RMI and the impact on Tail Risk: The European Story

(1)

RMI and the impact on Tail Risk:

The European Story

University of Groningen

MSc Finance

Author: Chris Maas

(2)

RMI and the impact on Tail Risk:

The European Story

Abstract

Since the financial crisis risk management has been at the centre of debate. In this paper I analyze the impact of independent, strong risk management function by using a risk management index (RMI), constructed with a unique database, to measure its impact on tail risk for European publicly listed bank holding companies (BHCs). I find that BHCs with higher RMI have higher excess returns compared to other banks. However, I do not find proof of a positive impact on tail risk.

JEL classification: G210, G280, G320

(3)

I. Introduction

Europe and its banks are still recovering from the shock of the financial crisis. In the fall of 2014 the European Supervisory Mechanism was introduced to centralize supervision within the European Union. An important moment for European banks, that leading up to this were thoroughly reviewed in the asset quality review. The importance of the risk management has become clearer than ever. As Kashyap (2010) stated it in a paper prepared for the financial crisis inquiry commission:”… Yet, risk management at almost all large banks is essentially a black box.”

In the past 5 years banks have adjusted their approach towards risk. The fourth annual report from EY (formerly Ernst & Young) and the Institute of International Finance shows that banks are making significant progress in adjusting the organization in light of changes in governance frameworks. A recent survey shows that Board Risk Committees are now installed almost everywhere and time (and resources) spent on risk have substantially increased. (EY, Remaking financial services: risk management five years after the crisis, 2013).

These measures were taken after the fallout of the financial crisis, which started in 2008, became clear. Pre-2008, banks’ risk managers had not taken adequate measures to curtail the excessive tail risks that were taken on by bank executives and traders. Risk managers either they failed to take note of exposures due to their methodology of assessing risks historically (Schleifer (2011)) or they were not in the position to restrain executives with compensation schemes that are heavily geared towards variable rewards. This all added to institutional arrangements that enhanced the status of risk takers and their influence at the expense of risk managers and their independence (Kashap, Rajan and Stein (2008), Senior Supervisors Group (2008)).

(5)

What becomes clear from the E&Y Report is the large difference in momentum between both sides of the Atlantic in assessing the internal risk culture. More than 85% of North American banks currently have programs for assessing the internal risk culture while only 60% of European banks have similar programs. Next to that, there is more unclear about actions taken within banks in Europe when comparing it to their North American colleagues (EY, Remaking financial services: risk management five years after the crisis, (2013)). Clearly, a transatlantic divide looms, which warrants research on the width of this divide by studying the European BHCs and comparing them to their North American counterparts. This leads to the following research question:

Does an independent and strong risk management function, as measured by RMI, limit excessive risk taking within European BHCs?

The main hypothesis is that BHCs with a strong and independent risk management function has lower tail risk, all else being equal. To answer test this hypothesis I first construct a RMI for BHCs for the years 2004-2013. Second, the impact on RMI of various financial characteristics is analyzed. Third, the impact of RMI on risk taking and various indicators of firm performance is analyzed. This approach is inspired by the methodology of Ellul and Yerramilli (2013).

They conduct their study based on publicly available regulatory filings for 74 of the top 100 BHCs from the US to investigate the relationship between risk taking by banks and the function of the risk management. First they collect seven proxy variables indicating the importance and relative importance of the risk management function. The second step consists of condensing these factors into a single RMI index factor by principal component analysis (PCA). The final step in their research is investigating the link between RMI and various indicators of risk taking behavior and performance of BHCs.

As noted by Ellul and Yerramilli (2013) the risk management function of a BHC may itself be endogenously reinforcing. Other factors, such as the risk culture or underlying business model of the BHC may influence the risk management system. This may mean that conservative BHCs take lower risks and also put in place stronger risk management systems. This is being referred to as the

business model channel. Alternatively, it is possible that some BHCs chose to undertake high (low)

risks and couple this with a strong (weak) risk management system. This is referred to as the hedging

channel since it corresponds with the core predictions of the theories of hedging (Smith and Stulz

(6)

Following Ellul and Yerramilli’s approach I constructed a sample of 50 European BHCs. These BHCs have the highest amount of total assets for publicly listed banks. They account for a total book value of total assets of 24.88 trillion Euro’s, or 94.41% of the total value of assets for all publicly listed BHCs within fifteen countries in the Euro zone. For these BHCs a RMI is constructed and related to their performance and risk taking. I find that BHCs who experienced higher lagged tail risk have a lower RMI in the subsequent year, everything else being equal. No evidence is found however for a relationship between lagged RMI and tail risk. I also do not find any support for either a hedging or a business model channel approach by BHCs, as a response to the financial crisis. Results with respect to firm performance are mixed, with RMI only showing a significant positive influence on excess peer returns when using a single factor capital asset pricing model. I am, based on these results, not able to give a decisive answer on the research question.

This research makes the following contributions. First, it offers a systematic examination of the organization of the risk management function within European BHCs, despite data limitations the quality of the risk management seems to be reasonably captured by RMI. Second, to the best of my knowledge, this research is the first to utilize the recently developed RMI on a different sample of banks. Third, it sheds some light on the relationship between risk management and tail risk for European banks, though no decisive answer can be found. Fourth, this paper contributes to the recent literature that examines risk-taking by banks and the overview of (European) banks (Laeven and Levine (2009), Fahlenbrach et al. (2012), Beltratti and Stulz (2012).

(7)

II. Literature Review and hypothesis development

Following the reasoning of Ellul and Yerramilli (2013), the main hypothesis of this paper is that European banking institutions with strong and independent risk management functions should have lower enterprise wide tail risk, all else being equal.

Agency problems became clear within banks during the crisis. CEOs were pressured to create high returns by taking excessive risks in the short run, which proved detrimental in the long run. Not only CEOs but also traders were pushing the limit in search for their own variable remuneration. Traders within BHCs were given incentives to earn returns above the appropriate market premium. The above market returns thus generated were not necessarily demonstrations of superior skills, but rather were a form of insurance premiums earned for taking on the risk of infrequent events. This lead to an increase in tail risk (Kashyap Rajan and Stein (2008), Rajan (2005)).

To counter this increase in tail risk for BHCs and the increase in systemic risk, direct supervisory of the financial sector is vital. However, more stringent rules, for example higher capital holding requirements, can lead to circumventing these rules, as has been seen in the case of triple A rated Collaterized Debt Obligations (CDOs). On the other hand, measurements to mop up the damage after an event may give unwanted incentives, such as the implicit guarantee of a bailout for ‘too big to fail’ banks or a customer deposit guarantee. These may reduce the scrutiny in the market and reduce market self-discipline. (Acharya, Phillipon et al. (2009)). Furthermore, the increase in use of derivative instruments and the technological developments in the financial sector, such as the move to non-interest income, increased the overall complexity of BHCs and reduced their transparency. Thus it became harder to monitor them. (Rajan (2006), Acharya, Phillipon et al. (2009), Kashyap, Rajan and Stein (2008).

When looking at risk management from a holistic perspective, its main objective can be seen as to maximize the firms value via the reduction of costs associated with different risks (Dionne, 2013). Or, in other words, to preserve firm value by avoiding high losses. How effective information, both qualitative and quantitative, flows within an organization, to accomplish the above, depends on its structure (Stein 2002). But even when information is allowed to flow, it is also essential that the risk management function is independent and strong within the organization (Kashyap, Rajan, and Stein (2008)).

(8)

the risk and the strength of the risk management, such that a conservative risk culture leads to the decision to take smaller risks and at the same time operate a strong risk management system. (Ellul and Yerramilli (2013)). Fahlenbrach et al. (2012) show that the stock performance of banks during the financial crisis of 1998 predict the stock performance in the 2008 crisis. This is consistent with the idea of a risk culture, in that banks do not alter their business model. Second, an alternative channel is referred to as the hedging channel. It states that banks actively decide on both their degree of risk exposure and level of risk management (Ellul and Yerramilli, 2013). Looking at the way that banks reacted on a crisis is a way of distinguishing between the channels, since they will respond differently. The business model channel predicts that banks do not or slowly adapt while the opposite goes for the hedging channel.

In line with Ellul and Yerramilli (2013), the alternative hypothesis is that the risk management function does not show any real impact on a BHCs tail risk. This may be due to the fact that the risk manager does not have a lot of influence on the CEO or is curtailed in the CEOs search for growth and thus is not able to remain independent. Another reason may be that the internal controls are not able to keep up with the developments on the trading floor.

With regards to previous research, Ellul and Yerramilli (2013) introduce and construct a Risk Management Index (RMI) with principal component analysis. This RMI is based on 7 proxies to measure the strength and independence of the risk management function within a BHC. They relate this to the degree of tail risk, as measured by the expected shortfall of the BHC’s stock, and stock return performance. Their research focuses on the United States and finds that BHCs with a higher RMI have lower tail risk and a higher return on assets, all else being equal.

Another study that uses a similar approach, but utilizes dummy variables in a panel data setting, is by Aebi et al. (2011). They collect data on 372 U.S. banks with regards to the position of the CRO and corporate governance mechanisms. They find that risk management related corporate governance mechanisms, such as the presence of a CRO in an executive board are associated with better bank performance during the financial crisis of 2007/2008.

(9)

There are other factors that have found to be of influence on tail risk . First, the earnings management that the bank utilizes within a crisis, although there is little evidence of the effect from earnings management from before a crisis (Cohen et al (2014)) as shown for a sample of US BHCs. Second, the way that expansion into non-interest income and non-deposit, wholesale funding create a higher vulnerability to systemic risk, at high levels of it. (Demirgüç-Kunt and Huizinga, (2010)). This in turn also influences the BHC tail risk. This is also proven by De Jonghe (2010) who uses tail beta to measure tail risk and finds that a shift towards non-traditional banking activities increases it. Third, a very important factor influencing tail risk is size, which is shown to be a key determent of a BHCs risk exposure (Keasy and Vallascas (2012), De Jonghe et al. (2014).

III. Data

The data for this thesis is hand collected from several sources. Given the effort involved I have limited the sample to include the 50 publicly listed BHC’s with the highest amount of book value of total assets in the EU-15 zone on .1 The following sections first describe the collection of data and the creation of the sample. Second, I describe the creation and collection of the data for the RMI. Third, I elaborate on the methodology used, detailing the PCA and the separate multi variable regressions.

A. Data collection

Based on Bureau Van Dijks Bankscope, data is collected on the amount of book value of total assets. For availability of the information of stock prices a limitation is made to publicly listed BHCs that have data available for the whole 2004-2013 period. A total of 227 BHCs were found with a combined book value of total assets of 24.88 trillion Euros.2 The top 50 banks account for 23.49 trillion Euros, or 94.41% of book value total assets at publicly listed bank holding companies. Within this sample of 50 a further elimination is made because of a full ownership relation between BHCs: a holding group and its subsidiary. Furthermore the central banks of Greece and Belgium were found within the sample of 50 BHCs. This leaves a sample of 46 banks with a combined book value of total assets of 23.22 trillion Euros, or 94.35% for the period 2004-2013.

The data on the presence of a CRO, his position within the BHC and information on the remuneration of the CEO, CRO and other executives and the installation of new CEOs is hand-collected from annual reports, corporate governance reports, fiscal reports and Basel Pillar III reports. These reports are

1_{The EU-15 zone is based on 2004, see appendix ….} 2

(10)

downloaded from the respective websites of BHCs, or their information page on

www.morningstar.com. For BHCs that are active within the United States the Edgar system was used. With regards to the construction of the Risk Management Index the difficulty was detecting the executives remuneration, especially in the earlier years of the sample. In total a RMI score was created for 393 year-BHC combinations, leaving an unbalanced panel. This is also due to the fact that two Spanish BHCs were created in 2009, Caixa Bank and Bankia. Appendix A shows a list of the included banks, their home country and their book value of total assets used for selection.

Information on balance sheet data is from Bureau Van Dijks Bankscope, collecting information on total assets, return on average assets, amount of gross loans, Tier 1 ratio, Non Interest Income / Gross Income, Impaired Loans (NPLs) / Gross Loans, Trading Securities at Fair Value through income, Held to Maturity Securities and Total Deposits, Money Market and Short-term Funding. The information is presented on a calendar-year basis and in thousands of Euros. A precise definition of the variables is to be found in appendix B1. The data on stock returns to calculate the expected shortfall, buy-and-hold-returns and excess returns are from Reuters Datastream. A total return index is used to incorporate dividends. The data is presented on a daily basis.

A special note must be made concerning the Tail Risk and the both Return variables when it comes to outliers. The sample contains several banks who experienced a severe financial distress situation. Amongst these are Dexia Bank (Belgium), National Bank of Greece S.A. and Pireaus Bank S.A (Greece). The Portuguese Bank holding company Espirito Santo Financial Group S.A. declared bankruptcy in 2014 and does no longer exist.

Tail risk is modeled by the Expected Shortfall measure. It is used to capture cumulative expected loss conditional on returns being less than some α-quintile. The absolute of this measure is taken, so that a higher tail risk means a larger the amount of expected loss (Acharya et al.(2010)). For an overview of the Expected Shortfall measure and different estimation methods an oversight is provided by Nadarajah et al. (2014).

B. The Risk Management index

(11)

RMI: CRO present, CRO executive, CRO top 5 paid and CRO centrality. This means a deviation from the RMI that was constructed in the work of Ellul and Yerramili (2013).

For CRO present a dummy variable is created. Due to the fact that not every BHC utilizes the same titles, similar titles are treated on as representing the same role.3 The CRO is the official that is exclusively charged with managing enterprise risk across all business segments of the BHC. CFO/CRO’s are thus not treated on the same level. The dummy variable CRO Executives identifies whether the CRO is part of executive management team (or an equivalent level, being the highest operating committee within the organization). Next to the fact that these two dummies capture the importance of the CRO, the executive dummy is, as found by Aebi et al. (2011), also an indication for the CRO’s independence, giving him direct access to the board of directors. The third variable is whether the CRO is within the top five of highest paid executives. The fourth is CRO centrality. This is defined as the ratio of the CRO’s compensation, excluding options and stock rewards, relative to the CEO’s compensation, excluding options and stock rewards. The idea of CRO centrality is based on Bebchuk et al (2011) who use a similar approach, CEO centrality, to investigate its relation with value, performance and behavior of public firms. The reason to exclude variable rewards is that CROs with a high proportion of variable compensation will have a conflicting incentive regarding their task (Ellul and Yerramilli (2013)).

As already mentioned, there are several difficulties with regards to collecting the data on compensation. First, the first year of the sample, 2004, contained several BHCs which did not compile their reports according to IFRS. Second, transparency is a factor that did not gain importance until after the crisis, although there is a difference between Northern European and Southern European countries, with the former divulging more details. Following Ellul and Yerramilli (2013), to deal with the case of a BHC not reporting having a CRO, the CFOs compensation is taken instead and a percentage point is deducted from the CRO centrality ratio. When the CRO is present but not among the 5 highest paid executives, again a percentage point is deducted. In the case that only the CEOs compensation is separately reported and the remainder of the executive teams compensation is reported together, an average compensation is calculated and two percentage points are deducted from the ratio. With regards to the compensation in a different currency than the Euro no problems should be expected, even though the compensation is calculated in Euros using the average exchange rate for the respective year. The RMI is created by taking the first principle component of the four variables above.

3_{Amongst others, Chief Credit Officer, Group Credit Officer, Vice President of Risk are titles that are used by}

(12)

IV. Methodology

This section will discuss the techniques used to come to the risk management index (RMI), the multivariate analysis of the determinants of RMI and the multivariate analysis of a high RMI on tail risk and of high RMI on different measurements of BHC performance.

A. RMI Index

The RMI Index is based on the four variables introduced earlier: CRO present, CRO executive, CRO top 5 paid and CRO centrality. Of these four variables the first principal component is taken, based on Principal Component Analysis (PCA). PCA is a mathematical factor model that translates k explanatory variables in to k uncorrelated new variables. In choosing the appropriate number of principal components Malhotra (2009) shows first that only components with an eigenvalue greater than 1 are of interest since they capture more information. Second, he shows that the chosen component should explain at least 60% of the total variance. With regards to the eigenvalue of the components, only those with a score of above 1 are of interest, containing more information. For the years 2012 and 2013 the first principal component shows a score lower than 1. This is due to the fact that CRO centrality has high loading on the second component. A combination of the two components is used for those years. The main advantage of using principal component analysis is that there are no subjective measurements necessary. As suggested by Tetlock (2007), the PCA is constructed on a year by year basis to avoid a possible look-ahead bias by using information from the future. The separate loading factors and their eigenvalues are shown in appendix C for every year and for the full sample.

B. Multivariate analysis

Determinants of RMI

To examine the determinants of RMI panel regressions of the following form are estimated:

(1)

(13)

Because the data for total assets is highly skewed, as can be seen in the section descriptive statistics, Size² is included to control for a possible nonlinear relationship.

Further control variables are Return on Average Assets to control for the BHC’s profitability and Buy-and-Hold return for BHC’s past performance. Controls for the balance sheet compromise Total

Deposits to Total Assets, Tier 1 Ratio, Total Loans to Assets and Impaired Loans to Gross loans. To

control for the quality of the loan portfolio the ratio Impaired loans / Gross Loans is used and Non-interest Income / Gross Revenue serves as a proxy for off balance sheet activity (Boyd and Gertler (1994)). Controls for the usage of derivatives is done through Derivative Hedging / Assets and

Derivative trading / Assets. Finally two dummies are used to identify a change in CEO in or a

large merger or acquisition in , being a year on year increase of more than 25% of total

assets.

A separate analysis by the form of equation (1) is being made for the period 2010-2013. This is because of the big increase of the median value of RMI after the crisis. Related to this a separate panel regression is done to estimate the change in RMI level after the crisis. This is being done with a first difference specification. The dependent variable in this regression is the first difference of RMI). The control variables are their respective first differences.

(2)

RMI and Tail Risk.

This section explains the panel regression used to investigates whether BHCs with a high RMI, meaning a strong and independent risk management function, had lower tail risk. In these regression controls are being used to capture the underlying risk of the BHC’s business activities. The panel regressions are of the form

(3)

In the above equation, j denotes the BHC and t denotes the year. The dependent variable is Tail Risk and the main independent variable is the BHC’s lagged RMI. The control variables are denoted by

, being lagged BHC characteristics with a possible influence on risk. Definitions of all variables

(14)

of a large merger or takeover. As described earlier, a higher RMI should lead to Tail Risk, all else being equal. Using equation (3) the following Hypotheses are test:

H0: RMI has no effect on Tail Risk.

H1: RMI has a negative effect on Tail Risk. RMI and Firm Performance

For the section describing the relationship between RMI and Firm performance equation (5) is being used with as dependent variable Return on Average Assets, Buy-and-Hold Return and Excess Return. Buy-and-Hold return is the natural logarithm of the return of a BHC stock in 1 year. The excess returns of BHC’s stocks are calculated over a market index, the Eurostoxx Banks Index. This index composes 32 bank stocks from the EU zone. The calculations are based on the classic Capital Asset Pricing (CAPM) model (Sharpe (1964), Lintner (1965)).

(4)

The β in equation (4) is calculated by dividing the covariance of the BHC stock with the market index by the variance of market index. The risk free rate used is the yearly average of the 10 year German bund. The used risk free rates can be found in appendix D. The full set of control variables used in equation (3) is being used.

The relationship between RMI and stock return performance is less clear-cut, depending on whether the risk function lowers the BHC’s systematic risk, idiosyncratic risk, or both. The modern portfolio theory predicts that investors hold well-diversified portfolios (Markowitz, 1952). Expected returns should only be based on systemic risk and not on idiosyncratic risk. If the reduction in risk due to a high RMI is with regards to idiosyncratic , there should not be a relationship between RMI and excess returns. Next to using the annual buy-and-hold return, the excess return over the market is calculated, based on equation (4). In this case the total return of the Eurostoxx Banks index is used.4 The results should be interpreted with cause, because the sample period is too short and the sample too small to give a precise answer regarding the nature of the risk. To estimate the relation between RMI and performance panel regressions of the following form are used.

(5)

Besides testing for Excess Returns, tests with ROAA and Buy-and-hold-Return as the dependent variable done. The following hypotheses are tested with equation (5).

4

(15)

H0: There is no relationship between RMI and Excess returns.

H1: There is a negative relationship between RMI and Excess Returns. Robustness Tests

First, the principal component analysis is, as stated above, carried out with yearly data to avoid a forward looking bias by using future data. An additional test has however been done with a alternative RMI which is constructed with a full sample PCA. This has been used

in equation (3) and replaces the RMI index which was used. The dependent variable in this panel regression is Tail Risk.

Second, to control for effects caused by the sovereign debt crisis which hit Europe in 2010 and 2011 a dummy variable is constructed which identifies banks from Portugal, Italy, Ireland, Greece and Spain. The panel regression in which this dummy is used is (2) with Tail Risk as dependent variable.

V. Results and Discussion

In the following section first the descriptive statistics of key variables will be presented. Second, a univariate analysis will be carried out. Third, the PCA is presented. This section concludes with the results from the multivariate analysis and robustness checks.

Descriptive Statistics

(16)

The mean value of Tier 1 Ratio is 0.097 or 9.7% with a minimum of below zero in case of the Greek banks during the sovereign debt crisis. The Total Deposits / Total Assets range between 1.5% and 98.8% shows that the sample has BHCs that are pure retail and others that focus on anything but retail banking. Finally, in 14.3% of the cases there has been a change in CEO and in 12.4% a merger or acquisition, of which almost 80% took place before 2007.

In table two the mean and median values of RMI are presented together with the mean values of the individual components. There is an gradual increase of mean RMI over the years until 2013 where it shows a decline, a movement that is noticeable across all components. This decrease may be explained by a softening of the debt crisis in the Eurozone.5 A steep increase in median RMI is seen from 2008 to 2009, following the credit crisis. In 2011, in the middle of the sovereign debt crisis, 84.8% of all BHCs had a dedicated CRO, a more than 100% increase from 2004. The average CRO centrality in this period went up by 22.6%.

Correlations among key variables.

In table 3 the pair-wise correlations between BHC’s Tail Risk, RMI, Size, Buy and Hold Return and the BHC’s other characteristics are presented. Tail Risk is negatively correlated to ROAA and buy and hold Return, indicating that more profitable BHC’s are less risky. Size shows a positive sign relating it to Tail Risk, indicating that bigger BHC have higher tail risk. Tier 1 Ratio is as expected negatively related and the quality of the loan portfolio, as indicated by Impaired Loans / Gross Loans is positively related, consistent with the idea that BHC’s with a healthy loan portfolio are less risky. Further, a change in CEO is positively related with higher risk.

Buy and hold Return shows, apart from Total Deposits / assets no surprising correlations. The fact that Total Deposits / Assets is negatively related to Buy and hold Returns is interesting when compared to the positive relation between Non Interest Income / Gross Revenue and Buy and hold Return, giving an argument for the fact that BHC’s with a larger share from non interest activities are more profitable.

Present RMI score is negatively related to tail risk and highly significant. To avoid this problem in the multivariate analysis the of the BHC is used. Furthermore it becomes clear that it is also

positively related to both Buy and hold Return and Size, indicating that bigger and in the same year better performing BHC’s have higher RMI scores,

5_{On july 26}th_{2012 Mario Draghi, president of the European Central Bank, delivered his ‘ Whatever it takes ‘}

(17)

Table 1

Summary statistics key variables

This panel presents descriptive statistics for the key variables used in the analysis in this research. All variables are defined in Appendix B Summary statistics (Entire Panel)

Tail Risk Excess

Return Buy and hold Return Total Assets in mln of Euros

Size ROAA Tier 1

(18)

Table 2

Summary statistics RMI and RMI components

Panel B presents a year wise distribution of the mean and median values of RMI and its components across all BHCs Year Mean RMI Median RMI Mean CRO Present Mean CRO Executive Mean CRO Centrality Mean CRO Top5 2004 0.426 0.127 0.370 0.196 0.479 0.087 2005 0.518 0.148 0.457 0.261 0.469 0.087 2006 0.754 0.759 0.587 0.413 0.433 0.152 2007 0.790 0.737 0.609 0.435 0.462 0.174 2008 0.837 0.665 0.674 0.435 0.533 0.239 2009 0.957 1.191 0.739 0.543 0.535 0.304 2010 1.033 1.262 0.783 0.565 0.541 0.261 2011 1.081 1.246 0.848 0.630 0.564 0.239 2012 1.287 1.214 0.804 0.652 0.590 0.283 2013 1.199 1.239 0.783 0.630 0.587 0.283 2004-2013 0.888 0.767 0.665 0.476 0.523 0.211

although ROAA shows a different sign than Buy and hold Return and is not significant. It is important to caution for over interpreting the results from table 3 since they are simply pair wise correlations that do not control for any characteristics of the BHC, especially size.

Multivariate Analysis

The following section is dedicated to the multivariate analysis of RMI and Tail Risk. The respective equations can be found in the methodology section.

Influences on RMI

The significant negative coefficient on Expected shortfall t-1 shows that, when controlling for variance over time, firms with higher tail risk in the previous year have a lower RMI, showing a sign for a negative relationship between tail risk and RMI. Moreover, the negative coefficient on Impaired loans/Gross Loans t-1 shows that, consistent with the Tail Risk t-1 coefficient, that BHCs who suffered higher losses on their loan portfolio have a lower RMI. The positive coefficient on Hedging

(19)

Table III

Correlations among key variables

This table presents pair-wise correlations between Tail Risk, Buy and hold Return, Size, RMI Index and other important BHC characteristics. Variable definitions are in Appendix B. * Denotes significance at the 5% level, ** denotes significance at the 1% level.

Tail Risk Buy and hold Return Size RMI Index

Tail Risk

Buy and hold Return -0.021

Size 0.148* 0.184**

RMI Index 0.215** 0.208** 0.238**

ROAA -0.500** 0.027 -0.069 -0.064

Tier 1 Ratio -0.133* 0.249** 0.179** 0.296

Impaired Loans / Gross Loans 0.378** 0.001 -0.071 -0.028

Deriv. Hedging / Assets -0.175** 0.132* 0.461** -0.013

Deriv. Hedging / Assets 0.100 -0.032 -0.162** 0.015

Total Deposits / Assets 0.070 -0.174** -0.406** -0.036

Total Loans / Assets -0.070 -0.153** -0.675** -0.119

Non Interest Income / Gross

Revenues -0.388** 0.161** 0.197** -0.027

Change in CEO 0.223** -0.029 -0.043 0.085

Large M&A -0.084 -0.007 -0.226** -0.078

(20)

Table IV

This table presents the results of panel regressions that examine the relationship between BHC's RMI and their financial characteristics. All variables are defined in Appendix B. The regression estimated is

The regression is estimated on a panel data that has one observation for each BHC-year combination and spans the 2005-2013 period for column (1) and (2) and 2010-2013 for column (3) and (4). In panel A Column (1) is with year fixed effects, column (2) repeats the regression in column (1) with additional BHC fixed effects. The columns (3) and (4) repeat the regression, but for the 2010-2013 period with respectively year fixed effects and year and BHC fixed effects. In panel B column (5) the equation in column (3) is repeated with a dummy variable indicating high tail risk in 2008. Column 6 contains a first difference regression. Redundant fixed effects tests have been conducted for all columns, with all values significant. Standard errors (reported in parentheses) are robust to heteroskedasticity. For statistical significance ***, ** and * indicate significance on the 1%, 5% and 10% level.

Panel A: RMI and BHC Characteristics (1) (2) (3) (4) 0.149 -0.449 0.389 -2.067 (0.504) (0.439) (1.996) (7.691) 0.001 -0.016 -0.004 0.043 (0.013) (0.010) (0.051) (0.201) _0.093 _0.013 _0.073 _-0.116 (0.121) (0.088) (0.139) (0.108) -4.635*** -1.483 -3.921** 0.815 (1.759) (1.412) (1.943) (2.130) _0.249 _0.556 _-0.771 _1.314 (0.377) (0.621) (0.522) (1.062) 0.053** 0.006 0.065** -0.009 (0.024) (0.021) (0.026) (0.028) _0.979** _1.055 _1.283** _2.647* (0.467) (0.822) (0.615) (1.473) _-0.064*** _-0.028** _-0.057*** _-0.012 (0.014) (0.013) (0.015) (0.017) -0.002 0.000 -0.001 -0.008** (0.003) (0.002) (0.004) (0.004) _2.319* _0.713 _5.044*** _0.545 (1.164) (1.539) (1.480) (3.011) _0.686 _2.904*** _-0.084 _-0.661 (0.901) (0.819) (1.661) (2.085) _-0.078 _-0.140* _-0.210 _-0.258*** (0.106) (0.072) (0.142) (0.093) _-0.192 _-0.072 _-0.105 _0.054 (0.214) (0.157) (0.250) (0.208) -0.043 0.037 -0.094 0.054 (0.050) (0.039) (0.055) (0.045) Constant -3.382 14.463** -5.759 (23.384 (5.154) (6.230) (19.731) (73.517) Observations 246 246 39 39 R² 0.298 0.781 0.365 0.862 Adjusted R² 0.229 0.708 0.270 0.763

Year FE Yes Yes Yes Yes

(21)

Table IV- continued

Panel B: Performance during 2008 and subsequent RMI

(5) (6)

RMI RMI

High Tail Risk 2008 0.276** High Tail Risk 2008 -0.115

(0.130) (0.096)

-0.980 Size 11.192

(2.069) (8.476)

0.031 Size ² -0.310

(0.052) (0.219)

_0.016 _{Buy and hold Return} _0.016

(0.139) (0.093)

-2.537 Tail Risk 4.793**

(2.021) (2.087)

_-0.864* Total Deposits / Assets -0.767

(0.516) (1.191)

0.063** Tier 1 Ratio

(0.026)

_1.738*** Total Loans / Assets 0.840

(0.643) (1.694)

_-0.058*** Impaired Loans / Gross Loans

(0.015)

_-0.003 Nonint.Income / Gross Revenues -0.004

(0.004) (0.004) _4.736*** _{Deriv.Hedging/Assets} _-5.879* (1.465) (3.401) _0.171 _{Deriv.Trading/Assets} _-0.935 (1.640) (1.406) _-0.198 _{Change in CEO} _0.149* (0.140) (0.083) _-0.045 _{Large M&A} _0.011 (0.248) (0.201) -0.090 ROAA -0.005 (0.054) (0.024) Constant 7.539 Constant 0.158** (20.416) (0.065) Observations 132 Observations 97 R² 0.389 R² 0.256 Adjusted R² 0.292 Adjusted R² 0.118

Year FE Yes Year FE Yes

(22)

RMI after the crisis and changes in RMI

As noted earlier, a sharp increase in the median value of the RMI can be noted after 2008. In 2008 the subprime crisis reached its peak, which may have caused a sharper focus on risk management within BHCs. In column (3) a separate analysis is carried out for the sample from 2009-2013. The significant variables in column (1), Tail Risk t-1, Tier-1-Ratio t-1, Total Loans/Total Assets t-1, Impaired Loans /Gross Loans t-1 and Derivative Hedging /Total Assets t-1 all remain significant and are showing similar signs. When controlling for BHC cross-section fixed effects, Total Loans/Total Assets t-1 is the only variable remaining significant as can be seen in column (4) when compared to column (3). Moreover, the dummy Variable CEO-Change t-1 and the variable non-interest income / gross revenue t-1 become significant. The negative sign on CEO-Change t-1 may be due to a renewed focus on growth instead of risk management after the crisis. The negative sign on non-interest income / gross revenue t-1 may be interpreted as a signal that BHCs came back to activities like investment banking and derivative trading after the crisis.

In column (5) a dummy variable is included to distinguish between the hedging channel and the business model channel. The dummy is defined as High Tail Risk 2008, being the year after which the increase in RMI can be seen. It identifies those BHCs whose tail risk exceeded the median value across all BHCs during that year. The positive coefficient on this dummy indicates that BHCs with the worst performance during the crisis, had a higher RMI subsequent years compared with other BHCs. This is consistent with the hedging model channel. In this regression only year fixed effects are included due to the limited sample. This may be the cause of finding an opposite sign on the dummy variable when compared to the research by Ellul and Yerramilli (2012) since variation within BHC’s is not accounted for.

When comparing column (5) to column (3) the coefficient on Tail Risk reduces in size and loses its significance. Tier 1 Ratio t-1, Impaired Loans/Total Assets t-1, Total Loans/ Total Assets t-1 and Derivative Hedging / Total Assets t-1 remain significant although their coefficients are reduced in size. Finally, the variable Total Deposits/Total Assets t-1 becomes significant at a 10% level showing that an increase in deposits leads to a higher RMI.

(23)

variable trap. The hedging channel predicts a positive sign on the coefficient for the 2008 High Tail Risk dummy and the business model the opposite. As can be seen from column (6), the sign is negative. This supports the business model theory, which contradicts column (5). However, in this case the coefficient is not significant. Once more, limitations must be made due to the relative low number of observations, which also does not allow any cross sectional fixed effects next to the year fixed effects.

The influence of RMI on tail risk.

In this section the RMI will be the independent variable and the objective is to see if BHCs that have a higher RMI, so a stronger position for the CRO, have indeed lower tail risk, controlled for amongst

others size and underlying risk of the business activities. Since Expected Shortfall is

measured in percentages (losses), a positive coefficient shows an increase in Tail Risk.

The regression is carried out with year fixed effects in column (1) and both year and cross-sectional effects in column (2). Lagged RMI Index score shows a positive coefficient, albeit a very small one, and is not significant in either column (1) or (2). The positive sign would suggest that an increase in RMI would lead to higher Tail Risk (as in higher losses). This is a contradiction with the findings of Ellul and Yerramilli (2008).

The variable (Total Loans/Total Assets) in column (1) indicates that BHCs with a larger fraction of their balance sheet dedicated to loans (BHCs which are focused on retail banking) have lower tail risk, although this effect loses significance when controlled for cross-sectional effects.

The M&A Dummy is significant for both column (1) as (2), indicating that BHCs who had an increase of over 25% in total assets in a year time enjoy lower tail risk. The effect becomes even larger when controlled for cross-sectional effects with a higher significance. A note must be made that almost 80% of the M&A dummies identified in this sample was before 2008, during the ongoing

consolidation of the European banking market (Goddard et al (2007)).

The variables Size and Size² become significant in column (2), however the signs are not similar. The negative coefficient on Size shows that larger BHCs enjoy a smaller tail risk. The positive sign on Size² does give an indication that the largest banks have significantly higher tail risk. This would be

(24)

Table V

Relationship between Tail Risk and RMI This table reports the results of panel regressions that examine the relationship between BHC's tail risk and RMI. The regression estimated is

The regression is estimated on a panel data that has one observation for each BHC-year combination and spans the 2005-2013 period for column (1) and (2). Column (1) is with year fixed effects, column (2) repeats the regression in column (1) with additional BHC fixed effects. Redundant fixed effects tests have been conducted for all columns, with all values significant. Standard errors (reported in parentheses) are robust to

heteroskedasticity. For statistical significance ***, ** and * indicate significance on the 1%, 5% and 10% level.

(1) (2) 0.003 0.004 (0.004) (0.005) -0.020 -0.082*** (0.032) (0.031) 0.001 0.002*** (0.001) (0.001) _-0.008 _0.002 (0.007) (0.006) _0.006 _-0.035 (0.024) (0.037) 0.000 0.000 (0.001) (0.001) _-0.064** _-0.036 (0.029) (0.054) _0.002*** _0.003*** (0.001) (0.001) _0.000 _0.000 (0.000) (0.000) _0.094 _0.295*** (0.074) (0.099) _-0.113** _-0.040 (0.055) (0.060) _-0.011* _-0.012** (0.007) (0.006) _-0.023* _-0.036*** (0.013) (0.013) 0.003 0.003 (0.003) (0.003) Constant 0.289 0.910** (0.322) (0.353) Observations 243 243 R² 0.453 0.711 Adjusted R² 0.396 0.614

Year FE Yes Yes

BHC FE No Yes

(25)

The variable Derivatives Trading / Total Assets is significant in column (1) and has a negative effect on tail risk, indicating that firms with a higher amount of derivatives held for trading would have a lower tail risk when controlled for time fixed effects. This effect disappears in column (2) when using cross sectional controls. The effect is also economically significant which would indicate that BHCs with larger derivatives desks enjoy lower tail risk.

The opposite is true for Derivatives Hedging / Total Assets which is not significant in column (1) but is in column (2). Furthermore, it shows a positive coefficient, indicating an increase in tail risk. The coefficients on CEO Changet-1 are negative and significant in both columns, suggesting that a change in regime can reduce the tail risk for BHCs. In the sample it becomes clear that many BHCs changed CEO during or directly after the crisis with 16 changes in 2008 and 2009. Finally, (Impaired Loans/Gross Loans) shows significant positive effect and leads to higher tail risk in both column (1) and (2), although the effect is small.

The influence of RMI on BHC performance

So far it has become clear that the relation between RMI and Tail Risk in this research is ambiguous. The following estimated regressions focus on the relationship between RMI and BHCs performance. If RMI allows for a more effective management of the risks the BHC is facing, a higher operating performance as measured by Return on Average Assets (ROAA).

(26)

Table VI

Relationship between RMI and Performance

This table reports the results of panel regressions that examine the relationship between BHC's tail risk and RMI. The regression estimated is

The regression is estimated on a panel data that has one observation for each BHC-year combination and spans the 2004-2013 period for all columns. The columns (1), (3) and (5) have year fixed effects. Columns (2), (4) and (6) have both year and BHC fixed effects. Redundant fixed effects tests have been conducted for all columns, with all values significant. Standard errors (reported in parantheses) are robust to heteroskedasticity. For statistical significance ***, ** and * indicate significance on the 1%, 5% and 10% level.

(1) (2) (3) (4) (5) (6) 0.003 -0.010 0.050 0.082 0.099*** 0.155*** (0.072) (0.104) (0.041) (0.059) (0.033) (0.049) 0.357*** 0.260* (0.102) (0.100) -0.006 -0.039 0.126*** 0.136*** (0.028) (0.031) (0.022) (0.025) 0.628* -0.464 -0.272 0.188 -0.098 0.293 (0.380) (0.623) (0.210) (0.366) (0.170) (0.298) 0.000 0.008 0.015 0.028* -0.007 -0.016 (0.020) (0.021) (0.013) (0.016) (0.010) (0.013) 0.782** 1.918** -0.064 -0.027 -0.269* -0.129 (0.338) (0.805) (0.193) (0.462) (0.155) (0.371) -0.020* -0.008 -0.008 -0.003 -0.017*** -0.030*** (0.011) (0.013) (0.008) (0.010) (0.006) (0.008) 0.003 0.002 0.001 0.000 0.000 -0.001 (0.002) (0.002) (0.001) (0.001) (0.001) (0.001) Constant -0.352 -0.453 -0.145 -0.552* -0.018 -0.316 (0.354) (0.534) (0.207) (0.320) (0.166) (0.259) Observations 303 303 328 328 324 324 R² 0.330 0.527 0.510 0.608 0.250 0.387 Adjusted R² 0.293 0.409 0.485 0.520 0.210 0.247

Year FE Yes Yes Yes Yes Yes Yes

(27)

Robustness tests

Two robustness tests are carried out. First, the dummy variable for the PIIGS countries. Second, an alternative RMI is being used, with the principal components based upon the entire sample period instead of the year by year calculations. The results can be found in table VII.

The Euro debt crisis left its marks on Europe with bail-outs of banks in Greece, Spain, Portugal and Ireland. Since these countries, together with Italy, suffered the most from this crisis, an additional test is created with a dummy representing the banks from the PIIGS-countries (Portugal, Ireland, Italy, Greece, Spain). The dummy shows a negative sign in the year fixed effects regression, indicating a higher tail risk, all else equal. It becomes positive when using cross-section controls, albeit the coefficients are not significant in both cases. Overall there is no prove for a higher tail risk for PIIGS countries with a lower RMI score. Moreover, the results are similar to the earlier regression in table V.

With regards to the alternative RMI score, using a PCA factorloading based on all years within the sample, it shows a similar sign as the regular RMI score with a smaller coefficient. As in the earlier tests shown above in table V, the RMI score is not significant, showing no proof for an effect on the tail risk.

VI. Conclusion

This study focuses on the relation between the strength and independence of the risk management function of BHCs in the 15 countries in the EU for the period of 2004-2013 and investigates its influence on the tail risk of these BHCs.

(28)

Table VII

Robustness Tests

This table reports the additional regressions that are used as robustness tests. Columns (1) and (2) include the dummy variable for the PIIGS countries. Column (3) utilizes an alternative RMI index score. The following regression formula is used.

The regressions in column (1) and (2) are based on a panel data having a BHC-Year observation for each of the years 2004-2013. Column (1) includes year fixed effects, columns (2) and (3) both year and BHC fixed effects. Redundant fixed

effects tests have been conducted for all columns, with all values significant. Standard errors (reported in parentheses) are robust to heteroskedasticity. For statistical significance ***, ** and * indicate significance on the 1%, 5% and 10%

level. (1) (2) (3) 0.003 0.004 (0.004) (0.005) -0.002 (0.006) -0.020 -0.085*** 0.094*** (0.032) (0.031) (0.032) 0.001 0.002*** -0.003*** (0.001) (0.001) (0.001) _-0.008 _0.002 _-0.002 (0.007) (0.006) (0.006) 0.006 -0.024 0.020 (0.024) (0.037) (0.039) -0.000 0.000 -0.001 (0.001) (0.001) (0.001) -0.065** -0.029 0.019 (0.029) (0.054) (0.056) 0.002 0.003*** -0.002*** (0.001) (0.001) (0.001) 0.000 0.000 0.000 (0.000) (0.000) (0.000) 0.096 0.292*** -0.280*** (0.075) (0.098) (0.100) -0.112 -0.033 0.043 (0.056) (0.060) (0.062) -0.011* -0.012** 0.012** (0.007) (0.006) (0.006) -0.023* -0.042*** 0.040*** (0.014) (0.013) (0.013) 0.003 0.002 -0.002 (0.003) (0.003) (0.003) PIIGS Countries 0.001 -0.039 (0.007) (0.026) C 0.285 0.927** -0.938*** (0.323) (0.351) (0.353) Observations 243 243 226 R² 0.453 0.715 0.729 Adjusted R² 0.393 0.616 0.630

Year FE Yes Yes Yes

(29)

However, this result has its limitations for interpretation, because of the fact that there is not accounted for within firm variance, because of a limited sample. To separate the change factor a first difference method is applied which finds no evidence for the business model channel. Overall no definite conclusion can be made with regard to the change in RMI after the crisis and whether firms are following a hedging or a business model. This fails to support the findings in Ellul and Yerramilli (2013) and Fahlenbrach et al. (2012) who find evidence for the business model channel.

Further it is shown that there is no clear relationship between RMI and Tail Risk for European BHCs, even when controlled for the banks in the PIIGS countries and within BHC variation. Other factors that do influence Tail Risk are a change in CEO, Size and M&A events, suggesting that BHCs who are on average larger, or changed their CEO or were involved in a noticeable size increase in the previous year experienced lower tail risk, all else being equal. The size effect is consistent with the findings of Ellul and Yerramilli (2013) for BHCs in the United States. RMI does however have an influence on the excess returns of BHCs over a market index containing peers. This suggest that a higher RMI can have a positive influence on higher returns when compared to peers. There is no significant relation found when related to other profitable measurements.

(30)

VII. Literature

Acharya, V. V., Pedersen, L. H., Philippon, T., & Richardson, M. P. (2010). Measuring systemic risk.

Acharya, V., Philippon, T., Richardson, M., & Roubini, N. (2009). The financial crisis of 2007‐2009: Causes and remedies. Financial markets, institutions & instruments, 18(2), 89-137.

Aebi, V., Sabato, G., & Schmid, M. (2012). Risk management, corporate governance, and bank performance in the financial crisis. Journal of Banking & Finance, 36(12), 3213-3226.

Bebchuk, L. A., Cremers, K. J., & Peyer, U. C. (2011). The CEO pay slice. Journal of Financial Economics,

102(1), 199-221.

Beltratti, A., & Stulz, R. M. (2012). The credit crisis around the globe: Why did some banks perform better?. Journal of Financial Economics, 105(1), 1-17.

Cohen, L. J., Cornett, M. M., Marcus, A. J., & Tehranian, H. (2014). Bank earnings management and tail risk during the financial crisis. Journal of Money, Credit and Banking, 46(1), 171-197.

De Jonghe, O. (2010). Back to the basics in banking? a micro-analysis of banking system stability.

Journal of financial intermediation, 19(3), 387-417.

De Jonghe, O., Diepstraten, M., & Schepens, G. (2014). Banks' size, scope and systemic risk: What role for conflicts of interest?. Available at SSRN.

Demirgüç-Kunt, A., & Huizinga, H. (2010). Bank activity and funding strategies: The impact on risk and returns. Journal of Financial Economics, 98(3), 626-650.

Dionne, G. (2013). Risk Management: History, Definition, and Critique. Risk Management and

Insurance Review, 16(2), 147-166.

Eken, M. H., Selimler, H., Kale, S., & Ulusoy, V. (2012). The Effects of Global Financial Crisis on the Behaviour of European Banks: A Risk and Profitability Analysis Approach. ACRN Journal of Finance

(31)

Ellul, A., & Yerramilli, V. (2013). Stronger risk controls, lower risk: Evidence from US bank holding companies. The Journal of Finance, 68(5), 1757-1803.

EY. (2013). Remaking financial services: risk management five year after the crisis Available at

http://www.ey.com/GL/en/SearchResults?query=remaking+financial+services%3A+risk+managemen t+five+years+after+the+crisis&search_options=country_name

Fahlenbrach, R., Prilmeier, R., & Stulz, R. M. (2012). This time is the same: Using bank performance in 1998 to explain bank performance during the recent financial crisis. The Journal of Finance, 67(6), 2139-2185.

Goddard, J., Molyneux, P., Wilson, J. O., & Tavakoli, M. (2007). European banking: An overview.

Journal of Banking & Finance, 31(7), 1911-1935.

Laeven, L., & Levine, R. (2009). Bank governance, regulation and risk taking. Journal of Financial

Economics, 93(2), 259-275.

Lintner, J., 1965, The valuation of risk assets and the selection of risky investments In stock portfolios and capital budgets, Review of Economics and Statistics 47 (1), 13-37.

Kashyap, A. K. (2010). Lessons from the Financial Crisis for Risk Management. Financial Crisis Inquiry

Commission.

Kashyap, A., Rajan, R., & Stein, J. (2008). Rethinking capital regulation. Maintaining stability in a

changing financial system, 431-471.

Malhotra, N. K. (2010). Marketing research: An applied orientation. Upper Saddle River, NJ: Pearson.

Markowitz, H. (1952). Portfolio selection*. The journal of finance, 7(1), 77-91.

Nadarajah, S., Zhang, B., & Chan, S. (2014). Estimation methods for expected shortfall. Quantitative

Finance, 14(2), 271-291.

(32)

Senior Supervisors Group. (2009). Risk Management Lessons from the Global Banking Crisis of 2008. Available at http://www.financialstabilityboard.org/2009/10/r_0910/

Sharpe, W., 1964, Capital asset prices: a theory of market equilibrium under conditions of risk, Journal of Finance 19 (3), 425-442.

Shleifer, A. (2011, January). The transformation of finance. In Lecture at American Finance

Association Meetings.

Smith, C. W., & Stulz, R. M. (1985). The determinants of firms' hedging policies. Journal of financial

and quantitative analysis, 20(04), 391-405.

Stein, J. C. (2002). Information production and capital allocation: Decentralized versus hierarchical firms. The Journal of Finance, 57(5), 1891-1921.

Tetlock, P. C. (2007). Giving content to investor sentiment: The role of media in the stock market. The

Journal of Finance, 62(3), 1139-1168.

(33)

VIII. Appendix

Appendix A: List of BHCs in sample Name of BHC 2014 Assets (€ bn) Time Span in Panel Country HSBC 1,937 2004-2013 GB BNP Paribas 1,800 2004-2013 FR Deutsche Bank 1,611 2004-2013 DE Barclays 1,567 2004-2013 GB Credit Agricole 1,537 2004-2013 FR Societé Generale 1,235 2004-2013 FR

Royal Bank of Scotland 1,227 2004-2013 GB

Banco de Satander 1,116 2004-2013 ES

ING 1,081 2004-2013 NL

Lloyds Banking Group 1,011 2004-2013 GB

Unicredit 846 2004-2013 IT

Nordea Bank 630 2004-2013 SE

Intesa Sanpaoli 626 2004-2013 IT

Banco Bilbao Vizcaya Argentaria SA 583 2004-2013 ES

Commerzbank 550 2004-2013 DE

Nataxis 510 2004-2013 FR

Standard Chartered 489 2004-2013 GB

Danske Bank 432 2004-2013 DK

Caixabank 340 2009-2013 ES

Skandinaviska Enskilda Banken AB 280 2004-2013 SE

Svenska Handelsbanken 280 2004-2013 SE

Bankia 251 _2012-2013 ES

KBC Group 241 2004-2013 BE

Credit industriel et commercial 233 2004-2013 FR

Dexia 223 2004-2013 BE

Swedbank AB 206 2004-2013 SE

Erste Group Bank AG 200 2004-2013 AT

Banca Monte dei Paschi di Siena SpA-Gruppo Monte dei Paschi di Siena

(34)

Banco de Sabadell SA 163 2004-2013 ES

Deutsche Postbank AG 162 2005-2013 DE

Banco Popular Espanol SA 148 2004-2013 ES

Exor Spa 133 2010-2013 IT

Bank of Ireland 132 2004-2013 IE

Raiffeisen Bank International AG 131 _2006-2013 AT

Banco Popolare - Società Cooperativa-Banco Popolare 126 2004-2013 IT

Unione di Banche Italiane Scpa-UBI Banca 124 2004-2013 IT

Allied Irish Banks 118 2004-2013 IE

National Bank of Greece SA 111 2004-2013 GR

Pireaus Bank SA 92 2004-2013 GR

Espirito Santo Financial Group SA 85 2004-2013 LU

Banco commercial Portugues 82 2004-2013 PT

Eurobank Ergasias SA 78 2004-2013 GR

Delta Lloyd 77 2010-2013 NL

Wüstenrot & Württembergische AG 75 2004-2013 DE

Alphabank 74 2004-2013 GR

Mediobanca 70 2004-2013 IT

Appendix B: Definitions of variables

Variable Definition

Bankscope code Total assets The book value of total assets for the BHC

in the respective year

Size The natural logarithm of total assets in the respective year.

Size ² The natural logarithm of total assets in the respective year, squared.

Buy and hold Return The return of holding the stock of the BHC for one year, including dividends.

Tail Risk The cumulative amount of the 5% worst return days for the bank stock.

(35)

term funding divided by total assets.

Tier 1 Ratio The percentage of Tier 1 regulatory Capital. 1850 Total Loans / Assets The ratio of gross loans to total assets.

Gross loans includes impaired loans.

2001

Impaired Loans / Gross Loans The percentage of nonperforming loans to the total amount of loans.

18200

Nonint.Income / Gross Revenues

The ratio of income not related to interest

activities to gross revenues. 18065 Deriv.Hedging/Assets The amount of securities and derivatives

held to maturity divided by total assets.

11180

Deriv.Trading/Assets The amount of trading securities held (and at fair value through Income) relative to total assets.

1150

Change in CEO A dummy variable that identifies BHCs that experienced a change in CEO in the

previous year.

Large M&A A dummy variable that identifies BHCs that experienced a year-on-year growth in the book value of total assets that exceeds 25%.

ROAA The return on average assets held by the

BHC during the year divided by net income. 4024 RMI Score Computed as the sum of the factor loadings for

the components in a principal component analysis that have an eigenvalue of higher than one. It is calculated on year-by-year basis Used are the following four risk management variables: CRO present, CRO Executive, CRO Top5 and CRO Centrality.

(36)

CRO Present A dummy variable that indicates whether the BHC had a Chief Risk Officer or equal position (Chief Lending Officer, Chief Credit Officer) in the respective year.

CRO Executive A dummy variable that indicates whether the CRO was a member of the executive management team.

CRO Top5 A dummy variable that indicates whether the CRO was among the 5 highest paid executives.

CRO Centrality The ratio of the CRO's total compensation, excluding variable rewards, to the CEO's total compensation, excluding variable rewards. When the CROs compensation was not reported the fifth highest

compensation was used and a percentage point deducted from the ratio. When there was no CRO in the respective year, the CFOs compensation is used and a percentage point deducted from the ratio. If the compensation for the executives, excluding the CEO is reported as a total amount, the average compensation for an executive is used and two percentage points deducted from the ratio. Excess Return The difference between the ' Buy and Hold' return and

(37)

Appendix C: Principal Component Analysis

Table C1: Principal Component Analysis for the four factors of risk management in 2004 Number Value Difference Proportion Cumulative

Value Cumulative Proportion 1 2.102956 1.145721 0.5257 2.102956 0.5257 2 0.957234 0.308162 0.2393 3.06019 0.765 3 0.649073 0.358336 0.1623 3.709263 0.9273 4 0.290737 --- 0.0727 4 1

Table C2: Principle Component Analysis; Eigenvector loadings 2004

Variable PC 1 PC 2 PC 3 PC 4 CRO 0.513274 0.034591 -0.80296 0.301011 CRO Executive 0.610124 -0.2166 0.097313 -0.755886 CRO Centrality 0.217192 0.961152 0.149949 -0.080805 CRO Top 5 0.563139 -0.167554 0.568595 0.57576 Table C3: Principal Components Analysis (PCA); ordinary correlations 2004

Variable CRO

CRO Executive

CRO

Centrality CRO Top 5

CRO 1

CRO Executive 0.534522 1

CRO Centrality 0.181039 0.106618 1

CRO Top 5 0.356348 0.666667 0.144867 1

Number Value Difference Proportion Cumulative Value Cumulative Proportion 1 1.99903 1.041575 0.4998 1.99903 0.4998 2 0.957456 0.287631 0.2394 2.956486 0.7391 3 0.669824 0.296135 0.1675 3.626311 0.9066 4 0.373689 --- 0.0934 4 1

Table E5: Principle Component Analysis; Eigenvector loadings 2005

Variable PC 1 PC 2 PC 3 PC 4

CRO 0.555926 0.208014 -0.577415 0.560597

CRO Executive 0.590534 -0.273228 -0.231891 -0.723079 CRO Centrality 0.335974 0.811059 0.445725 -0.175028

CRO Top 5 0.478892 -0.473562 0.643543 0.363667

Table E6: Principal Components Analysis (PCA); ordinary correlations 2005

Variable CRO

CRO Executive

CRO

Centrality CRO Top 5

(38)

CRO 0.543467 0.253311 -0.488633 0.63381

CRO Executive 0.554652 -0.250221 -0.401046 -0.684772 CRO Centrality 0.440506 0.669384 0.558339 -0.214796

CRO Top 5 0.45051 -0.652034 0.537268 0.288505

Table C9: Principal Components Analysis (PCA); ordinary correlations 2006 Variable CRO CRO Executive CRO Centrality CRO Top 5

CRO 1 CRO Executive 0.65 1 CRO Centrality 0.487707 0.350649 1 CRO Top 5 0.340693 0.524142 0.291281 1

CRO 0.544745 0.05176 -0.604375 0.579055

CRO Executive 0.575626 -0.279872 -0.212546 -0.738343 CRO Centrality 0.376717 0.868752 0.29772 -0.121313 CRO Top 5 0.479574 -0.405293 0.707755 0.323772 Table C12: Principal Components Analysis (PCA); ordinary correlations 2007 Variable CRO CRO Executive CRO Centrality CRO Top 5

CRO 1

CRO Centrality 0.364663 0.274979 1

(39)

CRO 0.514859 -0.056227 0.800148 0.302524

CRO Executive 0.604763 -0.236518 -0.121956 -0.750631 CRO Centrality 0.219129 0.967427 -0.02612 -0.124039 CRO Top 5 0.566713 -0.07059 -0.586692 0.574148 Table C15: Principal Components Analysis (PCA); ordinary correlations 2008 Variable CRO CRO Executive CRO Centrality CRO Top 5

CRO 1

CRO Centrality 0.167721 0.093859 1

CRO Top 5 0.382971 0.68313 0.192683 1

CRO 0.553635 0.04295 -0.686424 0.469538

CRO Executive 0.622345 -0.135377 -0.03338 -0.770225 CRO Centrality 0.102081 0.987086 0.079485 -0.094456 CRO Top 5 0.543834 -0.074086 0.722074 0.421148 Table C18: Principal Components Analysis (PCA); ordinary correlations 2009 Variable CRO CRO Executive CRO Centrality CRO Top 5

CRO 1

CRO Centrality 0.1119 0.018618 1

RMI and the impact on Tail Risk: The European Story

RMI and the impact on Tail Risk:

The European Story

University of Groningen

MSc Finance

RMI and the impact on Tail Risk:

The European Story

Contents

I.

Introduction

II.

Literature Review and hypothesis development

III.

Data

A. Data collection

B. The Risk Management index

IV.

Methodology

A. RMI Index

B. Multivariate analysis

V.

Results and Discussion

Descriptive Statistics

Correlations among key variables.

VII. Literature

VIII. Appendix