The risk disclosure quality of European banks

(1)

The risk disclosure quality of European banks

Word count: 11,962

June 2018

Master of Accountancy and Controlling

R.A. Visser

(2)

1

Abstract

This research explores risk disclosure quality in annual reports of European banks. The determinants of profitability, state aid, and the European Central Bank’s (ECB) stress test scores are examined to establish their relationship, if any, with risk disclosure quality. The quality of 75 annual reports from different European banks is measured by using a disclosure index. The findings show that there is no relationship between profitability or state aid and risk disclosure quality. The ECB stress test score consists of two primary sections. The transitional total capital delta show a weak negative relation with the risk disclosure quality. The fully loaded leverage delta has a stronger significant negative relationship with the quality of risk disclosure. This research provides evidence that banks who perform bad on the stress test, are providing higher quality of risk disclosure. These results are controlled for the bank size and the strictness of the supervisor. Future research should aim to identify determinants for analysis that account for risk disclosure quality and are unaffected by size.

(3)

2

1 Introduction

1.1 Introduction

The collapse of Lehman Brothers was the largest bankruptcy in United States history and one of the watershed moments in the global economic crisis of 2007–2012. The bankruptcy was partly triggered by massive investments in subprime mortgage-backed securities in the US. These securities where rated as “secure” investments by rating companies, because the mortgage coverage was deemed to be safe collateral. The ratings famously proved to be incorrect.

This was an illustrative example of the problems within the banking sector (The Economist, 2018). Shady Faulty constructions, such as the repackaging and the resale of mortgage-backed securities, were established to hide information from the outside world (Ospina and Uhlig, 2017). Banks were willing to invest in projects involving high risk (Martynova, Ratnovski and Vlahu, 2017) due to the complex construction of investments and the failure of rating agencies to assess properly the risk of such assets (Fratianni and Marchionne, 2009). These developments together with the deregulation tendency inevitably led to the banking crisis (Crotty, 2009).

The events described demonstrate the effect of the failing banking system on the whole economy (Tsoumas, 2017). The indirect effect of economic events in one industry on events in other, unrelated industries, is referred to as the spillover effect. Its workings became obvious with the influence of the banking sector on the economy. Therefore, a resilient banking industry is necessary to avoid contagion to other industries. The Edelman Trust Barometer survey showed an overall drop in the professionals’ trust in European banks from 2008 to 2013 (Edelman Berland, 2013). One way to rebuild the trust in the banking sector is by increasing transparency in their performance, operations, and thesubsequent risks (Brenna et al., 2009).

The increasing demand for more transparency of banks leads to an increased demand of good quality information disclosures, (Klepczarek, 2016). Risk disclosers must bridge the information gap of the risks faced by the European banks. The current developments are aimed at developing a framework that will improve the effectiveness of risk disclosure (Singh and Peters, 2013). Singh and Peters (2013) highlight the importance of focusing primarily on meeting the information needs

(6)

5

of stakeholders. Since the financial crisis, risk disclosure has gained prominence as an effective tool for monitoring banks. The monitoring process may prevent banking crises, or at the very least, make their occurrence less likely (Financial Stability Board, 2012). Also, the current regulation of the fourth Capital Requirement Directive (CRDIV) and the Capital Requirement Regulation (CRR), which are based on the guidelines of the Basel III Capital Accords, and the International Financial Reporting Standards (IFRS), have increasingly focused on the disclosure of private information. For example, disclosing credit loss techniques like the “roll forward reconciliations” is mandatory, given the level of management judgment and complexity (KPMG, 2015 & PWC 2015).

Research on disclosures is relevant due to the central role of banks in the economy and the economic crisis. With the stakeholders’ call for transparency and the renewed legislation of disclosures, banks could handle differently both the regulations and the public demands for risk disclosure. Therefore, the quality of private information disclosures may vary among different banks. The present study aims to shed light on the difference in the disclosure quality in annual reports. Upon identifying the difference in quality, if any, the determinants, i.e. the causes of the discrepancy in quality, are investigated. This study focuses on the profitability and two crisis-related determinants. The two crisis-crisis-related determinants are the effect of state aid and the effect of the ECB stress test score on the quality of risk disclosure. The following research question is thus formulated:

Do profitability, the ECB stress test score, and the state aid affect the risk disclosure quality of annual reports?

This study focuses only on the banks supervised by the European Banking Authority (EBA). The regulatory playing field for these banks is similar because they all have to comply with the CRR and have to report on the basis of IFRS (Huttenhuis and ter Hoeven, 2015). All of these banks have to obey the same regulations determined by the EBA. For example, both the implementation of the IFRS and the guidelines of the Basel Capital Accords are mandatory for the listed banks.

(7)

6

1.2 Scientific contribution

This study aims to contribute to the existing literature by researching the effect of profitability, state aid, and ECB stress scores on risk disclosure quality. Clear differences can be observed between the literature on the banking sector and on companies belonging to other sectors. The banking sector has been at the forefront of developments in risk management and risk measurement techniques (Linsley et al., 2006). Linsley et al. expect banks to have a higher quality of disclosures compared to other companies due to their forefront position. Consequently, as a result of their forefront position of risk management development and its expected higher quality, the banking sector needs to be studied independently.

Since 2000, the banking industry has seen many changes including those involving reporting. Recent studies have pointed to the improvement of risk disclosure. This was partly caused by the implementation of the Basel II standards and IFRS 7, which demanded risk disclosure in the annual reports of banks (e.g. Nahar, et al., 2016; Homölle, 2009). Henceforth, the research has focused on the determinants, and the risk disclosure quality has subsequently increased. Linsley et al. (2006) have revealed a clear link between the size and level of risk disclosure in the UK and Canadian banks. Other studies focused on the effect of governance on the risk disclosure quality (e.g. Barakat and Hussainey, 2013). They concluded that independency of the boards, a lower executive ownership, concentrated non-governmental ownership structure, and a more active audit committee provides higher quality of operational disclosure.

Several studies have already investigated the link between profitability and the risk disclosure quality with contrasting conclusions. Linsley et al. (2006) did not establish a relationship between profitability and the risk disclosure quality. However, Samanta and Dugal (2016) conducted a comparable study on Indian banks and uncovered a positive relationship between profitability and risk disclosure. Conversely, Kuranchie-Pong et al. (2016) identified a negative relationship between performance and disclosure quality for Ghanaian banks. The respective findings beg the question whether the culture of the geographical location of the bank affects the link between performance and the disclosure quality. The geographical influence is mitigated in this research by

(8)

7

using only European banks. The present study aims to investigate whether profitability can affect risk disclosure quality.

The second determinant that is investigated is the state aid that banks have received during the last 10 years. State aid comprises a novel element in the respective research area. The link between state aid and the disclosure quality previously not been studied. The research on bank state aid mainly focussed on the disruption of the market discipline and the increase of risks (Miller, 2011). The state aid affected the risks negatively, but the relationship with the risk disclosure quality remains unclear. Garhardt and Vennet (2016) carried out an event study exploring the difference before and after receiving state aid. Their results accounted for the relationship between state aid and profitability, equity leverage and size. For all that, the change in disclosure quality is yet to be explained. The present study aims to contribute to the field by providing evidence of the relationship, if such is established, between state aid and the risk disclosure quality.

The stress test determinant is yet another novel element in this research area. The stress test is introduced by EBA in 2009. Prior research on the stress test and disclosure has mainly focussed on the difference between disclosure before and after the implementation (e.g. Flannery et al., 2016). The respective studies uncovered a positive relationship between the implementation of the stress test and its impact on disclosure quality. The present study focuses on the impact of the stress test results on the risk disclosure of individual banks. These outcomes aim to providing evidence regarding the relationship, if any, between the stress test results and risk disclosure quality.

1.3 Practical relevance

Besides its original scientific contribution, the present study also aims to achieve practical relevance. Risk disclosure is of interest to both companies and stakeholders. Risk disclosure increases bank transparency and decreases volatility (Healy & Palepu, 2001; Baumann and Nier, 2004). Moreover, both transparency and volatility affect firm value. The disclosure benefits are discussed against the theoretical background of this thesis in chapter 3. Together with other studies in the field of determinants of the disclosure quality, the practical relevance of this study is reflected in its identifying the determinants that affect the disclosure quality. When these variables and their impact on the disclosure quality are revealed, stakeholders can use this information in their

(9)

8

decision-making. The findings of this study paint a finer grained picture of the risk disclosure quality as well as its link to the financial information. Shareholders can use the outcomes of this study to provide accounts of a number of disclosure policies. Additionally, they can demand higher disclosure policies from banks. For supervisors, these insights could help implement better suited regulations that would increase the transparency of European banks.

(10)

9

2 Current legislation, regulation, and guidelines

In this chapter the current legislation, regulation, and guidelines will be discussed. The first section focuses on the International Financial Reporting Standards, whereas the second discusses the Basel III capital accords. These capital accords form the basis of the European Capital Requirement Regulation (CRR) and the Capital Requirement Directive IV (CRDIV).

2.1 International Financial Reporting Standards (IFRS)

The IFRS is a single set of high quality, understandable, enforceable, and globally accepted accounting standards developed by the IFRS foundation (The IFRS Foundation and the International Accounting Standards Board, 2018). These standards are developed to be universal accounting standards for improving the conformity within annual reports. Since 2005, the EU has made these standards mandatory for all listed European banks (European Commission, 2013). Non-listed European banks are not obliged to implement the IFRS legislation. Nevertheless, most of the non-listed banks comply with the said standards. The mission statement is to bring transparency, accountability, and efficiency to the financial markets by developing the IFRS Standards (European Commission, 2013). The IFRS legislation is divided into seventeen chapters. In the present study, the seventh and ninth chapter are of particular interest. Especially for banks, these two chapters are of interest because they are focussing on the valuation of financial instruments and the disclosure of financial reporting related information.

On the first of January of 2018, the IFRS 9 title was improved and re-released. The IFRS 9 title specifies how entities should be classified and financial assets and financial liabilities measured (The IFRS Foundation and the International Accounting Standards Board, 2018). Moreover, the standards for contracts to buy or sell non-financial items are fixed in this legislation. IFRS 9 title is of interest for banks as their balance primarily consists of financial assets and liabilities. The real value of these assets is more volatile than of non-financial assets. Therefore, the fixed disclosure of information about these assets is of interest in increasing the transparency and the generality between different banks.

IFRS 7 title sets the standard for the disclosures of financial instruments. According to the IFRS 7 title, the respective entities are required to provide additional information in their financial

(11)

10

statements.These disclosures enable the users to evaluate the significance of financial instruments, the nature and the extent of risks arising from them, and the way entities manage the said risks (International Accounting Standards Board, 2005). According to the IFRS 7 title, the companies are demanded to evaluate their financial instruments at fair value. The IFRS 7 title regulation describes four main ways of disclosing information on the asset or liability (International Accounting Standards Board, 2010):

 The maximum exposure to credit risk;

 The amount by which any related credit derivatives or similar instruments mitigate that maximum exposure to credit risk;

 The amount of change, during the period and cumulatively, in the fair value of the financial asset;

 The amount of the change in the fair value of any related credit derivatives or similar instruments that has occurred during the period and cumulatively since the financial asset was designated.

Banks are directly affected by this legislation because almost all of their assets and liabilities are financial instruments. Therefore, banks need to disclose a substantial amount of internal

information in their annual reports. The disclosure of this information forms the starting point for this research.

2.2 Basel III Accords, CRR and CRDIV

The Basel Committee on Banking Supervision (BCBS) introduced the Basel III Accords with the goal of promoting a more resilient banking sector (Basel Committee on Banking Supervision, 2011). A resilient banking sector is one that has the ability to absorb shocks arising from financial and economic stress. This is necessary because of the high risk of the banking sector transferring financial shocks from the financial sector to the entire economy (Basel Committee on Banking Supervision, 2011). The revision on the Basel Capital Accords was a response to the global financial crisis of 2007–2012 (Basel Committee on Banking Supervision, 2017).

(12)

11

The Basel Capital Accord comprises three pillars: (1) the minimum capital requirements, (2) the supervisory review process, and (3) enforcing market discipline by developing a set of disclosure requirements. This research focuses on the third pillar of the risk disclosures in annual reports. The BCBS introduced the third pillar to improve bank transparency. According to Brown et al. (2009), transparency became particularly relevant due to the growing demand from supervisors, governments, and other stakeholders to increase the reporting of financial and business risks. The third pillar comprises a set of standards for developing the disclosure requirements to promote market discipline (Basel Committee on Banking Supervision, 2015).

The capital requirements for banks in the European Economic Area are known as the CRDIV, which consists of the Capital Requirement Regulation and the Capital Requirement Directives. The said regulation is a legal framework for implementing the Basel III capital accords in the EU.

(13)

12

3 Theoretical background

The purpose of this chapter is to present the theoretical background employed by this study. In the first section, the agent-principal relationship will be explained. The agency theory offers an explanation as to why risk disclosure quality is necessary within the banking sector. Second, the responsibility of and the external pressure on banks will be discussed. The reasons for disclosing private information will be discussed in this context. Finally, risk information disclosure will be addressed.

3.1 Agent-principal relationship

The agent-principal relationship is a contract under which the shareholders (the principal) engage another person (the agent) to manage the firm. The agent performs a service on the behalf of the principal that involves delegating decision-making authority to the agent (Jensen and Meckling, 1976). Ownership and management are thus separated, which means that ownership and control are separated as well. Consequently, we can observe a separation between risk bearing and decision-making (An et al., 2011). Whereas the shareholders bear risks, the management are the decision makers. The agency problem occurs due to the unaligned goals of the agent and the principal. Both parties aim to maximize their own returns, so there are good reasons to believe that their interests may be in conflict (Jensen and Meckling, 1976). It is argued that the interest of the manager is in obtaining a high remuneration and other private benefits. The interests of the shareholders are conversely mainly in creating value of the equity of the company.

Next to the separation of ownership and control, the conflict of information asymmetry needs to be considered. The information asymmetry, which is a key concept within agency theory (An et al., 2011), occurs when the management has an information advantage over the shareholders. This is a consequence of the management’s direct access to internal information from the daily operations of the business, also referred to as private information. The information asymmetry has an increasing effect on the agency problem (Subramaniam, 2006). The information asymmetry needs to be reduced for the purposes of building the investor trust (Scott, 2009).

The most important risks for banks are the liquidity and credit risks because of high leverage and volatility of financial assets (Donnellan and Rutledge, 2016). From the point of view of the

(14)

13

company, the risks affect the cost of capital directly. Shareholders and lenders demand a higher return for their investments when the risks are higher. Therefore, reducing risks positively affects the cost of capital (Embong et al., 2012). On the other hand, shareholders aim at a return on their investments that corresponds to the risk involved. When the risks decrease and the return on investment remains the same, the value of their investment increases. It is beneficial for shareholders to reduce their investment risks.

Disclosing private information is a common way of reducing the agency problem and the information asymmetry. The annual report disclosures are an important channel for companies to expose publicly their information (Ronen and Yaari, 2001; Barakat and Hussainey, 2013). Li et al. (2008) argue that risk disclosure can be used as an external technique for monitoring monitor the behaviour of the management. It affects the decision-making process in that it decreases the opportunistic behaviour of the management.

3.2 Responsibility and pressure

Due to the central place of banks in the economy, the interest of shareholders must not overrun other interests (Dunbar and Clunie, 2013). Freeman (1984) argued that next to the shareholders, other parties are of interest for banks as well. Banks have an important position in providing financial services to stimulate the economy. As mentioned above (§1.1), due to the spillover effect of the banking sector, the importance of the perspective of stakeholders is on the rise. To create long-term value for the company, the company needs to be accountable in two branches, namely, the positive and the ethical branch (An et al., 2011). The positive branch accounts for the way of dealing with various demands of stakeholders. Based on the significance of the respective stakeholder, the company should implement a strategy to maintain a long-term and healthy relationship that benefits both parties (Deegan and Samkin, 2009). The ethical branch describes how companies are accountable for “fair” treatment of all stakeholders. They should thus provide information to all stakeholders on how they are affected by the company activities (Deegan and Samkin, 2009). Overall, according to this theory, a company, and especially a bank, needs to ensure accountability for their actions and their effects on other stakeholders.

(15)

14

An important stakeholder that banks need to take into consideration is the public opinion. On the one hand, the public has a vested interest in well-functioning banks due to their vital role in society. On the other hand, the public indirectly influences decisions. According to the legitimacy theory, companies are affected by external pressure, that is, by public opinion (Tilling 2004). Moreover, companies take actions to ensure that their activities and operations are perceived as legitimate (Whiting and Millar 2008). In this context, “legitimate” means that their operations are aligned with social norms (Sinclair & Bolt, 2013). If the company activities and the public opinion are not in alignment, a legitimacy gap appears (Andreaus et al., 2014).

A legitimacy gap for instance occurred in 2018 with the proposed salary increase for the ING CEO Ralph Hamers. The bank’s board of directors proposed a salary increase by almost 50% in order for the CEO’s performance, they argued, to be in line with the market (ING, 2018). A number of politicians, members of the media, and other individuals expressed their disagreement, because ING had required a capital injection from the Dutch government to survive the financial crisis. The social pressure in the end resulted in ING deciding not to increase the CEO’s salary (ING.com, 2018).

This example is a case in point of how social pressure can affect the decision-making of companies. The described activities did not affect favorably the business of ING. When a brand name is “polluted”, it could have serious consequences for the company. For example, Shell does have problems with finding highly educated beta students who are willing to work at Shell. Due to the polluted name, professionals are less willing to work Shell (FD, 2018). To maintain long-term value creation and avoid scandals, it is vital to ensure that the legitimacy gap is as small as possible (Tilling, 2004). Decisions, however, can never be met with unanimous approval. Therefore, it is important to seek establishing equilibrium among the interests of stakeholders. According to Scott (2009), the information asymmetry between the society and banks is always present. Moreover, it is of banks’ interest to reduce the information asymmetry in building not only investor trust, but also trust in society at large. The reason for this is the fact that information asymmetry also affects the legitimacy gap. Banks can reduce the information gap by increasing transparency. In other

(16)

15

words, by disclosing risk information in annual reports, companies can introduce signals that show their activities to be aligned with social norms. In what follows, the said signals will be explained.

3.3 Disclosure of risk information

The signalling theory describes the actions to reduce the information asymmetry between shareholders and the company. This theory is concerned with reducing information asymmetry between two parties, which is caused by adverse selection (Connelly et al., 2011). As mentioned above, information asymmetry leads to uncertainty among stakeholders. Less performance certainty leads to higher risk, and higher risk results inhigher cost of capital, which reduces the company value. Due to the information asymmetry, investors may not be willing to pay the optimum price for shares and securities (An et al., 2011). The management aims to reduce information asymmetry by signalling the level of risk to shareholders. Such practices indicate that a bank is managing its risk exposure and, by doing so, reducing its return volatility. A lower return volatility results in building a more stable company and, therefore, more stable investments. Finally, shareholders are seeking to find stable investments with low volatile results. The said theory focuses on arguing for disclosing information to avoid the undervaluation of shares and to reduce information asymmetry (Giner, 1997).

Information disclosure is also useful for investors because it affects their decisions and provides more certainty regarding future performance (Nier & Baumann, 2004). Risk disclosure increases company transparency and it decreases the information asymmetry of a company (Healy & Palepu, 2001). Baumann and Nier (2004) also argue that a qualitatively high-risk disclosure lowers stock volatility.

Signalling theory suggests that, managers tend to reveal good news and withhold bad news (Verrecchia, 1983). Based on this assumption, Kirmani and Rao (2000) argue that there are two kinds of information signalling strategies. Namely, high quality signalling companies that benefit from signalling, and low quality signalling companies that do not benefit from signalling.

(17)

16

Based on this assumption and according to the signalling theory, the higher risk companies are willing to disclose more information (Elshandidy et al., 2013). High risks involve a higher degree of opacity, which can, however, be reduced by disclosing information. Elshandidy et al. (2013) also found that well-performing companies are willing to disclose more private information, and, consequently, they are more inclined to disclose their good performance. Khlif and Hussainey (2016) argue that under good ﬁnancial performance, risk disclosure may reduce uncertainty regarding future cash ﬂows and economic environment.

The demand for disclosure challenges market participants not only to provide information, but also to place that information in a meaningful context (Themistokles et al., 2008). Hereby, the quality of the disclosed information is as important as its quantity. Disclosures without the optimal level of quality are useless and do not lead to a decrease in risks. In this research the quality and quantity of risk disclosures in annual reports of European banks will both be measured. Therefore, this research will not only measures the quantity, but also the quality of the risk disclosures.

The selection of the signalled or disclosed information should be well defined. It should first be considered whether or not specific information should be included in the disclosure, due to it potentially building or destroying company value. For example, too much risk disclosure could lead to an “information overload” among investors, making it difficult for investors to distinguish between the valuable high quality information and the unnecessary low quality information. This will consequently decrease the intended effect of the disclosure. The risk of “boilerplate” risk disclosure also needs to be taken into account. Boilerplate risk disclosure is a standardized disclosure that leads to lower information quality and thus to its lower value. The big four accounting firms independently warn about the large quantity and the low quality of disclosure (EY, 2014, Deloitte 2013 & PWC, 2017). To maintain qualitatively good disclosures, the corporate reporting must be closely aligned with the investors’ and stakeholders’ versatile needs. For these needs is not a one-size-fits-all model. Every stakeholder has its own needs. To ensure a fair representation of the stakeholders needs, this research use the point of view that is presented by the supervisors and government agencies. The approach to measuring the disclosure quality will be discussed in chapter 5.

(18)

17

4 Hypotheses

The present study is divided according to three sub-questions. In this chapter, the sub-questions will be discussed and substantiated. The substantiation will be done based on the theoretical framework discussed in chapter 3.

4.1 Profitability

According to the signalling theory, the uncertainty about future profitability increases the potential impact of risk disclosure (Elshandidy et al., 2013). Investors and other stakeholders want to know what to expect regarding future company performance. The signalling theory states that well-performing organisations have more incentive to disclose information. Well-well-performing companies benefit more from the disclosures because their risk of mispricing is higher. Therefore, Whiting and Millar (2008) argue that the possible threat of mispricing is higher for well-performing companies. Moreover, managers are held responsible for the company performance. If a company performs well, managers are willing to share that information externally. They do so in order to demonstrate that they are capable of doing their job well. Their remuneration is performance-dependent, so they have a financial interest in disclosing positive performance. This theoretical claim, however, stands in contrast to the findings of Jordan et al.’s study (1999) of US banks. They found that twenty years ago, poorly performing US banks were also more likely to produce low quality risk disclosure. This was due to the fact, the researchers claimed, that poorly performing banks were attempting to reduce their loss of reputation. To reduce the reputation loss, they are willing to disclose more information to show that their performance is not as poor as it seems. Their finding is in line with a more recent study of Samanta and Dugal (2016) who reported similar patterns in Indian banks. However, other studies did not identify relationships between profitability and bank disclosure (e.g. Linsley et al., 2006; Sharif and Ming Lai, 2015). This effect could be demographically explained. The basic assumptions of the signalling theory contradict the findings of the other studies. This begs the question whether the signalling theory can actually be applied to European banks. In order to test the perspective of signalling theory, the following hypothesis is thus formulated:

(19)

18

4.2 Government interference / State aid

Too big to fail (TBTF) is a common concept within the banking industry. It describes a bank so large that a government will provide assistance to avoid its bankruptcy because not doing so would have a disastrous effect on the economy. It is also referred to as a systemically important financial institution (SIFI). In the banking crisis of 2007–2012, many banks faced considerable problems. Some banks required governmental assistance through state aid, because they were too important to go bankrupt. Altogether 114 European banks were provided with state aid (Gerhardt and Vennet, 2016). Banks aware of their TBTF status act less conservatively because of the lower risks of bankruptcy (Cabrera et al., 2016).

Governmental interference disturbs the agent-principal relationship explained in the agency theory. In a normal relationship, investors face losing their investments when the company goes bankrupt. Bailouts and financial aid guarantee or even enable governmental assistance. The risk thus remains the same, yet the actors bearing the risks shift due to the fact that TBTF banks are saved anyway. Investors could, however, lose some of their investments if the said banks are nationalized. Moreover, when a bank is already nationalized, the chance of bankruptcy is low. For example, during the nationalization of SNS Reaal (European Commission, 2013), the shareholders were essentially wiped out and lost all of their investments. The bond-holders where spared in this rescue. This case shows that even when a bank is nationalized, not all investors necessarily lose their investments.

The distortion in the relationship decreases the incentive to act in a risk-averse manner because the risk-bearing party are no longer the investors but the society at large. The higher probability of state aid decreases the risks involved and, thus, the incentiveto take risks increases. The willingness of companies to disclose information can thus be affected, which subsequently leads to a lower quality of risk disclosure. The term “market discipline” is used to refer to the relationship between the riskiness of banks and the stakeholders’ response. Cabrera et al. (2016) found that, when state aid is expected, market discipline decreases or even vanishes completely. Consequently, banks have lower incentives to disclose information, as it is already clear that the government will protect the bank. The agency risks are thus disrupted, so the incentive to act responsibly in the context of

(20)

19

the stakeholders’ demand for risk disclosure is not functioning. This situation is reflected in the following hypothesis:

The risk disclosure quality of European banks that are not subject to governmental interference is lower compared to banks that are subject to governmental interference.

4.3 EU-wide stress testing

During the financial crisis, 114 banks benefited from state aid, which were mainly expensive forms of support by governments (Gerhardt and Vennet, 2016). The turmoil thus caused was disastrous for the EU economy. The EU aimed to reduce the number of bank bailouts. One of the reactions by the EBA was the introduction of a stress test, a risk management tool that evaluates banks on their internal risk management. The stress test is used to evaluate the financial positions of banks at several macroeconomic scenarios as part of their internal risk management (EBA, 2018). The result is shown as a rating between the threshold value and the adverse economic performance hypotheses. The test results in a score between -1 and -1500. The more lower the score, the more venerable the bank is to economic downfall.

Flannery et al. (2016) compared single banks before and after the stress test implementation. They found that the implementation of the stress test lead to a decrease in the amount of private information. This shows that the stress test overall positively impacts transparency. The same group of researchers also found that the implementation of the stress test had a higher impact on banks with a high level of risk. As interesting as this finding is, it does not account for the relationship between the test results and the quality of disclosure.

Poor performance on the EBA stress test means indicates an increased risk for investors that want to invest in the respective bank. Flannery et al. (2016) argue that banks described as “highly risk” are willing to disclose more information due to the larger incentive. They are more inclined to disclose information as disclosures account for certain risks. Also, the disclosure impact could be higher when the investment risks are higher as well. Cho et al.’s (2012) study on corporate social responsibility (CSR) led to similar results. According to the researchers, the companies with poor performance on CSR issues disclose more information to decrease the effect of poor performance.

(21)

20

According to the legitimacy theory, to achieve long-term value creation, companies must strive to keep the legitimacy gap as small as possible (Tilling, 2004). A poor stress test score increases the said legitimacy gap. In case of poor performance, a company is willing to disclose more information to decrease the respective effect. The next hypothesis is thus formulated as follows:

(22)

21

5 Methodology

In this chapter, the research methodology is discussed. First, the overall design of this research will be discussed along with the basic design, sample, and the statistical model applied in the study. In the second section, the depended variables will be discussed, and the disclosure index in measuring the risk disclosure quality will be explained. In the third section, the independent variables will be discussed, the method used in measuring them, and the extent to which they contribute to testing the hypotheses or answering the research question. Last, the control variables will be discussed as well as their effect on the relationship between the dependent and the independent variables.

5.1 Research design and sample

This research investigates the relationships between determinants and the quality of risk disclosure in the context of European banks. As discussed in the introduction, the banks have a central role in the economy due to the spill over effect of negative events from the banking to other industries. This was demonstrated during the financial crisis of 2008. Partly due to this central role, supervisors made it mandatory for banks to produce higher quality annual reports. Also, the characteristics of banks differ from those of non-financial companies. A bank has a low equity percentage, and assets and liabilities more volatile in value than other companies. Such characteristics comprise risk-increasing factors. As discussed in chapter 3, a risk increase leads to higher disclosure of demand and supply. This research focuses only on European banks because of the comparable regulations across banks in European countries (Huttenhuis and Ter Hoeven, 2015).

For this research, a sample of altogether 75 European banks was selected. The banks were selected if they were deemed to be major EU banks. In total, 47 banks under the ECB supervision and 28 banks under local supervision were chosen for the analysis. Some banks included in the original sample did not publish their reports at the time when the research was conducted. These banks were thus replaced with other banks with comparable characteristics, based on country, size, and supervisor. No banks from Cyprus, Lithuania, Latvia, and Slovenia were included in the sample due to their insignificant role in the EU banking industry. Chart 1 illustrates the distribution of the sample.

(23)

22

Chart 1

Information could also be disclosed in a number of different ways, including, for example, additional reports. For the purposes of this research, only the annual reports were used. At the time this research was conducted, all banks mandatorily published their annual reports for the fiscal year 2017.

5.2 Dependent variables

The dependent variables in this research are the variables used for measuring the risk disclosure quality. For banks, it is mandatory to disclose certain information. Additional to these mandatory disclosures, banks could risk disclosure risk information. The present research is focussing on these voluntary risk disclosures. Seeing that voluntary risk disclosure quality is an abstract concept that cannot be directly measured, a disclosure index is initiated (Marston and Shrives, 1991). This disclosure index is a checklist of thirty points that should be included in a good quality annual report. This collection method is referred to a content analysis. According to Steenkamp and Northcott (2008), there are three main types of risks occurring in content analyses.

Distribution of the sample

Germany : 8 France : 7 Italy : 6 Netherlands : 6 United Kingdom : 6 Spain : 5 Sweden : 5 Belgium : 4 Denmark :3 Ireland : 3 Austria : 2 Portugal : 2 Estonia : 2 Norway : 2 Croatia : 2 Czech Republic : 2 Finland : 2 Greece : 1 Hungary : 1 Malta : 1 Luxembourg : 1 Bulgaria : 1 Poland : 1 Romania : 1 Slovakia : 1

(24)

23

First, the recording units are subject to bias, because the researcher determines the selection of index content. Measuring and weighting every aspect fairly is difficult. To mitigate the risk of bias, the disclosure index is based on standards used by supervisors. These thirty measurements are based on the Enhanced Disclosure Task Force (EDTF) of the financial stability board (FSB), and the third pillar of Basel Capital accords III. The checklist is divided into four categories:

 Seven elements for risk in general;  Eight elements for credit risks;  Five elements for liquidity risks;

 Ten elements for solvency or financial strength.

The composition of the disclosure index is included in appendix 1.

Second, coding repetitive messages could harm the validation. The method of evaluating the selected disclosure checks is potentially problematical. To mitigate the risk of incorrect valuating a specific item, the choice was made to measure in the item in a three-fold manner. Each item was assessed on a scale from:

0. Non-existing or insufficient

1. Minimal of average amount of information included 2. Well-elaborated

Third, the items were subjectively described. The difference between different scores on the scale was subject to researcher judgment. This should be considered with regard to the outcome. To mitigate bias, clear descriptions of the three classes were introduced in the beginning of the process.

In this study, five different researchers collected the data. Seeing that a situation can be judged differently, there is an increased chance of obtaining a biased result. To decrease this risk, we aimed to implement homogeneity in judging situations. First, the evaluation of the first two banks was discussed. After this, the outcome, process, and discussion were jointly analysed in the whole group. This process clarified the process of evaluating annual reports. In collecting the large sample, the first eight of fifteen evaluations were reviewed by a fellow researcher. The outcome of

(25)

24

both findings was discussed together with the respective colleague. This approach ensured an acceptable level of data validity.

5.3 Independent variables

The independent variables are the determinants of this research. Next, the independent variables of the three determinants, i.e. profitability, stress test, and state aid, will be discussed.

Profitability

The profitability is measured by the return on equity. The return on equity is measured by the profit after tax for the year 2017 and divided by total bank equity at the end of 2017. The return on equity assumes a great relevance to profitability as it measures how effectively capital is utilized to generate profit for company’s shareholders (Kharatyan, et al. 2017). The necessary information is obtained from the annual reports. The profit that is shown in currency other than euro is re-calculated to a euro standard according to the exchange rate on the 31st of December 2017.

State aid

There are several ways of government potentially interfering with the banking industry. Therefore, clear boundaries need to be drawn between different forms of aid received between January 2007 and January 2016. In the present study, state aid is divided into three groups. Although several governments used guarantee schemes to help banks, they alone were not considered to be forms of direct financial aid. The guarantee schemes did increase the moral hazard effects (Gruendl & Guettler, 2010). Because almost all banks in Europe have used directly or indirectly these guarantee schemes, the banks receiving only the government guarantees were placed in the first group. The second group comprised banks that received direct financial aid. The government provided direct financial assistance to banks through recapitalization, direct loans, or other forms of financial aid. Relevant for this type of aid is the fact that banks are required to repay it. The third group consists of nationalized banks or banks already owned by governments. For this research, a nationalized or government owned bank is one controlled by a government, which owns more than 50% of the bank shares.

(26)

25

In the present study, government interference was measured for the period between January 2007 and January 2016. The starting date was chosen due to it being the year of the beginning of the economic crisis. During the crisis, governments interfered multiple times in the banking industry. These events comprise the appropriate context for testing the hypothesis. The analysed period extends to January 2016 because the influence of the government interference requires time to affect the annual report quality. For example, the annual report for 2017 will not be affected by the state aid received in January 2018. The data was manually selected from the database “State Aid Cases” of the European commission (EC). When the information available in the database did not provide appropriate evidence, other sources were used such as the local government information sources, bank websites, and publications of the ECB, EC, banks or other reliable sources. The list of types of state aid per bank is included in appendix 2.

Stress test

The stress test variable is a score measuring the misalliance of the bank to economic events (European Banking Authority, 2016). The banks are evaluated on the basis of an adverse scenario on certain financial ratios. The score that defines the results is called “Delta Adverse 2018” and it is expressed in basis points (BPS). The test is performed by the European Banking Authority, and its results are publically available. The results of the 2016 test will be used, because the numbers of the 2018 test were not available at the time this research was conducted. The stress test variables will be collected from the report: “2016 EU-Wide Stress Test” (EBA, 2016). The EBA reports their results in six ratios. This study employs the transitional total capital ratio and the fully loaded leverage.

The transitional total capital ratio is the ratio of all paid-up capital. The transitional total capital ratio is calculated by the sum of the CET1, additional Tier 1 and the Tier 2 capital, divided by the risk weighted assets (RWA). The regulators use this ratio to determine the bank’s capital adequacy. The RWA is the value of the assets adjusted to the risks of the assets. The higher the risks on the assets of the bank, the higher the amount of the RWA. The level of the RWA determines the amount of total capital that must be held, to meet the regulators’ required capital ratio. The Tier 1 capital consists of the Common Equity tier 1 (CET1) and the additional tier 1. The CET1 is the core capital

(27)

26

of the bank, like common shares, and retained earnings. The additional tier 1 capital consists of instruments that are not common equity but are eligible to be included as CET1 (Basel Committee on Banking Supervision, 2011). For example, certain instruments that convert into equity when a trigger event occurs. Tier 2 consist of undisclosed reserves, revaluation reserves, general provisions, hybrid debt capital instruments, and subordinated term debt (Committee on Banking Supervision, 2011). The CET 1, Tier 1 and Tier 2 together forms the capital of the bank. This ratio provides a good overview of CET1, Tier 1 adjustment, and the Tier 2 capital.

The fully loaded leverage ratio is defined as “a capital measure divided by the exposure measure” (Basel Committee on Banking Supervision, 2016). The fully loaded leverage ratio is defined as the capital measure divided by the exposure measure. The capital measure for the leverage ratio is the Tier 1 capital of the risk-based capital. The exposure measures are the on-balance sheet, non-derivative exposures (Basel Committee on Banking Supervision, 2014). The fully loaded leverage ratio differs from the transitional leverage ratio, because the fully loaded leverage ratio is calculated by using the new, stricter regulations. The fully loaded leverage ratio gives good overview about the core capital related to the total assets.

5.4 Control variables

Two control variables were included in the present study. These variables affected the relationship between the determinants and the depended variables. The said relationships had already been established in previous research.

The first control variable used in this research is bank size, which positively correlated with the quality of risk disclosure (e.g. Linsley et al. 2006). In the present sample, there is a wide range in the size of the 75 banks. The smallest among the banks, UBI Banca, has a size of 127 million euro in total assets. The largest bank, HSBC has a size of 2,105,910 million euro in total assets. HSBC is thus 16,500 times the size of UBI Banca. Seeing that previous research has uncovered the effect of size on quality, this variable must be taken into account as control.

Second, supervisors can affect the risk disclosure quality. Supervisors and the required regulations affect the annual report quality (e.g. Palea, 2013. & EY, 2014). Strict supervisors are more likely

(28)

27

than lenient supervisors to impose strict rules on banks. Therefore, strict supervisors are expected to correlate positively with quality. Huttenhuis and ten Hoeven (2016) concluded that banks under supervision of the ECB are more likely to disclose information of certain risks. In their sample, the English banks were fell under the supervision of the ECB. Back in 2006, Llewellyn (2006) came conclusion that the ECB and the Bank of England (BoE) are strict regulators. Therefore, both the ECB and the BoE are defined as strict supervisors in this study.

(29)

28

6 Results

In the previous chapter, the methodology and the statistical tests for this research is discussed. In this chapter, the results of the performed tests will be discussed. To start with, data from the sample will be discussed. In this paragraph, the pre-tests and data restructuring are explained. This will be discussed in the first paragraph. In the second paragraph, the robustness will be secured, by performing a multi correlation test, and a Cronbach alpha test. This is followed by the third paragraph, where the results of the multiple regression tests will be discussed.

6.1 Descriptive statistics

The “index score: quality of risk disclosure” is the dependent variable in this study. The sample consists of 75 cases. The lowest score was 11 index points of the Volkswagen Financial series AG and the highest score was that of Standard Chartered Plc with 42 points. The mean of this sample is 26.41 points, with a median of 26 points. There were no missing values in the disclosure index, profitability and state aid variables in the cases. Thirty banks that were not subject to the ECB stress test were excluded for the ECB stress test analysis. This is due to the fact that a limited number of banks underwent the ECB stress test. Table 1 includes an overview of the observed cases, means, median, and the minimum/maximum of the dependent, independent, and control variables. The data per bank is included in appendix 2.

N Mean Median

Std.

Deviation Minimum Maximum

Dependent variable Valid Missing

Total Index Score 75 0 26.41 26.00 7.604 11 42

Independent variable

Profitability 75 0 7.08% 7.91% 8.72% -33.58% 34.22%

Transitional Total Capital Delta

45 30 -483.20 -451.00 247.884 -1,492 -10

Fully Loaded Leverage Data 45 30 -105.89 -76.00 106.753 -582 -1

State aid 75 0 0.59 0.00 0.660 0 2

Control variable

Supervisor 75 0 0.71 1.00 0.458 0 1

Total assets (In million Euro) 75 0 364,679 165,379 482,785 1,738 2,104,910

(30)

29

The proposed tests can only be used if the sample is normally distributed. The Shapiro-Wilk test was used to examine the normal distribution of the sample. The test results, summarized in Table 2 and Graph 1, show the sample to be normally distributed at a 0.1 significance level. A lower significant level could be occurred by merging quality scores into groups with a 5 point range. However, this would decrease the information value of the sample.

Shapiro-Wilk Test Statistic Df Sig. Total 0,970 75 0,077* * significant at a 0.1 level Table 2 Graph 1

Next, the control variables “strict supervisor” and “size” need to be computed for test use. For the control variable, “strict supervisor,”a dummy variable is created in order to test its relationship with disclosure quality. As mentioned above, strict supervisors are the ECB and the BoE. The strict supervisor is assigned the value “1,” whereas the other supervisors are assigned the value “0.” The control variable “size” needs to be recalculated, because bank size in these cases is expressed on a

(31)

30

logarithmic base. By recalculating the size from linear to logarithm, the control variable can be used in tests. This control variable is referred to as “LogSize.”

The test results of both the control variables are summarized in Table 3. Two single linear regressions were conducted in this test, which included the constant in equity. Both LogSize and Supervisor show individually a positive results significant at the 5% level. In other words, both size and strict supervisors correlate positively with the total score. When both the control variables are put together in an OLS regression, it gives a R squared of 0.267. The R squared is the total proportion of the variance for a dependent variable that is explained by an independent variable. This R squared will be compared to the R squared of models when more independent variables are included.

Control variables

Unstandardized Coefficients

Standardized

Coefficients t Sig. B Std. Error Beta `Beta

LogSize (constant) -2.974 5.808 -0.512 1.610 LogSize 5.693 1.287 0.517 5.085 0.000 Supervisor (constant) 23.591 1.584 14.892 0.000 Supervisor 3.994 1.884 0.241 2.119 0.037 Unstandardized Coefficients Standardized Coefficients t Sig.

B Std. Error Beta `Beta

(constant) -2.974 6.008 -0.449 .655

LogSize 5.588 1.241 0.507 4.502 0.000

Supervisor .374 1.859 0.023 .201 0.841

R Squared 0,267

(32)

31

6.2 Multi-correlation and the Cronbach’s Alpha

Next, the multi-correlations between the variables are tested. A multi-correlation means that different independent variables predict the same reaction in the dependent variables. If this is the case, the information value of the independent variables are at risk. This could bias the results in ways that one of the predictor variables gives an incorrect statement about the influence on the dependent variables. This effect is tested with the Pearson test. The outcome of this test is presented in the Pearson correlation matrix in table 4. The multi-correlation effect occurs when different variables are ma the score is significant. In general a score lower than 0,7 is acceptable. It is not allowed to put them into one model when the variables are multi-correlated. The Pearson Correlation shows that the two stress-tests are correlating with each other above 0,7. This means that a combined model of those two and the profit is not allowed and the model is removed from the research. This table also shows that the dependent variable, the index score, is correlating with the control variable size and supervisor. This is in line with the previous research, that is discussed in the previous paragraph.

To test the reliability of the disclosure index, the Cronbach's alpha test is used. The Cronbach's alpha tests the internal consistency of the dependent variables. The score could be between minus infinity to one. The test score increase as the inter correlations among test items of the disclosure checklist increase. The rule of thumb is that the Cronbach’s Alpha score above 0,7 is acceptable. The unreported Cronbach’s Alpha score for the dependent variables of this research is 0,733. This means that the dependent variable in this research is reliable.

(33)

32 Pearson Correlations Index score Profitabilit y ECB stress test 1 ECB stress test 2

State aid LogSize Supervisor

Index score 1

Profitability winst/equity -0,180 1

ECB stress test 1 Transitional total capital delta

-0,194 0,603** 1

ECB stress test 2 Fully loaded leverage delta

-0,086 0,627** 0,709** 1

State aid 0,194 -0,051 -0,459** -0,256 1

LogSize 0,517** -0,122 0,091 0,308* 0,033 1

Supervisor 0,367** -0,273 -0,312 -0,133 0,096 0,497* 1

* correlation significance at the 0.05 level ** correlation significance at the 0.01 level

Table 4

6.3 Regression

This paragraph discusses the results of the OLS regression analysis, which examines the relationship between the independent and the dependent variable. In section 5.1, the tested models were introduced. The results pertaining to the said models will be discussed in what follows.

Profitability

The hypothesis is that profitability positively affects disclosure. The hypothesis is formulated based on the assumption of the signalling theory that firms aim to signal good performance to their stakeholders. To test this hypothesis, linear regression is carried out on the entire sample consists of the observed banks. The results of linear regression are shown in Table 5.

(34)

33 Profitability Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta LogSize 5,609 1,238 0,509 4,532 0,000 Supervisor -0,324 1,942 -0,020 -0,167 0,868 Profitability -0,112 0,093 -0,129 -1,201 0,234 R Sqaured 0.282 Table 5

The value of 0.05 is taken as a cut-off for significance. The results indicate a small negative beta at a significance level of 0.234 The negative beta is in line with the hypothesis, however, the relationship between profitability and the risk disclosure quality is not significant. The R squared could be compared to the R Squared of the control variables. Here we see that the profitability increased the R squared with 0,015 point. This is not a large increase, what means that profitability does not explain a large part of the total quality.

The test was repeated after excluding the outliers in the dependent variable. The outliers in this research are the highest and lowest five cases in the sample. The said cases excluded from this test are included in appendix three. The outcome of the tests is shown in Table 6.

Profitability Unstandardized Coefficients Standardized Coefficients t Sig. B Std. Error Beta LogSize 5,269 1,377 0,487 3,826 0,000 Supervisor 0,696 2,233 0,043 0,312 0,756 Profitability 0,132 0,278 0,063 0,476 0,636 R squared 0.235 Table 6

(35)

34

Table 6 shows that even without the outliers, the test results are still not significant. The significance decreased from 0,234 to 0.636. This means that the relationship is less explainable when the outliers are excluded from the dataset. This model shows that the R squared is lower when the outliers are excluded. These results show that the exclusion of the outliers, decrease the proportion that is explained. Both Tables 5 and 6 show that the risk disclosure quality can mostly be explained by the control variable “LogSize,” i.e. bank size.

State Aid

The second hypothesis focuses on the effect of state aid on the risk disclosure quality. As discussed in chapter 5, state aid is a nominal variable divided in three groups. This variable was tested by using the one-way analysis of variance (ANOVA). The ANOVA is used to determine whether the means of the different samples differ statistically significantly from each other. First, Table 7 shows the statistics relating to different state aid classes. Table 8 summarizes the results of the ANOVA test.

State aid statistics

N Mean Median

Std.

Deviation Minimum Maximum

Valid Missing

No aid 38 0 25.26 25.50 7.277 11 42

Financial aid 30 0 26.93 26.50 7.930 14 38

Nationalisation 7 0 30.43 30.00 7.345 18 40

Table 7

One-way ANOVA test: State aid

Sum of Squares df Mean Square F Sig.

Between Groups 171.237 2 85.619 1.501 0.230

Within Groups 4,106.949 72 57.041

Total 4,278.187 74

(36)

35

Table 7 shows that the means of the “No aid,” “Financial aid,” and “Nationalization” categories are not equal in the analysed sample. The “State aid” mean is the lowest, followed by the “Financial aid” mean, which is 1.67 point higher, and the “nationalization” mean, which is the highest. The difference in means is considered to be significant at the cut-off point of 0.05. The differences in means are not significant, and it, therefore, cannot be concluded that there are differences across the sample. This can be explained by, on the one hand, the small sample size, and, on the other, by the comparable range between the minimum and maximum values of the classes. Based on these outcomes, it cannot be confirmed that state aid affects disclosure quality.

ECB stress test scores

The testing of the third hypothesis involves exploring whether the European stress test affects outcomes on risk disclosure quality. As explained in chapter 4, the implementation of the European stress test has increased the annual report quality (Flannery et al., 2016). Disclosing information decreases stakeholder risk due to lower information asymmetry. This effect increases when the risks are higher. The investment risks are higher when banks perform poorly on the stress test. This section explores whether poor performance increases disclosure quality.

Only 45 cases were included in the present analysis seeing that not the ECB stress test scores were not available for the entire sample. To test this hypothesis, a OLS regression was performed with both the stress test delta scores. The cut-off for significance is 0.05. The results of the OLS regression results are shown in Tables 9 and 10.

(37)

36

ECB stress test 1

Unstandardized Coefficients Standardized Coefficients

t Sig.

B Std. Error Beta

LogSize 9,365 2,304 0,560 4,066 0,000

Supervisor -0,605 2,543 -0,035 -0,238 0,813

Transitional total capital delta -0,008 0,004 -0,262 -1,860 0,070

R squared 0,339

Table 9

ECB stress-test 2

Unstandardized Coefficients Standardized Coefficients

t Sig.

B Std. Error Beta

LogSize 10,676 2,456 0,638 4,346 0,000

Supervisor -0,951 2,528 -0,055 -0,376 0,709

Fully loaded leverage delta -0,021 0,010 -0,314 -2,137 0,039

R squared 0,355

Table 10

The beta values of both the transitional total capital delta and the fully loaded leverage delta are negative. According to the hypothesis, the higher the score, the lower the disclosure index. The beta values are in line with the hypothesis. The stress test score of the transitional total capital has is significant at a 0.07 level. This score is close to the cut-off for significance is 0.05, but still not enough to accept the significant influence. By adjusting the transitional total capital delta, the R squared increases with 0,072. This variable shows that this stress test score does explain a part of the quality. The significance score of the fully loaded leverage delta is 0.039. This means that the “Fully loaded leverage delta” has a significant influence in the risk disclosure quality in annual reports. of the variables show significant results. This is also shown in the R squared score. By adjusting the fully loaded leverage delta, the R squared increased with 0,088. The small beta score explains that even though the score is significant or almost significant, and the R squared increases quite a bit, the effect of this independent variable is limited.

(38)

37

This test was repeated without the outliers. The removed cases are included in appendix 3. The results of the tests conducted on the dataset excluding the outliers are shown in Tables 11 and 12.

ECB stress test 1 without outliers

Unstandardized Coefficients Standardized Coefficients T Sig. B Std. Error Beta LogSize 8,958 2,788 0,566 3,213 0,003 Supervisorecb 0,117 3,111 0,007 0,038 0,970

Transitional total capital delta 0,000 0,024 -0,002 -0,014 0,989

R squared 0.321

Table 10

ECB Stress test 2 without outliers

Unstandardized Coefficients Standardized Coefficients T Sig. B Std. Error Beta LogSize 8,390 2,682 0,530 3,128 0,004 Supervisorecb 0,462 2,848 0,026 0,162 0,872

Fully loaded leverage delta 0,005 0,009 0,085 0,520 0,607

R squared 0,327

Table 11

The results of this test shows that the significant score the fully loaded leverage delta and the significance score of the transitional total capital delta increases. Now both scores fail to reach the significance threshold. Also the R squared score of both variables is lower without the outliers. This means that the test without the outliers does not provide any evidence that these score affect the risk disclosure quality. Whether or not this result is informative about the general population is questionable. The external validity of this test is weak due to the small sample size of only 45 banks. When the ten outliers are excluded, the external validity decreases further. Therefore, there is no evidence that the risk disclosure is affected by the stress test results.

(39)

38

Independent variables in one model

In this model, all the independent variables are include. This test is performed to measure the maximum proportion of the variance for a dependent variable that is explained by an independent variables used in this research. In paragraph 6.2 the multi correlation is tested. The results of this test was that both ECB stress test scores where correlating on a level of 0,7. This means that both variables are not allowed in the same model because these two variables correlate too much. Therefore, two models are performed including one of the stress test scores. The results of both tests are presented in table 12 and 13.The results of both tests show that the fully loaded leverage delta explains more about the risk disclosure quality than the transitional total capital.

Total model with stress test 1

Unstandardized Coefficients Standardized Coefficients T Sig. B Std. Error Beta LogSize 9,559 2,316 0,572 4,127 0,000 Supervisorecb -1,878 2,829 -0,108 -0,664 0,511 Transitional total capital delta -0,005 0,006 -0,179 -0,928 0,359 Profitability winst/equity -0,017 0,181 -0,017 -0,096 0,924

State aid 1 2,429 1,865 0,217 1,303 0,201

R squared 0.370

Table 12

Total model with stress test 2

Unstandardized Coefficients Standardized Coefficients T Sig. B Std. Error Beta LogSize 11,323 2,513 0,677 4,506 0,000 Supervisorecb -2,433 2,761 -0,140 -0,881 0,384 Profitability winst/equity 0,087 0,180 0,086 0,484 0,631 Fully loaded leverage delta -0,023 0,013 -0,343 -1,802 0,079

state aid 1 2,569 1,626 0,229 1,580 0,122

R squared 0.407

Table 13

The results of both tests show that the fully loaded leverage delta explains more about the risk disclosure quality than the transitional total capital delta. The independent and control variables that are used in this research, could have a maximum R squared score of 0,407. This means that 40,7 percent of the variance of the risk disclosure quality is explained by these variables.

The risk disclosure quality of European banks