ABSTRACT
The decreased confidence of society in the banking sector led to multiple regulatory bodies developing new requirements regarding risk related ratios and disclosures. This paper investigates whether the board characteristics gender diversity, independence and board size have an effect on the quality of risk disclosures by European banks. Based on prior research, this study hypothesizes that gender diversity and board size have a nonlinear, converted U-shaped relationship with the quality of risk disclosure and independence has a positive relationship with the quality of risk disclosure. This study takes into account the difference in board structure (one/two-tier) of banks. Regression analysis is used to analyse hand collected data from 59 banks from 19 different countries. After controlling for firm size, leverage and national culture, the results of this study suggest that a significant positive relationship exists between board independence and quality of risk disclosure. However, no evidence is found for the relationships between gender diversity and board size with the quality of risk disclosure. Interestingly, dividing the sample into two groups based on board structure, suggest different relationships. For the sample containing banks with one-tier boards the relationship with independence became insignificant, whereas the results for the sample containing banks with two-tier structures indicated a positive association between gender and quality of risk disclosure.
Keywords: risk disclosure quality, board characteristics, gender diversity, board independence, board size
Is More Always Better?
Board Size, Board Independence and Gender
Feikje H. Zijlstra
University of Groningen
Master Thesis Accountancy Supervised by J. Huttenhuis
20-01-2020 11709 words
Mutua Fidesstraat 13 Groningen f.h.zijlstra@student.rug.nl
0642317827 S2887274
1. INTRODUCTION
he banking industry received a lot of negative media attention over the past few years. Headliners like ‘Swedish bank joins list of EU dirty money scandals’ (Rettman, 2019); ‘More banks are caught up in money-laundering scandals: the aftershock of the Danske affair’ (The economist, 2019); and ‘Dutch bank ING fined $900 million for failing to spot money laundering (Sterling & Meijer, 2019), are not uncommon and they represent the turbulent times banks all over the world are going through.
Since the banking crisis in 2008, the confidence of the market in the banking industry declined (Basel Committee on Banking Supervision, 2014). Therefore, banks are trying to rebuild their reputation and regain the public trust.Findings of Grimmelikhuijsen and Klijn (2015) show that transparency via financial reporting affects social trust positively, and thus increasing transparency is important in regaining reputation and public trust. Over the past few years, supervision increased and more and higher conditions are to be met by banks (Kleymenova & Zhang, 2019). An example is Basel III, which has been legally enforced in the EU as of 1 January 2014. The regulation aims at improving the banking sectors’ ability to absorb shocks arising from financial or economic crisis (Basel Committee on Banking Supervision, 2014). In addition, it wants to improve risk management and governance and increase a bank’s transparency (Basel Committee on Banking Supervision, 2010). It consists of three pillars, where the third pillar is concerned with the disclosure requirements of banks. The Basel Committee on Banking Supervision introduced this pillar to overcome the problem of information asymmetry for the market participants (Basel Committee on Banking Supervision, 2014). The key concepts of this pillar are that the disclosures of banks should be explicit, complete, relevant to users, consistent over time and comparable (Basel Committee on Banking Supervision, 2010). It aims at promoting market discipline by requiring firms to disclose information about their risks and the corresponding objectives and policies (Basel Committee on Banking Supervision, 2014). However, currently there are no accounting requirements to include this regulatory information in the financial statements of banks. In addition, banks and their auditors have not agreed upon a universal set of relevant information. Banks, therefore, have a certain degree of freedom in the amount of risk information they want to include in their annual report and where to disclose this risk information (EDTF, 2012). Hence, there is a variation in the quality of risk disclosures across banks. Since generally only the financial statements are audited by an external auditor (Bedard, Sutton, Arnold & Philips, 2012) and normally risk information is placed outside the financial statements in the annual report, most risk information disclosed by banks is not audited. This makes the information less reliable, leading to reduced informativeness for the users of the information (Huttenhuis & ten Hoeven, 2016).
Several theories can be used to explain the importance and usefulness of risk disclosures by companies (in general) and banks. According to the agency theory (Jensen & Meckling, 1976), the demand for financial reporting and disclosure arises due to information asymmetry and agency conflicts between managers and outside investors (Healy & Palepu, 2001). The disclosure of risk reports by European banks will decrease the information asymmetry with their outside investors. However, according to the stakeholder theory, managers and firms will be better off if they also consider the needs of other groups in addition to their shareholders
(Freeman, 1984). Management, therefore, should consider the interests of other stakeholders, which can be achieved by the disclosure of accounting information in order to assist users in making appropriate decisions (An, Davey & Eggleton, 2002). As stated by the legitimacy theory and signaling theory, risk disclosures could also be a way for a company to signal that their operations are legitimate and, therefore, improve the public trust. In this proposal the focus is on risk disclosures by European banks.
Disclosure on risk information has not been the topic of many research articles so far. Research that does exist on the determinants of the quality of risk reporting has mainly focused on the mechanical effects of a bank’s contextual factors such as size, leverage, riskiness, profitability and capital adequacy on risk reporting quality (Barakat & Hussainey, 2013). Prior research suggests, however, that the board of a company also has an influence on the disclosure policy of the firm (e.g. Michelon & Parbonetti, 2012), since it is an outcome of the judgement, discretion and decision-making process of the board. This, in turn, is influenced by each members’ personal and professional characteristics (Katmon, Mohamad, Norwani & Farooque, 2017). Therefore, it is expected that the characteristics of board members have an impact on a company’s risk disclosures.
In general, a distinction can be made between two types of boards, namely a one-tier and a two tier-board structure. The difference is whether or not a separate supervisory board is present. The advantage of a one-tier board is reduced information asymmetry and faster decision making (Bezemer, Peij, Kruijs & Maassen, 2014), whereas a two-tier board performs better in their monitoring and resource provision role towards management (Bezemer et al., 2014). In this paper, both banks with a one-tier board and banks with a two-tier board will be taken into account. Regardless of the board structure (one/two-tier), the board of a company has two tasks, namely monitoring management and providing management with resources. The latter includes assisting in strategy formulation and in making other important decisions (Kaczmarek, Kimino & Pye, 2012). An example of such a decision is what risk information to include in the annual report and where to include it.
Disclosing more meaningful risk information could be beneficial to banks for several reasons. It could reduce stakeholder uncertainty regarding the expected future cashflows of the bank (Linsmeier, Thornton, Venkatachalam & Welker, 2002), which reduces the risk premium, resulting in a lower required rate of return by investors (Elshandidy & Neri, 2014). This leads to reduced cost of capital (Linsley & Shrives, 2005), since it enables the bank to lend bonds at lower interest rates and reduce the height of dividend payments to shareholders. Other reasons could be improved reputation, increased legitimacy (Oliveira, Craig & Rodrigues, 2011) and reduced stock price volatility (Baumann & Nier, 2004). However, providing more risk information could also have some negative effects for the bank, it could, for example, draw more attention to the firm’s riskiness (Elshandidy, Fraser & Hussainey, 2013). Increasing the attention of the market could lead to investors increasing their risk premium in order to compensate for the perceived higher risk exposure when disclosures reveal unknown contingencies and risk factors (Kravet & Muslu, 2013). Also, the disclosure of proprietary information could place a company at a competitive disadvantage when made public (Linsley & Shrives, 2005).
In conclusion, the decision regarding the form and content of the risk information in the financial statements is influenced by the composition of the board (Michelon & Parbonetti, 2012).Therefore, the research question
of this paper reads as follows: ‘What is the effect of board characteristics on the quality of risk disclosure by
European banks?’ In particular, the focus is on the board characteristics gender, board size and board
independence.
1.1 Scientific contribution
This paper contributes to risk disclosure literature in several ways. Firstly, it extends the limited existing literature regarding the quality of risk reporting by banks in response to the general call for more research into the determinants of the quality of risk reporting in financial institutions (Barakat & Hussainey, 2013). Previous studies show that a bank’s board characteristics have an influence on different types of reporting, like sustainability reporting (e.g. Michelon & Parbonetti, 2012; Dienes & Velte, 2015), greenhouse gas disclosure (e.g. Liao, Luo & Tang, 2015; Prado Lorenzo & Garcia-Sanchez, 2010), corporate social responsibility information (e.g. Jizi, Salama, Dixon & Stratling, 2014; Kiliç, Kuzey & Uyar, 2015) and voluntary disclosure in general (e.g. Bhasin, Makarov & Orazalin, 2015). This paper will contribute to this line of research by investigating whether board characteristics also have an influence on risk reporting by banks. Research on board characteristics as a determinant of the quality of risk reporting by banks is limited, but growing, stimulated by the increased demand for more disclosures. Over the past few years some studies conducted research about this subject, but mostly their empirical evidence is restricted to one country (Nahar, Azim & Jubb, 2016; Hemrit, 2018; Elgammal, Hussainey & Ahmed, 2018). This reduces the ability to generalize the findings across national boundaries (Samaha, Khlif & Hussainey, 2015). Therefore, this study will focus on banks all over Europe, which is in line with the research of Barakat and Hussainey (2013). This makes the findings more generalizable and it enables comparison between different countries.
Second, prior risk disclosure studies do not examine the impact of the board structure type in their research (e.g. barakat & Hussainey, 2013; Elgammal, Hussainey & Ahmed, 2018). In contrast, this study does investigate whether the type of board structure within a bank has an influence on the quality of their risk reporting. The impact of the variables under investigation in this study might be different in one- and two-tier structures, due to the difference in the decision making process of supervisory board members compared to the non-executive members of a board with a one-tier structure (Dienes & Velte, 2016).
Third, this study uses a unique dataset, obtained by hand collected data from annual reports of banks in Europe. This type of data collection tends to be more reliable, authentic and objective and it is specific for the variables investigated in this research (Kabir, 2016).
Finally, the findings of this study also have practical relevance, since it informs banks, banks’ stakeholders, users of the annual report, regulators, supervisors and other relevant policy makers concerned with developing disclosure rules about the impact of the investigated board characteristics on the quality of risk disclosures by banks. The outcomes of this study are especially important to the supervisory board of a bank in their nominations for new board members and to the shareholders that eventually appoint the members in the general meeting of shareholders. In addition, it is important to supervisors, like De Nederlandsche Bank (DNB), in assessing the expertise and integrity of a banks (nominated) directors (Lückerath-Rovers & Stavast-Groothuis,
2016). It informs these parties about what type of member is best for a bank in order to improve transparency regarding risk information.
The remainder of this paper proceeds as follows. Part 2 will discuss the theory and develops the research hypotheses. Part 3 discusses the method used to measure risk disclosure quality, describes the sample used and defines the independent and control variables. In part 4 the results will be presented and part 5 provides the conclusion, discussion, implications and limitations of the research.
2. LITERATURE REVIEW AND HYPOTHESIS DEVELOPMENT
The conceptual model used as a framework for this section is presented in Figure 1. First, the definition of risk reporting quality will be presented. Second, the theories relevant for this model will be described. Finally, the expected relationships between the percentage of women on the board, the amount of independent board members and board size with the quality of risk disclosure will be developed.
FIGURE 1 Conceptual Model
2.1 Definition risk reporting quality
The dependent variable of this research is quality of risk reporting. Risk disclosure can be defined as all information provided by a firm that describes its major risks and the expected economic impact of these risks on their future performance (Miihkinen, 2012). This information is important, since it provides insights about the continuity of the bank by presenting potential risks and uncertainties threatening the bank (Ibrahim & Hussainey, 2019). Risk disclosures by European banks are subject to complex regulations from different regulatory bodies like the International Accounting Standards Board (IASB), national central banks and national and European regulatory bodies like the Basel Committee on Banking Supervision, the European Banking Authority (EBA), the European Securities and Markets Authority (ESMA) and the Enhanced Disclosure Task Force (EDTF) (Jones, Melis, Gaia & Aresu, 2018). Although they all aim at improving transparency, banks still have a certain degree of freedom in the amount and place of risk information that they want to provide in their annual report (Linsley & Shrives, 2005). For example, as stated by International Financial Reporting Standard (IFRS) 7, set by the IASB, an entity must disclose information that enables the
users of its financial statements to evaluate the nature and extent of risks arising from financial instruments to which the entity is exposed at the end of the reporting period and how these risks are managed. Information about financial instruments is particularly important in the banking sector, because banks operate as an intermediary between lenders and borrowers in the capital market (Gebhardt, Riechardt & Wittenbrink, 2002). However, autonomy regarding how and where to disclose this information remains with the banks.
2.2 Theories regarding risk reporting quality
Researchers investigating corporate disclosure literature searched for explanations why companies decide to provide certain information in their annual report. Several social theories can be used to identify the motivations for corporate disclosure, namely agency theory, stakeholder theory, legitimacy theory, and signaling theory (Khalil & Maghraby, 2017).
The first theory that can be used in the area of research on reporting is the agency theory. The theory is based on the premise that a conflict of interest exists between the shareholders of a firm and its managers. One way of reducing this conflict of interest, and thereby the agency costs, is for the firm to disclose information (Jensen & Meckling, 1976). So, the disclosure of risk reports by European banks can decrease the information asymmetry with their outside investors, provided that a firm’s risk management is in no other way observable to them (Linsmeier et al., 2002). When considering risk, another agency problem exists in the banking sector, namely between the shareholders of the bank and its debt-holders. Shareholders prefer the bank to take high risk, since they can diversify the risk away. The preference of the debt-holders, on the other hand, is minimizing the risk (Felício, Rodrigues, Grove & Greiner, 2018). According to the traditional corporate governance approach, managers tend to be more aligned with shareholders and therefore, take greater risk (Felício et al., 2018). By disclosing information about risks, the bank will inform both parties and thereby reduce information asymmetry, which enables both parties to make better decisions.
The second theory, which expands the agency theory, is the stakeholder theory. This theory also includes the stakeholders to whom the bank has a responsibility, in addition to the shareholders (Freeman & Reed, 1983). Stakeholders are groups or individuals that are influenced by or can have an influence on the performance or actions of the firm (Freeman, 1984). So, they can affect the goals of the organization, which indicates that organizational performance might benefit from the activities and participation of its stakeholders (Kaur & Lodhia, 2018). According to this theory, banks might disclose risk information to more effectively interact and better communicate with influential stakeholders like large shareholders, governmental blockholders and bank supervisors (Barakat & Hussainey, 2013). Disclosing risk information could reduce stakeholders’ uncertainty by enabling them to better evaluate banks’ risk profiles (Jones, Melis, Gaia & Aresu, 2017).
A third theory, that is closely connected with stakeholder theory, is the legitimacy theory. It states that a social contract exists between the organization and those affected by the operations of the organization. The organization has to comply with the terms of this contract in order to continue its operations (Brown & Deegan, 1998). To ensure continuity and survival of the bank and to respond to pressure of society, banks voluntarily disclose detailed information in order to legitimize their operations (Barakat & Hussainey, 2013; Naser,
Al-Hussaini, Al-Kwari & Nuseibeh, 2006). According to Oliveira et al. (2011), banks disclose risk information to enhance their reputation among its stakeholders and, in turn, strengthen the confidence of its stakeholders. This is in line with Sánches-Ballesta and Lloréns (2010), who state that disclosing information is a market mechanism to create and sustain a bank’s reputation. In accordance, banks might disclose more voluntary risk information to reduce the risk of a bank run, where depositors suddenly withdraw their deposits when they suspect the bank to fail in the short future (Diamond & Dybvig, 2019). Banks disclosing more high-quality information enable customers and investors to make better decisions regarding whether or not to entrust the bank with their money (Huttenhuis & ter Hoeven, 2016). According to Jungherr (2018) another possible outcome of disclosing more risk information by banks is that the bank reduces its risk taking. The bank will anticipate that its portfolio choice is publicly observable and will therefore choose a relatively safe portfolio to prevent a bank run from happening.
The last theory, that is related to both legitimacy theory and agency theory, is the signaling theory. It is concerned with ways to address problems arising from information asymmetry (Kromidha & Li, 2019). Moreover, it is pursuant to legitimacy theory, because organizations should disclose information in order to signal that they are complying with societal expectations and norms (An et al., 2011). For banks it is important to signal their competence to society, since they need to regain the public trust. They have to convince society again that it is safe to entrust the bank with their money. For banks disclosing more risk-related information, uncertainty for investors reduces (Elshandidy, Guo & Neri, 2018) and the banks image will be improved (Bravo, 2017). Managers may choose to disclose risk information to signal their competence and capability about how they deal with risks and to distinguish themselves from their competitors (Al-Maghzom, Hussainey & Aly, 2016). In conclusion, since banks became more vulnerable to risks and disclosure could give a positive signal to society, it could improve societies’ opinion and improve the public trust.
These four theories are important, since they are all concerned with the survival of the bank (Khalil & Maghraby, 2017). As stated before, due to the financial crisis of 2008, society lost their confidence in the banking sector (Basel Committee on Banking Supervision, 2014). All over the world, the government had to save banks from bankruptcy by providing them with loans (Moise & Illie, 2012), or sometimes they even nationalized banks like ABN AMRO, Fortis and SNS bank, in order to rescue the financial sector (Kuo, 2013). Therefore, it is important for banks to clarify to society that they can be trusted again and that it is still safe to deposit money into their bank. They need to convince them that they still have a right to exist. This can be done by taking the interests of both shareholders and other stakeholders into account in order to keep their licence to operate and signal their competence and to avoid a bank run. Disclosing more and better risk information could help banks to regain trust of society (Cornelissen, 2014).
2.3 Board structure
A bank can have a one-tier or a two-tier board structure. The difference is whether or not a separate supervisory body exist. In a one-tier board, typical in for example the UK, Ireland and Spain (Jungmann, 2006), only one body exists, consisting of executive- and non-executive members, where the executive members participate in the daily management of the company and the non-executive members provide them with
independent oversight and advice (Millet-Reues & Zhao, 2010). In a two-tier board, typical in for instance Germany, Denmark and the Netherlands (Jungmann, 2006), two separate bodies exist, namely the management board and the supervisory board. Here, the management board runs the day-to-day operations (Millet-Reyes & Zhao, 2010). So, the executive members in a one-tier board can be compared with the members of the management board in the two-tier structure and the non-executive members in a one-tier board can be compared with the members of the supervisory board in the two-tier structure. In this paper, the latter are both addressed as board of directors.
The advantage of a one-tier board is the lower amount of information asymmetry due to less organizational layers. In a two-tier board information asymmetry is higher, which could lead to slower decision making (Bezemer et al., 2014). However, in a one-tier board, the ability to monitor executive directors and to provide independent advice to management may be weakened, which is less of a problem in a two-tier board (Bezemer et al., 2014).
Regardless of the board structure (one/two-tier), the board of a company has two tasks. First, it needs to monitor management to ensure they operate and make decisions in line with the interests of the shareholders. In a one-tier board, this is the task of the non-executive directors. In a two-tier board, the supervisory board oversees the management board (Millet-Reyes & Zhao, 2010). Second, they have a resource provision function, which includes assisting in strategy formulation and in making other important decisions (Kaczmarek, Kimino & Pye, 2012). An example of such a decision is what information to include in the risk disclosures and where to include it.
2.4 Board characteristics
The main role of the board of directors is providing management with advice, strategic support and guidance. In addition, they need to ensure that the firm operates in the best interest of the shareholders, while they have to consider the impact of the firm’s operations on society and environment at the same time (Dah & Jizi, 2018). Board diversity can be described as the variety in the composition of the board of directors (Kang, Cheng & Gray, 2007). Directors can, for example, differ on the basis of gender, age, nationality, ethnicity, tenure, professional experiences and education (Katmon et al., 2017). According to Post, Rahman and Rubow (2011), diversity in the board enhances the decision-making process due to the different types of knowledge domains, perspectives and ideas of the different members. It ensures that different viewpoints are considered, which increases creativity and innovation and leads to more effective problem solving and global relationships (Carter, Simkins & Simpson, 2003). According to Engen (2011), boards in the banking sector are generally too large in size, too male and that they are often dominated by insiders. Therefore, in this paper the focus is on the characteristics gender, size and independence. For all three characteristics, the effect of the board structure (one/ two-tier) will be considered. This enables comparison between the two different board structure types.
2.5 Gender
A much debated and researched topic is the influence of gender diversity in business and on corporate boards (Ellwood & Garcia-Lacalle, 2015). Gender diversity in the board refers to the presence of female directors on the board (Ellwood & Garcia-Lacalle, 2015). It can be measured by the percentage of women present in the board. Several studies in the area of psychology and management find gender-based differences regarding behavior towards for example risk taking and confidence (Byrnes, Miller & Schafer, 1999; Niederle & Vesterlund, 2007). Findings of Niederle & Vesterlund (2007) suggest that men are substantially more overconfident and competitive than women. According to Goel and Thakor (2008), overconfidence could lead to underestimating project risks. So, to reduce the risk of boards making value-destroying investment decisions and in order to temper the overconfidence of male board members, it could be beneficial to have some women on the board. Fernandez-Feijoo, Romero & Ruiz (2012) state that having only one woman on the board will not make a difference, because she will not feel free to give her opinion and will be less active and raise less issues. A board should at least contain three women to ensure gender is no longer a barrier for them to give their opinion (Konrad, Kramer & Erkut, 2008). Increasing the number of women could be beneficial for a bank for several reasons. First, Huse and Solberg (2006) state that women generally are more committed and involved, more diligent and that they create a better sphere within the board. In addition, Coffey and Wang (1998) state that female directors improve the decision-making process and increase board effectiveness, because it is less likely they put their own self-interest above that of others. Moreover, empirical evidence suggests that women are more concerned than men with taking actions to reduce perceived risk (Post et al., 2011) and they are better able to facilitate more informed decisions (Ben-Amar, Chang & Mcllkenny, 2017). According to the agency theory, the communication level between management and stakeholders can be improved when there are more female directors on the board (Katarachia, Pitoska, Giannarakis & Poutoglidou, 2018). In addition, as stated by stakeholder theory, the appointment of female board members could increase the image of the firm and, therefore, have a positive effect on the firm’s stakeholders. Further, signaling theory states that firms could use female representation on the board to signal that they are committed to create social value (Anca & Gabaldon, 2014) and thus build better public image (Saggar & Singh, 2017). In line with these statements, evidence of several studies shows a positive relationship between the presence of women on the board on the quality of risk disclosure in the annual report (e.g. Al-Maghzom et al., 2016; Saggar & Singh, 2017; Bravo, 2018). However, a board consisting of only women is not preferable, because then there will be a lack of good characteristics that men bring to the board, e.g. men provide the board with greater skill in math and spatial orientation tasks (Strydom, Yong & Rankin, 2017). In addition, Saggar and Singh (2017) found that a diverse boards, e.g. gender diversity, is expected to result in better governance. Moreover, Solal and Snellman (2019) state that firms hiring more female directors are perceived as more interested in social goals than their commitment to maximize shareholder value. In conclusion, it is expected that the different characteristics of men and women complement each other and therefore, the following hypotheses are drawn:
H1: The percentage of women on the board has a curvilinear, converted U-shaped relationship with the quality of risk disclosure.
H1a: The percentage of women on the one-tier board has a curvilinear, converted U-shaped relationship with the quality of risk disclosure.
H1b: The percentage of women on the supervisory board of the two-tier board has a curvilinear, converted U-shaped relationship with the quality of risk disclosure.\
2.6 Board independence
The board of directors usually consists of a mixture of inside- and outside directors, where the latter are presumed independent from management (Kang et al., 2007). They are not directly involved in the day-to-day operations, which allows them to provide management with more objective advice, focusing on firm performance and operations (Liao et al., 2015). In addition, CEOs have less influence over them, since their careers are not dependent on CEOs and they are less aligned with management than inside board members (Liao et al., 2015). In line with the agency theory, Carter et al. (2003), state that the presence of independent board members is critical to ensure the board acts in the best interest of shareholders. Outside board members, in contrast to inside members, have no close ties to agents and have less insider interests, which allows them to better monitor the behavior of managers and intervene when managers show opportunistic behavior (Post, Rahman & Rubow, 2011). In accordance, it enables them to question and evaluate management more objectively (Haque, 2017). Moreover, they are expected to represent shareholders in a less biased way due to the lack of conflicts of interest between the principal and the agent (Buckby, Gallery & Ma, 2015). In light of the stakeholder theory, independent board members are closer connected with stakeholders and are more aware of stakeholders’ expectations, which enables them to better address their demands (Dah & Jizi, 2018). In addition, outside directors are more stakeholder oriented because of their diverse backgrounds (Liao et al., 2015). In this study, directors are considered independent when they meet two criteria. They should not be employed by the company and they should not serve on the same board for more than ten years. Findings of, among others, Chen and Jaggi (2000) and Barakat and Hussainey (2013), show that the percentage of independent directors on the board is positively associated with the quality of performance-related disclosures. Taking into account the theory and previously conducted studies, the second hypotheses are as follows:
H2: The percentage of independent directors on the board has a positive relationship with the quality of risk disclosure.
H2a: The percentage of independent directors on the one-tier board has a positive relationship with the quality of risk disclosure.
H2b: The percentage of independent directors on the supervisory board of the two-tier board has a positive relationship with the quality of risk disclosure.
2.7 Board size
Past studies and several theories provide different arguments regarding the most favorable amount of members in the board. A board that is too small will experience high CEO influence, leading to higher agency costs. In addition, there will be a lack of members with adequate experience (Mokhtar & Mellet, 2013). When
the size of the board increases, stakeholder theory suggests that a larger stakeholder group will be represented, and the board will be in the possession of greater diversity in terms of expertise, experience and opinions. As demonstrated by Saggar and Singh (2017), this diversity leads to improved disclosure quality of banks. In addition, signaling theory suggests that firms with larger boards have a higher incentive to signal risk management performance to shareholders (Elzahar & Hussainey, 2012). However, when the board grows too large, there will be less coordination, worse communication and higher free rider behavior of directors (Jensen, 1993). Also, the effectiveness of the monitoring role of a board that is too large will decrease (Ntim, Lindop & Thomas, 2013) and the board will be less flexible, have less informal meetings and will have worse interpersonal relationships (Block & Gerstner, 2016). All in all, prior studies show that disadvantages exist, both when a board is too big and when it is too small. Therefore, the third hypotheses are formulated as follows:
H3: The size of the board has a curvilinear, converted U-shaped relationship with the quality of risk disclosure.
H3a: The size of the one-tier board has a curvilinear, converted U-shaped relationship with the quality of risk disclosure.
H3b: The size of the supervisory board of the two-tier board has a curvilinear, converted U-shaped relationship with the quality of risk disclosure.
3. METHODOLOGY
3.1 Sample
The sample used in this study is based on European banks included in the EBA transparency exercise (European Banking Authority, 2016). All disclosures provided by banks in the European Union have to comply to the same regulatory bodies, like IFRS, Basel III and ESMA, which enables comparison. The sample contains one bank from Norway, which is officially not a member state of the European Union. However, Norwegian banks have to comply to the same regulatory directives as European banks, due to the European Economic Area (EEA) agreement (Putnis, 2018) and will therefore be considered the same as an EU member country. The data comprises the years 2017 and 2018, because for these years sufficient data was available and it allows for comparison between these years. For the year 2018, using the annual reports of European banks, a risk disclosure quality index has been determined for each bank. For 2017 the risk disclosure quality index had already been gathered the same way. In addition, information about general financial results and governance are gathered for each bank. To enable comparison, all data regarding financial information was converted to euro’s by using the exchange rate of December 31 of the relevant year. To ensure no banks unnecessarily had to be excluded from the sample due to missing data, after completing the dataset, a last check was performed for both years to make sure the data was not available.
This resulted in an original sample consisting of 60 banks from 19 different EU member countries. One bank was excluded from this sample, because for 2017 no risk disclosure quality index information was available. So, the final sample consists of 59 banks from 19 different EU member countries. After including
the data of 2017, the sample consisted of 118 bank-year observations. Appendix A shows a breakdown of the final- and original sample divided by country. To investigate if it matters what type of board structure a bank has, the main sample is divided into two sample groups. The first, called the one-tier sample, only consists of banks with a one-tier board structure (n = 52) and the second, called the two-tier sample, consists only of banks with a two-tier board structure (n = 66). The breakdown per country for these samples is also presented in Appendix A.
3.2 Dependent variable
The dependent variable in this study is quality of risk disclosure by European banks. Prior accounting studies show that there are several ways to operationalize and measure the quality of risk disclosure in the annual report. For example, a method used by Allini, Rossi and Hussainey (2015) and Elshandidy et al. (2013), consists of identifying the number of sentences containing at least one risk related keyword. The higher the number of these sentences, the higher the risk disclosure quality. However, this refers more to the quantity of information, rather than the quality. Therefore, in line with Barakat and Hussainey (2013), this study uses another method, called content analyses. This method analyzes publicly available annual reports, enabling researchers to draw quantitative conclusions based on narrative documents (Vitouladiti, 2014). So, theoretical constructs are converted into quantitative data (Awasthy, Gopakuma, Gouda & Haldar, 2019), which provides this study with a unique dataset. A disclosure checklist was used to determine the disclosure quality index of each bank. The index consisted of thirty items which all carried the same weight, in order to mitigate subjectivity, and is based on Basel III, EDTF and ESMA. The focus was on general risk information, credit risk information, solvency information and liquidity information. For each item an ordinal score was granted. If certain information was not provided in the annual report, a zero was appointed. When limited information was provided, one point was given and when the information about a certain subject was provided comprehensively, two points were granted. The total of points given represents the risk disclosure quality index (RDQIndex). The higher the score, the better the quality of the risk disclosure in the annual report, with a maximum score of 60 points. The subjects taken into account to determine the risk disclosure score are described in appendix B. The data was hand collected by four students, which all analyzed 15 banks for the year 2018. To ensure consistency and objectivity, prior to the data collection, a meeting was held to discuss difficulties and to define unclear concepts. In addition, during the data collection, each student reviewed three or four banks of the other students to increase objectivity. Lastly, to ensure consistency between 2017 and 2018, checks were performed by two students for banks where the difference in the disclosure score between 2018 and 2017 was 10 points or more. If such a difference occurred, the score of 2017 was hand collected again and 2018 was checked again.
3.3 Independent variables
The independent variables of this study are all related to the board of the bank. Both the one- and two-tier boards will be considered and potential differences between the two types of boards will be investigated. First, no distinction will be made between the two types of boards, where both the supervisory board and
management board will be taken into account. Thereafter, the variables will be tested separately for banks with a one tier-board and for the supervisory board in banks with a two-tier structure. All data needed for this study is collected by consulting the annual reports or the governance reports of each bank.
The first independent variable of this study is gender. It refers to the mixture of male and female members on the board. In this study, the focus is on the relative amount of women present in the board, so the number of women is divided by the total amount of board members. One way of investigating the hypothesized curvilinear relation, is calculating an extra quadratic variable and add it to the linear model (Ganzach, 1997; Rapp, Rapp, Bachrach, 2013; Galen & Kloet, 2011). This is done by transforming the independent variable gender by multiplying it with itself and then performing a hierarchical multiple regression. The linear and quadratic terms are entered in the regression models sequentially. The gender squared variable represents the expected bend in the relationship. According to Patrashkova-Volzdoska, McComb, Green and Compton, (2003), a curvilinear relationship is present when both, the squared term and the regression model including the squared term, are significant.
The second variable is board independence. It refers to the relative number of independent members on the board, so the number of independent members divided by the total amount of board members. A member is considered independent when he or she has not served on the same board for more than ten years and he or she is not employed by the bank.
The final variable is board size. For a one-tier board this comprises the total amount of board members and for the two-tier structure, for each main hypothesis, the board members of the supervisory board and the management board will be taken together, whereas for hypothesis b, only the amount of the supervisory board will be taken into account. To investigate the expected curvilinear relationship with risk disclosure quality index, the same method is used as for the variable gender. So, an extra squared variable is calculated and added to the analysis.
3.4 Control variables
A few control variables will be included in this study, because previous studies about corporate risk disclosure show that they may affect disclosure quality of risk information. The first control variable is size of the bank (Barakat & Hussainey, 2013; Miihkinen, 2012; Baroma, 2014), which was measured as the log of total assets. The log was used because this allows for simpler representation of the otherwise large numbers. To keep the samples as large as possible, the extreme values were winsorized to reduce the effect of spurious outliers. The values for total assets of banks having total assets exceeding more than three times the standard deviation from the average were adjusted to the average plus three times the standard deviation. Prior studies mostly report a positive relationship between firm size and disclosure (e.g. Abraham and Cox,2007; Baroma, 2014; Miihkinen, 2012).
The second control variable is leverage of the bank (Naharet al., 2016; Maffei, Aria, Fiondella, Spanò & Zagaria, 2014; Amran, Bin & Hassan, 2009). The proxy used to measure this, is the leverage ratio (CRR) as proposed by the Basel Committee on Banking Supervision (2014), calculated by dividing the tier 1 capital by a bank’s average total consolidated assets. This control variable is taken into account since one of the key
drivers of the financial crisis of 2008 was banks funding a lot of long-term opaque loans with short term deposits, leading to an increased risk of a bank run (Dermine, 2015). During the financial crisis, this phenomenon gained a lot of attention, since it directly threatened the continuity of banks. It led to policy makers, like Basel, developing frameworks regarding capital structures of banks. The Basel Committee on Banking Supervision introduced the non-risk based leverage ratio as a new capital standard into the regulatory framework, where banks were required to fund at least 3% of their assets with equity (Basel Committee, 2014). As suggested by the agency theory, higher leveraged firms have to deal with higher agency- and monitoring costs. In order to reduce these costs, banks are expected to increase the level of information they provide in their annual reports (Jensen & Meckling, 1976). In line with this theory, most prior studies find a positive relationship between leverage and risk disclosure levels (e.g. Kolsi, 2012; Barako, Hancock & Izan, 2006; Egbunike & Tarilaye, 2017). However, Soyinka, Sunday & Adedejji (2017) find a negative relationship and Wang, Swon & Claiborne (2008) find no relationship.
Prior research reports that national culture influences the choices of management and the preferences of investors regarding risk reporting (e.g. Elshandidy et al., 2015; Elshandidy, 2011; Wong, 2012; Dobler, Lajili & Zéghal, 2016). It has an influence on the personal characteristics of board members, like their values, opinions, attitudes and approaches, and therefore, also on their orientation toward risk disclosure. In addition, it defines differences between various societies in terms of the desires, demands and preferences of the stakeholders within a certain society (Garía-Meca, Uribe-Bohórquez & Cuadrado-Ballesteros, 2018). It is possible that the orientation of individual members toward risk reporting is changed due to the cultural context. Therefore, the third, fourth and fifth control variables relate to culture, which is measured with the theory of Hofstede’s cultural dimensions. This theory allows for international comparison between cultures (Hofstede, 1983). In particular, this study will focus on three of Hofstede’s dimension, namely uncertainty avoidance, power distance and individualism, since prior studies found they were the most influential on risk reporting activities (Gray & Vint, 2012; Wong, 2012; Khlif, Hussainey & Achek, 2015; Tapang, Bessong & Effiong, 2012; Dobler et al., 2016; Orij, 2010). Appendix C shows the scores for the different countries regarding the three dimensions.
Besides these control variables, a dummy variable was included, namely reporting year, to investigate movements along the years.
3.5 Data analysis
Data analysis was used to determine the extent and direction of the relationships between the risk disclosure quality index and gender, board independence and board size. In order to analyze the data, SPSS was used. First, the descriptive statistics and the results of the Pearson correlation analysis will be presented, where the latter provides information about the degree of correlation between variables. Second, the data were further examined performing multiple regression analyses while controlling for relevant characteristics of the bank (firm size, leverage, national culture and year). The first and third hypotheses were examined by performing quadratic hierarchical linear regressions. The second hypothesis was examined by using an ordinary least
squares (OLS) regression. For the one- and two-tier sample groups the same regressions were performed for each variable.
The following model was used to test the hypothesis:
RDQIndex = β0 + β1GENDER + β2INDEP + β3BRDSIZE + β4FRMSIZE + β5LEV + β6UNAV + β7PDIS +
β8INDV + β9YEARDUMMY + ɛ
Where the β’s are the coefficients and ɛ is the error. Table 1 presents the description of the variables included in the model above. For the difference in board structure (hypotheses a and b) the formula is the same, only with different samples. The squared terms are left out of the formula for simplicity reasons.
4. RESULTS
4.1 Descriptive statistics and correlationsTable 2 shows the descriptive statistics per variable. In this table only the combination of both board structures (the main hypotheses) is included to provide better overview. In appendix D and E the descriptive statistics for both hypothesis a and b for each variable are presented respectively. Table 2 shows large variations exist among European banks regarding the risk disclosure quality index score, ranging from an 11-point minimum to a 45-11-point maximum. The average score is 28.20 11-points and the median 27.00 11-points, which are relatively low, considering the maximum score of 60 points.
As for the independent variables, on average, the board consists of 18 members, with 28% female and 51% independent members. Noteworthy is that there is a high variance in board size, with a minimum of 7 members
Table 1 Variables Definition
Variable Description Dependent variable
RDQIndex Risk disclosure score. Number of points a bank scored on its risk disclosure quality
Independent variable
GENDER Number of female directors divided by the board size INDEP Number of independent directors divided by the board size
BRDSIZE Number of directors on the board
Control variables
FRMSIZE Firm size, measured by the natural logarithm of total assets
LEV Leverage of the firm, measured by the ratio of tier one capital to consolidated assets
UNAV Uncertainty avoidance, degree to which a country is risk avoidant. The higher the score, the higher the preference for risk avoidance of a country
PDIS Power distance, degree to which members of society accept authority and associated (un)equal distribution. The higher the score the more unequal the distribution of power
INDV Individualism, degree to which society are focused on self-interest versus interest of a group. The higher the score, the more individualistic the society
and a maximum of 48 members. Developing a scatter plot, as presented in appendix F, shows that the sample contains only four bank years with such a high amount of board members. In addition, a low presence of female board members is observed, varying form a minimum of 5% to a maximum of 50%. Moreover, board independence shows a wide range of variation, with a minimum of 12.50% and a maximum of 81.82%. The sample consisted of banks with total assets varying between 26,454 million and 1,973,665 million euro, with an average of 438,205 million euro. The average leverage of the banks is 5.75% and the countries included in the sample score on average 64.93 on uncertainty avoidance, 44.24 on power distance and 69.02 on individualism versus collectivism.
Appendix D and E show some differences that are noteworthy. First, regarding the risk disclosure quality index, banks in the one-tier sample, on average, score better (μ = 30.9, M = 32.5) than the banks in the two-tier sample (μ = 26.12, M = 26.00). In addition, both samples show that the board on average has 14 board members. However, for the two-tier sample, this amount comprises only the members of the supervisory board. This indicates that the total board (supervisory- plus management board) in banks with a two-tier board structure are on average bigger than in banks with a one-tier board structure.
Table 3 presents the correlations among the independent, dependent and control variables. As stated by Yu, Jiang and Land (2015), multicollinearity problems could occur if the correlation score has a value of .7 or higher. Table 3 shows that the risk disclosure quality index is significantly positively correlated with gender (r = .327, p < .001) and independence (r = .287, p < .01) and negatively correlated with board size (r = -.085, p > .05), however, this last correlation is not significant. Further, the correlations among the independent variables and control variables do not show any problematic high values. What strikes is that some significant relationships exist between the independent and control variables, with the highest value for the association between individualism and gender (r = .383, p < .001), indicating that banks in more individualistic countries have a higher percentage of women in their board. The magnitude of the association is approximately low(.3 < r < .5) and is therefore not considered a problem. This is also true for the correlation between firm size and gender (r = .356, p < .001). The other significant correlations are not perceived as a problem, since the associations are all below .3, which indicates very low correlations. However, the values among the control variables do show some high significant correlations. For example, uncertainty avoidance and power distance (r = .808, p < .001), exceed the threshold of .7. This level of correlation could cause multi-collinearity, leading to reduced predictive power of the regression models (Orij, 2010). Appendix G and H show the correlations for the one-tier and two-tier sample respectively. For the one-tier sample, the correlation between uncertainty avoidance and power distance became even higher (r = .941, p < .001). Noteworthy is that, although they do not exceed the threshold of .7, the relationships between the percentage of independent board members and uncertainty avoidance (r = .516, p < .001) and power distance (r = .520, p < .001) became significant and increased to a higher value. In addition, the relationship between board size and power distance (r = .466, p < .001) became more significant and increased.
Lastly, a reliability analysis was performed on the disclosure index, comprising 30 items. Cronbach’s alpha was calculated and is presented in italics on the diagonal of table 3 (α = .736). As stated by Cronbach (1951),
an alpha above 70% is sufficient for a reliable disclosure index. Therefore, it can be concluded that high internal consistency of the disclosure index exists and that it reliably measures the risk disclosure quality.
N = 118 Table 3 Correlation Analysis Variable 1 2 3 4 5 6 7 8 1. RDQIndex .763 2. GENDER .327 *** 3. INDEP .287 ** .167 4. BRDSIZE (.085) (.145) (.199) * 5. FRMSIZE .591 *** .356 *** .171 .022 6. LEV (.289) ** (.251) ** (.144) (.026) (.353) *** 7. UNAV (.224) * (.215) * (.106) (.037) (.216) * .158 8. PDIS (.029) .088 (.139) (.204) * (.030) .016 .808 *** 9. INDV .265 ** .383 *** .191 * (.124) .338 *** (.257) ** (.559) *** (.247) ** *** p < .001, ** p < .01, * p < .05 Based on two-tailed tests
4.2 Regression analysis
To test the hypotheses, multiple regression analyses were performed. Table 4 presents the results of these analyses on the risk disclosure quality index. The table consists of five different models. The first model includes only the control variables firm size, leverage and the three culture dimensions. The second, third and fourth model include gender, independence and board size respectively. In line with Singh (1998), for the second and fourth model, priority was given to the linear relationships of gender and independence by entering the linear terms prior to the quadratic terms, because the linear associations are more generally accepted than the curves this study proposes. Since entering the squared terms to the regression led to nonsignificant regression models, only the betas of the squared variables are presented below the results of the linear
Table 2 Descriptive Statistics
Variabele Mean Median Standard Deviation Minimum Maximum
RDQIndex 28.21 27.00 7.79 11 45 GENDER 27.80 28.13 10.45 5 50 INDEP 51.14 53.33 19.39 12.50 81.82 BRDSIZE 18.63 16.50 7.54 7 48 FRMSIZE 438205 221004 498040 26454 1973665 LEV 5.75 5.20 2.00 2.50 15.60 UNAV 64.93 65.00 23.16 23 100 PDIS 44.24 38.00 15.89 11 68 INDV 69.08 71.00 13.59 27 89
regressions in models 2 and 4. The fifth model includes all independent and control variables and here only a linear regression is performed. In all five models, the dummy variable year is included. The table also includes the variance inflation factors (VIFs), which indicates the degree of multicollinearity between variables. Commonly it is accepted that the value of VIF should not exceed 10 (O’brien). The largest value of all variables included in the model is used as proxy of severe multicollinearity. As shown in table 4, none of the VIFs exceed this threshold, so no indication of multicollinearity exists.
4.3 Control variables
The first model in table 4 is the baseline model, which includes only the control variables firm size, leverage, uncertainty avoidance, power distance and individualism. The results show that this model significantly increases the explained variance of risk disclosure quality index (∆R2 = .363, p < .001). Besides, the results
show that only firm size has a significant positive relationship with the risk disclosure quality index (β = 8.472, p < .001), which is also true for the other five models with betas of 8.074, 8.029, 8.596 and 7.786 respectively. This indicates that the larger a bank, the higher the risk disclosure quality index. In addition, models 3 (β = -.124, p < .05) and 5 (β = -.111, p < .05) show a significant negative relationship between the level of uncertainty avoidance and the risk disclosure quality index, suggesting that banks in countries where uncertainty avoidance is lower, the risk disclosure quality is higher. Lastly, model 3 shows a positive significant relationship between power distance and the risk disclosure quality index (β = .143, p < .05), suggesting that banks operating in countries with a high degree of power distance, score better at the risk disclosure quality index.
The other control variables do not show significant associations with the risk disclosure quality index.
4.4 Gender
Hypothesis 1 suggests that a curvilinear, converted U-shaped relationship exists between the percentage of women on the board and the risk disclosure quality index. Model 2 includes the independent variable gender and the results show that this model significantly increases the explained variance in risk disclosure quality index (∆R2 = .366, p < .001). The results show, however, that neither the linear (β = .080, p > .05) nor the
curvilinear (β = .000, p > .05) association was significant, so no support is found for hypothesis 1. 4.5 Independence
Hypothesis 2 suggests that a positive relationship exists between the percentage of independent members on the board and the risk disclosure quality index. Model 3 includes the independent variable independence and the results indicate this model significantly increases the explained variance in risk disclosure quality index (∆R2 = .401, p < .001). The adjusted R square of this model is higher compared to the other models, indicating
that including the variable independence in the model leads to higher explanation power for the variance in risk disclosure quality index. In line, the results show that a significant positive relationship exists between the percentage of independent board members and the risk disclosure quality index (β = .086, p < .01). Consequently, an increase in the percentage of independent board members leads to an increase in the risk disclosure quality index, supporting hypothesis 2.
*** p < .001, ** p < .01, * p < .05 N = 118
4.6 Board size
Hypothesis 3 assumes that a curvilinear, converted U-shaped relationship exists between the size of the board and the risk disclosure quality index. Model 4 includes the independent variable board size and the results suggest that this model significantly increases the explained variance in risk disclosure quality index (∆R2 = .363, p < .001). In line with model 2, first the results of the linear regression are presented, suggesting a
negative relationship between board size and risk disclosure quality index (β = -.081, p > .05), however the association is not significant. Moreover, after entering the quadratic board size term to the regression, the results suggest that the relationship was insignificant (β = .008, p > .05), thus no support is found for hypothesis 3.
4.7 Board structure
Appendix I and J show the results of the regression analyses for the one-tier and two-tier samples respectively. Both appendixes show some noteworthy results. The results in Appendix I, addressing the banks
Table 4 Regression analysis
Model 1 Model 2 Model 3 Model 4 Model 5
Intercept (13.766) (13.629) (14.369) (12.117) (13.495) Controls FRMSIZE 8.472*** 8.074*** 8.029*** 8.596*** 7.786*** LEV (.272) (.241) (.196) (.293) (.183) UNCAV (.099) (.087) (.124)* (.090) (.111)* PDIS .106 .088 .143* .087 .122 INDV (.020) (.031) (.046) (.026) (.054) Independent GENDER .080 .061 INDEP .086** .080* BRDSIZE (.081) (.021) Squared GENDER x GENDER .000 BRDSIZE x BRDSIZE .008 R2 .396 .404 .437 .401 .442 ∆R2 .363*** .366*** .401*** .363*** .396*** F-value 12.104*** 10.659*** 12.187*** 10.520*** 9.513*** VIF 4.609 4.762 4.737 4.728 5.113
with a one-tier structure, show that the adjusted R squared for all models increased to values above .5 with the highest value of .566, indicating that the explanatory power of the regression analyses increased compared to the results of table 4 and that it is relevant to split the sample up into two individual samples. In addition, the results show that, besides the significant positive relationship of the control variable firm size, which is in line with the main variable, the level of uncertainty avoidance has a significant negative influence on the risk disclosure quality index for all models, not only models 3 and 5. Moreover, the association with power distance is significant for each model on the .001 level. This indicates that banks with a one-tier board, operating in higher power distant countries, show a higher risk quality disclosure index. In addition, the betas for both variables power distance and uncertainty avoidance in each model increase, indicating that they are more influential on the risk disclosure quality index. Also, although not significant, the betas for leverage increase and become positive instead of negative for all models. Another striking point in these results is that no significant relationship is found regarding the percentage of independent board members, in contrast with the main sample. Moreover, the directions of the relationships with gender and board size both change compared to table 4. However, Appendix H show very high VIF values for all models, with the highest value of 21.329, indicating that multicollinearity problems exist. This could reduce the reliability and predictive power of the results presented in Appendix H. Therefore, later in this study, a robustness check was performed, leaving out power distance as a control variable.
In addition, the results of Appendix J, addressing only the banks with a two-tier structure, also show deviant results compared to table 4. Here, besides the variable independence, the independent variable gender also shows a significant positive relationship (β = .241, p < .01). This suggests that an increase in the percentage of women in the supervisory board, leads to an increase in the risk disclosure quality index. However, the quadratic variable is insignificant (β = .004, p > .05), so no evidence is found for a bend in the relationship. Therefore, these results do not support hypothesis 1b, but they do suggest a positive linear association. In addition, model 2 shows a negative significant relationship with leverage (β = -.792, p < .05), indicating that an increase in leverage leads to a decrease in the risk disclosure quality index. Lastly, the relationships with uncertainty avoidance and power distance became insignificant and, except for model 3, the direction of the relationships with uncertainty avoidance changes to positive instead of negative compared to table 4.
4.8 Robustness test
To examine the models’ robustness, additional tests are performed. First, to overcome the problem of multicollinearity and reduced predictive power of the regression models that arose in the one-tier sample, another regression is performed, excluding the variable power distance from the analyses. This variable was excluded since the correlation analysis for the one-tier sample showed the highest number of significant correlations between power distance and the other variables, namely with board independence, board size, uncertainty avoidance and individuality. The results of the regression analyses are presented in Appendix K and they shows that the VIF values all decreased to an acceptable level, with the highest value of 2.914. So, this model resolved the problem of multicollinearity. The regression did lead to some differences compared to the original regression shown in Appendix I. First, the adjusted R squared decreased, but is still higher than in
the main analysis, with the highest value of .438, indicating that it is still relevant to divide the sample in the one- and two-tier sample. Second, the relationship with leverage became negative for all models, in line with the main regression analysis and the two-tier regression analysis. In addition, although with a lower value and on a lower significance level, uncertainty avoidance still shows significant negative relationships with the risk disclosure quality index, in line with the original one-tier regression analysis and different from the main- and two-tier regression analyses. Last, although not significant, model 3 suggests that the relationship with independence becomes negative.
In addition, since some researchers use a threshold of 5 for the maximum VIF value (Yu et al., 2015) instead of the threshold of 10 used in this study, a test is performed to control for multicollinearity in the main regression analysis. As shown in Table 4, model 5 exceeds the threshold of 5 (VIF = 5.133) and the other models also show VIF values close to 5. Therefore, the regression analyses are executed again, leaving out the variable power distance to investigate whether this leads to different results. The results are presented in Appendix L, showing that the VIF values decrease below the threshold of 5, with a highest value of 1.725. The results show that the significant relationship with uncertainty avoidance in model 3 is no longer significant. Moreover, the significance level of independence in model 3 devaluates to the .05 level. Since the VIF values for both the main regressions and the one-tier regression decreased, these results are expected to be more reliable and therefore, in the remainder of this paper, these results will be used.
Moreover, since Konrad et al. (2008) stated that the influence of the presence of women would only influence disclosure when there are at least three women in the board, another test is performed. All banks with less than three women were excluded from the sample, however, this did not lead to different results.
Last, since the scatter plot, presented in Appendix F, shows four bank years with exceptionally high board sizes, a check is performed to rule out the influence of outliers. Banks with more than three times the standard deviation from the mean were excluded. The unreported results did not show any differences from the reported results.
5. DISCUSSION AND CONCLUSION
Recent studies show that different factors, like board characteristics, have an influence on the risk disclosure quality of banks (e.g. Elgammal et al., 2018). This study contributes to this line of research by examining the effect of board gender diversity, board independence and board size on the quality of risk disclosures by European banks, while controlling for firm size, leverage and national culture. Predicted is that both gender diversity and board size have a nonlinear, converted U-shaped relationship with the quality of risk reporting and that board independence has a positive association with quality of risk reporting. The study also investigates whether the type of board structure (one/two-tier) influences these relationships. Based on hand collected data of 118 bank year observations for 59 banks from 19 different countries for the years 2017 and 2018, the current study finds evidence that board independence is indeed positively associated with the quality of risk disclosure. However, the findings of this study do not support the hypothesized relationships for gender
diversity and board size. Interestingly, dividing the sample into two groups in order to separately analyse the findings for banks with a one-tier board and banks with a two-tier board, show some different results.
First, for the main- and two-tier sample, no evidence is found for a relationship between gender and quality of risk reporting. However, for the two-tier sample, the results show a positive association, in line with, among others, Saggar and Singh (2017).A cause for the insignificance in the main- and one-tier sample regarding the relationship could be that female board members are still underrepresented in the boards included in the sample. Due to the limited number of boards having more than 33% female directors, analysing their influence on risk reporting is hard (Lückerath-Rovers, 2011). Perhaps if there had been more observations of boards with higher female representation, a significant relationship for all samples could have been established.
Second, this study shows that an increase in the percentage of independent board members is associated with an increase in the risk disclosure quality index. This is in line with Barakat and Hussainey (2013). However, for the one-tier sample, no evidence is found for the relationship, in line with Akbas (2016). A reason for this could be that in a one-tier board, the independence of the independent members could become questionable (Adams, Hermalin & Weisbach, 2010). It is possible that they establish closer relationships with the executive directors (Moor, 2014), since they operate in the same board as the executive members. Another explanation for these mixed results regarding gender diversity and independence could be, as stated by Michellon and Parbonetti (2010), that using tradition proxies like dependent versus independent and male versus female, might be insufficient for studies on disclosure quality. They suggest that focussing more on individual characteristics of board members like personality, mentality, risk aversion and other personal traits, might show different results.
Third, no significant effect is found for the association between board size and quality of risk disclosure. This is in line with Cheng and Courtenay (2006), who hypothesize no relationship exists between board size and voluntary disclosure. They state that a small board lacks monitoring capabilities, whereas the benefits of a larger board may be offset by the decreased quality of communication and decision-making.Therefore, they state that, no preponderance of theory for board size exists. This explanation suggests, however, that an optimal board size exists, but the results of the present study do not support this. The insignificance of the results may be driven by the fact that the sample did not include enough banks with larger board sizes, as shown in the scatter plot of Appendix F.
Altogether, the results of this study have some important theoretical and practical implications. From a theoretical perspective, it shows that characteristics of the board should be analysed with more detail. Proxies for board characteristics should go beyond the homogeneous distinction between male and female and dependent and independent. Members of the board should be assessed based on more detailed characteristics like for example their backgrounds, mentality, education and competences. Furthermore, since most banks included in the sample are large international corporations, it is questionable whether national culture of the country where the bank is positioned is a good proxy to measure the influence of culture on board members attitude towards risk disclosure. It is possible that the culture of the country where the mother company is seated influences the board of its daughter companies located in other countries, since a board can consist of national as well as international members.
