0
Master thesis Jeroen Vos
University of Groningen Faculty of Economics and Business
MSc Accountancy & Controlling June, 2016
Basel III and Bank Riskiness*
A study on the Consequences of Regulatory Intervention* The author wishes to thank dr. V.A. Porumb for his excellent and motivating guidance in writing this thesis, as well as drs. R. de Wilde RA for offering the unique opportunity to write this thesis in an excellent and stimulating environment, and providing inspiring feedback..
1
new leaks emerge. The only certainty with pay regulation is that new leaks will emerge in unsuspected places, and that the consequences will be both unintended and costly”.
Murphy, K. J. (2013). Regulating banking bonuses in the European Union: a case study
in unintended consequences. European
Financial Management, 19(4), 631-657.
“In this sense, risk management is rather like the “war on drugs”: it might help to reduce drugs coming in, but it can never conclusively win”.
Dowd, K. (2009). Moral hazard and the financial crisis. Cato Journal, 29, 149-166.
2 ABSTRACT
This thesis researches the effects of Capital Requirement Directive IV (CRD IV) implementation on bank risk-taking behaviour. Specifically, I focus on the effects of the new regulation on the risk-taking behaviour of executives, insolvency risk and the quality of bank portfolios. Analysis is done on unbalanced panel data, consisting of 37 to 42 banks in 15 to 16 different countries representing 612 to 723 observations within the European Union. My results show that bank risk does not significantly decrease after CRD IV enforcement, and stability increases. Most notably, results show CRD IV partly achieves its desired goals, with completely opposite than intended mechanisms. The results in this thesis are timely and have important policy implications in order to prevent potential and backfire with unintended consequences.
Keywords: Basel III, CRD IV, capital structure, executive compensation, risk, banks.
Data availability: the data used in this study is publicly available and extracted from BVD Bankscope, Eurostat, Capital IQ and banks’ annual reports.
Copyrights: the copyrights of this thesis are of the university and the author.
Author J.R. Vos Word count: 12.310
Student number S2802740
E-mail jeroenvos91@gmail.com
University University of Groningen
Faculty Faculty of Economics and Business
Degree Master of Science in Accountancy & Controlling
Specialization Accountancy
Supervisor Dr. V.A. Porumb
Co-assessor Prof. dr. I.J.J. Burgers Thesis internship Deloitte Netherlands
3 TABLE OF CONTENTS
1 INTRODUCTION 4
2 THEORETICAL BACKGROUND 7
2.1 REGULATORY INTERVENTION: THE KEY POINTS OF CRD IV 7
2.2 MECHANISM OF EXECUTIVE COMPENSATION: AGENCY THEORY 8
2.3 PAY FOR PERFORMANCE: FORMS OF COMPENSATION 9
3 LITERATURE REVIEW 10
3.1 AGENCY THEORY 10
3.2 EXPECTED EFFECTS OF CRD IV ENACTMENT 10
3.3 EMPIRICAL EVIDENCE ON REGULATORY INTERVENTION 11
3.3.1 REGULATING COMPENSATION 11
3.3.2 REGULATING CAPITAL 12
3.3.3 REGULATING CAPITAL STRUCTURE & EXECUTIVE COMPENSATION 13
4 HYPOTHESIS DEVELOPMENT 15
4.1 EXECUTIVE COMPENSATION & RISK-TAKING BEHAVIOUR AFTER CRD IV 15 4.2 CAPITAL STRUCTURE & RISK-TAKING BEHAVIOUR AFTER CRD IV 17
5 DATA & METHODOLOGY 19
5.1 SAMPLE SELECTION 19 5.2 ECONOMETRIC MODEL 19 5.2.1 RISK MEASURES 20 5.2.2 CONTROL VARIABLES 20 6 EMPIRICAL RESULTS 22 6.1 INITIAL ANALYSIS 22
6.2 EXECUTIVE COMPENSATION ON RISK-TAKING BEHAVIOUR AFTER CRD IV 23
6.2.1 RISK THROUGH INVESTMENTS IN RISK-BEARING ASSETS 23
6.2.2 RISK THROUGH THE QUALITY OF RISK-BEARING ASSETS 24
6.2.3 RISK THROUGH INSOLVENCY RISK 25
6.3 CAPITAL STRUCTURE ON RISK-TAKING BEHAVIOUR AFTER CRD IV 26
6.3.1 RISK THROUGH TOTAL EQUITY 26
6.3.2 RISK THROUGH TIER 1 REGULATORY CAPITAL 27
6.4 ROBUSTNESS TESTS 27
6.4.1 TESTS FOR STRONGER EXPECTED EFFECTS OF CRD IV’S ENACTMENT 28
6.4.2 TESTS FOR WEAKER EXPECTED EFFECTS OF CRD IV’S ENACTMENT 30
6.5 THE EFFECTS OF CRD IV 32
7 SUMMARY & RECOMMENDATIONS 35
4 APPENDICES
APPENDIX I CONTROL VARIABLE FORMULAS 43
APPENDIX II DESCRIPTIVE STATISTICS OF SET H1, H2 & H3 44
APPENDIX III DESCRIPTIVE STATISTICS FOR SET H4 45
APPENDIX IV DESCRIPTIVE STATISTICS FOR SET H5 46
APPENDIX V CORRELATION TABLE FOR SET H1, H2 & H3 (RAR & IZ) 47 APPENDIX VI CORRELATION TABLE FOR SET H1, H2 & H3 (NPL) 48 APPENDIX VII CORRELATION TABLE FOR SET H4 (RAR & IZ) 49
APPENDIX VIII CORRELATION TABLE FOR SET H4 (NPL) 50
APPENDIX IX CORRELATION TABLE FOR SET H5 (RAR & IZ) 51
APPENDIX X CORRELATION TABLE FOR SET H5 (NPL) 52
APPENDIX XI REGRESSION RESULTS FOR SET H1 (RAR) 53
APPENDIX XII REGRESSION RESULTS FOR SET H2 (NPL) 54
APPENDIX XIII REGRESSION RESULTS FOR SET H3 (IZ) 55
APPENDIX XIV REGRESSION RESULTS FOR SET H4 56
APPENDIX XV REGRESSION RESULTS FOR SET H5 57
APPENDIX XVI ROBUSTNESS (+) REGRESSION RESULTS FOR SET H1, H2 & H3 58 APPENDIX XVII ROBUSTNESS (+) REGRESSION RESULTS FOR SET H4 & H5 59 APPENDIX XVII ROBUSTNESS (-) REGRESSION RESULTS FOR SET H1, H2 & H3 60 APPENDIX XIX ROBUSTNESS (-) REGRESSION RESULTS FOR SET H4 & H5 61
5
1 INTRODUCTION
Economic turmoil highlighted the importance of banks for the worldwide economic stability, due to the financial crisis of 2008. Public, academic and regulatory attention quickly lead to the widespread opinion that the credit crisis was initiated by two main phenomena: (1) the absence of capital buffers in banks to absorb underperformance and liquidity runs, (2) and shortcomings in corporate governance leading to excessive risk-taking behaviour by banks executives (Brunnermeier, 2009; DeYoung et al., 2013). The root cause of excessive risk-taking behaviour was quickly appointed to executive compensation (Bebchuck & Spamann, 2008; Kashyap et al., 2008; Kirkpatrick, 2009), as “inappropriate remuneration policies were identified as one underlying factors of the financial risks” (European Banking Authority, 2014, p. 2). Therefore, remuneration policies were linked closely with banks’ risk appetites and long-term goals (Kirkpatrick, 2009).
The public narrative was that executive compensation was not to blame, as executives themselves experienced great losses (Fahlenbrach & Stulz, 2011). However, executives actually experienced positive cash flows from performance-based compensation, taking the downfall of their respective banks into account (Bebchuck et al., 2010). Empirical evidence actually finds an association between executive compensation and excessive risk-taking by banks, concluding that incentives do matter (Bebchuck et al., 2010; Bhagat & Bolton, 2014), not only for executives, but for the whole industry (Avgouleas, 2009).
New regulatory restrictions regarding executive compensation and capital structure emerged as a response to the crisis, in order to mitigate the effects of a possible recurring economic downfall. The most important regulation is the Capital Requirements Directive IV (CRD IV). This set of requirements is in fact the enactment of Basel III into European Union (EU) law and it brings significant changes for the banking sector. CRD IV imposes a cap on variable compensation to decrease the risk-taking behaviour of managers and to assure banks’ financial stability. However, the extent to which their endeavour is successful is an empirical question. This thesis comes to answer this question and studies if, after the 2013 adoption of CRD IV, risk-taking behaviour of executives is affected by banks’ capital and remuneration structure.
Research on remuneration of executives is prevalent in the corporate governance literature, with heavy emphasis on the conflicts between shareholders and managers. In this thesis I focus on agency theory as it represents a good fit for framing my research questions (Jensen & Meckling, 1976). The agency theory focuses on misalignment of interests in a principal-agent relationship (e.g. shareholders and executives), describing selfish behaviour by agents as a result of goal misalignments (Eisenhardt, 1989) and information asymmetry (Richardson, 2000). Due to the information asymmetry, agents can initiate opaque actions such as excessive high risk-taking behaviour, which are potentially triggered by variable incentive remuneration and even lead to excessive risk-taking behaviour (Sanders, 2001). In turn, the classic agency problem arises by creating incentives for CEOs and other corporate executives to increase the value of equity on short notice, but simultaneously decrease the firm value on the long term.
The most recent stream of research in executive compensation and capital structure in the banking sector brings mixed results and is sceptical about the effects of CRD IV. Following the credit crisis, a part of extant research makes recommendations regarding the composition and structure of variable (equity-based) compensation and its effective outcome on risk-taking behaviour (e.g. Bhagat & Bolton, 2014). In contrast, other studies document that the structural change CRD IV induces is not the desired one (e.g. Murphy, 2013). With CRD IV enforcing an increase in equity of banks, and capping incentive compensation as well as changing the structure of the incentive compensation, this thesis investigates whether these empirical
6 recommendations are effective and to see if, and how regulation affects these relations. Research after the enactment of CRD IV mainly focuses on capital structure and the developments regarding bank operations and profitability, but research on CRD IV’s effect on risk-taking behaviour is scarce. The results of this research contribute to existing literature by showing how banks’ risk-taking behaviour changes following regulatory adjustments, and whether the intended effects of the new regulation truly hit their target.
Using unbalanced panel data representing 37 to 42 banks in 15 to 16 countries from several public available sources of European banks for the period 2010-2014, I find mixed results regarding bank risk-taking behaviour following the enactment of CRD IV, not all in favour of the new regulation. Whereas CRD IV intended to limit risk-taking behaviour by capping remuneration, investments in risk-bearing assets and its quality rose simultaneously, following a rise in remuneration. This result is in line with CRD IV’s intentions, but via completely opposite mechanisms. Results on total capital show that total capital, contrary to common belief, actually increases risk measures, whereas the safer core of equity (Tier 1 regulatory capital) potentially leads to lower investments in risk-bearing assets and simultaneously decreases the quality of bank portfolios. The net effect found on insolvency risk is that CRD IV has a detrimental effect on bank stability, conditionally on the state of the bank.
The results of this thesis are of utmost importance as regulation entered enforcement relatively recent, and results show the regulation does not achieve its desired effect via the intended mechanisms. The regulation has effect, but results show it can also backfire quite rapidly if appropriate measures are not taken within a short amount of time. This thesis contributes to literature by warning regulators timely on results of regulatory intervention, enabling to intervene on a timely basis to prevent backfire with serious consequences, potentially aggravating the fruits of realized mechanisms without taking measures at the roots. The remaining of this thesis is as follows: in the next chapter, the fundamental theory used in this paper regarding this subject will be set forth. Following theory, relevant literature regarding these subjects will be discussed in the third chapter, to explain prior research and to see where research stands now. Following the existing knowledge, the fourth chapter encompasses hypotheses development based on prior literature and expectations. Moving to the fifth chapter, the methodology and data will be elaborated on, which includes the data collection and the analysis plan. Finally, the results will be described in the sixth chapter and discussed in seventh chapter.
7
2 THEORETICAL BACKGROUND
In order to obtain an understanding of what the regulatory intervention encompasses and how it influences the underlying mechanisms of the determinants of risk-taking behaviour, the following sections describe: (i) the important effects of CRD IV, (ii) the fundamental theory of executive compensation, the agency theory, and (iii) executive compensation itself.
2.1 Regulatory intervention: the key points of CRD IV
Basel III represents a body of recommendations from the Basel Committee on Banking Supervision (BCBS) aiming to build a stable banking sector. As the BCBS describes, “one of the main reasons the economic and financial crisis became so severe was that the banking sectors of many countries had built up excessive on- and off-balance sheet leverage” (Basel Committee on Banking Supervision, 2011, p. 1). The accord was enforced through the issue of the CRD IV by the European Commission on the 17th of July, 20131.
CRD IV encompasses reforms for banks, which are too vast to discuss in detail. In short, the main points of Basel III as opposed to Basel II are that banks need to improvise their capital base in both quality and quantity, banks will be obliged to strengthen their risk coverage of their exposure, a leverage ratio as a supplementary monitoring tool is introduced, build-up of capital buffers in positive economic environments which can be of support in negative economic environments is mandatory, and finally, the accord prescribes a short-term and long-term liquidity standard to ensure a solid liquidity structure. In addition and most relevant to this paper, incentive executive remuneration is limited to better (and more appropriately) fit the risk appetite of banks. Variable, incentive executive compensation is limited to one hundred percent of the fixed part of the compensation package, with a maximum of two hundred percent with shareholders’ approval and additional disclose is needed for individuals earning more than one million euro per year (CFA Institute, 2013). An overview of the consequences following enactment of CRD IV is presented in Table 1.
When assessing the new remuneration rules of CRD IV, it becomes clear that the Directive aims at aligning remuneration with banks’ risk appetite and that the bank should focus on long-term remuneration including the risks associated with the remuneration2. The Directive distinguishes between fixed remuneration and variable remuneration: payments, regular pension contributions and benefits granted without involvement of performance related variables are identified as fixed remuneration, whereas additional payments, benefits depending on performance and other contractual elements other than routine employment packages are accounted for as variable remuneration3. A maximum of 25% of the total variable remuneration can be paid out in less than five years4, whereas at least 40% of the variable remuneration is deferred over a period between three to five years, unless the remuneration is
1 Directive 2013/36/EU, available at
http://ec.europa.eu/finance/bank/regcapital/legislation-in-force/index_en.htm.
2 See CRD IV, supra note 1, article 63.
3 See CRD IV, supra note 1, article 64.
4 See CRD IV, supra note 1, article 94(1)(g)(iii).
TABLE 1
Phase-in arrangements of CRD IV succeeding CRD III
Phases 2010 2011 2012 2013 2014 2015 to 2019
CRD III III III IV IV IV
Ratio fixed to variable compensation N/A 1:1 - 1:2* Minimum Tier 1 regulatory capital 4,0% 4,5% 5,5% 6,0%
Minimum Total Capital 8,0%
8 exceptionally high. In that case, at least 60% must be deferred over a period which suits the business cycle of the bank5. Variable compensation is allowed to be discounted with a factor of 25% when it is deferred over a period of more than five years. Finally, clawback clauses are now mandatory, meaning that managers need to pay back their bonus ex-ante in case of bad performance6.
2.2 Mechanism of executive compensation: agency theory
The main theory in corporate governance related to incentive executive compensation, is the agency theory, describing conflicts between stakeholders by a principal-agent relationship (Coase, 1937; Jensen & Meckling, 1976; Fama, 1980; Fama & Jensen, 1983a,b). The agency theory posits that an organisation is technically a nexus of contracts, where separation of ownership and control of a firm creates monitoring costs, called agency costs, between the owner of the firm (the principal as the shareholder) and the controller of a firm (the agent as a manager) (Berle & Means, 1932). Agents can engage self-serving behaviour as a result of goal misalignments (Eisenhardt, 1989) and information asymmetry (Richardson, 2000). Due to information asymmetry, agents can initiate opaque actions such as excessive high risk-taking behaviour when a principal does not have the power to monitor and control the agent (Fama, 1980; Eisenhardt, 1989; Dalton et al., 2008). The agency theory identifies two ways information asymmetry can be exploited by the manager: ex ante by adverse selection, meaning that the agent has an information advantage over the principal in a contractual transaction or activity and ex post by moral hazard, meaning the agent can take decisions a principal is unable to monitor and which can be detrimental to the principal (Eisenhardt, 1989). The agency theory knows three kinds of agency conflicts. The first being the conflicts between shareholders and managers due to the separation of ownership and control of the firm, denoted as the Type I agency problem. Shareholders are able to diversify their risk by the ability to invest in multiple firms, but managers are bound to organisation they work for. Therefore, managers are more risk-averse than the shareholders due to the inability to diversify their risk. In addition, managers tend to underperform in their duties. The second agency problem, denoted as the Type II agency problem, confines conflicts between large shareholders versus dispersed shareholders. Large shareholders (blockholders) have the tendency to extract private benefits from which dispersed shareholders do not benefit. On the other hand, large shareholders show more risk-aversion than dispersed shareholders since the financial interest is way higher in firms large shareholders invest in. Third, agency conflicts exist between creditors and shareholders. This agency problem is denoted as the Type III agency conflict. Again, risk-aversion plays a significant role in differences in interests between the two parties: creditors have a lower risk-appetite than shareholders since the former prefers their claims being redeemed in time, while the latter wants to invest in risky project to realise a high return on investments.
Banks are different than other nonfinancial firms, both in corporate governance and thus agency problems, and operations. First, banks have significantly more stakeholders than nonfinancial firms due to being debt-financed up to twice as much as nonfinancial firms, with depositors being the majority of debtholders. Therefore, banks are of high systemic importance and have to deal with a lot of stakeholders due to their socioeconomic impact (Macy & O’hara, 2003; Mülbert, 2009; Mehran et al., 2011). As depositors are among a bank’s debtholders, and post-crisis also governments and thus tax payers, the whole social system bear risks taken by banks (Becht et al., 2012). Second, banks operational activities and balance sheets are less
5 See CRD IV, supra note 1, article 94(1)(m).
9 transparent and more complex than nonfinancial firms, as their financial products are not as easy to understand as a nonfinancial firms’ assets in both substance and quality. In addition, a bank is able to shift between assets very quickly unlike nonfinancial firms as banks’ assets are highly liquid and their information on financial assets is highly aggregated, contributing to the opacity of their operations and balance sheets (Mülbert, 2009; Mehran et al., 2011; Becht et al., 2012). Third, banks produce added value with financial assets (liquidity) instead of operating assets by exploiting an intentional disproportionality of maturity on both sides of their balance sheets. This makes a bank prone to liquidity runs by their depositors, as happened at the start of the 2008 crisis (Macy & O’hara, 2003; Mülbert, 2009; Mehran et al., 2011). Finally, banks are subject to stringent regulation and supervision because of their socioeconomic and systemic interest. As described in the previous part concerning CRD IV, banks are limited in the amount of risk they take (Mülbert, 2009).
2.3 Pay for performance: forms of compensation
An executive’s remuneration package is mostly composed of short-term remuneration and long-term remuneration. Short-term compensation is composed of the base salary and an annual incentive bonus payed at the end of, and based on the results of, a fiscal year. Long-term remuneration includes, from a firm’s point of view, equity in the form of (vested and/or restricted) stock (options), or debt in the form of retirement pensions (Goergen & Renneboog, 2011).
Long-term, incentive executive compensation is employed in the agency theory, in a compensation contract, as a means to align the interests of the principals and the agents, and thus to reduce agency costs (Jensen & Meckling, 1976). The popularity of incentive executive compensation rose enormously in the 1980’s (Frydman & Saks, 2010), to create a pay-performance relationship by granting stock and stock options to managers. However, empirical research report on both the insignificance of a pay-performance relationship (Jensen & Murphy, 1990) and the significance of it (Hall & Liebman, 1998). The main differences between stock and stock options, however, is the potential for earnings. Since stock infers a current ownership of equity, stock options infer a future ownership of equity. Stock has limited potential for earnings as the value of stocks is dependent on market valuation, while options have a very high potential for earnings due to their pre-set exercise price (Chen et al., 2006).
Executive compensation in the banking industry differs from nonfinancial firms. First, relative to other industries, CEOs in banks compensation in cash is lower, are less inclined to accept variable remuneration in stock options and hold less stock options, and the amount of equity in both options and stock relative to their total compensation is lower (Houston & James, 1995). Second, since the banking industry is strongly regulated compared to the nonfinancial industry, compensation in banks is less reactive to performance than in lesser regulated firms (Smith & Watts, 1992; Mayers & Smith, 1992).
10
3 LITERATURE REVIEW
In the previous chapter, the fundamental theory of this study is laid out. In the following sections, the applicable literature on those subjects will be discussed to see where research stands now, and what relations are to be expected according to prior work. The literature review will be used to form testable hypotheses in the following chapter.
3.1 Agency theory
A manager is risk-averse due to the inability to diversify investments in both human and monetary capital (Pathan, 2009), as the agency theory posits in the Type I agency problem (principal-agent or shareholder-manager). The higher risk appetite of shareholders compared to managers is also empirically verified (Laeven & Levine, 2009). Variable compensation is therefore used by shareholders to motivate the manager to employ higher risk-bearing investment strategies, enabling the firm to generate more profit.
The agency theory, however, is based on a normal (nonfinancial) firm with no long-term debt, thus no debtholders. This phenomenon leads to an agency conflict between managers and debtholders in banks when managers are rewarded with equity in any form. Due to leverage and the managers’ capability of influencing endogenous risk, which in turn influences volatility of earnings, managers are incentivized to take on more risk than is desirable (Becht et al. (2012). This is mainly due to the context of asset substitution (also known as the risk-shifting problem): value can be shifted from other stakeholders to holders of equity (Jensen & Meckling, 1976). This can initiate the classic agency problem by creating incentives for CEOs and other corporate executives to increase the value of equity on short notice, but simultaneously decrease the firm value on the long term. Due to banks being primarily debt financed, the risk-shifting problem dominates: investments in risky projects by managers, following the demands of shareholders, allows shareholders to reap the returns of those investments, while the risk is borne by the debtholders (Jensen & Meckling, 1976).
Regarding capital structure, the agency theory suggests there are two ways how capital structure can influence risk-taking behaviour. First, firms with a high debt-to-equity or a low equity-to-assets ratio feature better alignment between shareholders and managers (Jensen & Meckling, 1976; Williams, 1987), and better alignment between both means a decrease in managers’ risk-aversion. Second, when debt is already high but keeps increasing, the problem of risk-shifting becomes significant and the incentives to keep default risk in control declines (Berger & Di Patti, 2006). Both imply higher risk-taking behaviour as a result, as shareholders in themselves have a higher risk appetite than managers and risk-shifting has the same effect. 3.2 Expected effects of CRD IV enactment
Murphy (2013) conducts a case study on the effects of CRD IV on remuneration and its effect on banks’ risk-taking behaviour. The propositions he makes, although not statistically verified, are alarming.
First, the cap on variable remuneration following the enactment of CRD IV logically leads to an increase of fixed compensation, thereby creating the possibility to increase the variable compensation as a means for banks in the European Union to stay competitive on the international labour market, as banks outside the European Union are not subject to the effects on remuneration of CRD IV. Second, as the pay-performance relationship decreases in effect due to the intended lower variable compensation, executives will not be incentivized to take good risks, as in value-increasing risks, but executives will take risks detrimental to long-term value creation as there are no personal gains to strive for. Third, he argues lowering the variable
11 compensation will not lead to less excessive risk-taking behaviour. For variable compensation to sufficient appropriately effective, the compensation should be symmetric in reward: upside risk and the following pay-off is positive for the executive, but vice versa is also true. Capping the variable compensation will strengthen the asymmetric reward effect. Fourth and finally, capping variable compensation will lead to an increase in default risk. As the fixed compensation is very likely to rise as per the first point of this paragraph, the fixed costs of banks will increase by a significant amount. As banks operations are very cyclical in nature, variable compensation fits this profile by variating in height parallel to the business cycles. Higher fixed compensation eliminates this advantage for banks and makes compensation a stable high expense, even in business cycles of low performance (due to independency of the performance of the manager [Thanassoulis, 2012]). Further analysis proposes that caps on bonuses are effective regarding risk-shifting, but not on incentivizing managers to exert effort (Hanekes & Schnabel, 2010).
3.3 Empirical evidence on regulatory intervention
3.3.1 Regulating compensation
Empirical research on executive remuneration in the banking sector following regulatory intervention is scarce, but the results on variable remuneration in the financial sector in general are crystal clear: a greater amount of equity-based compensation in the composition of the variable compensation leads to more risk-taking behaviour by managers (Saunders et al., 1990; Chen et al., 2006; Bai & Elyasiani, 2013), but as Sanders (2001) warned and several researchers conclude (e.g. Bebchuk & Spamann, 2010; Bebchuck et al., 2010; DeYoung et al., 2013; Cheng et al., 2015), it triggers excessive risk-taking behaviour. For example, as shareholders already are highly motivated to pursue high-risk strategies, managers are even more motivated to do so by granting stock options in their remuneration package. Shareholders are already covered for a big part when losses occur due to the leverage effect in banks, but managers are even more protected from losses since they have future and not current claims on the bank via stock options (Bebchuck & Spamann, 2010). There is research implying that executive compensation did not contribute to the financial crisis since executives suffered great losses (Fahlenbrach & Stulz, 2011), but Bebchuck et al. (2010) show through an in-depth analysis that executives ultimately benefited from their variable compensation, despite the major losses. The variable compensation structure eventually even loses its effectiveness the more risk the manager takes, leading to higher firm share volatility and a lower pay-performance relationship (Coles et al., 2006; Bulan et al., 2010). Regulatory intervention is also shown to decrease the pay-performance relationship (Crawford et al., 1995; Hubbard & Palia, 1995). That is, to the extent there is a pay-performance relationship (John & John, 1993; Houston & James, 1995) as some argue that the absence of a pay-performance relationship is needed to mitigate the risk-shifting issue (John & Qian, 2003; John et al., 2010). In that light, it is remarkable that firms with high financial leverage in the period 1992-2006, like banks, show a positive correlation between equity-based incentive compensation and financial leverage (Mehran, 1992; Lin et al., 2012). As per the theoretical framework, executives in the banking sector are severely underpaid opposed to other sectors, creating an argument for a positive pay-performance relationship by using low fixed compensation and high variable compensation (Houston & James, 1995). Moreover, in nonfinancial firms, an inverse relationship exists between fixed compensation and risk-aversion, as executives with higher fixed compensation are better diversified and thus are less risk-averse (Guay, 1991).
While variable remuneration tends to focus mainly on stock and stock options, executives are also remunerated with debt in the form of pension benefits. In empirical
12 research, this is often denoted as inside debt, and evidence shows that inside debt has a risk-tempering effect on risk-taking behaviour. For example, Sundaram & Yermack (2007) argue that CEOs with high debt incentives are more conservative in their management, which is later confirmed by Cassell et al. (2012). These results of Bennet et al. (2015) show that inside debt in remuneration packages of CEOs in bank holding companies lead to better capital positions and management, stronger earnings and a more solid positions in period of underperforming markets. Gerakos (2010) found that executives swap cash compensation and equity grants for pension benefits, which is in line with the findings of Sundaram & Yermack (2007): CEOs tend to prefer pension benefits as they grow older. Inside debt might be an effective way to reduce agency costs by since it lessens the Type III agency problem (managers-creditors), mainly by reducing risk-shifting at the expense of creditors (Edmans & Liu, 2011). The same argument is used by Bolton et al. (2010), who in turn developed a theoretical model which applies a bank’s credit default swap spread to tie executive compensation default risk instead of the use of debt. However, while the use of the spread may have advantages, it is also more complex and opaque to apply. If debt is applied, however, empirical evidence shows that convertible debt is a better way to strengthen the pay-performance relationship (John & John, 1993; Ortiz-Molina, 2007).
3.3.2 Regulating capital
Empirical evidence regarding regulatory intervention on capital structure and the following effects on risk-taking behaviour, however, isn’t that conclusive. One stream of research show that banks lower risk-taking as a consequence of increasing capital as long as asset risk and size remain equal and do not increase (Furlong & Keeley, 1989; Keeley, 1990; Chortareas et al., 2011). In contrast, another stream shows that banks might induce higher risk-taking following capital increases (Flannery; 1989; Koehn & Santomero, 1980; Gennote & Pyle, 1991), as the quality (risk) of banks’ portfolios is negatively (positively) influenced by regulatory intervention (Koehn & Santomero, 1980; Kim & Santomero, 1988; Gennote & Pyle, 1991; Besanko & Kanatas, 1993; Boot & Greenbaum, 1993), implying an increase in bank riskiness. This is more recently confirmed by González (2005). On the opposite, there is also evidence that capital increases lead to a decrease in portfolio risk (Jacques & Nigro, 1997). Blum (1999) also shows that an increase in bank risk is to be expected, as he argues that “in order to raise the amount of equity tomorrow it may be optimal for a bank to increase risk today” (p. 768). Using models that imply full liability and treating equity in the same way as debt, regulatory intervention leads to a lower volume of banks’ risky assets (Koehn & Santomero, 1980; Kim & Santomero, 1988), but introducing limited liability and treating equity different than debt shows that commercial banks’ risk-taking behaviour increases (Rochet, 1992). Specific to Switzerland, banks do increase their equity positions due to regulatory intervention, but that does not change their risk-taking behaviour (Rime, 2001). The same argument is used by Admati et al. (2011) as they state that banks would just raise equity instead of lowering (size of the) risk (assets).
Contrary to one-directional theories in the previous paragraph, Calem & Rob (1999) created a two-directional model which describes the relation between capital land risk-taking behaviour as U-shaped: when a bank is undercapitalized, it takes excessive risk. When capital starts to rise, the risk-taking behaviour starts to decline. However, when the capital keeps rising, so will the risk-taking behaviour eventually. This is confirmed by Gorton & Rosen (1995) on the lower end of equity, who show that when equity becomes low enough, banks will proceed to take excessive risks. In theory, a higher capital-to-assets ratio is assumed to reduce the risk on equity, however, research by Berger (1995) shows the opposite. Shrieves & Dahl (1992) agree with this, as they show that capital and risk are positively related. The U-shaped model
13 of Calem and Rob (1999) is shown in figure 1. The graph denoted as (a) shows the portfolio share of risky assets as a percentage of the total portfolio, related to the ratio of capital to total assets, without a premium charge if a bank is undercapitalized. Graph (b) shows the same, but with a premium charge if a bank is undercapitalized.
FIGURE 1. U-shaped relation between capital and risk. Graph (a): no premium charge for undercapitalization, graph (b): premium charge for undercapitalization (Calem & Rob, 1991, p. 336).
3.3.3 Regulating capital structure & executive compensation
When taking both capital structure and equity and debt compensation in consideration, empirical evidence shows that the composition of the firm’s debt and equity, and the composition of debt and equity in compensation packages are related regarding investment decisions. The difference between the two is denoted as the leverage gap and evidence is found that the greater the leverage gap is, the more distorted the investment decisions are. Managers with more debt relative to equity in their remuneration packages under-invest, while managers with more equity relative to debt in their remuneration packages over-invest (Eisdorfer et al., 2013). These results imply that the composition of the remuneration package influence the incentives for risk-taking behaviour, in such a way that more debt (instruments) in compensation reduces risk-taking behaviour, in accordance with the previous discussed effects of debt in compensation packages. From a theoretical point of view, there should be a balance between debt and equity to align the risk appetite of both the bank and the manager. As discussed before, equity-based compensation can induce excessive risk-taking behaviour, but too much debt in a compensation contract could lead to excessive risk-reducing behaviour as Low (2009) shows. She concludes that high risk-aversion, just like low risk-aversion, could
14 potentially lead to equal problems in investment decisions. Lewellen (2006) shows similar results, as she found that ownership of in-the-money options (meaning that these options now have a positive value when exercised in the future) could discourage risk-taking behaviour.
As discussed in the theoretical framework, compensation in banks is less reactive to performance than in lesser regulated industries (Smith & Watts, 1992; Mayers & Smith, 1992). Post-crisis regulatory protection for depositors lessens incentives for monitoring excessive risk-taking behaviour as their savings are protected, independent of the investment strategies of banks. Since regulation removes incentives for depositors as debtholders, shareholders experience less resistance in practicing their preferred high risk strategies (Macey & O’hara, 2003). Government protection for depositors, leading to “insured debt”, worsens risk-taking behaviour by banks (Hanekes & Schnabel, 2010).
15
4 HYPOTHESIS DEVELOPMENT
Using the literature on agency theory, I develop the main hypotheses of the paper in this chapter. The hypotheses will be tested in the following chapter.
4.1 Executive compensation & risk-taking behaviour after CRD IV
After the enactment of CRD IV, which imposes a cap on variable remuneration by linking it to fixed compensation, it became likely that variable remuneration will drop, ceteris
paribus. The agency theory posits that managers are risk-averse as described in the agency
Type I problem, and variable compensation is employed as a mechanism to decrease the managers’ risk-aversion. This initial assumption will therefore lead to an increase of managers’ risk-aversion in comparison to the period prior to the adoption of CRD IV.
However, by capping variable compensation, a rise in fixed remuneration is also expected, in order to maintain banks’ competitiveness in the international executive labour market (Murphy, 2013). Applying empirical evidence of nonfinancial firms to this effect, the increase of fixed compensation will lead to less risk-aversion due to better diversification possibilities (Guay, 1991), leading to more investments in risky assets, ceteris paribus. In addition, the cap on variable compensation is enforced to limit excessive risk-taking behaviour. This reasoning is effective, conditional on banks keeping fixed compensation equal, or keeping total compensation equal prior to the enactment of CRD IV and rebalance the weights of fixed and variable compensation in executives’ remuneration packages to comply with CRD IV. However, if the fixed compensation does rise and the variable compensation exceeds the 1:1 ratio, the variable compensation after the enactment of CRD IV might exceed the variable compensation prior to CRD IV. To clarify this effect, Table 2 below illustrates possible scenarios prior CRD IV and post CRD IV.
According to the agency theory, shareholders have higher risk appetites than managers. Therefore, shareholders have incentives to decrease an executives’ risk-aversion to strive for maximal firm value. This translates to higher variable compensation for executives and as such, I expect both fixed and variable compensation to rise after the enactment of CRD IV, in such way that the third scenario presented in the table is more likely to arise than the second.
In addition, a higher ratio of fixed to total compensation is to be expected after CRD IV’s enforcement, leading to a pay-performance relationship with lesser strength but which is also more asymmetrical in upside versus downside rewards (Murphy, 2013). Therefore, I expect executives to decrease investments in risk-bearing assets as the ratio of fixed to total compensation increases, as executives with relatively higher fixed compensation have lesser personal gains to realize and are to a lesser extent affected by downward risk. This also applies for absolute fixed compensation. I expect to find opposite effect for higher absolute variable compensation and variable to total compensation ratios, as executives have more to gain when variable compensation is relatively higher than fixed compensation. In extension, I expect to
TABLE 2
Possible scenarios prior and after the enactment of CRD IV
Pre-CRD IV Post-CRD IV (rebalance) Post-CRD IV (maximum compensation) Fixed compensation 2 4 4 Variable compensation 6 4 8 Total compensation 8 8 12
16 find a positive relationship between investments in risk-bearing assets and the ratio of variable to fixed compensation.
Hypothesis 1a: After the implementation of CRD IV, executives with higher fixed compensation
decrease their risk-taking behaviour by decreasing investments in risk-bearing assets.
Hypothesis 1b: After the implementation of CRD IV, executives with higher variable compensation
increase their risk-taking behaviour by increasing investments in risk-bearing assets.
Hypothesis 1c: After the implementation of CRD IV, executives with a higher ratio of fixed to total
compensation decrease their risk-taking behaviour by decreasing investments in risk-bearing assets.
Hypothesis 1d: After the implementation of CRD IV, executives with a higher ratio of variable to total
compensation increase their risk-taking behaviour by increasing investments in risk-bearing assets.
Hypothesis 1e: After the implementation of CRD IV, executives with a higher ratio of variable to fixed
compensation increase their risk-taking behaviour by increasing investments in risk-bearing assets.
Investing in risk-bearing assets is not necessarily detrimental for banks as long as the quality of the assets is sufficient. To be able to draw conclusions from the previous proposed hypotheses, I will test for the quality of the risk-bearing assets. Most literature concludes that high fixed compensation leaves executives indifferent about the positive or negative returns on investment, as the pay-performance relationship is non-existent when only applying fixed compensation. This would logically lead to a negative relationship between the quality of bank portfolios and fixed compensation. Variable compensation, by contrast, has the opposite effect, in accordance with the agency theory.
However, theory mainly focuses on normal and steady business environments. The current situation is that banks are under regulatory scrutiny following a global economic recession. Therefore, I expect fixed compensation in absolute numbers to influence investment decisions negatively, leading to a lower quality of bank portfolios as executives have no personal gains to realize as a pay-performance relationship is not evident. The opposite effect applies to absolute variable compensation and I expect variable compensation to influence bank portfolio quality positively. However, when comparing the ratios of both fixed and variable compensation to total compensation, I expect to find a lower quality in bank portfolios for banks’ executives having relatively higher fixed to total compensation as the pay-performance relationship argument still applies, despite the pressure from regulatory attention. In extension, I expect the opposite effect for the variable to fixed compensation ratio and the variable to fixed compensation ratio.
Hypothesis 2a: After the implementation of CRD IV, executives with higher fixed compensation
increase their risk-taking behaviour by decreasing the quality of investments in risk-bearing assets.
Hypothesis 2b: After the implementation of CRD IV, executives with higher variable compensation
decrease their risk-taking behaviour by increasing the quality of investments in risk-bearing assets.
Hypothesis 2c: After the implementation of CRD IV, executives with a higher ratio of fixed to total
compensation increase their risk-taking behaviour decreasing the quality of investments in risk-bearing assets.
Hypothesis 2d: After the implementation of CRD IV, executives with a higher ratio of variable to total
compensation decrease their risk-taking behaviour increasing the quality of investments in risk-bearing assets.
17 Hypothesis 2e: After the implementation of CRD IV, executives with a higher ratio of variable to fixed
compensation decrease their risk-taking behaviour increasing the quality of investments in risk-bearing assets.
The expected rise in fixed compensation will logically lead to more fixed costs for banks as opposed to the period prior to the enactment of CRD IV. This effect of CRD IV will lead to a higher insolvency risk and thus default risk as banks’ operations are cyclical in nature (Murphy, 2013). Higher fixed costs in an underperforming business cycle aggravates the downside effects that come along with underperformance and as such, this will lead to more instability in banks. Following this logic, I expect a positive relationship between fixed compensation and insolvency risk, and the opposite for variable compensation.
Hypothesis 3a: After the implementation of CRD IV, banks compensating executives with higher fixed
compensation increase insolvency risk.
Hypothesis 3b: After the implementation of CRD IV, banks compensating executives with higher
variable compensation decrease insolvency risk.
However, looking at proportional differences in banks, I expect banks that remunerate executives with a higher fixed to total compensation to show higher insolvency risk. In contrast, I expect banks that remunerate executives with a variable to total compensation to show lower insolvency risk, which logically then also applies to the ratio of variable to fixed compensation. Hypothesis 3c: After the implementation of CRD IV, banks compensating executives with higher fixed
to total compensation increase insolvency risk.
Hypothesis 3d: After the implementation of CRD IV, banks compensating executives with higher
variable to total compensation decrease insolvency risk.
Hypothesis 3e: After the implementation of CRD IV, banks compensating executives with higher
variable to fixed compensation decrease insolvency risk.
4.2 Capital structure & risk-taking behaviour after CRD IV
Literature did not reach a consensus regarding the relation between capital structure and risk-taking behaviour. However, most research concludes that an increase of capital following regulatory intervention does increase risk-taking behaviour, as the relation between capital and risk-taking behaviour is U-shaped (Calem & Rob, 1999). Banks have shown to be undercapitalized before and during the crisis, and it were those years in which their risk-taking behaviour was utterly excessive. Following the model of Calem & Rob (1991), I expect that the regulatory intervention of CRD IV leads to an increase in risk-taking behaviour by banks when taking total equity into account, as banks’ regulatory equity requirements are on the rising line of the U-shaped model and a part of this capital (total capital net of Tier 1 regulatory capital) can be exploited in business operations. I expect this effect to show in increasing investments in risk-bearing assets, but an increase in bank portfolio quality as banks have to increase earnings to increase equity. Finally, I expect total capital to be positively related with insolvency risk as banks have more value-at-risk when investments in risk-bearing assets increases.
18 Hypothesis 4a: After the implementation of CRD IV, executives of banks with higher total capital
increase their risk-taking behaviour by increasing investments in risk-bearing assets.
Hypothesis 4b: After the implementation of CRD IV, executives of banks with higher total capital
decrease their risk-taking behaviour by increasing the quality of investments in risk-bearing assets.
Hypothesis 4c: After the implementation of CRD IV, bank with higher total capital show an increase
of default risk.
The application of CRD IV requires bank to hold more risk-free capital in the form of Tier 1 regulatory capital. This capital is risk-free in the sense that banks are restricted from exploiting this capital in their business operations and therefore, this capital serves as a safeguard for bad performance in both operational and general economic terms. Following this logic, Tier 1 regulatory capital should lead to a decrease in risk-bearing assets, simultaneously increasing the quality of those assets and lowering default risk. However, Tier 1 regulatory capital is calculated by the ratio of Tier 1 regulatory capital to risk-weighted assets. Tier 1 regulatory capital serves as a capital buffer for value-at-risk and therefore, Tier 1 regulatory capital increases parallel with risk-bearing assets. This leads to the expectation that Tier 1 regulatory capital is positively related to investments in risk-bearing assets. At the same time, due to the regulatory scrutiny and banks’ incentives to increase capital, the quality of investments is expected to increase. Finally, as with the increase in total capital, an increase in insolvency risk is expected as the value-at-risk increases simultaneously with the parallel increase in Tier 1 regulatory capital.
Hypothesis 5a: After the implementation of CRD IV, executives of banks with higher Tier 1 regulatory
capital increase their risk-taking behaviour increasing investments in risk-bearing assets.
Hypothesis 5b: After the implementation of CRD IV, executives of banks with higher Tier 1 regulatory
capital decrease their risk-taking behaviour by increasing the quality of investments in risk-bearing assets.
Hypothesis 5c: After the implementation of CRD IV, banks with higher Tier 1 regulatory capital show
19 5 DATA & METHODOLOGY
This chapter describes the method employed and data used in this study. The sample and methods of testing are discussed first, followed by the econometric model which includes the variable construction and the regression model. Finally, the control variables are explained. 5.1 Sample selection
The sample starts with remuneration data of 66 banks in the European Union from the years 2008 to 2014, for a total of 2.468 observations. Selecting the years 2010 to 2014 results in a sample of 1.698 observations. Next, additional data for dependent variables, independent and control variables is collected and matched with the existing remuneration data. Initially, the sample consists of 60 banks which are affected by CRD IV regarding compensation and capital adequacy rules. Eliminating missing values, the final and definitive sample consists of 57 banks representing 1.433 observations. As not all data is available for both remuneration and capital effects, the sample will be tested by unbalanced panel data, which leaves 41 banks representing 703 observations in 16 countries for compensation using the risk asset ratio (RAR) and iZ-score (iZ) as dependent variables, and 39 banks representing 675 observations in 16 countries for compensation using the non-performing loans ratio (NPL) as dependent variable. Next, regarding capital adequacy rules, two regressions are employed using total equity in the first one, and Tier 1 regulatory capital in the second. For total equity, this results in 42 banks representing 723 observations in 15 countries for RAR and iZ as the dependent variable, and 38 banks representing 690 observations in 15 countries for NPL as the independent variable. For Tier 1 regulatory capital, this results in 39 banks representing 633 observations in 15 countries for RAR and iZ as the dependent variable, and 37 banks representing 612 observations in 15 countries for NPL as the dependent variable. The data is extracted from BVD Bankscope for all independent variables with the exception of the annualized three-month interbank rate, which is extracted from Eurostat. Remuneration data is obtained through Capital IQ.
Due to the vast size of the descriptive statistics tables and the correlation tables, the descriptive statistics for the sets of hypotheses can be found in the appendices. The descriptive statistics for the hypotheses can be found in the first to the fourth appendix, the correlation tables can be found in the fifth to the tenth appendix.
5.2 Econometric model
The basic econometric model is shown in equation (1) as a multiple linear regression, where the dependent variable RISKi,t is measured through three types of risk, explained in the
following section. The variable IVi,t stands for the independent variable which will differ for the first and second hypotheses, and will be used to measure the relation and the direction of the relation between RISKi,t and IVi,t. The interaction term between CRD IV and the dependent variable RISKi,t represents the incremental effect of CRD IV adoption on bank riskiness. Finally, CONTROLSi,t represent a set of control variables at bank and country level.
20
5.2.1 Risk measures
In measuring risk-taking behaviour by banks, I use several measures to determine the magnitude of riskiness, adopted from prior work. First, to measure investments in risk-bearing assets, the ratio of risk assets is computed as used by Delis & Kouretas (2011) and is shown below in equation (2). This proxy for risk-taking behaviour is calculated by the ratio of risk assets to total assets for bank i in year t. This variable will be denoted as the risk asset ratio (RAR). The nominator, risk assets, include all assets in banks’ portfolios except for cash and cash equivalents, government securities and balances due from other banks, as they are not influenced by market risks and/or credit quality.
𝑅𝐴𝑅𝑖,𝑡 = 𝑅𝑖𝑠𝑘 𝑎𝑠𝑠𝑒𝑡𝑠𝑖,𝑡 𝑇𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡𝑠𝑖,𝑡
The second ratio for risk, also derived from Delis & Kouretas (2011), represents the quality of banks’ portfolios as a proxy. The variable is computed by the ratio of non-performing loans to total loans, denoted as non-performing loan ratio (NPL), and is shown below in equation (3). As the non-performing loans increase, the quality of banks’ portfolios decrease and thus increase the risk position of the bank as this encompasses higher losses on outstanding loans. Important to note, though, is that his variable is more limited than risk assets as executives cannot influence this ratio as it is subject to external influences (systemic risk) and the ratio technically reflects the credit risk, while the risk assets ratio is a direct measure for banks’ risk-taking level. The variables are computed for bank i in year t.
𝑁𝑃𝐿𝑖,𝑡 =𝑁𝑜𝑛 − 𝑝𝑒𝑟𝑓𝑜𝑟𝑚𝑖𝑛𝑔 𝑙𝑜𝑎𝑛𝑠𝑖,𝑡 𝑇𝑜𝑡𝑎𝑙 𝑙𝑜𝑎𝑛𝑠𝑖,𝑡
Finally, the banks’ insolvency or default risk will be measured. Following Laeven & Levine (2009), I use the banks’ natural logarithm of the Z-score to measure this type of risk, called the inverse Z-score. The measure is calculated by the natural logarithm of return on assets plus the capital asset ratio, divided by the standard deviation of asset returns. The inverse Z-score is employed as the asset returns are not normally distributed. The equation is shown below in equation (4). This measures the distance from insolvency, with a higher Z-score indicating more stability. The inverse Z-score will be denoted as iZ. The variables are computed for bank i in year t.
𝑖𝑍𝑖,𝑡 = 𝑙𝑛 (𝑅𝑂𝐴𝑖,𝑡+ 𝐶𝐴𝑅𝑖,𝑡 𝜎(𝑅𝑂𝐴𝑖,𝑡) )
5.2.2 Control variables
Risk is not only influenced by the composition and height of compensation and capital levels, but also by bank- and country-level variables. As such, at the bank-level, I control for several variables used in prior literature. Following Leaven & Levine (2009), I control for earnings volatility, loan loss provisions, liquidity, growth, size, Tobin’s Q, and large shareholder ownership. Following Delis & Kouretas (2011), I also control for bank efficiency, profitability, and capitalization. Not only are control variables from previous literature used, the final variables employed are the annual interbank interest rate, Tier 1 regulatory capital, dummy variables for each individual year and dummy variables for each country. The
(2)
(3)
21 justification for these variables is now explained, the formula for each variable is added in appendix I.
Earnings volatility serves as an alternative measure for bank risk. Executives have the ability to exercise influence on this variable for opportunistic (e.g. compensation motives) and contract efficiency (e.g. political costs) motives. Loan loss provisions are expenses to form allowances on bad loans of which the maturity passed 90 days. The main difference with the dependent variable non-performing loans is that loan loss provisions are already acknowledged as default loans, while non-performing loans are not. As such, this variable measures the accounted outcome of instead of the accounted presence of risk. Liquidity follows logic reasoning, lower liquidity increases default risk. Bank growth, size, efficiency, profitability and capitalization are a two-edged sword and interrelated: values of the variables indicate risk, both downside and upside risk as they are respectively lower or higher. The Tobin’s Q measure is shown to be affecting the bank’s valuation and regulatory capital requirements are shown to influence bank valuation (Leaven & Levine, 2009). A higher Q-value would then indicate higher bank stability. The same authors also find that large ownership (higher than ten percent of total shares) can potentially reduce the regulatory capital stringency effect, as “this shows that the stabilizing effects of an intensification of capital stringency regulations diminish when the bank has a large owner with the incentives and power to increase bank risk” (Leaven & Levine, 2009, p. 269). The annual interbank interest rate is the interest rate for which banks can mutually borrow money and thus serves as a cost price for interbank loans. A lower cost price makes it less costly to take risk with potential negative outcomes, lowering the threshold to gamble and potentially increase risk. Tier 1 regulatory capital is taking into the equation as this serves as mandatory risk-free capital, i.e. banks are not allowed to exploit this capital to invest in its operations. However, the amount of Tier 1 regulatory capital a bank should hold is dependent on the amount of risk-weighted assets through a positive relationship. Therefore, Tier 1 regulatory capital should show a risk-increasing effect. Finally, dummy variables are included for both years and countries. Every observation is coded with a one for the country the bank is situated to eliminate country-fixed effects and the same applies to each year, to eliminate time-fixed effects. The year dummy of 2014 also serves as the indicator for the enactment of CRD IV in interaction terms. The fourth and fifth set of hypotheses have a control variable for respectively total equity and Tier 1 regulatory capital, which takes a value of one if their total capital or Tier 1 regulatory capital is above the mean of the sample. The fourth and fifth hypotheses are tested for banks with “higher capital”, which is operationalized to banks with (Tier 1 regulatory) capital levels higher than the mean of the corresponding sample. This dummy is also included in an interaction term to test for banks which have above sample mean total equity and Tier 1 regulatory capital. As per the first to the third set of hypotheses, fixed and variable compensation to total assets is included in the control variables for the fourth and fifth hypotheses, as compensation has both increasing and decreasing effects on executives’ risk-aversion.
22
6 EMPIRICAL RESULTS
This chapter describes the results from the regression models as explained in the previous chapter. First, results regarding compensation in relation to risk are laid out. Next, the results regarding capital structure and risk are described. Finally, the robustness of the results will be tested by regressions computed for countries in which strong and weak reactions to CRD IV are expected. The results will serve as input for the next chapter, the conclusions. 6.1 Initial analysis
As argued in the hypotheses development, the effectiveness of CRD IV regarding the cap on variable executive compensation depends on the banks’ response to the enactment. If banks rebalance the weights of fixed and variable compensation while keeping total compensation equal, or do not adjust fixed compensation after the enactment, variable compensation will drop. However, if banks to choose to rebalance the weights of compensation but increase the ratio of fixed to variable compensation above 1:1, variable remuneration will ultimately rise, realising the opposite effect of CRD IV. Table 3 shows the mean values of the different samples used in this research, of compensation but also of total equity to assets and Tier 1 regulatory capital to total assets as these are the main variables of interest in this research. The figures show that compensation, both fixed and variable, increased in 2014, the first full year of CRD IV is applied. Although the average ratio of variable to fixed compensation is closer to 1:1 than 1:2, it significantly rose after 2013 as well as absolute fixed and variable compensation. Therefore, definitive conclusions cannot be drawn from these figures alone. The intention of CRD IV was to limit and cap variable compensation, but it only achieved the former while the latter was, and is still not utterly excessive in terms of ratios.
To discover what CRD IV did achieve, regression results are discussed in the coming paragraphs. First, the effects of compensation on risk are discussed, followed by the capital structure.
TABLE 3
Average mean values of samples
2010 2011 2012 2013 2014
Fixed compensation 5.593.318 6.354.673 6.275.312 6.547.148 9.837.040 Variable compensation 7.795.530 5.703.939 5.860.626 7.841.883 12.264.104 Total compensation 13.388.848 12.058.612 12.135.938 14.389.031 22.101.144
Fixed to total compensation 0,61 0,70 0,64 0,66 0,59
Variable to total compensation 0,39 0,30 0,36 0,34 0,41
Variable to fixed compensation 1,60 0,69 0,95 0,88 1,18
Capitalization 0,0676 0,0662 0,0690 0,0756 0,0730
Tier 1 regulatory capital 0,1167 0,1235 0,1313 0,1323 0,1321 Year
23 6.2 Executive compensation on risk-taking behaviour after CRD IV
6.2.1 Risk through investments in risk-bearing assets
Results (Table 4) show that the direction of the absolute fixed compensation coefficient is in line with the hypothesis, however, it has no significant effect on investments in risk-bearing assets, finding no support for H1a (model 1). This means that fixed compensation has no significant influence on investments in risk-bearing assets, without influence of CRD IV at all. In contrast, absolute variable compensation, has a positive relation with investments in risk-bearing assets at five percent significance. Initially, this supports H1b (model 2), meaning that higher variable compensation leads to higher investments in risk-bearing assets. However, the interaction term is not statistically significant, meaning that variable compensation has a significant relationship with investments in risk-bearing assets, but CRD IV has no effect in the relationship between absolute variable compensation to investments in risk-bearing assets. Remarkably, absolute total compensation is tested for robustness and shows a positive significant relation at five percent significance, despite the absolute components of compensation are not both significant. Here too, the interaction term is not statistically significant (model 3).
TABLE 4
M ain regression results of the first set of hypothesis
M odel (1) M odel (2) M odel (3) M odel (4) M odel (5) M odel (6)
RAR RAR RAR RAR RAR RAR
0,000 (0,000) 0,000** (0,000) 5,573** (0,000) -0,098*** (0,012) 0,098*** (0,012) 0,005*** (0,001) FCxCRD -0,000 (0,000) VCxCRD -0,000 (0,000) TCxCRD -0,000 (0,000) FCTCxCRD 0,038 (0,029) VCTCxCRD -0,038 (0,029) VCFCxCRD -0,002 (0,004) Variable to fixed compensation Variable to total compensation Fixed to total compensation Total compensation
Standard errors in parentheses. ***: significant at 1%, ** significant at 5%, * significant at 10%. Full regression table in appendix XI. Significant results highlighted. Variable
compensation Fixed compensation
24 The ratio of fixed compensation to total compensation shows a negative, significant relation at one percent significance, indicating that banks compensating executives with higher fixed to total compensation decrease investments in risk-bearing assets. Again, the interaction term for this regression is not significant, finding no support for H1c (model 4) which means that the enactment of CRD IV has no influence on this relationship. In contrast, a positive relation is found for bank compensating executives with more variable to total compensation at one percent significance. Executives with higher variable to total compensation invest more in risk-bearing assets. No statistical significance is found for the interaction term, finding no support for H1d (model 5). CRD IV has no influence on the relationship between the ratio of variable to total compensation. Finally, a positive relationship for executives receiving a higher ratio of variable to fixed compensation is found at one percent significance, leading to the same conclusion of H1d. No significance applies to the interaction term thus influence of CRD IV, not supporting H1e (model 6).
6.2.2 Risk through the quality of risk-bearing assets
Results (Table 5) show that the quality of bank portfolios is positively affected by executives receiving higher absolute fixed compensation at one percent significance, as higher absolute fixed compensation leads to lower performing loans. In turn, lower non-performing loans show better quality. However, this is not due to CRD IV as the interaction
TABLE 5
M ain regression results of the second set of hypothesis
M odel (7) M odel (8) M odel (9) M odel (10) M odel (11) M odel (12)
NPL NPL NPL NPL NPL NPL -0,000*** (3,362) -0,000** (1,578) -0,000*** (1,203) 0,0216*** (0,006) -0,021*** (0,006) -0,001* (0,000) FCxCRD -0,000 (0,000) VCxCRD -0,000*** (0,000) TCxCRD -0,000*** (0,000) FCTCxCRD 0,084*** (0,015) VCTCxCRD -0,084*** (0,015) VCFCxCRD -0,009*** (0,002) Variable to fixed compensation
Standard errors in parentheses. ***: significant at 1%, **: significant at 5%, *: significant at 10%. Full regression table in appendix XII. Significant results highlighted. Fixed compensation Variable compensation Total compensation Fixed to total compensation Variable to total compensation
