What Matters in Swedish Corporate Governance?
Master’s Thesis 30 credits
Department of Business Studies Uppsala University
Spring Semester of 2018
Date of Submission: 2018-08-08
Axel Edholm Ludvig Karlsson
Supervisor: Adri De Ridder
1
Abstract
By using five measures of corporate governance, this paper sheds light on the relationship between corporate governance, firm performance and firm valuation in a sample of large
Swedish firms between 2013-2016. The study is conducted on the grounds of the Agency Theory as proposed by Jensen and Meckling (1976) and influenced by corporate governance research by Bhagat and Bolton (2008). Using Tobin’s Q and return on assets (ROA) as estimates of firm valuation and firm performance respectively, we find mixed results compared to prior research concerning the effects of good corporate governance. Our study shows that greater equity holdings of board members are significantly and positively impactful on Tobin’s Q as well as ROA. Furthermore, we find that a larger board size has a significant inverse relationship with both Tobin’s Q and ROA, which is consistent with prior research suggesting that smaller boards are more effective. Interestingly and partly inconclusive with prior research however, we find that greater equity holdings of the CEO is significantly and negatively impactful on Tobin’s Q as well as ROA. These results are robust for multiple controls and various models.
Keywords: Corporate governance, firm performance, Agency Theory, board size, gender diversity,
director ownership and dividend policy.
2
Table of Contents
1. Introduction ... 3
2. Literature Review ... 6
2.1. Agency Theory and Corporate Governance ... 6
2.2. Regulatory Environment ... 7
2.3. Board Size ... 8
2.4. CEO Ownership ... 9
2.5. Board Ownership ... 11
2.6. Dividend Policy ... 12
2.7. Gender Diversity ... 14
3. Methodology ... 16
3.1. Data Retrieval ... 16
3.2. Independent Variables ... 16
3.3. Performance Variables ... 17
3.4. Control Variables ... 18
3.5. Hypothesis Testing ... 19
4. Data Analysis ... 22
4.1. Summary Statistics ... 22
4.2. Heteroscedasticity Testing ... 23
4.3. Correlations ... 24
5. Results ... 25
5.1. Main Results ... 25
5.2. Controls ... 32
5.3. Robustness Checks ... 33
6. Conclusion and Implications for Further Research ... 37
6.1. Conclusion ... 37
6.2. Implications and Suggestions for Further Research ... 38
7. References ... 39
3
1. Introduction
Jensen and Meckling’s (1976) contributions to Agency Theory brought attention to the costs
occurring due to incomplete alignment of the agent’s and owner’s interests. Agency Theory has also brought the roles of external and internal monitoring mechanisms to the forefront of empirical research. Corporate governance considers Agency Theory and addresses the problems that occur due to the separation of ownership and control in corporations, and in which ways the suppliers of capital to corporations ensure themselves of getting return on their investments (Shleifer and Vishny, 1997).
There was a surge of interest in corporate governance following the collapse of Enron in 2002, in the
hope that a closer scrutiny of firms’ governance could prevent future disasters (Roberta, Bhagat and
Bolton, 2008). Also, the global financial crisis of 2008-2009, triggered by the sudden failure of
Lehman Brothers and the subprime mortgage problems, renewed the interest in corporate governance
in the financial sector. The development of a strong corporate governance framework is important in
order to protect stakeholders, and the examples of massive corporate collapses like Enron and
Parmalat that resulted from weak corporate governance systems have highlighted the need to
improve and reform corporate governance (Dibra, 2016). Prior to the surge of interest in corporate
governance, three financial economists – Gompers, Ishii and Metrick – created what was called the
GIM index (Gompers, Ishii and Metrick, 2003). The authors in the oft-cited paper created an index of
corporate governance quality consisting of 24 variables for a large number of publicly traded firms
on the U.S. stock exchange. Following the collapse of Enron, the research by Gompers et al. (2003)
was of great interest to not only scholars but a far wider audience (Roberta et al. 2008). The GIM
index showed a striking relationship between corporate governance and stock returns during the
1990s. Buying well governed firms and selling weak ones as according to their index proved to yield
an abnormal annual return of 8,5 percent (Gompers et al, 2003). Also, the study found that well
governed firms were highly correlated with firm value. Since then, a number of researchers (see
Renders, Gaeremynck and Sercu (2010), Brown and Caylor (2004) and Bebchuk, Cohen and Ferrell
(2008)) have created their own corporate governance indices related to firm performance showing
similar results. However, Roberta et al. (2008) argue that, although the aforementioned researchers
found positive associations between their indices’ rankings of corporate governance quality and firm
performance, correlations are not causation. Further, Roberta et al. (2008) argue that subsequent
research on the topic has questioned whether a positive association truly exists between the
4
indices and firm performance and state that while identifying measures of governance quality is admirable in theory, the existing indices fail to capture the diverse ways in which governance operates in firms. Instead, a single, simple governance variable could perform better than the leading academic indices as the indices are constructed to treat all variables as complements, while the data suggests that several mechanisms are substitutes to each other. Roberta et al.
(2008) conclude that one size does not fit all since governance is highly context-specific, something which limits the effects of corporate governance indices. Given the aforementioned critique, we find it interesting to research corporate governance variables according to the research by Bhagat and Bolton (2008). Their study considers seven corporate governance measures as opposed to a larger index, and find that one specific measure, namely stock ownership of board members, should be the focus of boards, policy makers and researchers.
Board ownership is positively related to both future operating performance and to the probability of disciplinary management turnover in poorly performing firms, whilst the GIM index is only positively related with future operating performance. (Bhagat and Bolton, 2008). These results suggest that there are in fact standalone variables that could prove more valuable than
governance indices when relating corporate governance to firm performance.
Furthermore, prior literature evaluating the relationship between corporate governance, firm performance and firm valuation focuses largely on firms listed on the US market, a market which separates itself from others in terms of legal protection. Legal researchers such as David and Brierley (1985) stated that the commercial legal systems of most countries stem from a few legal
“families”, such as common-law, French-civil-law and German-civil-law. Scandinavian countries however, do not originate from any of the aforementioned legal families, but has a legal system that origins from their own traditions (La Porta, Lopez-de-Silanes, Shleifer and Vishny, 2000a).
Different legal families have a different set of rules concerning shareholder protection, thus the
effects of corporate governance should differ amongst the legal families. Common-law countries,
such as the US and the UK, have the strongest protection of outside investors, whereas French-
civil-law has the weakest. Scandinavian countries fall in between, meaning that they have weaker
protection for outside investors than common-law countries, but
5
stronger than French-civil-law countries. However comparatively, Scandinavian countries have stronger protection for creditors (La Porta et al, 2000a).
The different approaches in how to relate corporate governance to firm performance in previous studies, along with the previous focus on the US market, has left the Scandinavian market in need of further research along the lines of Bhagat and Bolton (2008). Researchers have targeted the Scandinavian market to some extent, (see Eisenberg, Sundgren and Wells (1998)) however not in the fashion of Bhagat and Bolton (2008), as previous research often focuses on a single measure. This study aims to fill that void by providing empirical evidence on the relationship between firm performance, firm valuation and several corporate governance measures on the Swedish market. Our results suggest that a greater board size as measured by the number of directors that sit on the board, has a significantly negative relationship with firm performance and firm valuation as measured by return on assets (ROA) and Tobin’s Q respectively.
Conversely, we also find that a greater total equity holding of all board members have a positive and significant relationship with Tobin’s Q and ROA. Also, interestingly, we find that a greater ratio of equity holdings by the CEO has a negative relationship with both Tobin’s Q and ROA, which is, to a certain extent, contrary to the findings of prior research. This is also inconsistent with theories regarding agency costs as proposed by Jensen and Meckling (1976). Furthermore, results concerning gender diversity amongst board members as well as the dividend payout ratio show no significant results.
The remainder of this paper is organized as follows. Section 2 reviews Agency Theory, corporate governance and the importance of the regulatory environment. Section 2 also includes a
theoretical scrutiny of each corporate governance measure considered in this paper as well as a hypothesis regarding each measure. Section 3 presents the methods used to carry out the analysis including descriptions of all variables as well as the models employed in this thesis. Section 4 present the analysed sample, with descriptive statistics, a bivariate analysis as well as other tests.
Section 5 presents the results from the employed models as well as its robustness to several
controls. Section 6 concludes the paper and highlights its contribution and implications for
further studies.
6
2. Literature Review
This study addresses the relationship between corporate governance, firm performance and firm valuation. Forthcoming in the literature review, we will scrutinize previous research and
highlight the corporate governance variables suited for our research. Furthermore, we address the importance of our samples legal origin as discussed by La Porta et al. (2000a) which shows that corporate governance varies between countries as a result of differences in the regulatory environment. As previous research has focused largely on US firms, differences in regulatory environment gives us incentive to study the effects of corporate governance on companies listed elsewhere.
2.1. Agency Theory and Corporate Governance
Agency Theory is a central theme in corporate governance, as it discusses the alignment of interests of different parties within a company (Jensen and Meckling, 1976). Jensen and Meckling (1976) define an agency relationship as a contract under which principals engage an agent to perform some service on their behalf and thus delegate some decision-making authority to the agent. If both parties are utility maximizers, there is a good reason to believe that the agent will not always act in the best interest of the principal, but in the interest of himself. However, the principal could establish incentives for the agent in order to limit divergences from his interest. In agency relationships the principal will often implement monitoring devices, such as budget restrictions, compensation policies and operating rules to name a few, in order to reduce the agency costs of diverging interests. (Jensen and Meckling, 1976).
Applying the Agency Theory in the context of a corporation, the shareholders and its
representatives act as a principal whereas the manager act as an agent. A typical example of
agency costs within a corporation exists when a shareholder is concerned about a manager’s
misuse of provided capital. For instance, a CEO might pursue growth at the cost of profitability
because it lies within his or her interest. (Stulz, 1999). Corporate governance could be used as a
way to mitigate the costs stemming from divergences of interests. Essentially, corporate
7
governance manages the rules by which the agent plays and by implementing monitoring devices and incentives, principals could realign the interest of the agent with their own.
2.2. Regulatory Environment
La Porta et al. (1998) provide evidence that the level of shareholder protection and quality of law enforcement differs across judicial system. For instance, judicial systems in common-law
countries like the US have a stronger shareholder protection than other systems. However, Scandinavian-law, French and German-civil-law differs in certain aspects, resulting in fewer rights and less influence for common shareholders. In judicial systems where shareholder protection is weak, there should be greater opportunities for agents to expropriate value from shareholders. La Porta et al. (1998) find that because of different legislation, countries with a legal framework that does not provide enough shareholder protection often need to incorporate additional corporate governance mechanisms to ensure that firms are well governed and expropriation of minority shareholders is limited. La Porta et al. (2000) distinguish between French-civil-law, German-civil-law, common-law, and Scandinavian-law to scrutinize in what regard different legal systems affect shareholder protection. Shareholder protection in
Scandinavian countries is weaker than that of common-law countries due to lacking mechanisms to protect oppressed minority shareholders. For instance, according to Scandinavian-law, a shareholder cannot call an extraordinary shareholder meeting if he or she owns less than 10 percent of the outstanding shares. Also, the legal system in Scandinavian countries does not allow cumulative voting or proportional representation of minorities on the board of directors.
Cumulative voting allows a group of minority shareholders to combine their voting power to ensure that the minority shareholders’ interests are represented within the company. Proportional representation of minorities of the board of directors allows minority shareholders to name future members of the board of directors. (La Porta et al, 2000a). The findings of La porta (1998;
2000a) are important for our research because, as mentioned above, Scandinavian-law offers
better opportunities for expropriation than common-law countries where most prior studies have
been conducted. This should present an opportunity for Scandinavian firms to implement and
make use of corporate governance mechanisms effectively. That is, in a country like Sweden,
corporate governance could be of more importance as a way to reduce agency costs since,
compared to a common-law country, shareholder protection is not as well embedded in the
8
current legal system. A study by Renders et al. (2010) find that the effects of improved corporate governance are greater in a country with a weak judicial system than in one which has a strong judicial system.
2.3. Board Size
Papers such as Lipton and Lorsch (1992) and Jensen (1993) have criticized the efficiency of boards on account of their size. Jensen (1993) argue that keeping boards small can help improve their performance, as boards that grow beyond seven or eight people are less likely to function efficiency due to the coordination and processing problems that arise. Lipton and Lorsch (1992) argue similarly that the monitoring advantages associated with a larger board size are
outweighed by the disadvantages of slower decision-making, less unbiased discussions of managerial performance, free-rider problems and biased risk-taking. Further, Lipton and Lorsch (1992) argue that these problems increase with the number of board members and recommend limiting the size of the board to ten people, although preferably only eight or nine. Neither Jensen (1993) or Lipton and Lorsch (1992) conduct statistical analysis on the effect of board members, something which Yermack (1996) mentions in his study. Yermack (1996) empirically confirms that smaller boards are more likely to dismiss CEOs following periods of inferior performance and that the threat of CEO dismissal decreases with the size of the board. By
studying large US corporations, he shows that there is an inverse relationship between board size and firm value as measured by Tobin’s Q during 1984-91. The negative relationship extends to financial ratios such as return on assets (ROA) and return on sales which further supports his findings. Similar research conducted on the European market, such as Eisenberg et al. (1998) and Conyon and Peck (1998), have found results consistent with Yermack (1996). Eisenberg et al. (1998) present evidence that the negative correlation between board size and firm
performance as measured by ROA extends to small firms in Finland. Conyon and Peck (1998)
find empirical evidence of the costs associated with a large board size and argue similar to
Lipton and Lorsch (1992) that the monitoring advantages of larger boards are outweighed by the
problems concerning communication and decision-making. The study find that board size has a
negative correlation with measures of corporate performance, such as return on equity (ROE)
and Tobin’s Q for five European countries between 1990-1995 (Conyon and Peck, 1998). More
recent studies by Guest (2009) and Nguyen, Rahman, Tong and Zhao (2015) find further
9
evidence to support the findings of the aforementioned scholars. Guest (2009) find support in the UK market and conclude that board size has a negative impact on several measures of firm performance such as Tobin’s Q, share returns and ROA. Guest (2009) also find the strongest negative board size – firm performance relation in large firms, which are more likely to have larger boards. Nguyen et al. (2015) find evidence that larger boards affect firm valuation
negatively in Australian firms between 2001-2011 for several measures of performance, such as Tobin’s Q and ROA. The results suggest that in an average firm of 5,2 directors, adding one director is likely to cause a decrease in Tobin’s Q of about 7 percent. These results are
economically large and stronger than the results obtained by Yermack (1996), whose average firm includes 12.25 directors. Thus, Nguyen et al. (2015) conclude that smaller boards
experience a stronger negative impact on firm value as the board increases in size compared to firms with larger boards. Based on the research mentioned above, we formulate our first hypothesis as follows.
Hypothesis 1: A larger number of directors on the board will have a negative relationship with ROA and Tobin's Q
2.4. CEO Ownership
To align the interests of managers and the shareholders is crucial in order to reduce agency costs.
Jensen and Meckling (1976) picture a firm completely owned by the manager and argue that as the owner-manager sells equity claim on the firm, agency costs will be generated by the
divergence between his own interests and the interests of the shareholders. As the manager’s equity claim decreases, he will only bear a fraction of the costs of any private benefits he takes out in order to maximize his own utility. Conversely, the costs of deviations in interest between outside shareholders and managers decrease as the manager’s equity holdings increase, a
hypothesis hereafter referred to as the convergence-of-interest hypothesis. (Jensen and Meckling, 1976). However, Morck, Shleifer and Vishny (1988) argue that there is a countervailing
hypothesis, the entrenchment hypothesis, which states that a manager who controls a substantial
fraction of the firm’s equity may have enough influence to guarantee his employment despite
unsatisfying performance. While the Convergence-of-Interest hypothesis argues that larger
managerial equity ownership should result in better performance, the Entrenchment Hypothesis is
10
not as positive. Morck et al. (1988) find evidence to support their previous concerns of
entrenchment. The study show that managerial ownership is positively related to Tobin’s Q when the combined ownership of the top two officers ranges from 0 to 5 percent. However,
interestingly, when the combined ownership exceeds 5 percent, each increase in ownership affects Tobin’s Q negatively which indicates that an entrenchment effect of large managerial ownership could exist (Morck et al, 1988). Equivalent results have been found by McConnell and Servaes (1990) and Griffith (1999). McConnell and Servaes (1990) find that there is significant curvilinear relationship between Tobin’s Q and the common stock ownership of corporate insiders for large US firms in 1976 and 1986. The results suggest that insider ownership has a positive effect on Tobin’s Q until ownership reaches 49.4 percent and 37.6 percent for 1976 and 1986 respectively, after which the effect turns negative (McConnell and Servaes, 1990). Griffith (1999) finds a similar non-monotonic relationship between Tobin’s Q and CEO ownership, also on the US market. He finds that CEO ownership is positively correlated with Tobin’s Q when the CEO own between 0 and 15 percent of the firm’s capital and declines as the ownership increases to 50 percent, after which it increases again. The results of the aforementioned studies are
consistent with the Convergence-of-Interest theory at the lower levels however the decline in Tobin’s Q at higher levels of ownership supports the entrenchment hypothesis. However, studies have been conducted regarding CEO ownership and firm performance that support the
Convergence-of-Interest hypothesis. Chen, Guo and Mande (2003) find that Tobin’s Q increases monotonically in relation to managerial ownership, and therefore suggest that as managerial ownership increase, there is a greater alignment of interest between managers and stakeholders.
Mehran (1995) studies the effects of stock and stock options held by the CEO and their effect on
firm performance. He finds that both the stock holdings of the CEO, as well as the combined
stock and stock option holdings of the CEO are positively correlated with Tobin’s Q. Similar
research conducted by Barnhart and Rosenstein (1998) find a positive relation between the
combined percentage of shares held by officers and directors with Tobin’s Q. Although the
findings of the aforementioned scholars are somewhat inconclusive, the majority find that the
effects of CEO ownership are positive. Due to this, we formulate our second hypothesis as
follows.
11
Hypothesis 2: Larger equity holdings of the CEO will have a positive relationship with ROA and Tobin’s Q
2.5. Board Ownership
The board of directors has the final responsibility for the functionality of the firm and sets the rules to which the CEO comply, and therefore play an important monitoring role. The jobs of the
board is to hire, select the compensation structure of the CEO and to provide counsel. Jensen (1993) argues that few boards in the past decade have done their monitoring well, largely
because board members do not hold substantial equity in the firm which board they are sitting on (Jensen, 1993). Therefore, he argues that board members should hold equity in order to provide better incentives and align their interests with the shareholders’ interests. Morck et al. (1988) provide some insight into the relationship between board members’ equity ownership and firm performance. The study finds that the combined ownership of equity among board members, excluding the CEO, show a non-monotonic relationship with Tobin’s Q, similar to their findings regarding managerial ownership as previously mentioned. Board equity ownership has a positive effect on Tobin’s Q and profit rate for ownership levels between 0 and 5 percent, however negative for ownership levels between 5 and 25 percent (Morck et al, 1988). These results suggest that board members, like managers, contribute to corporate wealth when they have their own money at stake but also that they might become entrenched at high levels of equity
ownership. Yermack (1996) studies the relationship between officer and director stock
ownership and measures of performance. He finds that the stock ownership of officers and
directors has a monotonically positive relationship with Tobin’s Q as well as sales to assets. The
findings of McConnell and Servaes (1990) formerly discussed in section 2.4 is also important
when discussing the equity ownership of boards, as they concluded that insider ownership, which
includes board members’ equity ownership, exhibited a non-monotonic relationship with Tobin’s
Q. Bagnani, Milonas, Saunders and Travlos (1994) find a significant relationship between bond
return premia and officer and director equity ownership in the 5 to 25 percent range. The authors
argue that bondholders require higher premiums when managers and directors act in the interests
of shareholders who are more prone to risk than bondholders. Weak evidence is found which
confirms the hypothesis that when managers’ and directors’ equity holdings are large, they
become more risk averse due to their non-diversifiable wealth. Consequently, in high ownership
12
ranges, managers’ and directors’ interests become more closely aligned with those of the bondholders rather than the shareholders, thus showing the same results as prior research on insider equity holdings (Bagnani et al, 1994). A more recent study by Guest (2009) shows that there is a positive significant relationship between the aggregate percentage of share ownership of board members and return on assets (ROA). Similar to the relationship between CEO equity ownership and firm performance measures, studies concerning the equity ownership of the board are somewhat inconclusive. However, the overall findings suggest that there is a positive
relationship between board of director ownership, firm performance and firm valuation, although perhaps non-monotonic. Therefore, we formulate our third hypothesis, as follows.
Hypothesis 3: Larger combined equity holdings of the board members will have a positive relationship with ROA and Tobin’s Q
2.6. Dividend Policy
The payout of cash to shareholders creates a major conflict between outside shareholders and corporate insiders as it reduces the resources under the managers’ control and thereby reduces the managers’ power (Jensen, 1986). Managers have incentives to manage their firms to grow
beyond their optimal size, as growth is positively related to the resources at the managers’
disposal as well as their compensation. The free cash flow generated in corporations creates a conflict of interest in how to motivate managers to disgorge cash instead of retaining earnings and using it for private benefits. (Jensen, 1986). Dittmar (2008) argues that high levels of cash holdings can be a cause for concern, since a considerable body of research has provided evidence that high levels of cash flow can lead to corporate waste and loss of value. Managers working under weak governance systems are likely to use excess resources for private benefits, thus giving rise to substantial agency costs (Dittmar, 2008). Agency costs that occur due to excess cash holdings can take the form of perk consumption and ill-advised, overpriced acquisition.
Dittmar and Smith (2007) show that in companies with weak governance systems, a dollar of
cash can be worth significantly less than a dollar, because the cash is considered to be used
inefficiently. Similarly, Kalcheva and Lins (2007) show that outside investors discount cash
holdings in firms that are likely to have agency problems due to weak protection against
expropriation. Also, when external shareholder protection is weak, dividend payouts are
13
positively related to firm valuation however when external shareholder protection is strong, cash holdings are unrelated to firm value (Kalcheva and Lins, 2007). By paying dividends, insiders return corporate earnings to their investors, and are therefore no longer capable of using retained earnings to benefit themselves (La Porta et al, 2000b). Accordingly, Dittmar (2008) argue that dividends can work as a corporate governance mechanism to reduce excess cash holdings and thereby reduce agency costs. Also, both La Porta et al. (2000b) and Jensen (1986) argue that the payment of dividends exposes managers to monitoring, should the manager want to raise
external funds from the capital market, and hence give outside investors an opportunity to exercise some control over the insiders. As a result, outside shareholders might prefer dividend payments over retained earnings since a failure to disgorge cash could lead to diversion, waste and less monitoring (La Porta et al, 2000b; Jensen, 1986). Harford (1999) provide evidence that support the free cash flow hypothesis, which predicts that agency conflicts arise between
managers and shareholders when the firm hold excessive cash. The study show that acquisitions carried out by cash-rich firms are value decreasing, as seen by a subsequent decline in stock returns following the acquisition. Also, mergers in which the bidder is cash-rich are followed by abnormal declines in operating performance (Harford, 1999). Further evidence is presented by Adjaoud and Ben-Amar (2010), who study the relationship between corporate governance quality and dividend policy in Canada. They find that better governed firms pay higher
dividends, and thus conclude that effective corporate governance attenuate agency conflicts and limit managers’ opportunistic behaviour in payout policy. Lastly, La Porta et al. (2000b) find that firms operating in countries with strong minority shareholder protection tend to pay higher dividends, and conversely, that firms operating in countries with weak minority shareholder protection tend to pay less dividends. These findings are consistent with the notion that dividend payouts are more uncommon in firms where expropriation is viable due to poor legal protection.
Given prior research, we predict that the payout ratio will have a positive relationship with firm performance and firm valuation. We formulate our fourth hypothesis as follows.
Hypothesis 4: Higher dividend payout ratios will have a positive relationship
with ROA and Tobin’s Q
14
2.7. Gender Diversity
Studies such as Cox and Blake (1991) and Robinson and Dechant (1997) have argued in favour of diversity in the workplace. Cox and Blake (1991) argue that diversity in organizations may lead to competitive advantages in terms of creativity, problem solving and flexible adaptation to change. Additionally, Robinson and Dechant (1997) argue that since the demographic in the marketplace is diverse, matching the diversity of a company to the diversity of the marketplace increases the ability to penetrate markets successfully. Also, while heterogeneity may produce more conflict in decision making processes, the variety of perspectives that emerges ensure that differing views and perspectives are considered. Robinson and Dechant (1997) states that heterogeneous groups outperform homogeneous groups in identifying problems and generating alternative solutions. While these scholars study workplace diversity, the arguments are similar for board diversity. Carter, Simkins and Simpson (2003) find empirical evidence that more gender diverse boards increase firm performance, and argue that diverse boards are better at solving problems due to their ability to offer diverse perspectives on certain issues. The study also finds significant relationships between the ratio of females or minorities on the board and firm value as measured by Tobin’s Q on firms listed on Fortune 1000. Even after controlling for size and industry, there are significant differences in Tobin’s Q between two groups of firms, where firm with highly diverse boards outperform those that are less diverse. (Carter et al., 2003). Further, Adams and Ferreira (2009) study the frequency of females in the boardroom and its impact on corporate governance and firm performance. The results of the study suggest that female board participation has a positive impact on performance in those firms that
otherwise have weak governance, as measured by their ability to resist takeovers. In firms with
strong governance however, gender diversity could decrease shareholder value possibly due to
over monitoring. The studies by Carter et al. (2003) and Adams and Ferreira (2009) speak of
the equivocal results concerning the effects of gender diversity in the boardroom. Further,
Carter, D’Souza, Simkins and Simpson (2010) find a significantly positive relationship between
the number of female board members and return on assets (ROA), however no significant
relationship between gender diversity and Tobin’s Q. Conversely, Chapple and Humphrey
(2012) do not find a strong case for gender diversity in boards, as they find no differences in stock
performance in gender diverse and all-male board firms. Instead, weak evidence is found which
suggests that more than one female on a board is associated with lower returns. Francoeur, Labelle
15
and Sinclair (2008) examine how the participation of females in the board of directors affect financial performance and find that firms operating in complex environments whilst having a high proportion of female directors have significantly higher returns. Erhardt and Werbel (2003) as well as Scholtz and Kieviet (2017) have found similar results to those of Carter et al. (2003). Due to the results of aforementioned research, which indicate that a gender diverse board have positive impact on firm performance and firm valuation, we formulate our fifth and last hypothesis as follows.
Hypothesis 5: A greater ratio of female board members to total board members
will have a positive relationship with ROA and Tobin’s Q
16
3. Methodology
3.1. Data Retrieval
Our analysis uses a panel of firms listed on the OMX Stockholm Large Cap which constitutes of the largest firms listed on the Swedish stock exchange based on market capitalization. We collect data using mainly Thomson Reuters Datastream accessed via Uppsala University. Datastream is a database for financial and economic research data which provide information for time series.
However, some data have been collected manually due to insufficient data in Datastream. We use a sample selection rule that requires each firm to have complete data representing each variable each year between 2013 and 2016. This requirement represents the need for full data in order to fully conduct our analysis. In total, we obtain a final sample of 292 observations for 73 firms across four years.
3.2. Independent Variables
For brevity, we shorten the name of each forthcoming variable in models and tables, and the shortened name is presented in parentheses next to the concerned variable in the coming sections.
Data regarding Board Size (BS) and Dividend Payout Ratio (DivP) is collected using Datastream.
Consistent with Bhagat and Bolton (2008) we estimate Board Size as the total number of board members at the end of the fiscal year, while Dividend Payout Ratio is estimated as the dividend payments out of net profits. The remaining governance predictors CEO Ownership (CEO), Board Ownership (BO) as well as Gender Diversity (GD) are collected manually by reviewing the annual reports of each firm at the end of the fiscal year. CEO Ownership as well as Board
Ownership are estimated as the fraction of the share holdings of the CEO and the aggregate share
holdings of all board members respectively to total outstanding shares. These estimations are in
line with prior research such as Yermack (1996), McConnell and Servaes (1990), Guest (2009)
and Bhagat and Bolton (2008) and do not include stock options. Lastly, consistent with Carter,
Simkins and Simpson (2003) as well as Francoeur , Labelle and Sinclair (2008) we estimate
Gender Diversity as the ratio of female board members to total board members.
17
3.3. Performance Variables
The dependent variables return on assets (ROA) and Tobin’s Q (TQ) are estimates of firm performance and firm valuation respectively. We estimate Return on Assets (ROA) by using Bebchuk et al’s (2008) definition, where:
𝑅𝑒𝑡𝑢𝑟𝑛 𝑜𝑛 𝐴𝑠𝑠𝑒𝑡𝑠 (𝑅𝑂𝐴) = 𝑁𝑒𝑡 𝑖𝑛𝑐𝑜𝑚𝑒
𝐸𝑛𝑑 𝑜𝑓 𝑦𝑒𝑎𝑟 𝑡𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡𝑠
Furthermore, we estimate Tobin’s Q by using Kaplan and Zingales’ (1997) definition, which subsequently has been used by Gompers et al. (2003), Bebchuk et al. (2008) as well as Bhagat and Bolton (2008), where:
𝑇𝑜𝑏𝑖𝑛
′𝑠 𝑄 = 𝑀𝑎𝑟𝑘𝑒𝑡 𝑣𝑎𝑙𝑢𝑒 𝑜𝑓 𝑎𝑠𝑠𝑒𝑡𝑠 𝐵𝑜𝑜𝑘 𝑣𝑎𝑙𝑢𝑒 𝑜𝑓 𝑎𝑠𝑠𝑒𝑡𝑠
According to this estimation, Tobin’s Q is equal to the market value of assets divided by the book value of assets, where the market value of assets is estimated as the book value of assets plus the market value of common stock less the sum of book value of common stock and deferred taxes.
This estimation differs from scholars such as Yermack (1996) as he estimates Tobin’s Q based on replacement costs of assets, while we use a simpler market-to-book ratio. Research by Perfect and Wiles (1994) suggests that the improvements from more advanced computation of Tobin’s Q are limited. Also, Bebchuk et al. (2008) argue that even simpler estimations that drop deferred taxes have been increasingly used in favour of more sophisticated measures of Tobin’s Q due to the high correlation between this proxy and more advanced measures such as the one Yermack (1996) employ. All data needed to estimate ROA and Tobin’s Q have been collected using Datastream and is estimated at the end of each fiscal year without lag.
18
3.4. Control Variables
We consider several variables for control; firm size, leverage, the age of the firm, capital expenditures to assets as well as the standard deviation of stock returns – all variables commonly used in relevant research. Yermack (1996) as well as Bebchuk et al. (2008) argue that firm size is an important determinant of firm performance and as well as board size and should consequently be considered in analyses like ours. We estimate firm size (Size) as the log of total assets at the end of the fiscal year, in accordance with Bhagat and Bolton (2008).
Total assets is collected from Datastream. Consistent with Bhagat and Bolton (2008) as well as Bebchuk et al. (2008), we estimate Leverage (Lev) as the long term debt plus the current portion of long term debt (long term debt due within one year) divided by total assets. Long term debt and the current portion of long term debt is collected using Datastream and estimated at the end of the fiscal year. We also include the age of the firm (Age) as a control variable following Gompers et al. (2003) and Bebchuk et al. (2008) and is estimated as the log of months that the firm has existed. As some firms are the results of mergers, we compute Age starting from the last merger. The necessary information needed to carry out our estimation did not exist in Datastream. Instead we used multiple sources to gather the information such as annual reports and firm websites. Furthermore, the standard deviation of stock returns
(STDDEV) is employed as a control variable in research similar to ours, such as Guest (2009) and Bhagat and Bolton (2008) to account for risk. STDDEV is computed as the average standard deviation of the monthly stock return over the 12 months preceding the financial year end. The data needed to compute STDDEV was collected using Datastream. Lastly, consistent with Gompers et al. (2003) as well as Bebchuk et al. (2008) we also include capital
expenditures to total assets (Cap/A) to control for differences in investment opportunities. The necessary data is also collected using Datastream. Yermack (1996) argue that controlling for investment opportunities in some way are important since Tobin’s Q can capture the value of future investment opportunities.
19
3.5. Hypothesis Testing
In order to test the results in reference to the hypotheses, we employ several regressions similar to Yermack (1996), Bhagat and Bolton (2008) and Bebchuk et al. (2008). All papers employ
ordinary least square (OLS) regressions with robust standard errors to account for potential
heteroscedasticity as part of their analysis. Also, OLS regressions with fixed effects estimators
are often used in relevant literature (see Bhagat and Bolton (2008), Yermack (1996), Bebchuk et
al. (2005) etc.). Petersen (2008) state that OLS standard errors are biased when residuals are
correlated across observations in panel data, and argue in favour of using alternative techniques,
such as fixed effects, in order to improve the efficiency of estimations. Furthermore, Thompson
(2011) argue that OLS standard errors is unlikely to be uncorrelated across both firms and time in
a finance panel data set. For instance, market-wide shocks will induce correlation between firms
at a specific time. Likewise, Bhagat and Bolton argue that when analysing panel data, the
residuals for a specific firm may be correlated across years or across the sample firms for a
specific year. Yermack (1996), Bebchuk et al (2008) as well as Guest (2009) all employ time-
fixed effects by implementing time dummy variables. Consequently, each year is assigned its
own unique intercept and thus take account of differences between time periods. We mimic this,
and employ dummies for each year between 2013-2016. Furthermore, Bebchuk et al. (2008)
study is of utter importance to us as their panel data is similar to ours. Their panel data set is
divided into two separate time frames; 1992-2002 (t=10) as well as 1998-2002 (t=4), where time
as well as firm-fixed effects are both included in the longer time frame however firm-fixed
effects are neglected in the shorter time frame. Other research, such as Yermack (1996) and
Bhagat and Bolton (2008), who employ firm-fixed effects, have data sets that span over longer
periods of time. Our sample spans, just like Bebchuk et al’s (2008) shorter time frame, over four
years. Similar research which focus on short time periods, such as Bhagat and Black (2001) and
McConnell and Servaes (1990) control for industry specific differences rather than firm specific
differences. Bhagat and Black (2001) study two a two-year period and control for industry
differences by using the mean of the dependent performance variable in each regression as a
control variable. Consequently, they can observe how the industry characteristics affect the
overall results of the regressions. McConnell and Servaes (1990) employ a similar method, but
however calculate the average value for each variable in the specific industries and subtract it
from the original values of each firm. We mimic the model by Bhagat and Black (2001) to reduce
20
the issues of endogeneity that is a common issue in these types of studies and consequently to account for differences between industries. Each firm is allocated to an industry group using Industry Classification Benchmark (ICB) which is the official industry classification on the NASDAQ stock exchange, where our sampled firms are listed. Thereafter, average values for Tobin’s Q and ROA are computed and included in the corresponding regression as a control variable. All variables which will be included in the forthcoming models can be displayed in Table 1 in Appendix 1, along with summary of how each variable is estimated. Altogether, we employ three separate models with two dependent variables each, resulting in a total of six regressions. All regressions are carried out using Stata, and all regressions employ robust standard errors. Firstly, 3.1 and 3.2 below show the OLS models.
TQ
it= Constant + 𝛽
1BS
it+ 𝛽
2GD
it+ 𝛽
3BO
it+ 𝛽
4CEO
it+ 𝛽
5DivP
it+ 𝛽
6Cap/A
it+
𝛽
7Size
it+ 𝛽
8Lev
it+ 𝛽
9Age
it+ 𝛽
10STDDEV
it+ ℇ
it(3.1) ROA
it= Constant + 𝛽
1BS
it+ 𝛽
2GD
it+ 𝛽
3BO
it+ 𝛽
4CEO
it+ 𝛽
5DivP
it+ 𝛽
6Cap/A
it+
𝛽
7Size
it+ 𝛽
8Lev
it+ 𝛽
9Age
it+ 𝛽
10STDDEV
it+ ℇ
it(3.2)
Where i represents the individual (firm) of the observation, t the time (year) of the observation and ℇ is the error term.
Next we include yearly time fixed effects to account for time-specific effects, as shown in model 3.3 and 3.4 below.
TQ
it= Constant + 𝛽
1BS
it+ 𝛽
2GD
it+ 𝛽
3BO
it+ 𝛽
4CEO
it+ 𝛽
5DivP
it+ 𝛽
6Cap/A
it+
𝛽
7Size
it+ 𝛽
8Lev
it+ 𝛽
9Age
it+ 𝛽
10STDDEV
it+ Year Dummies
t+ ℇ
it(3.3) ROA
it= Constant + 𝛽
1BS
it+ 𝛽
2GD
it+ 𝛽
3BO
it+ 𝛽
4CEO
it+ 𝛽
5DivP
it+ 𝛽
6Cap/A
it+
𝛽
7Size
it+ 𝛽
8Lev
it+ 𝛽
9Age
it+ 𝛽
10STDDEV
it+ Year Dummies
t+ ℇ
it(3.4) Where Year Dummies capture the time fixed effects over time. For brevity, we simply include Year Dummies in the model rather than each year’s dummy variable.
Lastly, we include control for industry specific differences as shown below in model 3.5 and 3.6.
21
TQ
it= Constant + 𝛽
1BS
it+ 𝛽
2GD
it+ 𝛽
3BO
it+ 𝛽
4CEO
it+ 𝛽
5DivP
it+ 𝛽
6Cap/A
it+
𝛽
7Size
it+ 𝛽
8Lev
it+ 𝛽
9Age
it+ 𝛽
10STDDEV
it+ 𝛽
11IND
it+ Year Dummies
t+ ℇ
it(3.5) TQ
it= Constant + 𝛽
1BS
it+ 𝛽
2GD
it+ 𝛽
3BO
it+ 𝛽
4CEO
it+ 𝛽
5DivP
it+ 𝛽
6Cap/A
it+
𝛽
7Size
it+ 𝛽
8Lev
it+ 𝛽
9Age
it+ 𝛽
10STDDEV
it+ 𝛽
11IND
it+ Year Dummies
t+ ℇ
it(3.6) Where β
11IND
itis the industry average on the pertained dependent variable as accounted for
earlier.
22
4. Data Analysis
4.1. Summary Statistics
Table 1 display the mean, median, standard deviation as well as minimum and maximum values for all the dependent, independent and control variables in our sample. In order to control for influential outliers, the data have been winsorized at a 2.5 percent level, meaning that the top and bottom 2.5 percent values in each variable are replaced by their closest corresponding
observation value.
Table 1.
Descriptive Statistics
Variable Mean Median Standard
deviation
Min Max
Performance Variables
Tobin’s Q 1.94 1.54 1.34 0.51 6.77
ROA 0.07 0.06 0.07 -0.05 0.33
Governance Variables
Board Size 9.5 10 2.34 5 16
Gender Diversity 0.31 0.31 0.12 0.13 0.57
CEO Ownership 0.014 0.0003 0.065 0.00 0.38
Board Ownership 0.072 0.004 0.12 0.00 0.51
Dividend Payout 0.49 0.48 0.29 0.00 1.00
Control Variables
Size 186713 36592 551867 8004 2619317
Leverage 0.203 0.53 0.15 0 0.80
Capex/Assets 0.036 0.023 0.046 0.00 0.33
Age 65.77 55.50 53.92 1 327
Standard deviation of stock returns
1.64 1.56 0.59 0.93 7.12
Firstly, Tobin’s Q seem to be highly dispersed in our sample. The differences in mean and
median values regarding suggest that the values do not have a symmetrical distribution. The
minimum and maximum values further speak of the asymmetric distribution. Return on assets
(ROA) follows the same pattern with a substantial difference in minimum and maximum values
as well a relatively high standard deviation. CEO and Board Ownership are also highly dispersed,
with large differences in mean and median values as well as minimum and maximum values. We
23
also note that there are large differences in Size, as the mean and median values differ heavily as well as the min and max. Throughout our sample, Size, as measured by total assets is heavily skewed towards financial firms.
Furthermore, in Table 2 below we present the average industry values for Tobin’s Q and ROA as well as the proportion each industry make up in our sample. As observed, Tobin’s Q and ROA are industry sensitive, meaning that their values vary across industries. This suggests that the inclusion of industry control as formerly discussed is important in our research in order to deal with the specific characteristics of each industry. Additionally, in accordance with Bhagat and Black (2001), we exclude industries which only constitutes of one firm, meaning that the Oil &
Gas industry is excluded when running the regressions which includes industry control.
Table 2.
Performance values per industry
Industry Tobin’s Q ROA % of total
observations
Basic Materials 1.70 0.039 9.6%
Consumer Discretionary
3.04 0.128 12.3%
Consumer Staples 2.47 0.106 6.8%
Financials 1.06 0.052 16.4%
Health Care 2.92 0.032 5.5%
Industrials 1.75 0.067 27.4%
Real Estate 0.98 0.059 11%
Technology 2.93 0.071 9.6%
Oil & Gas 1.95 -0.036 1.4%
4.2. Heteroscedasticity Testing
Since we analyze panel data, we test the models for heteroscedasticity in order to see whether the variance of the residuals is constant for the dependent values or not. We test for
heteroscedasticity visually by constructing a separate plot diagram for Tobin’s Q and return on
assets (ROA) respectively and quickly observe that there are clear signs of clustering. The plots
can be observed in Appendix 2. Further, in order to test whether the heteroscedasticity is
significant, we conduct both a Breusch-Pagan test as well as a Cook-Weisberg test. Both test
24
show that the regressions are significantly heteroskedastic, which confirms the use of standard errors that are robust to heteroscedasticity formerly mentioned in light of previous relevant research. The results of the Breusch-Pagan test as well as a Cook-Weisberg test can be observed in Table 2 in Appendix 3.
4.3. Correlations
Table 3 below show the correlation between the variables used in our models. We note that the correlation between Tobin’s Q and ROA is relative strong of 0.49. This is reasonable since better profitability should be highly valued by the market. However, the non-perfect correlation indicate that certain governance variables might impact ROA and Tobin’s Q differently, which could further our analysis. Furthermore, Board Size is positively correlated with Size which suggests that larger firms often have larger boards. Also, according to the matrix, larger firms also tend to pay a larger proportion of their net profits as dividends. Interestingly, CEO Ownership as well as Gender Diversity are negatively correlated with each performance variable, which is contrary to our hypotheses.
Table 3.
Correlation Matrix
TQ ROA BSize GD CEO_ BOow DivP Size D/Aa Cap/A Age STDDEV
TQ 1
ROA 0.49 1
BS -0.14 -0.16 1
GD -0.06 -0.04 0.02 1
CEO -0.16 -0.04 -0.32 0.006 1
BO 0.01 0.11 -0.28 0.18 0.57 1
DivP -0.03 -0.02 0.34 0.09 -0.25 -0.05 1
Size -0.15 -0.20 0.39 0.12 -0.05 -0.05 0.19 1
Lev -0.14 -0.24 -0.08 0.14 0.03 -0.11 -0.11 0.002 1
Cap/A 0.09 -0.03 -0.17 0.13 0.13 0.07 -0.18 0.06 0.31 1
Age -0.25 -0.03 0.18 -0.03 -0.08 -0.05 0.16 0.03 -0.02 -0.25 1
STDDEV 0.35 0.06 -0.16 -0.15 -0.05 -0.13 -0.15 -0.14 -0.11 -0.03 -0.07 1