Does Brexit reverse the Impact of
Compensation Regulation? Evidence from the
European banking system
BSc Economics & Business Economics
Anouk D. Bijman
Supervisor: Ekaterina Seregina
University of Amsterdam, June 30
th2020
Abstract
This paper assesses the impact of Brexit on compensation regulation. This is done by both looking at previous research, as well as performing an empirical analysis. The main focus in the empirical part of the research is on multiple difference-in-difference regressions. These regressions are based on multiple variable to fixed remuneration ratio thresholds. For banks that show a variable to fixed remuneration ratio that is in the top 50%, the difference-in-difference variables are significant to a bank's risk level. For banks that show several levels of variable to fixed remuneration ratio, the difference-in-difference variable is significant to the additional risk to the financial system a bank is contributing. The literature shows that the UK is in a position in which it can change remuneration legislation. However, it is not concise if the UK will do this.
Statement of Originality
This document is written by Student Anouk Divera Bijman who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Abbreviations
CEO – Chief Executive Officer
CRD IV – Capital Requirements Directive IV EPS – Earnings per Share
EU – European Union
G-SIB – Global Systemically Important Bank NII – Net Interest Income
NPL – Non-Performing Loan ROE – Return on Equity ROA – Return on Assets
TSR – Total Shareholder Return UK – United Kingdom
1. Introduction
When the people of the United Kingdom (UK) voted to leave the European Union (EU) in 2016, no one could fully comprehend or know what the consequences of Brexit were to be. Brexit led to uncertainty in terms of what the effects would be for the economy in general. It is still unsure what Brexit will mean for the bonus culture in the UK. The European regulation on caps for executive pay has faced many protests within the UK. Should there be an option to diverge from the bonus regulation post-Brexit, the UK could potentially loosen the standards for its banks.
Executive pay has been a controversial topic for an extended period. The most recent wave of attention to the topic has started after the Great Financial Crisis in 2008: a significant part of the public expressed the opinion that the bonus system in the financial sector is not adequate. The public thought it was unbelievable that banks needed bailout money while these same banks granted their employees enormous bonuses. (Lu, 2016) According to Inman (2009), Citigroup, one of the banks that received a large amount of bailout money, granted their employees $5.3 billion in bonuses in 2008, even though the bank lost more than $18 billion that same year. In order to prevent such a failure of the financial system from happening again, the EU implemented new measures to prevent similar situations from happening. These measures include the cap on variable remuneration. This bonus cap implies that the variable to fixed remuneration ratio has a maximum of 100%. If shareholders grant their approval, the maximum can increase to 200% of the fixed remuneration level.
Much controversy still exists concerning executive remuneration. (Lui, 2018) The main question is whether bonuses increase an executive's performance or incentivize
excessive risk-taking by an executive. It is essential to know if a relationship exists between performance and compensation, and how strong it is because such a relationship affects the financial stability in a country. (Cerasi, Deininger, Gambacorta, & Oliviero, 2020) If a correlation exists between excessive bonuses and the level of risk, one can influence the banking system outcomes by adjusting the pay structure of executives.
Now that the UK can change its legislation concerning the bonus system, the question arises to what extent Brexit has an impact on the link between performance and compensation concerning banks. This will be researched in this paper, explicitly looking at multiple variable to fixed remuneration ratios and risk levels. In my study, I will look at a sample of European
banks and see whether the risk levels around the Brexit announcement are significantly different for high bonus banks.
According to certain studies, the goal of bonus restrictions is to decrease excessive risk-taking. However, this goal is not likely to be achieved. (Murphy, 2013) Others, such as Colonnelllo, Koetter, & Wagner (2019) and Lui (2018), agree with this finding. Therefore, it is essential to research how compensation schemes influence the risk structure banks are willing to take. This is to make sure that compensation legislation achieves what it is aiming to achieve.
In order to analyze the correlation between performance and compensation, a panel regression is run based on a dataset concerning data about market-based risk indicators, control variables, and the variable to fixed remuneration rate of EU banks. After the data cleaning procedure, multiple regressions are run to examine to what extent high-bonus banks show different outcomes concerning the comparable low-bonus banks. The focus of the dependent variable is on the CoVaR, a measure for systemic risk suggested by Adrian & Brunnermeier (2011). The regressions are also run with Systemic Risk to Market
Capitalization as the dependent variable, in order to make the research more robust.
In the first regression, the baseline regression is run with only the control variables as independent variables. The second regression that is run takes a dummy variable into account that indicates whether a bank belongs to a treated or non-treated group. The third model is a set of difference-in-difference models. The model is based on a Brexit variable, a dummy variable that indicates a certain level of variable to fixed remuneration ratio, and an
interaction variable that is the product of these two dummy variables. The dummy variable that concerns the level of variable to fixed remuneration ratio can either refer to whether a bank is treated, in a high bonus group, or in the biggest half of the group. The last regression that is run is a difference-in-difference model based on time period dummy variables.
The main results following from the regressions are that Brexit has a significant effect on both dependent variables, based on the Biggest Half variable. Brexit has a positive effect on the CoVaR and a negative effect on the Systemic Risk to Market Capitalization. In other words, there is significant evidence that the option to diverge from EU regulation matters for banks that are in the biggest half group. According to theory, the effect of a bonus cap is not concise. It can, therefore, not be predicted to what extent the UK will loosen their
My paper will be structured as follows: Part 2 covers a literature review, including the discussion of remuneration regulation as well as the current status of UK legislation
concerning remuneration. It also covers the previous research and explains how my study contributes to this. After the literature review, the methodology is explained in Part 3. Then, Part 4 shows the results obtained by following the method described in the previous Part. Lastly, in Part 5, the discussion of the result is explained.
2. Literature Review
The correlation between performance ratios and remuneration ratios is very important because this correlation affects a bank's risk behavior and therefore affects the financial stability of that bank. (Cerasi, Deininger, Gambacorta, & Oliviero, 2020) Consequently, if many banks display high-risk behavior, it affects the stability of the entire sector. An explanation for this effect on financial stability is the fact that, according to Lui (2018), bonuses that depend on performance levels have the consequence that executives have no incentive to care about the long-term consequences of their decisions.
Lu (2016) discusses that this is a typical principal-agent problem: an executive is supposed to perform his job in the interest of his shareholders, while at the same time, it is more profitable for that executive to perform his job in his interest. The excessive focus on short-term results, as a result of the relationship between the value of variable remuneration and performance, leads to damaging effects in the long-term and excessive risk-taking. (Lui, 2018) According to Lu (2016) and Cerasi et al. (2020), excessive risk-taking and short-term thinking were central to bankers' decision making before the Financial Crisis in 2008.
Cerasi et al. discuss that the suboptimal financial performance visible for some banks during the Financial Crisis was associated with a combination of a substantial level of variable remuneration and lax regulations. Short-term thinking by executives arose because executives were incentivized to think like their shareholders. The latter group is prone to short-term thinking and high risk-taking because their risk is limited by the value of their investment. In practice, this means that in case of bankruptcy, the maximum amount a shareholder can lose equals the value of their investment. Because executives' risks are also limited in this way, they exhibit an excessive focus on term results. This focus on short-term thinking led to banks putting themselves in a dangerous position in short-terms of the level of risks they were taking. Research performed by Fahlenbrach & Stulz (2011) shows that banks that used incentive schemes related to the value of shareholder wealth exhibited worse performances during the Financial Crisis. Colonello, Koetter, & Wagner (2019) confirm this result and, above that, relate the correlation between performance and compensation to financial stability.
Contrary to the results of the researches and papers mentioned above, equity-based compensation does not merely result in negative consequences. Equity-based compensation aligns bankers' incentives with the incentives of their shareholders, according to Lu (2016).
Lu (2016) explains that by using an equity-based compensation scheme, the interests of executives and shareholders are aligned so that executives are rewarded for taking high risks. High risks usually lead to higher shareholder value and should, therefore, be incentivized. Using equity-based compensation schemes solves the problem of an executive not willing to take high-enough risks that are necessary for high performance.
This finding is emphasized by Kuo, Li, & Yu (2013). They performed research concerning the correlation between Chief Executive Officers (CEO) equity-based compensation and performance. Kuo, Li, & Yu (2013) concluded that an equity-based compensation scheme has a positive influence on firm performance. This theoretical framework is the reasoning behind the loose regulations concerning variable remuneration that was implemented before the Financial Crisis.
Colonello et al. (2019) strengthen this reasoning by discussing that a decrease in incentive pay leads to a decrease in executives' effort and a possible increase in systematic risk. The latter can be explained by the fact that a high level of fixed remuneration creates sort-of insurance for risk-averse bankers who refuse to take the risks that are needed for a high level of performance. By not taking high-risk decisions, the performance of a firm is likely to be lower than optimum. If an executive receives a higher level of fixed remuneration, the potential personal loss related to the lower performance of the firm is minimal. Therefore, the presence of a high level of fixed remuneration insures executives against a lower than optimum level of performance.
Currently, executive remuneration in the European Union is regulated by various measures. The regulations concerning these measures were implemented after the Financial Crisis and are meant to align the interests of both bankers and overall stakeholders. Lui (2018) discusses that these current measures are based on the fourth Capital Requirements Directive (CRD IV). Specifically looking at the UK, their legislation concerning executive pay is derived from both domestic legislation as well as EU legislation. As a result, current UK legislation is largely based on CRD IV. For example, all UK banks and building societies need to adhere to the requirements set by CRD IV.
According to Lui, the CRD IV legislation is based on the four following principles. The first principle is the use of malus, which refers to the fact that the level of variable
remuneration can be adjusted before the bonus has been awarded to an executive. The second principle is the use of clawback, which refers to the fact that bonuses can be reclaimed. The
third principle is the application of bonus caps, which means that the amount of variable remuneration an executive can receive can be restricted by putting a ceiling to the level of bonuses. The fourth principle is the requirement of remuneration disclosure, which means that a certain level of disclosure is required with respect to the level of executives' remuneration.
Lu (2016) also elaborates on these principles. Lu (2016) explains that clawback is a contractual clause that can force employees to return money that has already been rewarded to them. This clause allows firms to request their employees to return bonuses under certain circumstances, such as in case a certain level of performance has or has not been achieved. According to Lu (2016), clawback plays an essential part in the transformation of the excessive bonus culture of the UK. Furthermore, Lu (2016) also elaborates on the caps on bonuses. Implemented in 2014, the cap on bonuses restricts bonuses to either a maximum of 100 percent of an employee's fixed salary or 200 percent of an employee's fixed salary with the shareholders' approval. Lu (2016) states that by implementing these bonus caps, the negative consequences related to excessive bonuses could be reduced.
Similar to Lu (2016), Murphy (2013) believes that the goal of the CRD IV is to lower the excesses in banking bonuses and reduce the excessive risk-taking that is associated with these excessive bonuses. However, Murphy (2013) discusses that a cap on bonuses is not likely to contribute to achieving these objectives. Murphy (2013) notes that as a result of a restriction on the variable remuneration is, the level of fixed remuneration is likely to increase. This results in the excessive risk-taking being incentivized because the negative consequences in terms of monetary terms related to poor performance are reduced.
This finding is confirmed by Colonnello et al. (2019). Based on their research, they conclude that risk-taking significantly increased after the bonus cap was introduced. In contrast, the traditional bonus system without a cap on bonuses encourages the avoidance of risks with a high likelihood of leading to a negative result. Instead, the traditional bonus system incentivizes the taking of risks with a high likelihood of leading to a positive result.
An additional negative effect of the restrictions of bonuses is explained by Murphy (2013). He explains that a system in which bonuses are capped may result in managers being incentivized to be dishonest about reporting the period in which they achieved certain levels of performance. Managers may undervalue good performances in a specific period and overvalue bad performances in another period. Executives are incentivized to do this because reporting their performances in this way means that the excess of value related to a good
performance is shifted to a period with low performance and vice versa. This is beneficial to an executive because no additional bonus will be rewarded to them to reward them for an excessively good performance, and no fixed remuneration will be taken from them to penalize them for their lousy performance. Therefore, by using this fraudulent tactic, their net reward increases. Another negative consequence related to bonus restrictions is the fact that these restrictions could cost jobs in the London financial sector. (Lui, 2018)
Colonnello et al. (2019) elaborate on this notion as well, by discussing that Financial Stability Board guidelines concerning compensation may reduce banks' opportunities to hold on to their CEOs. However, Colonnello et al. (2019) conclude, based on their research, that this concern is not valid. The turnover of executives is not significantly higher after banking regulations changed, and banks were forced to lower their variable remuneration.
Contradictory to the beliefs of researchers such as Murphy (2013), the research of Colonnello et al. (2019) concludes that bonus caps neither reduce risk-taking nor increase the stability of the financial system. Colonnello et al. (2019) conclude that a bonus cap only has the required effect if the bonus is not reliant on obtaining a specified performance target. In conclusion, it is not clear what effect a bonus cap will have on the level of risk-taking and the financial system.
Even though the United Kingdom has fully implemented CRD IV, it has made clear that it opposes the restriction of the level of bonuses. (Lu, 2016) After the Financial Crisis, the UK investigated the bank failures and the banking characteristic that led to these failures. This investigation shows that one of the main reasons for the Financial Crisis was excessive risk-taking. This is a big part of the reason why the UK did implement the CRD IV into their regulations and implemented even more stringent regulations concerning clawback and deferral. However, Lu (2016) discusses that the UK does apply the CRD IV rules in a flexible way. EU legislation is applied at the minimum level, and as much as possible is left to the previous UK regulations. For example, the UK left out certain types of firms, such as smaller financial institutions, from having to adhere to the bonus restrictions. According to UK banking regulators, the restriction on bonuses can be harmful to the competitiveness and sustainability of the financial industry in the UK. This is because it decreases the
competitiveness of London as a financial center.
Since the UK has left the EU, an opportunity has formed for the UK to change and improve their remuneration legislation. Lu (2016) discusses that it is essential that the UK ensure that the remuneration legislation remains coherent during the adjustment period in
which the UK leaves the EU. Furthermore, Lu (2016) explains that it is crucial for the UK to merge their regulation with EU regulation in order to ensure trade relations between the two in the future.
Lu (2016) discusses the fact that many commentators, regarding the remuneration legislation, have predicted that the UK will abolish the bonus cap. Lu (2016) states that equity-based remuneration was indeed related to the issue of excessive risk-taking during the Financial Crisis. However, he notes that the percentage of banks that used equity-based remuneration was not the underlying problem. Lu (2016) explains that the type of
performance metric used to define the level of remuneration was the underlying problem. The bonus cap, therefore, focuses on the wrong issue. He also notest that, due to Brexit, the UK financial sector will have to face more significant struggles than the bonus cap. Therefore, some banks do not want remuneration legislation to change in order to make sure that trade relations with the EU are not damaged even more than they already will be. Lu (2016) concludes that the UK will probably remove the bonus cap. This is because it will probably not affect the trade relations between the UK and the EU that much, but it will remove the negative consequences that come along with implementing a bonus cap.
Previous research concerning the effect of performance on compensation, its relation to financial stability, and the effect of a sudden shock to this relation is done by researchers such as Cerasi et al. (2020) and Colonnello et al. (2019). Cerasi et al. (2020) research whether banks' CEO compensation changed after the remuneration legislation changed as a result of the Financial Crisis. They analyzed the sensitivity of the compensation of banking CEOs to performance. In their research, performance is measured by short-term profit and risk. Based on their data, Cerasi et al. (2020) conclude that the remuneration legislation significantly changed how the compensation of banking CEOs is structured. For commercial banks, the share of variable compensation is positively correlated to short-term profit and negatively correlated to risk.
Colonnello et al. (2019) investigate the effect of executive compensation regulation on multiple factors. According to their research, restricting the proportion of variable
compensation does not lead to executive directors leaving the EU banking industry. At the same time, risk-adjust performance worsens if variable remuneration is restricted. This can be explained by both the fact that a lower share of variable remuneration decreases incentives to put in a high level of effort, as well as by the insurance effect. The insurance effect refers to the fact that a larger share of fixed compensation functions as a type of insurance for
executives against achieving submaximal performance. Furthermore, Colonnello et al. (2019) conclude that executives who are more affected by the bonus restriction show a bigger
systematic as well as systemic risk.
In terms of when exactly the event Brexit took place, multiple dates can apply. (Clarke, 2017) The referendum concerning the EU membership of the UK happened on the 23th of June 2016. Nine months later, on March 29th 2017, the UK triggered an Article that indicated that the UK was formally planning on leaving the EU. Because of this, the UK initially planned on leaving the EU on March 29th 2019. Then, on June 19th 2017, the first round of negotiations between the UK and the EU Commission began. Because of struggles concerning the negotiation process, the initial leaving date of March 29th 2019 could not be achieved. Therefore, the leaving date was extended until October 31st 2019. Again, this leaving date was not feasible as well. Therefore, it was extended until January 31st 2020. This is the date that the UK left the EU for good. Up until December 31st 2020, the UK is in a transition period in which the UK will adjust to not being part of the EU.
The research performed in this thesis will add to the existing research by showing to what extent Brexit, and therefore a possible change in regulation, has on the correlation between risk and performance. It is an excellent opportunity to analyze this effect because the UK is in a similar financial climate as the remaining part of the EU.
3. Method
3.1 Sample data
The analysis is done for a sample of large European banks for which the remuneration data is available. According to Cerasi et al. (2020), disclosure of the remuneration data is quite common for large and publicly traded banks. However, it is not mandatory for all banks. This is the reason why not all banks have remuneration data available. As a result, my sample is limited to banks for which remuneration data is available. In this research, three main datasets are used: one concerning the variable to fixed remuneration ratio, one concerning control variables, and one concerning risk metrics. The dataset concerning the ratio of variable to fixed remuneration concerns 159 banks. In contrast, the dataset concerning panel data about the control variables about 149 banks ranging from the year 2000 to 2017, depending on the variable.
Furthermore, the sample is limited to banks that are publicly traded on the stock market. This is because the risk indicators are calculated based on the banks' stock market data. The data concerning the risk metrics concerns 94 banks. After banks that do not correspond to any value of the variable to fixed remuneration ratio, as well as banks that did not correspond to a region were dropped, 75 banks were left. The data used for the analysis conducted is strongly balanced on a yearly basis: all banks have data for all years of interest. However, the data is unbalanced on a quarterly basis: not all banks have data for all quarters. This probably does not affect the results of the analysis because the banks remain to be balanced on a yearly basis. The use of fixed effects controls for variables that cannot be observed or measured, such as differences in business ethics across banks. Panel data setup also controls for variables that do change over time but do not change across different entities. This is referred to as time fixed effects. The banks present in the dataset can be classified into three regions: the UK, the EU, and Switzerland. The data from all three banks are used in the analysis because the legislation in Switzerland is comparable to that of the EU and the UK.
The data concerning the variable to fixed remuneration ratio is given on a yearly basis and only available for the year 2013. The ratios are computed based on data extracted from Capital IQ. Even though the data is not the most recent data, it is still a good indicator of the variable to fixed remuneration ratio in the years that are of interest to this research. The ratios represent the remuneration structure, which would be present in case of absence in a change of regulation. The focus in this research is on the variable to fixed remuneration ratio. By
focusing on this, it is possible to group the banks in different groups, such as high bonus versus low bonus groups.
The dataset concerning risk metrics holds data about monthly risk metrics, quarterly risk metrics, quarterly covariance values, and monthly covariance values. This runs from 2000 to 2017 on both a quarterly and a monthly basis. For the analysis conducted in this paper, the focus is on the quarterly risk metric data. This is to avoid the pooling together of data points. The data concerning the risk metrics are extracted from the Volatility and Risk Institute, which is a center that researches and analyzes financial and non-financial risks. This data includes the Systemic Risk to Market Capitalization ratio, the maximum of this ratio, the beta, the long-run marginal expected shortfall, the sigma, the correlation, and the leverage of a firm. In this paper, the focus is on the following variables: the capital shortfall to market capitalization ratio, the long-run marginal expected shortfall, and the leverage. Firstly, this is because the maximum of the capital shortfall to market capitalization ratio is not a good indicator of risk of a firm, because it could be an outlier. Therefore, it is preferable to look at the average capital shortfall to market capitalization ratio.
The last part of the dataset contains data concerning controls, including Assets, Return on Assets (ROA), Tier 1 level, and Non-Performing Loans (NPL). These variables run from 2009 to 2016. Because data is needed that runs up until 2017, all variables, including the risk metrics and other control variables, are lagged by one year. This is a standard practice that aims to limit the reverse influence of dependent variables on the controls. Other control variables include Operating Income, Deposits, Loans, Net Income, Net Interest Income (NII), Other Income, and Liquidity. These variables range from 2008 to 2015. All data concerning the control variables are given on a yearly basis and extracted from FactSet.
3.2 Data Preparation
All variables are winsorized at level 1% and 99% in order to mitigate the potential impact of the outliers. According to Bentley, Keyton, & Reifman (2010), winsorizing is preferable to trimming if the sample size is relatively small. In this research, the number of observations is quite high. However, after dropping the banks that did not provide the
necessary information, the number of banks left is 75. This number is not that high compared to the number of listed banks in the EU and UK, 148. (Köhler, 2015) The number of banks in my sample is even smaller compared to the total number of banks in the EU and UK
combined, which is 3214. Another advantage of winsorizing compared to trimming, is the fact that by using winsorizing, information that a specific data point had concerning the frequency distribution is preserved. At the same time, by using winsorizing, the harmful effects on the reliability of the regression of outliers are removed. Based on these relative advantages, the variables in this research are winsorized instead of trimmed.
After that, all monetary values are transformed into Euros, so that the values can be compared in a correct way. The currency of the UK data is initially displayed in Pounds. The average exchange rate from Pound to Euro between 2009 and 2016 was 1.29. Therefore, by dividing the monetary data by 1.29, for every data point for which the variable region corresponds to the UK, the data is transformed into Euros. The currency of the Swiss data is initially displayed in Swiss Francs. The average exchange rate from Swiss Franc to Euro between 2009 and 2016 was 0.93. Therefore, by dividing the monetary data by 0.93, for every data point for which the variable region corresponds to Switzerland, the data is transformed into Euros. The monetary data includes the following variables: Assets, Return on Assets, Tier1, Non-Performing Loan, Operating Income, Loan, Net Income, Net Income after Interest, Other Income, and Liquidity.
After that, Liquidity is divided by 100 to transform it into a ratio between 0 and 1 instead of a percentage. In order to transform the monetary values into millions, the following variables are divided by 1,000,000: Assets, Tier1, Non-Performing Loans, Loan, Net Income, Net Income after Interest, and Other Income. Then, several ratio variables are created. By using ratios in the analysis, initial differences in bank size can be controlled. All ratio variables are all scaled by Deposits, except for the already given ROA. The following variables are scaled by Deposits: Tier 1, Operating Income, Net Income after Interest, Non-Performing Loan, and Loan. The variables are scaled by Deposits, rather than Assets because the standard deviation of Deposits is lower than the standard deviation of Assets. This means
that the ratio variables are less influenced by the deviations of the denominator when the variables are divided by Deposits instead of Assets.
Then, various dummy variables are generated. The variable RegionUK is generated, which takes value 1 if a bank is situated in the UK. Besides that, two Brexit variables are generated, which indicate if the time variable is pre- or post-Brexit. The first variable, Brexit1, takes value 1 if the date is after the referendum date, June 23rd 2016. Because June 23rd is at the end of the second quarter, the Brexit1 variable is set as the second quarter of 2016. The data after the second quarter of 2016 is referred to as post-Brexit data, and the data up until and including the second quarter of 2016 is referred to as pre-Brexit data. The second variable, Brexit2, takes value 1 if the date is after the start of the negotiations, March 29th 2017. Because March 29th is at the end of the first quarter, the Brexit2 variable is set as the first quarter of 2017.
After that, three dummy variables concerning the variable to fixed remuneration ratio are generated. The first dummy variable divides the banks into a treated and non-treated group. The Treated variable takes value 1 if the ratio variable to fixed remuneration ratio is above 100 percent, because above this threshold shareholders' approval needs be obtained as 100 percent is the legal maximum of the variable to fixed remuneration ratio. The second dummy variable is a High Bonus variable, which indicates whether a bank is in the top 10 percent highest remuneration ratios. This means that the High Bonus variable takes value 1 if the variable to fixed remuneration ratio is above 2.17. The third dummy variable is the
Biggest Half variable, which indicates whether a bank is in the top 50% highest remuneration ratios. This means that the Biggest Half variable takes value 1 if the variable to fixed
remuneration ratio is above 0.2658228. The quantile distribution of the variable to fixed remuneration ratio variable is shown in Table 2 in Part 4.
Lastly, a dummy variable concerning the global systemically important banks (G-SIB) is created. According to Magnan & St-Onge (1997), the level of executive compensation is positively correlated to the level of executive compensation. Therefore, it is important to include a variable that takes the bank size into account. The GSIB variable is based on the list of G-SIBs that the Financial Stability Board published in November 2016. (Financial Security Board, 2016) This publication list is chosen because it is most relevant for the data that is used in my research. The sample includes 14 banks that are mentioned in the G-SIB list. The GSIB variable takes value 1 if a bank is mentioned in the G-SIB list.
In Table 3, shown in Part 4, the correlations between the GSIB variable and the remuneration variables are shown. Based on this, the GSIB variable is not included in the regressions that include the Treated variable. This is because the Treated variable is relatively strongly correlated to the GSIB variable. In order to research the impact of the GSIB variable itself, a difference-in-difference regression is run based on the GSIB variable, the Brexit1 variable, and an interaction variable that is the product of these two variables. Because the GSIB is not as strongly correlated to the High Bonus and Biggest Half variable, it is used as a control variable in the regressions that include these remuneration variables.
3.3 Analysis
For the first part of the analysis, descriptive statistics are generated. These statistics are displayed over the time period 2008-2016 and over the time period 02.2015-02.2017. This second time period reflects the time period is a year before and a year after the referendum date. By doing this, it is shown that there is a difference in means for the variables for the different time horizons. Based on this, the focus in the regression part of the analysis is on the time period 02.2015-02.2017. This is because this data reflects the period before and after the Brexit more accurately. The data from 2008-2016 is likely to be influenced by more events than just Brexit, which is the reason why the time period in my research is limited.
The data is divided into the original three categories: variable to fixed remuneration ratios, control variables, and risk metrics. The descriptive statistics of both the control variables and risk metrics are calculated as a whole group and on a grouped basis. These groups are based on the variable RegionUK, the two different Brexit variables, and the remuneration ratio dummy variables. Similarly, the descriptive statistics of the variable to fixed remuneration ratios are calculated as a whole and on a grouped basis. These groups are based on the variable RegionUK, High Bonus variable, the Treated variable, and the Biggest Half variable. The statistics do not need to be grouped based on the two different Brexit variables, because the data originates from 2013 and therefore does not differ per quarter. After the descriptive statistics have been generated, multiple t-tests for all control variables are run to see if they respond significantly to any of the dummy variables.
The second part of the analysis is focused on multiple panel regressions. The focus for the dependent variable in the regression analysis is on CoVaR, as defined by Adrian &
Brunnermeier (2011). According to their theory, CoVaR refers to the level of risk a financial system faces, conditional on the institutions in the system being under distress. The main advantage of using CoVaR as a dependent variable is the following: CoVaR focusses on the additional risk an institution is contributing to the overall system risk, whereas a traditional risk metric focusses on the risk of an individual institution.
All regressions are run based on the Systemic Risk to Market Capitalization ratio variable as a dependent variable as well. This variable is chosen because it reflects the systemic risk of a bank well while controlling for initial differences between banks' size and costs. This makes the Systemic risk to Market Capitalization ratio variable a reliable
for the Systemic Risk to Market Capitalization ratio is over 6,700 times smaller than the standard deviation of the maximum of that same ratio. This makes sense as the maximum of a ratio is likely to be influenced by outliers, whereas the average value of a ratio is not. As a result, the focus is on the Systemic Risk to Market Capitalization ratio rather than the maximum of that ratio.
The regressions are run based on the following baseline model. The variable that indicates whether a bank received government support in 2011 is included in the control variables. This is because a firm that received government support in the past has shown that it has had problems in the past with taking too high levels of risk. Therefore, I suspect that it affects the Systemic Risk to Market Capitalization ratio.
𝑅𝑖𝑠𝑘 𝑀𝑒𝑡𝑟𝑖𝑐𝑖𝑡 = 𝛼𝑖+ 𝜆𝑡+ 𝛽1𝐶𝑜𝑛𝑡𝑟𝑜𝑙 𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑖𝑡+ 𝜀𝑖𝑡 (1) In this context, 𝛼𝑖 refers to the fixed effect relating to the bank variables, and 𝜆𝑡 refers to the
fixed effect relating to the time variables. The control of interest is 𝛽1, which captures the effect of the control variable on the risk metric. The model above can be extended to the following model.
𝑅𝑖𝑠𝑘 𝑀𝑒𝑡𝑟𝑖𝑐𝑖𝑡 = 𝛼𝑖+ 𝜆𝑡+ 𝛽1𝐶𝑜𝑛𝑡𝑟𝑜𝑙 𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑖𝑡+ 𝛽2𝑇𝑟𝑒𝑎𝑡𝑒𝑑𝑖𝑡+ 𝜀𝑖𝑡 (2) This model takes one of the remuneration ratio variables into account. This means that the effect of having a certain level of variable to fixed remuneration ratio is reflected in 𝛽2. The
variable Treated can be replaced by the High Bonus variable or the Biggest Half variable, in order to confirm that the variable to fixed remuneration ratio affects the dependent variable. The third model that will be analyzed is the following difference-in-difference model.
𝑅𝑖𝑠𝑘 𝑀𝑒𝑡𝑟𝑖𝑐𝑖𝑡 = +𝛼𝑖 + 𝜆𝑡+ 𝛽1𝐶𝑜𝑛𝑡𝑟𝑜𝑙 𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑖𝑡+ 𝛽2𝑇𝑟𝑒𝑎𝑡𝑒𝑑𝑖 + 𝛽3𝐵𝑟𝑒𝑥𝑖𝑡𝑡+
𝛽4𝐷𝐼𝐷𝑖𝑡+ 𝜀𝑖𝑡 (3, 4, 5, 6)
In order to run this regression, a new dummy variable DID, needs to be created. This variable is the product of one of the Treated variable multiplied by Brexit1. The variable Treated can be replaced by the High Bonus variable or the Biggest Half variable. When the High Bonus variable is used, the DID variable is the product of the High Bonus variable multiplied by Brexit1. When the Biggest Half variable is used, the DID variable is the product of the Biggest Half variable multiplied by Brexit1. 𝛽2 indicates the underlying difference between the two groups and 𝛽3 indicates the macro effect of the Treated variable. The parameter of
Lastly, the regression model can be extended into a difference-in-difference model based on multiple period dummy variables. This model represents the following regression (7).
𝑅𝑖𝑠𝑘 𝑀𝑒𝑡𝑟𝑖𝑐𝑖𝑡 = 𝛽0+ 𝛽1𝐶𝑜𝑛𝑡𝑟𝑜𝑙 𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑖𝑡+ 𝛽2𝑇𝑟𝑒𝑎𝑡𝑒𝑑𝑖 + ∑−1−2𝛿𝑘𝐵𝑟𝑒𝑥𝑖𝑡1𝑡+ ∑ 𝛿21 𝑘𝐵𝑟𝑒𝑥𝑖𝑡1𝑡+ ∑−2−1𝛾𝑘𝐵𝑟𝑒𝑥𝑖𝑡1𝑡𝑇𝑟𝑒𝑎𝑡𝑒𝑑𝑖+ ∑ 𝐵𝑟𝑒𝑥𝑖𝑡121 𝑡𝑇𝑟𝑒𝑎𝑡𝑒𝑑𝑖+ 𝜀𝑖𝑡 (7) In order to run this difference-in-difference model, new dummy variables related to the two periods before and the two periods after Brexit need to be created. For example, the Q2 variable takes value 1 if the date is in the fourth quarter of 2016, and the -Q2 variable takes value 1 if the date is in the first quarter of 2016. Additionally, several interaction variables need to be created. These interaction variables are the product of the period dummy variable and the DID variable. The DID variable is constructed in the same way as in the third model. If 𝛾−2 and 𝛾−1 is insignificant and 𝛾1 and 𝛾2 is significant, it shows that the impact of Brexit is significant.
4. Results
4.1 Descriptive Statistics
Below, the summary statistics for the variables are shown over the time period of interest.
Table 1
Summary statistics: 06.2015-06.2017
Number of Observations Mean
(Standard Deviation)
CoVaR 666 0.27
(0.22)
Systemic Risk to MC 675 0.67
(1.16)
Maximum Systemic Risk to Market Capitalization 675 307.00 (7867.86) ROA 675 0.24 (1.12) Tier1 to Deposits 502 1.68 (9.98) NPL to Deposits 502 5.50 (35.82)
Operating Income to Deposits 502 3.47
(21.99) NII to Deposits 502 3.47 (21.99) NPL to Deposits 502 0.24 (0.71) Loan to Deposits 502 125.82 (761.91) Liquidity 675 0.26 (0.20) Government Support 2011 675 0.12 (0.33) Government Support 2013 675 0.20 (0.40) GSIB 675 0.18 (0.39)
Variable to Fixed Remuneration Ratios
Variable to Fixed Remuneration Ratio 675 0.64
(0.87) Treated 675 0.23 (0.42) High Bonus 675 0.09 (0.29) Biggest Half 675 0.49 (0.50)
UK & Brexit variables
RegionUK 675 0.08 (0.27) Brexit1 675 0.44 (0.50) Brexit2 675 0.11 (0.31)
Interestingly, only 8% of the sample consists of UK banks. 44% of the sample
consisted of data after the Brexit referendum date. Furthermore, 23% of the sample is affected by the bonus regulations. The average variable to fixed remuneration ratio is 0.69. This means that, on average, the value of the bonus of a bank CEO in this sample is 69% of their fixed remuneration.
As mentioned in Part 3, the control variables are capped by Deposits, in order to control for initial differences between banks. The variable Biggest Half should have a mean of 0.50, as it indicates whether a bank's fixed to variable remuneration ratio is above the median or not. However, the mean of Biggest Half is 0.49. This is probably because the median value that is used for the definition of Biggest Half is rounded off. The same explanation applies to the occurrence that the variable High Bonus has a mean of 0.09, whereas it should have a mean of 0.10. The first Brexit date, June 2016, is chosen as the cutoff date, because the sample includes data for all variables for the entire time period between one year before June 2016 and one year after. In contrast, the sample does not include data for all variables for the entire time period between one year before the second Brexit date, March 2017, and one year after. Therefore, the variable Brexit2 should not be used to determine the effect of Brexit. Table 2
Detailed summary statistics of the variable to fixed remuneration ratio
10% 25% 50% 75% 77.17% 90%
Variable to Fixed Remuneration Ratio
0 0 0.2658228 0.97 1 2.17
In Table 2, the detailed summary statistics of the variable to fixed remuneration ratio are shown. The definitions of high bonus and Biggest Half are based on these values. The median is at 0.2658228, and the top 10% is at 2.17. Interestingly, the lowest 25% has a variable to fixed remuneration ratio of 0. This means that at least 25% of the banks does not pay any variable compensation to their CEOs.
In Table 3, the correlation between the G-SIB variable and the Treated, High Bonus, and Biggest Half variables is shown. Noticeably, the correlation between Treated and GSIB is the highest, followed by the correlation between Biggest Half and GSIB. The smallest
correlation is between Biggest Half and GSIB. Based on this, regressions run based on Treated do not include GSIB as a control variable, whereas regressions run based on High Bonus or Biggest Half do include GSIB as a control variable.
Table 3
Correlation table between GSIB and Treated, High Bonus, and Biggest Half
Treated High Bonus Biggest Half
GSIB 0.4762 0.3168 0.3486
In Table 4, the summary statistics of the variables used in the regression, categorized by pre- and post-Brexit, are depicted. In these statistics, the variable Brexit1 is used, meaning that the data is classified based on the referendum date: June 2016. This is because a larger part of the sample has a value of 1 for the Brexit1 variable compared to the Brexit2 variable. Therefore, taking Brexit1 should be used in the regression because it results in more reliable results. All control variables decreased after the Brexit referendum date. Noticeably,
Operating Income to Deposits decreased from 6.82 to 2.09, dropping to lower than a third of its original value.
Table 4
Summary statistics before and after the Brexit referendum date in the time period 02.2015-02.2017
Pre (Before 06.2016) Post (After 06.2016)
Mean Standard Deviation Mean Standard Deviation
ROA 0.24 0.06 0.23 0.07 Tier1 to Deposits 2.03 0.59 0.78 0.45 OpInc to Deposits 6.82 2.14 2.09 1.43 NII to Deposits 4.24 1.30 1.47 1.02 NPL to Deposits 0.26 0.04 0.18 0.04 Loan to Deposits 147.50 43.29 69.78 48.24 Liquidity 26.22 1.01 25.72 1.13
Lastly, in Table 5, the summary statistics categorized by UK and non-UK banks are categorized. Interestingly, the variable to fixed remuneration ratio is 1.93 for UK banks and 0.53 for non-UK banks. This means that non-UK banks have an average Variable to Fixed remuneration ratio worth only a fourth of the average variable to fixed remuneration ratio that UK banks have. Furthermore, it is interesting that the Operating Income to Deposits is 6.00 for non-UK banks and 0.05 for UK banks. This might be due to the fact that the sample does not include a high number of UK banks.
Table 5
Summary statistics categorized by UK versus non-UK banks
Non-UK Banks UK Banks
Mean Standard Deviation Mean Standard Deviation
Variable to Fixed Remuneration Ratio 0.53 0.03 1.93 0.14 Treated 0.17 0.02 0.83 0.05 High Bonus 0.06 0.01 0.50 0.07 Biggest Half 0.46 0.02 0.83 0.05 ROA 0.23 0.05 0.36 0.09 Tier1 to Deposits 1.83 0.49 0.10 0.00 OpInc to Deposits 6.00 1.74 0.05 0.00 NII to Deposits 3.78 1.07 0.03 0.00 NPL to Deposits 0.25 0.03 0.03 0.00 Loan to Deposits 137.24 37.07 0.74 0.02 Liquidity 25.40 0.80 32.94 1.51
4.2 T-tests
Several t-tests are performed in order to establish significance for the differences noticeable in Table 4 and Table 5. The p-values of these t-tests can be seen in Table 6. In the case of the pre-Brexit versus post-Brexit t-tests, the hypotheses are that the post-Brexit variables are greater than, equal to or smaller the pre-Brexit. In the case of the UK banks versus non-UK banks, the hypotheses are that the values for the UK banks are either greater than, equal to, or smaller than the values for the non-UK banks. None of the differences in the control variables caused by the Brexit time variable are significant. Noticeably, all variables to fixed remuneration ratio variables are significantly higher for UK banks compared to non-UK banks.
Table 6
T-tests concerning the control variables and variable to fixed remuneration ratio, categorized by either the Brexit1 variable or the RegionUK variable
Post-Brexit versus Pre-Brexit
Smaller than Equal to Greater than
ROA 0.4629 0.9257 0.5371 Tier1 to Deposits 0.1034 0.2069 0.8966 OpInc to Deposits 0.0923 0.1847 0.9077 NII to Deposits 0.1031 0.2062 0.8969 NPL to Deposits 0.1425 0.2850 0.8575 Loan to Deposits 0.1529 0.3058 0.8471 Liquidity 0.3712 0.7425 0.6288
Non-UK Banks versus UK Banks
Smaller than Equal to Greater than
Variable to Fixed Remuneration Ratios 0.0000 0.0000 1.0000 Treated 0.0000 0.0000 1.0000 High Bonus 0.0000 0.0000 1.0000 Biggest Half 0.0000 0.0000 1.0000 ROA 0.1932 0.3865 0.8068 Tier1 to Deposits 0.8584 0.2833 0.1416 OpInc to Deposits 0.8485 0.3031 0.1515 NII to Deposits 0.8551 0.2897 0.1449 NPL to Deposits 0.9761 0.0479 0.0239 Loan to Deposits 0.8666 0.2668 0.1334 Liquidity 0.0032 0.0065 0.9968
4.3 Regression Results
The regression results with the CoVaR as the dependent variable are depicted in Table 7 over the time period 06.2015-06.2017.
Table 7
Regression results with CoVaR as dependent variables over the time period 06.2015-06.2017
CoVaR (1) (2) (3) (4) (5) (6) (7) ROA -0.001 (0.005) 0.000 (0.005) -0.000 (0.005) 0.002 (0.005) 0.001 (0.005) 0.000 (0.005) -0.001 (0.005) T1 to Deposits 0.029 (0.113) 0.020 (0.111) -0.012 (0.109) -0.004 (0.109) 0.006 (0.107) 0.004 (0.107) 0.076 (0.104) OpInc to Deposits 0.013 (0.039) 0.012 (0.039) 0.001 (0.038) 0.005 (0.036) 0.009 (0.037) 0.008 (0.037) -0.020 (0.036) NII to Deposits -0.042 (0.136) -0.035 (0.133) 0.003 (0.130) 0.010 (0.126) -0.023 (0.129) -0.019 (0.128) 0.079 (0.125) NPL to Deposits 0.068 (0.074) 0.089 (0.072) 0.091 (0.071) 0.119 (0.069) 0.104 (0.071) 0.098 (0.071) 0.106 (0.068) Loan to Deposits 0.000 (0.000) 0.000 (0.001) -0.000 (0.001) 0.000 (0.001) 0.000 (0.001) 0.000 (0.001) -0.000 (0.001) Liquidity 0.001 (0.001) 0.000 (0.001) 0.001 (0.001) 0.001 (0.001) 0.001 (0.001) 0.000 (0.001) 0.001 (0.001) Government Support 2011 0.034 (0.067) 0.075 (0.066) 0.081 (0.065) 0.049 (0.060) 0.051 (0.061) 0.026 (0.065) 0.090 (0.065) Treated 0.176*** (0.052) 0.159** (0.052) 0.121* (0.052) GSIB 0.221*** (0.053) 0.233*** (0.055) 0.252*** (0.055) High Bonus 0.019 (0.070) Biggest Half -0.051 (0.047) Brexit1 0.011 (0.006) 0.009 (0.005) 0.018*** (0.005) 0.010 (0.007) DID 0.044*** (0.011) 0.064*** (0.012) 0.035* (0.02) 0.024* (0.010) -Q2 0.011 (0.007) -Q1 0.002 (0.007) Q1 0.018* (0.007) Q2 0.011 (0.007) -Q2*Treated 0.070*** (0.015) -Q1*Treated 0.077*** (0.015) Q1*Treated 0.075*** (0.015) Q2*Treated 0.071*** (0.015) Within R2 0.000 0.002 0.076 0.106 0.057 0.058 0.206 Observations 495 495 495 495 495 495 495
First, the dependent variable is regressed on various control variables, similar to regression (1) stated in Part 3. After that, the Treated variable is added to the regression, similar to
regression (2) stated in Part 3. Additionally, the difference-in-difference regression is shown, similar to regression (3, 4, 5, 6) stated in Part 3. Lastly, a difference-in-difference regression based on period dummy variables is depicted, similar to regression (7) stated in Part 3. In Table A1 these same regressions are shown with the Systemic Risk to Market Capitalization as the dependent variable
Table 7 shows that none of the control variables have a significant effect on the Systemic Risk to Market Capitalizations ratio. However, all DID variables do show a
significant correlation to CoVaR. The difference-in-difference variables based on Treated and GSIB are both significant at a 0.1% significance level, whereas the difference-in-difference variables based on High Bonus and Biggest Half are significant at a 5% significance level. The GSIB variable is significant at a 0.1% significance level in every regression it is included. Interestingly, all interaction variables are significant in regression (7). Regression (7) has the highest within R2, 0.206, and regression (1) has the lowest within R2, 0.000.
5. Discussion
The results shown in Table 1 indicate the average result for various variables that are included in the regression analysis and show which variables should and which variables should not be included in the regressions. The number of firms after Brexit1, after the
referendum date, is bigger than the number of firms after Brexit2. Therefore, using Brexit2 as the Brexit variable is prone to make the regression analyses less reliable. As explained in Part 3, this is the reason why the variable Brexit1 is used in the regressions, instead of Brexit2.
The deviation in the number of firms that received government support in 2013 is bigger than the deviation in the number of firms who received government support in 2011. Therefore, the data for government support in 2011 is used as a control variable, rather than the data for government support in 2013. Furthermore, the standard deviation of the Systemic Risk to Market Capitalizations is many times smaller than the standard deviation of the maximum of the Systemic Risk to Market Capitalization. Therefore, as explained in Part 3, the Systemic Risk to Market Capitalizations is used as the dependent variable in the
regressions in the Appendix, rather than the maximum of that same variable.
In Table 4, the control variables differ when comparing the pre-Brexit and post-Brexit values. This result is not confirmed by the p-values shown in Table 6. This indicates that the control variables are not significantly different before Brexit compared to after Brexit. This is beneficial because this finding contributes to avoiding the influence of the Brexit variable on the control variable in the regressions. By doing this, it avoids collinearity in the regression part of this research. It is noticeable that, as shown in Table 5, the average percentage of banks that granted bonuses higher than 100% of the fixed remuneration level for non-UK banks is 17%, whereas the average percentage of banks that granted bonuses higher than 100% of the fixed remuneration level for UK banks is 83%. This indicates that UK banks are more affected by CRD IV and the corresponding legislation. It is interesting to note that the variable to fixed remuneration ratio is almost four times as high for UK banks, compared to non-UK banks.
In the second regression (2), the variable Treated is significantly correlated to the CoVaR at a 0.01% level. This indicates that the bonus structure does affect the risk structure of a bank. In the third regression (3), the DID variable is significant at a 0.01% level as well. In other words, the Brexit affects the level of risk a bank is taking positively. This is in line with the theory shaped in Part 2. Past research states that the more CEOs are rewarded for
taking risks, the higher the risks they are willing to take. The same result is visible for the difference-in-difference regressions when taking High Bonus or Biggest Half as the remuneration variable.
The difference-in-difference variable related to Biggest Half has the lowest correlation to CoVaR, whereas the difference-in-difference variable related to Treated has the highest correlation to CoVaR. This shows that the level of variable remuneration affects the level of risk a bank is willing to take. However, this finding also suggests that a certain threshold exists concerning the level of risk. After this threshold, a CEO might be incentivized to decrease the level of risk. The highest correlation exists between GSIB and CoVaR. This might indicate that the CoVaR is affected by the size of a bank, rather than its level of CEO compensation.
The highest value of within R2 is related to regression (7), the difference-in-difference
regression when taking the time period dummy variables into account. The lowest value of within R2 is related to regression (1). This makes sense because adding variables contribute to
the explaining power of regression. The mere increase in explaining power alone does not mean that regression (7) is a better model than regression (1). However, the fact that
regression (7) has a higher R2 does indicate that the extension of the baseline regression helps to explain the trend in a better way. Similarly, the fact that the regular difference-in-difference regressions all have a higher R2 than the baseline regression shows that the explaining power of those models is higher than that of the baseline model.
In regression (7), the difference-in-difference regression based on the period dummy variables is shown. The only time period dummy variable that is significant is the variable that indicates whether a time period is in the first quarter after the Brexit date. Interestingly, all interaction variables between the time period dummy variables and the DID variable are significant at a 0.01% level. This is not the result that was expected, as outlined in Part 3. An explanation for this occurrence could be that the positive correlation between Treated and CoVaR in the two periods before Brexit is bigger than the negative correlation between the period dummy variables and CoVaR in the two periods before Brexit. This could cause the interaction variable to be significant, while it should be insignificant.
The results of Table 7 are both similar and different from Table A1 in various ways. First of all, the difference-in-difference variable related to Biggest Half is significantly correlated to CoVaR at a 0.01% significance level in both regressions. In Table 7, the DID
variable is positively correlated, whereas, in Table A1, the DID variable is negatively
correlated. As mentioned in Part 3, CoVaR focusses on the additional risk a bank attributes to the financial system, whereas the Systemic Risk to Market Capitalization focusses on the additional risk a bank attributes to its own risk profile. This suggests that whether a bank is in the top 50% highest variable remuneration affects the bank's own risk structure positively but affects the additional risk that a bank contributes to the financial system negatively.
Therefore, it makes sense that certain variables have a different effect on these variables. Because Biggest Half is the only DID variable that is significant in Table A1, the difference-in-difference regression based on multiple period dummy variables is based on Biggest Half, instead of Treated. This regression shows a significant result for the 𝛾1 and 𝛾2
variables and an insignificant result for 𝛾−1 and 𝛾−2. This is according to the expectations set in Part 3. It indicates that the DID variable is significant after Brexit and insignificant before Brexit. The variables are negatively correlated to CoVaR, at roughly the same level as in regression (6). Interestingly, the DID variable has a stronger correlation in the first period after Brexit compared to the second period after Brexit. This suggests that Brexit had a strong initial effect on the risk profile. This initial effect decreased in the second period.
6. Conclusion
The difference-in-difference analyses show significant results for several DID
variables. Brexit has a significant effect on the additional risk a bank attributes to the financial system as well as on the additional risk a bank attributes to its own risk level. Brexit has a positive effect on the first and a negative effect on the latter. The level of variable to fixed remuneration affects this correlation. The analytical part of this research shows that Brexit can affect the risk structure of banks, as well as affecting the risk structure of the financial system. According to literature, it is not clear what UK legislation concerning variable remuneration will be after the UK has had a chance to change its remuneration legislation. It is possible that the UK abolishes the bonus cap. A bonus cap is meant to decrease excessive risk-taking by executives. However, past research has shown that this relationship is not necessarily true. My research concerning the impact of Brexit on the compensation regulation can be improved in multiple ways. Firstly, it could be improved by including a larger sample: only 75 banks were left after the data had been adjusted. In particular, a larger share of UK banks is likely to improve the reliability of this research. Furthermore, more recent data on the variable to fixed remuneration ratio could improve the quality of the research. Using data from 2013 does give a good indication for what the ratios were in the following years, yet more recent data could improve the quality of the research. Similarly, it would improve the quality of the research if more data is available concerning the way in which the variable remuneration is computed. This gives more insight into the variable remuneration and allows for a better quality of research. Future research should also focus on later Brexit dates. By doing this, the full effect of Brexit on the compensation regulation can be researched.
Several previous research papers have focused on the correlation between CEO compensation and risk and how this relationship is affected by a sudden shock. For example, Fahlenbrach & Stulz (2011) investigated the relationship between CEO incentives and bank performance during the Financial Crisis in 2008. Fahlenbrach & Stulz (2011) concluded that some evidence is present that suggests that the more CEO incentives and shareholder
incentives are aligned, the worse CEOs perform. Similarly, Gregg, Jewell, & Tonks (2012) researched the relationship between a company's performance and the cash compensation that is awarded to executives. They focused on to what extent this relationship is affected by a company being in the financial sector or not. Gregg, Jewell, & Tonks (2012) concluded that this relationship is not significantly higher for the financial sector in comparison to other sectors. Cerasi & Oliviero (2015) conducted research concerning the relation between CEO
incentives, financial regulation, and the level of risks banks take. They concluded that the combination of harmful regulation and a high level of CEO compensation was related to the poor financial performance of some banks during the Financial Crisis.
Several other papers have researched the influence of the changed CRD IV regulation. Cerasi et al. (2020) researched the influence of the changed regulations after the Financial Crisis on the level of CEO compensation. They concluded that the level of CEO
compensation is more strongly affected by the banks' risk after the regulations changed. The novelty of the research conducted in my paper is that it focusses on the effect of Brexit on the compensation regulation. This is done by both focusing on the theoretical aspect and on the evidence from the European banking sector.
My research adds the element of Brexit to this existing research. It investigates the relationship between risk structure and compensation structure and the effect of a sudden shock. Usually, this shock relates to the Great Financial Crisis in 2008. My research combines empirical analysis with existing literature on the possible effects of Brexit on the
compensation legislation. In conclusion, based on the literature review and the statistical research, Brexit is likely to influence the correlation between CEO compensation and performance in the banking industry.
Appendix Table A1
Regression results with CoVaR as dependent variables over the time period 06.2015-06.2017
CoVaR (1) (2) (3) (4) (5) (6) (7) ROA -0.001 (0.005) 0.000 (0.005) -0.000 (0.005) 0.002 (0.005) 0.001 (0.005) 0.000 (0.005) -0.040 (0.049) T1 to Deposits 0.029 (0.113) 0.020 (0.111) -0.012 (0.109) -0.004 (0.109) 0.006 (0.107) 0.004 (0.107) 4.417*** (0.564) OpInc to Deposits 0.013 (0.039) 0.012 (0.039) 0.001 (0.038) 0.005 (0.036) 0.009 (0.037) 0.008 (0.037) 1.592*** (0.207) NII to Deposits -0.042 (0.136) -0.035 (0.133) 0.003 (0.130) 0.010 (0.126) -0.023 (0.129) -0.019 (0.128) -5.566*** (0.696) NPL to Deposits 0.068 (0.074) 0.089 (0.072) 0.091 (0.071) 0.119 (0.069) 0.104 (0.071) 0.098 (0.071) 2.146*** (0.500) Loan to Deposits 0.000 (0.000) 0.000 (0.001) -0.000 (0.001) 0.000 (0.001) 0.000 (0.001) 0.000 (0.001) 0.027*** (0.004) Liquidity 0.001 (0.001) 0.000 (0.001) 0.001 (0.001) 0.001 (0.001) 0.001 (0.001) 0.000 (0.001) 0.007 (0.004) Government Support 2011 0.034 (0.067) 0.075 (0.066) 0.081 (0.065) 0.049 (0.060) 0.051 (0.061) 0.026 (0.065) 0.274 (0.244) Treated 0.176*** (0.052) 0.159** (0.052) GSIB 0.221*** (0.053) 0.233*** (0.055) 0.252*** (0.055) 0.564** (0.213) High Bonus 0.019 (0.070) Biggest Half -0.051 (0.047) 0.193 (0.184) Brexit1 0.011 (0.006) 0.009 (0.005) 0.018*** (0.005) 0.010 (0.007) DID 0.044*** (0.011) 0.064*** (0.012) 0.035* (0.02) 0.024* (0.010) -Q2 0.119 (0.107) -Q1 -0.054 (0.107) Q1 0.265* (0.107) Q2 0.319** (0.107) -Q2*Treated -0.005 (0.150) -Q1*Treated -0.039 (0.150) Q1*Treated -0.381* (0.150) Q2*Treated -0.353* (0.150) Within R2 0.000 0.002 0.076 0.106 0.057 0.058 0.1613 Observations 495 495 495 495 495 495 495
References
Adrian, T., & Brunnermeier, M. K. (2011). CoVaR (No. w17454). National Bureau of Economic Research.
Cerasi, V., & Oliviero, T. (2015). CEO compensation, regulation, and risk in banks: Theory and evidence from the financial crisis. International Journal of Central
Banking, 11(3), 241-297.
Cerasi, V., Deininger, S. M., Gambacorta, L., & Oliviero, T. (2020). How post-crisis
regulation has affected bank CEO compensation. Journal of International Money and
Finance, 102153.
Colonnello, S., Koetter, M. & Wagner, K. (2019). Compensation Regulation in Banking: Executive Director Behavior and Bank Performance after the EU Bonus Cap.
Available at SSRN 3104869.
Fahlenbrach, R., & Stulz, R. M. (2011). Bank CEO incentives and the credit crisis. Journal of
financial economics, 99(1), 11-26.
Financial Security Board. (2016, 21 November). 2016 list of global systemically important
banks (G-SIBs) [Press release]. Retrieved from https://www.fsb.org/2016/11/2016-list
of-global-systemically-important-banks-g-sibs/
Gregg, P., Jewell, S., & Tonks, I. (2012). Executive pay and performance: Did bankers' bonuses cause the crisis?. International Review of Finance, 12(1), 89-122.
Inman, P. (2009, July 31). Wall Street bonuses under fire. The Guardian. Retrieved from https://www.theguardian.com/business/2009/jul/31/wall-street-bonuses-bailout Köhler, M. (2015). Which banks are more risky? The impact of business models on bank
stability. Journal of Financial Stability, 16, 195-212.
compensation on firm performance–An application of a panel threshold regression model. The British Accounting Review, 45(3), 203-214.
Lu, L. (2016). The End of Bankers' Bonus Cap: How Will the UK Regulate Bankers' Remuneration after Brexit. European Business Law Review, 27, 1091.
Lui, D. A. (2018). Post-Brexit Britain and the pay culture: challenges and opportunities.
NILQ, 69(2), 107-25.
Magnan, M. L., & St‐onge, S. (1997). Bank performance and executive compensation: A managerial discretion perspective. Strategic Management Journal, 18(7), 573-581. Murphy, K. J. (2013). Regulating banking bonuses in the European Union: a case study in
unintended consequences. European Financial Management, 19(4), 631-657. Philip Clarke. (2017). Brexit timeline so far.(News)(Chronology). Farmers Weekly, (1181). Reifman, A., & Keyton, K. (2010). Winsorize. Encyclopedia of research design, 3,
