Risk reporting and the cost of equity for
European banks: A content analysis
Master thesis, Msc Accountancy & Controlling, specialization Controlling University of Groningen, Faculty of Economics and Business
DEFINITIVE VERSION 20 January 2020 TIJMEN MENNINK Studentnumber: 2711346 Jacobijnerstraat 12b 9712 HZ Groningen Tel.: +31 6 24191944 e-mail: tpmennink@gmail.com
Primary supervisor/ university J.G. Huttenhuis
Secondary supervisor/ university R.L. ter Hoeven
1
ABSTRACT
Excessive risk taking of banks was one of the key causes of the financial crisis in 2007. Risk reporting has since been identified as one of the key tools to increase the resilience of the banking industry and should help to prevent such economical disasters from happening in the future. Standards, like Basel III, stress the importance of risk reporting and focus for a large part on market discipline to control risk taking from banks. Market discipline relies on the market to reward (punish) banks for avoiding (taking) unwanted risks. A way to do this is by requiring a lower (higher) return on their investment, decreasing (increasing) the cost of equity of banks. In this paper I investigate if investors indeed reward banks for better risk disclosure quality and less ambiguous reports by lowering the cost of equity. My sample consists of 129 annual reports from 38 European banks between 2015 and 2018. Risk reporting quality is measured using a content analysis, while disclosure ambiguity is measured using the analyst forecasts dispersion of earnings per share forecasts. Cost of equity is measured using the modified PEG-model from Easton (2004). The results of this research contradict the expectations created by market discipline and suggest a positive relation between risk reporting quality and the cost of equity for European banks. The expected relation between disclosure ambiguity and the cost of equity is confirmed and found to be positive. Banks should therefore not expect an increase in disclosure quality to be rewarded. Decreasing ambiguity however, should help in lowering the cost of equity. If standard setters and regulators wish to further increase market discipline and transparency, they should consider implementing stricter standards as, at the moment, market discipline seems to be ineffective at increasing risk reporting.
Keywords: Risk reporting, cost of equity, disclosure ambiguity, content
analysis, Basel III, market discipline
2
1. INTRODUCTION
Banks have been at the heart of the financial crisis that started in 2007. Starting in the United States, where banks were giving out an increasing amount of mortgages and were growing less careful with their quality, leading to a lot of bad loans. The bad loans, declining house prices and increasing short-term interest rates led to a significant increase in mortgage defaults and weakening the financial position of banks. For some banks the impact was so big that they were in danger of collapsing. In the United Kingdom, Northern Rock requested liquidity support from the UK government to avoid a collapse, starting a bank run from customers that were afraid to lose their savings. In the United States several banks needed to be bailed out by the government to avoid a collapse, also spurring bank runs. After spending a lot of tax payer money to save banks, the US government was no longer able to bail them out, leading to the collapse of the Lehman Brothers bank and marking the beginning of the global financial crisis (Johnson and Mamun, 2012). In the maelstrom of the crisis, banks everywhere in North America and Europe collapsed or had to be nationalized (Gertler and Gilchrist, 2018). This heavily damaged the economy in most developed countries and lead to the Great Recession.
The financial problems of banks at the time of the crisis were caused by excessive risk taking (Johnson and Mamun, 2012). Especially firms that relied heavily on mortgage loans were negligent in checking the creditworthiness of households that applied for loans. When housing prices started to fall, and more households were unable to pay their mortgage, banks started to make losses to the point that their continuity was in jeopardy. On top of that, banks build portfolios with risky long-term maturity assets that were financed with short-long-term maturity liabilities, while using a high leverage financing structure (Kerkemeyer, 2019). This structure did not allow banks to convert their assets in cash quick enough to respond to the market turmoil caused by the crisis (Poole, 2010). This means that a lack of proper liquidity management prevented banks from absorbing the shocks that arose from the financial crisis (Basel Committee on Banking Supervision, 2008).
Besides the financial position, the role of banks in the crisis also damaged their image. Since the crisis, the banking industry is faced with distrust from the public and investors (Stevenson and Wolfers, 2011). A lack of trust can severely damage the functioning of a market, especially in the case of financial markets (Guiso, Sapienza and Zingales, 2008), as distrust can trigger bank runs and turbulent stock markets (Knel and Stix, 2015). Restoring trust in the financial industry is therefore vital for the functioning of the global financial system. (Financial Stability Board, 2012). Standard setters and regulators have identified transparency as one of the main key areas to focus on to accomplish this. That is because banks are generally seen as opaque organizations, which, in combination with their role in the crisis, provides an image of shady business. The perceived opaqueness of a bank is therefore a determining factor for the amount of trust in a bank (Jansen, Mosch and Van der Cruijsen, 2015).
In their attempt to increase transparency, the banking industry should first focus on risk reporting. Given that the public was unaware of the increasing amount of excessive risks that banks were taking before the crisis, this particular area will be scrutinized by the public and investors. The importance of risk reporting is highlighted by the introduction of new regulations and standards like Basel III, which requires an increase in capital buffers and increased transparency on risk exposure. However, the current regulations, including Basel III, are mostly principle based and leave banks with a certain amount of freedom. Banks therefore have the ability to determine their own level of risk disclosure quality to an extent, creating differences in risk disclosure quality between banks (Barakat and Hussainey, 2013).
The room for flexibility and the variation in quality is why recent studies have focused a lot on the determinants of risk disclosure quality (Khlif and Hussainey, 2016). So far, less attention has been given to the consequences of risk reporting. Recent studies have therefore suggested that a closer look at the economic consequences of risk reporting quality would be helpful in further
3
understanding the importance of transparency for banks (Elshandidy et al., 2015; Khlif and Hussainey, 2016). In this paper I will try to contribute to the literature by investigating these economic consequences. The results of this research will help bank managers understand the benefits or drawbacks of risk reporting. Standard setters can use the increased understanding of management incentives for risk reporting when developing new standards and regulations.
Using legitimacy theory, banks are expected to use risk reporting as a tool to regain society’s trust (Barakat and Hussainey, 2013). Major public scandals or impactful events, like the financial crisis, are proven to increase reporting behavior of organizations that were involved because organizations want to convey that they adhere to generally accepted social values (Dube and Maroun, 2017). This aligns with the argument of Gray, Bebbington and Walters (1995) that organizations that do not meet society’s expectations have economic incentives to increase disclosure. One of these incentives is a decrease in the cost of capital (Fiechter and Zhou, 2016). That is because firms with higher levels of voluntary disclosure have lower cost of equity capital (Francis, LaFond, Olsson and Schipper, 2005).
Applying the logic of agency theory provides the assumption that bank managers (agent) have more knowledge of the organization than investors (principal), resulting in information asymmetry between them. Bank managers can take advantage of this and purse their own interests rather than that of the investors (Johnson and Mamun, 2012. The risk-taking activities of banks prior to the financial crisis is a prime example of this. It is in the interest of investors to decrease the information asymmetry so they can better control what bank managers are doing (Nier and Baumann, 2006). Investors are therefore expected to reward banks that provide higher levels of transparency.
Stakeholder theory suggests that banks should take to the desires from the public and investors into account when choosing a reporting strategy (Kaur and Lodhia, 2018). Failing to do so could ultimately lead to an increase in the cost of equity. In order to maintain a good relationship with both stakeholders, banks are expected to use voluntary reporting to enhance their communication in an attempt to meet the their expectations (Barakat and Hussainey, 2013). With better access to timely, comparable information on risks, investors are able to better assess risks and protect their interests (Silva, Chavez and Lopez-Lubian, 2013). Knowing what risks they are facing, investors will be able to lower their required risk premium, which should result in a lower cost of equity for banks (Easley and O’hara, 2004).
Legitimacy theory, agency theory and stakeholder theory all three suggest that increasing risk disclosure quality will result in a lower cost of equity for banks. Thus, increasing risk reporting quality is expected to have beneficial economic consequences for banks. Either by an increase in society’s trust in a firm, less information asymmetry or a better relationship with stakeholders.
The economic consequences of risk reporting will be determined based on the impact of disclosure quality on the cost of equity. Disclosure quality is measured through a disclosure index and disclosure ambiguity. The disclosure index is based on standards from Basel III, the Enhanced Disclosure Task Force (EDTF), the European Securities and Market Authority (ESMA) and other similar bodies. The content analysis will be conducted on the annual reports of a sample of European banks that were included in the European Banking Association (EBA) Transparency Exercise. Similar data has been collected for the annual reports from 2015, 2016 and 2017 by student in previous years, using the same disclosure index. This data will also be available to use in my research. Disclosure ambiguity is measured as the standard deviation of analyst forecasts. Cost of equity is measured using the modified price-earnings growth (PEG) model as introduced by Easton (2004). The year of the report, the country in which the bank is located, bank size and bank profitability will be used as control variables.
This rest of this paper is structured as follows. Section two contains an explanation of the theoretical framework used to develop the hypotheses. Section three provides an overview of the methodology and operationalization of the variables. Section four presents the results of this research. Section five concludes with additional analysis and suggestions for further research.
4
2. THEORY
Scientific contribution
The global financial crisis that started in 2007 was for a large part caused by the excessive amounts of risks that banks were taking. Especially in the US, banks were under the assumption that housing prices would continue to rise (Hunt, 2011). In such a setting, even households that were not creditworthy would receive credit as long as they could offer real estate as collateral. When prices stopped to rise, subprime borrowers were no longer able to fulfill their obligations. As a result, banks that were involved in subprime mortgages got in serious financial trouble or even collapsed (Kerkemeyer, 2019). This ultimately led to the collapse of big banks like Lehman Brothers (Johnson and Mamun, 2012). After the collapse of Lehman Brothers, the crisis spread quickly to other developed economies in Europe and Asia (Sinhania and Anchilia, 2013; Kerkemeyer, 2019). In the wake of the crisis, regulators and bank supervisors concluded that operational risks should be controlled better. As society was unaware of risks that banks were taking, regulators and standard setter decided that bank risk reporting should be one of the main mechanisms to prevent future financial crises (Financial Stability Board, 2012). With the introduction of Basel III came more detailed standards for the capital levels of banks to increase protection against operational risks and an emphasis on the importance of risk reporting (Basel Committee on Banking Supervision, 2013). However, the standards from Basel III are principle based and leave banks with room to determine their level of risk disclosure to a certain extent (Barakat & Hussainey, 2013). This leads to differences in reporting quality between banks.
So far, research has focused a lot on the determinants of risk disclosure quality like firm size, firm performance, governance, regulation, supervisions, country characteristics and culture characteristics (Barakat and Hussainey, 2013; Elshandidy et al. 2015; Elbannan and Elbannan, 2016). Research on the consequences of risk reporting quality is however still scarce. This research will contribute to the literature by increasing the understanding of the economic consequences of risk reporting. Gaining insights in the economic consequences could help both banks and standard setters in their efforts to restore trust in the banking industry. Bank managers will be able to use the results in decision-making processes when determining their risk disclosure level. Standard setters and regulators on the other hand will be able to use the results in their practice of developing new standards and regulations. Insights in the economic consequences of risk reporting helps them to understand the management incentives for risk reporting, which is important in developing new regulations and standards (Miihkinen, 2012). By investigating the economic consequences of risk reporting, this research answers to the call from previous research (Barakat and Hussainey, 2013; Khlif and Hussainey, 2016; Elshandidy, Shrives, Bamber and Abhraham, 2018).
Banks and regulation
Banks operate in a heavily regulated industry where they have to comply with numerous detailed regulations and standards. That means that during the crisis, not only did banking institutions fail, the legal framework surrounding the banking industry also failed (Kerkemeyer, 2019). This means that the legal framework needed to be revised. The regulations in the banking industry are primarily designed by the Basel Committee on Banking Supervision (BCBS), a global committee of central banks and supervisory bank authorities. Since its installation, the BCBS introduced a series of accords and principles with the aim to design proper capital adequacy rules for banks. As a standard setter, their publications are not enforceable by law (BCBS, 2018). However, most standards are adopted by governments and supervisory authorities to enhance bank supervision, either by implementing new laws or incorporating them in new regulations (Pugliese, 2014).
Prior to the crisis the BCBS had introduced Basel I in 1988 and Basel II in 2004 to develop rules that ensure the capital adequacy of banks. With the introduction of Basel II came a new capital
5
adequacy framework that is comprised of three pillars: the Pillar 1 minimum capital requirements, the Pillar 2 supervisory review process and the Pillar 3 market discipline (BCBS, 2004). Pillar 1 required banks to maintain higher levels of capital buffers, with specific measurements for credit risk, market risk and operational risks. Pillar 2 increased the involvement of supervisory authorities in banks, requiring continuous assessment of risks and capital adequacy, and communication between banks and supervisors. Pillar 3 focusses on increasing transparency to allow the market to better assess risks and to increase comparability.
The BCBS also published a document called Principles for Sound Liquidity Risk Management and Supervision (2008) to provide banks with a way to develop a proper liquidity management framework and to realign bank risk tolerance with actual risk taking (BCBS, 2008). This was deemed necessary because a lack of proper liquidity management was one of the main reasons that banks were unable to respond to the market turmoil that was created by the financial crisis. This document was first response to the chaos of the financial crisis by the BCBS.
To further increase the resilience of the banking industry, the BCBS introduced Basel III in 2013. Building on the existing capital adequacy framework from Basel II, Basel III introduced new metrics for bank liquidity and funding in the Liquidity Coverage Ratio (LCR) and the Net Stable Funding Ratio (NSFR). The goal of the LCR is to increase short-term liquidity and the goal of the NSFR is to increase long-term stability (BCBS, 2013). Both ratios try to achieve their goal by increasing the loss-absorbing capital. That means banks need to replace a portion of their debt financing with equity financing. Banks, like most business organizations, usually prefer to use debt financing because the interest is tax-deductible. This benefit makes debt financing cheaper than equity financing. Increasing equity financing relative to debt financing will thus result in higher financing costs (Miles, Yang and Marcheggiano, 2013). The introduction of Basel III standards in Europe is therefore expected to lead to an increase of the cost of capital for banks.
Disclosure freedom
In Europe, Basel III has been implemented by means of the Capital Requirements Regulation (CRR) and the Capital Requirements Directive (CRD) as of January 2014. Because the European banks and authorities have a rather large involvement in developing the Basel standards, the Basel III standards are closely reflected in the European legislation after the adoption. Although Basel III is very strict on adhering to liquidity and funding ratios to reduce capital adequacy risks, banks are left with a certain amount of freedom in determining their level of disclosure on those risks. The Pillar 3 standards of Basel III are specifically designed to allow banks to interpret reporting requirements in a way that is aligned with their internal risk management (BCBS, 2009). With this approach, banks can interpret the standards in a way that benefits them. This gives banks flexibility in determining to what extent and how they want to report on their risks.
However, banks still need to adhere to the International Financial Reporting Standards (IFRS) when preparing their annual report. IFRS 7 requires all entities to disclose information in their financial statements regarding the “significance of financial instruments for the entity’s financial position and performance” (IFRS, 2005). In doing so, entities also need to disclose an overview of the risks that that are created by their financial instruments. Since almost all bank products are financial instruments, banks are required to disclose quantitative and qualitative disclosures on pretty much all their individual activities and the risks that result from them. But like the Basel III standards, the requirements of IFRS 7 are principle based. They leave room for interpretation which gives banks flexibility in their risk reporting.
Given the lack of detailed enforceable regulations, banks are able to determine their own level of risk disclosure to a certain degree. The quality and comparability of risk disclosure is therefore more dependent on a banks willingness to provide that information rather than the ability of banks to gather and report the information. This flexibility allows a lot of divergence between banks regarding risk disclosure quality.
6 Market Discipline
As the regulations are not airtight, the Basel framework looks to the market to help deter banks from taking unnecessary risks. By requiring banks to disclose risk information in Pillar 3, the capital adequacy framework aims to facilitate market discipline. Market discipline is a collection of actions that stakeholders like investors can take that could impact the decision-making processes within banks (Flannery, 2001). It provides stakeholders with tools to reduce the opportunity for banks to take excessive risks and is essential in managing bank risks (Nier and Baumann, 2006). The fact that market discipline is incorporated in the Basel framework highlights its importance.
The main way in which investors are able to discipline banks is by requiring a higher interest rate on loans or a higher rate of return on equity when the perceived risks of banking activities are high. However, in order to accurately judge how much risk a bank is taking, investors need information (De Araujo and Leyshon, 2017). Risk disclosure therefore plays an important role in facilitating market discipline in the banking industry. Because of the agency problem, there is an information asymmetry between banks and investors. Bridging the information gap requires banks to be more transparent and provide more information to investors. While reducing the information asymmetry, transparency also increases investors ability to properly assess a banks risks. Investors are therefore expected to require a lower risk premium on their investments (Easley and O’hara, 2004; Elbannan and Elbannan, 2015).
Conceptual theories
Marston and Shrives (1991) divide information in annual reports into required disclosure and voluntary disclosure. Required disclosure follows from laws and regulations whereas voluntary disclosure occurs when organizations have the impression that providing more information than required will benefit them.
As the standard from Basel III and the IFRS are incorporated in most European laws, bank risk reporting is required. However, as mentioned earlier, the standards from Basel III and the IFRS are principle based and leave banks with flexibility in their reporting to a certain extent. The amount of freedom that banks have in their risk disclosure practices means that the quality of the disclosed information is more dependent on a banks willingness to provide information, rather than the ability to gather the information. The amount of risk disclosure that a bank provides is therefore dependent on the perceived benefits of a certain level of risk disclosure. This makes bank risk disclosure similar to voluntary disclosure and allows the use of legitimacy theory, agency theory and stakeholder theory to explain bank reporting behavior.
Legitimacy theory
Legitimacy theory (Dowling and Pfeffer, 1975) tries to explain why organizations provide voluntary reporting and relies on the concept of legitimacy. In this context, legitimacy refers to the public perception that an organization acts in line with generally accepted value systems (Suchman, 1995). If the actions are perceived to be in line with these value systems, an organization gains trust from the public regarding its ability to contribute to society and its likelihood of continuity (Higgins and Walker, 2012). Increasing transparency through more disclosure helps to convey that massage. An organization with high quality disclosure signals that it is capable of handling current, but also future, standards and regulations (Fonseka, Rajapakse and Tian, 2019).
At the start of the crisis, banks got in trouble because of their excessive risk taking, something that was not in line with generally accepted value systems. They lost their legitimacy and customers were afraid to lose their savings because of a collapse. This triggered several bank runs across Europe (Northern Rock, Landsbanki, DSB Bank) and the US (Countrywide Financial, IndyMac, Bear Stearns), demonstrating the importance of legitimacy in the banking industry.
In the past, such major events have been proven to positively affect organizational reporting behavior because organizations feel the need to restore their legitimacy (Patten, 2002; Dube and Maroun, 2017). Banks are therefore expected to be driven to restore their organizational legitimacy
7
by using more voluntary reporting (Barakat and Hussainey, 2013). This enhances legitimacy because it assures stakeholders that market discipline is effective by fulfilling institutional pressures (Oliveira, Rodrigues and Craig, 2011).
The value of trust should not be neglected. Gray et al. (2015) mentions that organizations that are not perceived to act in line with social values have economic incentives to increase disclosure quality. That is because firms with higher levels of voluntary reporting are found to have a lower cost of equity as a result of increased trustworthiness (Francis et al., 2005) and decreases the total cost of capital (Fiechter and Zhoud, 2016). The expectation is therefore that higher quality risk disclosure will result in a lower cost of equity for banks.
Agency theory
Agency theory (Jensen and Meckling, 1976) focusses on the problems that arise from the separation of ownership and control. This problem emerges when the agent is hired by the principal to act on his (their) behalf. In these situations the agent receives decision-making authority from the principal to be able to effectively manage the organization (banks). The agent can use his granted power to deter from what is best for the principal and instead focus on decisions that maximize his own wealth. For example, the agent could alter accounting numbers to maximize his bonus (Jensen and Meckling, 1976). This is possible because the separation of ownership and control creates information asymmetry between the principal and the agent. The principal can therefore not properly control the actions of the agent.
Translated to the banking industry, shareholders are the principal and bank managers are the agent. The information asymmetry between shareholders and bank managers allows the bank mangers to take decisions that are not in the best interest of the shareholders. In the financial crisis it became clear that banks had been taking unnecessary and undesirable amounts of risks. They took advantage of the existing information asymmetry in pursuit of earning big bonuses (Johnson and Mamun, 2012). At the same time, they were not bearing the downsides of the risks. In case of loses, shareholders and investors would be the ones that were heavily affected. The risks that bank managers were taking were therefore not in line with the interests of investors. By reducing information asymmetry and increasing transparency, this misalignment of desired risk and taken risk should be better (Nier and Baumann, 2006). Shareholders and investors are then in a better position to castigate bank managers when taking excessive and undesirable amounts of risk.
Stakeholder theory
Stakeholder theory (Freeman, 1984) focusses more on the relationship between organizations and their stakeholders. Freeman (1984) defines a stakeholder as anyone that can affect or is affected by the organizations actions. His definition indicates that the relationship between organizations and their stakeholders works two-ways. Stakeholders are affected by the actions that organizations take in pursuing their goals and stakeholders can influence the goals of the organization (Goa and Zhang, 2001). This premise suggests that organizations can benefit from the participation of stakeholders (Kaur and Lodhia, 2018).
Translated to the banking industry, we can define the public and investors as the biggest stakeholders. Both are heavily affected by a banks actions. How the public is affected by banks was demonstrated in the financial crisis. A failure in the financial industry can quickly snowball out of control and affect a lot of people. These people can affect banks by either depositing their savings or withdrawing them. Incorporating the needs of the public in their goals is therefore a legitimate strategy. Investors on the other hand can affect a banks actions by determining the amount of return that they require. If investors require a lower rate of return, the cost of equity will be lower for banks. A lower cost of equity may sway banks to use more equity financing instead of debt financing, which is a safer alternative (Miles et al., 2013). Since the crisis, the general call from the public and investors towards banks is for more information and transparency. Banks can respond to these request by improving their risk disclosure. Thus, based on stakeholder theory,
8
banks are assumed to use risk disclosure as a way to better communicate with the public and investors (Barakat and Hussainey, 2013).
Hypotheses
The introduction of Basel III in the regulations of the European banking market requires banks to make safer financing decision. Banks have to maintain higher levels of equity capital to meet the LCR and NSFR ratios. This means that banks need to lower their leverage and replace some debt capital with equity capital. Because equity financing is more expensive than debt financing, banks are afraid for an increase in their cost of capital (Miles et al., 2013), which will make financing more expensive and lowers their ability to generate returns (Toader, 2015).
However, there are ways for banks to limit the increase in financing costs. The reason financing costs would rise is because the cost of equity is higher than the cost of debt. If banks are able to lower their cost of equity, the overall impact of more equity financing is weakened. Increasing transparency can help banks in this regard. As banks disclose more detailed information, information asymmetry between banks and investors will be reduced (Diamond and Verrecchia, 1991). This allows investors to better asses risks, providing them with the ability to better protect themselves. Because of that, investors will require a lower risk premium on their investments, reducing the cost of equity for banks (Easley and O’hara, 2004; Elbannan and Elbannan, 2015).
This interaction between banks and investors is an example of market discipline, which is one of the main areas that Basel III focusses on. Because their standards on risk disclosure are principle based, the Basel III framework needs help from the market to increase the resilience of the banking sector. The idea is that large stakeholders, like investors, can punish banks for undesirable behavior and create benefits for desirable behavior (Flannery, 2001). Investors can do this by requiring a higher or a lower risk premium to impact the financing costs of bank. A bank that is taking undesirable risks can therefore expect a higher financing cost than a bank whose risks are perceived to be less undesirable (Nier and Baumann, 2006).
To be able to properly determine if a banks risks are desirable or not, investors need information (De Araujo and Leyshon, 2017). Thus, the information asymmetry between banks and investors needs to be reduced. However, because the standards from Basel III and the IFRS are principle based, rather than strict regulations, banks are left with some flexibility when it comes to risk disclosure. Risk disclosure is therefore not uniform and there are significant difference in reporting quality between banks (Barakat and Hussainey, 2013).
If banks increase their disclosure quality, investors have better access to information that they can use to assess and compare risks of different banks. This allows investors to determine the perceived riskiness and impact of an investment (Silva et al., 2013). Through the mechanisms of market discipline, investors are then able to adjust the required rate of return. Investors are expected to require a lower rate of return from banks with better risk disclosure quality, as the risks are better identifiable (Akhigbe, McNulty and Stevenson, 2013). Therefore, the expectation is that increasing risk disclosure quality will result in a lower cost of equity for banks. This leads me to the following hypothesis:
H1. There is a negative relation between risk reporting quality and the cost of equity.
However, every user group of annual reports and the disclosed information in them has its own needs and focusses on different things (Marston and Shrives, 1991). Investors might value certain disclosure items different than other user groups. The disclosure index that I use contains disclosure items centered around four themes: general risk disclosure, credit risk disclosure, solvability risk disclosure and liquidity risk disclosure. It would be interesting to see if one of these disclosure types has a bigger influence on the cost of equity than others. Therefore I will also test the influence of each individual risk disclosure type that is included in my disclosure index on the cost of equity. The expectation for the relation between the types of disclosure and the cost of equity
9
remain the same as that of H1. Each individual disclosure item is therefore expected to decrease the cost of equity. This line of thought provides the following hypotheses:
H2a. There is a negative relation between general risk disclosure and the cost of equity. H2b. There is a negative relation between credit risk disclosure and the cost of equity. H2c. There is a negative relation between solvability risk disclosure and the cost of equity. H2d. There is a negative relation between liquidity risk disclosure and the cost of equity.
Providing useful information to investors is one step towards decreasing information asymmetry. However, more information does not necessarily mean that the information is unambiguous. Investors place numbers in their own context when creating an understanding of what the information says. Investors can therefore have different interpretations of the same numbers. Ambiguity is a source of interference on the communication between banks and their investors, lowering their ability to comprehend the provided information (Ertugrul, Lei, Qiu and Wand, 2017). Providing information is therefore not enough. The disclosures should provide context in order for investors to accurately judge the information. Decreasing ambiguity can thus have consequences in the form of valuation volatility (Loughran and McDonald, 2013). With less ambiguity, investors should be able to better judge the risks of a bank. Just like with risk disclosure, investors are expected to reward banks for this by decreasing the cost of equity. Based on this expectation, I developed the following hypothesis:
H3. There is a positive relation between disclosure ambiguity and the cost of equity.
3. METHOD
Data collection
Disclosure data has been collected from the annual reports of 60 different banks in the European Union and European Economic Area for the year 2018. The sample of 60 was pulled from a total of 131 European banks from 25 countries that are included in the Transparency Exercise from the EBA. The EBA selects banks for the Transparency Exercise based on their relevance and importance for the functioning of the European banking sector. Using banks that participate in the Transparency Exercise ensures that the included banks are of significant size and importance.
The 60 banks were divided among me and three other students, each barring the responsibility to review annual reports from 15 different banks. Because of time constraints, including more than 60 banks would not benefit the quality of reviews. The data that we gathered was added to an existing dataset with the same data for the years 2015, 2016 and 2017. Therefore the final dataset contained results from 215 different annual reports.
For the purposes of this research I was only able to use the data for publically listed firms as my measure for the cost of equity includes share price data. Besides, the analyst forecasts in the International Broker’s Estimate System (I/B/E/S), that I used in determining the cost of equity, are only provided for listed banks. This meant that I could only include 145 results from 40 banks and 15 countries in my analysis. However, the forecasts for 14 reports did not allow me to calculate the cost of equity properly because the estimated growth was negative. Therefore my final dataset consisted of 129 reports from 38 banks and from 15 countries, spread across 4 years. The distribution of the number of reports per year and country can be found in table 1.
Cost of equity
Cost of equity is often used in financial decision-making. Banks use it for investment decisions and raising capital, while investors use it to determine the value of securities when building a balanced portfolio (King, 2009). The cost of equity represents the required return on investments from investors and is based on the risks that investors face while providing additional equity.
10
When studying cost of equity, there are three possible valuation models. These models are the residual income valuation (RIV) model (Gebhardt, Lee and Swaminathan, 2001), the abnormal earnings growth (AEG) model (Gode and Mohanram, 2003), and the modified price-earnings growth (PEG) model (Easton, 2004). For the purpose of studying the relationship between the level of disclosure and the cost of equity, the PEG model is preferential method according to Mangena, Li and Tauringana (2016). They argue that the PEG model is the most suitable model to evaluate the relative variations in estimates of the cost of equity and the required data is easier to obtain. The PEG model from Easton (2004) uses the 1-year-ahead analyst’ earnings forecast (𝑒𝑝𝑠1), the 2-year-ahead analysts’ forecast (𝑒𝑝𝑠2) and the current share price (𝑃0) to estimate the cost of equity (COE). This results in the following model:
𝐶𝑂𝐸 = √𝑒𝑝𝑠2− 𝑒𝑝𝑠1 𝑃0
For the model it is important that 𝑒𝑝𝑠1 and 𝑒𝑝𝑠2 are both positive and that 𝑒𝑝𝑠2 is greater than 𝑒𝑝𝑠1 (Easton, 2004). The analysts forecasts were collected from the I/B/E/S and the share price was gathered using the historic share price from Yahoo Finance. To ensure that the information in the annual reports is taken into account, I only used analyst forecasts that were made after the release of the annual reports. Since all banks had publish their annual report before the end of the first half of march, I used the mean 𝑒𝑝𝑠1 and 𝑒𝑝𝑠2 from the second half of march. For the share price I used the historic closing share price from the day of the publication of the analyst forecasts.
Risk disclosure quality
There is no universally accepted measure of quality when it comes to disclosures. Measuring disclosure quality is therefore not straightforward and requires some special attention. While it has certain limitations, using a disclosure index to proxy disclosure quality is the most popular method. So far, no other method has been developed that can do a better job (Marston and Shrives, 1991).
TABLE 1 Distribution of reports
This table provides an overview of the distribution of the number of reports that were included in this research.
Country Short 2015 2016 2017 2018 Total Percentage
Austria AUS 2 2 2 2 8 6,2% Belgium BEL 1 1 0 1 3 2,3% Denmark DEN 2 2 2 2 8 6,2% France FRA 2 3 3 3 11 8,5% Germany GER 2 2 2 2 8 6,2% Greece GRE 0 0 0 1 1 0,8% Hungary HUN 1 1 1 1 4 3,1% Ireland IRE 1 0 1 1 3 2,3% Italy ITA 5 4 3 5 17 13,2% Netherlands NED 2 1 2 2 7 5,4% Norway NOR 1 1 1 1 4 3,1% Poland POL 1 1 1 1 4 3,1% Spain SPA 4 4 4 5 17 13.2% Sweden SWE 4 4 4 4 16 12,4% United Kingdom UK 5 3 5 5 18 14,0% Total 33 29 31 36 129 Percentage (%) 25,6% 22,5% 24,0% 27,9%
11
However, it is important to note that, because of the abstractness of measuring quality, the disclosure index essentially measure the quantity of reporting instead. The disclosure index is thus a testing device with the aim of measuring the underlying value of quality. Therefore, to be able to use the disclosure index, a lot of researchers make the assumption that quantity and quality are positively related (Beattie, McInnes and Fearnley, 2004). I make the same assumption and use a disclosure index to proxy for risk disclosure quality.
Risk disclosure index
The articles of Marston and Shrives (1991) and Beattie et al. (2004) were used to identify pitfalls and best practices in the development of the disclosure index. Marston and Shrives (1991) note that the number of items that have the potential to be disclosed is close to infinite. Therefore, the items that are included in the disclosure index are crucial for the usability of the index. The disclosure index that I used consists of 30 items and is previously used by students in 2015, 2016 and 2017. All included items are based on the current guidelines and regulations from the standard setting and governing bodies, like the BSBC, the ESMA, the EDTF and others, with the aim to approximate the general idea of high quality risk disclosure.
An ordinal scale was used to measure the index score for each individual item. A single item could thus receive a score of 0, 1 or 2 points, based on the required information being not included (0), limited (1) or comprehensive and clear (2). An ordinal scale was chosen over a nominal scale to allow the assessment of the quality of the specific disclosure item to be assessed, rather than its mere existence (Beattie et al., 2004). Because not all disclosure items are equivalents of each other in terms of weight, it is not possible to gather data on an interval scale with this disclosure index. Despite this, parametric statistical tests are still useable when handling disclosure index data according to Marston and Shrives (1991).
As the disclosure index contains 30 items that can all score 2 points, the maximum is 60. The 30 items are also categorized in four different groups. The first group focusses on general risk disclosure (GRD) and consists of 7 items. The second group focusses on credit risk disclosure (CRD) and consists of 8 items. The third group focusses on solvability risk disclosure (SRD) and consists of 10 items. The fourth group focusses on liquidity risk reporting (LRD) and consists of 5 items. To confirm that the four types of risk disclosure show sufficient correlation to be grouped as one total score of risk disclosure quality, I calculated the Cronbach’s alpha. As shown in table 2, the Cronbach’s alpha for the four types of risk disclosure is 0,700. This means that the internal consistency between the scores is acceptable for grouping them together.
Marston and Shrives (1991) raise the question of how to weight the items in a disclosure index. They note that it is important to identify the user group of the disclosure in order to be able to determine a certain weighting system. This way, it is possible to assign more weight to items that
TABLE 2 Cronbach’s alpha
This table contains a reliability analysis of the four types of risk disclosure. The table shows the value of Cronbach’s alpha for
GRD, CRD, SRD and LRD. The table also shows what impact
removing one of the types of disclosures from the analysis.
Reliability statistic Cronbach’s alpha
GRD, CRD, SRD and LRD 0,700
Without GRD 0,658
Without CRD 0,606
Without SRD 0,536
12
are perceived to be more important to the user group. However, Marston and Shrives (1991) also note that is hard to determine what the user group will see as important. Besides, they provide examples of numerous examples of research that found no difference in outcomes for weighted and unweighted scores. Spero (1979) explains this phenomenon as it is likely that organizations that are better at reporting bigger or more important items are also better at disclosing smaller or less important items. In my research I found no difference in weighted or unweighted results. Therefore I will only use unweighted disclosure scores in my analysis.
Intra-coder reliability
In order to preserve scientific integrity, we had to ensure that all student that collected data would produce the same results when coding the same content (Beattie et al., 2004). Because the disclosure index was applied by other students to annual reports from 2015, 2016 and 2017, we had to mimic their process as much as possible to be able to compare our results. Before gathering the data, we discussed these issues and came up with u number of measures.
First, every student had to perform two practices reviews which were discussed altogether. We made sure to eliminate all ambiguity about the disclosure index and that our interpretation was in line with the intended measurement. The goal was to create a shared understanding of the disclosure index items. Most importantly, understanding what the item was supposed to measure and when information disclosure was sufficient to award 1 or 2 points. Because our supervisor also guided the students that collected the data for previous years and designed the disclosure index that we were using, he could ensure that our application of the disclosure index for 2018 was as similar to that of previous students for 2015, 2016 and 2017 as possible.
Second, after the data collection process, we reviewed each other’s results. Every student performed a total of 10 reviews, whereby at least 2 result from every student were included. In the reviewing process, all items were a slight or major disagreement showed were marked and later discussed with the original reviewer. The original reviewer then got the chance to change his results, depending on what both students agreed upon. This process was meant to further increase our consensus on the application of disclosure index.
Finally, all results were sent to our supervisor, who reviewed a number of results per student. In this process, he was able to look for possible discrepancies or unusual outcomes. This was a last check to assure that there were no big differences between this year’s results compared to previous results.
Disclosure ambiguity
Ambiguity, like risk disclosure, is not a straightforward concept and can be measured in different ways. Once again, there is no universal measurement for ambiguity. Entugrul et al. (2017), for example, use file size as a proxy for ambiguity and Penno (2008) analyses the vagueness of specific words. I will follow the example of Akhigbe et al. (2013) and use analyst forecasts dispersion (AFD) to proxy ambiguity. Akhigbe et al. (2013) formulate that a smaller dispersion between analyst forecasts represents more transparency, meaning that the provided information should be clearer. The amount of forecast dispersion between analysts therefore resembles the amount of ambiguity. For my research, I used the standard deviation of the analysts 𝑒𝑝𝑠1 and 𝑒𝑝𝑠2 forecasts that I obtained from I/B/E/S.
Control variables
Control variables that will be included are the year of the annual report, bank country, bank size and bank. The year of the annual report (YEAR) is included to prevent market wide developments from impacting the results.
Bank country (CTRY) will be expected to influence the cost of equity as there are strong differences in the market premium and risk reporting practices between countries (King, 2009; Elshandidy et al., 2015). For example, in studying the impact of the mandatory adoption of IFRS
13
in 2005, Li (2010) found strong differences in the impact of the adoption on the cost of equity for different countries. Countries with higher enforcement mechanisms experience a bigger impact on the cost of equity than other countries.
Larger firms usually have a lower cost of equity (Alberts and Archer, 1973). Bank size (SIZE) is therefore expected to be negatively associated with the cost of equity and will be measured as the log of total assets (Francis, Khurana and Pereira, 2005).
Bank profitability (PRFT) is also expected to affect the cost of equity as investors take the anticipated profitability into account when determining the cost of equity (Toader, 2015). As an organization with high profitability should be more attractive for investors, the relationship between bank profitability and is expected to be negative (Elshandidy et al., 2015). Bank profitability will be measured as net income divided by equity.
Logistic regression models
For the analysis of the data I used ordinary least squares regression models. To test the relation between risk disclosure quality and the cost of equity, I used two similar, but slightly different, models with independent variables and control variables. The control variables are the same for both models. The first model tests the effect of risk disclosure and disclosure ambiguity on the cost of equity. The first regression model looks like this:
𝐶𝑂𝐸 = 𝛽0+ 𝛽1𝑇𝑅𝐷 + 𝛽2𝐷𝐴𝑀 + 𝛽3𝑌𝐸𝐴𝑅 + 𝛽4𝐶𝑇𝑅𝑌 + 𝛽5𝑆𝐼𝑍𝐸 + 𝛽6𝑃𝑅𝐹𝑇
The second model categorizes risk disclosure quality into four categories of risk disclosure to see if a particular type of risk disclosure is more important for investors. These categories are general risk disclosure, credit risk disclosure, solvability risk disclosure and liquidity risk disclosure. Therefore, the second regression model is as follows:
𝐶𝑂𝐸 = 𝛽0+ 𝛽1𝐺𝑅𝐷 + 𝛽2𝐶𝑅𝐷 + 𝛽3𝑆𝑅𝐷 + 𝛽4𝐿𝑅𝐷 + 𝛽5𝐷𝐴𝑀 + 𝛽6𝑌𝐸𝐴𝑅 + 𝛽7𝐶𝑇𝑅𝑌 + 𝛽8𝑆𝐼𝑍𝐸 + 𝛽9𝑃𝑅𝐹𝑇
4. RESULTS
Judging from the results of the content analysis, the risk disclosure quality has been improving steadily over the years. As can be seen in table 4, median risk disclosure quality has risen from 24
TABLE 3 Variables
This table gives an overview of the variables I use in this paper and how they are operationalized and obtained.
Variable Description Clarification
COE Cost of equity (𝑒𝑝𝑠2 - 𝑒𝑝𝑠1) / 𝑃0*, obtained from I/B/E/S and Yahoo Finance
TRD GRD CRD SRD LRD DAM YEAR CTRY SIZE PRFT
Total risk disclosure General risk disclosure Credit risk disclosure Solvability risk disclosure Liquidity risk disclosure Disclosure ambiguity Year of annual report Bank country Total assets Bank profitability
Score from 0 to 60, obtained from content analysis Score from 0 to 14, obtained from content analysis Score from 0 to 16, obtained from content analysis Score from 0 to 20, obtained from content analysis Score from 0 to 10, obtained from content analysis
Standard deviation of analyst forecast, obtained from I/B/E/S The year with which the annual report is concerned, dummy variable The country from which the bank is operated, dummy variable Log of total assets of a bank at year end, obtained from annual reports Net income / equity, obtained from BankFocus
*𝑒𝑝𝑠1 represents the 1-year-ahead analysts forecast, 𝑒𝑝𝑠2 represents the 2-year-ahead analyst forecast and 𝑃0 represents the current share price
14
in 2015 to 32 in 2017 and 2018. Both the mean and the median improved for each individual type of risk disclosure in the same time. This can either be a result from improving standards and regulations or an improved perception of the importance of risk reporting. As for the country specific scores, banks from the United Kingdom, the Netherlands and Austria show the best risk disclosure quality, while banks from Hungary, Greece and Poland have the lowest risk disclosure quality. However, it should be noted that all three of the later countries were only represented by one bank through the years.
Table 5 presents descriptive statistics for all variables. With a mean of 10,38% the cost of equity for the average bank in this research is pretty normal. However, the difference between banks is pretty big. The cost of equity ranges from very low (0,67%) to very high (27,27%). This shows that the difference in the cost of equity between banks is pretty big.
The disclosure ambiguity of a report is measured as the standard deviation of analyst forecasts on the earnings per share for a bank. The average height of the earnings per share forecasts was 3,540. The average forecasts dispersion between analysts was 0,218. To put that number into context, that is an average difference between analysts of 6,16% on the average earnings per share forecasts. The report with the least amount of ambiguity only had an analyst dispersion of 0,010 (or 0,28% of the average earnings per share forecasts). The report with the highest amount of ambiguity had a standard deviation of 0,780 (or 22,03% of the average earnings per share forecast). The fact that the median (0,110) of analyst forecasts dispersion is almost half of the mean (0,218), indicates that the mean is skewed by some reports with a high score on ambiguity. The average bank showed less disclosure ambiguity than expected based on the mean.
In terms of size, the average total assets of banks was 687,7 million (687.674.000). The smallest included banks was OTP Bank Nyrt. from Hungary with total assets of 26,5 million (26.479.000). The largest included banks was HSBC Holdings plc from the United Kingdom with total assets of 3,3 billion (3.268.960.000).
TABLE 4 Disclosure index results
This table provides an overview of the median risk disclosure score per country and per year.
Group Mean Median TRD Mean Median GRD Mean Median CRD Mean Median SRD Mean Median LRD
YEAR 2015 24,0 24,0 6,7 6,0 6,7 7,0 7,3 8,0 4,1 4,0 2016 26,4 26,0 7,4 8,0 7,6 7,0 7,4 7,0 4,0 4,0 2017 31,0 32,0 7,4 7,0 9,8 9,0 8,9 9,0 4,9 5,0 2018 31,2 32,0 7,4 8,0 9,8 10,0 9,4 9,0 4,5 5,0 CTRY AUS 29,5 30,0 7,8 7,5 10,4 10,0 7,0 7,0 4,4 4,0 BEL 25,3 22,0 6,7 6,0 6,3 4,0 6,7 7,0 5,7 6,0 DEN 20,4 23,0 4,8 5,0 7,3 7,0 4,1 4,5 4,1 4,0 FRA 31,5 31,0 7,5 7,0 7,3 7,0 11,6 12,0 5,0 5,0 GER 31,1 31,0 7,8 8,5 8,3 7,0 9,8 9,5 5,4 6,0 GRE 16,0 16,0 2,0 2,0 12,0 12,0 2,0 2,0 0,0 0,0 HUN 12,0 13,5 3,0 3,0 3,8 4,0 3,0 3,0 2,3 2,5 IRE 27,0 26,0 7,0 7,0 8,0 7,0 8,0 8,0 4,0 4,0 ITA 27,2 26,0 7,8 7,0 8,6 8,0 8,2 8,0 2,6 3,0 NED 35,7 36,0 8,4 9,0 11,6 12,0 12,1 12,0 3,6 4,0 NOR 19,8 21,0 6,8 7,0 5,0 4,5 5,5 5,5 2,5 2,5 POL 17,5 19,0 3,8 3,5 6,0 6,5 4,5 5,5 3,3 3,0 SPA 27,1 26,0 6,5 6,0 8,7 9,0 7,5 7,0 4,4 5,0 SWE 28,3 28,5 7,6 7,5 7,4 7,0 8,0 8,5 5,2 5,0 UK 36,2 36,0 8,1 9,0 11,1 10,5 11,0 11,0 6,0 6,0
15
With an average of 7,50%, most banks were found to be profitable. From the 129 included reports, 111 were from banks that showed a profitability higher than 0%. The remaining 18 reports were from banks with a negative profitability.
The median of the reporting scores from banks was 28 where 60 would have been the maximum. The bank with the highest score even had a score of 43. The bank with the lowest score, however, only managed to receive a score of 7. The large range, along with a standard deviation of 7,8, shows that the current regulations and standards from Basel III and the IFRS indeed leave banks with some freedom to determine their level of risk disclosure. With stricter regulations and standards, the scores could have been much closer.
TABLE 5 Descriptive statistics
This table provides an overview of the obtained data. The table is split up in continuous variables, disclosure scores and disclosure scores for the binary control variables.
Continuous variables Mean Median Minimum Maximum St. Deviation
COE 10,38% 10,16% 0,67% 27,78% 5,23%
DAM 0,218 0,110 0,010 0,780 0,232
SIZE 11,617 11,645 10,423 12,514 0,475
PRFT 7,50% 7,44% -2,90% 19,09% 5,51%
Disclosure scores Mean Median Minimum Maximum St. Deviation
TRD 28,2 28,0 7 43 7,8
GRD 7,1 7,0 1 12 2,4
CRD 8,5 9,0 1 15 3,0
SRD 8,3 8,0 1 16 3,3
LRD 4,4 4,0 0 9 1,9
Dummy variables COE DAM Mean SIZE PRFT
YEAR 2015 11,7% 0,256 11,601 7,1% 2016 10,7% 0,235 11,623 5,7% 2017 9,4% 0,190 11,638 10,6% 2018 9,7% 0,191 11,611 6,7% CTRY AUS 10,0% 0,321 11,218 8,9% BEL 5,1% 0,363 11,413 14,6% DEN 10,5% 0,246 11,280 7,0% FRA 10,2% 0,415 12,208 9,6% GER 17,4% 0,263 11,930 1,9% GRE 21,0% 0,055 10,785 0,7% HUN 10,7% 0,227 10,537 13,5% IRE 10,5% 0,085 11,099 7,9% ITA 16,3% 0,192 11,493 5,0% NED 7,6% 0,114 11,785 10,4% NOR 2,9% 0,075 11,440 10,8% POL 10,% 0,045 10,835 6,9% SPA 11,8% 0,043 11,648 6,8% SWE 6,6% 0,048 11,626 10,1% UK 6,8% 0,472 12,080 5,1%
16 Regression results
The results of the logistics regression are presented in table 6. In my analysis I used six different models. The tables shows the results for all six. The first model provides the basis for the analysis and only tests the relation between the cost of equity and the control variables year, country, size and profitability. These variables are subsequently used in all five other models. In my analysis I used the year 2015 as the base year and Spain as the reference country. I used Spain as the reference country as it was one of the four countries with more than 15 available reports and the cost of equity was cost of equity was pretty average. The second and third model both test the relation between risk disclosure and the cost of equity. However, the second model uses total risk disclosure, while risk disclosure is split in the four types of risk disclosure quality for the third model. The fourth model tests the relation between disclosure ambiguity and the cost of equity. The fifth and sixth model include all variables. However, like before, the fifth model includes total risk disclosure and the sixth model includes the four types of risk disclosure instead.
From the results of model 1, we see that the cost of equity is significantly lower in 2017 and 2018 compared to previous years. The market wide development therefore seems to be that
TABLE 6 Regression results
This table contains the results from the regression for the first model where I test the relation between risk disclosure, analyst forecast dispersion and the cost of equity. ***,** and * represent statistical significance at respectively the 1 percent, 5 percent and 10 percent level.
Variable Model 1 Model 2 Model 3 Model 4 Model 5 Model 6
YEAR 2016 -0,088 -0,146 ** -0,134 * -0,070 -0,109 * -0,103 2017 -0,153 ** -0,377 *** -0,329 *** -0,093 -0,221 *** -0,223 *** 2018 -0,197 *** -0,383 *** -0,369 *** -0,137 ** -0,266 *** -0,265 *** CTRY AUS -0,122 -0,181 ** -0,173 ** -0,255 *** -0,281 *** -0,278 *** BEL -0,203 ** -0,223 * -0,233 *** -0,304 *** -0,307 *** -0,309 *** DEN -0,099 -0,026 -0,035 -0,193 *** -0,135 ** -0,141 ** FRA -0,009 -0,059 -0,062 -0,260 *** -0,268 *** -0,261 *** GER 0,276 *** 0,250 *** 0,245 *** 0,157 ** 0,151 ** 0,154 ** GRE 0,134 ** 0,198 *** 0,208 *** 0,139 *** 0,181 *** 0,180 *** HUN -0,107 -0,015 -0,012 -0,154 ** -0,088 -0,087 IRE -0,070 -0,081 -0,076 -0,074 -0,081 -0,078 ITA 0,259 *** 0,249 *** 0,279 *** 0,161 ** 0,165 ** 0,181 ** NED -0,159 ** -0,274 *** -0,251 *** -0,202 *** -0,274 *** -0,264 *** NOR -0,305 *** -0,252 *** -0,245 *** -0,314 *** -0,278 *** -0,271 *** POL -0,088 -0,023 -0,022 -0,069 -0,028 -0,029 SWE -0,323 *** -0,365 *** -0,372 *** -0,238 *** -0,356 *** -0,352 *** UK -0,274 *** -0,415 *** -0,418 *** -0,616 *** -0,675 *** -0,675 *** SIZE -0,200 * -0,280 *** -0,267 ** -0,136 -0,195 ** -0,192 ** PRFT -0,048 0,039 0,039 -0,028 0,027 0,030 TRD (H1) - 0,437 *** - - 0,289 *** - GRD (H2a) - - 0,108 - - 0,058 CRD (H2b) - - 0,144 - - 0,124 SRD (H2c) - - 0,165 - - 0,111 LRD (H2d) - - 0,175 ** - - 0,099 DAM (H3) - - - 0,504 *** 0,454 *** 0,452 *** Observations 128 128 128 128 128 128 Adjusted R-squared 0,531 0,591 0,582 0,686 0,710 0,702 F-value 8,632 10,238 8,758 14,977 15,896 13,572 Highest VIF 3,203 3,539 3,433 3,234 3,752 3,488
17
investors require a lower return on their investment than in the years before. In terms of countries, it appears that banks from Belgium, The Netherlands, Norway, Sweden and the United Kingdom have a significantly lower cost of equity than banks from Spain. Banks from Germany, Greece, and Italy on the other hand, have a significantly higher cost of equity than banks from Spain. Banks from Austria, Denmark, France, Hungary, Ireland and Poland show no significant differences with those from Spain. Furthermore, bank size also seems to decrease the cost of equity. However, this relation is only significant at a 10 percent level and therefore not very strong. Profitability is the only control variable in my analysis that shows no significant relation with the cost of equity.
The main takeaway from the results from model 2 is that there seems to be a significant relation between the total risk disclosure score and the cost of equity. This relation is positive, contrary to my expectations beforehand. The results indicate that a higher risk disclosure score will lead to a higher cost of equity. This suggests that analysts do not reward more transparency on risks with a lower cost of equity.
Model 3 provides more are less the same results as model 2. However, of the four types of risk reporting, only liquidity risk reporting seems to impact the cost of equity significantly. Like total risk disclosure, liquidity risk disclosure impacts the cost of equity positively.
Based on the results from model 4, it appears that disclosure ambiguity impacts the cost of equity. The results indicate that more disclosure ambiguity leads to a higher cost of equity. The relation between disclosure ambiguity and the cost of equity is therefore positive, as I predicted. This relation is statistically significant at the 1 percent level.
The results from model 5 do not show a lot of differences with those from model 2 and model 4. Both the relation between total risk disclosure and the cost of equity and the relation between disclosure ambiguity and the cost of equity remain the same.
In model 6, the only notable difference with model 3 and model 5 is that there is no longer a significant relation between liquidity risk disclosure and the cost of equity when disclosure ambiguity is introduced.
Overall model fit
The overall fit of the regression model is calculated using the R-squared measure. Generally, a higher adjusted R-squared value means a better fit of the data and the regression model. Therefore, the stronger the model, the higher R-squared is. However, R-squared has limitations. The most important limitation is that adding variables will always result in a higher R-squared. To account for this problem, I use the adjusted R-squared to estimate the strength of my regression models, as the adjusted R-squared adjust the R-squared for the number of included variables.
With an adjusted R-squared of 0,531, model 1 shows that 53,1% of the variance in the cost of equity is explained by the control variables. The overall strength of the model increases when I add more variables. The addition of risk disclosure increases the strength to 59,1% for model 2 and 58,2% for model 3. Adding disclosure ambiguity to the control variables results in a model with a predictive strength of 68,6%. These results indicate that the relation between disclosure ambiguity and the cost of equity is stronger than the relation between risk disclosure and the cost of equity. 1 The models with all variables, model 5 and model 6, show the highest adjusted R-squared (respectively 0,710 and 0,702). More than 70% of the changes in the cost of equity are explained by the independent variables in these models.
Collinearity
To check my results for multicollinearity, I looked at the variance inflation factor (VIF). The VIF is an indicator of the effect of independent variables on the standard error. In other words, a high VIF means that a large part of the variables impact can be explained by other variables. Multicollinearity becomes a problem when the VIF is higher than 10 (Hair, Black, Babin and Anderson, 2014). In my regression models, the highest VIF was 3,752. As this number is far below the cutoff point of 10, multicollinearity does not seem to be a problem in this research.
18
However, it is interesting that the highest VIF increases when adding more variables. For example, the highest VIF in model 1 is 3,203 and the highest VIF in model 2 is 3,539. In model 5, when disclosure ambiguity is added to the model, the highest VIF is even 3,752. This means that there is a small amount of multicollinearity between the independent variables. However, it is not enough to spur additional modifications of the model.
Analysis
My first hypothesis is tested in model 2 and model 5. The results show a significant positive relation between total risk disclosure quality and the cost of equity (β = 0,437, p < 0,01 for model 2; β = 0,289, p < 0,01 for model 5). That means that H1 should be rejected. It therefore appears that increasing risk disclosure quality is not beneficial for banks. These results contradicts the theory that increasing risk disclosure quality will be rewarded by investors with a lower cost of equity. It even appears that the opposite holds true for the banks in my sample.
The second hypothesis is tested in model 3 and model 6. General, credit and solvability risk disclosure are not found to have a significant impact on the cost of equity. This hold true for both model 3 and model 6. Liquidity risk disclosure seems to be the only type of risk disclosure that has more impact on the cost of equity. In model 3, the relation between liquidity risk disclosure and the cost of equity is found to be positive and significant (β = 0,175, p < 0,05). This significance of the relation seems to disappear when disclosure ambiguity is added in model 5 (β = 0,099, p > 0,1). Overall that means that there is no clear evidence that one type of risk disclosure that is included in this research is valued more than others by investors.
The third hypothesis is tested in model 4, model 5 and model 6. In all three models, disclosure ambiguity seems to positively impact the cost of equity. The relation appears to be the strongest in model 4 (β = 0,504, p < 0,01). However, model 5 (β = 0,454, p < 0,01) and model 6 (β = 0452, p < 0,01) also provide strong evidence that disclosure ambiguity seems to impact the cost of equity for banks. That means that investors do reward banks that provide unambiguous risk disclosure.
5. CONCLUSION
The role of banks in the financial crisis had a big impact on the trust of society in banks and the morals of bank managers are questioned as a results. For the functioning of the global financial system, restoring that trust is vital. Risk reporting has been identified by governing and standard setting bodies as one of the main tools to do that. Research has focused a lot on what factors contribute to better risk reporting quality. This research however, investigates the economic consequences of risk reporting for the cost of equity for banks. Legitimacy theory, agency theory and stakeholder theory all suggest that increasing risk reporting quality is rewarded by investors that require a lower return on equity. Beforehand, the expectations were that risk disclosure quality would decrease the cost of equity and that disclosure ambiguity would increase the cost of equity.
The expected relation between disclosure ambiguity and the cost of equity is confirmed by the results from this research. Less ambiguous reports were tied to a lower cost of equity. The relation between disclosure quality and the cost of equity appears to be inverse of my expectations. Banks that got a higher risk reporting score showed a significantly higher cost of equity than banks with lower risk reporting scores.
My findings suggest that banks need to focus on conveying clear, information with a minimal amount of ambiguity if lowering the cost of equity is their goal. Increasing their risk reporting quality does appear to only increase the cost of equity. An explanation for this phenomenon might be that through better risk reporting, investors realize that banks face more risks than that they were taking into account. Another explanation is that the relation is actually the other way around. Maybe, banks with a high cost of equity increase their risk reporting quality to try to lower their cost of
19
equity. Banks with a lower cost of equity might not be similarly incentivized to increase risk reporting and neglect their risk reporting somewhat. However, further research would be necessary to determine if this could be a valid explanation.
Standard setters should be able to use this information as it tells them that banks do not seem to have a financial incentives to increase risk reporting. If they wish to increase risk reporting behavior they should consider implementing more standards and regulations that provide banks with less freedom.
Limitations
This research has a number of limitations. The first limitation is that the disclosure index was originally not developed for this application. The disclosure index was developed to measure the impact of other variables on risk disclosure. According to Marston and Shrives (1991), the user group should be considered when determining what items should be included in the index. Developing a disclosure index that is more focused on investors perception of risk reporting might therefore be a good idea for follow-up studies. Because this requires more insight in the consensus of important risk reporting factors according to investors, this was not done for this research.
Another limitation is that I was not able to determine the publication dates of all annual reports. Uncovering these would have allowed me to use the analyst forecasts directly after the publication of the reports. This would improve the overall accuracy of the impact of risk disclosure on the cost.
Future research
For future research I recommend looking at the impact of risk disclosure on other types of economic consequences. Future research may also use the capital asset pricing model (CAPM) to determine the cost of equity instead of the modified PEG model. A benefit of that CAPM is that it allows the inclusion of the cost of debt as well to determine the effect of risk disclosure on the cost of capital. Another suggestion for future is investigating what type of risk reporting items investors perceive to be useful information.
