Supervisory stress tests conducted in recent years by the European Banking Authority (EBA) provided potentially useful information to banks’ management. This paper examines empirically the hypothesis that the stress test results published have influence on banks’ ongoing risk mitigation efforts. The hypothesis rests on the evidence that stress tests have induced a reaction in bank equity and debt markets, which in turn sends a clear market signal to banks’ management. The concept of market discipline implies that market prices can influence the banks’ management in terms of their risk mitigation decisions. Using a difference-in-differences approach, it is measured how banks reacted to their respective stress test results. The evaluation shows a general rise in risk mitigation efforts of EU banks when looking at average total capital ratios since 2007. However, the results do reject the hypothesis that banks with low stress test scores have a stronger incentive to mitigate risk after the stress test publication. Yet low score banks demonstrate that adequate capital buffers were already set in place long before the stress test results. The higher capital buffers are reflected in above average total capital ratios of low score banks. This fact indicates that these banks might have reacted already before the stress test publication to make up for the higher asset risk exposure by growing stronger capital buffers.
Table of Contents
Figures and Tables
List of Acronyms
1 Introduction
1.1 Motivation
1.2 Hypothesis
1.3 Content Structure
2 Stress Tests and Related Literature
2.1 Purpose of Stress Tests
2.2 EBA Supervisory Role
2.3 EBA Stress Testing Methodology
2.4 Relevant Literature
3 Theoretical Framework
3.1 Market Disciplinary Effects
3.2 The Role of Subordinated Bank Debt
3.3 Incentives of Market Participants
3.3.1 Market Perception
3.3.2 Investors
3.3.3 Regulatory Authorities
3.3.4 Institutional Investors
3.3.5 Government
3.3.6 Conclusion
3.4 Reaction of Banks
3.4.1 Capital Adequacy
3.4.2 Recapitalisation
3.4.3 Asset Reallocation
3.4.4 Conclusion
4 Empirical Analysis
4.1 Implementation of the Hypothesis
4.2 Methodology
4.3 Difference-in-differences Approach
4.3.1 Parallel Trend Assumption
4.3.2 Empirical Specification
4.4 Reaction of Banks to Supervisory Stress Tests
4.4.1 Summary Statistics - Good Performers
4.4.2 Summary Statistics - Bad Performers
4.4.3 Findings
4.4.4 Deviations from the Hypothesis
5 Final Remarks
5.1 Conclusion
5.2 Critical Appraisal
5.3 Research Outlook
Data Directory
Reference List
Figures and Tables
Fig. 1 Evolution of mean total capital ratios between 2007 and 2013
Fig. 2 Evolution of mean total capital ratio between 2007 and 2013 for good performing banks
Fig. 3 Evolution of mean total capital ratio between 2007 and 2013 for bad performing banks
Table 1 Summary of 2011-2014 capital structure changes for good performing banks
Table 2 Summary of 2014-2016 capital structure changes for good performing banks
Table 3 Summary of 2011-2014 capital structure changes for bad performing banks
Table 4 Summary of 2014-2016 capital structure changes for bad performing banks
List of Acronyms
Abbildung in dieser Leseprobe nicht enthalten
Executive Summary
The following graduation thesis The Reaction of European Banks to Supervisory Stress Tests provides an evaluation of the supervisory bank stress tests introduced by the European Banking Authority (EBA). Supervisory stress tests were implemented in the aftermath of the 2007 financial crisis to prevent excessive risk-taking activities of financial institutions in the EU. The EBA, together with the participating banks, publishes risk relevant information in regular intervals. This information contributes to the stability of the financial system by enabling investors and depositors to monitor the activities of banks. To date, the EBA has waived defining minimum requirements to pass the test. Instead, the focus is on educating the market about the risks of individual banks (Committee on Banking Supervision (2018)). As a result, market participants would subsequently induce market mechanisms that exert indirect pressure on banks that show excessive risk-taking. Such negative market reactions are reflected, for example, in falling stock prices and rising bank funding costs. Studies such as Georgescu et al. (2017) and Petrella and Resti (2013) provide evidence on negative abnormal returns of credit defaults swaps and negative stock market returns for banks that present low stress test scores. This thesis is linked to this evidence and further analyses the impact of such negative market attention on banks. The empirical investigation shows risk-mitigating reactions of banks as a response to their performance in stress tests. The hypothesis claims that banks with low stress test scores have a stronger incentive to reduce asset risk and to increase capital buffers. The validation of this hypothesis would underpin the proclaimed benefits of publishing stress test results.
To understand the stress test results in the first place, it is necessary to capture the precise meaning of risk in banking. In principle, bank risk can be broken down into two sub-aspects, namely the riskiness of the asset allocation one side and capital buffers on the other side. The common concept of capital adequacy juxtaposes both risk aspects and suggests that a higher risk exposure can be assured by increasing capital buffers adequately. The risk-weighted assets (RWA) approach is a widely accepted measure for asset risk exposure. By weighing the assets on the balance sheet according to their risk, this approach adjusts for those assets that are partly or entirely hedged against default. When the available capital buffers are divided through the RWA, one obtains common capital adequacy indicators like Tier 1 or CET1 ratios. The EBA stress tests take the actual CET1 ratios of banks and apply projections based on an adverse macroeconomic scenario. A possible scenario would be e.g. an escalating debt crisis in the EU.
The incurred losses in this scenario are accordingly put into relation to the capital buffers of the bank. Hence, the main risk indicator in the ECB stress tests is the resulting capital depletion reflected in the projected changes of banks’ CET1 ratios.
The results provided by the EBA stress tests allow for a comparison between different banks, since the stress tests are standardized and rely on the same assumptions and scenarios. This important benefit sets the EBA stress tests apart from traditional stress tests, where banks model each their own scenarios. For investors and depositors, the comparability of the results means that they can better discriminate between banks. The enhanced information transparency should in theory increase market discipline because managers are under pressure to maintain relations with investors and depositors. Equity holders can put a bank’s managers under pressure by selling their stake and, thus, reducing equity prices. Since equity price fluctuations tend to correlate with bank funding costs, debt holders also enforce market discipline. By directly influencing banks’ funding costs debt holders can exert more imminent pressure on banks’ liquidity and cash flows. Apart from investors and depositors, the government itself is striving for financial stability. Together with the regulatory authorities, it lays the foundation for efficient capital markets. The EU continues its efforts to reduce the burden of failed banks on tax payers. Governments need to be committed to adopt a no-bailout policy. The commitment relies on a standardized insolvency law that foresees bankruptcy in time to ensure the continuity of the banks’ critical functions. Overall, there are signs that banks are increasingly aware of their risk exposure and their market perception. Regulation by the EBA, including publicly accessible stress tests, seems to improve market discipline. Whether the market pressure on banks is strong enough to trigger risk-mitigating reactions, will be tested as follows.
This paper is interested in banks’ risk mitigation efforts after the publication of the stress test results. Risk mitigation strategies are reflected in capital structure changes of banks. Therefore, the interest lies in changes of the capital structure during the three-year period after the stress test publication. Recapitalization through retained earnings, equity issuances or contingent convertible bonds leads to higher capital buffers, which, in turn, strengthens the risk-bearing ability of banks. Furthermore, banks can change their risk exposure by reallocating their assets to less risky assets with e.g. higher collaterals or less volatility. To measure changes in risk mitigation of banks, the total capital ratio seems appropriate because it comprises changes in the capital structure as well as changes in asset risk exposure. Based on the public data provided by the
EBA, it is tested whether banks react differently after having obtained positive or negative market attention in terms of their stress tests results. For the 2011 and 2014 stress tests, banks are classified into well- and badly-performing banks based on the stress test results. To measure if badly- or well-performing banks show a special reaction, their changes in the total capital ratio are compared with banks that performed about average in the stress test. A difference-in-differences approach is used because it depicts deviations from general capital structure trends. The experimental setting should lead to results that explain whether or not banks increased their risk mitigation on average and how badly-performing banks reacted in contrast to well-performing banks.
The results reject the hypothesis that banks with low stress test scores have a stronger incentive to mitigate risk after the stress test publication. When comparing badly- and well-performing to average banks, no significant deviations in banks’ risk mitigation in the period after the stress tests could be found. The results in Fig.1, however, show a general rise in risk mitigation efforts of EU banks, which is reflected in increasing total capital ratios from 2007 onwards.
Abbildung in dieser Leseprobe nicht enthalten
Fig. 1 Evolution of mean total capital ratios between 2007 and 2013
Increasing total capital ratios can be interpreted as an underlying trend where banks grow capital to reach future capital requirements imposed by regulatory authorities. The banks with average results in the 2011 stress test managed to increase their total capital ratios by 3.5% on average over the six-year period between 2007 and 2013. Fig. 1 also shows significant differences when it comes to comparing total capital ratios in the four years before the 2011 stress test. These pre-treatment differences in the banks’ capital buffers are an important aspect when interpreting the risk mitigation of banks. Banks with bad test scores seem to have had already reacted long before the publication of the 2011 stress tests by building up high capital buffers compared to the average group. Thus, badly-performing banks might have seen no necessity for further risk-mitigating measures after the 2011 stress test. The early reaction of banks could be explained by the fact that the market and banks had already been using other models for risk assessment that delivered similar results like the EBA stress test.
The hypothesis suggests that the stress test results would induce market discipline and, thus, enforce risk mitigation on the side of banks. Although the publication of the stress tests led to measurable market reactions, the induced market pressure might be sufficient to influence bank decisions in terms of their respective capital structure. The research on supervisory stress tests is still in its early stages and so are the methodologies and scenarios behind the EBA stress tests. The EBA stress test results should therefore be considered as complementary to traditional and well-proven risk assessment methods.
Abstract. Supervisory stress tests conducted in recent years by the European Banking Authority (EBA) provided potentially useful information to banks’ management. This paper examines empirically the hypothesis that the stress test results published have influence on banks’ ongoing risk mitigation efforts. The hypothesis rests on the evidence that stress tests have induced a reaction in bank equity and debt markets, which in turn sends a clear market signal to banks’ management. The concept of market discipline implies that market prices can influence the banks’ management in terms of their risk mitigation decisions. Using a difference-in-differences approach, it is measured how banks reacted to their respective stress test results. The evaluation shows a general rise in risk mitigation efforts of EU banks when looking at average total capital ratios since 2007. However, the results do reject the hypothesis that banks with low stress test scores have a stronger incentive to mitigate risk after the stress test publication. Yet low score banks demonstrate that adequate capital buffers were already set in place long before the stress test results. The higher capital buffers are reflected in above average total capital ratios of low score banks. This fact indicates that these banks might have reacted already before the stress test publication to make up for the higher asset risk exposure by growing stronger capital buffers.
Key words: supervisory stress test, European Banking Authority, EBA, capital structure, risk mitigation, risk exposure, RWA, capital depletion, capital ratio, Tier 1, capital requirements
1 Introduction
1.1 Motivation
The supervisory stress tests introduced by the European Banking Authority (EBA) in the aftermath of the 2007 financial crisis provide a standardized measure for banks’ risk exposure. Testing the robustness of banks to specific, extreme scenarios provides information that is potentially complementary to other methods of risk management such as VaR, and SRISK can, hence, be useful to management, shareholders, clients and supervisors. Furthermore, since - other than, for example, in Switzerland - results were made public, they also triggered certain market reactions which banks’ management had to respond to. The publication of the standardized results offers unprecedented opportunities to compare the financial soundness of banks in and between different regions. Looking back at a decade of published data and awaiting the results of the 2018 stress tests, the academic interest in the EU stress tests continues to rise.
Recent research has mainly emphasized the effect of the stress test results on banks’ stock prices. The modelled risk resulting from the applied extreme scenarios evidentially has an influence on market perception, and thus, on stock prices. The information content of supervisory stress tests depends on how accurately the stress impact can be modelled. Camara, Pessarossi, and Philippon (2017) undermine the credibility of common stress tests by testing the accuracy of the model forecasts. The comparison between model predictions and ex-post realized losses suggests that supervisory stress tests are informative. The proven informative value of supervisory stress tests for investors poses the question to banks’ management whether it is necessary to adapt by applying further risk mitigation measures so that a positive perception on the side of the investors can be maintained. The goal of this MA thesis is, thus, to investigate empirically how banks have responded to their performance in stress tests. Based on data availability, the focus will be on the stress tests conducted by the EBA in 2011, 2014 and 2016.
1.2 Hypothesis
The present hypothesis is based on the results of recent studies on supervisory stress tests. Empirical evidence states that past supervisory stress tests had a measurable price impact on the stock market. This observation leads to the question whether published stress test results provoke market disciplinary reactions on the side of the banks. It seems to be a natural reaction of banks’ management to increase risk mitigation measures, especially when stress test results have led to a devaluation of equity. Furthermore, such a reaction coincides with the EBA’s desired effect in terms of supervisory stress tests. Disclosing an adequate and comparable picture of financial institutions should enhance the efficiency of market disciplinary effects (Committee on Banking Supervision (2018)), particularly because the current EBA bank stress tests are not disciplinary in the sense of requiring minimal stress test scores. Market discipline, therefore, should play an important role in increasing the stability of banks. Originating from the model on market discipline in regulating bank risk by Avery, Belton, and Goldberg (1988) this paper differentiates between market participants and their contribution to market discipline. Here one may discover different views in terms of bank stress test results. By identifying the different interests of management, shareholders, creditors as well as regulation authorities and governments, this MA thesis tries to explain possible risk mitigation reactions.
Once the cause-effect chain behind banks’ market discipline has been specified, the paper analyses by way of statistical tests if and how banks adapted their capital structure in response to stress test results. The hypothesis that banks with low stress test scores might have a stronger incentive for risk mitigation arises from the recent findings of Georgescu et al. (2017). Based on the 2014 and 2016 EBA stress tests, they provide evidence that the publication of stress test results induced price discrimination as the price impact on equity tended to be stronger for banks with low stress test scores. Accordingly, one would expect that banks with low stress test scores will react stronger by increasing risk mitigation measures. The evaluation of the published EBA data should show whether riskier banks reacted stronger to stress test results, with the rationale of avoiding negative market attention in future. The evaluation should also show if a convergence in risk exposure between different banks can be observed.
1.3 Content Structure
The present MA thesis is structured into four main parts:
In the first part, a definition of the concept of bank stress tests is provided. The most common bank stress test approaches will be presented and compared. Subsequently, the implications of the supervisory character of the EBA stress tests will be specified. Based on this conceptual background, the findings of related studies in this field will then be aggregated.
In the second part, the existing theoretical framework to explain market disciplinary effects within the subject of supervisory stress tests will be used. By analyzing the incentives of market participants, a causal interpretation of banks’ reaction to stress test results will be provided. An interpretation of stress test results will be analysed from the investors’ perspective, including that of stockholders and bondholders. An explanation of the crucial role of regulating authorities when it comes to defining the goals behind stress tests follows. Finally, the position of the government is looked at since government authorities are often involved in dealing with stress situations in the financial system.
The third part of this thesis consists of an empirical evaluation of the hypothesis. Based on EBA data from 2011 to 2016, we will perform several trend analyses. In order to obtain conclusive findings, the rationale behind the choices made regarding data categorization and evaluation will be disclosed. The parallel trend assumption is verified by depicting capital structure trends before and after the stress test publication. Subsequently, the findings of the statistical evaluation will be depicted. These results underline the hypothesis made. Important deviations from that hypothesis will be answered by way of new evidence.
The fourth part completes the research on banks’ reactions to stress tests with a brief summary of the results, followed by an explanation of the implications of the research and stress tests in general. To conclude this paper, the focus will be on the future importance of stress tests and some interesting lines of research.
2 Stress Tests and Related Literature
2.1 Purpose of Stress Tests
The goal of stress tests performed by financial institutions is to predict the sensitivity of the business to hypothetical fluctuations in asset prices. A stress test serves as an instrument for the preventive detection of vulnerabilities of a financial institution. The results of a stress test give rise to appropriate measures for approaching deficiencies (Vlad (2012)).
The term ‘stress test’ is generic and includes a variety of methodologies with different scope. Stress tests performed at the internal level have proven to be useful and play an important role in portfolio and treasury management. Depending on the methodology, a stress test covers different risks. Macro stress tests predict the effects of market risks. In addition, credit, liquidity and operational risks can be included in stress tests (Blaschke et al. (2001)). A survey conducted by Fender, Gibson, and Mosser (2001) compares the stress tests used by large international financial institutions. The results show that, at the time, banks applied stress test methodologies that were significantly different in their assumptions. Because the tests were performed internally, the stress test scenarios appeared to be tailored to the respective bank’s business strategy, portfolio composition and risk appetite. Stress test scenarios are insofar more extensive than simple sensitivity stress tests as that they not only predict single but also simultaneous extreme moves in portfolio asset values. Such asset shocks extend from equity prices to exchange and interest rates. Depending on the granularity of the modelled price fluctuations, stress test scenarios can even specify shocks for different maturity levels (Fender, Gibson, and Mosser (2001)).
The assumed price fluctuations in stress tests depend on the modelled extreme scenario. Evidently, different banks use different extreme scenarios when performing stress tests. In a given stress test scenario like e.g. a mortgage crisis, the underlying assumptions about price fluctuations should be realistic or at least comprehensible. For a stress test scenario claiming to be extreme and realistic at the same time might, at first glance, seems contradictory. How can an extreme event that is unlikely to happen be a realistic assumption? The answer lies in the word ‘unlikely’, more specifically in the probability of an extreme scenario over time. While the extreme scenario is not likely to happen at the time of the stress test, the probability that a series of reactions lead to an extreme scenario increases over time. Apparently, extreme scenarios occur at irregular intervals and can extend from a national level, such as the Chinese stock market crash in 2015, to an international level, like the subprime mortgage crisis in 2007. As mentioned above, extreme scenarios may well become reality in the period between stress tests. Furthermore, for such extreme scenarios to be realistic, the modelled directions of asset price shocks should be chosen carefully. To obtain realistic parallel movements between different asset classes it is useful to justify the rationale behind such movements or, at least, to base those assumptions on past market observations. A survey regarding stress testing practices (Committee on the Global Financial System (2001)) shows that most stress test scenarios rely on historical as well as hypothetical events. The sheer quantity of different events to rely on makes the decision of defining a future extreme event difficult, if not arbitrary. Alexander and Sheedy (2008) accordingly criticise that many extreme yet plausible scenarios are not even considered. By defining a whole spectrum of scenarios and assigning scenario probabilities they outline a promising stress test methodology. Breuer and Jandacka (2012) present a methodology whereby bank stress tests are implemented in a quantitative risk-management framework. To obtain useful stress scenarios they recommend to identify key risk factors in a first step and plan risk-reducing actions in a second step.
Having presented some basic methodologies for stress testing, it is time to focus on the main topic of this MA thesis and elaborate on a more comprehensive view by studying particular features of the EBA stress tests. While showing some methodological similarities to the internal stress tests, the EBA supervisory stress tests show significant differences in the underlying assumptions and objectives.
2.2 EBA Supervisory Role
The EBA is a regulatory agency of the EU that emerged from the former Committee of European Banking Supervisors in 2011. The main objective of the EBA consists of establishing supervisory standards and practices for financial institutions in Europe. The EBA monitors European banks and publishes relevant supervisory information, thus contributing to the stability of the financial system in the EU. Through collaboration with national banking authorities, the EBA aims to strengthen the consistency of supervisory outcomes. (European Banking Authority (2017)).
In the aftermath of the 2007 financial crisis, the bank bailouts performed by government authorities were harshly criticized by the public. While the governments argued that they were forced to save big financial institutions because of the “too big to fail” difficulty, critics insist that excessive risk-taking activities of financial institutions should by no means be borne by the tax payer. State aid cases in the period from 2007 to 2008 amount to 35 interventions in the EU alone. In addition to bank bailouts, such interventions consist of equity, credit and refinancing guarantees, all of which implicitly shift risk from the financial sector to the tax payer (European Commission (2018)). One of the main tools used by the EBA to enforce self-responsibility of banks is the EU-wide stress test exercise (European Banking Authority (2017)).
The EBA supervisory stress tests were implemented in 2009 and are carried out approximately every two years. The tests performed evaluate banks’ resilience to severe financial shocks. The focus is on Core Tier 1 banks that, per definition, add to the systemic risk in financial markets. Such banks can cause contagion effects during a financial crisis due to their relevant size and infrastructure (Vlad (2012)). This definition coincides with the definition of “too big to fail” institutions. Initially, the participating banks, as well as the EBA, needed time to adapt to the new undertake of supervisory stress tests. Therefore, the five published stress tests starting from 2009 have been gradually customized over time. Furthermore, the number of participating banks has risen from 22 to 51 within 10 years, representing about 70% of EU banks’ total assets (European Banking Authority (2017)). The transparency regarding macroeconomic scenarios improved and the methodologies used evolved over time.
2.3 EBA Stress Testing Methodology
Similar to the internal stress tests as depicted in chapter 2.1, the EBA implies an adverse macroeconomic scenario that impacts assets on the banks’ balance sheets. The shocks modeled reveal the expected total decrease in asset value given the adverse scenario. Unlike most stress tests performed internally by banks, the EBA stress tests go further in the interpretation of the results. In particular, they place incurred losses in relation to the risk capital of the bank. This important step arises from the Basel 2 Accord stipulating that banks provide sufficient capital in order to withstand the results of economic shocks (Basel Committee on Banking Supervision (2006)). By complementing stress test results with capital adequacy measures the EBA not only allows for comparison between different banks, but also prevents unsustainable risk taking that would lead to undesirable state aid measures. The standardized and comparable ratios resulting from this methodology also carry another important benefit in that the difficult and often criticized assumptions regarding the severity and amplitude of shocks become less binding since the comparability of the results at least expresses the relative financial soundness of banks.
Based on such advantages the EBA opted for “the overall impact on the CET1 capital ratio” as the key measure to summarize the results of their stress tests. As will be shown later, this measure is important for the evaluation of this paper’s hypothesis and, hence, needs further explanation.
To understand the dynamics behind the CET1 capital ratio one looks at both the nominator and the denominator. The nominator represents the common equity Tier 1 Capital, which includes the banks’ common shares, retained earnings, stock surpluses from share issuances (The European Commission (2014)). The term ‘Tier 1’ stands for the capital that is depleted at first in the event of a crisis (Basel Committee on Banking Supervision (2006)).The denominator represents banks’ risk-weighted assets (RWA). By weighing the assets in the balance sheet according to their risk, banks adjust for those assets that are partly or entirely hedged against default. Sovereign debt and collateralized debt, for example, are unlikely to vanish entirely since they are secured by the central bank or by a collateral, respectively. The participating banks assess their risk weights with their own credit risk models. In 2017, the EBA published a report asking for a more consistent approach for risk weighting high default portfolios. In addition, the EBA reserves the right to change risk weights resulting from inadequate internal models, as it was the case with Finland’s OP Financial Group (Turk-Ariss (2017)). The combination of nominator and denominator in the CET1 capital ratio is suitable for risk assessment purposes because it emphasizes those asset values that would impact equity most directly in case of a shock.
In order to measure the impact on the CET1 capital ratio, the EBA starts by defining the adverse scenario. Such adverse scenario should reflect those systemic risks identified by the authorities. In 2016, those risks were stated as follows: an abrupt reversal of the compressed global risk premia could be amplified by low secondary market liquidity and, thus, impact banks’ refinancing operations. In addition to that, low nominal growth prospects would decrease the profitability of banks. Finally, the risk of competitors and spillovers resulting from the growing shadow banking sector should also be implemented in the adverse scenario. In a second step, the EBA captures how these risks would materialize in the international capital and goods markets. Only by considering common market dynamics can one predict different price effect on individual countries and banks. Price effects on banks’ assets are always measured as a deviation from the baseline scenario. In the case of endogenous interest rate hikes due to low market liquidity, not all sovereign spreads of EU countries would be affected equally. The EBA, for example, states that nominal exchange rates of eastern European countries against the euro would depreciate sharply. This in turn leads to an even stronger increase in sovereign spreads for those countries. Similar dynamics are also considered in the equity and commodity markets. The scenario modeled even implements monetary policy decisions that absorb part of the impact (European Systemic Risk Board (2016)).
The previous section described how the model, as applied by the EBA, explains price effects of an adverse scenario on the financial and non-financial assets in the banking book. The individual price effects finally account for the cumulative impairment losses resulting from the adverse scenario. Such cumulative impairment losses have a direct impact on the banks’ operating profit. The incurred profit or loss leads to changes in common equity, which are of interest in evaluating the stress tests. So far, it has only been considered how the adverse scenario affects the price of banks’ financial assets. In order to fill the gap between economic shocks and the impact on capital, the EBA stress tests require trading losses that result from adverse economic shocks to be projected and taken into account. Since the adverse scenario implies lower trading volumes and more competition in the banking sector, trading income is likely to decrease or even turn negative as a result. For some banks incurred trading losses almost add up to the impairment losses of such banks’ assets. As an example, banks like Barclays and Deutsche Bank would incur severe trading losses due to the adverse scenario because they broadly engage in investment banking activities (European Banking Authority (2011)). With the forecast trading and impairment losses one can finally derive the total impact on CET1 capital and therefore compare the stress test results of different banks.
Even though the presented approach considers dynamic asset price fluctuations to determine impairment losses, it remains somewhat static in terms of banks’ business strategies. For comparability purposes the EBA stress tests foresee a stable business mix, assuming no changes in geographical range, product strategies and operations for the time horizon of the stress scenario. Therefore, assets and liabilities that mature within the time horizon are replaced with similar financial instruments in the model (European Banking Authority (2016)). The EBA neither defined mandatory hurdle rates nor capital thresholds for the stress tests between 2011 and 2016. This suggests that market participants should form their own opinion on whether or not the published stress test results are acceptable. While collecting the relevant data for the stress test, other risk factors were also recorded and published to provide a holistic picture to investors. For example, asset quality indicators, such as non-performing or defaulted loans, are also listed in the stress test dataset. Last but not least, the EBA identified sovereign debt exposures as a systemic risk worth considering in the stress test. In view of the imminent European sovereign debt crisis, the EBA published the concentration of banks’ exposures towards the respective public sector. A high concentration of sovereign investments, also known as banks’ home-bias, can reinforce unwanted financial market spillovers from a sovereign crisis (Enria, Farkas, and Overby (2016)). While asset quality indicators and sovereign risk exposures do capture the risk of a bank, this MA thesis will mainly focus on the stress test results arising from the adverse scenario. The impact of the modeled scenario on the CETl-ratio of banks is the main stress test result and therefore a relevant topic in current banking regulation research.
2.4 Relevant Literature
The widespread introduction of supervisory stress tests in 2009 has created a broad field of research. Many papers examine the EBA and the US Dodd-Frank Act stress tests because of their publicly available data, covering the world’s most important banks. In a first step, regulatory institutions themselves have provided valuable research input in refining the goals, assumptions and methodologies of current stress tests. Most of the developing literature supports supervisory stress tests by evaluating stress test effects and suggesting improvements.
Recent research examining the market reactions to the EBA stress tests resulted in mixed evidence. Petrella and Resti (2013) for example assessed if and how stress test publications affect bank stock prices. Based on the EBA stress test from 2011 they discovered that the market attached considerable importance to variables like liquidity risk and the modeled risks, whereas exposure to sovereign debt holdings showed little effect. Other authors like Bird et al. (2015) and Morgan, Peristiani and Savino (2014) find similar results. Research on more recent stress tests shows by means of event studies that price effects have declined significantly. Candelon and Sy (2015) as well as Neretina, Sahin and De Haan (2015) only detected weak effects on stock prices for the most recent stress tests. Glasserman and Tangirala (2016) attempt to explain this observation by stating that stress test results have become more predictable. They suggest that greater diversity in the modeled scenarios would result in an increased information content. Better predictability of stress test results must not necessarily be interpreted negatively. Predictable stress test results over the last 5 years could be attributed to the moderate but constant banking environment with positive equity market trends in contrast to the burden caused by low-interest rates. Another likely explanation is that banks have had sufficient time to adapt their business practices, aiming for a positive or at least average stress test result compared to their competitors. This explanation should be supported or rejected by the results of this thesis. Accordingly, the research carried out by Ellahie (2013) indicates that transparency and credibility are the most important factors for a successful stress test. The EBA’s effort to standardize stress tests by requiring transparency also contributes to improved market information and confidence. By defining new stress test methodologies one year before the test deadline the EBA provides the banks with sufficient time to deliver proper and transparent results.
On a more theoretical basis, Camara, Pessarossi, and Philippon (2017) assess the credibility of common stress tests by testing the accuracy of the model forecasts. The comparison between model projections and ex-post realized losses suggests that the models behind supervisory stress tests are based on valid assumptions. Steiner and Marra (2017) examines significant factors that influence the CET1 ratio spread resulting from the adverse scenario. These factors present a significant, negative correlation between the size of a bank and the corresponding CET1 ratio spread. However, factors such as leverage ratios, loans on assets, net interest margins and profitability are not sufficient to explain differences in the severity of the stress test outcomes between different banks.
For banks the information about the determinants behind the stress test results is crucial because it allows them to focus on specific risk mitigation measures. Even though the determinants might not be categorized by the factors used in Steiner and Marra (2017), banks are clearly able to identify those portfolio characteristics that led to the most severe stress test impact. The investigative work of tracing risk origination lies within the responsibility of the individual banks. Only with this intermediate step can banks find measures that truly reduce risk, hence leading to better stress test results in the future. The following chapter therefore depicts the strengths and limitations of banks’ risk mitigation measures. Having an overview of feasible reaction measures in turn allows to define the figures of interest for the subsequent empirical analysis.
3 Theoretical Framework
3.1 Market Disciplinary Effects
The results of the 2011-2016 EBA stress tests not only show a significant impact on equity markets but on bank funding costs in general. The 2014 stress test shows that those banks with a large CET1 impact report significantly negative higher CDS abnormal returns than banks with a low impact (Georgescu et al. (2017)). Usually this impact translates into higher bank funding costs. The new information revealed in the stress test results has provoked a change in investors’ perceptions. The observed price discrimination between banks can be interpreted as a market disciplinary effect. In this case, the market, consisting of equity and debt holders, exerts price pressure on the banks that performed poorly in the stress test.
The concept of market discipline relies on the assumption that market prices can affect the status of managers and their decisions. Market discipline, hence, represents a mechanism that complements governance efforts to align the interests of shareholders and management. Changes in equity prices of companies are signals resulting from firm evaluations of the individual market participants (Nier and Baumann (2006)). Equity and debt prices are relevant indicators not only to outside investors, but to the managers themselves. Managers’ compensation plans are often linked to the company’s equity price, thus exerting an explicit disciplinary function. In addition, a decrease in a company’s market value can be partly due to bad management decisions, which in turn can lead to the replacement of managers. The indirect pressure on managers exerted by equity prices and the direct pressure on banks’ liquidity exerted by varying debt funding costs, both contribute to the market discipline of a bank.
3.2 The Role of Subordinated Bank Debt
As far back as 1988 researchers contributed to the important role of market discipline in regulating bank risk (Avery, Belton, and Goldberg (1988)). Subordinated bank debt seemed promising for regulatory purposes, based on the rationale that investors would monitor the banks and charge appropriate interest rates. In the case of a bank failure, subordinated debt investors absorb the first loss after the equity is depleted (Birchler and Hancock (2003)). In general, regulators welcome the presence of subordinated investors because they are willing to take a higher default risk and vice versa profit from a higher yield. Accordingly, subordinated debt acts as an additional capital buffer, protecting depositors and senior lenders. The EBA acknowledges this advantage by defining special types of subordinated debt as Tier 1 capital. Therefore part of the subordinated debt is included in the Tier 1 ratio of the banks tested (The European Commission (2014)).
Avery, Belton, and Goldberg (1988) examined the potential of subordinated bank debt to enhance market discipline. They analysed whether common fundamental measures of bank risk were reflected in subordinated debt interest rate premiums. Surprisingly, their findings only discovered a weak correlation between subordinated debt risk premiums and Moody’s ratings and no significant correlation regarding balance-sheet risk indicators. However, other studies using more recent data did find correlations, indicating that subordinated debt holders might be relevant in monitoring bank behavior. Ashcraft (2006) used a different approach by comparing banks with similar regulatory capital ratios, but different amounts of subordinated debt. This experimental design allows for a separation between disciplinary pressure exercised by regulators and pressure from subordinated debt investors. The results of that study suggest that the amount of subordinated debt has a positive effect in helping banks to recover from financial distress. These findings also support the strong efforts of regulatory authorities in incentivising subordinated debt. By using subordinated debt, European banks have found a suitable instrument to support their CET1 ratio. The year 2013 has seen a strong increase in subordinated debt issues by European banks. Cumulative debt issues of $90bn show that European banks are willing to strengthen their capital buffers. At the same time, this demonstrates a strong demand by investors seeking higher yield for increased risk-taking (Thompson (2013)).
3.3 Incentives of Market Participants
3.3.1 Market Perception
Having shown that equity and debt prices can affect managers’ decisions, the following step seeks to identify the cause of price effects that happen in the first place. An important premise of market discipline is that stakeholders, including investors, are able to evaluate a bank’s “true” condition (Bliss and Flannery (2002)). The true condition of a bank is a subjective opinion resulting from the assessment of publicly available information about the bank. Such information reaches from press and financial statements to analyst forecasts and is now complemented with the stress test results published by the EBA. Only through periodic information assessment are investors able to monitor banks and change price perceptions accordingly. It is essential that the information retrieved is relevant, up-to-date and accurate. Furthermore, information comparability between periods and banks is desirable, since it allows for price discrimination, which in turn is a key aspect of efficient markets. For the European banking industry, the introduction of stress tests has enabled investors to access relevant and comparable information about banks’ risks. Bank’s main tasks are focused on managing risks. Thus, the stress test results are likely to be of particular interest in order to assess a bank’s condition and value.
Since the assessment of the stress test results still is impacted by investors’ subjectivity, differing interpretations are the result. Therefore, it is useful to consider the incentives and opinions of different market participant.
3.3.2 Investors
In comparison to stockholders, bondholders are less interested in business growth. Bondholders are more interested in banks that avoid a credit default, while maintaining enough earnings to disburse interest payments (Bliss and Flannery (2002)). As a result of such divergence between stock and bondholders, market disciplinary effects might vary depending on the respective capital structure. Accordingly, Ellis and Flannery (1992) present a relief for this concern. By comparing bank equity prices and the according credit default rates they provide evidence that changes in the value of equity and bonds are mainly caused by deviations in expected asset payoffs and not by changes in price volatility. This could explain the parallel movement of stock and bond returns for banks. Especially in the case of subordinated debt, the price correlation with stock returns makes sense. If a bank presents positive risk figures, bondholders might accept lower interest rates, which in turn decreases the interest burden of equity holders. This close link between both security types might explain the co-movement in prices. The co-movement between banks’ stocks and bonds implies that the market influence on management decisions moves in the same direction for both assets. While it is difficult to assign the exact portion of market influence to a security category, one still can measure the combined effect. This study concentrates on equity and subordinated debt. Both types of investors are likely to exert most of the market discipline because their respective stakes have the most direct risk exposure.
3.3.3 Regulatory Authorities
The parties invested in a bank comprise of shareholders, complemented by senior and subordinated debt holders. The main investment objective is to achieve capital gains, given the individual timeframes of investors. In contrast, regulatory authorities, like e.g. the EBA, are not invested in banks nor do they seek capital gains. They pursue a different target that exceeds the foresight of most investors. The overall objective of the EBA is to maintain the financial stability in the EU and safeguard the integrity of the banking sector (European Banking Authority (2016)). The stress tests help the EBA assess data concerning the financial stability of individual banks. Only if such data can be accurately assessed can the EBA start monitoring the risk and, where applicable, implement requirements and thresholds. Studies by Fry et al. (2017) and Schneider et al. (2017) forecast a certain downward pressure on CET1 ratios related to the gradual phasing-in of Basel 3 rules. After the 2016 stress tests the Basel Committee proposed a standardisation in banks’ risk-weighting approaches. The risk-weighted assets (RWA) defined by banks’ internal models showed a wide variability, which going forward will be limited by so-called IRB output floors. McKinsey and KPMG, based on those studies, both estimate a CET1 decrease of up to four percentage points in the next five years, mainly because of higher overall risk weights (Fry et al. (2017)). Tightening regulation affecting the banking sector is not necessarily desirable for investors. In an extreme case, where regulation causes excessive costs, investors might be disinclined to invest in the banking sector. Furthermore, private investors, often being overconfident, tend to ignore warning signs indicating risks (Toft (2002)). Thus, market disciplinary effects might vanish for those investors that regard the methods and assumptions behind stress tests as exaggerated and irrelevant.
3.3.4 Institutional Investors
Apart from private investors, who only represent a small fraction of banks’ capital, one should also consider the market discipline imposed by institutional investors. Institutions that accept funds from third parties and invest them on such parties’ behalf are called institutional investors. Because of their relevant size they represent a major force in many capital markets (OECD (2011)). Studies such as, McCahery, Sautner, and Starks (2016), Theurillat, Corpataux, and Crevoisier (2008), provide evidence that companies with high institutional ownership show better governance. According to Doidge et al. (2017), the incentive to monitor and engage in firms increases with ownership size. Therefore, the likelihood that firms adopt governance reforms increases with institutional ownership. The results suggest that institutional investors attach a high value to proper governance practices. As institutional investors make long-term investments, they must ensure that long-term strategy and governance are sustainable. Accordingly, the EBA stress test results may be of special importance for institutional investors owning large stakes in banks’ capital. Likewise, institutional investors buy and sell large stakes of companies and are therefore well capable of enforcing market discipline.
3.3.5 Government
The relevance of adverse stress scenarios also relies on the credibility of governments promising not to bail-out banks in future. Explicit and implicit government guarantees might decrease the sensitivity of banks’ debt yields since a part of the financial distress risk is transferred to the government. The financial crisis of 2007 showed that the existence of big financial institutions heavily relied on government interventions, financed by tax payers. As a result, supervisory institutions imposed stricter bank capital regulations and transparency measures, like e.g. supervisory stress tests, in order to prevent bank bailouts. In the end, governments might still be tempted to pursue a bailout policy in systemically important situations because of simple cost- benefit considerations. Such decisions obviously depend both on the banks in question and on the government. In Switzerland, the bailout of UBS hit the headlines in 2008 and attracted a lot of public opposition. The rescue act consisting of a $31.8 billion loan was accepted by the Swiss National Bank (SNB). Only five years later the bailout turned out to be a success story, delivering extraordinary earnings of approximately $6 billion to the SNB (Hug (2013)). According to the assessment of the Financial Stability Board, UBS remains on the list of global systemi- cally important banks (FSB (2017)). Due to the small size of Switzerland the bank is even more important on a national level. Although the success was not predictable back in 2008, this example illustrates the positive potential of bank bailouts. The UBS bailout may have impaired the credibility of the government, but the decision benefited ordinary account holders. While the UBS case represents a success, other bailouts, like e.g. Lloyds Bank in the UK, created costs for the tax payer in the amount of $7.5 billion in total (Allen (2013)). To prevent such shortcomings going forward, the EU continues its efforts to reduce the burden of failed banks on tax payers. The bank recovery and resolution directive (BRRD) includes proceedings for setting up a national resolution fund to overcome financial distress. Elke Konig, chair of the Single Resolution Board, highlights the necessity for a standardized insolvency law that foresees bankruptcy in time to ensure the continuity of the banks’ critical functions (Holtschi (2018)). National insolvency law often deviates from the BRRD and makes the banking system more vulnerable to bailouts.
The rating agency Fitch assesses national and bank-specific factors regularly and rates governments’ propensity and ability to support national banks. The so-called ‘bank support rating’ covers a scale of 1 to 5, indicating how likely a bailout is for a given bank. Gropp, Vesala, and Vulpes (2002), Nier and Baumann (2006) tested whether banks with strong government support (i.e. a low Fitch rating) experience less market discipline through subordinated debt yields. Indeed, only banks with weak government support have subordinated debt yields that truly reflect bank risks. This result implies that weak government support creates incentive effects of market discipline. When imposing bailout policies, governments should be aware of possible losses in market efficiency that can deteriorate market disciplinary effects in the long run.
3.3.6 Conclusion
This chapter has analysed the interests of different investors and how those investors might interpret stress test results. Investors can indirectly or directly exert market pressure by adjusting equity prices and bank funding costs. The studies presented show how the government and regulatory authorities provide the basis for efficient capital markets that enable market discipline. Regulation by EBA, including the publicly accessible stress tests, prove to be a trustworthy source of information for investors, among them institutions, bond- and stockholders.
3.4 Reaction of Banks
3.4.1 Capital Adequacy
Stress tests and banking regulation in general evolve around the notion of capital adequacy. Apart from the traditional aspect of capital, as base input for business or production, it has a special significance for financial institutions. The capital invested in a bank provides a quasiinsurance policy against losses to depositors (Hughes and Mester (1998)). Thus, the capital should stand up for the risk exposure of a bank. In other words, there should be an adequate amount of capital depending on the risk exposure. Furthermore, capital attracts depositors and is, thus, part of the main value proposition of a bank. Although the capital ratios have a special meaning for financial institutions, they are not higher when compared to non-financial institutions. On the contrary, listed European financial institutions have mean leverage ratios of 15 (US 10) compared to the 2.8 (US 2.4) ratios of non-banks (Kalemli-Ozcan, Sorensen, and Yesiltas (2012)). Many economists share the opinion that present bank leverage is excessive, and they therefore recommend capital buffers for banks (Admati and Hellwig (2013) and Pflei- derer (2010)). Confronted with stricter Basel 3 capital adequacy regulations and the growing request for risk disclosure, banks have reacted accordingly and European banks have significantly decreased leverage since 2008 (OECD (2018)).
The justification for the relative high leverage ratios in the financial sector rests on the costs and the scalability of bank operations. An increase in deposits does not require a proportional increase in tangible assets, such as servers or working space. Because of this scalability, capital investments tend to be less prevalent than changes in operating costs. Therefore, increases in capital might have little impact on business. Often banks’ excess capital is solely used as a capital buffer. Even though capital buffers are necessary to cover the insurance aspect of a bank, they might create high opportunity costs for shareholders. A justification for high leverage ratios of banks might become less relevant if one considers the one-sidedness of this indicator. The asset side, respectively the riskiness of the portfolio allocation, varies widely between banks. For a holistic view of bank risk, it is therefore imminent to consider asset risk. The RWA approach that results in the CET1 capital ratio represents a widely accepted measure for bank risk. This measure is also used by the EBA to summarize the results of their stress tests. This paper’s hypothesis claims that hypothetical adverse impacts on the CET1 capital ratio influence investors to exert pressure on banks. In the previous section it was shown how investors might interpret stress test results. Subsequently, the possibilities of banks how to cope the with market pressure will be analysed.
3.4.2 Recapitalisation
Investors can send indirect signals to management by trading shares and bonds of banks, thus determining the fair market value of the bank. According to the market discipline paradigm, these signals should be able to influence the banks’ (managers’) decisions (Bliss and Flannery (2002)). The case at hand raises the question if the market signals resulting from the stress test are strong enough to influence banks’ decisions. A bank will voluntarily react with risk mitigation measures only if the action is profitable. Among other strategies, banks can opt for recapitalisation to mitigate risk.
One way for banks to increase the risk-adjusted capital ratio is to use retained earnings as capital. Instead of paying out the yearly earnings as dividends, shareholders can reach an agreement to reinvest part of such earnings. Well-performing banks with strong earnings are in a better position to retain earnings without cutting dividends. However, badly-performing banks might be reluctant to retain earnings, especially when that strategy would require dividend cuts. Dividend cuts tend to induce negative abnormal returns and are announced only if inevitable (Bessler and Nohel (2000)). The period between 2009 and 2012 is characterised by the high amount of retained bank earnings. According to a study carried out by Cohen (2013), these retained earnings where the main driver behind growing risk-weighted capital ratios for that period. The recapitalisation by retained earnings leaves the investor base unchanged, thus avoiding the cost of engaging in equity markets. Furthermore, the intermittent character of earning retentions allows for steady capital growth and flexibility regarding business cycles.
Another way of dealing with increased capital requirements is the issuance of new equity. The capital flowing into the bank either expands total assets or replaces liabilities. If existing bank liabilities are retired without being rolled over, they can be replaced by equity. That strategy has a stronger positive impact on the capital ratio since it reduces bank debt. (Admati et al. (2013)). Furthermore, secondary equity offers (SEO) allow for a faster rebalancing of the capital structure compared to the accumulation of retained earnings. Dinger and Vallascas (2016) analysed the capitalisation strategies of poorly capitalised banks. Given an extensive sample of banks and SEOs they discovered that poorly capitalised banks, for the period between 1993 and 2011, were more likely to issue equity. For banks being very close to the minimum capital ratio they found less SEOs. They conclude that equity issuance is not primarily motivated by stricter capital regulation but rather by market pressure on poorly capitalised banks. One could argue that poorly capitalised banks issue more SEOs because it is less costly for them. Such hypothesis has been refused by Krishnan et al. (2010), instead providing stronger evidence for the influence through market pressure. In their study, they show that both undercapitalised and well-capitalised banks experience similar negative equity shocks after SEO announcements and, hence, face the same implied costs when issuing new equity. A popular opinion regarding the explanation of negative announcement effects that can not only be observed in the banking sector, is based on the information asymmetry between insiders and outside investors. According to Myers and Majluf (1984), managers with superior information about the firm are tempted to issue overvalued equity. That strategy is anticipated by investors and generally results in negative stock price adjustments. Moreover, Krishnan et al. (2010) measured long-term SEO abnormal stock returns. Similar results for well-capitalised and undercapitalised banks indicate that capitalisation levels do not significantly influence the implied cost of SEOs.
One last important risk mitigation strategy that enables banks to recapitalise are Contingent Convertible Bonds (CCB). This financial instrument, being a special form of subordinated debt, converts into common stock if the bank’s capital ratio or share price falls below some prespecified level. As shown in the chapter about subordinated debt, CCBs act as an additional capital buffer, protecting depositors and senior lenders. Based on Article 52 regarding prudential requirements for credit institutions, the EBA defines CCBs as additional Tier 1 capital (European Parliament (2013)). The use of CCBs enables banks to maintain the equity ratio over a wide range of asset value losses. By issuing CCBs ranging from low to high (conservative) triggers, a given solvency standard can be maintained over long periods (Flannery (2005)). Once CCB’s are issued, capital is replenished automatically by reaching the trigger. Such contractual rules act well ahead of financial distress, thus preventing interference with regulatory thresholds. Furthermore, CCBs should incentivise managers to avoid reaching trigger scenarios (Haldane (2011)). Issuing CCBs can therefore strengthen market disciplinary effects.
3.4.3 Asset Reallocation
While equity and recapitalisation are the key measurements when analysing the capital structures of companies, there is another measure that is especially important in the context of banking. For the assessment of banks, one generally uses risk-weighted capital ratios instead of absolute capital ratios. Therefore, the left side of the balance sheet plays an important role when assessing banks’ capital structure. By choosing the composition of total assets, banks can influence the amount of collateral, the value at risk, and the general volatility of the asset portfolio. Those measures, to name a few, characterise the RWA and, thus, indirectly influence the risk-weighted capital ratios of banks.
A holistic view on bank risk can only be obtained if asset risk and allocation are considered. With RWA as a main component of the CET1 capital ratio, one obtains a widely accepted measure for bank risk. In the years after the 2007 financial crisis, the riskiness of banks’ total assets has on average declined (Le Lesle and Avramova (2012)). Such risk decrease is reflected in the decrease of RWA density. RWA density measures RWAs as a percentage of total assets. Because RWA is an indicator that combines many risk characteristics of different assets, it is difficult to separate the drivers behind changes in RWA density. The RWA density decrease can be partly explained by gradual reallocation of bank assets to assets that allow for lower risk-weights. At the same time, economic conditions can also influence RWA density. An economic downturn increases the default probability of assets and, thus, increases risk-weights on average (Le Lesle and Avramova (2012)).
Because credit risk accounts for the largest part of RWA, it is within the credit portfolio that banks can best achieve RWA reductions. A bank facing stricter capital regulations has two main options to decrease RWAs. The first option encompasses changes in the business mix (Le Lesle and Avramova (2012)). Switching from unsecured retail and corporate loans to mortgages and collateralised commercial loans, for instance, can significantly decrease the value at risk of the bank book. The second option consists of shortening the maturity of new loan issuances (Paudel (2007)). A shorter maturity decreases uncertainty regarding credit repayment and therefore has a lower risk-weighting. Both options can also be applied simultaneously when rolling over existing loans.
3.4.4 Conclusion
Because this paper’s hypothesis evolves around the reaction of banks to the EBA stress test, it is necessary to specify the whole range of possible reactions. Banks’ capital structures can be actively developed by the banks’ management. Other than the non-financial sector, banks apply risk-weightings to their capital ratios, thereby considering the riskiness of their balance sheet assets. This augmented perspective on the capital structure extends the scope of managing the risk of a bank. The first choice as to how to enhance the capital structure is recapitalisation. The bank’s equity can be increased by retained earnings or issuance of new equity capital. The second choice, which is also positive for the capital structure, consists of reallocating banks’ existing assets into less risky ones. Both these drivers behind capital ratios have had a strengthening effect on the overall resilience of banks against risk since 2007.
4 Empirical Analysis
4.1 Implementation of the Hypothesis
The present study aims to depict changes in capital structures of banks in response to the stress test results. The EBA stress tests were performed three times during the 2011-2016 period. The regular intervals between the stress tests provide comparisons of the capital structures before and after. Furthermore, the intervals of two to three years enable the banks to react in time with their risk mitigation strategies, which was referred to in chapter three. The chosen period exhibits significant changes in capital structures of banks. This empirical analysis will tackle the difficulty of separating and explaining capital structure trends.
The hypothesis in this paper suggests that the stress test results published will influence the banks’ management’s ongoing risk mitigation efforts. This hypothesis rests on the evidence that stress tests have induced a reaction in bank equity and debt markets, which in turn sends a clear market signal to banks’ management. If the market disciplinary effect resulting from such market signals is strong enough to influence management, it would partly explain capital structure trends. To test the hypothesis, one needs to separate banks that obtained different market signals resulting from stress test results. Moreover, the capital measures that are relevant for the overall capital structure of a bank must be specified. The comparison over time and between different banks ought to shed light on the effectiveness of the EBA stress tests.
4.2 Methodology
The statistical evaluation is mainly based on the public data provided by the EBA. In contrast to the prevalent event study techniques used to measure market reactions on bank stress tests, this examination will apply linear trend estimations since convergence mechanisms are being measured over a predefined period. Having the data of a total of five performed stress tests, only observations from 2011, 2014, 2016 are taken into account. The data from prior years simply lacks in quantity and comparability since the stress tests underwent several modifications (3,200 data items in 2011 compared to 149 data items in 2010). Based on the three datasets, two consecutive periods are being analysed. The presented methodology will be applied equally for both periods.
In both periods, it is tested whether banks react differently after having obtained positive or negative market attention regarding their stress tests results. For statistical comparison, a classification of banks is proposed. Many banks performed well in the stress tests. These banks are, thus, categorised by their strong risk resilience in case of an adverse scenario and are named “well-performing banks”. Contrasting those are “badly-performing banks” which featured bad scores in the stress tests. The classification of well- and badly-performing banks relies on the stress test results. More specifically, the main risk indicator resulting from the ECB stress test is being used, namely the capital depletion resulting from the adverse stress scenario measured in percentage points (pp). It is important to clarify why capital depletion is perceived to be the most informative risk indicator. As a matter of fact, capital ratios like CET1, tier 1, tier 2 and total capital ratio have been used as bank risk indicators long before the EBA stress tests because these ratios are based on the RWA approach with the riskiness of different bank assets being clearly considered. The only shortfall of the RWA approach is that banks do not have standardised models that define realistic risk weights. The EBA stress test results should deliver a more conclusive risk assessment by considering country-specific stress scenarios that apply equally for each bank. This also meets the EBA’s goal to provide market-relevant information.
Having categorised the banks by their stress test score, a way to capture the banks’ reaction regarding their capital structure is specified. The question to ask is, how banks can improve their risk resilience in the limited timeframe between two stress tests. As it seems difficult and counterintuitive for banks to sell off large stakes of their national high-risk mortgages, the banks would rather shift their portfolio structure from equity to more secure assets with collateral or sovereign bonds. This prudent action would reduce the volume of risk-weighted assets compared to total assets. This would also increase the total capital ratio. To measure the reaction of banks, one should focus on total capital ratio changes over time. The total capital ratio captures both, changes in RWA and changes in equity capital through recapitalisation or retained earnings. The total capital ratio consists of tier 1 and tier 2 capital and, thus, includes hybrid instruments as well as subordinated term debt, both of which are effective risk mitigation instruments for banks that should be considered in the capital structure. To determine the main driver behind changes in the total capital ratio, the development of RWA are also measured, with tier 1 and tier 2 capital being measured separately. In order to compare the RWA between banks, a relative measure that divides the RWA through total assets of the banks is used. A high ratio would therefore imply a high percentage of risky assets in the bank book. The data regarding RWA has been retrieved from Orbis Intelligence (2018) and SNL Financial Data (2018).
4.3 Difference-in-differences Approach
For the measurement of changes in banks’ capital structure one could simply apply linear regressions to the data. Because the case at hand compares two bank categories, namely the good and bad performers, the main focus is on the differences between them. A difference-in-differences approach is suitable because it depicts deviations from general capital structure trends. The last two decades show a gradual upward trend in the total capital ratio of EU banks. The difference-in-differences regression used in this paper’s approach shows whether banks developed their capital buffers more or less than the average.
In order to perform the difference-in-differences regression, the population of EU banks is divided into three subsets, whereby two are treatment groups and one is a control group. The two treatment groups are “well-performing banks” and “badly-performing banks”. The control group settles in between the treatment groups and consists of banks that show an average performance in the adverse stress scenario. The two treatment groups and the control group contain approx. the same number of banks. For the period 2011-2014, 20 banks with the lowest CET1 depletion in the 2011 adverse scenario are defined as good performers and those 20 banks with the highest CET1 depletion as bad performers. The control group consists of the 22 banks in between. For the 2014-2016 period, there is a slightly smaller dataset totaling 43 banks. For this period, new groups are also created in line with the 2014 stress test scores.
To depict the changes in capital structure, difference-in-differences regressions are being run for each treatment group while using the same control group. The proposed experimental setting should finally lead to results that explain whether banks’ increased risk mitigation measures on average and how badly-performing banks reacted in contrast to well-performing banks. This approach does not intend to explain the development of specific banks.
4.3.1 Parallel Trend Assumption
The comparison of different bank groups over time only delivers meaningful results under a certain condition. The difference-in-differences approach requires a comparable setting for both the treatment and the control group, meaning that apart from the treatment (good or bad market reaction) both groups should experience similar outside factors influencing the dependent variable. In the regression, the total capital ratio was defined as a dependent variable. It can be argued that treatment and control group banks experience a similar regulatory environment and apply similar risk reporting standards according to the Basel and EBA framework. Since all banks analysed are based in the EU, they should experience similar business cycles over time. The sovereign exposure of certain banks might induce country-specific variations in the dependent variable. As long as these variations are not correlated with receiving treatment and change over time, one can still rely on the average trends that result from the difference-indifferences regression. The only factor that really differs between both groups is the treatment of receiving good or bad capital market reactions. Comparing the pre- and post-treatment period should therefore reveal the causality between market reactions and changes in banks’ capital structures.
By excluding possible confounders like differing regulatory environments and business cycles, one can logically explain a comparable setting for the treatment and the control group. Nevertheless, there could be other factors influencing the dependent variable, which are not so obvious. To exclude other influencing factors, the pre-treatment trend of the dependent variable for the treatment and the control group is being compared. If the trends in total capital are similar for both groups, then there is a strong indicator that treatment and control group have and will continue to experience a comparable setting. This so-called parallel trend assumption is required for running difference-in-differences regression. To verify the parallel trend assumption, Fig. 2 and 3 show the evolution of mean-weighted total capital ratios for the treated and non-treated banks. The vertical line indicates the last datapoint before the 2011 EBA stress test and therefore separates the pre- and post-treatment periods.
Since two separate difference-in-differences regressions are run, the parallel trend assumption is verified twice. Fig. 2 shows well-performing banks, which have received positive market attention in 2011, plotted against the control group. Badly-performing banks are depicted in Fig. 2 accordingly.
Abbildung in dieser Leseprobe nicht enthalten
Fig. 2 Evolution of mean total capital ratio between 2007 and 2013 for good performing banks
Comparing the well-performing banks with the control group in Fig. 2, one can see a similar development of the total capital ratio. The year 2009 shows a very steep curve, meaning that both groups managed to increase their total capital ratio by approx. 2 pp in one year. Spanning over the whole pre-treatment period, an increase in the total capital ratio of 3.5 pp for the “good performers” and 2.3 pp for the control group can be observed. Although the treatment group has increased their capital ratio somewhat faster than the control group, one can argue that the parallel trend assumption is roughly met for this regression. However, for the evaluation one should remember that there might be an outside factor inducing a slightly higher capital growth for the treatment group in general. Apart from the trends seen before, Fig. 2 shows a clear difference in mean total capital ratios between both groups. The difference in means increased from 2007 resulting in a gap of 3.1 pp between treatment and control group in 2010. This indicates that well-performing banks not only have less risky assets according to the stress test but present a larger capital base on average. This two-part risk mitigation approach consisting of high capital buffers and low asset risk applies particularly to the Nordic banking sector. Nordic banks presented good profitability in recent years while having a solid capital adequacy by European standards (Koskinen et al. (2016)). Therefore, unsurprisingly, the categorisation of “well-performing banks” includes many Nordic banks. Out of the 19 best performing banks in the stress test, there is a total of 8 Swedish and Danish Banks.
Abbildung in dieser Leseprobe nicht enthalten
Fig. 3 Evolution of mean total capital ratio between 2007 and 2013 for bad performing banks
The trend analysis in Fig. 3 compares the badly-performing banks with the control group. Apart from the year 2008, where “bad performers” did not increase their capital ratio, there are comparable trends for both groups in the pre-treatment period. Spanning over the whole pre-treatment period, one can observe an increase in the total capital ratio of 1.8 pp for the “bad performers” and 2.3 pp for the control group. Originating from pre-treatment trends, one can argue that the parallel trend assumption is met for this regression. Similar to Fig.2, one can also observe a difference in mean total capital ratios between both groups. This difference should be interpreted as follows: According to the stress test results, the “bad performers” have riskier assets compared to the control group. Hence, the “bad performers” make up for the higher risk by having stronger capital buffers. In comparison to the control group, the “bad performers” have on average a higher total capital relative to their risk-weighted assets. Apparently, this could be interpreted as an overcompensation through capital by “bad performers” to hedge their high-risk assets. The trends continuing from 2010 will be further discussed by way of statistical evidence in the main part of this study.
4.3.2 Empirical Specification
The two-period specification for the difference-in-differences regressions is obtained by using point observations for the year before the stress test publication and for the subsequent years. For the 2011-2014 regression, the total capital ratios at the end of 2010 are being used as pretreatment observations and total capital ratios at the end of 2013 as post-treatment observations. For the 2014-2016 period, 2013 and 2015 total capital ratios are being considered accordingly. The two-year (resp. three-year) period has been chosen in consideration of the time it takes for banks’ management to apply changes in the capital structure. The difference-in-differences regressions are then performed by specification of interaction terms. Accordingly, the estimation of the treatment effect can be extracted from following ordinary least squares regression (OLS):
ATotCapiit =pQ + pTTreatedi + pPPostt + pTPTreatediPostt + sit (1)
To determine the main driver behind capital structure trends, the main components behind the total capital ratio were separated. The terms Tier1 and Tier2 are capital ratios that, when combined, form the total capital. The term RWA is a relative measure showing the RWA divided by the total assets of the banks. These additional OLS regressions were performed as follows:
ATierlit =fi0 + pTTreatedi + pPPostt + pTPTreatediPostt + sit (2)
ATier2iit =fi0 + pTTreatedi + fiPPostt + fiTPTreatediPostt + siit
ARWAiit =pQ + pTTreatedi + pPPostt + pTPTreatediPostt + sit (4)
The output variables used are ratios that are represented in decimals. The index i refers to treatment or control group and the index t refers to the period before or after the treatment. The variable Treatedi is a dummy variable, which takes the value 0 for the control group and 1 for the treatment group. The variable Postt is a period dummy taking the value 0 pre-treatment and the value 1 post-treatment. The interaction term TreatedPostt is a dummy variable being either 0 or 1. A bank in the treatment group post-treatment is assigned with the interaction term 1 and 0 otherwise. The term £i t represents the error inherent in an OLS regression and should be uncorrelated with the interaction term. The coefficient fiTP should represent the key information in the difference-in-differences regression. It shows the average treatment effect of the treated group.
4.4 Reaction of Banks to Supervisory Stress Tests
4.4.1 Summary Statistics - Good Performers
The difference-in-differences regression outputs for the treatment group “well-performing banks” are reported in Table 1 and Table 2. Throughout all tables the standard errors of the coefficients are stated in parentheses. For each output variable (1) to (4), the regression outputs are listed in columns. The sample size in Table 2 for the 2014-2016 period is smaller than the previous period, because there were less banks participating in the 2014 stress test. Furthermore, Regression (4) has a smaller sample size because of data gaps in the databases used. The r-squared interpretation in accordance with Cohen (1992) has widespread acceptance and indicates a medium effect size for the regressions (1, 2), being in the 0.10 to 0.30 r-square range. The regressions (3, 4), however, have a low r-squared value, thus explaining only little of the total data variation.
To start by analyzing the regression outputs in Table 1 and 2, these regressions cover the changes in capital structure for the “well-performing banks”, which per definition have experienced positive market signals. For the coefficient fiTP, one can only find a statistical significance at the 10% level for regression (1) in Table 1. The coefficient of 0.019 indicates a parallel movement between capital structure and being treated. Accordingly, treated banks show an additional total capital ratio rise of 1.9 pp after having been treated when compared to the control group. However, since the significance of the coefficient is too low and the regressions in Table 2 are not significant, one cannot conclude a causal relation between being treated with good market reaction (treatment) and changes in capital structure.
Table 1
Summary of 2011 -2014 capital structure changes. Comparing banks with good scores in stress test to banks with avg. scores with a DiD-regression. The DiD-estimator is represented with interaction term Treated-Post.
Abbildung in dieser Leseprobe nicht enthalten
Data: Copyright © 201*, S&P Global Market Intelligence, Copyright ©, Orbis. Bureau van Dijk, European Banking Authority (EBA) published results
The time trend in the control group represented by pP is positive and significant at the 5% level for the regressions (1) in Table 1 and (1, 2) in Table 2. This can be interpreted as an underlying trend where banks grow capital to reach future capital requirements imposed by regulatory authorities. This trend is ongoing since 2007 and is assumed to be similar for the treatment and the control group according to the evidence in the parallel trend chapter.
For the 2011-2014 period represented in Table 1, one can observe a special property of the RWA (4) from the control group. The coefficient stands for the difference between the two groups’ pre-treatment average RWA ratios. The “well-performing banks” had 11.9 pp higher pre-treatment RWA in relation to total assets compared to the control group. The difference is significant at the 5% level, meaning that the “well-performing banks” had riskier assets than the control group before the stress test. Nevertheless, they performed better in the stress test. This is a clear indicator that the RWA ratio and the stress test have a diverging interpretation of banks’ risk.
Table 2
Summary of2014-2016 capital structure changes. Comparing banks with good scores in stress test to banks with avg. scores with a DiD-regression. The DiD-estimator is represented with interaction term Treated-Post.
Abbildung in dieser Leseprobe nicht enthalten
Data: Copyright © 201*, S&P Global Market Intelligence, Copyright ©, Orbis. Bureau van Dijk, European Banking Authority (EBA) published results
4.4.2 Summary Statistics - Bad Performers
The difference-in-differences regression outputs for the treatment group “badly-performing banks” are reported in Table 3 and Table 4. For each output variable (1) to (4), the regression outputs are listed in columns. Note that regression (4) again has a smaller sample size because of data gaps in the database used. The r-squared interpretation in accordance with Cohen (1992) indicates a medium-sized effect for the regressions (1, 2) only for the 2014-2016 period. Therefore, one should more focus on Table 4 for a causal interpretation later on.
The regressions in Tables 3 and 4 show the changes in capital structure for the “badly-performing banks”. These banks performed poorly in the stress test and therefore experienced negative market signals. Against the hypothesis, one can find that the coefficient fiTP is not statistically significant. The treatment, referred to as a negative market signal, seems to have no positive impact on banks’ total capital ratio. Moreover, looking at regression (2) in table 3, one can even find statistical significance against the hypothesis. For the 2011-2014 period, the “bad performers” showed a 2.8 pp lower Tier 1 capital increase than the control group, hence signaling an even higher risk to investors.
Table 3
Summary of 2011-2014 capital structure changes. Comparing banks with bad scores in stress test to banks with avg. scores with a DiD-regression. The DiD-estimator is represented with interaction term Treated-Post.
Abbildung in dieser Leseprobe nicht enthalten
Data: Copyright © 201*, S&P Global Market Intelligence, Copyright ©, Orbis. Bureau van Dijk, European Banking Authority (EBA) published results
One feature that stands out in Tables 3 and 4 is the unequal starting position with regard to the capitalisation of the treatment and the control group before the treatment. The coefficient in regression (1, 2) stands for the difference between the two groups pre-treatment capital ratios. The “badly-performing banks” showed 3 pp respectively 5.4 pp higher total capital ratios for the first and the second period compared to the control group. This difference is significant at the 10% respectively 5% level for the first and the second period. The difference is mainly attributed to the differences in Tier 1 capital between treatment and control group. This difference amounts to 2.2 pp for the first period and 4.5 pp for the second period as shown in regression (2) of Tables 3 and 4.
Table 4
Summary of2014-2016 capital structure changes. Comparing banks with bad scores in stress test to banks with avg. scores with a DiD-regression. The DiD-estimator is represented with interaction term Treated-Post.
Abbildung in dieser Leseprobe nicht enthalten
Turning to regression (4) in Tables 3 and 4, where changes in the RWA ratio are being measured, one did not find any statistically significant differences between the treatment and the control group. The control group indicates that there is a negative time trend in the RWA ratio, represented by . While not being significant, the RWA ratio of the control group still has decreased on average over both time periods. This means that the riskiness of the assets has decreased slightly over the 5-year time horizon, resulting in lower risk weights and RWA ratios.
4.4.3 Findings
In this chapter, it has been statistically tested how banks adapted their capital structure in response to the stress test results. By means of a difference-in-differences regression deviations from general capital structure trends have been analysed to test whether the treatment provoked a significant reaction of banks.
The hypothesis that banks with low stress test scores have a stronger incentive to increase their total capital has been rejected. Moreover, one can even find weak statistical significance against this hypothesis when looking at the Tier 1 capital ratio. This ratio decreased in the period 2011-2014 for the average “badly-performing bank”. Neither did the difference-in-differences regression of “well-performing banks” reveal the expected results. The “well-performing banks” even succeed in increasing their total capital ratio further on average than the control group, although, as pointed out in this chapter, the statistical significance is too weak to deduct a causal explanation of that result. Hence, one must remain uncertain in terms of possible reactive measures conducted by banks after the publication of the stress tests. While banks most certainly have reacted, the average reaction might not be reflected solely in capital structure changes.
It has also been analysed if a convergence in risk exposure between the different bank groups can be observed. Therefore, the risk exposure of the asset side has been captured by dividing RWA through the total assets of the individual banks. The coefficient capturing the difference- in-difference for this RWA ratio shows significant differences between treatment and control groups.
The informative value of the results is possibly affected by the limited sample size of this study. 84 respectively 56 participating banks in the 2011 and 2014 stress tests represent a lower limit to perform a quantitative analysis, especially when considering the size and country-specific heterogeneity. Despite the limited sample size, significant differences between treatment and control group could be determined with regard to pre-treatment statistics. The differences between the two groups’ pre-treatment capital ratios are an important aspect when interpreting capital structure dynamics.
As an example, the “well-performing banks” had a 11.9 pp higher pre-treatment RWA in relation to total assets compared to the control group. Furthermore, a repeated significance for pretreatment differences in total capital ratios and Tier 1 ratios between “badly-performing banks” and the control group was found. The “badly-performing banks” appear to have had stronger capital ratios already before the 2011 stress test. This fact suggests that the “badly-performing banks” had already reacted in advance with capital increases due to their generally higher risk exposure. The fact that “bad performers” have a higher total capital relative to their risk- weighted assets on average is also visible in chapter 4.3.1 (parallel trend assumption). The divergence in the capital structures between treatment and control group remained reasonably constant for the 2007-2010 period, therefore not indicating a significant reaction of banks for the four years before the first comprehensive stress test in 2011. The causes that have led to an overcompensation of the risk with capital might therefore be located further back in time.
While no significant changes in capital ratios could be determined in the period between stress tests, a significant pre-treatment divergence in capital ratios between treatment and control groups could be found. To conclude, a disproportionately positive relation between capital buffers and risk exposure of banks could be observed. This indicates that higher capital buffers are an important tool for bad performers to justify or at least maintain increased risk exposure. Furthermore, the risk exposure captured by the RWA ratios disclosed no significant differences between treatment and control groups. This observation coincides with the results of (Matejasak (2015)). Using the example of Czech banks, he shows that changes in risk exposures alone explain only a small portion of changes in the overall capital structure.
4.4.4 Deviations from the Hypothesis
The empirical evaluation of the hypothesis has given insight into differences in risk exposure and capital between those banks that performed well and those that performed badly in the stress test. What the statistics do not show, however, is the banks’ reaction within the tested time period. Since this is the main effect that the hypothesis intended to test, it is important to discuss possible reasons for such deviation from the hypothesis. In that respect, it is important to question the assumptions made, but also to critically examine the information content of the risk indicators used for this examination. Based on the theory chapter, it will now be outlined why the expected reaction of banks was not observed.
The premise of this evaluation emerges from the ongoing research in bank regulation. Several studies reveal that banks with a severe CET1 impact in the adverse stress scenario experienced certain market reactions. Georgescu et al. (2017) reveal significantly negative higher CDS abnormal returns for badly-performing banks. Petrella and Resti (2013) describe a significant reaction of stock markets after the publication of the 2011 EU stress tests. Abnormal returns reward banks that experienced, among other factors, a lower drop in the CET1 ratio in the adverse scenario. Typically, such impact translates into higher bank funding costs. The new information revealed in the stress test results has provoked a change in investors’ perceptions. The price discrimination observed between banks exerts pressure on the banks that performed poorly in the stress test. It was such pressure that was thought would influence banks’ behaviour in the period between the stress tests. Clearly, the price pressure caused by the market would need to be sufficiently strong to effect a change in the banks’ capital structures.
Insufficient market pressure might explain the absence of banks’ reactions in the evaluation period. EU banks may simply not have been impressed by the market reaction or they may have accepted the changed market perception. There are, for instance, banks that pursue a riskier asset allocation because of strategic reasons or perhaps because the national mortgage market is riskier. Those banks as well as the market are aware of the increased risk exposure, which in turn leads to relatively higher bank funding costs. These funding costs may well be served by the bank if the riskier asset allocation yields a higher profit accordingly. Such behaviour can be viewed as a type of market positioning where banks are able to choose and maintain the risk of their assets. The wide use of CET1 and total capital ratios of EU banks shows that, even before the stress tests, banks were able to sustain their business over long periods without the need of converging to a common risk exposure.
If the market pressure induced by the stress tests was insufficient to influence bank managers’ decisions, one should consequently query whether the informative value of the stress test scores is relevant. One crucial result in the evaluation is an explanatory approach in this respect. The evaluation revealed significant differences between treatment and control groups in terms of pre-treatment statistics. Especially the “badly-performing banks” showed 3-5 pp higher total capital ratios than the control group before the stress test results were publicised. It appears peculiar that the group performing worst in the stress test still exhibits an above average total capital ratio. This observation underlines the two-sidedness in the risk assessment approach of banks. Such two-sidedness in terms of asset side risk and capital side protection becomes complex when combined in a single measure like e.g. CET1 ratio or total capital ratio. Relying solely on such combined risk measures for the risk assessment of a bank would imply that a higher risk exposure on the asset side could simply be compensated by increasing capital. This paper’s findings, however, show that banks with risky assets choose to have more capital in proportion to the risk-weighted assets. This could be regarded as a strategy to reassure debt holders that the bank still has sufficient capital in case of an adverse scenario. The exact tradeoff between the asset and the capital side is hard to capture. The risk exposure on the asset side is controlled by the bank and relies on the micro- and macroeconomic situation, while bank capital mainly represents a risk limitation function for debt holders, depositors and for the financial system. Eventually, both sides of a bank’s balance sheet are susceptible to risk mitigation measures.
In the evaluation, the emphasis was on the asset side of the banks tested by interpreting the impact on the CET1 ratio in case of an adverse scenario. The impact is on the CET1 ratio and arises from the assets and their corresponding risk exposure. The measure used, also referred to as CET1 depletion, was applied to classify the banks into treatment and control groups. Besides the CET1 depletion for the baseline scenario, the CET1 depletion for the adverse scenario is referred to as a key measure summarizing the results of the stress tests. In fact, the total effect of the adverse scenario simulations is compiled in the CET1 depletion measure. It was expected that this measure would be the main trigger of market reactions, followed by bank reactions because it contains or at least summarizes all of the new information provided by the EBA stress tests. On closer inspection, however, the CET1 depletion measure alone may be an incomplete risk indicator for the purpose of this examination. In order to judge the risk resilience of a bank, it is necessary to consider CET1 depletion respectively asset risk exposure as well as the size of banks’ capital buffers. In the case at hand, those banks classified as “bad performers” had relatively high capital buffers already before the stress test. The classification might have turned out differently if the initial size of the banks’ capital buffers had somehow been integrated in the used stress test scores.
A holistic interpretation of stress test results is feasible by considering both asset risk exposure and capital buffers. This, however, would present new difficulties regarding the experimental design. Petrella and Resti (2013) faced a similar decision when measuring the stock price reaction to stress test results. By compiling the initial capital ratio and the capital depletion in the adverse scenario they obtained the so-called stressed capital ratio. That ratio is calculated by subtracting the capital depletion amount from the initial capital ratio. The problem is that this measure combines publicly available information (initial capital ratio) as well as new information (capital depletion) as presented in the stress test results. As expected, the initial capital ratio did not turn out to be a significant determinant of the market reaction. The new information concealed in the capital depletion alone could explain some of the stock market reactions. In this study, these market reactions did, however, not create sufficient market discipline for banks to react with changes to their capital structure. Especially those banks with a high initial capital ratio might not see any necessity to improve their risk mitigation measures.
One last possible explanation for the deviation from the hypothesis deals with the risk modeled in the adverse scenario. The capital depletion gives insight into the risk exposure of total bank assets. Before the stress tests were implemented in 2009 there has already been common models to simulate market risks and deduce the riskiness of banks’ assets. Models such as the VLAB and VaR continue to be widespread tools in risk management. As these models have proven themselves over time, they are still applied by market participants. Accordingly, it is conceivable that the market will tend to adhere to long-established risk assessment models. This would reduce the information relevance of the EBA stress test results. In a study by Acharya, Pierret and Steffen (2016), the authors compared the 2016 EBA stress testing methodology with a market-based risk assessment approach. Using the VLAB methodology, they measured the downside risk of European banks. Such downside risk was then compared to the capital depletion projected by the EBA stress tests. Finally, they compared the capital shortfalls resulting from these models and discovered that the VLAB methodology applied larger stressed capital losses than the EBA methodology. However, the VLAB methodology ranks banks similarly to the EBA stress test results (Acharya, Pierret and Steffen (2016)). A comparison of the two approaches suggests that investors apply alternative risk assessment methods that can be based on different assumptions. Therefore, the informative value of the stress test results might only persist if the underlying methodologies stand the test against long-established risk assessment methods.
In summary, three possible explanations for the deviation can be identified. The first explanation is that the market pressure induced by the stress tests was insufficient to influence the capital structure of the banks. This in turn would imply that the informative value of the stress test scores is not relevant enough. The second explanation emphasizes the shortcomings in the information content of the risk measure used and called capital depletion. Accordingly, a possible holistic interpretation of stress test results was proposed, taking into account asset risk exposure and initial capital buffers of banks. The third explanation deals with the risk modeled in the adverse scenario. There it is taken into consideration that other risk assessment models might compete with the stress test scenarios used by the EBA. The popularity of other risk assessment models alone can attract market attention, thus displacing the information published in the EBA stress tests. The three explanations presented may all apply to a certain degree, although the limited evidence might come from omitted variables or from the experimental setting itself.
5 Final Remarks
5.1 Conclusion
Since banks’ risks are difficult to perceive, difficult to compare and also change rapidly, investors and depositors are more reliant on regular information on a bank’s risks. The standard approaches consisting of regulatory capital ratios and internally performed risk assessment lost credibility during the 2007 financial crisis. The EBA stress tests introduced in 2009 aim to provide clarity by disclosing results subdivided in different risk categories and by presenting a standard approach for modeling risk exposure of European banks. The EBA stress tests are performed and disclosed approx. every two years. After several studies published, analysing the market reactions to stress test disclosures, one took it a step further by looking at another important reaction. The investigation aims to measure and explain how banks reacted with regard to their performance in the EBA stress tests.
The EBA plays a significant role in monitoring European banks. Among other publications, the stress test results should contribute to the transparency of financial institutions. A key result of the EBA stress tests is the impact of a hypothetical adverse scenario on CET1 capital ratio. The adverse scenario is modeled based on historical data and considers individual parameters like, for instance, equity prices, exchange rates, housing market prices and exchange rates. In a crisis, such parameters have an overall negative impact on valuation and profits, thus depleting the available capital buffer.
The theoretical framework suggests that the stress test results would induce market discipline. The concept of market discipline relies on the assumption that market prices can affect the status of a bank and the decision of its managers. The bank’s management acts in the interest of investors and depositors, while the regulation authorities specify the scope of action of banks. The EBA stress tests prove to be a trustworthy source of information for investors, among them institutional investors, bond- and stockholders. In order to see whether the capital market reaction of these investors requires counter-measures by banks, a statistical evaluation has been performed.
A difference-in-differences approach has been used to measure capital structure trends for the 2011-2016 period. These capital structure trends should shed light on risk mitigation measures performed by banks in the aftermath of the stress tests. Possible risk mitigation strategies include the rebalancing of the portfolio to more secure assets as well as increasing capital buffers. These portfolio and capital structure changes are reflected in the total capital ratios of banks. Banks have been categorised into those with good and those with bad scores in the stress tests. The hypothesis suggests that banks with bad stress test scores might have a stronger incentive to mitigate risk. Risk mitigation measures would subsequently be reflected in the capital structure. The analysis of the banks’ capital structures eventually delivered the results for the interpretation.
The hypothesis that banks with low stress test scores have a stronger incentive to increase their total capital has been rejected. Neither banks with good nor bad stress test scores showed significant changes in the capital structure compared to the control group. However, it has been discovered that pre-treatment differences in capital buffers exist between treatment and control groups. The pre-treatment capital ratios are an important aspect when interpreting the capital structure of banks. Banks with low stress test scores seem to have had already reacted long before the publication of the 2011 stress tests by building up disproportionately high capital buffers. This primary reaction could be explained by the fact that the market and banks already had been using other models for risk assessment and mitigation for some time.
5.2 Critical Appraisal
The goal of the EBA stress tests is well defined and consists of providing market-relevant information about banks’ risk. The EBA sets market discipline above direct intervention when it comes to ensure the stability of EU banks. The concept of market discipline is quite pervasive in the banking literature because of the powerful property of self-regulation. In the course of this thesis, there have been repeatedly challenges in measuring the market discipline of banks.
The difference-in-differences approach chosen to measure capital structure changes requires a classification of the banks into treatment and control groups. The definition of treatment is the negative or positive market attention that banks experienced after the stress test. In this regard, the classification could have been done more precisely, if it had been known exactly which banks experienced significant stock market shocks. Instead, a classification of best or worst in class banks was defined, assuming they on average experienced positive or negative market reactions. While this assumption relies on the results of other studies, that approach might, for that reason, be limited in expressive power.
The results presented are possibly affected by the limited sample size of this study. Although the EU has a heterogeneous banking industry, the number of participating banks is limited and has diminished between 2011 and 2014. Furthermore, the concentration of Nordic banks within one of the categorizations signals that national particularities possibly have an impact on the banks’ capital structures. This could, in the worst case, lead to an omitted variable bias in the worst case.
5.3 Research Outlook
Supervisory stress tests such as those performed by the EBA and the US Federal Reserve Board are set to become key drivers in safeguarding international financial stability. The macro perspective of the stress tests should disclose vulnerabilities not yet considered by market participants. The research in this field is still in its early stages and so are the methodologies and scenarios behind the EBA stress tests themselves. In this regard, the EBA supervisory board established a review panel that periodically conducts peer reviews. These reviews contribute to the course of research by testing the technical supervisory standards and by analysing best practices. Previous EBA bank stress tests were generally not binding in the sense of requiring minimal stress test scores. This testing strategy allowed the EBA to identify to what extent banks can regulate themselves through market transparency without fearing disciplinary measures. Future research is needed to define effective disciplinary measures that complement the stress test results. Among other things, these measures involve standardized and binding insolvency laws that ensure the continuity of stressed banks’ critical functions. The latest Fed stress test results have already led to disciplinary measures that restrict the use of future profits. Future financial downturns will show if minimum capital requirements serve their purpose.
Data Directory
European Banking Authority, 2011, Results of the 2011 EBA EU-wide stress test from website, https://www.eba.europa.eu/risk-analysis-and-data/eu-wide-stress-testing/2011, 22.03.2018.
European Commission, 2018, State Aid Cases from website, http://ec.europa.eu/competition/state_aid/register/, 16.06.2018.
OECD, 2018, Banking sector leverage (indicator) from website, https://data.oecd.org/corpo rate/banking-sector-leverage.htm, 26.04.2018.
Orbis Intelligence, 2018, Bank and Financial Data from website, https://orbis2.bvdep.com/, 12.06.2018.
SNL Financial Data, 2018, Bank and Financial Data from website, https://www.snl.com/, 12.06.2018.
Reference List
Acharya, Viral V, Diane Pierret, and Sascha Steffen, 2016, Introducing the “Leverage Ratio” in assessing the capital adequacy of European banks, Technical Report (August 01), Mimeo ZEW.
Admati, Anat R., Martin Hellwig , 2013. The Bankers’ New Clothes: What's Wrong with Banking and What to Do about It (Princeton University Press. Princeton, NJ).
Admati, Anat R., Peter M. Demarzo, Martin F. Hellwig, and Paul Pfleiderer, 2013, Fallacies, Irrelevant Facts, and Myths in the Discussion of Capital Regulation: Why Bank Equity is Not Socially Expensive, Graduate School of Business Research Paper No. 13-7, Stanford University.
Alexander, Carol, and Elizabeth Sheedy, 2008, Developing a stress testing framework based on market risk models, Journal of Banking & Finance 32, 2220-2236.
Allen, Matthew, 2013, Swiss bank bailout turned poison into profit from website, https://www.swissinfo.ch, 16.06.2018.
Ashcraft, Adam B., 2006, Does the Market Discipline Banks? New Evidence from Regulatory Capital Mix, Journal of Financial Intermediation 17, 543-561.
Avery, Robert B., Terrence M Belton, and Michael A Goldberg, 1988, Market Discipline in Regulating Bank Risk: New Evidence from the Capital Markets, Journal of Money, Credit and Banking 20, 597-610.
Basel Committee on Banking Supervision, 2006, International Convergence of Capital Measurement and Capital Standards from website, https://www.bis.org/publ/bcbs128.htm, 14.06.2018.
Bessler, Wolfgang, and Tom Nohel, 2000, Asymmetric information, dividend reductions, and contagion effects in bank stock returns, Journal of Banking & Finance 24, 1831-1848.
Birchler, Urs W., and Diana Hancock, 2003, What Does the Yield on Subordinated Bank Debt Measure? from website, https://www.federalreserve.gov/pubs/feds/ 2004/200419/200419pap.pdf, 09.06.2018.
Bird, Andrew, Stephen A Karolyi, Thomas G Ruchti, and Austin Sudbury, 2015, Bank regulator bias and the efficacy of stress test disclosures from website, http://dx.doi.org/10.2139/ssrn.2626058, 14.04.2018.
Blaschke, Winfrid, Matthew T. Jones, Givanni Majnoni, and Maria S. Martinez, 2001, Stress Testing of Financial Systems: An Overview of Issues, Methodologies, and FSAP Experiences, Working Paper no. 01/88, IMF.
Bliss, R. R., and M. J. Flannery, 2002, Market Discipline in the Governance of U.S. Bank Holding Companies: Monitoring vs. Influencing, Review of Finance 6, 361-396.
Breuer, Thomas, and Martin Jandacka, 2012, How to Find Plausible, Severe, and Useful Stress Scenarios. 1st ed. (Bibliogov, Vienna).
Camara, Boubacar, Pierre Pessarossi, and Thomas Philippon, 2017, "Back-testing bank stress tests.", Working Paper No. 23083, NBER.
Candelon, Bertrand, and Amadou N Sy, 2015, How Did Markets React to Stress Tests?, Working Paper no. 15/75, IMF.
Cohen, Benjamin H., 2013, How have banks adjusted to higher capital requirements?, BIS Quarterly Review (September 2013), 25-41.
Cohen, Jacob, 1992, A power primer. Psychological bulletin, 112(1), 155-159.
Committee on Banking Supervision, 2018, Pillar 3 disclosure requirements from website, https://www.bis.org/bcbs/publ/d432.htm, 13.06.2018.
Committee on the Global Financial System, 2001, A survey of stress tests and current practice at major financial institutions from website, https://www.bis.org/publ/cgfs18.htm, 08.05.2018.
Dinger, Valeriya, and Francesco Vallascas, 2016, Do Banks Issue Equity When They Are Poorly Capitalized?, Journal of Financial and Quantitative Analysis 51, 1575-1609.
Doidge, Craig, Alexander Dyck, Hamed Mahmudi, and Aazam Virani, 2017, Collective Action and Governance Activism, Working Paper No. 2635662, Rotman School of Management.
Ellahie, Atif, 2013, Capital Market Consequences of EU Bank Stress Tests from website https://ssrn.com/abstract=2157715, 22.06.2018.
Ellis, David M., and Mark J. Flannery, 1992, Does the debt market assess large banks’ risk? Time series evidence from money center CDs, Journal of Monetary Economics 30, 181— 502.
Enria, Andrea, Adam Farkas, and Lars Overby, 2016, Sovereign Risk: Black Swans and White Elephants, Sovereign and Banking Risks: What Policies? 1, 51-72.
European Banking Authority, 2016, 2016 EU - Wide Stress Test Methodological Note from website, https://www.eba.europa.eu, 02.05.2018.
European Banking Authority, 2017, Report on Convergence of Supervisory Activities from website,https://www.eba.europa.eu, 02.05.2018.
European Parliament, 2013, Regulation (EU) No 575/2013 from website, https://eur- lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:32013R0575&from=en, 19.04.2018.
European Systemic Risk Board, 2016, Adverse macro-financial scenario for the EBA 2016 EUwide bank stress testing exercise. https://www.eba.europa.eu, 09.03.2018.
Fender, Ingo, Michael S. Gibson, and Patricia C. Mosser, 2001, An International Survey of Stress Tests, Current Issues in Economics and Finance 7, 34-39.
Flannery, Mark J., 2005, Capital Adequacy beyond Basel: Banking, Securities, and Insurance No pain, no gain? Effecting market discipline via reverse convertible debentures (Oxford Univ. Press, Oxford).
Fry, Fiona, Daniel Quinten, Clive Briault, and David Nicolaus, 2017, Basel Committee finalises output floor, credit risk and operational risk from website https://home.kpmg.com/xx/en/home/insights/2017/12/basel-committee-finalises-standa rds-fs.html, 13.06.2018.
FSB, 2017, List of global systemically important banks (G-SIBs) from website http://www.fsb.org/wp-content/uploads/P211117-1.pdf, 02.06.2018.
Georgescu, Oana-Maria, Marco Gross, Daniel Kapp, and Christoffer Kok, 2017, Do stress tests matter? Evidence from the 2014 and 2016 stress tests, Working Paper Series no 2054/May 2017, European Central Bank.
Glasserman, Paul, and Gowtham Tangirala, 2016, Are the Federal Reserve’s stress test results predictable?, The Journal of Alternative Investments, 18(4), 82-97.
Gropp, Reint, Jukka Vesala, and Giuseppe Vulpes, 2002, Equity and bond market signals as leading indicators of bank fragility, Working Paper Series no. 150, European Central Bank.
Haldane, Andrew, 2011, Capital Discipline - Speech given at the American Economic Association, Denver, Colorado.
Holtschi, Rene, 2018, Europas oberste Bankenabwicklerin zieht erste Lehren, NZZ online, April 05.
Hug, Daniel, 2013, UBS-Rettung zahlt sich aus, NZZ, August 04.
Hughes, Joseph P., and Loretta J Mester, 1998, Bank capitalization and cost: Evidence of scale economies in risk management and signaling, Review of Economics and Statistics 80, 314325.
Kalemli-Ozcan, Sebnem, Bent Sorensen, and Sevcan Yesiltas, 2012, Leverage Across Firms, Banks and Countries, Journal of International Economics 88, 284-298.
Koskinen, Kimmo, Hanna Putkuri, Pertti Pylkkonen, and Eero Tolo, 2016, Nordic financial sector vulnerable to housing market risks, Bank of Finland Bulletin 2, 12.
Krishnan, CNV, O Emre Ergungor, Paul A Laux, Ajai K Singh, and Allan A Zebedee, 2010, Examining bank SEOs: Are offers made by undercapitalized banks different?, Journal of Financial Intermediation 19, 207-234.
Le Lesle, Vanessa, and Sofiya Yurievna Avramova, 2012, Revisiting Risk-Weighted Assets, Working Paper no. 12/90, IMF
Matejasak, Milan, 2015, Did the Czech and Slovak Banks Increase their Capital Ratios by Decreasing Risk, Increasing Capital or Both?, Procedia Economics and Finance 25, 256263.
McCahery, Joseph A., Zacharias Sautner, and Laura T. Starks, 2016, Behind the Scenes: The Corporate Governance Preferences of Institutional Investors, The Journal of Finance 71, 2905-2932.
Morgan, Donald P, Stavors Peristiani, and Vanessa Savino, 2014, The Information Value of the Stress Test, Journal of Money, Credit and Banking, 46, 1479-1500.
Myers, Stewart, and Nicholas Majluf, 1984, Corporate Financing and Investment Decisions When Firms Have Information That Investors Do Not Have, Journal of Financial Economics 13, 187-221.
Neretina, Ekaterina, Cenkhan Sahin, and Jakob De Haan, 2015, Banking stress test effects on returns and risks, Working Paper no. 419, De Nederlandsche Bank.
Nier, Erlend, and Ursel Baumann, 2006, Market discipline, disclosure and moral hazard in banking, Journal of Financial Intermediation 15, 332-361.
OECD (2011), The Role of Institutional Investors in Promoting Good Corporate Governance, Corporate Governance (OECD Publishing).
Paudel, Youbaraj, 2007, Minimum Capital Requirement Basel II - Credit Default Model& its Application, BMI Paper July 2007, Vrije Universiteit.
Petrella, Giovanni, and Andrea Resti, 2013, Supervisors as information producers: Do stress tests reduce bank opaqueness?, Journal of Banking & Finance 37, 5406-5420.
Pfleiderer, Paul, 2010. On the relevancy of Modigliani and Miller to Banking: A parable and some observations, Working Paper no. 93, Stanford University.
Schneider, Sebastian, Gerhard Schrock, Stefan Koch, and Roland Schneider, 2017, Basel “IV”: What’s next for banks?, Global Risk Practice April 2017, Mckinsey.
Steiner, Margaux, and Marjolaine Marra, 2017, Determinants of the spread of CET1 for European Banks - Quantitative study based on the 2016 EU-wide Stress test, Master Thesis, Umea School of Business and Economics.
The European Commission, 2014, Commission Delegated Regulation (EU), Official Journal of the European Union 241.
Theurillat, Thierry, Jose Corpataux, and Olivier Crevoisier, 2008, The Impact of Institutional Investors on Corporate Governance: A View of Swiss Pension Funds in a Changing Financial Environment, Competition & Change 12, 307-327.
Thompson, Christopher, 2013, Big rise in subordinated debt issuance by EU banks, Financial Times, December 12.
Toft, Brian, 2002, Financial Risks, Decisions and Behaviour, Risk management: An International Journal 4, 7-15.
Turk-Ariss, Rima, 2017, Heterogeneity of Bank Risk Weights in the EU: Evidence by Asset Class and Country of Counterparty Exposure, Working Paper no. 17/137 , IMF.
Vlad, Costica, 2012, Bank Stress Tests Methodology Used in the Euro Area, Ovidius University Annuals, Series Economic Sciences 12, 1756-1761.
-
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X. -
Upload your own papers! Earn money and win an iPhone X.