Regulatory bank bonus interventions and short-termism within the
banking sector
Dennis van Hofwegen – student number: 11852542
Economics and Business Economics
Finance
Ekaterina Seregina
15-07-2020
Abstract
This paper investigates whether the bank bonus interventions that were implemented in response to the Great Recession were efficient in reducing the short-termism in Europe. In addition, this paper analyzes whether the Brexit has an opposing effect on the short-termism relatively. Financial stability is important, mainly because the banking sector is proven to be highly interconnected. A failure of a systemically important bank could lead to a collapse within the global banking sector. Some studies criticize the effectiveness of bank bonus interventions. By contrast, this paper shows that the relevant bank bonus intervention studied was effective in reducing the individual risk-taking of banks and improving exposure to systemic risk. However, in line with other studies, this paper suggests that this process may be run more efficiently when implemented with capital regulations and restrictions on the
curvature of pay. Lastly, this paper provides evidence that the Brexit made banks in general less bounded to legislation leading to an increase in the short-termism along with a
deterioration of the exposure to systemic risk, indicating a reversed effect on bonus legislation.
Statement of Originality
This paper is written by Dennis van Hofwegen who declares to take full responsibility for the contents of this paper.
I declare that the text and the work presented in this paper 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.
1. Introduction
The Great Recession, which emerged in the United States and lasted from 2007 to 2010, has been one of the biggest shocks to the global economy since the Great Depression. It all started with the extreme high rates and ease at which (subprime) mortgages, very complex financial products (such as collateralized debt obligations) and other mortgage-related securities were sold (Grusky et al., 2011). This led to a high engagement in excessive risk taking which was mainly incentivized by attractive variable pay of bank managers (Bhagat & Bolton, 2014; Deyoung et al. 2013). The strong linkage of incentives with short-term engagement led to very high leverage ratios. Thereby, a very unstable financial industry was created. In addition to the interconnectedness of the financial sector, this was one of the main reasons why only a few shocks could lead to a destructive collapse of the global economy (Kleymenova & Tuna, 2018; Bennet et al. 2016).
People have been blaming the executive pays of banks incentivizing managers into excessive risk-taking as one of the causes for the Great Recession (Kleymenova & Tuna, 2018). Efing et al. (2015) conclude that bonuses provided by European banks were not a realistic product resulting from an appropriate trade-off between risk and return. Moreover, Anginer et al. (2018) state that privately optimal contracts that are in line with the interest of shareholders could still lead to socially excessive risk-taking. Therefore, some of the costs of excessive risk-taking were to be borne by taxpayers and banks their depositors which were not fully taken into account in the bank managers’ decision making.
The society was in need for governmental intervention and regulation on variable compensation provided in the financial industry. The United Kingdom was the first country to act accordingly by introducing the Remuneration Code, followed by the United States with its Dodd-Frank Act and the European Union with the Capital Requirements Directives (CRD) III and CRD IV where after many more restrictive laws and regulations were to follow. These will be explained briefly in further detail in the theoretical framework. The importance of a stable financial industry has also been proven by the way banks and other financial
institutions are treated differently from non-financial firms. The government has implemented safety nets for the important financial institutions. Examples of these safety nets are deposit insurance schemes and bailouts of “too-big-to-fail banks”, reflecting the systemic relevance of banks stress which can lead to consequences conflicting with the public interest (Acharya, 2009).
After years, there still seems to be conflicting findings regarding the effectiveness of the bonus regulations by the government on reducing the excessive risk-taking in the financial industry (DeYoung et al. 2013). For instance, Colonello et al. (2019) have been questioning whether banks in the new situation would still be able to incentivize enough to get to better levels of risk-taking after implementing regulations that restrict market mechanism from working. Kleymenova and Tuna (2018) state that certain bonus regulations led to unintended consequences, such as increased manager turnover because of the highly mobile
characteristics of the financial labor force industry. The aim of this paper is to find out
whether regulations on variable pay had a significant effect and which factors mattered for the significance and size of the impact. As a result, this paper can support or contradict current theories proposed by other studies. In particular, this paper studies the impact of the
announcement of the public consultation on the draft of CRD IV and the introduction of the Brexit.
This paper is in line with other researches that study the effect of regulations on excessive risk-taking within the financial industry by studying the effects on both equity and debt holders, measured by commonly used indicators such as cumulative abnormal returns and long run marginal expected shortfall respectively (Kleymenova & Tuna, 2018; Colonello et al. 2019). Moreover, this paper differs from current researches by studying the effect of the outcome of the Brexit referendum. Normally, researches look at events that are of restrictive effect on bonus payout policies. However, the UK leaving the EU may possibly revert such effects and may even produce spillover effects to other EU banks. Therefore, it is interesting to find out whether banks will act in an opposite way, or whether regulations have already led to full commitment in new policies that cannot easily be changed back.
Main findings of this paper indicate that high bonus banks are hit harder by the public consultation of CRD IV. This resulted in a deterioration of their performance of equity and an improvement of their exposure to systemic risk. Tests on the Brexit referendum suggest that it provides incentives for banks to increase short-termism as cumulative abnormal returns increased. Systemic risk increased alongside, though, for banks with var-to-fix ratios between 100% and 200% only. Thereby, it seems that the Brexit referendum has an opposing effect compared to that of the public consultation.
Throughout the rest of this paper, existing literature and different perspectives on regulatory bonus pay will be discussed and why this is relevant for this paper. After that, some descriptive information on the indicators and hypotheses are provided on which the
methodology will be based. Then, the findings will be presented and shortly discussed. Finally, a conclusion is formulated.
2. Related literature
A number of studies argue that in the case of financial distress, shareholders’ optimal response would be to incentivize and engage in excessive risk-taking that would lead to an “all or nothing situation”. As shareholders are limited liability owners, they would only bear the positive load of such activities and the costs were to be borne by banks’ creditors or long-term shareholders and in case of bailouts even by the taxpayer (Anginer et al., 2018;
Kleymenova & Tuna, 2018; Thanassoulis & Tanaka, 2018). Thereby, incentives were closely linked to short-termism which led to highly leveraged financial institutions (DeYoung et al., 2013). Hence, regulations on variable pay were mainly intended to reduce the excessive risk engagement as they are blamed to be one of the main reasons for the Great Recession (Klemeynova & Tuna, 2018). The great impact of this crisis on the global economy along with the necessity to rebuild a stable financial industry has led to many studies and researches closely monitoring and testing the effects of relevant regulations. While many regulations attempted to alter the alignment between incentives and excessive risk-taking, there still are some studies casting doubts on their effectiveness and show conflicting results. This chapter continues by describing some of the big regulations. Second, it discusses the main ideas and findings on the effects of the bonus caps by other studies. Third, why certain regulations seemed to have more impact relatively to others and to what extent bonus regulations led to other (unintended) consequences.
Many regulations introduced after the Great Recession were mainly focused on decreasing the maximal variable-to-fixed pay and ultimately the short-termism. Employees were financially awarded if they performed well, but were not “punished” in case of poor results, since bonuses cannot turn into negative values. Excessive risk taking could either lead to high returns or minimal losses for banking managers and the engagement in excessive risk taking was inevitable (Kleymenova & Tuna, 2018). The first country to react with a
regulation was the UK with its Remuneration Code. They already had an existing shareholder vote (say-on-pay regulation), but this was rather consultative than binding. The Remuneration Code implicitly required financial institutions to defer a larger amount of their bonus
payments which would become conditional on performance requirements. Higher rewards were made more uncertain and it implied a financial punishment for managers in case of bank failures (Thanassoulis & Tanaka, 2017). In 2010, the United States came up with the Dodd-Frank Act in response to the crisis. The Act mainly required banks to defer half of the bonuses which were to be paid out over a timespan of three years and to stop all kinds of incentive pay that were aligned with excessive risk-taking. Simultaneously, the European FSB reacted
accordingly with CRD III, which was a “clawback” rule and delayed the payments of bonuses by making it more dependent on long-term goals (similar to the UK remuneration code). In 2013 the European Parliament implemented CRD IV which capped the bonuses at a
maximum variable to fixed (var-to-fix) compensation ratio of 100% up to 200% upon
shareholders’ approval. In short, there were mainly two ways through which regulations tried to reduce the short-termism, either directly with a cap or indirectly by deferring variable pay.
Some studies done on bonus regulation within the EU argue whether they were effective in decreasing systematic and systemic risk. They state that regulating variable pay could lead to risk-averse managers becoming less risk averse. Some regulations are said to have created an “undesirable insurance mechanism”, since many financial institutions indemnified their managers for the bonus caps by fixing a relative higher pay by the amount the variable pay decreased (Carlson and Lazrak, 2010; Albuquerque et al., 2019; Colonello et al., 2019). Managers are not as incentivized to exert effort and caution as high as before (as they will not gain as much from this as before), most likely leading to reluctancy and less optimal decision making. So, exposure to systemic as well as systematic risk may increase after such regulations (Carlson & Lazrak, 2010). In addition, Colonello et al. (2019) found that risk-adjusted banks’ exposure to firm-specific risk increased, while their performance deteriorated. This is possibly due to the inefficient levels of exerted effort by managers which is stimulated by inferior incentives. A possible solution according to Kolm et al. (2017) would be to implement a compensation regulation accompanied by a capital regulation.
Compensation regulation prevents managers from proposing excessively risky projects but does not stimulate an active shareholder board necessarily to accept strategies that actually reduce risk. They speak of an underinvestment problem in projects that reduce risk taking. A way to reduce risk more effectively is to simultaneously implement capital regulations that decreases a bank’s ability to leverage and stimulates lower-risk investments.
However, many of the previously studies were mainly conducted on the EU as a whole, at the international level. Schäfer et al. (2013) argue that regulations on an
international level, where a lot of parties were involved in, would directly be updated within market expectations. Event-studies on such regulations might not be able to fully capture its effects and results might even be affected by other shocks and confounding events that occurred internationally. Moreover, regulations and shocks on a national level may also manipulate the results. It might be hard for such studies to isolate the field of study and make inferences about the effects of regulations on an international level. Besides, international
regulation often is less binding relatively to national regulations. Therefore, the total effect of regulations on a national level is easier and better captured by an event-study.
Kleymenova and Tuna (2018) studied the effect of the Remuneration Code on systemic risk for UK financial institutions. They were able to isolate their study of field and their findings might be more revealing. In short, the Remuneration Code tried to link incentives to long-run performance, by deferring bonuses. Kleymenova and Tuna (2018) compared UK financial institutions to other large UK firms, thereby controlling for macro-economic shocks. In turn, this is one of their limitations of their study, because both groups operate in different sectors. Therefore, they have also created two other control groups, which are similar sized/characterized banks from 1) the United States and 2) Europe. They
concluded that firms affected by the Code (the treated sample) played a relatively smaller part in their contribution to systemic risk and became less sensitive to systemic shocks than
before. These findings still hold when they compare their treated sample with large US banks of comparable size. Though, when compared with European banks, contributions to systemic risk did not change relatively. The Code succeeded in decreasing systemic risk, but their results also show that it was not able to decrease risk-taking, possibly due to increased shareholding requirements by managers imposed by shareholders. This would incentivize managers to increase equity value as much as possible. Therefore, Thanassoulis and Tanaka (2017) mention that clawback rules would be more effective when there are restrictions on the curvature of compensation. Initially, clawback rules would prevent managers from proposing risky projects as their financial rewards are dependent on the succession of the projects. However, shareholders could scale up risk-taking by incentivizing managers with pay that is convex in equity prices via granting equity options for instance. Hence, restrictions on the curvature of pay in addition to clawback rules would stop shareholders from being able to provide excessive risk-taking incentives.
The drawback of such a study might be its possible weak external validity. However, due to the interconnectedness of the banking sector, which is proven to be true if you look at the global collapse during the Great Recession, positive results of bonus regulations on a national level could contribute to a decrease in total systemic risk globally. Moreover, positive effects in one country might lead to positive spillover effects to other countries. On top of that, prove of a positive effect of regulation in a country could be an example and trigger for other countries to implement such regulations.
The financial industry is characterized by higher returns to skill relative to other sectors. Therefore, some researches state that the cap on bonuses also led to an unintended
consequence such as increased managerial turnover. Philippon and Reshef (2012) argue that an industry with such characteristics would have to deal a lot with the risk of not being able to retain their people when not rewarding them enough. The likeliness of turnover increases even more when incentives are weakened because managers within this competitive industry contain skills and knowledge that are highly portable. Therefore, personal switching costs in case of leaving are relatively low (Weinberg, 2001). However, Colonello et al. (2019) show that there is not enough evidence for a significant increased turnover of managers who are subject to heavier regulation. They rather link the dismissal of managers to underperformance than because of a dissatisfaction motive of managers (because of the lower pay), who would have then been said to leave voluntarily. Moreover, they have shown that managers who left after the bonus cap, mostly switched to an inferior function. Implying that in case of dismissal it has probably been not because of the decrease salary as they are not better off leaving since they are worse off salary-wise.
In conclusion, bonus regulations attempt to alter the short-termism of banks by either implementing bonus caps or bonus deferral policies. When capping bonuses, banks try to indemnify their managers for the reduction in variable pay by increasing the fixed amounts. This may lead to agency problems in which managers are not incentivized as much as before by creating an undesirable insurance mechanism. In addition, it may lead to an
underinvestment problem where shareholders are not willing to accept low risk low rewards investments. Also, shareholders can incentivize risk-taking by granting equity holdings to managers. Many studies point towards a solution where bonus regulations are implemented in combination with capital regulations and restrictions on pay curvature. Bonus regulations are also believed to lead to unintended consequences, mainly increased manager turnover. Theories are supporting this belief as the banking sector is very competitive which makes it hard to retain its managers. However, some studies provide contradicting results and blame the increased turnover to underperformance and even show that managers who leave, switch to inferior positions/jobs.
3. Data description and methodology
In general, the two most important groups of interest of banks are shareholders and creditors. Both are claimants of the total asset value of the bank but have different payoff functions. Shareholders are junior claimants and are more likely to stimulate risky projects with high rewards as their expected payoffs increase from this, since they only get the remains of the profits after creditors have been paid. In contrast, creditors hold senior claims and are more risk averse (Colonello et al., 2019). Especially in case of financial distress, this difference in risk-preference is very likely to generate agency conflicts between the two. To measure a regulation’s impact, this paper studies the performance of equity and the changes in risk around the introduction of the public consultation of the draft of CRD IV (26/02/2010) and the referendum of the Brexit (23/06/2016). This paper rather uses the announcement dates than the moments at which they are actually implemented, as the efficient market hypothesis states that information is constantly updated in prices. Hence, announcement dates would be considered more of an exogenous shock in comparison to the dates at which CRD IV and the Brexit are actually implemented. Which is often considered a slow process in which small pieces of information are updated in prices on a daily basis (Schäfer et al. 2013).
The impact on the equity side is generally measured by changes in the abnormal returns on equity (Schäfer et al., 2013). Firstly, normal returns are estimated over the estimation window with equation 1). From equation 1) the alfa and beta are retrieved, with which expected normal return on equity is computed over the relevant event window. Equation 2) measures the abnormal returns, which is the difference between actual return on equity during the event window and the expected normal return over the event period (De Haan et al., 2015). 𝑅!,# comprises the daily required rate of return on equity for bank i at time
t and 𝑅$,# is the return of the market portfolio at time t (MSCI return index). To estimate the required rate of return the Capital Asset Pricing Model is used. Finally, the abnormal returns over the event window are accumulated on which tests will be run to find out which factors mattered for the significance and size of the impact of the events on the cumulative abnormal returns (CAR).
𝑅!,# = 𝛼 + 𝛽%𝑅$,#+ 𝜀!,# (1)
Creditors are also known as debt holders and have priority claims on the returns of a bank for their debt repayments. The claims of creditors are bound to the obligations a bank has towards them. Therefore, creditors often benefit less from risky projects with high expected returns and sometimes do not even gain anything at all from riskier projects while their repayments become more uncertain. In general, they benefit from minimal exposure to risk. So, to measure the effects on the performance of the credit side the changes in systemic risk are examined. In particular this paper looks at the impact on the long run marginal expected shortfall (LRMES) as its systemic risk indicator. It is a commonly used indicator that measures how much resources a bank should possess to survive a systemic risk event (Acharya et al., 2017; Colonello et al., 2019). The LRMES is modelled according to the constant return model. This approach is consistent with analyses done on credit default swaps spreads that may be comparable to this analysis, since they also contain no market series (Campbell J. Y. et al., 1997; Schäfer et al., 2013). In contrast to the market model that is applied to the CAR, there will be no market return used during the estimation of the “normal level” of LRMES. Equation 3) represents the statistical hypothesis for this, where
𝐸_𝐿𝑅𝑀𝐸𝑆!,# is the estimated LRMES for bank i at time t and 𝜇! is the constant for firm i during the estimation window. With equation 4), the abnormal change of LRMES during the event is computed. Where 𝐴_𝐿𝑅𝑀𝐸𝑆!,# is the actual level of LRMES at time t for bank i.
𝐸_𝐿𝑅𝑀𝐸𝑆!,# = 𝜇! + 𝜀!,# (3)
𝐴𝐵_𝐿𝑅𝑀𝐸𝑆!,# = 𝐴_𝐿𝑅𝑀𝐸𝑆!,#− 𝜇! (4)
The estimation window is capped at 220 trading days and 30 days before the
announcement dates. This is proven to be of big enough size to estimate normality using daily returns according to MacKinlay (1997). Since it is hard to decide whether the total effect of an event could be best captured immediately or spread over a few days, robustness is tested of the results by running tests over varying event windows: [-1;1], assuming the event has an immediate total effect and [0;5], assuming the information needs some time to be processed and absorbed. Though, they should not be too big as confounding events could also
manipulate results or effects could become biased downwards as it may then capture more days without any news (Schäfer et al. 2013).
The dataset that is used for the empirical analysis comprises of European banks only. Since the CRD IV applies to every bank in the EU, this paper is not in the possession of a counterfactual sample of banks who would be completely unaffected by European legislation when studying the effect of the consultation of CRD IV. Therefore, two groups are separated by their var-to-fix compensation ratio. One group is considered to be insensitive to
regulations. They are considered to be the control group. These are the banks that have a var-to-fix ratio smaller than 100%. Regulations will probably not affect their policies and
performances as they are not obliged to cut down on variable pay (CRD IV cuts the max possible var-to-fix to 100% up till 200% upon shareholder approval). Whereas the treated group (also called high bonus banks) consists of banks who have a var-to-fix ratio of at least 100%. Generally, the regulations try to aim at these banks to decrease excessive risk-taking by trying to reduce its short-termism incentives. The statistical approach distinguishes between one another with dummy variables being either 0 or 1 (control and treated group respectively). However, the current situation may also find banks in the treated sample who do not really differ that much from the control group, for instance, banks with a var-to-fix ratio of 110%. This would lead to cuts of 10% in the variable pay and may only lead to weak changes in and effects on performances, potentially leading to a downward bias in the results of the effectiveness of regulations. Therefore, to get a clearer distinction between low variable paying banks and high variable paying banks and to reduce the amount of “false positives” within the treated group, tests are also run for dummies who divide the group between below and above 200%. But most importantly because this threshold is possible upon shareholder approval. So, running the tests for two groups separated from each other based on the 200% var-to-fix limit would completely terminate the possibility that the treated sample contains banks that actually are not directly affected by legislation and vice versa. When studying the effect of the referendum on Brexit a logical split would be made between banks either being from the UK or not, but the dataset contains over only 6 banks from the UK out of a total sample of 81 banks. Since the banking sector is highly interconnected and systemic banks even more, this paper may be able to measure the effect of the referendum on the Brexit on the same split between treatment and control group as before. However, by including a dummy variable for UK banks as well this paper tries to measure the impact of being a bank in the UK on equity and credit performance around the Brexit.
The main control variables are lagged, so that the predictor can use the
information/patterns retrieved from the varying amounts of the control variables in its forecast. Mainly bank-specific variables are controlled for, such as a bank’s total assets as a
size parameter; the net interest income (nii) as a way to measure profitability; the npl ratio, which shows which percentage of the bank’s outstanding loans are in default by its debtors. Lastly, tier1 capital to assets is controlled for, which reflects how much of a bank’s assets consists of valuable capital from a regulator’s point of view, the higher, the more stable a bank is considered by regulators. Kleymenova and Tuna (2018) even state that these are covariates of systemic risk, hence it is important to include them in the prediction. To prevent random noise from extreme observations, all the variables are winsorized at the 1st and 99th
percentile that were unnaturally skewed. More information on the variables is to be found in tables 1, 2 & 3.
The sample consists of 81 banks over a timespan from 2009 up until 2017 of which 21 have a var-to-fix ratio equal to or higher than 100% and 11 of those even have a ratio of at least 200%. Within the descriptive statistics table 3 panels A-D distinguishes between either control or treatment group and split the periods between before and after the public
consultation of CRD IV. In general, the table shows that before the event banks with a relative lower var-to-fix ratio are characterized by a lower LRMES, implying they are less sensitive to systemic shocks since they need less capital to withstand such shocks. Their lower average betas also show that they contain less systematic risk. By contrast, higher variable paying banks are of bigger size and generate higher profits in absolute terms. On average, they also have lower deposits to assets ratios and higher levels of liquidity. This is in line with the idea that higher variable paying banks are engaged in riskier activities, because, their relative levels of stable sources of funding (deposits to assets) are lower and their liquidity is higher due to the higher incentivized short-termism. For the period after the public consultation of CRD IV, we see a sharp decrease in the net income of high bonus banks whereas their liquidity remains nearly the same. A possible reason may be the underinvestment problem mentioned by Kolm et al. (2017), where shareholders do not necessarily accept risk low-return projects. Thereby remaining the liquidity level by not actively decreasing the short-termism. And on the other hand, due to underinvestment, income decreases in absolute terms. However, the deposit ratios increased, which shows that banks did partially engage in saver activities after regulations. The LRMES increases for all banks after the event, and even harder for low bonus banks. The general idea that exposure to systemic risk has not decreased as well as an increase in systematic risk represented by higher betas and deterioration of bank performance is in line with studies suggesting that bonus regulations are not implemented correctly and create an undesirable insurance mechanism for managers. Therefore, in the following chapter will be tested whether banks with high var-to-fix ratios (the main targets of
bonus regulations) are hit effectively by regulations, measured by changes in equity and credit performance or that we find adverse results in line with previous related studies. More
interestingly, whether the announcement of Brexit has a reversive effect on bonus regulations will also be tested. Since the Brexit makes the UK independent from EU bonus legislation and might produce spillover effects to systemic banks in the rest of EU.
4. Results & discussion
In this chapter the results of the empirical analysis are presented and discussed. Starting off with the impact of both events on equity performance around the events, which is measured with changes in the CAR. This paper tests whether the CARs contained significant values; analyzed whether the var-to-fix ratios of banks played a role in the impact of the event on the performance of their CAR and included more controls and varied around event windows to check whether the results would still hold. The same regression tests were run on LRMES but based on the constant return model instead of the market model to investigate whether being a high bonus bank mattered for the changes in LRMES around the same events. Finally, the drawbacks of this paper are discussed.
Since bonus regulations ultimately aim at reducing the short termism by restricting its incentives and possibilities, banks will probably have to change their activities accordingly. Consequently, it is expected that their returns will decrease and that the CAR will be zero or lower around the public consultation of CRD IV. Because high bonus banks are characterized by their higher engagement in short-term activities, it is expected that they will be hit even harder. Table 4, Panel B provides significant evidence for a reduction in returns during the event, as the CAR indicates a significant negative value, meaning that their actual returns during the event were unnaturally lower than during “normal” times. For low bonus banks the CAR is of positive value (Table 4, panel B: ratio<200%) as they are not particularly hit by the event that may restrict them from outperforming normal performance. When the split is made at 100% between bonus banks (Table 4, panel A), not any revealing difference in the CARs is found. However, they do differ from each other because the bonus banks who are in between 100% and 200% ratio possibly push the CAR harder up than the banks with more than 200% ratio drag it down. Simultaneously, banks with a ratio of below 100% have lower abnormal returns than banks between the 100% and 200% ratio, as they are considered to be safer banks, thus lower (abnormal) returns during the event. Table 4, panel C and D show the CARs around the referendum of the Brexit. A Brexit would free the UK from European legislation. Hence, ceteris paribus, UK banks may engage in more risky activities after the event that might lead to higher returns than before. Because the banking system is
interconnected and systemic banks are even more, spillover effects may even take place to other EU banks. Thus, in contrast to the public consultation of CRD IV, we now expect a CAR of at least zero or higher. Table 4, panel C and D support this expectation as the CAR for high bonus banks is positive and higher than for low bonus banks. As high bonus banks were the ones actually affected and targeted by regulations, their restrictions will now be
more reduced relatively, and they will be the ones who benefit more from this. According to these results, it seems that the Brexit referendum does indeed have a reversed effect on the CAR in comparison to that of the public consultation. The question follows, which factors actually mattered for the significance and size of the impact on CAR and the LRMES.
4.1 The effect of the public consultation on CAR and systemic risk
The aim of this paper is to investigate whether the bonus regulations were efficient in reducing risk in the banking sector. This section will discuss which factors mattered for the impact of the public consultation of CRD IV on the CAR and LRMES. Based on that, there will be argued whether the regulation was efficient. Reduction in the risk engagement is accompanied by a decrease in returns on the equity side, ceteris paribus. Table 5 panel A and B contain the results of the OLS regressions that test whether banks with a high var-to-fix ratio were hit harder than low bonus banks. This is measured by the effects of the var-to-fix ratios on the CARs during the event, controlled for bank specific variables. Column 2 and 4 of Panel A both show that being a high bonus bank (>=200% var-to-fix) during the event, had a significant negative effect on the CAR. Though, this effect is weakened by the significant positive coefficient of the interaction term. But overall, high bonus banks are negatively hit by the event which forces banks to reduce their risky activities (represented by a decrease in equity returns). According to column 1 and 3 of panel A, the split made between high and low bonus banks on 100% does not provide significant results. Most likely the cause being that the vast majority of both groups are not directly forced to change policies by the event, thereby their returns are not per se abnormally different. The output of panel A assumes that the event is processed and absorbed in prices immediately. However, this may not be the case and prices may even be more affected by the event. Therefore, panel B tests for robustness of the result by running the OLS regression on a larger event window. Following panel B, column 2 and 4 still prove that being a high bonus bank has a significant negative effect on the CAR during the event, though, at a stronger magnitude when compared to panel A. Yet again, these effects are weakened by the interaction term. The interaction terms in column 1 and 3 of panel B show a significant positive effect on the CAR during the event but are canceled out by the negative impact of the var-to-fix ratio on the CAR. The negative sign of the coefficient of the dummy variable (“X>=ratio”) still indicates that being a “high” bonus bank has a negative impact on the CAR, however they are insignificant. In terms of signs and relationships, the results are robust over varying event windows, however, the effect of the event seems to be stronger over a longer event window as being a high bonus bank (ratio>=200%) becomes of a
higher negative effect on the CAR. In conclusion, the results support findings of studies that high bonus banks’ performance deteriorated around the events, as their usual levels of short-termism are restricted by decreasing its incentives.
Besides effectively decreasing the engagement in short termism for high bonus banks, it is interesting and important to find out whether exposure to systemic risk improved
alongside. As legislation tries to improve financial stability, LRMES should improve after the event. E.g. a possible decrease in the LRMES for the treated banks. In table 6, column 2 and 4 of panel A show that being a high bonus bank (ratio>=200%) has a significant negative effect on the LRMES during the public consultation of the CRD IV. This means that the expected loss of equity during a systemic shock would be significantly lower for high bonus banks. However, when the split is made on a 100% ratio between high and low bonus banks, only column 1 and 3 provide evidence for significant positive impacts on the abnormal levels of LRMES during the event. This may be due to the fact that banks between a ratio of 100% and 200% mainly are negatively affected in terms of their LRMES for a reason that may be beyond the effect of the event(thus their LRMES increased above normal levels during the event). Panel B tests for robustness over a varying event window. In terms of signs and relationships they provide the same significant results. However, the magnitude of the
positive coefficients of the high/low bonus bank dummy variable in column 1 and 3 increased whereas they decreased for column 2 and 4 relative to the ones in panel A. All in all, its findings suggest that the public consultation had its desired effect on high bonus banks. Though, low bonus banks seemed to endure a higher exposure to systemic risk during the event. Whether this is due to the event is questionable, as low bonus banks are not the ones particularly hit by the (future) regulations of the CRD IV.
In conclusion, results show that the public consultation of CRD IV was effective on high bonus banks. Firstly, short-termism seemed to have decreased during the event, as being a high bonus bank had a significant negative impact on their CARs. Secondly, banks with a high var-to-fix ratio (ratio >=200%) were proven to be of negative impact on the cumulative abnormal levels of LRMES, thus a decrease in systemic risk. Though, it seems that being a low bonus bank has an increasing effect on the levels of LRMES (deterioration of systemic risk). This may be due to a reason which may be beyond the scope of this paper.
4.2 The effect of the Brexit referendum on CAR and systemic risk
Most studies on bank bonuses look at restricting events towards bonuses, however, little research is done on events that may offset the effects of such events, for instance bank bonus
regulations. This section will discuss the results of the Brexit referendum again on the CAR and LRMES of banks. This event may be of opposite effect compared to the public
consultation as it sets UK banks free from European legislation. Table 5, panel C and D present the outputs of the tests done on the impact of relevant factors on CAR, that may be the result of the Brexit referendum. Panel C shows that being either a high or low bonus bank seems to be of no real effect (as most coefficients are nearly zero) on the CARs during the event. Moreover, all values are insignificant. However, the “GB” dummy variable shows a significant negative effect in the first two columns. This indicates that being a UK bank around the referendum negatively affected its equity returns. Which is in opposition to what was thought that would happen. However, according to Breinlicht et al. (2018) this may not have much to do with UK banks becoming more independent from bonus regulations, but because market participants act according to the future (negative) impact of Brexit on its economy. Switching to panel D, where tests are run on a larger event window, provides positive significance of the high/low bonus banks dummy variable in column 1, 2 and 4. This may indicate a delayed reaction by the market on the referendum. Since the GB indicator becomes insignificant in all columns, the negative impact of being a UK bank in panel C may have indeed been a (negative) shock due to future economic uncertainty for UK banks (which was of short duration to be restored). Then, the positive effect on CAR (column 1, 2 and 4) being a high bonus bank may be the actual result of UK banks becoming relatively less restricted by regulations and engage in more profitable, riskier activities. Due to the interconnectedness of the banking sector, some of these effects (mainly for >=200% ratio banks, as their coefficients are of greater magnitude) may be spilled over to European Banks. Thence a stronger effect and the positive signs in column 1, 2 and 4 for “X>=ratio”.
So far, it seems that the Brexit referendum does have an opposite effect for high bonus banks on CAR relative to the public consultation of CRD IV. Table 6, panel C and D tested the effects of the referendum on the LRMES. Column 2, 3 and 4 of panel C show a significant increase in the LRMES for UK banks specifically. Column 1 and 3 also provide significant results for banks (mainly with a bonus ratio between 100% and 200%) of a deterioration of the LRMES. Overall, LRMES increased for “high” bonus banks (mainly with ratios between 100% and 200%) and banks from the UK, which is in line with the expectations that banks become less restricted and engage in riskier activities. However, as stated previously, the effect of the referendum seemed to be delayed on the CAR. Hence, stronger effects were possibly found in panel D (larger event window). All columns indeed show a larger increase in the LRMES for UK banks. Also, column 1 and 3 show an increase for high bonus banks,
but again mainly for banks between 100% and 200% bonus ratios as the values of the “X>=ratio” turn negatively insignificant in column 2 and 4. Indicating that the positive
coefficients does not come from bonus banks with ratios above 200%. This may be logical, as banks with ratios above 200% were already engaging in risky activities at a level they were satisfied with. Hence, for them to become less restricted does not really affect their current activities.
In short, results suggest that the Brexit referendum indeed had an opposite effect on the CARs for high bonus banks compared to the public consultation of CRD IV. This may imply that the event increased the short-termism by making bonus policies relatively less restricted. Also, the LRMES increased simultaneously, though mainly for bonus banks with ratios between 100% and 200%. Lastly, it turns out that UK banks in particular became systemically riskier.
The drawbacks of the research of this paper are mainly about the sample used and the conduction of the counterfactual sample. Firstly, the sample of banks used mainly consists of banks with a var-to-fix ratio below 200%. Therefore, the effects that occurred due to being a bonus bank with a ratio of above 200% might have been biased downwards. A better sample would have been distributed more equally according to the to-fix ratios. Besides, the var-to-fix ratios do not comprise of continuous values and are only used to distinguish between either high or low bonus banks. Whereas in reality, banks change their var-to-fix ratios (according to bonus regulations). Therefore, this paper is not able to estimate the effects of bonus regulations on the bonus policies of banks. Next to analyses on changes in CARs and LRMES, this can be considered important for someone to argue whether bonus regulations were effective. As stated by Schäfer et al. (2013), an event study on an international regulation may lack ability to capture its full effects. This may also apply to the statistical analysis of this paper. Moreover, if the database contained more banks from the UK to compare against a counterfactual sample of EU banks, effects of the Brexit referendum on CAR and LRMES could have been captured more effectively. Lastly, the objective of this paper is to decide whether regulations were efficient in reducing the short-termism. It was not able to properly measure this, as many studies argue that bonus regulations lack of either capital regulations or restrictions on pay curvature. This paper suggests that one should compare the effects of a “simple” bonus regulation to the effects of a bonus regulation that is accompanied with either a capital regulation or a restriction on pay curvature. Thereby, efficiency might be better defined and measured.
5. Conclusion
The aim of this paper was to investigate whether bank bonus regulations were efficient in reducing the short-termism within the banking industry. In particular, which factors mattered for the significance and size of the impact of the events. This paper looked at the effects of the public consultation of CRD IV and the Brexit referendum on both the cumulative abnormal returns and the long run marginal expected shortfall.
The general perception of studies that question the effectiveness of bank bonus regulations is that these regulations would create an undesired insurance mechanism for managers. Ultimately, this would lead to an increase in risk engagement and exposure to systemic risk which is the opposite desired result of bank bonus regulations. However, the findings of the empirical analysis of this paper suggest that the public consultation of the CRD IV did have its desired effect decreasing the short-termism. The CARs decreased for banks with a var-to-fix ratio of at least 200% indicating a decrease in the short-termism and at the same time, their LRMES improved (systemic risk decreased). This was not the case for banks with var-to-fix ratios below 200%, of which their exposure to systemic risk even increased. Indicating that the significance of the positive impact of the events comes from being a bank with a bonus ratio of above 200%, as they are mainly considered the ones to be hit by the events. These findings indicate that bank bonus regulations effectively decreased short-termism of their targeted group (high bonus banks) by decreasing its incentives, which are mainly considered to be the excessive bonus pay. However, the question whether this process could have been run more efficiently remains unanswered. Other studies do argue that bonus regulations could be implemented more effectively in line with capital regulations and restrictions on pay curvature, suggesting current regulations could be run more efficiently.
A Brexit would dissolve UK banks from European legislation and among other create a situation with relatively less safety nets for UK banks. Therefore, the Brexit referendum would make the UK banking sector more uncertain which typically results in a decrease in equity performance and an increase in systemic risk exposure. The empirical analysis on the Brexit referendum produced somewhat mixed results, but overall, it seemed that the
referendum had an opposing effect compared to the CRD IV. CARs increased for high bonus banks, indicating that banks short-termism incentives were increased relatively. However, the LRMES increased (deteriorated) for banks with bonus ratios between 100% and 200% and banks from the UK only. Whereas tests produced insignificant values for bonus banks above the 200% threshold and their levels of systemic risk seemed insensitive to the referendum. Finally, these effects were only able to be measured over a delayed time.
References
Acharya, V. V. (2009) “A theory of systemic risk and design of prudential bank regulation.” Journal of Financial Stability 5:224–225
Anginer, D., Demirgüç-kunt, A. Huizinga, H., & Ma, K. (2018) “Corporate Governance of Banks and Financial Stability.” Journal of Financial Economics 130 (2018):
327-346.
Albuquerque, R., Cabral, L., & Guedes, J. (2019) “Incentive Pay and Systemic Risk.”
Review of Financial Studies (Forthcoming).
Bennett, B., Gopalan, R., & Thakor, A. (2016) “The Structure of Banker’s Pay.” Working
Paper, Ohio State University.
Bhagat, S., & Bolton, B. (2014) “Financial Crisis and Bank Executive Incentive Compensation.” Journal of Corporate Finance 25:313–341.
Breinlich, H., Leromain, E., Novy, D., Sampson, T., & Usman, A. (2018) “The economic effects of Brexit: evidence from the stock market.” Fiscal Studies, 39(4), 581-623.
Carlson, M., and A. Lazrak. (2010). “Leverage Choice and Credit Spreads when Managers Risk Shift.” Journal of Finance 65:2323–2362.
Campbell, J. Y., Champbell, J. J., Campbell, J. W., Lo, A. W., Lo, A. W., & MacKinlay, A. C. (1997). “The econometrics of financial markets.” princeton University press.
Colonnello, S., Michael K., & Konstantin W. (2019) "Compensation Regulation in Banking: Executive Director Behavior and Bank Performance after the EU Bonus Cap." Available at SSRN 3104869.
De Haan, J., Neretina, E., & Sahin, C. (2015) “Banking Stress Test Effects on Returns and Risks.” Journal of Banking and Finance 117.
DeYoung, R., Peng E. Y., & Yan, M. (2013) “Executive Compensation and Business Policy Choices at U.S. Commercial Banks.” Journal of Financial and Quantitative
Analysis 48:165–196.
Efing, M., Hau, H., Kampkötter, P., & Steinbrecher, J. (2015) “Incentive Pay and Bank Risk-Taking: Evidence from Austrian, German, and Swiss Banks.” Journal of
International Economics 96:123–140.
Grusky, D. B., Western, B., & Wimer, C. (Eds.). (2011) “The great recession.” Russell
Sage Foundation.
Kleymenova, A., & Tuna, A. (2018) "Regulation of Compensation and Systemic Risk: Evidence from the UK." Chicago Booth Research Paper 16-07.
Kolm, J., Laux, C., & Lóránth, G. (2017) “Bank Regulation, CEO Compensation, and Boards.” Review of Finance 21:1901–1932.
MacKinlay, A. C. (1997) “Event studies in Economics and Finance.” Journal of
Economic Literature 35 (1), 13–39.
Mayer, C., Pence, K., & Sherlund, S. M. (2009) "The Rise in Mortgage Defaults." Journal
of Economic Perspectives, 23 (1): 27-50.
Philippon, T., & Reshef, A. (2012) “Wages and Human Capital in the U.S. Finance Industry: 1909-2006.” Quarterly Journal of Economics 127:1551–1609.
Schäfer, A., Schnabel, I., & Weder di Mauro, B. (2013) “Financial Sector Reform After the Crisis: Has Anything Happened?” CEPR Discussion Paper No. DP9502.
Available at SSRN: 2278079.
Tanaka, M., & Thanassoulis, J. (2017) “Optimal Pay Regulation for Too-Big-To-Fail Banks.” Bank of England Working Paper No. 558
Weinberg, B. A. (2001) “Long-Term Wage Fluctuations with Industry-Specific Human Capital.” Journal of Labor Economics 19:231–264.
Appendix
Table 1: Variable definitions.
Variable Definition Source
Beta Represents the sensitivity to systematic risk. Vlab
LRMES
Long Run Marginal Expected Shortfall shows how much
capital an institution requires to survive a systemic shock. Vlab
Log returns The logarithm of stock returns. Datastream
CAR
Cumulative abnormal return is the difference between actual return during event window and estimated normal return.
Computed by author
Vartofix Variable to fixed compensation ratio. SNL financial
GB Dummy variable indicating whether bank is from the UK. ROA Return on assets.
SNL financial
Tier 1
capital Is a bank's primary source of funding. An important indicator for financial stability from a regulator's point of view. SNL financial
NPL
Non-performing loans are loans that are in default by its borrowers. The NPL ratio tells you how much of a bank's
outstanding loans are in default. SNL financial
Liquidity
The ease in which a security or asset can be turned into money without influencing its market price.
SNL financial
Deposits
Money held at bank, considered to be a stable funding source. The ratio deptoas tells you how much of the assets consists of
deposits. SNL financial
Loan Outstanding loans.
SNL financial
Netinc Net income. SNL financial
Nii
Net interest income, the differnce between interest receivements and payments.
SNL financial Opinc Operating income, income after operating expenses deducted.
SNL financial Otherinc Other income.
SNL financial
Divers
Indicator [0;1] showing how much income sources are diversified. Reaching the extreme values means only few different sources, reaching the middle means well balanced.
Computed by author T1toas Tier 1 capital to assets.
Computed by author
Table 2: Variable-to-fixed compensation ratios.
This table shows the differing averages of variable-to-fixed compensation ratios amongst all control and treatment groups. These values are mainly used to distinguish between the control and treatment group.
Variable No. of banks Mean S.D. Min .25 Mdn .75 Max
Var-to-fix ratio Ratio>=100% 21 2.16 0.77 1.00 1.54 2.01 2.77 3.56 Ratio<100% 60 0.21 0.29 0.00 0.00 0.03 0.37 0.97 Var-to-fix ratio Ratio>=200% 11 2.81 0.45 2.01 2.42 2.77 3.24 3.56 Ratio<200% 70 0.39 0.53 0.00 0.00 0.12 0.62 1.98 Overall sample 81 0.72 0.98 0.00 0.00 0.27 1.08 3.56
Table 3: Descriptive statistics
This table provides an overview of the statistics of the dependent and control variables used during the statistical analyses for all the different groups, control and treatment, separately. All values are denominated in Euros and the split in periods are done around the public consultation of CRD IV. To prevent random noise from extreme observations, all variables are winsorized at the 1th and 99th percentile. Panel A, B, C & D show the descriptive
statistics of all variables for banks with a var-to-fix ratio below 100%; above and including 100%; below 200% and above and including 200% respectively. “N” indicates the number of banks that are within the relevant group. All variables are defined in table 1.
Panel A: Banks with a var-to-fix ratio < 100% (N=60).
2009- 2010 2011- 2015
Variable Mean S.D. Min Mdn Max Mean S.D. Min Mdn Max Lrmes 0.39 0.17 0.01 0.42 0.78 0.42 0.18 0.00 0.44 0.78 Beta 1.06 0.56 0.02 1.05 3.01 1.15 0.64 0.01 1.13 3.01 LogRet 0.00 0.03 -0.08 0.00 0.08 -0.00 0.03 -0.08 -0.00 0.08 Tier1 7.39 11.89 0.04 2.38 62.45 8.55 13.37 0.04 3.21 73.48 Assets1 177.57 308.66 3.38 54.15 1737.37 181.44 314.88 3.40 58.17 1907.05 Roa 0.34 0.68 -3.84 0.43 2.24 0.10 1.08 -3.84 0.38 2.24 Dep1 66.59 118.09 0.02 15.49 616.38 73.65 134.02 0.00 19.59 700.31 Nii1 4.35 7.35 0.04 0.89 32.49 4.45 8.04 0.00 0.87 32.49 Netinc1 0.95 2.99 -7.25 0.16 14.06 1.13 3.94 -7.25 0.13 14.78 Npl1 6.10 9.02 0.00 2.52 54.72 8.70 11.72 0.01 3.00 62.84 Liq 0.27 0.15 0.09 0.24 0.95 0.26 0.15 0.06 0.24 0.95 T1toas 0.06 0.02 0.00 0.06 0.13 0.06 0.02 0.00 0.06 0.16 Npltoas 0.03 0.02 0.00 0.03 0.09 0.07 0.09 0.00 0.03 0.57 Lognpl -3.79 0.96 -7.71 -3.69 -2.45 -3.50 1.53 -8.04 -3.38 -0.56 Logassets 17.83 1.56 15.03 17.81 21.28 17.94 1.49 15.04 17.88 21.37 Deptoas 0.49 0.22 0.00 0.51 0.94 0.51 0.23 0.00 0.54 1.92 Divers 0.43 0.25 0.08 0.41 2.35 0.41 0.17 0.02 0.40 1.26
Panel B: Banks with a var-to-fix ratio >= 100% (N=21).
Panel C: Banks with a var-to-fix ratio <200% (N=70).
2009- 2010 2011- 2015
Variable Mean S.D. Min Mdn Max Mean S.D. Min Mdn Max Lrmes 0.41 0.16 0.01 0.43 0.78 0.43 0.18 0.00 0.46 0.78 Beta 1.12 0.56 0.02 1.12 3.01 1.21 0.64 0.01 1.21 3.01 LogRet 0.00 0.03 -0.08 0.00 0.08 -0.00 0.03 -0.08 -0.00 0.08 Tier1 9.33 14.28 0.04 3.20 68.54 10.53 16.00 0.04 3.70 76.85 Assets1 240.58 416.98 3.38 54.47 1998.16 246.83 422.41 3.40 60.49 1998.16 Roa 0.37 0.66 -3.84 0.43 2.24 0.12 1.04 -3.84 0.35 2.24 Dep1 86.08 143.11 0.02 21.36 616.38 94.12 156.42 0.00 23.73 700.31 Nii1 4.86 7.53 0.04 1.28 32.49 4.90 8.06 0.00 1.17 32.49 Netinc1 1.11 2.99 -7.25 0.19 14.06 1.19 3.79 -7.25 0.14 14.78 Npl1 7.11 10.12 0.00 2.71 54.72 9.45 12.14 0.00 3.66 62.84 Liq 0.29 0.17 0.09 0.25 0.95 0.29 0.17 0.06 0.25 0.95 T1toas 0.06 0.02 0.00 0.06 0.13 0.06 0.02 0.00 0.06 0.16 Npltoas 0.03 0.02 0.00 0.02 0.10 0.07 0.08 0.00 0.03 0.57 Lognpl -3.89 1.37 11.24 -3.72 -2.26 -3.58 1.76 11.62 -3.41 -0.56 Logassets 18.03 1.65 15.03 17.81 21.42 18.12 1.59 15.04 17.92 21.42 Deptoas 0.47 0.21 0.00 0.50 0.94 0.50 0.23 0.00 0.54 1.92 Divers 0.43 0.24 0.08 0.41 2.35 0.41 0.17 0.02 0.40 1.26 2009- 2010 2011- 2015
Variable Mean S.D. Min Mdn Max Mean S.D. Min Mdn Max Lrmes 0.50 0.09 0.13 0.51 0.78 0.51 0.11 0.00 0.51 0.78 Beta 1.38 0.37 0.27 1.40 3.01 1.44 0.47 0.01 1.39 3.01 LogRet 0.00 0.03 -0.08 0.00 0.08 -0.00 0.02 -0.08 -0.00 0.08 Tier1 29.67 25.68 0.65 27.25 82.15 31.10 26.86 0.04 33.94 82.15 Assets1 789.62 704.92 6.68 694.66 1998.16 767.27 701.00 7.33 621.13 1998.16 Roa 0.51 0.50 -0.33 0.44 2.01 0.32 0.78 -3.84 0.28 2.24 Dep1 298.77 224.97 3.24 265.35 700.31 311.50 223.64 0.00 330.14 700.31 Nii1 9.99 8.14 0.20 10.02 29.47 9.37 7.83 0.28 9.09 31.39 Netinc1 2.33 2.99 -2.58 1.75 10.60 1.41 4.06 -7.25 0.96 13.88 Npl1 15.00 18.91 0.01 3.66 62.84 15.33 18.88 0.00 6.19 62.84 Liq 0.43 0.17 0.17 0.48 0.65 0.43 0.17 0.11 0.49 0.71 T1toas 0.05 0.03 0.02 0.04 0.12 0.05 0.02 0.00 0.05 0.12 Npltoas 0.03 0.03 0.00 0.02 0.10 0.04 0.04 0.00 0.02 0.23 Lognpl -4.39 2.01 -11.24 -3.76 -2.26 -4.38 2.14 -11.62 -3.87 -1.49 Logassets 19.49 1.89 15.71 20.36 21.42 19.46 1.85 15.81 20.25 21.42 Deptoas 0.38 0.13 0.10 0.39 0.62 0.42 0.15 0.00 0.39 0.77 Divers 0.51 0.16 0.12 0.51 0.84 0.50 0.17 0.19 0.50 0.81
Panel D: Banks with a var-to-fix ratio >200% var-to-fix ratio (N=11).
2009- 2010 2011- 2015
Variable Mean S.D. Min Mdn Max Mean S.D. Min Mdn Max Lrmes 0.48 0.09 0.22 0.49 0.78 0.48 0.10 0.18 0.47 0.78 Beta 1.30 0.34 0.49 1.33 3.01 1.31 0.39 0.38 1.24 3.01 LogRet 0.00 0.03 -0.08 0.00 0.08 -0.00 0.02 -0.08 0.00 0.08 Tier1 36.86 27.55 0.65 30.12 82.15 39.18 27.15 0.71 41.41 82.15 Assets1 949.90 678.77 6.68 928.76 1998.16 893.68 690.47 7.33 860.43 1998.16 Roa 0.52 0.54 -0.11 0.49 2.01 0.42 0.80 -1.55 0.28 2.24 Dep1 363.91 215.59 3.24 344.22 700.31 366.95 214.29 3.81 401.47 700.31 Nii1 11.28 8.27 0.20 10.02 29.47 10.35 8.00 0.28 9.09 31.39 Netinc1 2.43 3.15 -2.58 2.33 10.60 1.24 4.96 -7.25 0.68 13.88 Npl1 17.73 22.63 0.05 3.64 62.84 17.14 22.59 0.03 4.70 62.84 Liq 0.46 0.13 0.22 0.51 0.64 0.47 0.14 0.15 0.50 0.64 T1toas 0.05 0.03 0.02 0.05 0.12 0.06 0.02 0.02 0.05 0.12 Npltoas 0.03 0.02 0.00 0.01 0.06 0.02 0.02 0.00 0.01 0.08 Lognpl -4.39 1.41 -6.98 -4.25 -2.77 -4.72 1.50 -7.70 -4.50 -2.58 Logassets 19.78 2.02 15.71 20.65 21.42 19.70 1.93 15.81 20.57 21.42 Deptoas 0.39 0.12 0.23 0.40 0.62 0.45 0.14 0.24 0.42 0.77 Divers 0.59 0.14 0.41 0.57 0.84 0.56 0.14 0.27 0.54 0.81
Table 4: Significance tests for CAR.
This table provides significance test results for the cumulative abnormal returns around both events, for each of the control and treatment groups used throughout the empirical analyses. Significance is tested via a ttest, whether 1) the relevant CAR value differs significantly from zero and 2) whether the CAR of group 1 differs significantly from group 2. Respectively ***, ** and * indicate significance at a level of <1%, <5% and <10%.
Panel A: Significance test for CAR – Introduction consultation CRD IV
Group Mean Std. Err. Std. Dev. [95% Conf. Interval]
Ratio<100% .0054877*** .0001112 .0364483 .0052698 .0057056 Ratio>=100% .0086082*** .00018 .034934 .0082553 .0089611
Combined .0062971 .0000947 .0360875 .0061114 .0064828
Difference -.0031205*** .000216 -.0035438 -.0026972
Panel B: significance test for CAR – Introduction consultation CRD IV
Group Mean Std. Err. Std. Dev. [95% Conf. Interval]
Ratio<200% .0077062*** .0001024 .0362778 .0075054 .0079069 Ratio>=200% -.0026539*** .0002384 .0334982 -.0031213 -.0021866 Combined .0062971 .0000947 .0360875 .0061114 .0064828
Difference .0103601*** .000275 .0098211 .0108991
Panel C: Significance test for CAR – Referendum Brexit
Group Mean Std. Err. Std. Dev. [95% Conf. Interval]
Ratio<100% .0300882*** .0002145 .0722926 .0296678 .0305087 Ratio>=100% .0500395*** .0002496 .0506288 .0495503 .0505287
Combined .0353934 .0001724 .0677923 .0350556 .0357312
Panel D: Significance test for CAR – Referendum Brexit
Group Mean Std. Err. Std. Dev. [95% Conf. Interval]
Ratio<200% .0344274*** .0001888 .068954 .0340574 .0347974 Ratio>=200% .0414486*** .0004088 .0596467 .0406473 .04225
Combined .0353934 .0001724 .0677923 .0350556 .0357312
Table 5: Effects of public consultation CRD IV and Brexit referendum on CAR.
This table shows the output of the OLS regressions to measure which factors mattered for the impact of the event on cumulative abnormal returns. “X>=ratio” relates to the ratio’s mentioned in the top row, for instance,
X>=ratio in the first column indicates a dummy variable for banks having a var-to-fix ratio of at least 100% or below 100%, being 1 or 0 for the dummy variable respectively. Interaction is the interaction term between var-to-fix and the dummy variable “X>=ratio”. Panel A and B are regression outputs on the CAR around the public consultation of CRD IV for different event windows: [-1;1] and [0;5] respectively. Panel C and D are regression outputs on the CAR around the referendum of Brexit on the same varying event windows. Lastly, the first 2 columns of each panel are regressions on the basic controls only and the last 2 columns are regressions run on full controls. Respectively ***, ** and * indicate significance at a level of <1%, <5% and <10% and robust standard errors in parentheses.
Panel A: Introduction Consultation CRD IV [1]
CAR Ratio = 100% Ratio = 200% Ratio = 100% Ratio = 200%
(1) (2) (3) (4) Var-to-fix 0.02*** 0.01** 0.00 0.00 (0.007) (0.004) (0.007) (0.003) X >= ratio 0.00 -0.07*** 0.00 -0.09*** (0.012) (0.020) (0.013) (0.030) Interaction -0.02** 0.02*** -0.00 0.03*** (0.009) (0.007) (0.009) (0.001) Log(assets)t-1 -0.01*** -0.01*** -0.01*** -0.02*** (0.004) (0.004) (0.003) (0.003) Log(nii)t-1 0.02*** 0.02*** 0.02*** 0.02*** (0.003) (0.003) (0.004) (0.003) Log(nplratio)t-1 0.00 0.00 0.00 0.00 (0.001) (0.001) (0.002) (0.001) (Tier1/assets)t-1 -0.17 -0.09 -0.30* -0.129 (0.147) (0.143) (0.166) (0.172) Roat-1 0.01 0.01 (0.004) (0.004) Diverst-1 0.02 0.02 (0.015) (0.014) Dep/Assetst-1 0.00 0.00 (0.016) (0.015) Liquidityt-1 -0.02* -0.01 (0.012) (0.013) Intercept 0.04 0.04 0.04 0.02 (0.035) (0.032) (0.045) (0.044) R-squared 0.1818 0.1971 0.2745 0.3242
Panel B: Introduction Consultation CRD IV [2] CAR Ratio = 100% Ratio = 200% Ratio = 100% Ratio = 200% (1) (2) (3) (4) Var-to-fix -0.03*** 0.00 -0.02 0.01*** (0.009) (0.003) (0.009) (0.004) X >= ratio -0.00 -0.12*** -0.01 -0.13*** (0.009) (0.032) (0.009) (0.032) Interaction 0.03*** 0.04*** 0.02** 0.04*** (0.010) (0.011) (0.010) (0.011) Log(assets)t-1 -0.01*** -0.01*** -0.00 -0.00 (0.003) (0.003) (0.004) (0.004) Log(nii)t-1 0.00 0.01** -0.01** -0.00 (0.003) (0.003) (0.003) (0.003) Log(nplratio)t-1 0.01*** 0.01*** 0.01*** 0.01*** (0.001) (0.001) (0.001) (0.001) (Tier1/assets)t-1 -0.42* -0.38 -1.16*** -0.95*** (0.212) (0.239) (0.251) (0.319) Roat-1 0.04*** 0.04*** (0.005) (0.005) Diverst-1 -0.06*** -0.06*** (0.017) (0.016) Dep/Assetst-1 -0.04*** -0.05*** (0.016) (0.015) Liquidityt-1 -0.11*** -0.11*** (0.015) (0.017) Intercept 0.16*** 0.14** 0.27*** 0.26*** (0.054) (0.053) (0.065) (0.060) R-squared 0.0864 0.1907 0.3289 0.3526
Panel C: Referendum Brexit [1] CAR Ratio = 100% Ratio = 200% Ratio = 100% Ratio = 200% (1) (2) (3) (4) Var-to-fix -0.00 0.00 0.00 0.01 (0.013) (0.004) (0.020) (0.004) X >= ratio -0.01 0.00 0.01 0.03 (0.012) (0.043) (0.016) (0.050) Interaction 0.01 0.00 -0.00 -0.01 (0.014) (0.015) (0.022) (0.019) GB -0.02** -0.02** -0.01 -0.02 (0.011) (0.011) (0.013) (0.012) Log(assets)t-1 0.01** 0.01** 0.01 0.01* (0.005) (0.005) (0.007) (0.006) Log(nii)t-1 -0.00 -0.00 0.00 0.00 (0.004) (0.004) (0.005) (0.005) Log(nplratio)t-1 0.01*** 0.01*** 0.00** 0.00* (0.002) (0.002) (0.002) (0.002) (Tier1/assets)t-1 -0.09 -0.07 0.05 0.07 (0.296) (0.294) (0.319) (0.320) Roat-1 -0.01 -0.01 (0.008) (0.008) Diverst-1 0.06* 0.06* (0.035) (0.034) Dep/Assetst-1 0.05** 0.05** (0.022) (0.021) Liquidityt-1 -0.01 -0.01 (0.030) (0.028) Intercept -0.09 -0.10 -0.25*** -0.26*** (0.075) (0.071) (0.086) (0.085) R-squared 0.1805 0.1793 0.1984 0.1990
Panel D: Referendum Brexit [2] CAR Ratio = 100% Ratio = 200% Ratio = 100% Ratio = 200% (1) (2) (3) (4) Var-to-fix -0.01 -0.00 -0.04*** -0.01 (0.012) (0.004) (0.015) (0.004) X >= ratio 0.05*** 0.09** 0.02 0.08** (0.012) (0.034) (0.016) (0.039) Interaction -0.02 -0.04*** 0.02 -0.03** (0.013) (0.012) (0.018) (0.016) GB 0.01 0.01 -0.01 -0.01 (0.009) (0.011) (0.010) (0.012) Log(assets)t-1 0.01 0.01* 0.014*** 0.02*** (0.004) (0.004) (0.005) (0.005) Log(nii)t-1 -0.00 -0.00 -0.01* -0.01** (0.003) (0.003) (0.004) (0.004) Log(nplratio)t-1 0.00** 0.00** 0.00 0.00 (0.001) (0.001) (0.002) (0.002) (Tier1/assets)t-1 -0.26 -0.22 0.05 0.15 (0.317) (0.314) (0.262) (0.261) Roat-1 0.02** 0.02** (0.007) (0.007) Diverst-1 -0.13*** -0.13*** (0.030) (0.030) Dep/Assetst-1 0.05** 0.04** (0.022) (0.021) Liquidityt-1 0.07*** 0.07*** (0.024) (0.021) Intercept -0.04 -0.07 -0.15* -0.19** (0.060) (0.058) (0.078) (0.076) R-squared 0.0687 0.0686 0.2378 0.2258
Table 6: Effects of public consultation of CRD IV and Brexit referendum on LRMES This table shows the output of the OLS regressions to measure which factors mattered for the impact of the event on cumulative abnormal levels of LRMES. “X>=ratio” relates to the ratio’s mentioned in the top row, for instance, X>=ratio in the first column indicates a dummy variable for banks having a var-to-fix ratio of at least 100% or below 100%, being 1 or 0 for the dummy variable respectively. Interaction is the interaction term between var-to-fix and the dummy variable “X>=ratio”. Panel A and B are regression outputs on the LRMES around the public consultation of CRD IV for different event windows: [-1;1] and [0;5] respectively. Panel C and D are regression outputs on the LRMES around the referendum of Brexit on the same varying event windows. Lastly, the first 2 columns of each panel are regressions on the basic controls only and the last 2 columns are regressions run on full controls. Respectively ***, ** and * indicate significance at a level of <1%, <5% and <10% and robust standard errors in parentheses.
Panel A: Regression output – Introduction Consultation CRD IV [1]
C.A. LRMES Ratio = 100% Ratio = 200% Ratio = 100% Ratio = 200%
(1) (2) (3) (4) Var-to-fix -0.01 0.07*** -0.10** 0.03 (0.041) (0.022) (0.048) (0.024) X >= ratio 0.20*** -0.36*** 0.18*** -0.55*** (0.055) (0.130) (0.050) (0.076) Interaction -0.07 0.04 -0.01 0.11*** (0.046) (0.047) (0.051) (0.034) Log(assets)t-1 0.03* 0.03* 0.10*** 0.10*** (0.019) (0.018) (0.021) (0.018) Log(nii)t-1 -0.04** -0.03* -0.09*** -0.08*** (0.017) (0.017) (0.018) (0.017) Log(nplratio)t-1 0.03 0.03*** 0.02** 0.02*** (0.007) (0.007) (0.008) (0.007) (Tier1/assets)t-1 1.04 1.33 0.67 1.93** (0.793) (0.815) (0.959) (0.894) Roat-1 0.17*** 0.16*** (0.030) (0.029) Diverst-1 0.11 0.06 (0.074) (0.072) Dep/Assetst-1 -0.02 -0.05 (0.101) (0.084) Liquidityt-1 -0.17** -0.14* (0.074) (0.073) Intercept -0.10 -0.22 -0.57* -0.72*** (0.223) (0.205) (0.300) (0.233) R-squared 0.1626 0.2134 0.3785 0.4403
Panel B: Regression output – Introduction Consultation CRD IV [2] C.A. LRMES Ratio = 100% Ratio = 200% Ratio = 100% Ratio = 200% (1) (2) (3) (4) Var-to-fix -0.04 0.13*** -0.25*** 0.06** (0.054) (0.027) (0.056) (0.029) X >= ratio 0.42*** -0.25* 0.39*** -0.50*** (0.062) (0.141) (0.056) (0.141) Interaction -0.12** -0.06 0.03 0.03 (0.060) (0.055) (0.065) (0.056) Log(assets)t-1 0.04* 0.05** 0.17*** 0.18*** (0.024) (0.023) (0.028) (0.024) Log(nii)t-1 -0.05** -0.05** -0.14*** -0.15*** (0.021) (0.021) (0.022) (0.021) Log(nplratio)t-1 0.07*** 0.07*** 0.04*** 0.04*** (0.009) (0.008) (0.010) (0.009) (Tier1/assets)t-1 1.35 1.57 0.48 1.85 (1.049) (1.122) (1.151) (1.267) Roat-1 0.30*** 0.28*** (0.034) (0.033) Diverst-1 0.28*** 0.17* (0.083) (0.086) Dep/Assetst-1 0.18 0.09 (0.124) (0.109) Liquidityt-1 -0.43*** -0.40*** (0.088) (0.090) Intercept 0.11 -0.11 -1.14*** -1.32*** (0.274) (0.044) (0.360) (0.313) R-squared 0.2067 0.2169 0.4364 0.4295
Panel C: Regression output – Brexit referendum [1] C.A. LRMES Ratio = 100% Ratio = 200% Ratio = 100% Ratio = 200% (1) (2) (3) (4) Var-to-fix -0.03 0.08** 0.03 0.11*** (0.055) (0.035) (0.073) (0.033) X >= ratio 0.18** -0.05 0.35*** 0.11 (0.072) (0.162) (0.084) (0.198) Interaction -0.00 -0.04 -0.15* -0.15* (0.065) (0.062) (0.087) (0.077) GB 0.04 0.08* 0.10** 0.14*** (0.043) (0.046) (0.047) (0.050) Log(assets)t-1 -0.01 -0.00 -0.01 0.00 (0.024) (0.024) (0.027) (0.027) Log(nii)t-1 0.02 0.02 0.06** 0.06** (0.021) (0.021) (0.025) (0.025) Log(nplratio)t-1 0.02* 0.02* -0.01 -0.01 (0.008) (0.002) (0.011) (0.011) (Tier1/assets)t-1 1.54 1.78 2.03 2.12* (1.130) (1.091) (1.300) (1.273) Roat-1 -0.04 -0.04* (0.022) (0.022) Diverst-1 0.53*** 0.50*** (0.148) (0.145) Dep/Assetst-1 0.42*** 0.39*** (0.122) (0.123) Liquidityt-1 -0.43*** -0.41*** (0.147) (0.138) Intercept 0.12 -0.03 -1.00** -1.05** (0.336) (0.312) (0.456) (0.454) R-squared 0.0917 0.0809 0.1742 0.1670
