The influence of governance committees on the quality of risk disclosure: a European banking perspective

(1)

The influence of governance committees

on the quality of risk disclosure:

a European banking perspective

Master thesis, MSc Accountancy

University of Groningen

Faculty of Economics and Business

A. Bonestroo

S2732203

January 17, 2020

Korte Jufferstraat 42

3512 EZ Utrecht

Supervised by

J.G. (Job) Huttenhuis

prof. dr. R.L. (Ralph) ter Hoeven

(2)

Page | 1

ABSTRACT

The objective of this research paper is to assess the influences of governance

committee characteristics on the quality of risk disclosure for European banks after the

implementation of Basel III. The proxies used are size, activity, independence and expertise

of the audit- and risk committee. The dependent variable, risk disclosure quality, is measured

by a disclosure index, comprised of elements of relevant laws and regulations. This research

is performed based on an unique dataset since the data is collected manually. The study takes

into account the years 2016-2018, focusing on an original sample of sixty banks across

Europe, by investigating the relationships with a multiple linear regression model. Our

findings provide evidence that risk committees fully occupied by independent members have

a significant positive impact on risk disclosure quality.

Keywords: Banking, Basel III, Risk Disclosure Quality, Audit Committee, Risk Committee,

Voluntary Disclosure Theories, Resource Dependence Theory.

(3)

Page | 2

4.7 Statistical model ... 23 FINDINGS ... 25 5.1 Descriptive statistics ... 25 5.2 Correlation analysis ... 26 5.3 Regression analysis ... 30 5.3.1 Hypothesis 1 ... 30 5.3.2 Hypothesis 2 ... 30 5.3.3 Hypothesis 3 ... 31 5.3.4 Hypothesis 4 ... 31 5.3.5 Hypothesis 5 ... 31 5.3.6 Hypothesis 6 ... 31 5.3.7 Hypothesis 7 ... 32

5.4 Validity disclosure index ... 36

CONCLUSION ... 36

(4)

Page | 3

6.2 Limitations and future research... 38

REFERENCES ... 39

APPENDICES ... 48

(5)

Page | 4

INTRODUCTION

A significant regulatory failure is a precondition for a major financial crisis (Coffee, 2009). According to Blundell-Wignall and Atkinson (2008), this is exactly what happened in the global financial crisis of 2008: instead of acting as a line of defense, the regulatory banking framework contributed or even exacerbated the crisis. The regulatory framework under which banks conducted their activity prior to the onset of the financial crisis has been subject of substantial criticism (Abreu & Gulamhussen, 2015). One of the main criticisms was the quality and level of the required capital that banks had to held, which turned out to be insufficient for the survival of some banks. Taking risks is a way to achieve returns and so is the case for the banking industry. The downside of the choices related to risk taking, is that risks can result in losses, which have to be absorbed by the banks. The difference with other businesses and what made it especially for banks so important to hold capital, is that the risks taken are, among others, funded with (consumer) savings and bank deposits. Banks therefore have to hold capital and meet with the laws and regulations. As a reaction to the regulatory failure, the European Union adopted a legislative package of reform measures in 2013, consisting of the Capital Requirements Regulation and the Capital Requirements Directive, respectively CRR (2013) and CRD IV (2013), commonly referred to as Basel III 1_{. These include measures with the purpose to strengthen} the regulation, supervision and risk management of banks.

On the 11th_{of February, 2019 the European Systemic Risk Board (ESRB) issued a report with a} remarkable conclusion: “Banks have not become measurably safer, despite the new increased capital requirements.” Ernest Dautović, economist and researcher at the ESRB concluded that European banks tend to show moral hazard when faced with higher capital requirements. As a result of new requirements banks have to wield stricter loan criteria, which reinforces the competition in the banking industry and causes increased risk-taking, a phenomenon also described by Bolt and Tieman, 2004. Especially large banks with low profitability increased their risk taking and managed to hide those risks (Dautović, 2019). In other words, the increased risk taking crowds-out the favourable development of increased capital. It looks like that banks have not learned from the previous financial crisis and the question now is: do banks at least improve the quality of risk reporting to inform their stakeholders since the increased risk taking?

According to Madison, Gregory, Friesen and Horner (2015) it is the responsibility of a bank’s board of directors to set an appropriate tone at the top about organizational risk appetite, taking into account the regulatory environment related to risks. In addition, risk oversight, risk assessment, monitoring and advising the company’s day-to-day managers are responsibilities of the ‘full board’ (COSO, 2009).

(6)

Page | 5 Full board responsibility also improves the alignment about risk management and the company’s

overall strategy (Beasley, Frigo & Litman, 2007). On the 11th_{of June, 2018 the Harvard Law School on corporate governance and financial regulation}

published a report about the 25 largest banks of Europe, showing that boards have become smaller and that the number of supporting board committees has consistently grown over the years

(Andersson, 2018). Since board size is decreasing and there is no reason to believe that the amount of work for board members has decreased too, the topic originates of what to delegate and to whom. This results in the question: should risk management still be a full board function, or should it be appointed to either the audit committee or a risk committee (Protiviti, 2011). A reason for the growing number of committees is partly driven by legal requirements, like the implementation of a risk- and/or audit committee as required by CRD IV and also the increasing complexity of the business environments (Kolev, Wangrow, Barker & Schepker, 2019). Specialized committees enable boards to cope with this complexity, because work is divided into committees’ expertise’s (Kolev et al., 2019). Boards are therefore discharging their responsibilities to supporting committees. The importance of these committees and the role of disclosure are as such increasing.

In the field of accounting, disclosure is defined as “informing the public by financial statements of the firm” and is about the communication of economic information (Ağca & Önder, 2007). The content of information can be financial or non-financial and can take quantitative or qualitative forms (Owusu-Ansah, 1998). Besides this grouping, there is a distinction between mandatory and voluntary

disclosure (Shehata, 2014). According to Shehata (2014), mandatory disclosure is when information is communicated when it is required by laws and regulations. On the opposite, voluntary disclosure is about providing information deemed relevant to decision needs of users (Meek, Roberts & Gray, 1995) and that is “recommended” by an authoritative code or body (Hassan & Marston, 2010). In this research we focus on the quality of voluntary risk disclosures at European banks and the role of supporting governance committees therein. As a starting point, this study will use the agency theory (Jensen & Meckling, 1976), stakeholder theory (Freeman, 1984), legitimacy theory (Dowling & Pfeffer, 1975), signalling theory (Spence, 1978) and resource dependence theory (Pfeffer & Salancik, 1978). These theories will explain and elaborate the multiple functions disclosures can fulfil and provides the incentives for banks to disclose information about risks. The key premises in this research are about reducing information asymmetry, discharging accountability and earning a license to operate. Accountability refers to the responsibility of an organization to disclose information regarding its performance, financial position, financing and investing, and compliance in order to assist users to make appropriate decisions (An, Davey & Eggleton, 2011). Banks might disclose risk information to effectively communicate with important stakeholders, like shareholders and regulators (Barakat & Hussainey, 2013).

(7)

Page | 6 The quality of risk reporting is therefore an important variable for boards to discharge accountability, reduce information asymmetry and enhance transparency. The research question central in this study is therefore: “To what extent and how do governance committees influence the quality of risk disclosures of European banks?”

SCIENTIFIC CONTRIBUTION

Business risks have always existed, but due to major recent scandals there is an increased interest in risk reporting (Oliveira, Rodrigues & Craig, 2013). The results about the determinants of risk disclosures are mixed, according to empirical literature (Khlif & Hussainey, 2016). Prior research finds different determinants of risk disclosure quality, like corporate size (Ashbaug-Skaife, Collins & Kinney Jr, 2007; Dobler, Laijili & Zéghal, 2011), profitability (Elzahar & Hussainey, 2012; Ntim, Lindop & Thomas, 2013; Al-Shammari, 2014), board independence and ownership concentration (Garcia-Meca & Sanchez-Ballesta, 2010) and industry type (Cooke, 1992; Madrigal, Guzmán & Guzmán, 2015; bin Shamsuddin, 2018).

As stated by Hines, Masli, Mauldin and Peters (2015), the global financial crisis prompted an increasing number of firms to create board-level risk oversight committees. In the field of supporting governance committees, as in this research is the audit committee and the risk committee, not much research has been done in relation to specifically the quality of risk disclosures. This is strange, because regulators promote internal committee use to facilitate board work (De Andrés, Arranz-Aperte & Rodriguez-Sanz, 2017) and the EU Commission launched an initiative to intensify the use of board committees (European Commission, 2005).

Studies about audit committees are in most cases focused on the effect on earnings quality and earnings management (Wild, 1996; Baxter & Cotter, 2009; Puat Nelson & Devi, 2013), financial restatements (Abbott, Parker & Peters, 2004; Schmidt & Wilkins, 2012; Wan Mohammad, Wasiuzzaman, Morsali & Zaini, 2018) or accounting conservatism (Lim, 2011; Sultana, 2015). Research focused on the impact of the audit committee on the quality of risk disclosure at banks after the financial crunch has not been performed yet. With respect to the risk committee, some research has been done on the effect on firm performance (Abdullah & Shukhor, 2017) and cost of equity capital (Al-Hadi, Hussain, Al-Yahyaee & Al-Jabri, 2018). However, the effectiveness of the committees is still unclear related to risk disclosure and empirical evidence to support the

recommendation of implementing risk committees is scarce (Abdullah & Shukor, 2017). Even large companies, like Lehman Brothers, went bankrupt despite the fact they had implemented risk

committees (Bates & Leclerc, 2009). All in all, the extant literature provides inconclusive evidence with respect to the characteristics of board sub committees and transparency related to risks (Buckby, Gallery & Ma, 2015; Abdullah, Shukor & Rahmat, 2017; Al-Hadi, Al-Yahyaee, Hussain & Taylor, 2017).

(8)

Page | 7 In this research we contribute to existing literature by investigating the impact of the risk committee and the audit committee after implementation of the Basel III requirements. One of the requirements of the Basel Committee is: “The risk committee should meet periodically with the audit and other

risk-relevant committees to ensure effective exchange of information and effective coverage of all risks, including emerging risks and any needed adjustments to the risk governance framework of the bank in the light of its business plans and the external environment.” Additionally, the Federation of

European Risk Management Associations (FERMA) state: “Cooperation between the audit committee and the risk committee is crucial to ensure a common risk management approach.” Concluding from the abovementioned statement: there is a high emphasis on the committees, which opens up areas for research.

If a relationship is determined between the quality of risk disclosure and the proxies used for the governance committees, this will be a valuable insight for the European banking sector and their supervisors, because the usefulness of the committees in their ‘risk-role’ has been confirmed. Besides, it is a confirmation that discharging of the risk accountability from the board to the governance committees successfully took place. Society has its expectations after the financial crisis and may regain trust once the influence and usefulness of the committees is proved. Besides the usefulness of the committees itself, the composition of how the committees should be structured (e.g. independence, expertise) and operating (e.g. number of meetings) is also a new insight and contributing to existing literature on risk disclosure. This study also contributes to current research due to the multi-year focus (2016-2018) to see the development of the disclosure quality in the years after implementation of the Basel requirements. Again, once this development is established shareholders, regulators, clients, society and other stakeholders may regain their trust in the banking sector if it is shown that the new regulation does make sense and that banks have become more transparent. As the effectiveness is established, this might open up areas for new research regarding board supporting committees and may be of interest for other industries as well.

2.1 Structure

The paper is structured as follows: the theoretical framework is elaborated in chapter 3. The main concepts of the theories are explained and are used as the foundation of this research. In combination with findings from previous research this will lead to the development of the hypotheses. Chapter 4 defines the methodology section and will explain the research design. Among other things, the proxies for the different variables, the data collection process and the statistical analyses will be treated. Chapter 5 covers the results of the analyses. Lastly, chapter 6 discusses the conclusion, limitations and recommendations for further research.

(9)

Page | 8

THEORETICAL FRAMEWORK

3.1 Theories

As a basis for this research, common and widely-accepted scientific theories are used. These are the agency theory, stakeholder theory, legitimacy theory and signalling theory, which are frequently used in disclosure research.The reason for these numerous and particular theories are the overlapping key premises between them. Since the similarities and to emphasize on the coherence, some of the

theories are combined in our integrated overarching framework. Next to the beforementioned theories, brief attention is given to the resource dependence theory. The structure of the theoretical foundation starts from a single perspective (agency theory, i.e. shareholder view), broadens by taking

stakeholders and society into consideration (stakeholder and legitimacy theory, i.e. stakeholder management) and ends up with a wide understanding (signalling and resource dependence theory) of the underlying incentives for banks to disclose voluntary information.

3.1.1 Agency theory

A renowned theory in the field of corporate governance and reporting is the agency theory (Jensen & Meckling, 1976). Central in this theory is the principal-agent relation. Jensen and Meckling (1976) refer to this as an entity (the principal) who appoints and engages another entity (the agent) to act on behalf of the principal. The concerns of the principal and agent are not aligned, resulting in a “conflict of interest” (Shehata, 2014). A key concept that plays a vital role here is information asymmetry, which is an information advantage over the other party (An et al., 2011). Information asymmetry may lead to two well-known phenomena in business research: moral hazard and adverse selection. In the study of Gershkov and Perry (2011) they define these concepts as follows: “moral hazard arises when the principal cannot observe the agent’s effort and adverse selection from the fact that the principal cannot observe the agent’s quality or the task’s complexity” (Gershkov & Perry, 2011). Consequently, this results in agency problems and related agency costs: monitoring costs (principal), bonding costs (agent), and residual loss (Jensen & Meckling, 1976). By making use of reporting and disclosure, the information gap shrinks, transparency enhances and opportunistic behaviour reduces (Li, Pike & Haniffa, 2008). Increasing voluntary disclosure by managers is a way of mitigating agency problems, reducing the agency costs (Barako, Hancock & Izan, 2006; An et al., 2011) and also a convincing factor for the external users for the optimal working of managers (Watson, Shrives & Marston, 2002), since corporate transparency is a signal of management quality (Katarachia, Pitoska, Giannarakis, Poutoglidou, 2018). Reducing information asymmetry is the driving force behind voluntary disclosure (Ben Prince & Dwivedi, 2013). This has become a necessity for the banking industry since the

(10)

Page | 9 3.1.2 Stakeholder & legitimacy theory

As an extension to the agency theory, Freeman (1984) developed a theory that concerns about more than just the shareholders. The influence of other stakeholders is increasing and therefore firms have to focus on their needs as well (Huang & Kung, 2010). The underlying idea of the theory is that the organization’s success is dependent on the way it manages the relationship with key groups

(stakeholders), such as employees, customers, financiers and society (Freeman & Phillips, 2002). Stakeholders derive their power from possessing resources that are needed by the organization (Freeman, 1984). Therefore, the relationships between the firm and their stakeholders is reciprocal (Wernerfelt, 1984) and have to be strategically managed (Herbohn, Walker & Loo, 2014). Besides the concept of information asymmetry, a key concept is accountability, which is already defined in the introduction section. To discharge accountability to the various stakeholders and to maintain their relationships, (accounting) information is considered as an important means for organizations (An et al., 2011). By communicating that information, the firm provides in the need of the stakeholders (Wernerfelt, 1984). According to Benlemlih, Shaukat, Qiu and Trojanowski (2018) the stakeholder theory suggests that if firms make extensively use of disclosure, it can help in building a positive reputation and gain trust among their stakeholders. Related to the financial crisis, this is one of the problems that banks currently have to cope with. One of the key concepts of the stakeholder theory refers to accountability and the responsibility of banks in disclosing information. Since this research aims at the disclosure of risks (determinants), this theory is fundamental to develop the hypotheses. It is focusing on information asymmetry as well and is therefore complementing to the agency theory. From stakeholder theory, it is only a small step towards the legitimacy theory of Dowling and Pfeffer (1975).

Legitimacy theory is an important driver for corporate disclosure, following from extensive research (Neu, Warsame & Pedwell, 1998). The bottom-line of the theory is that organizations have to be sure that their operations are perceived as legitimate by the stakeholders and society (An et al., 2011). When a company’s values are not matching with societies’, the theory assumes that the firm has no right to exist (Shehata, 2014). Stakeholder management is, as in line with the stakeholder theory, also for this theory an important means. There is a ‘social contract’ between society and the company, which can be revoked by society when the values are no longer aligned. By disclosing expected risk information, companies fall within the boundaries of the stakeholders and will be perceived as legitimate. Because the attitudes and values of society may change over time, organizational legitimacy is dynamic and may change too (Deegan, 2006). Firms are continuously seeking to align their activities with the requirements of society, but there often appears to be a gap between the organizations’ acting and the perception of society: the legitimacy gap (An et al., 2011). To keep this gap as small as possible, firms communicate and legitimise themselves via mandatory or voluntary disclosures (Lightstone & Driscoll, 2008).

(11)

Page | 10 Related to the banking industry, management policies should connect the expectations of stakeholders and society via transparency (Geraats, 2002). Besides management of expectations it is also about regaining trust in the banking sector (Braun, 2016). Communication about risk-taking is a vital part in narrowing the legitimacy gap and demonstrating the organizations’ legitimacy after the reputational damage and the decrease in trust. Disclosure and transparency have a negative intonation for banks, but on the other hand, they are also benefitting from higher disclosure quality, since they become more attractive for clients. Banks are dependent on the resources possessed by their stakeholders and therefore want to be as attractive as possible. Bank’s resources are derived from different channels (i.e. stakeholders), which can have multiple forms. The most common ones are savings (e.g. private, governmental or company savings), deposits, interest on loans/mortgages and loans taken out by the bank itself. By reinvesting these resources against more favourable interest rates than they pay to their capital providers, banks aim to make a profit. Explicitly for banks, those resources are indispensable to continue their business and operations. In the absence of these resources, there are no longer any means to invest, with the result of no inflow of cash. Stakeholder management is therefore crucial, which is the overlapping attribute between the legitimacy theory and the stakeholder theory. 3.1.3 Signalling & resource dependence theory

The signalling theory is about two parties: the sender and the receiver of the signal. These parties can be defined as the entity who possesses information (signaler, the bank) and the stakeholders for whom this information is not observable (receiver) (Spence, 1978). Between these parties there is

information asymmetry (Urquiza, Navarro, Trombetta & Lara, 2010). To reduce this information asymmetry, the sender signals information to mitigate this problem (An et al., 2011). According to research of Bini, Giunta and Dainelli (2010), signalling theory is therefore, because of its explanatory power, a useful extension to the agency theory. A high-performing company has an incentive to signal its superior excellence to prevent a so-called opportunity loss. For low performing companies this is the other way around: they have an opportunity gain. Under the assumption of information

asymmetry, signalling inside information can be beneficial (opportunity gain; being able to signal some information) or detrimental (opportunity loss; not being able to signal excellence). Apparently, firms only signal information when their performance is above average (Bini et al., 2010; An et al., 2011). There are different means to signal voluntary information. According to Watson et al. (2002) and Xiao, Yang and Chow (2004) one of the most effective ways to signal is via voluntary disclosure of positive accounting information. To get an insight of why banks are willing to signal information, another widely-known theory is used: the resource dependence theory.

The resource dependence theory of Pfeffer and Salancik (1978) argues that organizations are

dependent on vital resources possessed by the environment (e.g. stakeholders), which is referred to as a concept of power (Ulrich & Barney, 1984). This dependence comes with an amount of uncertainty, which managers try to reduce (Hillman, Withers & Collins, 2009).

(12)

Page | 11 After all, a bank’s business operations are dependent on for instance savings of their clients that are available to reinvest, interest income and other income derived from financial assets provided by the environment. Referring back to the signalling theory and translated it to a business environment, management can be seen as the sender who is signalling information (e.g. risk appetite statement, financial risks etc.) to the stakeholders via annual reports. Signalling this information can benefit the firm by enhancing relationships with stakeholders and improve the corporate reputation (Verrecchia, 1983; Bini et al., 2010; Barakat & Hussainey, 2013). The theory helps to get an understanding of the incentives for information disclosure (i.e. reputation recovery, stakeholder management). The power of the environment and the dependence on resources is mitigated by signalling this information. Banks are recovering from the financial crisis and want to perform outstandingly, signal this to stakeholders and prevent opportunity losses, because their surviving is dependent on their environment.

3.2 Integrated framework

The fundamental theories contribute to our study in a way so we can understand the incentives of banks to communicate about their risk taking. The turbulent times that banks have been through create motives for disclosing risk information with the aim of reducing information asymmetry, discharging accountability and earn their license to operate. These three key premises are interrelated concepts between the agency, stakeholder, legitimacy and signalling theory (An, Davey & Eggleton, 2011). Therefore, these theories are generally referred to as ‘voluntary disclosure theories’. Together with the resource dependence theory, which incorporates the complete environment of the banking industry, the voluntary disclosure theories create a comprehensive, substantiated and holistic theoretical foundation for this study. As different theories and concepts are discussed and to maintain structure, an overview of the overlapping attributes of the voluntary disclosure theories are visualised in table 1.

TABLE 1

Voluntary Disclosure Theories

Theory Related to Interrelated concepts

Agency Stakeholder Information asymmetry Stakeholder management Stakeholder Legitimacy Stakeholder management Accountability

Legitimacy Signalling Accountability Organizational legitimacy Signalling Agency Organizational legitimacy Information asymmetry

(13)

Page | 12

3.3 Disclosure and regulatory context

This research focuses on the quality of voluntary risk disclosures at European banks. To examine this quality and to draw right conclusions, it is important to get an understanding of the regulatory context. This section elaborates the context of the disclosure requirements with respect to laws, regulations and directives.

The banks in this research are required by law to publish annual reports with their financial statements, generally consisting of a balance sheet, an income statement, a statement of changes in equity and a cashflow statement. To shape this, banks are subject to International Financial Reporting Standards (IFRS), enhancing the usefulness of the financial information by making it verifiable, comparable, timely and understandable. One of these standards is IFRS 72_{, which requires disclosure} of information about financial instruments (i.e. monetary contracts between parties) in quantitative (e.g. data on exposures and concentrations) and qualitative terms (e.g. risk policies, objectives). Since financial instruments, such as cash, savings and (mortgage) loans are the core of the banking industry, IFRS 7 is particularly important for banks. The required information is about the significance of the instruments and the nature and extent of the risks that arise from them. These risks are divided into three categories: credit-, liquidity- and market risks (IFRS 7.32, Appendix III). Credit risk is the potential risk a borrower cannot meet the payment obligations that arise from the agreement. An entity having difficulties in paying its financial liabilities is referred to as liquidity risk. Market risk refers to fluctuations in value of financial instruments due to changes in market circumstances, like currency risks or interest rate risks. To communicate about this information and to enhance

transparency, banks often have included a risk paragraph in their annual report.

Next to IFRS, banks are required by Basel III to disclose risk information. This reform is a result of a process that has been going on for many years. In 1930 the Bank of International Settlements (BIS) was founded with the purpose of improving monetary and financial cooperation, by functioning as a bank for central banks worldwide. Currently sixty central banks (or other organisations with direct supervisory authority) are a member of this international financial institution, which is located in Basel, Switzerland. One of the committees that is operating under the BIS, is the Basel Committee on Banking Supervision (BCBS). The aim of this committee is to strengthen the regulation and supervise banks all over the world, with the purpose to enhance financial stability. This has been done by issuing several key reforms, starting with the Basel Capital Accord (Basel I) in 1988, which was focusing on capital adequacy (Basel Committee on Banking Supervision, 2014). Sixteen years later, the committee released a revised capital framework (Basel II) based on three pillars: (I) minimal capital requirements, (II) supervisory review process, and (III) transparency and market discipline (Basel Committee on Banking Supervision, 2004).

(14)

Page | 13 A few years later the financial crisis occurred and revealed the shortcomings of Basel II, especially in pillar 3: the market participants were not able to assess risks.

Largely in response to the crisis, the committee came up with a strengthened framework as a foundation for resilient banking (Basel III). The foundation and legal base for implementation arises from the CRR and the CRD IV. The CRR has a direct impact on the Member States via European laws, while the CRD IV locally have to be implemented and incorporated into national laws. According to the BIS, the banking sector had too much leverage and inadequate liquidity buffers when it entered the financial crisis (Basel Committee on Banking Supervision, 2014). It therefore extended the framework with new requirements regarding to maximum leverage, liquidity requirements, countercyclical capital buffers and tier-1 capital (Basel Committee on Banking Supervision, 2010). The new main goal of pillar 3: increasing comparability and consistency of disclosures. The BCBS used five guiding principles to achieve this, based on lessons learned from the financial crisis: clarity, comprehensiveness, usefulness, consistency and comparability (Basel

Committee on Banking Supervision, 2019). Although publishing pillar 3 information is not mandatory (and not subject to statutory audits), it is highly desirable to disclose since the relevance of the

information for the users. More transparent reporting (i.e. by using the guiding principles) enables clients and investors to determine the financial position of the bank and assess whether their funds can be safely entrusted. Additionally, the minimum requirements and key figures discussed above (e.g. liquidity requirements) are a means to establish trust. An example of the introduced requirements to gain trust is the liquidity coverage ratio (LCR). This ratio tries to ensure a banks survival by holding sufficient reserves to cope with liquidity stress during thirty days. The net stable funding ratio (NSFR) is an associated measure, focusing on a stabilized funding profile over a longer time horizon.3

To conclude, Basel III issued measures to enhance security and transparency as what was a

shortcoming of pillar 3 in Basel II. It aims at enabling users to perform better risk assessments. The purpose of disclosures according to the reporting standards (IFRS 7.32A, Appendix III) is to link qualitative and quantitative information and form a better picture of the nature and extent of risks that result from the financial instruments. However, to link these disclosures in order to assess a bank’s risk profile, users have to understand the implications of the Basel III ratio’s. A linkage of these implications and/in IFRS 7 would be beneficial to the users and contribute to trust. Trust that is essential since banks are dependent on the resources users provide. In the absence of trust, there are no funds to issue loans or grant mortgages and banks no longer collect the resulting interest, threatening their survival.

(15)

Page | 14

3.4 Hypotheses development and background literature

3.4.1 Audit Committee

Overseeing the transparency of financial reports and ensuring the objectivity of the external audit is one of the main responsibilities of the audit committee according to the FERMA (2014). As an addition to this, Mangena and Pike (2005) define the role of the audit committee as being the representative for the board and oversee matters that are related to the credibility of reporting,

auditing and corporate governance. Besides the previous proceedings, the audit committee selects and meets with the external auditor, questions internal auditors and monitors if management acts in the firm’s best interest (Klein, 2002). The audit committee is therefore often described as ‘the ultimate monitor’ (Blue Ribbon Committee, 1999; Krishnan & Visvanathan, 2008; Bhasin, 2012). The added value of the audit committee with regards to financial reporting has been proved by different

empirical examinations: the presence of an audit committee leads to more extensive disclosure (Ho & Wong, 2001), lower chances of financial reporting fraud (Peasnell, Pope & Young, 2001; Dechow, Sloan & Sweeney, 1996) and improved reporting quality (Turley & Zaman, 2003). We can conclude from this research that in general the audit committee is doing what it is supposed to do.

The aspects that turned out to be of importance to establish an effective audit committee, are

independence and financial expertise (Mangena & Pike, 2005). Since the aforementioned requirement of the Basel Committee about the periodically meetings between the committees, the focus is also on committee activity. Committee size is a common proxy as well, despite the empirical evidence is mixed (Mangena & Pike, 2005).

AC Size

Prior research finds different conclusions regarding the size of the audit committee. Arguing from the agency perspective, Sultana, Singh and van der Zahn (2015) state that a smaller audit committee enhances the dynamics and the cohesion of the group. Another argument is that if the size of the committee is larger, this may result in impaired monitoring and control functions. A larger audit committee also increases the probability of opportunistic behavior (Mintzberg, 1983). On the other hand, Bédard, Chtourou and Courteau (2004) argue that larger committees are more likely to resolve and uncover problems in the reporting process. Also, the more members on the committee, the more expected expertise available, and thus an increasing efficiency regarding its risk management (Dionne & Triki, 2005). With respect to quality of reporting, Felo, Krishnamurthy and Solieri (2003) found a positive relation with committee size. Despite the mixed and inconclusive results, more value is attached to last mentioned studies in developing the hypotheses. This study is specifically aiming at the reported disclosure quality, which is more in line with the latter studies. The first hypothesis is therefore:

(16)

Page | 15 AC Activity

Activity is defined as the number of annual meetings. Frequent meetings lead to improved risk management communication, according to Tao and Hutchinson (2013). Studies by Kent and Stewart (2008) find increasing corporate disclosure when audit committees meet on a regular basis. Finally, there is also evidence suggesting that meeting frequency positively is associated with voluntary disclosure (Abdullah et al., 2017). Another aspect of meetings that has been considered for this research is the attendance rate at the meeting. Since the importance and functioning of the audit committee partly lies in meeting and discussing current issues, it is assumable that there is awareness about the responsibility among the committee members and the importance of the meetings. Since attendance rates generally are high, the presence rates are excluded from this research. The second hypothesis is as follows:

H2: There is a positive relation between audit committee meeting frequency and the quality of risk disclosure.

AC Independence

According to Goodwin (2003), when the audit committee is composed of a majority of independent members, this results in improved financial reporting quality. Abraham and Cox (2007) add that the disclosure of specifically risk management is also improving. Since risk disclosure is about

transparency and integrity, the findings of McMullen and Raghunandan (1996) strengthen the third hypothesis: independent audit committees have a preventive role in fraudulent reporting and misleading financial statements. Despite the fact that insiders have more knowledge and access to each department’s risk activities, which the independent directors have not (Ismail & Rahman, 2011), the following relation is expected:

H3: There is a positive relation between audit committee independence and the quality of risk disclosure.

AC Expertise

In firm’s voluntary disclosure practices, financial expertise plays a central role (Kelton & Yang, 2008). In the extant literature it is suggested that more financial expertise amongst the members of the audit committee enhances the committee’s effectiveness (Krishnan & Visvanathan, 2008; Dhaliwal, Naiker & Navissi, 2010). Research by Cohen, Hoitash, Krishnamoorthy and Wright (2013) and Kusnadi, Leong, Suwardy and Wang (2016) show a positive association between financial expertise and the quality of financial reporting. Expertise also contributes in developing effective risk

management processes (McDaniel, Martin & Maines, 2002). This evidence is the basis for the fourth hypothesis:

H4: There is a positive relation between financial expertise on the audit committee and the quality of risk disclosure.

(17)

Page | 16 3.4.2 Risk Committee

Some companies charge the audit committee with all risk oversight (Protiviti, 2011). Since the audit committee is assigned with more than risk monitoring (e.g. internal control and financial reporting monitoring), there is a strong demand for the establishment of risk committees (Hines et al., 2015). By using a citation of Hines et al. (2015), retrieved from research by Moore and Brauneis (2008) and Schlich and Prybylski (2009), the function of the risk committee is: “An RC’s function is to oversee

an organization’s risk management framework, including the process for identifying, assessing, and responding to all current and future risks that threaten an organization’s existence, including

financial, credit, operational, market, and compliance risks.” Researchers in the field of RC’s are not

unanimous in their judgments on the effectiveness of the committee. Some argue that the single focus on risks contributes to the total risk oversight (Whyntie, 2013) and that specifically assigned resources improve the ability to continuously monitor the risk profile, appetite and corresponding management of the firm (Moore & Brauneis, 2008). Others see the committee as an extra layer of bureaucracy (Protiviti, 2011) and argue that the risk oversight should stay a responsibility of the full board (COSO, 2010). Research by Gordon, Loeb and Tseng (2009) show that larger boards are closer to enterprise risk management (ERM) ‘best practice’ effectiveness, but this is exactly the problem as identified in the introduction: boards are becoming smaller.

To determine risk committee effectiveness, three common proxies are used. They overlap with the proxies for the audit committee, except the variable expertise. For risk committees we therefore look at the size, activity and independence. The reason for the exclusion of expertise regarding the risk committee lies in the nature of the requirements of the different committees. The audit committee requires at least one member with competence in accounting and/or auditing (European Parliament Directive 2014/56/EU, 2014; Deloitte EU audit legislation, 2015), which will be measured by their expertise. For the risk committee it is recommended to assign members having specific industry knowledge and experience, but it is hard to determine whether someone has expertise in the field of specifically industry related risks. Therefore, expertise regarding the risk committee is excluded. Literature from comparable research (i.e. board- and committee composition evidence) is used to arrive at the hypotheses, since recent evidence on (risk) committees is scarce. Table 2 shows an overview of a literature review performed by Kolev et al. (2019), indicating that recently not much research has been done on board committees. Note that not all worldwide research on committees is taken into account in this literature review, but it shows at least a certain tendency over the last three years. Moreover, the risk committee was not even considered as a main governance committee in this review. While nearly every bank has a risk committee, this is apparently not the case (yet) for other businesses and industries.

(18)

Page | 17

TABLE 2

Studies on Board Committees

Year

Studies

Before 1990

5 1990 to 1995

4 1996 to 2000

16 2001 to 2005

32 2006 to 2010

40 2011 to 2015

39 2016 to 2018

6

Reprinted: “Board Committees in Corporate Governance : A Cross-Disciplinary Review and Agenda for the Future.” by Kolev et al., 2019. Journal of Management Studies, 56(6), 1138-1193.

RC Size

Similar and as discussed before about the audit committee, a larger committee is likely to have more competence, skills and experience. This is confirmed by Al-Hadi et al. (2018), who showed a larger risk committee improved risk management skills and knowledge, leading to more risk disclosure. Due to the lack of risk committee research, we rely on comparable AC evidence. The fifth hypothesis is therefore in line with the first one:

H5: There is a positive relation between the risk committee size and the quality of risk disclosure.

RC Activity

To be informed about the latest developments regarding risk management, frequent meetings are expected to increase that ability (Abbott, Parker & Peters, 2004). The effectiveness of the risk

committee increases by the number of meetings dedicated to risk oversight (Ellul & Yerramilli, 2013). Research in 2002 by DeZoort, Hermanson, Archambeault and Reed, demonstrates that the frequency of meetings of audit committees is clearly related to less financial reporting problems and higher quality. For this hypothesis we argue in line with the AC activity hypothesis and do not take the attendance rate into scope. Similar to audit committee findings, a positive impact on the disclosure quality is hypothesized.

H6: There is a positive relation between the risk committee meeting frequency and the quality of risk disclosure.

RC Independence

Arguing from the agency and stakeholder theory, independent directors should be able to make unbiased judgments, increase the monitoring of managers and thus enhancing the interests of stakeholders (Al-Hadi et al., 2018). Outside directors also have the incentive to protect their

reputation as monitors and therefore are willing to criticize management (Fama & Jensen, 1983). On the contrary and as explained before, the difference with the audit committee is that the risk

committee needs to be familiar with specific industry and company-related risks. Banks have to deal with totally different risks compared to for instance the construction or IT-sector.

(19)

Page | 18 To determine, control and report about these specific risks, knowledge and experience in the sector are needed. Banks are complex and subject to lots of laws and regulations, thereby assuming that insiders possess the required tools over the outsiders. Research by Eng and Mak (2003) indeed found a negative relation with independent directors and voluntary disclosure, which was also demonstrated by Barako et al. (2006). Despite the assumption that insiders will have more firm- and industry-specific experience compared to independent directors, the seventh hypothesis is non-directional due to the mixed evidence:

H7: There is a relation between the independence of the risk committee and the quality of risk disclosure.

METHODOLOGY

4.1 Sample selection and data collection

Basel III only has legal effect in the European Union. Therefore the population is limited to European banks. The European Banking Authority (EBA) performs EU-wide transparency exercises to provide data about the largest, most influential and leading European banks. This study is based on the same population as selected for the yearly transparency exercise of the EBA. The effect of the Basel III developments will be visible and reflected most on this leading banks and therefore gives us more valuable insights. The population consists of hundred-thirty banks representing twenty-five different countries. The original sample is made up of sixty of these banks and covers nineteen countries (table 3). For an oversight of the sample selections, see appendix II. The data is manually collected for all variables and retrieved from annual reports that are published by the banks (i.e. archival data). This means the study relies on secondary data (i.e. making use of disclosed information). By reason of the multi-year focus, data is collected for 2016, 2017 and 2018. Reasons for this timespan are that these are the years after the implementation of Basel III and the lack of recent board committee studies (table 2). The aim is on multiple observations for several variables in different points in time, resulting in a dataset consisting of panel data.

4.2 Research design

Since the number of complete observations differs among the years, banks with missing values had to be removed (see paragraph 4.6). Although, to analyze as much as possible of the available data, a three-stage-model was designed. With this model it is possible to get the most out of the data, without unnecessary removal of observations. This is done by starting with a single aim on 2018 (Model 1: N=52). Second, the incomplete observations of 2017 are removed to get two maximal identical samples for 2018 and 2017 (Model 2: N=44 per year; Total N=88). Lastly, the same principle was applied to 2016, ending up with a dataset of 34 identical banks for all three years (Model 3: N=102). An overview of the represented countries and their share per model is shown in table 3.

(20)

Page | 19

TABLE 3

Sample Distribution

Country

Original

Model 1

Model 2

Model 3

Austria

2 Belgium

4 Czech Republic

1 Denmark

3 Finland

1 France

6 Germany

8 Greece

1 Hungary

1 Ireland

2 Italy

5 Luxembourg

1 Netherlands

6 Norway

1 Poland

1 Portugal

2 Spain

5 Sweden

4 United Kingdom

6 Total

60 4.3 Independent variables

(3.3%)

2 (3.8%)

2 (4.5%)

2 (5.9%)

(6.7%)

4 (7.7%)

4 (9.1%)

2 (5.9%)

(1.7%)

1 (1.9%)

1 (2.3%)

-

(5.0%)

3 (5.8%)

3 (6.8%)

3 (8.8%)

(1.7%)

1 (1.9%)

1 (2.3%)

1 (2.9%)

(10.0%)

5 (9.6%)

4 (9.1%)

4 (11.8%)

(13.3%)

4 (7.7%)

3 (6.8%)

3 (8.8%)

(1.7%)

1 (1.9%)

1 (2.3%)

-

(1.7%)

-

(3.3%)

2 (3.8%)

1 (2.3%)

1 (2.9%)

(8.3%)

3 (5.8%)

2 (4.5%)

1 (2.9%)

(1.7%)

1 (1.9%)

1 (2.3%)

-

(10.0%)

6 (11.5%) 5 (11.4%) 4 (11.8%)

(1.7%)

1 (1.9%)

1 (2.3%)

1 (2.9%)

(1.7%)

1 (1.9%)

1 (2.3%)

1 (2.9%)

(3.3%)

2 (3.8%)

-

(8.3%)

5 (9.6%)

5 (11.4%)

4 (11.8%)

(6.7%)

4 (7.7%)

3 (6.8%)

3 (8.8%)

(10.0%)

6 (11.5%) 6 (13.6%)

4 (11.8%)

(100%)

52 (100%)

44 (100%)

34 (100%)

The effects of committee characteristics are extensively researched. The measures used are therefore derived from prior studies and are equal for both committees. Committee size is defined as the number of members on the committee during the fiscal year (Mangena & Pike, 2005; Pucheta-Martinez & García-Meca, 2014; Wilbanks, Hermanson & Sharma, 2017; Abad & Bravo, 2018). To measure the committees’ activity, we use data on the amount of annual committee meetings that took place during the fiscal year (Aldamen, Duncan, Kelly, McNamara & Nagel, 2012; Sultana, 2015; Lee & Fargher, 2017). Previous research points out multiple options for measuring committee

independence. A frequently used proxy is the amount of independent members measured as a percentage of the total committee size (Chen, Moroney & Houghton, 2005; Sultana, 2015). Since a majority of independent members is more the rule than the exception, a committee is only considered independent as all members are (Abbott, Parker, Peters & Raghunandan, 2003; Hines et al., 2015). A fully independent committee is assigned a value of one and a zero otherwise. Regarding the audit committee, the role of expertise amongst the members is studied as well. Financial expertise is defined as having experience as a public accountant, auditor, controller, CFO, CAO (chief accounting officer) or CEO (Sultana, 2015). Members complying with this criterium are measured as a

percentage relative to the total committee size (Badolato, Donelson & Ege, 2014; Lee & Fargher 2017). Table 4 shows an overview of the measures, including their acronyms.

(21)

Page | 20

4.4 Dependent variable

Risk disclosure quality (RDQ jt) is measured by a disclosure index, which is a partial form of content analysis. This index is composed of thirty elements, based on Basel III, the Enhanced Disclosure Task Force (EDTF) and the European Securities and Markets Authority (ESMA). The elements are divided into four areas: general information about risks (7 items), information on credit risks (8 items), liquidity risks (5 items) and solvency/financial strength risks (10 items). The selection of items in these categories correspond to the minimum of categories distinguished and required by IFRS 7.32 (see appendix III). The items are assigned an ordinal score varying from zero (not disclosed), one (disclosed, but limited) to two (fully disclosed), leading to a possible maximal score of sixty points (see appendix II). A more detailed scale (e.g. ranging from 1-10) would entail nuance problems. In addition, since the freedom of choice for companies of how to give substance and design annual reports, a more detailed distribution would make allocation of points even harder and increase subjectivity (e.g. when to award 3 points and when 4?). On the other hand, a scale of only two values would exclude ‘the gray area’. This would give a distorted reflection of reality, showing extreme outcomes. With the level of measurement used, this issue was tackled as much as possible. The risk disclosure items itself are also sensitive to subjectivity. For instance, item 20 till 24

(appendix II) are straightforward and leave no space for differences in interpretation (i.e. allocation of points, a ratio is clearly disclosed or not). However, the opinion about whether the different risk sections are easily identifiable (item 2) may differ per user or researcher. To increase consistency and to reduce the impact of this subjectivity, measures were taken at two different points in time: before the data collection and afterwards. In advance, a workshop about the data-collection process was organized by the research supervisor. This meeting showed the impact of subjectivity and the possible angles to look at reported information. The workshops gave us insights that reduced the differences in our perceptions and aligned our allocation of points. All involved researchers attended the workshop. After the data collection, mutual data reviews have been performed on two-third of the original sample, resulting in the second check on subjectivity. Occurred dissimilarities were discussed and adjusted when upon agreed.

It would be very complicated to include multiple dimensions of quality in the measurement of financial reporting quality (Knoops, 2001). This study therefore uses ‘a measure of quality’, and not ‘the measure of quality’. The assumption underlying the index: a higher score indicates better quality and vice versa. In addition, there is an equal weighting for each disclosure item. The rationale behind this comes from a literature review on disclosure indices by Marston and Shrives (1991). They argued that weighting the importance of items is highly dependent on the user group. Since the stakeholder-wide view in this study, the importance preferences amongst different user groups would average each other out (Marston & Shrives, 1991).

(22)

Page | 21

TABLE 4

Variable Definitions

Proxies Acronym Measure

Disclosure Quality RDQ jt Disclosure index score of firm j during period t

AC Size AC_SIZE jt Number of audit committee members of firm j during period t AC Activity AC_ACT jt Number of audit committee meetings held of firm j during

period t

AC Independence AC_IND jt 1 if AC is fully occupied by independent members, 0 otherwise for firm j during period t

AC Expertise %AC_EXP jt Number of audit committee members with financial expertise divided by committee size of firm j during period t*

RC Size RC_SIZE jt Number of risk committee members of firm j during period t RC Activity RC_ACT jt Number of risk committee meetings held of firm j during

period t

RC Independence RC_IND jt 1 if RC is fully occupied by independent members, 0 otherwise for firm j during period t

Firm Size F_SIZE jt Natural log of total assets for firm j during period t

Firm Leverage %LEV jt Ratio of total liabilities to total assets for firm j during period t Firm Profitability %ROE jt Return on equity (ROE) for firm j during period t

Uncertainty avoidance

UNCAVOID Score on Hofstede dimension per country Individualism INDIVID Score on Hofstede dimension per country

4.5 Control variables

Previous research on the topic of disclosure reveals firm size to be consistently associated with disclosure levels (Linsley & Shrives, 2006; Barakat & Hussainey, 2013; Abdullah et al., 2017). Firm size is measured as the natural log of total assets at year end. Since high leveraged firms are expected to be more strictly monitored by creditors (Al-Hadi, Hasan & Habib, 2016) and may have less

incentives to be transparent in case of high financial risk (Bokpin, 2013), we also control for leverage. As stated by Elshandidy, Fraser and Hussainey (2015), high profitable firms are likely to provide more risk information. An incentive for this is to support the organization’s continuation and keep their current (profitable) position (Al-Hadi et al., 2016). A third control variable is therefore profitability, measured by the return on equity. Lastly, cultural influences are taken into account. In general, cultural values are essential in harmonizing accounting practices (Ding, Hope, Jeanjean & Stolowy, 2007). Research by Hope (2003) demonstrated that national culture influences managers’ choices (in preparing) and investors preferences (in using) with regard to financial reports.

(23)

Page | 22 To make culture measurable, Hofstede (1984) developed five dimensions which are extensively used in accounting research. Two of Hofstede’s dimensions are uncertainty avoidance and individualism. Gray (1988) adjusted this dimensions for accounting purposes and came up with another one: secrecy vs transparency. This seamlessly fits with the core of this research and is found to be significantly related to two of Hofstede’s dimensions: a higher degree of uncertainty avoidance and a lower degree of individualism, would lead to more secrecy (Gray & Vint, 1995). Since the close link between transparency and the core of this study, data on these two dimensions of Hofstede is used to control for cultural influences.4

4.6 Data consistency and modifications

To create a reliable, comparable and consistent dataset, several adjustments and checks are applied to the raw data. First of all, missing values are removed from the initial sample (N=60) for every year. This resulted in different amounts of observations for model 1, 2 and 3 (see paragraph 4.2 and table 3). Second, the risk disclosure scores for 2017 and 2016, available from prior year, were linked and combined to the collected financial and governance data. Since the inequalities in reporting currencies a correction was applied and converted all amounts to euro’s (EUR; €). However, this was only the case for the control variable firm size, because this is the only financial variable affected (total assets). The other financials both are ratio’s (i.e. leverage and return on equity), which are not sensitive for a correction. The conversion factors are derived from the website of the Dutch tax authorities 5_.

TABLE 5

Currency Conversion Rates

Country Currency Rate per 31/12/18

Czech Republic CZK € 1 = 26,002 CZK

Denmark DKK € 1 = 7,4617 DKK

Norway NOK € 1 = 9,729 NOK

Poland PLN € 1 = 4,2995 PLN

Sweden SEK € 1 = 10,3143 SEK

United Kingdom GBP € 1 = 0,89108 GBP

United Kingdom USD € 1 = 1,1409 USD

Next, a check was executed to see whether the dataset consisted of unexpected or illogical values (e.g. leverage ratio’s greater than 1 or huge differences in total assets between years). A similar check was done for the scores on the disclosure index. Large variances over the years are unexpected, especially for declining scores, since it is likely that banks are reporting more extensively due to new legislation. The threshold for extra reviews was set at a difference of 10 points per year. Therefore a total of four banks were subject to an additional check, resulting in adjusted scores. These reviewed scores have been agreed upon with an involved researcher and are determined in line with the four-eyes principle. Subsequently, the total assets were replaced by their natural logarithms.

4_{Data derived from https://www.hofstede-insights.com/product/compare-countries/}

(24)

Page | 23 This was done by means of making the distribution less skewed and convert this data to a normal distribution. Lastly, the continuous variable of total assets was subject to winsorizing. Data beneath the 2.5th_{percentile is set to the 2.5}th_{percentile, which is also the case for data above the 97.5}th percentile. This amounts to the mean plus/minus two times the standard deviation, since the total assets via logging were converted to a normal distribution. The rationale behind this is to mitigate the effect of outliers.

4.7 Statistical model

By use of multiple linear regression the hypotheses are tested. The statistical formula is represented as follows:

RDQjt = β0 + β1*AC_SIZEjt + β2*AC_ACTjt + β3*AC_INDjt + β4*%AC_EXPjt + β5*RC_SIZEjt + β6*RC_ACTjt + β7*RC_INDjt + β8*F_SIZEjt + β9*%LEVjt + β10*%ROEjt + β11*UNCAVOID + β12*INDIVID + ε

Since three different models are involved, it is required to use various approaches regarding the regression. The first model is built up of one year (2018), allowing an ordinary multiple linear regression. For model 2 and 3 different statistics are required, because this data consists of multiple observations at different points in time for the same group of banks. To see whether it is desirable to organize the data in a panel-format, a Breusch-Pagan Lagrange Multiplier (LM-test) was calculated. In other words, this test determines whether there are company- or time-specific effects within the dataset. Since this is the case for model 2 and 3 (p < .001), a panel data set is preferred. To determine if the company- and time-effects are correlating with the explanatory variables, a Hausman-test was executed. Because the outcome of this test had a very high p-value, (p > .9) a random-effects-model was preferred over a fixed-effects approach. The regressions of model 2 and 3 are therefore suited best with a random-effects model. Another argument for this approach is that these results are more generalizable to the whole population compared to the fixed-effects model (i.e. this model is more informative for just the sample, not the whole population). The statistical formula for the panel regressions of model 2 and 3 look like the first formula, despite the fact that the error term is pulled apart in three elements. The formula:

RDQjt = β0 + β1*AC_SIZEjt + β2*AC_ACTjt + β3*AC_INDjt + β4*%AC_EXPjt + β5*RC_SIZEjt + β6*RC_ACTjt + β7*RC_INDjt + β8*F_SIZEjt + β9*%LEVjt + β10*%ROEjt + β11*UNCAVOID

(25)

Page | 24 Where:

ujt = μj + λt + νjt,

μj= unobservable company heterogeneity;

λt= unobservable time heterogeneity;

νjt= residual random error.

In addition, the descriptive statistics are calculated, a correlation analysis is performed and the Cronbach’s Alpha for the disclosure index has been determined. The descriptive statistics give an insight in the general values of the dataset, like averages, maximum/minimum values etc. The correlation matrix indicates if there is a statistical relation between the variables. However, this does not have to be evidence for causality (yet). The Cronbach’s alpha is a statistical test, computing a coefficient of internal consistency and reliability of the disclosure index.

(26)

Page | 25

FINDINGS

5.1 Descriptive statistics

The table below shows the descriptive statistics per model after winsorizing. The logging of the total assets is not applied yet in this table, since unlogged assets are more informative. For the same reason the dummy variables (independence indicators) are excluded from this table. Information on these variables is set out in table 7.

TABLE 6 Descriptive Statistics

Variable N Min Max Mean St. Dev.

M O D EL 1 RDQ 52 15 45 30.5 7.77 AC_SIZE 52 3 9 4.90 1.51 AC_ACT 52 3 25 8.54 4.31 %AC_EXP 52 0.2 1 0.64 0.23 RC_SIZE 52 3 10 5.21 1.59 RC_ACT 52 4 25 9.37 4.68 F_SIZE 52 26,953 2,040,836 460,842 528,823 %LEV 52 0.85 0.98 0.93 0.02 %ROE 52 -0.35 0.17 0.07 0.07 UNCAVOID 52 23 100 63 25 INDIVID 52 27 89 68 15 M O D EL 2 RDQ 88 13 45 30.0 7.46 AC_SIZE 88 3 9 4.97 1.63 AC_ACT 88 2 18 7.92 3.43 %AC_EXP 88 0 1 0.59 0.30 RC_SIZE 88 3 11 5.23 1.69 RC_ACT 88 4 21 8.97 4.03 F_SIZE 88 26,953 2,040,836 501,552 533,680 %LEV 88 0.84 0.97 0.93 0.02 %ROE 88 -0.01 0.19 0.08 0.04 UNCAVOID 88 23 100 62 24 INDIVID 88 35 89 70 13 M O D EL 3 RDQ 102 14 45 29.4 7.10 AC_SIZE 102 3 9 5.15 1.74 AC_ACT 102 2 17 7.47 3.12 %AC_EXP 102 0 1 0.58 0.30 RC_SIZE 102 3 11 5.44 1.76 RC_ACT 102 4 38 9.41 5.52 F_SIZE 102 75,417 2,104,910 609,288 571,472 %LEV 102 0.85 0.97 0.93 0.02 %ROE 102 -0.26 0.18 0.08 0.05 UNCAVOID 102 23 94 60 24 INDIVID 102 51 89 71 11

For 2018, the 52 companies achieved an average score of 30.5 out of the maximum possible 60 on the quality of their risk disclosure. Model 2 and 3 show lower means, respectively 30.0 and 29.4. Since these models include prior years, this development indicates that banks are scoring slightly higher in more recent years.

(27)

Page | 26 The highest value scored is 45 and is represented in all three samples. The lowest was a score of 13 in the second model. The size of the audit committee varies between 3 and 9 members, which is 3 and 11 for the risk committee. On average in all three models, the risk committee have more annual meeting compared to the AC. The average expertise represented by the audit committee members varies from 58% up to 64%. Regarding the sizes of the banks, it becomes clearly visible that model 3 on average consists of larger banks. Since the removal of missing values the average sizes of the banks have increased. The assumption that smaller firms provided more missing values seems to be legitimate. However, this is not the same or no evidence yet for a relationship with poorer disclosure practices. The debt to assets ratio hardly shows no changes: 93% for each model. With respect to the

independence proxies, an overview of the frequencies are shown in table 7. A value of 1 means the committee is fully independent; a 0 otherwise. For each model there nearly is an equal distribution for audit committees: the frequencies deviate maximally with 1 from the mean. The risk committee is more dispersed with a maximum deviation of 8 from the average.

TABLE 7 Dummy Variables

Variable Value Model 1 Model 2 Model 3

AC_IND 0 27 43 50 1 25 45 52 Total 52 88 102 RC_IND 0 31 52 58 1 21 36 44 Total 52 88 102

5.2 Correlation analysis

To see if there is a statistical relation between the variables of interest, an analysis on the correlation coefficients took place. Correlation is the statistical coherence among variables, expressed in a coefficient varying from -1 up to +1, where +1 is a perfect positive, -1 a perfect negative and 0 no correlation at all. However, correlations do not have to be evidence for causality. The matrices for each model show no significant correlations between the dependent and independent variables. This indicates that there is no coherence between the several committee proxies and the quality of risk reporting. The control variables of firm size and individualism do show a statistical significant positive relation with the RDQ in each model. The highly significant negative relationship between the cultural variables was expected as stated in paragraph 4.5. The correlation analysis is also used to determine multicollinearity: a phenomenon of highly correlated explanatory variables (i.e. an

independent variable can be used to predict another). Coefficients exceeding the value of 0.7 are an indication for multicollinearity. Since the similar proxies for the committees, this is the case for some independent variables. However, an extra analysis on the variance inflation factors (VIF) showed no multicollinearity issues, which is further discussed in paragraph 5.3.

(28)

TABLE 8

Correlation Matrix – Model 1

1 2 3 4 5 6 7 8 9 10 11 12 13 1. RDQ 1.00 2. AC_SIZE .143 .313 1.00 3. AC_ACT .043 .761 -.173 .221 1.00 4. AC_IND .088 .536 -.194 .168 -.176 .213 1.00 5. %AC_EXP -.007 .958 -.340** .014 .028 .846 .154 .275 1.00 6. RC_SIZE .186 .187 .788*** .000 -.140 .322 -.007 .960 -.138 .330 1.00 7. RC_ACT .107 .450 -.016 .909 .749*** .000 -.159 .261 -.126 .373 .021 .882 1.00 8. RC_IND .121 .394 -.204 .148 -.021 .881 .620*** .000 .133 .348 -.161 .256 -.183 .194 1.00 9. F_SIZE .567*** .000 .207 .141 -.033 .819 -.105 .458 .029 .841 .238* .090 .060 .673 -.107 .449 1.00 10. %LEV .108 .448 -.065 .650 -.278** .046 -.232* .098 .026 .858 -.049 .733 -.315** .023 -.068 .631 .221 .115 1.00 11. %ROE .123 .387 .052 .716 -.306** .027 .149 .292 .088 .535 .100 .481 -.174 .217 .099 .484 .076 .592 .056 .693 1.00 12. UNCAVOID -.154 .275 .127 .371 .163 .248 -.146 .302 -.446*** .001 -.056 .692 .187 .184 -.269* .054 -.124 .381 -.138 .330 -.263* .060 1.00 13. INDIVID .380*** .005 -.084 .552 -.189 .180 .209 .138 .176 .212 -.017 .902 -.229 .103 .309** .026 .359*** .009 .455*** .001 .283** .042 -.578*** .000 1.00

(29)

TABLE 9

Correlation Matrix – Model 2

1 2 3 4 5 6 7 8 9 10 11 12 13 1. RDQ 1.00 2. AC_SIZE .045 .681 1.00 3. AC_ACT -.035 .746 -.221** .039 1.00 4. AC_IND .083 .443 -.259** .015 .031 .778 1.00 5. %AC_EXP .028 .796 -.372*** .000 .126 .243 .267** .012 1.00 6. RC_SIZE .125 .247 .803*** .000 -.157 .144 -.030 .781 -.178* .096 1.00 7. RC_ACT .036 .737 .002 .988 .622*** .000 -.082 .447 -.127 .239 .057 .600 1.00 8. RC_IND .165 .124 -.296*** .005 .026 .809 .721*** .000 .183* .088 -.208* .051 -.189* .078 1.00 9. F_SIZE .564*** .000 .118 .272 .011 .918 -.057 .597 -.048 .655 .130 .228 .125 .248 -.050 .645 1.00 10. %LEV .152 .158 -.065 .545 -.276** .010 -.233** .029 -.062 .564 -.141 .190 -.238** .026 -.113 .296 .301*** .004 1.00 11. %ROE -.154 .153 -.331*** .002 -.176 .102 .093 .389 .221** .039 -.128 .235 -.144 .180 .051 .634 -.330*** .002 .059 .584 1.00 12. UNCAVOID -.220** .040 .156 .148 .063 .563 -101 .350 -.400*** .000 -.028 .794 .107 .320 -.166 .121 -.158 .141 -.137 .204 -.122 .257 1.00 13. INDIVID .371*** .000 -.252** .018 -.164 .127 .198* .065 .126 .242 -.121 .263 -.257** .016 .272** .010 .325*** .002 .442*** .000 -.091 .399 -.552*** .000 1.00

The influence of governance committees on the quality of risk disclosure: a European banking perspective