Firm diversification and equity liquidity

Firm diversification and equity liquidity

Manuel Kampman∗

October 2008

Abstract:

This paper examines the relation between firm diversification and equity liquidity. By acknowledging that the degree of adverse selection differs between diversified and focused firms this link is established. It is theorized that diversification both increases and decreases equity liquidity. The accompanying empirical evidence also shows mixed results. This study provides additional empirical evidence by means of panel analysis. Similar to Gilson at el. (2003) and Krishnaswami and Subramaniam (1999) a positive relationship between diversification and equity liquidity is found. In addition, this study shows that in comparison to 2001 the diversification effect was stronger in 2002 and 2003 and weaker in the following years. However, due to low levels of significance, results should be interpreted with caution. The positive relation between diversification and equity liquidity does not provide an explanation for the recent tendency to increase firm focus. Compared to diversified firms, focused firms have less liquid securities.

Keywords: diversification, equity liquidity, bid-ask spread

Jel Codes: G10, G34

∗

∗∗ ∗

- 1 -

After the seminal paper of Demsetz (1968), research on the determinants of financial market liquidity is still of importance. In the real world, markets do not operate without costs and frictions. An important characteristic that investors look for is liquidity. Liquidity is the ability to buy or sell significant quantities of a security quickly, anonymously, and with relatively little price impact (Gregoriou et al. (2005)). Equity liquidity is important for several reasons. For individual firms, increased equity liquidity facilitates trade and price forming which makes stocks more attractive as investment opportunity. In addition, illiquid stocks require higher rates of return compared to more liquid stocks. This will increase the firms` cost of capital as investors are faced with further trading frictions and transaction costs (Amihud and Mendelson (1986)).

A number of studies have looked at firm characteristics that determine (or at least correlate with) equity liquidity. For instance, Loeb (1983) found that equity liquidity increases with market capitalization. In Huang and Stoll (1987) it is shown that more frequently traded stocks are more liquid compared to less frequently traded stocks. Copeland and Galai (1983) and McInish and Wood (1992) suggested that equity liquidity increases with stock volatility.

One of these firm characteristics is the level of firm diversification. A diversified firm constitutes of multiple business segments which can be clearly be distinguished with respect to economic activities and/or risk profile. In the last decades there has been the tendency of firms to focus on their core business and divest non-core business units, decreasing the level of diversification1. Several theoretical and empirical studies (e.g. Hudson and MacKinnon (2003), Chang and Yu (2004) and Lipson et al. (2007)) have stressed the link between diversification and equity liquidity. These models state that diversification influences equity liquidity through the adverse selection problem.

1

- 2 -

The idea is that uninformed market makers cannot distinguish between the types of trader (liquidity or information motivated). As a result, the market maker increases the bid-ask spread (commonly used measure of liquidity) to compensate for expected losses due to adverse selection problems. As will be illustrated in section two, it is theoretically and empirically uncertain whether diversification increases or decreases equity liquidity. Given there is no theoretical and empirical consensus, this paper aims to provide additional evidence on the relation between firm diversification and equity liquidity. Therefore, the central question in this paper is:

Does firm diversification influence equity liquidity ?

This paper contributes to the existing literature in several ways. First, additional empirical evidence is provided. Since the existing empirical evidence is mixed, this is especially of interest. Second, the results are obtained using a new dataset consisting of 99 Dutch firms (whereas other studies use predominantly US data). In addition, this study uses data from recent years ranging from 2001 up to and including 2007. These years include both bull and bear markets. Third, an alternative approach is used to capture the effect of diversification on equity liquidity. By using panel analysis, several statistical problems can be addressed. For instance, other studies dealing with this issue (e.g. Welker (1995) and Heflin et al. (2005)) face endogeneity problems. It is likely that firm diversification is also prone to endogeneity issues. Unobserved effects such as differences in management, strategy and ownership concentration are likely to be correlated with diversification (specification bias). In addition, diversification could partly be determined by equity liquidity (simultaneity bias). By adopting a fixed effects panel model (as opposed to cross-section analysis) these endogeneity problems are reduced.

- 3 -

In the first part of this section the different components of the bid-ask spread are introduced. Subsequently, the relative importance of adverse selection component is discussed because it is theorized that diversification and equity liquidity are connected through adverse selection. Next, the theoretical arguments are presented and empirical evidence is reviewed.

2.1 Equity liquidity and its components

Many exchanges use market makers to maintain liquidity. Securities are bought at the bid price and sold at a (higher) ask price, the difference between the two represents the bid-ask spread. This is basically the compensation for the market maker for providing liquidity. The bid-ask spread as proposed by Demsetz (1968) is often used to represent the degree of liquidity of a firms’ security2. The spread can be divided into three main components: order processing, inventory holding and adverse selection costs. Order processing costs are associated with providing the market making service and include items such as computer costs, labor costs and floor space rent. Inventory holding costs arise because market makers face the risk of price changes of stock in inventory. Moreover, the market maker is confronted with opportunity cost if trades cannot be made as the result of illiquidity. As described by Kunmar (2004), the market maker seeks to create an optimal inventory level by balancing the inventory carrying costs (excess inventory) against opportunity cost of sales (inventory shortage). As first put forward by Bagehot (1971) and formalized by Copeland and Galai (1983) and Glosten and Milgrom (1985), adverse selection costs arise when the market maker cannot distinguish ex ante between the types of traders. It is theorized that market makers face two types of traders, namely liquidity and information based traders. Informed traders possess non-public information and will buy (sell) at the bid (ask) price if they have information justifying a lower (higher) price. Liquidity traders have no special information and trade for liquidity purposes only. The market maker optimizes his position by setting spreads that maximize the difference between the

2

- 4 -

revenues earned from liquidity motivated traders and expected losses from the informed traders (Coughenour and Shastri (1999)).

Several studies investigated the relative importance of the three bid-ask spread components3. These studies found that the adverse selection costs component is most likely to constitute a significant part of the bid-ask spread (see Appendix A). For instance, Stoll (1989) demonstrates that 43 percent of the spread can be attributed to adverse selection costs where inventory costs account for 47 percent of the total spread. The remaining 10 percent can be attributable to inventory costs. The study of Kim and Ogden (1996) reports similar percentages, here adverse selection costs consist of approximately 50 percent of the total spread. Affleck-Graves et al. (1994) show that adverse selection costs range from 50 to 10 percent depending on the sample and model specification used.

2.2 Theoretical models of diversification and adverse selection

This subsection discusses the theoretical models that link diversification and adverse selection. The models of Subramanyam (1991), Gorton and Pennacchi (1993) and Hadlock et al. (2001) demonstrate that diversification decreases the adverse selection problem. Habib et al. (1997) and Nanda and Narayanan (1999) show that this does not have to be the case, and present a model in which diversification increases the adverse selection problem.

With respect to the adverse selection problem, Subramanyam (1991) and Gorton and Pennacchi (1993) examine whether it is optimal to trade securities on a stand-alone basis or in a basket of securities. There are neutral market makers operating in a speculative market for multiple securities and expect to earn zero profits. Market makers deal with liquidity and information motivated traders. Information based traders possess superior information and can profit at the expense of the liquidity motivated traders. Furthermore, it is assumed that information motivated_traders only have superior information about individual securities.

3

- 5 -

Subramanyam_(1991) demonstrates that when orders are submitted by informed traders to a basket of securities, these orders tend to offset each other (as traders are better informed about individual securities only). Gorton and Pennacchi (1993) obtained a similar result by illustrating that it is optimal for liquidity traders to trade in a basket of securities (as opposed to individual securities). This reduces the information advantage for insiders, minimizing losses for liquidity traders. Based on this result, Gorton and Pennacchi (1993) argue that information based traders will prefer to trade in individual securities whereas liquidity traders prefer to trade in baskets of securities. Note that the arguments are presented in the context trading securities. As will be illustrated next, Hadlock et al. (2001) argue that these arguments also hold in the context of firm diversification (as diversified firms are in essence a basket of focused firms). In that perspective Subramanyam (1991) and Gorton and Pennacchi (1993) present two arguments that are also applicable in a corporate diversification context. First, private information is diversified away when firms have multiple business divisions. Second, the market maker is able to identify the type of trader with more accuracy as liquidity traders prefer to trade in diversified firms. The market maker will change the bid-ask spread accordingly.

As described, Hadlock et al. (2001) directly present their arguments in a corporate-diversification context. The logic of trading in a basket of securities is used to interpret the differences between focused and diversified firms. In this way the model can assert under which conditions diversification reduces the adverse selection problem. The argument put forward by Hadlock et al. (2001) is that compared to focused firms, the market makes more errors in valuing diversified firms. If every market participant make more mistakes in valuing diversified firms, the overall error will be smaller as errors tend to cancel each other out. Consequently, the adverse selection problem is diversified away because the information advantage of informed traders decreases in value.

- 6 -

(each with a separate stock market listing) the price system becomes more informative4. Compared to diversified firms, focused (spun-off) firms have a more informative price system which reduces the uninformed investors’ uncertainty about firm value. Less uncertainty reduces the adverse selection problem as non-public information decreases in value. The opposite holds for diversified firms where the price mechanism is less informative. In a similar line of reasoning, Nanda and Narayanan (1999) present a model with a (diversified) firm consisting of two divisions. Each divisions` cash flow depends on the “quality” of the division. It is assumed that the market can only observe aggregate cash flows where management is in the position to obtain information about the cash flow of individual divisions. This creates an adverse selection problem because market makers cannot distinguish between the (informed) management and the (uninformed) market. When each division is able to trade independently (i.e. becomes more focused), price discovery is facilitated and the adverse selection problem is diminished. In both models there is a transfer of information from informed to uninformed investors.

2.4 Empirical evidence

In this subsection the empirical evidence is reviewed, an overview is presented in Table 1. Recall that it is theorized that the degree of diversification influences the level of adverse selection which in turn changes equity liquidity. Each study uses a different indicator to establish the effect of diversification in this relation. For instance, Hudson and MacKinnon (2003) and Lipson and Mortal (2006) use the bid-ask spread as proxy for liquidity, where Clarke et al. (2004) estimate the adverse selection component of the bid-ask spread which is subsequently used to establish the effect of diversification. In Krishnaswami and Subramaniam (1999), Thomas (2002) and Gilson et al. (2003), the behavior and features of financial analysts are used to proxy the degree of information asymmetry. Lower asymmetric information problems indicate a more effective price mechanism which reduces adverse selection problems. The bottom line is that the mixed empirical findings could be the result of the kind of indicator used.

4

- 7 -

Table 1 – Overview of Empirical Literature

In this Table an overview the empirical literature is given. Each study is described with respect to the research methodology, the data period used and the characteristics of the observations. In the last column (findings) the established effect of diversification on equity liquidity is shown.

Study Methodology Data Period Observations Findings

Hadlock et al. (2001) Equity offerings, comparing

abnormal returns 1983 - 1994 217 diversifying and 424 focusing equity offerings Increase Clarke et al. (2004)

Comparison between diversified and (a portfolio of) focused firms with respect to information asymmetry

1993 - 1997 2,225 diversified and

4,596 focused firms Increase

Hudson and MacKinnon (2003)

Corporate spin-offs, effect on the

information environment 1984 - 1994

59 focusing and 25

non-focusing spin-offs Increase

Gilson et al. (2003)

Corporate breakups, investigating changes in quantity and quality of financial analysts

1990 - 1995 103 breakups Decrease

Krishnaswami and Subramaniam (1999)

Corporate spin-offs, investigating

changes in analyst’ forecast errors 1979 - 1993 118 spin-offs Decrease

Thomas (2002)

Comparison between diversified and (a portfolio of) focused firms with respect to analyst’ forecast errors

1985 - 1993 3,814 diversified and

8,468 focused firms No effect

Lipson and Mortal (2006)

Comparing diversifying and focusing M&As with respect to equity liquidity

1993 – 2003 339 diversifying and

124 focusing M&As No effect

The last column in Table 1 shows that the empirical results are mixed with respect to the effect of diversification on equity liquidity. Each paper will be discussed in more detail.

The papers of Hadlock et al. (2001) and Clarke et al. (2004) and Hudson and MacKinnon (2003) find that diversified firms face less information asymmetry problems. As a result, the adverse selection problem is reduced as there are less informational differences between the information (managers) and liquidity traders (shareholders). Each paper uses a different approach to assess information asymmetry problem which are described next.

- 8 -

abnormal returns compared to focused firms because the information asymmetry between managers and shareholders is smaller. Hadlock et al. (2001) show that diversified firms have less negative announcement effects compared to focused firms. Even after controlling for other variables that are known to influence abnormal stock returns, this relation holds. This indicates that diversified firms face less information asymmetry problems between managers and shareholders. In Clarke et al. (2004) results are obtained by comparing the degree of asymmetric information of a diversified firm to a portfolio of individual firms that resembles the diversified firm in several dimensions (e.g. scale, stock price, line of activities). Both the adverse selection component of the bid-ask spread and the price impact are used to proxy the degree of asymmetric information. The price impact (change in price due to trade) is used because it is a function of the market liquidity: a large price impact signals an illiquid market (Damodaran (2005)). Both measures indicate that diversified firms have lower asymmetric information costs. However, the results should be interpreted with caution as they are sensitive to changes in the specification of the model. In Hudson and MacKinnon (2003) it is argued that a spin-off changes the information environment with respect to the degree of asymmetrical information between liquidity and information motivated traders. Results are obtained by comparing the degree of information asymmetry between focusing and non-focusing spin-offs before and after a spin-off5. Evidence is twofold. First, it is found that after a focusing spin-off, the average and median residual variance (a proxy for the level of asymmetric information) increases significantly. Second, the bid-ask spread increases significantly after a focusing spin-off. Both effects are only observed at focusing spin-offs. This makes these findings consistent with the notion that the information advantage of better informed traders is diversified away.

5

- 9 -

Both Gilson et al. (2003) and Krishnaswami and Subramaniam (1999) find that diversification increases the effectiveness of the price mechanism. As described, a more effective price mechanism reduces the adverse selection problem as the value of private information is reduced. Following stock breakups that eliminate unrelated business lines (making firms less diversified), Gilson et al. (2003) analyze changes in the quantity and quality of financial analysts. The idea is that stock breakups are expected to increase the number of analyst following the firm as new investment banking opportunities are created. In addition, more specialized analysts are attracted as analysts often focus on just one particular field. Hypothetically, a greater number of (more) specialized analysts provide better information which results in more accurate forecast reports. In this way the quantity and quality of financial analysts acts as a proxy for the degree of price informativeness. Gilson et al. (2003) find that both the number (and quality) of analysts increased following a stock breakup. In the first three years the number of analysts increased on average by 45 percent and the accuracy of the analysts’ forecast reports increased by 30 to 50 percent. In Krishnaswami and Subramaniam (1999) both the analysts’ forecast errors and the corresponding standard deviations are used to proxy the informativeness of the price system. These proxies are compared before and after focusing spin-offs. Results show that after completion of a focusing spin-off, asymmetric information problems are significantly reduced indicating a more informative price mechanism. In addition, diversified companies engaging in spin-off have more information problems compared to their industry- and size-matched peers. This could indicate that firms engaging in focusing spin-offs are motivated by the lower levels of information asymmetric problems after the spin-off.

- 10 -

Nothing indicates that diversified firms have higher levels of information asymmetry compared to focused firms. Lipson and Mortal (2006) investigate changes in equity liquidity resulting from mergers and acquisitions (hereafter: M&A)6. By comparing diversifying with non-diversifying M&A, differences in equity liquidity can be attributed to firm diversification. Empirical evidence is twofold. First, by comparing the bid-ask spread before and after the merger (or acquisition), it is found that on average equity liquidity is improved by M&A. Second, Lipson and Mortal (2006) perform a cross-sectional analysis which controls for changing firm characteristics (changes in the number of analyst, number of shareholders and size). The cross-sectional analysis shows there is no remaining effect of M&A on equity liquidity: the difference in equity liquidity is completely captured by the changing firm characteristics.

Next to the different approaches and indicators, the conclusions can to certain extent be influenced by the sample selected. For example, Krishnaswami and Subramaniam (1999) select firms that have chosen to separate their assets whereas Hadlock et al. (2001) include diversified firms which have chosen not to separate their assets. This could reinforce a possible selection bias because in these studies high (low) information asymmetry could be the very reason why these firms are focused (diversified).

3. Methodology

In this section the methodology will be outlined. First, the methodological framework is presented which describes how the diversification effect on equity liquidity is captured. Then, related issues concerning the bid-ask spread are discussed. Subsequently, the different measures of diversification are presented followed by the control variables. The final part of this section presents the functional form.

6

- 11 -

In this paper equity liquidity is captured through transaction costs. As described, the cost of transacting is commonly represented by the bid-ask spread as proposed by Demsetz (1968). Other proxies of liquidity are less suitable in the context of this paper (e.g. the trading volume and firm size) or require data which is not available (for instance to estimate the price impact, detailed trading data is required).

The empirical model is build on the micro structure literature which theorizes that the bid-ask spread is composed of order processing, inventory and adverse selection costs (e.g. Stoll (1989)). The bid-ask spread is regressed to individual firm characteristics that have been found to explain a significant part of the variation in the bid-ask spread. Studies like Atkins and Dyl (1997), Heflin and Shaw (2000) and Menyah and Pudyal (2000) have identified these firm characteristics to be market value (represented by the share price), the risk of a security (often price volatility is used as a proxy), trading volume and firm size. To capture the diversification effect on equity liquidity, the bid-ask spread is (in addition to the control variables) regressed on a proxy for the degree of diversification. This methodology is based on Welker (1995), Heflin et al. (2005), Attig et al. (2006) and Gregoriou et al. (2005). However, these papers use this methodology in another perspective. Three variables are used to proxy for the degree of firm diversification. First, a dummy variables indicating whether a firm is diversified or not. Second, the number of divisions acts like a proxy for firm diversification. Third, an assets based Herfindahl index is used to represent the degree of diversification. These proxies are described in more detail in subsection 3.3. 3.2 Issues concerning the bid-ask spread

- 12 -

the bid price from the ask price. The relative spread is obtained by dividing the quoted spread by the price. The quoted spread can a priory reveal information about the market makers’ estimation of receiving informed orders. However, the quoted spread is often not the actual cost of trading: after observing trade characteristics like order size and volatility, market makers often execute trades lower than the quoted spread, reducing the spread actually paid (Lin et al. (1995) and Huang and Stoll (1997)). Relative spreads (spreads based on executed prices) therefore better represent the actual cost of trading and are generally preferred to absolute spreads because they represent the cost of transacting per traded dollar. Though, Miller and McConnell (1995) demonstrate that the quoted spread is more sensitive to changes in the trading environment (like a change in the degree of the adverse selection problem) whereas the relative spread is more sensitive to changes in the share price. This can be explained by the fact that the bid-ask spread is deflated by the share price to obtain the relative spread. As a result, changes in the share price can alter the relative spread even though equity liquidity has not changed.

Two problems could surface when closing bid-ask spreads are used (Acker et al. (2002)). First, the closing spread could not be representative for the behavior of market makers during the day. Though, this is not of interest in the context of this paper. Second, the (level of the) closing spread is generally not representative for the spread throughout the day. Acker et al. (2002) show that the mean daily spread is an unbiased predictor for the closing spread and is therefore legitimate to use as a proxy for the spread throughout the day7.

3.3 Measures of diversification

A diversified firm is defined as a firm with multiple segments. A segment is defined as part of the firm that can be clearly distinguished with respect to economic activities and/or risk profile. To determine the number of segments in each firm, U.S. Standard Industrial Classification (SIC) codes are used. A SIC code consists of a four digit

7

- 13 -

number assigned to identify a firm based on the type of business. The first two digits represent the main business groups (e.g. manufacturing vs. trade) where the last two digits denote the subgroups (e.g. constructing highways vs. constructing homes). Following Hadlock et al. (2001), Hudson and MacKinnon (2003) and Clarke et al. (2004), a firm is considered to be diversified when the company has segments that differ from each other with respect to the first two digits of the SIC code.

As described, three measures are used to represent the degree of diversification. First, following Lim et al. (2008) and Ruland and Zhou (2005) a dummy variable has the value of one if the firm is active in multiple business segments and zero otherwise. The second variable that proxies the degree of diversification is the number of business segments (see e.g. Lim et al. (2008)). Third, a revenue based Herfindahl index is used which is obtained by computing the sum of squares of each segments’ percentage of total sales (see for instance Berger and Ofek (1995) and Chen and Guo (2005)). The sales based Herfindahl index is represented by

,

where Herfit represents the Herfindahl index of firm i in year t; TSjit symbolizes the total sales in segment j of firm i in year t. Note that a focused firm has a value of one where a firm consisting of two segments, each generating 50 percent of total sales, has the value of a half [(0.5)2+(0.5)2]. The higher the degree of diversification, the lower the sales based Herfindahl index. Papers like Hadlock et al. (2001) and Clarke et al. (2004) use a similar index to proxy the degree firm diversification where each segments’ assets are used instead of sales.

Unfortunately, there is no indicator that perfectly captures the degree of diversification, all proxies have their shortcomings. The dummy variable and the number of business segments both have a significant limitation: differences in segment size are not accounted for. Some business segments might be very insignificant in contrast to other segments and are therefore unlikely to contribute to the creation of asymmetric information problems. To a large extent, this problem is solved by the

- 14 -

sales based Herfindahl index which uses each segments’ sales. However, SIC-codes are also imperfect in determining the degree of diversification: some business units could be closely related to other business segments. For instance, one could say that a firm with a division manufacturing meat processing machines and a division producing airplane materials is fully diversified. Whereas a firm including a division that provides engineering services and a division that engages in drilling activities is less diversified. Therefore, some of the presented arguments will only hold in part when some overlap exists in business activities. In case divisions are closely related, it will be easier for outsiders to assess the proper value of each segment.

3.4 Control Variables

The trading volume, price volatility, share price and firm size are known to control for a large part of the variation in the bid-ask spread. Each of the control variables captures (part of) the order processing and inventory costs.

- 15 -

When controlling for the inventory and order processing costs, the risk of a security, the share price and firm size also capture part of the adverse selection costs. An increase in the riskiness of a security could signal an increased interest of informed traders (Kim and Verrecchia (1994)). However, in the literature (e.g. Hadlock et al. (2001)) it is described that a change in riskiness signals an increased interest by informed traders in the short term (e.g. by the arrival of news). A structurally high level of riskiness (in this paper measured by the yearly average of daily volatility observations) probably has other causes. As described in Gregoriou et al. (2005) and Kim and Verrecchia (1994), the volatility only captures price risk associated with holding the stock, but is does not fully capture the information asymmetry between traders and market markets. Firms with a high share price traditionally enjoy more market depth: the stock is more closely watched by analysts which makes the price more informative. This reduces the potential information asymmetry (Gregoriou et al. (2005)). Firm size indirectly captures part the adverse selection costs (Heflin et al. (2005)). Larger firms may have smaller spreads because there is more information available (e.g. due to more press coverage and greater analyst following).

To sum up, the trading volume, share price and firm size are expected to negatively influence the bid-ask spread. The volatility is expected to positively influence the bid-ask spread.

3.5 Model specification

- 16 -

variables could partly be determined by the dependent variable (simultaneity bias). For instance, the bid-ask spread, trading volume and firm size are all measures of liquidity.

In this paper, panel analysis is used which brings about several advantages8. First, using panel analysis gives more informative data, more degrees of freedom and results in more efficient estimators. Second, a fixed effects panel model is used (as opposed to a random effects model) which partly resolves the mentioned endogeneity problems: it especially helps to avoid the specification bias. A fixed effects model is able to control for unobserved time-invariant variables where a time-series or a cross-section analysis cannot. This is especially relevant because it is believed that several unobserved variables are not included in the model (e.g. differences in firm management, location, strategy, disclosure quality, ownership concentration). An additional advantage of the fixed effects model is that it is easier to interpreted unit effects. However, there are also disadvantages. Fixed effects models are not able to include variables that are constant over time (because they are perfectly collinear with the fixed effects), and if variables vary only slightly over time, the effect will be hard to estimate precisely. To a certain extent, this problem is also present in this study because the selected companies are fairly stable (and in some cases constant) over time with respect to the degree of diversification. To allow for these time-invariant variables to be included in the model, the degree of diversification is interacted with a year dummy. In this way the model provides additional information about changes over time (in comparison to a base year).

The main part of the analysis consists of a panel analysis. However, using a fixed effects model gives rise to several statistical problems (discussed above). Therefore the model is first estimated using a pooled regression specification. The main advantage of using a pooled regression model is that it only raises minor statistical complications. It is assumed that the dataset consists of independent sample observations. In this way, correlations in the error term for different observations can be ruled out. However, this is also the weakness of pooling cross-sections as the obtained observations most likely dependent on each other. Therefore results should

8

- 17 -

be interpreted with caution. An additional advantage is that the pooled regression model does not eliminates time-invariant variables (as opposed to the fixed effects model).

As described in subsection 3.1, the bid-ask spread is regressed on a set of control variables, year and interaction dummies and a measure of diversification. This leading to the following structural form:

     


 !_" _  !_$" #_$% _& _' (_,

where Spreadit represents the yearly average of the daily closing bid-ask spread (€) for stock i in year t; Priceit is the yearly average share price (€) for stock i in year t; Volit symbolizes the yearly average of the daily volatility of stock i in year t; TrdVolit represents the yearly average of the number of daily transactions for stock i in year t;

Sizeit is the average firms size measured in total assets (€) for stock i in year t; Divit represents one of the three measures of firm diversification described in subsection 3.3 for stock i in year t. The year dummy for year t is represented Yeart. Note that similar to Welker (1995), Heflin et al. (2005) and Attig et al. (2006) the simple yearly average is used for both the bid-ask spread and the control variables. In this way these variables match the time interval of the degree of diversification which is measured on a yearly basis.

4. Data

4.1 Data collection

The dataset consist of firms that are located in a single country and are quoted on a single exchange index. This is important as these factors are known to influence equity liquidity. For instance, Frost et al. (2006) demonstrate that the liquidity of the market can differ substantially between indexes and countries. Differences in the accounting system, corporate governance rules, investors protection, legal system (common vs. civil law) and stock exchange size determine a large part of the differences in liquidity.

- 18 -

The final dataset consist of 99 Dutch firms which all have a quotation on the Amsterdam Stock Exchange (Euronext)9. There are no discontinuations and no missing observations. Data is available from 2001 up to and including 2007 which adds up to a total of 693 observations.

Daily stock prices, bid prices, ask prices, trading volumes are all collected from DataStream. The Amadeus database is used to obtain the U.S. SIC-codes for each firm. According to the SIC-codes classification, the annual reports are used to obtain each years’ sales.

4.2 Descriptive statistics

Table 2 presents the summary statistics for diversified firms (Panel A) and focused firms (Panel B). Each panel includes summary statistics for the aggregate sample and for the individual years.

The sample contains 99 firms of which 32 diversified firms (in 2007 this number is 31). Diversified firms have on average two and a half divisions, with a maximum of four (not shown). In the period under consideration just one firm changed from a multiple division to a single division firm. However, there are also firms that changed with respect to the number of divisions but remained diversified. From the annual reports it is possible to capture the cause of each change: there are six divestures, two acquisitions and two reorganizations. These changes are not concentrated in particular years or particular sectors and are expected not to interfere with the results. Note that because of the limited changes in the division structure, it is not possible make a an analysis as in Hudson and MacKinnon (2003) or Lipson and Mortal (2006) without producing unreliable results.

9

- 19 -

Table 2 – Summary Statistics

This Table contains summary statistics for diversified firms (Panel A) and focused firms (Panel B) with respect to the price (€), total assets (million €), price volatility, trading volume and the Herfindahl index. Both panel A and B begin with the aggregate of all available years, followed by summary statistics for each individual year ranging from 2001 up to and including 2007. An (*) indicates that the null hypothesis of no skewness and kurtosis is rejected at the 5% level. Tests used for skewness and kurtosis are those proposed by Jacque and Bera (1980).

Panel A:diversified firms Panel B: focused firms

All ‘01 ‘02 ‘03 ‘04 ‘05 ‘06 ‘07 All ‘01 ‘02 ‘03 ‘04 ‘05 ‘06 ‘07 Price (€) Mean 18.1 20.1 16.2 11.9 13.4 16.8 21.5 26.9 16.4 16.3 12.7 10.8 13.5 16.9 20.9 23.9 Median 16.9 15.7 13.9 10.1 11.4 15.1 20.1 27.1 12.1 12.9 10.6 7.9 10.7 12.5 16.6 18.4 St. Dev. 11.9 15.8 11.4 8.2 8.1 8.7 10.5 12.3 15.9 12.8 11.0 10.4 12.5 15.9 18.9 22.1 Min. 0.6 2.4 1.1 0.6 1.5 1.7 2.1 5.6 0.3 0.5 0.3 0.4 0.3 0.3 0.7 0.4 Max. 73.9 73.9 49.2 31.6 35.2 36.8 43.6 57.4 109.5 54.0 53.8 53.9 67.7 82.3 92.7 109.5 Skewness 1.1* 4.5* 0.9* 0.7* 0.9* 0.4* 0.2* 0.5* 2.1* 1.1* 1.6* 2.1* 2.0* 2.0 1.8 1.6 Kurtosis 5.1* 5.4 3.8* 2.6* 3.5* 2.7* 2.6* 3.8* 9.2* 3.2* 5.9 8.1 8.1 7.8 6.8 6.1 N 223 32 32 32 32 32 32 31 470 67 67 67 67 67 67 68

Total Assets (million €)

- 20 -

Table 2 – Summary Statistics (continued)

Panel B: diversified firms Panel A: focussed firms

- 21 -

Table 2 shows that diversified firms are on average considerably larger compared to focused firms: the average (median) diversified firm has an assets total of 37.8 (0.8) million (€), focused firms have an average (median) assets total of 6.2 (0.3) million (€). Diversified firms enjoy higher average (median) trading volumes compared to focused firms. Diversified firms have an average (median) trading volume of 1805 (63) versus 705 (40) for focused firms. This does not come as a surprise because on average diversified firms are larger. As described in the literature review, large firms enjoy on average greater market depth which manifests in higher trading volumes. The average volatility is lower for diversified firms compared to focused firms (7.8 vs. 11.7) where the median volatility is higher (2.3 vs. 1.7). The standard deviation of the volatility is larger for diversified firms (17.7 vs. 90.1). This implies that within the focused group, there are large variations. Over time, the average share price appears to be influenced by the economic business cycle which holds for both the diversified and focused firms. Growth of the GDP in 2002 and 2003 was close to zero percent where in the following years growth increased to approximately 3.5 percent in 2007 (CBS, 2008). The years 2002 and 2003 could be considered bear markers where the following years could be regarded as bull markets. The total assets increased over time for both panels. The volatility and the trading volume have similar patterns for both diversified and focused firms. The average (median) Herfindahl index is 0.58 (0.56) and is fairly stable over time: firms are relatively stable with respect to the degree of diversification.

- 22 -

Table 3 - Bid-Ask Spreads

This Table contains summary statistics for diversified firms (Panel A) and focused firms (Panel B) with respect to the absolute and relative bid-ask spread. The absolute bid-ask spread is obtained by simply subtracting the bid price from the ask price. The relative spread is obtained by dividing the quoted spread by the share price. Both panel A and B begin with the aggregate of all available years, followed by summary statistics for each individual year ranging from 2001 up to and including 2007. An (*) indicates that the null hypothesis of no skewness and kurtosis is rejected at the 5% level. Tests used for skewness and kurtosis are those proposed by Jacque and Bera (1980).

Panel A:diversified firms Panel B: focused firms

All ‘01 ‘02 ‘03 ‘04 ‘05 ‘06 ‘07 All ‘01 ‘02 ‘03 ‘04 ‘05 ‘06 ‘07

Absolute bid-ask spread (€)

Mean 0.24 0.29 0.52 0.24 0.15 0.14 0.17 0.20 0.28 0.40 0.27 0.22 0.21 0.20 0.20 0.25 Median 0.08 0.12 0.10 0.06 0.06 0.06 0.08 0.09 0.11 0.19 0.12 0.10 0.10 0.10 0.10 0.10 St. Dev. 0.77 0.47 1.83 0.52 0.22 0.20 0.27 0.36 0.41 0.62 0.43 0.30 0.32 0.30 0.30 0.43 Min. 0.01 0.04 0.02 0.01 0.01 0.01 0.01 0.01 0.01 0.02 0.02 0.01 0.01 0.01 0.01 0.01 Max. 10.43 2.28 10.43 2.79 1.07 0.86 1.10 1.93 3.77 3.77 2.51 1.62 1.33 1.86 1.73 2.11 Skewness 10.9* 3.1* 5.2* 4.0* 2.8* 2.9* 2.5* 3.9* 3.9* 3.5* 3.9* 2.8* 2.6* 3.3* 3.2* 3.0* Kurtosis 140.9* 12.6* 28.7* 20.0* 11.0* 8.3* 8.5* 18.9* 23.7* 17.3* 16.0* 11.3* 9.0* 16.3* 14.2* 12.4* N 223 32 32 32 32 32 32 31 470 67 67 67 67 67 67 68

Relative bid-ask spread (€)

- 23 -

The descriptive statistics in Table 2 and 3 indicate that diversified firms enjoy a higher level of equity liquidity compared to focused firms. First, diversified firms are on average larger. In general, the information supply of larger firms is of better-quality (e.g. more press coverage, increased analyst following and regulators interest) which increases the effectiveness of the price mechanism. Second, diversified firms have on average higher trading volumes which signals higher liquidity. Third, the bid-ask is on average lower for diversified firms, which indicates that diversified firms enjoy a higher degree of equity liquidity. In the next section is will be investigated whether these differences between diversified and focused firms can be attributed to firm diversification.

Note that Table 2 and Table 3 also provide information about the presence of skewness and kurtosis. Many variables show signs of non-normality (as indicated by an asteriks). Similar to Gregoriou et al. (2005) and Welker (1995) this problem is minimized by taking the natural logarithm of all variables in the regression (excluding the measures diversification).

Before proceeding to the results, the multicollinearity issue needs to be addressed. In case explanatory variables are highly correlated with each other, estimations could be less precise: standard errors will be higher and t-statistics lower. For most of the explanatory variables presented in Appendix B, low correlations are detected. However, the price volatility (Vol) and the share price (Price) together with the total assets (TA) and the trading volume (TrdVol) show high correlations (0.77) and (0.73). The latter should be taken into account when interpreting results.

5. Results

- 24 -

As described in subsection 3.5, the diversification effect on equity liquidity will be captured using both pooled and panel analysis.

First, each variant of the pooled regression model will be estimated and checked for heteroskedasticity using the Breusch-Pagan test. In case heteroskedasticity is present, a Huber/White error term specification will be used.

- 25 -

To control for heteroskedasticity, a Huber/White/Sandwich error structure will be used10. In case heteroskedasticity is detected in combination with autocorrelation it is hard to control for both simultaneously. It is possible to use an model that uses an AR(1) process, unfortunately this model cannot be used in combination with Huber/White/Sandwich estimates. Consequently, the decision is made that in case both autocorrelation and heteroskedasticity are present the model is only corrected for heteroskedasticity.

5.2.1 Results pooled regression

Table 4 contains the pooled regression results. In regression (1), (2) and (3) the absolute spread is used as dependent variable whereas in regression (4), (5) and (6) the relative bid-ask spread is used. In regression (1) and (4) Dummy is used where regression (2) and (5) use the number of divisions (No. of Div.) as proxy for diversification. Regression (3) and (6) use the Herfindahl index (H-Index). All estimates of diversification find a negative influence of diversification on both the absolute spread and relative bid-ask spread (remember that an increase in the Herfindahl index implies that an increase in firm focus). A negative influence on the spread implies a positive effect on equity liquidity. However, all are highly insignificant.

Year dummies are negative and significant at a 1 percent level. The year 2002 in regression (3) and (6) are an exception, here the year dummy is positive (but insignificant). The interaction dummies show a consistent pattern. Compared to 2002, the diversification effect was higher in 2003 and 2004, where in the following years the effect was lower. Apart from the year 2002 in regression (2) and (4), all variables are insignificant.

With exception of the share price, the control variables all have the expected sign in regression (1), (2) and (3): trading volume (TrdVol) and total assets (TA) are both negative where the price volatility (Vol) has a positive sign. The share price (P) was

10

- 26 -

expected to be negative but has a positive sign. The control variables are significant at a 1 percent level, excluding the price volatility in regression (2) and (3) which is significant on a 5 percent level. In regression (4), (5) and (6) the control variables all have the expected sign. Trading volume (TrdVol), share price (P) and firm size (TA) all have negative signs where price volatility (Vol) has a positive sign. The use of the relative bid-ask spread probably caused the change of the share price (P) (as the spread is deflated by the share price).

Because the proxies that represent diversification could distort the estimates of the interaction dummies, equation (1) till (6) is also estimated without the measure of diversification. Similar results are obtained compared to the equations including the proxies for diversification. The signs of the coefficients are similar and the significance levels are practically identical. Some interaction variables only showed marginal improvements in the significance level. Because results are closely related, these equations are not shown as no additional information is provided.

Note that the pooled results should be interpreted with care. As described, the pooled regression model makes several assumptions (e.g. that observations are independent of each other) that could produce unreliable results: t-statistics are likely to be too high and standard deviations too low.

5.2.2 Results panel analysis

- 27 -

Table 4 – Pooled Regression Results

Estimates in this Table are generated using a linear regression model. Heteroskedasticity is present in all equations, therefore a Huber/White error structure is used. In the left (right) hand of this table, the absolute (relative) bid-ask spread is regressed on the share price Ln(P), the stock price volatility Ln(Vol),the trading volume Ln(TrdVol), firm size Ln(TA), and (Div) to measure diversification. (Div) constitutes of a dummy variable (Dummy) or the number of divisions (No. of. Div.) or the Herfindahl index (H-Index). Coefficients significant at a 10 percent level are marked with a single asterisk (*). Coefficients significant at a 5 percent level are marked with a double asterisk (**), while coefficients significant at a 1 percent level are marked with a triple asterisk (***). Standard errors are in parentheses.

Absolute spread Relative spread

- 28 -

Table 5 – Panel Regression Results

Estimates in this Table are generated using a OLS fixed effects model, to control for heteroskedasticity Huber/White/Sandwich estimates are used. In the left (right) hand of this table, the absolute (relative) bid-ask spread is regressed on the share price Ln(P), the stock price volatility Ln(Vol), the trading volume Ln(TrdVol), firm size Ln(TA), and to a measure of diversification (Div). (Div) constitutes of a dummy variable (Dummy) or the number of divisions

(No. of. Div.) or the Herfindahl index (H-Index). The probability that the null hypothesis of the Hausman (1978) test is

rejected is presented by P-Hausman. Coefficients significant at a 10, 5 and 1 percent level are marked with a single (*), double (**) and triple asterisk (***). Standard errors are in parentheses.

Absolute spread Relative spread

- 29 -

- 30 -

Similar to the pooled regression analysis, the equations are also estimated excluding the proxies for diversification. In this way the measure of diversification does not interfere with the interaction dummies. Again, no significant differences are detected. All coefficients have identical signs and similar levels of significance.

5.3 Interpretation

Regression results indicate that there is a negative relation between diversification and the bid-ask spread. This implies that diversification increases equity liquidity. The regressions show that after controlling for differences in individual firm characteristics (price, size, volatility and trading volume) differences in liquidity can be attributed to the degree of diversification. These finding are in line with the descriptive statistics. Here is was found that diversified firms are on average larger, have higher trading volumes and lower bid-ask spreads which all indicate that diversified firms enjoy a higher level of equity liquidity. In addition, is found that in comparison to 2001 the diversification effect was higher in 2002 and 2003, and lower in the successive years. Although one should keep in mind that the interaction dummies are insignificant in most cases. A possible explanation lies in the relative importance of adverse selection costs in contrast to the inventory and order processing costs. For instance, if due to changing economic conditions the inventory and order processing costs decrease relative to the adverse selection component, the relative importance of the adverse selection component increases. Since the diversification effect manifest itself through the adverse selection component, a change in the relative importance could also imply a change in the relative strength of diversification. The relative impact of the diversification effect constitutes an interesting field for future research.

- 31 -

lower order processing costs (Huang (2004)). Second, corporate governance rules could functioned increasingly better over time. As a result, firms could have been obliged to give more detailed information which lowered the value of non-public information. This might have reduced the adverse selection component of the bid-ask spread.

6. Discussion and conclusions

- 32 -

Unfortunately several caveats are in order. First, data limitations exist. There are only a small number of firms that change with respect to the number of business divisions. An analysis similar to Hudson and MacKinnon (2003) or Lipson and Mortal (2006) would therefore produce unreliable results. These studies only select firms that change to a more diversified/focused firm structure. In addition, the fixed effects model drops all time-invariant observations. This causes problems for the variables that proxy the degree of diversification as these variables are fairly constant over time. This could result in less robust estimators. Second, there is a significant loss of information embedded in the control variables: the yearly average is taken from daily observations to match the degree of diversification (measured on a yearly basis). Third, there could be a self-selectivity bias in firm diversification. The observed level of firm diversification may well be a rational choice, asymmetric information problems could be the very reason why firms are diversified or focused. The self-selectivity bias makes it difficult to determine causation between diversification and equity liquidity.

- 33 -

between diversified and focused firms could subsequently be linked to the degree of diversification. An additional advantage is that changes in relative importance may explain the relative strength of diversification over time. Due to data limitations this technique could not be used in this paper.

- 34 -

Acker, D. and Stalker, M. and Tonks, I. (2002), Daily Closing Inside Spreads and Trading Volumes Around Earnings Announcements, Journal of Business

Finance & Accounting, Vol. 29(9/10): 1149-1179.

Affleck-Graves, J. and Hedge, S.P. and Miller, R.E. (1994), Trading Mechanisms and the Components of the Bid-Ask Spread, The Journal of Finance, Vol. 49(4):

1471-1488.

Affleck-Graves, J. And Callahan, C.M. and Chipalkatti, N. (2002), Earnings Predictability, Information Asymmetry, and Market Liquidity,

Journal of Accounting Research, Vol. 40 (3): 561–583

Amihud, Y. and Mendelson, H. (1986), Asset Pricing and the Bid-Ask Spread,

Journal of Financial Economics, Vol. 17: 223-249.

Atkins, A.B. and Dyl, E.A. (1997), Transaction Costs and Holding Periods for Common Stocks, The Journal of Finance, Vol. 52(1): 309-325.

Attig, N. and Fong, W-M. and Gadhoum, Y. and Lang, L.H.P. (2006), Effects of Large Shareholding on Information Asymmetry and Stock Liquidity, Journal of

Banking and Finance, Vol. (30): 2875-2892.

Bagehot, W. (1971), The only game in town, Financial Analysts Journal, Vol. 27(2): 12-14.

Baltagi, B.H. (1995), Econometric Analysis of Panel Data, Wiley and Sons, Chichester (1995).

Berger, P.G. and Ofek, E. (1995), Diversification’s Effect on Firm Value, Journal

of Financial Economics, Vol. 37: 39-65.

Brooks, R. and Masson, J. (1996), Performance of Stoll’s Spread Component Estimator: Evidence from Simulations, Time-Series, and Cross-Sectional Data, The

Journal of Financial Research, Vol.19 (4): 459-476.

Chang, C. and Yu, X. (2004), Investment Opportunities, Liquidity Premium, and Conglomerate Mergers, Journal of Business, Vol. 77 (1): 45-74.

Chen, H-L. and Guo, R-J. (2005), On Corporate Divestiture, Review of

Quantitative Finance and Accounting, Vol. 24: 399-421.

Clarke, J.E. and Fee, E. and Thomas, S. (2004), Corporate Diversification and Asymmetric Information: Evidence From Stock Market Trading Characteristics,

- 35 -

Comment, R. and Jarrell, G.A. (1995), Corporate Focus and Stock Returns,

Journal of Financial Economics, Vol. 37: 67-87.

Copeland, T.E. and Galai, D. (1983), Information Effects on the Bid-Ask Spread,

The Journal of Finance, Vol. 38 (5): 1457-1469.

Coughenour, Y. and Shastri, (1999), Symposium on Market Microstructure: a Review of Financial Research, The Financial Review, Vol. 34: 1-28.

Damodaran, A. (2005), Marketability and Value: Measuring the Illiquidity Discount, Available at SSRN: http://ssrn.com/abstract=841484.

Demsetz, H. (1968), The Cost of Transacting, Quarterly Journal of Economics, Vol. 83 (3): 33-53.

Drukker, D.M. (2003), Testing for Serial Correlation in Linear Panel-Data Models,

Stata Journal, Vol. 3:168-177.

Frieder, L. and Martell, R. (2006), On Capital Structure and the Liquidity of a Firm’s Stock, Available at SSRN: http://ssrn.com/abstract=880421.

Frost, C.A. and Gordon, E.A. and Andrew, F.H. (2002), Stock Exchange Disclosure and Market Development: An Analysis of 50 International Exchanges, Journal of Accounting Research, Vol. 44(3): 437-483. George, T.J. and Kaul, G. and Nimalendran, M. (1991), Estimation of the

Bid-Ask Spread and its Components: A New Approach, The Review of

Financial Studies, Vol. 4(4): 623-656.

Gilson, S.C. and Healy, P.M. and Noe, C.F. and Palepu, K. G. (2001), Analyst Specialization and Conglomerate Stock Breakups, Journal of Accounting Research, Vol. 39 (3): 565-582.

Giouvris, E. and Philippatos, G. (2008), Determinants of the Components of the Bid-Ask Spread on the London Stock Exchange: The Case of Changes in Trading Regimes, Journal of Money, Investment and Banking, Vol 1: 49-61.

Glosten, L.R. and Harris, L.E. (1988), Estimating the Components of the Bid/Ask Spread, Journal of Financial Economics, Vol.21: 123-142.

Glosten, L.R. and Milgrom, P.R. (1985), Bid, Ask and Transaction Prices in a Specialist Market With Heterogeneously Informed Traders, Journal of

Financial Economics, Vol. 14: 71-100.

- 36 -

Gregoriou, A. and Ioannidis, C. and Skerratt, L. (2005), Information Asymmetry and the Bid-Ask Spread, Journal of Business, Finance and Accounting, Vol. 32 (9): 1801-1826.

Habib, M.A. and Johnsen, D.B. and Naik, N.Y. (1997), Spinoffs and Information,

Journal of Financial Intermediation, Vol. 6: 153-176.

Hadlock, C. And Ryngaert, M. And Thomas, S. (2001), Corporate Structure and Equity Offerings: are there Benefits to Diversification?, Journal of Business, Vol. 71: 613-635.

Harris, L.E. (1994), Minimum Price Variations, Discrete Bid-Ask Spreads, and Quotation Sizes, The Review of Financial Studies, Vol. 7(1): 149-178.

Hausman, J.A. (1978), Specification Tests in Econometrics, Econometrics, Vol. 46: 1251-1271.

Heflin, F. and Shaw, K.W. (2000), Adverse Selection, Inventory-Holding Costs, and Depth, The Journal of Financial Research, Vol. 24(1): 65-82.

Heflin, F. and Shaw, K.W. and Wild, J.J. (2005), Disclosure Policy and Market Liquidity: Impact of Depth Quotes and Order Sizes, Contemporary Accounting

Research, Vol. 22(4): 829-865.

Huang, R.D. and Stoll, H.R. (1997), The Components of the Bid-Ask Spread: A General Approach, The Review of Financial Studies, Vol. 10 (4): 995-1034. Huson, M.R. and MacKinnon, G. (2003), Corporate Spinoffs and Information

Asymmetry Between Investors, Journal of Corporate Finance, Vol. 9: 481-503. Jacquemin, A. and Berry, C. (1979), Entropy Measure of Diversification and

Corporate Growth, The Journal of Industrial Economics, Vol. 27: 359-369. Jarque, C.M. and Bera, A.K. (1980), Efficient Tests for Normality,

Heteroskedasticity and Serial Independence of Regression Residuals, Economic

Letters, Vol. 6: 255-259.

John, K. and Ofek, E. (1995), Assets Sales and Increase in Focus, Journal of

Financial Economics, Vol. 37: 105-126.

Kim, O. and Verrecchia, R.E. (1994), Market Liquidity and Volume Around Earnings Announcement, Journal of Accounting and Economics, Vol. 17: 41- 67.

- 37 -

Kothare, M. and Laux, P.A. (1995), Trading Costs and the Trading System for Nasdaq Stock, Financial Analyst Journal, Vol. 51(2): 42-53.

Krishnaswami, S. and Subramaniam, V. (1999), Information Asymmetry,

Valuation, and the Corporate Spin-Off Decision, Journal of Financial Economics, Vol. 53: 73-112.

Kunmar, P.C. (2004), Bid-Ask Spreads in the U.S. Equity Markets, Quartely

Journal of Business and Economics, Vol. 43: 85-102.

Lakhal, F. (2006), Ownership Structure and Voluntary Earnings Disclosures: The Case of French-Listed Firms, Available at SSRN:

http://ssrn.com/abstract=877709.

Lim, C.Y. and Thong, T.Y. and Ding, D.K. (2008), Firm Diversification and Earnings Management: Evidence from Seasoned Equity Offerings, Review of

Quantitative Finance and Accounting, Vol. 30: 69-92.

Lin, J-C. And Sanger, G.C. and Booth, G.G. (1995), Trade Size and Components of the Bid-Ask Spread, Review of Financial Studies, Vol. 8(4): 1153-1183.

Lipson, M.L. and Mortal, S. (2007), Liquidity and Firm Characteristics: Evidence from Mergers and Acquisitions, Journal of Financial Markets, Vol. 10: 342-361. Loeb, T. F. (1983), Trading Costs: the Critical Link Between Investment

Information and Results, Financial Analyst Journal, Vol. 39: 39-44.

McInish, T.H. and Wood, R.A. (1992), An Analysis of Intraday Patterns in Bid- Ask Spreads for NYSE Stocks, The Journal of Finance, Vol. 47 (2): 753-764. Menyah, K. and Paudyal, K. (2000), The Components of Bid-Ask Spreads in the

London Stock Exchange, Journal of Banking & Finance, Vol. 24: 1767-1785. Miller, J.T. and McConnell, J.J. (1995), Open-Market Share Repurchase

Programs and Bid-Ask Spreads on the NYSE: Implications for Corporate Payout Policy, The Journal of Financial and Qualitative Analysis, Vol. 30(3): 365-382.

Nada, V. and Narayanan, M.P. (1999), Disentangling Value: Financing Needs, Firm Scope, and Divestitures, Journal of Financial Intermediation, Vol. 8: 174-204. Roll, R. (1984), A Simple Implicit Measure of the Effective Bid-Ask Spread in an

Efficient Market, The Journal of Finance, Vol. 39(4): 1127-1139.

Ruland, W. and Zhou, P. (2005), Debt, Diversification, and Valuation, Review of

- 38 -

Stoll, H.R. (1989), Inferring the Components of the Bid-Ask Spread: Theory and Empirical Tests, The Journal of Finance, Vol. 44(1): 115-134.

Stoll, H.R. (2000), Friction, The Journal of Finance, Vol. 55(4): 1479-1514. Subrahmanyam, A. (1991), A Theory of Trading in Stock Index Futures, Review

of Financial Studies, Vol. 4 (1): 17-51.

Thomas, S. E. (2002), Firm Diversification and Asymmetric Information: Evidence From Analysis’ Forecasts and Earnings Announcements, Journal of

Financial Economics,Vol. 64: 373-396.

Wei, P. (1992), Intraday Variations in Trading Activity, Price Variability, and the Bid-Ask Spread, Journal of Financial Research, Vol. 15(3): 265-276.

Welker, M. (1995), Disclosure Policy, Information Asymmetry, and Liquidity in Equity Markets, Contemporary Accounting Research, Vol. 11(2): 801-827. Wooldridge, J. (2002), Econometric Analysis of Cross Section and Panel Data,

- 39 -

Appendix A : Relative Importance Adverse Selection Component.

This appendix presents an overview of studies investigating the relative importance of the bid-ask spread components. The bid-ask spread is divided in adverse selection, inventory and order processing costs. The relative importance of each cost component is presented in percentages. Part of this Table is based on Brooks and Masson (1996).

Study Sample Adverse

selection (%) Inventory (%)

Order**** processing (%)

Stoll (1989)* NASDAQ 43 47 10

George et al. (1991)** NASDAQ – Weekly 30 70 -

George et al. (1991)** NASDAQ - Daily 4 96 -

Wei (1992) – A* NYSE/AMEX 82 18 -

Wei (1992) – B** NYSE/AMEX 23 77 -

Affleck-Graves et al. (1994) – A* NYSE/AMEX 51 1 48

Affleck-Graves et al. (1994) – B* NASDAQ 36 47 17

Affleck-Graves et al. (1994) – C** NYSE/AMEX 29 71 -

Affleck-Graves et al. (1994) – D** NASDAQ 10 90 -

Kim and Ogden (1996)** NYSE/NASDAQ/AMEX 50 50 -

Huang and Stoll (1997)*** NYSE – Daily (20 largest) 21.5 68.9 9.6 Giouvris et al. (2008) – A** FTSE100 – Order driven 67-80 20-33 - Giouvris et al. (2008) – B** FTSE100 – Quote driven 24-32 68-76 - Giouvris et al. (2008) – C** FSTE250 – Order driven 59-66 34-41 - Giouvris et al. (2008) – D** FSTE250 – Quote driven 61-67 33-39 -

*Methodology as proposed by Stoll (1989).

**Methodology as proposed by George et al. (1991) (or based on).

***A combination of the methodologies as proposed by Stoll (1989) and George et al. (1991).

- 40 -

Appendix B: Correlation Matrix of Explanatory Variables.

Ln(P) Ln(Vol) Ln(TrdVol) Ln(TA) Dummy No. Of Div. H-Index

Ln(P) 1.00 Ln(Vol) 0.77 1.00 Ln(TrdVol) 0.09 0.08 1.00 Ln(TA) 0.47 0.34 0.73 1.00 Dummy 0.15 0.10 0.18 0.20 1.00 No. Of Div. 0.16 0.10 0.15 0.19 0.88 1.00 H-Index -0.18 -0.14 -0.19 -0.24 -0.90 -0.89 1.00