Internal Corporate Governance and Stock Market Liquidity:

Internal Corporate Governance and Stock Market Liquidity:

A European Analysis

K.J. van Raan

Double Degree MSc International Financial Management

University of Groningen and University of Uppsala

Supervisor:

Prof. Dr. C.L.M. Hermes

Co-assessor:

Prof. Dr. L.J.R. Scholtens

January 9

, 2015

For helpful comments and discussions I would like to express my sincere gratitude to Prof. Dr. C.L.M. Hermes.

2

2

Abstract

In this paper I examine the relation between internal corporate governance (ICG) and stock market liquidity (hereafter liquidity). Grounded in the principal-agent theory and based on prior literature I hypothesize that ICG positively affects liquidity. I create a corporate governance index, consisting of nine governance standards, to measure the quality of a company’s ICG. As such, I investigate 222 European publicly listed companies that are listed at the FTSEurofirst 300 index over the period 2003-2012. My findings suggest that ICG has a positive effect on liquidity. Yet, the relation is only significant during times of financial turmoil (2008-2012). I also find that the country-specific characteristics law enforcement and stock market development have a negative effect on the link between ICG and liquidity; thereby acting as substitutes for ICG in influencing liquidity. Corruption has a positive impact on the link between ICG and liquidity. Finally, I suggest that the relation between ICG and liquidity is of a contemporaneous nature.

Key words: internal corporate governance, stock market liquidity, law enforcement, corruption, stock

market development, Europe

3

1. Introduction

Over the last two decades the relations between internal corporate governance (ICG) and firm-level outcomes like firm value, cost of debt financing and executive compensation have been investigated extensively3. This stream of research has been growing rapidly mainly due to the establishment of the 2002 Sarbanes-Oxley Act in the United States (US). This piece of corporate governance legislation was introduced to restore the public confidence after major fraud scandals at Enron or WorldCom were brought to light. One year later it was Europe’s turn. Accounting scandals at Royal Ahold and Parmalat caused capital markets to lose faith in the governance practices of numerous external auditors. As a response, the European Commission (EC) issued several recommendations and directives to improve and harmonize corporate governance practices within the European Union (EU). Although vast amounts of literature have been written about ICG, its link with stock market liquidity (hereafter liquidity) has only received limited academic attention. Yet, I argue that liquidity is important for companies as liquidity reduces the company’s cost of capital through lower required expected returns of investors (Amihud and Mendelson, 1986) and lower issuance costs for raising external capital via seasoned equity offerings (SEOs) (Butler, Grullon and Weston, 2005). Hence, this paper sheds light on whether, to what extent and under which circumstances ICG affects liquidity. As such, my aim is to offer valuable insights that are relevant for the academic and business world.

Grounded in the principal-agent theory, Chung, Elder and Kim (2010) maintain that effective ICG improves financial and operational transparency and thereby mitigates information asymmetry between insiders (managers, employees and majority shareholders) and outsiders (creditors, minority shareholders and other stakeholders) as well as among outsiders. Through effective ICG, managers are being monitored better, thereby mitigating opportunistic managers to conceal information from outsiders (Prommin, Jumreornvong and Jiraporn (2014). With respect to the link between ICG and liquidity, Diamond (1985) argues that increased voluntary information disclosure and corporate transparency reduce information asymmetry between managers and market participants4 as well as among market participants. Thereby, the risk beliefs of market participants become more aligned. Through increased voluntary disclosure market participants have fewer incentives to obtain private company information; consequently, there are less well-informed market participants who can employ private company information to their advantage at the expense of other market participants. Market participants therefore face less risk that they do not receive the fair value of the share when trading with other market participants. This decreases the adverse selection costs market participants perceive.

3 _{See La Porta, Lopez-de-Silanes, Shleifer and Vishny (2000), Gompers, Ishii and Metrick (2003), Brown and Caylor (2006),}

Ciceksever, Kale and Ryan (2006), Fahlenbrach (2008) and Bebchuck, Cohen and Ferrell (2009).

4

4 Due to mitigated adverse selection costs market participants offer more liquidity for shares of well-governed organizations (Diamond, 1985; Glosten and Milgrom, 1985; Rindi, 2008).

Within the literature there is a general consensus that ICG positively influences liquidity. Coffee (1991) and Bhidé (1993) argue that shareholders encourage internal corporate processes that increase liquidity, as these processes reduce the adverse selection problems perceived by market participants. However, they do not support their assertions with any empirical evidence. It was not until Chung, Elder and Kim (2010) that the relation between ICG and liquidity was explored empirically. The authors created a corporate governance index (CGI) to measure the effect of transparency-related ICG on liquidity for US companies and find that there is a positive association between ICG and liquidity. These findings are in line with the findings of Tang and Wang (2011), Prasanna and Menon (2012) and Prommin, Jumreornvong and Jiraporn (2014), who all constructed corporate governance indices to study this relation in China, India and Thailand, respectively.

Alternatively, Farooq and Seffar (2012) employed three proxies for ICG to estimate the link between ICG and liquidity in the Middle East and Northern African continent, namely 1) the number of analysts following, 2) the presence of high ownership concentration and 3) the attendance of a Big Four auditor as official external auditor. They find significant results for all proxies of ICG. Additionally, Bar-Yosef and Prencipe (2013) examined the joint effects of ICG, as measured by board independence and chairman-Chief Executive Officer (CEO) separation, and earnings management on liquidity. In their study the authors exclusively investigated Italy and find a positive relation between ICG and liquidity. Nevertheless, since the link between ICG and liquidity, as explored in previous investigations, is possibly affected by distinct institutional and regulatory features of national stock markets, prior findings cannot be generalized to Europe. To the best of my knowledge there are no studies examining the relation between ICG and liquidity on a European level; therefore, this study contributes to the literature by filling this gap.

5 are present in the FTSEurofirst 300 index as of December 31st, 2012. The investigation period concerns the years 2003-2012. Yet, due to data limitations my final sample entails 222 companies.

Although I hypothesize a positive association between ICG and liquidity I also suggest that the context of a country affects the relation between ICG and liquidity. Based on the limitations of Chung, Elder and Kim (2010), Bar-Yosef and Prencipe (2013) and Prommin, Jumreornvong and Jiraporn (2014) I focus on three country-specific characteristics, namely law enforcement, corruption and stock market development. As such, I stipulate whether these country-specific characteristics have a positive or negative effect on the relation between ICG and liquidity.

The results of my study suggest that ICG has a positive effect on liquidity, but only in times of financial distress. I argue that this positive relation is only apparent in times of financial distress, because stronger ICG decreases information asymmetry and thereby takes away (part of the) market uncertainty among market participants that is caused by the sharp drop in stock prices. Since there is less market uncertainty during ‘normal’ economic times I argue that the presence of higher market uncertainty during times of financial distress enhances the importance of ICG in increasing liquidity. The analysis concerning the effect of the separate corporate governance categories on liquidity provides evidence that there is a general association between ICG and liquidity. After checking the robustness of my results I find that the relation between ICG and liquidity is of a contemporaneous nature, i.e. ICG affects liquidity the same year.

With respect to the effect of the country-specific characteristics on the relation between ICG and liquidly I find that law enforcement negatively affects the link between ICG and liquidity. This implies that for countries with a weak law enforcement system ICG plays a more important role in influencing liquidity than for countries with a well-developed law enforcement system, i.e. law enforcement substitutes ICG in affecting liquidity. Corruption has a positive effect on the link between ICG and liquidity, thereby implying that ICG has a larger effect in explaining liquidity for corrupt countries than for non-corrupt countries. Stock market development is negatively related to the relation between ICG and liquidity. As such, for countries that have an underdeveloped stock market ICG is more essential in affecting liquidity than for countries with more advanced stock markets. For this reason, stock market development acts as a substitute for ICG in increasing liquidity.

6

2. Literature Review and Hypothesis Development

2.1. Corporate governance and liquidity defined

In order to avoid misinterpretations concerning corporate governance and liquidity it is important to define these concepts first. According to Mustapha and Ahmad (2011) corporate governance is “a term often used to explain the processes and structures used to direct and manage the business activities of a company in order to enhance its shareholders’ wealth”. In other words, corporate governance refers to a set of mechanisms that establish clear structures concerning accountability, responsibility and transparency at the management of the company. There are two sorts of corporate governance mechanisms, namely internal- and external corporate governance mechanisms. ICG mechanisms are organizational mechanisms such as the functioning and structure of the BOD. Examples of external corporate governance mechanisms are capital market monitoring practices, external auditors as well as a country’s regulatory and institutional system (Huyghebaert and Wang, 2012). In this paper I focus my attention on investigating the effect of ICG on liquidity.

7 2.2. Theoretical relation between ICG and liquidity

From a theoretical perspective the relation between ICG and liquidity can be linked to the principal-agent theory. This theory finds its origin in the US nearly two centuries ago. At that time US railroad companies wanted to expand their business across the country. To finance these expansions US railroad companies needed large investments from other external parties. Together with the growth of US banks this separated wealthy investors from those who were running the day-to-day operations of the railroad company (Steger and Amann, 2008). In this way the evolution of modern corporations led to an increasing separation between ownership and control. Consequently, the principal-agent theory started to become more evident in the academic and business world. Nowadays the principal-agent theory is referred to as the encompassment of two different actors and the link between these actors, namely the principal (shareholder) and the agent (manager) of a company (Berle and Means, 1932; Jensen and Meckling, 1976; Shleifer and Vishny, 1997). As such, the duty of the agent is to maximize the wealth of the principal (Stulz and Wasserfallen, 1995). However, the interests of the agent and the manager are not always the same. Consequently, due to opportunistic behavior managers might act in self-interest, rather than in the interest of the shareholder, thereby resulting into agency problems. In order to better align these interests and therefore maximize the principal’s wealth, so-called agency costs might arise. These agency costs can be minimized through effective ICG via monitoring (e.g. the BOD and the board committees) and/or bonding (e.g. pay-performance compensation programs) mechanisms (Jensen and Meckling, 1976). These days most countries have corporate governance codes that contain recommendations for companies on how to structure its ICG so as improve accountability, responsibility and transparency at the management of the company.

In firms where ICG is weak opportunistic behavior by managers is likely to present itself faster (Jensen and Meckling, 1976). When this is the case, managers are inclined to disclose information selectively so as to conceal wealth expropriation as well as to withhold information concerning financial and/or operational inefficiencies. For instance, managers might unnecessarily increase salary standards and take part in wasteful investments. As a result, corporate transparency is reduced, thereby increasing information asymmetry between insiders and outsiders as well as among outsiders (Chung, Elder and Kim, 2010; Bar-Yosef and Prencipe, 2013). Conversely, companies employ ICG mechanisms (e.g. the BOD and the board committees) to mitigate the management’s incentive and ability to disclose selective information, thereby enhancing financial and operational transparency and mitigating information asymmetry (Leuz, Nanda and Wysocki, 2003).

8 asymmetry, market participants perceive less adverse selection problems and therefore face less adverse selection costs (Diamond, 1985; Glosten and Milgrom, 1985; Rindi, 2008). This is because market participants have fewer opportunities to use specific private company information to their advantage at the expense of the other party. Since more voluntary information disclosure leads to less trading opportunities, market participants face less risk of trading with better informed market participants. In this way there is a higher chance that market participants receive the fair value of a share. As a consequence of this risk reduction market participants offer more liquidity, via a smaller B_A spread, for stocks of corporations with stronger ICG (Bar-Yosef and Prencipe, 2013).

2.3. Prior literature concerning the relation between ICG and liquidity

Within the corporate governance literature there is always the problem of endogeneity. One source of endogeneity is reversed causality. With respect to the relation between ICG and liquidity the implication of reversed causality is that it is unclear whether ICG affects liquidity (ICG is the cause) or is affected by liquidity (ICG is the effect). In this manner there is a two-way relationship between ICG and liquidity. From a theoretical perspective, Coffee (1991) and Bhidé (1993) reason that shareholders encourage ICG processes that increase liquidity. This is because shareholders face lower costs of selling their shares when the shares are liquid rather than illiquid. These selling costs consist out of a) direct trading costs (such as investment commissions and brokerage fees) and b) adverse selection costs (Amihud and Mendelson, 2012). Whereas direct trading costs are relatively stable for each selling transaction adverse selection costs deviate due to the perceived information asymmetry of market participants. Following similar argumentation, increased information asymmetry results into larger B_A spreads, thereby impairing liquidity. Coffee (1991) and Bhidé (1993) argue that it is beneficial for shareholders to promote ICG mechanisms that reduce information asymmetry between market participants as this reduces the costs of selling their shares. Yet, the authors do not support their assertions with any empirical evidence.

9 down, thereby hurting managers who are compensated based on the share price of their company. Therefore, liquidity allows blockholders to facilitate governance via two ways: a) the threat and b) the act of selling shares once blockholders have acquired a stake in the company (Edmans, Fang and Zur, 2013). Overall, the academic world’s perception is that highly liquid stocks progress the well-functioning of capital markets (Bar-Yosef and Prencipe, 2013).

Whereas there is ample and consistent cross-country evidence that external corporate governance mechanisms (such as the regulatory and political systems of a country) positively influence liquidity5, the link between ICG and liquidity has only received limited academic attention. To the best of my knowledge there are no academics that have ever investigated the empirical relation between ICG and liquidity for Europe as a whole. However, I regard analyzing the link between ICG and liquidity in Europe as a fundamental contribution to the corporate finance and market microstructure literature as this could clarify how ICG ultimately affects the wealth of shareholders.

Chung, Elder and Kim (2010) were the first who answered the empirical question whether and how ICG affects liquidity. In their study they investigated firms listed on the NASDAQ and the NYSE, which are both US-based stock exchanges. The authors created a CGI, consisting of 24 transparency-related governance standards, to measure the effect of ICG on liquidity. They find that corporations with better ICG exhibit high liquidity in the form of narrower spreads, higher market quality index, smaller price impact of trades and lower probability of informed trading. In line with Chung, Elder and Kim (2010), Prasanna and Menon (2012) also created a CGI, consisting of 13 governance attributes, to investigate the relation between transparency-related ICG and liquidity. Based on a sample of 100 Indian companies listed at the Bombay Stock Exchange the authors find significant support for a positive relation between ICG and liquidity.

Tang and Wang (2011) studied the Chinese stock market to investigate the effect of ICG on liquidity. The research of Tang and Wang (2011) differs from that of Chung, Elder and Kim (2010) and Prasanna and Menon (2012), because the authors examined the relation between the overall quality of ICG and liquidity, rather than the link between transparency-related ICG and liquidity. In order to do so, the authors built a CGI and classified governance-related company information into five groups, namely expropriation of minority interests, BOD structure and process, supervisory board structure and process, ownership structure and financial transparency and disclosure. Similar to prior findings the authors find a positive and significant relation between ICG and liquidity.

Additionally, by investigating companies from the Middle East and Northern African continent, Farooq and Seffar (2012) reason that ICG has a positive effect on liquidity. Yet, instead of using a CGI to measure ICG, the authors employed three proxies for ICG, namely 1) the number of analysts following, 2) the presence of high ownership concentration and 3) the attendance of a Big Four auditor

10 as official external auditor. They find significant results for all proxies of ICG. Besides, Bar-Yosef and Prencipe (2013) examined the joint effects of ICG, as measured by board independence and chairman-CEO separation, and earnings management on liquidity. In their study the authors investigated Italy and find a positive relation between ICG and liquidity. More recently, Prommin, Jumreornvong and Jiraporn (2014) studied the relation between transparency-related ICG and liquidity in Thailand. After building a CGI the authors find that over time within firms, when the quality of its ICG increases, liquidity also significantly increases.

In short, based on theory and prior literature my reasoning is that ICG positively influences liquidity. This is because strong ICG increases monitoring activities and voluntary information disclosure, thereby improving corporate transparency. Through the increased financial and operational transparency information asymmetry between insiders and outsiders as well as among outsiders is mitigated. Accordingly, due to less trading opportunities market participants face less risk of trading with better informed market participants, thereby facing less adverse selection problems. In turn, mitigated adverse selection costs make the stocks of well-governed organizations more liquid. I translate this logic into my first hypothesis.

Hypothesis 1: ICG has a positive effect on liquidity.

2.4. The effect of country-specific characteristics on the relation between ICG and liquidity

Even though I expect a positive association between ICG and liquidity, its effect might differ across countries due to the country context. For instance, it is possible that country-specific characteristics increase the importance of ICG in influencing liquidity. In this way country-specific characteristics positively affect the relation between ICG and liquidity. Conversely, country-specific characteristics might also have a negative effect on the link between ICG and liquidity, thereby decreasing the importance of ICG in influencing liquidity. As such, country-specific characteristics act as substitute for ICG, thereby making the role of ICG in increasing liquidity more or less obsolete. Another possibility is that country-specific characteristics do not influence the above-mentioned relation. Based on the limitations of Chung, Elder and Kim (2010), Bar-Yosef and Prencipe (2013) and Prommin, Jumreornvong and Jiraporn (2014) I arrive at three country-specific characteristics that I think might influence the link between ICG and liquidity, namely the law enforcement, corruption and stock market development.

11 rationale when scrutinizing the study of Chung, Elder and Kim (2010). In their study the authors exclusively examine US firms. Taking into consideration the significant difference between the legal system of the US and those of European countries, it is unknown whether the positive relation between ICG and liquidity for US firms also holds for European firms.

One important determinant of investor protection is the quality of law enforcement (La Porta, Lopez-de-Silanes and Shleifer, 1998). When a country effectively enforces the rights of investors, minority investors are better protected against expropriation by insiders. In this case opportunistic behavior by insiders is mitigated, thereby decreasing agency problems and increasing the company’s financial and operational transparency (Chung, Elder and Kim, 2010). This in turn reduces information asymmetry between insiders and outsiders as well as among outsiders (La Porta, Lopez-de-Silanes, Shleifer and Vishny, 2000). Following similar argumentation, when the risk of adverse selection declines, market participants offer more liquidity (via lower B_A spreads) to stocks traded at stock markets that are associated with a higher quality of law enforcement. Just like ICG, law enforcement prevents expropriation of outsiders by insiders as well as decreases information asymmetry. As such, law enforcement negatively affects the relation between ICG and liquidity, thereby acting as substitute for ICG in influencing liquidity. This negative effect implies that the effect of ICG on liquidity is high when law enforcement is low, but low when law enforcement is high. This suggestion forms the reasoning for my second hypothesis.

Hypothesis 2: Law enforcement has a negative effect on the relation between ICG and liquidity.

12 argumentation, when the risk of adverse selection increases, market participants offer less liquidity (via higher B_A spreads) to stocks traded at stock markets of corrupt countries. Consequently, in corrupt countries the role of ICG to prevent resource expropriation, to decrease information asymmetry and therefore to increase liquidity becomes more important. It is for this reason that I suggest that corruption within a country has a positive effect on the link between ICG and liquidity. This means that the effect of ICG on liquidity is high when corruption is high and low when corruption is low. I translate this suggestion into my third hypothesis.

Hypothesis 3: Corruption has a positive effect on the relation between ICG and liquidity.

The results of the study of Prommin, Jumreornvong and Jiraporn (2014) find low generalizability on a European level, because the authors exclusively investigated the Thai stock market. They claim that underdeveloped stock markets are not as sophisticated in terms of information efficiency as more advanced stock markets. This is because financial intermediaries, such as analysts and/or (central) banks, provide less company information in underdeveloped stock markets compared to more advanced stock markets. Consequently, investors that are present in underdeveloped stock market depend more on voluntary information disclosed by the company itself. Contrariwise, since European stock markets are better developed, investors have more and better access to voluntary company information, thereby reducing information asymmetry between market participants (La Porta, Lopez-de-Silanes, Shleifer and Vishny, 2000). Following similar argumentation, when the risk of adverse selection declines, market participants offer more liquidity (via lower B_A spreads) to stocks traded at more advanced stock markets. Based on this rationale I suggest that the effect of ICG on liquidity is high when stock market development is low, but low when stock market development is high. I hypothesize that stock market development mitigates the role of ICG in influencing liquidity. This reasoning forms the fundament for my fourth hypothesis.

13

3. Data and Methodology

3.1. Sample

In order to investigate the relation between ICG and liquidity I study a sample of publicly listed companies included in the FTSEurofirst 300 index over the period 2003-2012. As such, I select the companies that are present in this index as of December 31st, 2012. The FTSEurofirst 300 index measures the performance of the 300 largest publicly listed European companies based on market capitalization (FTSE, 2014). Since the companies in this index are a) diversified in terms of country origin and industry and b) stand for circa 70% of Europe’s total market capitalization, I consider the FTSEurofirst 300 index to be a representative index for the whole of Europe6. Using this index allows me to make comparisons between European countries. Sufficient data are available on both company and country-level as the countries present in this index have broad disclosure requirements (Frost, Gordon and Hayes, 2006)7. Since ICG is a slow-moving variable and therefore changes only moderately over time (Bauer, Gunster and Otten, 2004) I use a ten-year investigation period to observe variation in the relation over time. I retrieve the data for ICG, liquidity and control variables from Datastream. I collect daily data as much as possible; however, for some variables only yearly data are available. Accordingly, I convert the daily data into yearly averages for data consistency reasons.

The construction of my sample, as can be seen in Table 1, is as follows. In Datastream I select the companies that are present in the FTSEurofirst 300 index as of December 31st, 2012. These initial 310 companies are divided over seventeen European countries. However, ten firms offer two classes or

6

Refer to Appendix A and B for an overview of the sample distribution classified by country and industry, respectively.

7

In my paper I focus on the effect of voluntary information disclosure on liquidity, rather than the effect of mandatory information disclosure on liquidity. Since most European countries impose disclosure requirements that are based on a comply-or-explain principle, this principle allows companies to differ in terms of voluntary information disclosure and therefore ICG. It is for this reason that I do not consider the broad disclosure requirements to have a large effect on the relation between ICG and liquidity. Refer to section 3.2 for an explanation of the comply-or-explain principle that is applied by most European corporate governance codes.

Table 1 – Sample construction

Table 1 presents the construction of my sample. I employ the FTSEurofirst 300 index, which measures the performance of the 300 largest publicly listed European firms based on market capitalization. This table lists the firms present in the FTSEurofirst 300 index as of December 31st, 2012.

Criterion Number of firms

FTSEurofirst 300 index 310

Firms with two types of share listed in the index - 10

Financial firms - 63

Firms with no available data - 15

14 types of shares that are both listed in this index8. In order to increase comparability, I leave out preferred shares and shares with extra voting rights from my sample. I do this because these types of shares are different equity instruments compared to common shares with normal voting rights; therefore, they do not have the same meaning with respect to liquidity. Based on the sector categorization of the Financial Times (2013) I eliminate firms from the financial sector (being banks, financial services, insurance services and investment trusts). I remove financial firms from my sample, because liquidity has different implications for financial firms than for non-financial firms and is therefore not comparable (Durnev and Kim, 2005). After I observe the data I find that the amount of missing observations is high for research and development (R&D) costs. For data maximization purposes I therefore assume that a company does not have R&D costs when it is not reported in Datastream. Apart from this exception I remove firms from my sample when they do not have any available data for one or more firm-level variables. This leaves me with 222 firms spread out over 16 countries. Due to missing observations my final sample contains 2,086 observations9.

3.2. Internal corporate governance

As mentioned by Harris and Raviv (2006) and Larcker, Richardson and Tuna (2007) there is no well-developed theory explaining corporate governance in a conceptual manner. Moreover, there is no universal model that specifies relevant governance characteristics that are needed to analyze the impact of corporate governance, let alone ICG. Though, in order to measure corporate governance and relate it to other firm-level variables, academics have created corporate governance indices. For instance, Gompers, Ishii and Metrick (2003) have applied an index of 24 firm-specific governance rules as a proxy for shareholder rights to measure the influence of corporate governance on firm valuation. Brown and Caylor (2006) have used 51 firm-specific governance provisions to investigate the effect of corporate governance on firm value. Later, Bebchuck, Cohen and Ferrell (2009) find that only six of the 24 corporate governance variables used by Gompers, Ishii and Metrick (2003) have a significant negative effect on firm value. The other 18 variables are uncorrelated. For a complete overview regarding the use of corporate governance indices see Bozec and Bozec (2012).

A CGI that explains the relation between ICG and liquidity needs governance data that would, in theory, increase financial and operational transparency and voluntary disclosure so that information asymmetry is mitigated. Due to absent theory concerning this relation I create my own CGI. I use the

8

For instance, A.P. Moller-Maersk offers two classes of common shares, namely class ‘A’ and ‘B’. Different voting rights are assigned to these classes, making class ‘A’ shares superior over the class ‘B’ shares. Another example considers Volkswagen. The company has listed two types of shares in the FTSEurofirst 300 index, namely common shares and preferred shares.

9

15 structure and functioning of the BOD and the board committees (audit, compensation and nomination committees) as a proxy for ICG. I employ this proxy because shareholders consider the act of monitoring by the BOD and the board committees as the most suitable mechanism to increase the quality of ICG (Fama and Jensen, 1983; De Andres, Azofra and Lopez, 2005).

Based on the governance standards that are used by Sawicki (2009) to create her CGI, I construct my own CGI. Sawicki (2009)’s governance standards have been used separately and/or jointly in the literature as structural indicators for ICG (Larcker, Richardson and Tuna, 2007). When looking at the governance standards regarding BOD and board committee independence, Sawicki (2009) stipulates these to be independent when at least one-third of the total number of directors is independent. Taking into consideration that the five South-East Asian countries investigated by Sawicki (2009) do not have extensive requirements concerning the level of independent directors within companies, I understand the minimum independence level of 33%. However, with respect to Europe I regard this minimum level to be fairly low for the following reason.

In a reaction to the 2002 Sarbanes-Oxley Act in the US the EC introduced action plans to harmonize corporate governance regulations in the EU. In 2004, the EC published a recommendation stating a number of criteria that companies should employ to increase the amount of independent directors. At the same time the EC provided recommendations concerning the structure of the BOD and the board committees (European Commission, 2004). Two years later the EC issued the 8th Directive, thereby declaring that at least one member of the company’s audit committee has to be independent (European Commission, 2006). I believe that because of these recommendations and directives European companies fulfill the director independence standard of 33% rather easily. Thus, in order to distinguish European companies in terms of the quality of ICG, I consider the BOD and the board committees to be independent when at least 50% of its directors are outside directors. I cannot benchmark this criterion with other ICG-related European studies as there are no European studies that employ board committee independence standards to measure ICG. Yet, this criterion is in line with the US-based investigation of Chung, Elder and Kim (2010).

I set up the CGI so that the governance standards are based on a ‘yes or no’ principle. By applying this principle in constructing my index I increase the objectivity of my study. The ‘yes or no’ principle has been used before in the corporate governance literature10. If a firm fulfills a governance standard one point will be awarded, with a maximum of nine points. The higher the total points, the stronger ICG. For each year I calculate the governance score of a firm. I state the CGI in Table 2.

10

16 The existence of the board committees as well as the expertise of the audit committee plays an essential role in affecting the quality of ICG. The reason is that board committees actively and independently monitor the company’s management so as to ensure the stakeholder’s interests and thereby maximize the shareholder’s wealth (Hermanson, Tompkins, Veliyath and Ye, 2012). Through increased monitoring activities more and better information is publicly disclosed, thereby decreasing information asymmetry between insiders and outsiders (Bar-Yosef and Prencipe, 2013). The existence and expertise of board committees also limit expropriation by insiders from outsiders (Chung, Elder and Kim, 2010). Therefore, I conjecture that the monitoring activities of active, experienced and diligent board committees contribute to financial and operational transparency and therefore stronger ICG.

When looking at the level of independence of the BOD and the board committees, Fama and Jensen (1983) state that independent directors are better able to exercise more stringent monitoring activities than affiliated directors are able to. For instance, independent directors play a more active role in replacing underperforming CEOs and in monitoring the process of collecting and processing financial information (Beasley, 1996). Since independent directors are more critical than affiliated directors this limits the manager’s ability and incentive to conceal wealth expropriation and hold back information regarding financial and operational inefficiencies (Patelli and Prencipe, 2007). Consequently, information asymmetry between insiders and outsiders is mitigated and corporate transparency is increased.

Regarding the separation of the CEO and chairman, Finkelstein and D’Aveni (1994) argue that no separation makes the CEO too powerful. They claim that when there is a dual CEO-chairman there is a higher probability that corporate information is either being withhold, delayed or biased. As a result, no separation negatively influences the disclosure of voluntary company information. Bar-Yosef and

Table 2 – Corporate governance index

Table 2 lists nine governance standards used to construct my corporate governance index. The audit committee possesses financial expertise when at least one financial expert within the meaning of the 8th Directive of the European Commission sits in the audit committee. The audit, compensation and nomination committee are independent when at least 50% of the total number of directors are outside directors.

Category Governance standard

Audit

1. Existence of audit committee

2. Financial expertise of audit committee 3. Independence of audit committee

Compensation 4. Existence of compensation committee 5. Independence of compensation committee

Nomination 6. Existence of nomination committee 7. Independence of nomination committee

Board of Directors 8. Independence of Board of Directors

17 Prencipe (2013) also claim that CEO duality reduces board independence and increases information asymmetry. The presence of a dual CEO-chairman also hinders the chairman to objectively and independently monitor the roles and interests within the BOD and the organization (Ajinkya, Bhojraj and Sengupta, 2005). Therefore, I regard the separation of the CEO and chairman as an essential mechanism to increase the quality of ICG.

One might question whether the governance standards used in my index are required via (national) corporate governance codes. Yet, due to the comply-or-explain principle of most corporate governance codes EU companies are allowed to deviate from the governance standards set in these corporate governance codes (International Finance Corporation, 2008). The comply-or-explain principle entails that publicly listed companies can either comply with recommended governance standards, or if they do not comply, they have to explain why and to what extent they do not. Through the act of publicly disclosing this information, the market decides which set of standards is appropriate for companies to follow by trading their shares accordingly (Enriques and Zetzsche, 2013). This implies that market participants sell their shares when they disagree with the company’s explanation of non-complying, thereby imposing a market sanction rather than a legal sanction (Cromme, 2005). Because the majority of the EU countries have corporate governance codes that are characterized by a comply-or-explain principle, companies still have the possibility to deviate in terms of their corporate governance quality. Next to this, I still observe sufficient variation in the CGI scores as presented in Table 4. It is for these reasons that the nine governance standards used in my index are suitable indicators for the measurement of ICG.

3.3. Liquidity

Brunnermeier (2008) argues there are three sub-forms of liquidity: 1) the B_A spread, which measures how much market participants lose when they buy one unit of an asset and sell it back right away; 2) market depth, which indicates how many market participants can engage in a transaction at the current bid or ask price without moving the price; and 3) market resiliency, which determines how long it takes for bid or ask prices that have temporarily fallen to return to the normal level. Although I regard all sub-forms to be eligible proxies for liquidity I notice that the B_A spread is used predominantly within the market microstructure literature. In this respect I follow the literature.

18 of the times the bid and ask price do not match. The next transaction will therefore be at a higher ask price, resulting in a difference between the bid and ask price. Most stock exchanges around the world work with electronic order-driven systems that match buyers and sellers automatically. At these stock exchanges there are no official market makers, yet they nevertheless exist (Amihud and Mendelson, 1986; Nowak, 2008).

The magnitude of this spread depends on a) the order-processing cost, b) the inventory holding costs and c) the information asymmetry costs (Amihud and Mendelson, 1986; Bar-Yosef and Prencipe, 2013). The last aspect of the spread, referring to the adverse selection risk faced by market participants, has a tendency to become higher when the market participants perceive a greater chance of trading with better informed market participants. Therefore, adverse selection costs reflect the extent of information asymmetry risk perceived by the market participants (Bar-Yosef and Prencipe, 2013). Since it is unclear what part of B_A spread is driven by information asymmetry costs I cannot say that the B_A spread is exclusively linked to adverse selection risk and therefore to ICG. Still I regard the B_A spread the most appropriate measurement that links ICG to liquidity. In this respect, the B_A spread has a negative relation with liquidity; the smaller the B_A spread the higher liquidity and vice versa. In line with Amihud and Mendelson (1986) and Chung, Elder and Kim (2010) I suggest that the wider the B_A spread is the higher the information asymmetry between management and market participants.

I clean the B_A spread data of errors, which is according to Chung, Elder and Kim (2010) common in the market microstructure literature. In order to clean the data I a) delete bid and ask prices if its values are negative and b) remove B_A spreads from my sample if its value is negative. The B_A spread is standardized by the mean of the bid and ask price in order to allow for cross-company comparison and to overcome nonlinearity problems. As it is common in the literature to express the B_A spread in basis points I multiply the B_A spreads by 10,000. The B_A spread is formulated as:

(1)

where is the ask price for stock i at time t, is the bid price for stock i at time t and is the mean of and .

3.4. Control variables

19 I include Size in the research model, because the larger the firm is the higher the availability of information, resulting into smaller adverse selection risks faced by market participants and thus higher liquidity (Harris, 1994). Another reason to include Size is that larger companies generally have a higher quality of ICG mechanisms as a result of greater interest by investors (Chung, Elder and Kim, 2010). With respect to Leverage, Udomsirikul, Jumreornvong and Jiraporn (2011) find that firms with liquid stocks are generally less leveraged than firms with illiquid stocks. This is because higher liquidity reduces the cost of equity and therefore increases the attractiveness of equity relative to debt. According to Amihud and Mendelson (1986) stock returns also influence liquidity. If investors value stocks based on the return net of trading costs, they demand higher expected returns for stocks with wider B_A spreads. Investors want to compensate for the higher cost of trading related to the purchase of illiquid stocks relative to liquid stocks. Therefore I add Return as a control variable to my regression.

Harris (1994) finds that liquidity is also affected by the share price of a company. However, I prefer the reciprocal of share price (1/Price) rather than the share price. The reason is that it better captures the share price effect on liquidity. Due to the minimum price variation between the bid and ask price, the B_A spread for low-priced stocks is relatively large compared to high-priced stocks. Consequently, market participants tend to get more protection for low-priced stocks than for high-priced stocks (Harris, 1994). In order to control for this effect I include 1/Price in my regression. Assets is taken as a control variable in my regression model as the payoffs of tangible assets, rather than intangible assets, are easier to observe by investors. As such, this increases organizational transparency and decreases information asymmetry (Prommin, Jumreornvong and Jiraporn, 2014). In contrast, firms with high R&D expenditures have more asymmetric information problems, as payoffs resulting from R&D expenditures are harder to foresee (Chung, Elder and Kim, 2010). Therefore I include R&D as my final control variable in my regression.

3.5. Country-specific characteristics

I include Enforcement, Corruption and Development in my analysis to measure the impact of law enforcement, corruption and stock market development on the relation between ICG and liquidity. In order to be able to do this I also add the interaction terms between the respective country-specific characteristic and the CGI11.

First of all, I add Enforcement to estimate the law enforcement quality of a country. Here, I follow La Porta, Lopez-de-Silanes and Shleifer (1998). In their research they use rule of law as a proxy for the quality of law enforcement. In order to measure rule of law I employ the Worldwide Governance Indicators. As of 1996, the World Bank publishes six governance dimensions for more than 200 nations around the world. What makes these governance indicators special is that they form the

20 aggregate of 32 individual data sources, thereby constructing holistic governance indicators. With respect to rule of law, it captures the individual’s perception of the degree to which they have confidence in the rules of society. In particular, this governance indicator measures to what extent individuals abide the quality of (contractual) law enforcement. The higher the score for rule of law, the higher the degree to which individuals act according to the law and place trust in them (Kaufmann, Kraay and Mastruzzi, 2010). These indicator scores refer to percentile rank terms ranging from 0 (lowest) to 100 (highest) among all countries worldwide12. I retrieve the data from the World Bank (2014a).

My second country-specific characteristic is Corruption. I make use of the Corruption Perception Index (CPI) of Transparency International to estimate corruption. Whereas in 1995 Transparency International produced its first CPI for only 41 countries, it nowadays it publishes indices for more than 175 countries around the world. Transparency International refers to corruption as the abuse of entrusted power for private gain. The CPI measures corruption as perceived by individuals and covers all sorts of corruption, ranging from small (petty) corruption to grand corruption (Transparency International, 2014). Each country is scored on a scale between 0 and 100, where 0 indicates that a country is perceived to be highly corrupt and 100 indicates very clean. For the years 2003-2012 I obtain the CPI scores for all countries that are present in my sample as published by Transparency International (2014).

I include Development as my third country-specific characteristic. Following prior research13, I use the stock market capitalization to Gross Domestic Product (GDP) ratio as a proxy for stock market development. I employ this ratio because countries with better fundamentals (e.g. a stable macro economy or the requirement of extensive information disclosure) have more advanced stock markets as measured by the stock market capitalization as a percentage of GDP (Djankov, La Porta, Lopez-de Silanes and Shleifer, 2008). Therefore, the higher the stock market capitalization to GDP ratio, the more developed the stock market. I collect the data from the World Bank (2014b).

3.6. Regression models

I start with using a pooled Ordinary Least Squares (OLS) to estimate the effect of ICG on liquidity. By employing panel data, OLS simultaneously captures variation between companies and over time. I add Size, Leverage, Return, 1/Price, Assets and R&D as control variables to isolate the impact of ICG on liquidity. Since Size is skewed to the right and the residuals of Size are not normally distributed14 I take the log of Size. This rescales the control variable by pulling in extreme observations. I formulate my standard regression model as follows.

12

Refer to Kaufmann, Kraay and Mastruzzi (2010) for a detailed explanation concerning the calculation of rule of law scores.

13

See Shleifer and Wolfenzon (2002) and Djankov, La Porta, Lopez-de Silanes and Shleifer (2008).

21

+

(2)

where is the B_A spread for stock i at time t, is the CGI, is the firm’s size, is the leverage ratio, is the stock return, ⁄ is the reciprocal of the share price, is the asset tangibility ratio, is the R&D expenditures ratio and is the error term.

I add Enforcement, Corruption and Development separately to the regression model to estimate the effect of these three country-specific characteristics on the relation between ICG and liquidity. I also include the interaction term between the country-specific characteristic and the Gov_index to the regression model. This interaction term is the product of the country-specific characteristic and the

Gov_index. Since Development is not normally distributed I take the log15. The regression model that

estimates the effect of Enforcement, Corruption and Development is therefore as follows.

+ (3)

where is the B_A spread for stock i at time t, is the CGI, is the firm’s size, is the leverage ratio, is the stock return, ⁄ is the reciprocal of the share price, is the asset tangibility ratio, is the R&D expenditures ratio, is the country’s law enforcement, is the country’s level of corruption, is the country’s stock market development, is the interaction term between and , is the interaction term between and , is interaction term between and and is the error term.

22 3.7. Descriptive statistics

Table 3 demonstrates the study’s descriptive statistics (part A) and the Spearman rank correlation matrix (part B). The mean of the B_A spread is 21.660 basis points and its standard deviation is 36.928 basis points. The average CGI score is 5.927. This figure indicates that for the years 2003-2012 the average European company almost meets six of the nine governance standards. The CGI score range from 0 to 9 where 7 is the median, indicating the distribution is skewed to the right. The average firm size is € 38.259.000.000. After taking the log of firm size the mean becomes 7.307. On average a company is financed by 21.41% debt and 78.59% equity, implying that the companies in my sample rely relatively much on equity financing. Yet, whereas certain (holding) firms in my sample do not have any debt in their capital structure (0%), there are also firms that finance their assets almost exclusively by debt (95.09%). The average stock return is 17.42% and the reciprocal of share price is 0.046. The mean of the asset tangibility ratio and R&D ratio is 0.292 and 0.026 respectively. With respect to the country-specific characteristics, the average score for law enforcement is 92.136, while the average corruption score is 78.835. The distribution of both variables is skewed to the right, indicating that the average European country has a highly developed law enforcement system and clean public sector16. Finally, the stock market development mean is 95.915, indicating that the average stock market capitalization in Europe is nearly as big as the country’s GDP. The log of stock market development is 1.911.

As can be seen in the Spearman rank correlation matrix, Gov_index is negatively correlated to the B_A spread (-0.135). This indicates that firms with strong ICG tend to be more liquid than firms with weak ICG. I observe that there is a low correlation between the country-specific characteristics law enforcement, corruption and stock market development and the control variables leverage ratio, asset tangibility ratio and reciprocal of share price. However, this does not cause any problems for further analysis. As a final point, there is a high and positive correlation between Enforcement and Corruption (0.887), thereby implying that there is a very close link between the two variables. This high level of correlation is of no surprise as both variables measure the same phenomenon. Since I independently measure the effect of law enforcement and corruption on the relation between ICG and liquidity I do not add these variables to the same regression model, thereby preventing any problems concerning multicollinearity.

16

23

Table 3 – Descriptive statistics and Spearman rank correlation matrix

Table 3 shows the descriptive statistics and the Spearman rank correlation matrix. With respect to Panel A, the B_A spread indicates the difference between the ask price and the bid price divided by the mean of the bid and ask price in basis points. Gov_index denotes the corporate governance index and is calculated based on nine governance standards. If a firm fulfills a governance standard one point will be awarded, with a maximum of nine points. For each year I obtain a firm-specific corporate governance score. Size is the book value of total assets in billions of Euros. Leverage is the book value of total debt divided by the sum of the book value of total debt and the market value of total equity. Return is the end of year share price minus the begin of year share price plus yearly dividends paid, divided by the begin of year share price. 1/Price is the reciprocal of the yearly average share price. Assets is the net property, plant and equipment divided by the total assets. R&D is the research and development expenditures divided by total sales. Enforcement is the rule of law World Governance Indicator as measured by Kaufmann, Kraay and Mastruzzi (2010). The rule of law scores are scaled in percentile rank terms ranging from 0 (lowest) to 100 (highest) among all countries worldwide. Corruption is the Corruption Perception Index score as measured by Transparency International (2014). The Corruption Perception Index scores are scaled between 0 and 100, where 0 means the country is perceived to be highly corrupt and 100 means it is perceived to be very clean. Development is the stock market development of a country as measured by stock market capitalization as percentage of the Gross Domestic Product of a country. With respect to Panel B, the Spearman rank correlation matrix illustrates the paired correlation coefficients.

Panel A – Descriptive statistics

Variable Observations Mean Std. dev. Minimum Median Maximum B_A spread 2,086 21.660 36.928 2.933 12.853 1221.283 Gov_index 2,086 5.927 2.518 0.000 7.000 9.000 Size (€ billions) 2,086 38.277 60.128 0.219 16.455 762,903 Log(Size) 2,086 7.307 0.596 5.341 7.353 8.822 Leverage 2,086 0.214 0.208 0.000 0.175 0.951 Return 2,086 0.174 0.377 -0.881 0.149 3.517 1/Price 2,086 0.046 0.010 0.001 0.020 1.543 Assets 2,086 0.292 0.201 0.001 0.244 0.947 R&D 2,086 0.026 0.050 0.000 0.003 0.496 Enforcement 2,086 92.136 7.127 60.287 93.301 100.000 Corruption 2,086 78.835 10.909 39.000 79.000 97.000 Development 2,086 95.915 51.418 15.174 84.539 281.388 Log(Development) 2,086 1.911 0.241 1.181 1.917 2.449

Panel B – Spearman rank correlation matrix

24 Table 4 shows the CGI scores for each country present in my sample. Apart from the overall corporate governance scores the table also presents separate scores for each of the corporate governance categories used to construct my index. I notice that Austria has the highest score (7.1), while Denmark and Germany score the lowest (3.6). What is surprising is that the corporate governance scores for Austria and Germany are relatively far apart, while both countries have German origins. Part of the reason why Denmark and Germany have low overall corporate governance scores is that these countries score low on the board committee categories. For instance, the low score for the nomination committee for Germany can be explained by looking at the German Corporate Governance Code. As of 2007 this Code recommends German companies to form a nomination committee. While the average nomination committee score is 0.3 for the years 2003-2006, this increases to 1.0 for the years 2007-2012. Since the difference between these average scores is significant I suggest that the above mentioned recommendation encourages German companies to indeed establish a nomination committee and/or increase the independence level of the committee.

I observe that French-origin countries such as France, Italy, Spain and Portugal score on average (5.4) lower than English-origin countries like Ireland and the United Kingdom (6.8). However, by looking at the standard deviation of the Gov_index means of the French- and English-origin countries I find that there is no significant difference between these means. When looking at the scores for the separate corporate governance categories I observe similar patterns. Unexpectedly, France and Luxembourg score the highest on the category BOD (1.3), while their overall governance score is below average. It is important to notice that not every country is equally represented in my sample. For example, there are only two Austrian and three Portuguese companies in my sample. One of the implications of this unequal country representation is that relatively few companies drive the corporate governance score of a whole country. When comparing the score of countries this limitation has to be taken into consideration17. Note that Table 4 refers to the average corporate governance score based on the country corporate governance scores. This differs from the Gov_index mean as presented in Table 3, because the latter corporate governance score is the average of the corporate governance scores for each company.

17

25

Table 4 – Corporate governance index scores per country

Table 4 presents the corporate governance index scores per country. Gov_index denotes the corporate governance index and is calculated based on nine governance standards. If a firm fulfills a governance standard one point will be awarded, with a maximum of nine points. For each year I obtain a firm-specific corporate governance score. The categories audit committee, Board of Directors, compensation committee and nomination committee show the corporate governance index scores for the respective category. The maximum score for the audit committee is 3, while the maximum score for the Board of Directors, compensation committee and nomination committee is 2. Average refers to the average corporate governance score, thereby weighing each country’s score equally.

Country Gov_index Audit

26

4. Analysis

4.1. Pooled OLS regression

I use a pooled OLS regression to estimate the effect of ICG on liquidity and show the results in Table 5. It is interesting to see that B_A spread is negatively related to Gov_index at a 1% significance level. Since Gov_index negatively affects B_A spread I suggest that better ICG increases liquidity. This suggestion supports my first hypothesis that ICG positively influences liquidity. The notion that stronger ICG leads to higher liquidity is also supported by prior literature18.

Size relates negatively to the B_A spread at any significance level. This outcome indicates that larger firms have smaller B_A spreads and is consistent with the findings of Harris (1994). Leverage relates positively and significantly to B_A spread at any level. This is in line with the results of Udomsirikul, Jumreornvong and Jiraporn (2011) who find that companies with liquid stocks are leveraged less than companies with illiquid stocks. B_A spread is positively related to Return and 1/Price at 1% significance level. These results are uniform to prior literature19. Yet, Assets and R&D do not significantly relate to B_A spread. Therefore, I argue that the company’s capital spending decisions in property, plant, equipment and R&D do not significantly affect information asymmetry between market participants and thus do not influence liquidity. This result is in line with Chung, Elder and Kim (2010)’s US-based investigation. The regression model includes 2,086 observations and explains 7% of the variation of the B_A spread.

18

See Chung, Elder and Kim (2010), Tang and Wang (2011), Farooq and Seffar (2012), Prasanna and Menon (2012), Bar-Yosef and Prencipe (2013) and Prommin, Jumreornvong and Jiraporn (2014).

19

27

Table 5 – Pooled Ordinary Least Squares regression

Table 5 demonstrates the pooled Ordinary Least Squares regression results of equation (2). B_A spread indicates the difference between the ask price and the bid price divided by the mean of the bid and ask price in basis points. Gov_index denotes the corporate governance index and is estimated based on nine governance standards. If a firm fulfills a governance standard one point will be awarded, with a maximum of nine points. For each year I obtain a firm-specific internal corporate governance score. Size is the book value of total assets in billions of Euros. Leverage is the book value of total debt divided by the sum of the book value of total debt and the market value of total equity. Return is the end of year share price minus the begin of year share price plus yearly dividends paid, divided by the begin of year share price. 1/Price is the reciprocal of the yearly average share price. Assets is the net property, plant and equipment divided by the total assets. R&D is the research and development expenditures divided by total sales. White (1980)’s t-statistics for standard errors are reported in parentheses. ***, ** and * indicate significance at the 1%, 5% and 10% levels, respectively.

B_A spread Intercept 109.582*** (10.73) Gov_index -1.452*** (-4.61) Log(Size) -11.901*** (-8.16) Leverage 15.764*** (3.46) Return 9.862*** (4.65) 1/Price 30.138*** (3.50) Assets 1.801 (0.44) R&D -19.087 (-1.16) R2 0.073 Adjusted R2 0.070 Observations 2,086

4.2. Fixed effects regressions

28 For both fixed and random effects it is possible to allow for cross-sectional variation as well as time variation. The cross-sectional fixed effects regression model controls for omitted variables that differ across firms but are constant over time. In this manner the model allows for different intercept terms for each company that is present in my sample; however, the intercept terms are constant over time. Including cross-sectional fixed effects is the same as including a dummy variable for each company in the regression model. For example, it is possible that the country in which the company is headquartered influences liquidity. Since the headquarter location of companies is not likely to change, this effect remains constant over time. Meanwhile, time fixed effects control for omitted variables that affect the relation between ICG and liquidity over time; however, this effect is the same for all companies. The intercept terms are allowed to vary over time, yet they are assumed to be the same for each firm for each given point in time (Brooks, 2008). Including time fixed effects is the same as including a dummy variable for each year in the regression model. For instance, when all stock markets around the world are subject to a new set of regulatory changes, this change of environment can influence ICG and liquidity, yet the effect is the same for all companies.

Unlike the fixed effects regression model, the random effects regression model allows for different intercepts that arise from a common intercept plus a random variable that is different across firms but remains constant over time (allowing for cross-sectional variation) or different across time but do not change between companies (allowing for time variation). More importantly, the random effect regression model initially assumes there are no omitted variables that influence liquidity. Yet, if there are any, it assumes that these omitted variables are uncorrelated with the explanatory variables. Since the random effect regression model does not allow for correlation between the omitted and explanatory variables, the occurrence of correlation among these variables leads to biased results. Yet, the fixed effects regression a) assumes there are omitted variables and b) allows these omitted variables to correlate with the explanatory variables. In this way correlation among these variables do not lead to biased results. Since there is always a chance that the omitted variables are correlated with the explanatory variables, the random effects regression model most likely shows biased results. It is for this reason that I prefer the fixed effect regression model over the random effects regression model.

cross-29 sectional and time variation in the intercept terms and thereby account for potential endogeneity problems I include cross-sectional and time fixed effects separately in my regression model.

Table 6 demonstrates the results for the cross-sectional and time fixed effects regressions20. Similar to the outcome of the OLS regression, the cross-sectional fixed effects regression also finds a negative and significant relation between ICG and liquidity at a 1% level. This means that after controlling for omitted variables that differ across firms but remain constant over time I still find that a higher quality of ICG increases liquidity. Furthermore, the signs of the beta coefficients as well as the significance levels of Size, Leverage, Return and 1/Price do not change compared to the OLS regression. The relation between Assets and B_A spread has changed from a positive to a negative effect. This relation is significant at a 10% level. R&D does still not influence B_A spread.

However, the findings for the time fixed effects regression are contrary to the outcome of the OLS regression and the cross-sectional fixed effects regression. That is, the time fixed effects regression shows that Gov_index does not influence B_A spread. This indicates that when controlling for omitted variables that influence the relation between ICG and liquidity over time but affect all companies the same way, ICG does not have an effect on liquidity. Thus, I suggest that there are omitted variables that both drive ICG and liquidity over time. More importantly, this finding rejects my first hypothesis that ICG positively influences liquidity. This conclusion contradicts Chung, Elder and Kim (2010) and Bar-Yosef and Prencipe (2013), who find that after allowing for cross-sectional and time variation ICG has a positive impact on liquidity. Meanwhile, my results are in line with the results of Prommin, Jumreornvong and Jiraporn (2014). However, I disagree with the reasoning of these authors. By examining the relation between transparency-related ICG and liquidity the authors also control for time fixed effects and thereby find insignificant results. Still they conclude that within firms, governance quality is related to liquidity over time. In my opinion one cannot draw a conclusion like this based on insignificant results. Finally, the signs of the beta coefficients, together with the significance levels of Size, Leverage, Return and 1/Price are similar to the results of the OLS as well as cross-sectional fixed effects regression. Assets stops having a significant effect on B_A spread. The coefficient of R&D becomes significant at a 10% level, indicating that the level of R&D expenditures negatively impacts liquidity.

20