. They also did research on the optimal price theory.

Most of the data for their study was collected from the Fact Book published by the Stock Exchange of Hong Kong; all stock splits and reverse stock splits completed for the period 1986 through 1992. The Fact Book covers capital changes and financing activities in Hong Kong-listed companies. Wu and Chan used a sample of 67 stock splits and 29 reverse stock splits. There are no limitations imposed by the Stock Exchange of Hong Kong for the usage of either stock splits or reverse stock splits.

Most of their data was found in the Fact Book. Data concerning the company names, effective dates, and split factors was collected from the Fact Book. Wu and Chan (1997) also

collected confounding events around the announcement dates that might impose valuation effects on the splitting stock from the Fact Book. However, the announcement dates for the stock splits were collected from the Securities Bulletin published by the Stock Exchange of Hong Kong and from the Data Base of the University of Rhode Island PACAP the

researchers collected the daily stock returns, the market return, trading volume, and bid-ask spreads (Wu & Chan, 1997).

Wu and Chan (1997) show the distribution of the split factors. For the reverse stock splits in their sample the majority of the companies chose to have a 1:2 split (31.3%). Also, 1:4 and 1:5 splits were popular in their sample, 16 splits (23.9%) and 17 splits (25.4%), respectively. Nine 1:10 splits (13.4%) were made, while there was only one 2:5 split. Reverse stock splits exceeding 10 were rarely used. One 1:20 split and two 1:50 are shown in their sample. For the normal stock splits the 5:1 and the 4:1 were used by the majority of the companies in their sample. The 5:1 stock split accounts for 11 splits (37.9%) and the 4:1 stock split accounts for 8 splits (27.6%). The next in line of popular stock splits is the 10:1; five were shown in their sample. There was one of each of the 2:1, 6:1 and 15:1 types (Wu & Chan, 1997).

In their study Wu and Chan (1997) found that their hypothesis about the optimal price range theory in the stock split sample is significant. According to their study there is a

positive effect of the relative price of the pre-split price to the market average price (Wu & Chan, 1997).

Lakonishok and Lev (1987) suggest that stock splits are mainly aimed at restoring stock prices, which increased considerably during an unusual growth period, to a normal range, defined in terms of market and industry-wide price and firm-specific prices (Lakonishok & Lev, 1987).

In their study Lakonishok and Lev (1987) used data in a twenty-year period, 1963 to 1982. University of Chicago’s CRSP tape, the Merged Annual Compustat tape, and the Compustat Prices-Dividends-Earnings (PDE) monthly tape are resources for their data (Lakonishok & Lev, 1987).

The average stock prices of the two samples are close during the five years before the split. Whereas a gap exists in favor of the splitting firms thereafter. After the stock split this trend is reversed. The gap narrows and the average stock price of the two samples becomes equal again (Lakonishok & Lev, 1987).

Angel (1997) implies in his model that firm size, idiosyncratic risk, and the fraction of investors who "know about" a firm could be important considerations in determining the optimal stock price. These implications are testable by looking at the prices that companies choose immediately after a stock split, before their stock prices have had time to wander far from the optimal price. Companies may choose a target price below the optimal price so that the stock spends more time close to its optimal price before it is time to split again.

Nevertheless, the target price should be closely related to the optimal price (Angel, 1997). In his study Angel (1997) did a regression of idiosyncratic risk, age, the natural logarithm of firm size, the number of analyst who ‘’know about’’ and the regulated industry on price. Idiosyncratic risk is the variance of the residual from a 60 month beta regression using the CRSP Value weighted Index. Age represents the length of time for which price information is available for the firm on the CRSP NYSE/AMEX monthly file, the natural logarithm of the book value of the firm (firm size) is obtained from COMPUSTAT. The number of analysts is obtained from Zacks. The regulated industry dummy is set to one for industries with an SIC code in the 4000 range (Angel, 1997).

The idiosyncratic risk is based on the variance of the residuals from a 60-month beta regression using the CRSP value weighted index. The model predicts that firms with a higher idiosyncratic risk should choose a lower stock price, so the predicted coefficient is negative, as is the case for both split firms and control firms. Age, defined as the number of years the firm has been listed on the CRSP NYSE/AMEX monthly tape, is used as one proxy for the fraction of investors who "know about" the firm. Since better-known firms should choose a higher stock price, the coefficient should be positive, as it is in both cases. The number of analysts is another proxy for how many investors "know about" the firm, with the positive coefficients predicted in the model. However, the coefficient is not significant for the

splitting firms. Firm size is proxied by the natural logarithm of the book value of equity. Here the result is the opposite of that predicted by the partial derivatives in the model: Large firms choose higher share prices even after correcting for the other explanatory factors. It may be that firm size is capturing other aspects of how well known the firm is beyond age and number of analysts. A better proxy for the number of investors who "know about" a firm is needed. Finally, a dummy for regulated industries is set to one for SIC codes in the 4000 range. The negative coefficient shows that regulated firms choose lower share prices at split time than nonregulated firms, although the result is not significant for the control firms. A

regulated firm may desire to give incentives to market makers and brokerage firms to widen the firm's investor base so that it will have more political clout with regulators (Angel, 1997).

III. Research design

Data

The most data for this study, covering from 1987 till 2012, were drawn from DataStream. This paper used the data from 1987 till 2012 because that was all the available data.

DataStream was the source of data for the split date, split factor, price at the announcement date and the index price at the announcement date for the sixteen selected companies in the AEX index. Most selected companies had more than once splitted their stocks. The

announcement date was untraceable; therefore this paper used three months before the split date to find the price at the announcement date with respect to weekends and other days that the index was closed. Three months is chosen because in the most cases three months before the stock split companies announce a stock split. The variable Firm Size is the total

common/ordinary equity. Firm Size was drawn from Compustat. Some characteristics of stock splits of the distributing firms are presented in table 1 in the appendix.

Table 1 in the appendix shows the found split factors and the date of the split factors. The biggest split factor is 10 and the smallest is 1.003, the most frequent used split factor is 2. In table 1 also is shown that the price at the announcement date is ranging from 2.803 to 48.626 and that the index price at the announcement date is ranging from 104.9 to 630.98.

Methodology

It appears that stock prices of splitting firms are adjusted to a certain target, but what is this target? The simplest hypothesis consistent with the "price range" motivation for stock splits appears to be that, by splitting their firms' stock, managers attempt to adjust prices to a market-wide price average. The alternative hypothesis would, of course, be that the split factor or size of split distribution is chosen randomly. Accordingly, the "implicit target price," namely the stock price of a splitting firm j before the split announcement ( ), divided by the size of the split ( , is set equal to the market-wide average price at that date, PM:

( )/ ( =PM, (1)

where ( is the size of the split distribution (e.g., 1.5 if the split increased the number of shares by fifty percent). This expression can be rewritten as:

( ) = ( ⁄ (2) The interpretation of the second expression is that the larger the relative deviation of a firm's stock price from the market-wide average price, ( ⁄ , the larger the size of stock split

( ) (Lakonishok & Lev, 1978). In log form the regression3 looks like:

( ) ⁄ (3)

Table 2 in the appendix shows the natural logarithms of ( ) and ( ⁄ . ( ) varies from 0.024693 to 2.302585 and ln ⁄ varies from -1.49131 to 0.915064.

The second regression contains the control variable of Firms Size in the natural logarithm:

( ) ⁄ (4)

The variable ln( ) is the natural logarithm of the book value of the firm (Firm Size).

IV. Results

Descriptive statistics

Here under is a summary of the data, showing the mean, the standard deviation and the 95 percent confidence interval of the various variables. As can be seen there are thirty

observations. The means range between 1.172758 and 9585.759. The standard deviations of ( ) and ( ) are quite high, whereas the standard deviations of ( ⁄ ), ( ) and ( ) are quite low. As can be seen in the table ( ), ( ⁄ ) and ( ) have a relative small

range. The ranges are respectively 1.837474 – 3.284926; 0.9727072 – 1.37281 and 3315.195 – 15856.32. The other two variables, ( ) and ( ), have a relative big range. The ranges

are respectively 11.81445 – 20.51322 and 10.89005 – 15.17227.

Observations Mean Std. Deviation [95% Conf. Interval]

( ) 30 2.5612 1.938175 1.837474 – 3.284926

( ⁄ ) 30 1.1727758 0.5357471 0.97272072 – 1.37281

( ) 30 16.16383 11.64787 11.81445 – 20.51322

( ) 30 13.03116 5.733993 10.89005 – 15.17227

(FS) 27 9585.759 15851.29 3315.195 – 15856.32

The next table of the descriptive statistics shows the correlations between the various

variables. The table shows that ( ) has little correlation with ( ⁄ ), ( ) and ( ).

( ⁄ ) and ( ) have a high correlation with ( ), but a small correlation with each other. ( ) has a small correlation with all the other variables.

( ) ( ⁄ ) ( ) ( ) (FS) ( ) 1.0000 ( ⁄ ) 0.1606 1.0000 ( ) 0.1315 0.8409 1.0000 ( ) -0.0242 0.2968 0.7177 1.0000 (FS) 0.0178 -0.0713 -0.0665 0.0072 1.0000

Finally, in the scatterplot below is seen that the variables ( ) and ( ⁄ ) move with

each other. As becomes a higher value, ⁄ becomes a higher value and vice versa.

The two variables move with each other.

( )

Regression

A cross-sectional regression estimate of the expression described in the previous section yielded the following result:

(1) (2) ln ⁄ 0.3579221 0.4477899 ln(FS) -0.0097367 Const. 0.7239196 * 0.7747666 R-squared 0.093 0.1476 Adj. R-squared 0.0607 0.0766 n 30 27

The interpretation of the first regression is that the larger the relative deviation of a firm's stock price from the market-wide average price, ( ⁄ , the larger the size of stock split ( ). A cross-sectional regression estimate in log form yielded the following result:

( ) ⁄ .

(6.66) (1.69)

(t-values in parentheses below the coefficients)

The interpretation of the second regression is that the larger the relative deviation of a firm's stock price from the market-wide average price, ( ⁄ , the larger the size of stock split

( ) and the larger the relative equity of the firms (FS), the smaller the size of stock split ( ). A cross-sectional regression estimate in log form yielded the following result:

( ) ⁄ .

(1.13) (2.03) (-0.12)

(t-values in parentheses below the coefficients)

*

Thus, the market-wide price average indeed appears to be a target for the size of stock splits, and the ratio of the firm's own presplit stock price to the market average "explains" about one third of the cross-sectional variability in the split factors in the first regression. When the natural logarithm of the Firm Size is added to the regression the ratio of the firm's own presplit stock price to the market average "explains" almost half of the cross-sectional variability in the split factors. A one percent deviation of the stock price from the market-wide average price is associated with a (1) 0.3579221 percent and a (2) 0.447789 increase in the size of the split. The natural logarithm of Firm Size has a negative effect on the cross-sectional variability in the split factors. The larger the relative Firms size, the smaller the stock split.

The t-value of is weak in both regressions. I cannot conclude with a significance of 5 percent that this value is different from zero. In the first regression I can conclude with a significance of 10.1 percent that this value is different from zero. In the second regression I can conclude that with a significance of 5.4 percent that this value is different from zero. The reason for this marginal significance is that the observations are not enough. Researchers that want to conclude with a significance of 5 percent should increase the observations by

increasing the used period for this study.

In this paper stock splits in The Netherlands are studied. This paper used the data from 1987 till 2012 because that was all the available data. In this analysis I focus on how firms choose the split factor to split their stocks. A lot of research has been done on this topic, but most studies used the U.S. stock market to research the split factor. This study extends this to the Netherlands.

V.

Conclusion

In this paper I have conducted a study of the Dutch stock market. My analysis of the cross-sectional distribution of the split factor provides support for the “optimal price range” hypothesis in the stock split sample of the Netherlands between 1987 and 2012. This study finds that the coefficients of the relative price of the pre-split price to the market average price are positive.

Thus, the market-wide price average indeed appears to be a target for the size of stock splits, and the ratio of the firm's own presplit stock price to the market average "explains" about one third of the cross-sectional variability in the split factors. A one percent deviation of the stock price from the market-wide average price is associated with a 0.3579221 percent increase in the size of the split and a 0.447789 percent increase in the size of the split when the control variable Firm Size is added.

VI. References

Angel (1997). Tick Size, Share Prices, and Stock Splits. The Journal of Finance, 2, 655-681. Baker, H.K. & G.E. Powell (1993). Further evidence on managerial motives for stock splits.

Quarterly Journal of Business and Economics, 32, 20-31.

Bar-Yosef, S., and L. Brown (1977). A Reex ami nation of Stock Splits Using Moving Betas.

The Journal of Finance, 32, 1069-1080.

Brennan, M.J., and T.E. Copeland (1988). Beta Changes Around Stock Splits: A Note.

Journal of Finance, 34, 1009-1014.

Brown, S. & J. Warner (1980). Measuring security price performance. Journal of Financial

Economics, 8, 205-258.

Dravid, A. (1988). Effect of Bid- Ask Spreads and Stock Price Discreteness on Distributions of Stock Returns. Stanford University, working paper.

Fama, E. (1970). Efficient capital markets: A review of theory and empirical work. Journal of

Finance, 25, 383-417.

Fama, E., L. Fischer, M. Jensen & R. Roll (1969). The adjustment of stock prices to new information. International Economic Review, 10, 1-21.

Kent, H. Baker and Gary E. Powell (1993). Further Evidence on Managerial Motives for Stock Splits. Quarterly Journal of Business and Economics, 32, 20-31.

Lakonishok, J., and Lev, B. (1987). “Stock Splits and Stock Dividends: Why, Who, and When.” Journal of Finance. 49, 913-32.

Lamoureux, C, and P. Poon (1987). The Market Reaction to Stock Splits. Journal of Finance,

42, 1347-1370.

Ohlson, J.A., and S.H. Penman (1985). Volatility Increases Subsequent to Stock Splits.

Schrama, R.J.M., H. G. Eijgenhuijsen (2006). Aandelensplitsingen in Nederland. Economisch

Statistische Berichten, 91, 250-252.

Wu, L. & Chan, B. (1997). On Existence of An “Optimal Stock Price”: Evidence from Stock Splits and Reverse Stock Splits in Hong Kong. International Journal of business, 2,

45- 70.