Incentivizing depositor discipline: are stress tests really helping?
An empirical study on depositor behavior following the 2010 and 2011 EU-‐wide stress testsMSc Thesis Economics
Specialization: Monetary Policy and Banking
Name: Harriëtte ten Brinke
Student number: 6132030
Date: July 24
th, 2015
Supervisor: prof. dr. S.J.G. van Wijnbergen
Second reader: dr. W.E. Romp
STATEMENT OF ORIGINALITY
This document is written by Student Harriëtte Willemijn ten Brinke who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
ABSTRACT
This thesis studies the effect on the 2010 and 2011 EU-‐wide stress tests on depositor discipline. The intention of the disclosure of stress tests results is to increase transparency and market discipline, but the tests were widely thought to lack credibility. Depositors exert discipline via the quantity and price of deposits. This study focuses at the former, in which discipline is measured by the deposit growth rate. Using panel data from 295 tested and non-‐tested European banks, an event study and regression analysis by pooled OLS are performed. In general, evidence suggests that depositor discipline based on bank fundamentals exists but its effectiveness is questionable. Furthermore, in line with the theory, insured depositors seem to exert no discipline, except at stress-‐tested banks, whereas uninsured depositors do exert discipline. Concerning the stress tests, only inclusion in the stress test of 2010 has a significant effect on the overall deposit growth rate. When a distinction is made between insured and uninsured depositors, evidence suggests neither insured nor uninsured depositors react to stress tests.
TABLE OF CONTENTS
1. Introduction 1
2. Stress tests 3
2.1 The need for stress tests 3
2.2 Stress test design 3
2.3 The European Union stress tests of 2010 and 2011 5
3. Market discipline & depositor discipline 8
3.1 Market discipline 8
3.2 Depositor discipline: a model 9
3.3 Depositor discipline: selective literature review 10
3.3.1 Transparency 11
3.3.2 Too Big To Fail and Too Protected To Fail 11
3.3.3 Deposit insurance 13 4. Methodology 14 4.1 Event study 14 4.2 Estimation strategy 15 4.3 Identification 16 5. Variables description 18 5.1 Deposit variables 18
5.2 Bank fundamental variables 19
5.3 Macroeconomic variables 20
5.4 Stress test variables 21
5.5 Other variables 21
6. Data & descriptive statistics 23
6.1 Data 23
6.2 Descriptive statistics 24
7. Empirical results & discussion 27
7.1 Event study 27 7.2 Regression analysis 30 8. Robustness 37 9. Conclusion 38
10. Limitations & Recommendations 40
References 42 Appendix A 46 Appendix B 47 Appendix C 48 Appendix D 49 Appendix E 52
INTRODUCTION
"The EBA continues its efforts to enhance transparency in EU banking. Developing common definitions and regular disclosure are fundamental to support effective market discipline.’’
Andrea Enria, chairperson of the EBA1 In the recent years, the European Banking Authority (EBA) has been on a quest to increase transparency in the EU banking sector. As such, it has undertaken several actions to decrease the opacity surrounding EU banking. An example is the stress testing of financial institutions, in which the resilience of banks is tested by subjecting holdings to multiple, often adverse, scenarios. The test result, i.e. a capital ratio, indicates which banks are weak and which can survive such adverse market conditions. It signals market participants about the situation of the bank and its risk-‐taking behavior (EBA, 2015, EBA, 2011a). This is where market discipline comes into play: market participants should discipline banks that are too risky. Market discipline has several advantages: it mitigates the incentive for excessive risk-‐taking and encourages efficiency in banking, as well as reducing the costs of regulatory supervision through the existence of another disciplining force (Martinez Peria and Schmukler, 2001). In theory, it is the perfect collaboration between the market and the regulators. In practice, however, there are some caveats. Although the majority of banks passed the 2011 EU-‐wide test, the market still reacted negatively. This may be influenced by the results of the 2010 and 2011 tests’ perceived lack of credibility. Banks that passed in the hypothetical scenario, such as Dexia or Irish banks, failed in real life not very long after disclosing the results (Candelon and Sy, 2015). The scenarios were found to be too lenient; the possibility of sovereign default was not included in any scenario, the adverse scenario was seen as ‘mild’ and the benchmark rate was low (Ong and Pazarbasioglu, 2014). All in all, this creates unfavorable conditions for effective market discipline, as illustrated by the decrease in stock prices for stress-‐tested banks in 2011 (Candelon and Sy, 2015). However, stockholders are not the only actors in market discipline; it can also be exerted by depositors, either through the withdrawal of funds or the demand of a higher interest rate on deposits. In the EU deposits are insured against bank failure, with a maximum of €100.000 per depositor, per bank under the so-‐called Deposit Guarantee Scheme (DGS) (EC, 2014). This creates an extra caveat for effective market discipline in relation to stress test results: if some deposits are insured, why bother about the state of the bank?
The aim of this thesis is to find whether effective depositor discipline prevailed after disclosure of the EU stress test results of 2010 and 2011. An event study is conducted to analyze the path of deposits in relation to disclosure of stress test results. Furthermore, the effects of
1 As found on https://www.eba.europa.eu/-‐/eba-‐publishes-‐outcome-‐of-‐2013-‐eu-‐wide-‐transparency-‐exerci-‐1 2 A regression similar to Martinez Peria and Schmukler (2001) will also be performed. It excludes stress test variables
inclusion in the stress tests and the outcome of the stress tests on the deposit growth rate are studied. In doing so, unbalanced panel data of 295 financial institutions from the first quarter of 2008 to the third quarter of 2014 is used to estimate the effects by pooled Ordinary Least Squares (OLS), while controlling for other factors influencing the deposit growth rate. In addition to the effect on the overall deposit growth rate, a distinction is made on the effects on the insured deposit growth rate and the uninsured deposit growth rate, to compare discipline between insured and uninsured depositors.
There exists ample literature on depositor discipline. The main approach of existing literature is to assess whether bank risk-‐taking has an influence of the price and/or quantity of deposits. Overall, the evidence of the existing literature suggests that bank risk-‐taking has an effect on effective depositor discipline (e.g. Martinez Peria and Schmukler, 2001, Park and Peristiani, 1998, Barajas and Steiner, 2001). Several papers study the effect of factors influencing depositor discipline, such as transparency, size, government ownership and deposit insurance. Overall, evidence regarding the relation of these aspects to depositor discipline is mixed (e.g. Wu and Bowe, 2012, Berger and Turk-‐Ariss, 2014, Barajas and Steiner, 2001, Oliveira et al., 2001, Demirgüç-‐Kunt and Huizinga, 1999, Martinez Peria and Schmukler, 2001).
This thesis contributes to the existing literature in several ways. First, it takes into account all the afore-‐mentioned factors: risk-‐taking, size, ownership, transparency and insurance. Second, this study specifically looks at the effects of the 2010 and 2011 EU stress tests on depositor discipline. The stress tests are modeled as variables concerning inclusion and outcome. As such, the effects of increased transparency via disclosure and the effect of outcome of the tests as an alternative measure for risk-‐taking can be estimated.
To summarize the findings: the evidence suggests that depositor discipline on the basis of bank fundamentals exists. Banks included in the stress test of 2010 experience a higher overall deposit growth, whereas the inclusion in 2011 has an insignificant effect. Results are different when a distinction is made between insured and uninsured deposits. Insured depositors do not seem to exert any discipline, neither in relation to changes in risk-‐profiles (except at stress-‐tested banks) nor to the stress tests. The uninsured deposit growth rate of a bank is responsive to joint bank fundamentals, implying these depositors react to risk-‐taking, but neither inclusion in the tests nor outcome of stress tests have a significant effect.
The remainder of this thesis is outlined as follows: sections two and three discuss the relevant literature on stress tests and depositor discipline. The methodology is explained in section four. A description of the variables and data are given in sections five and six. Section seven discusses the results of the event study and estimation by OLS, followed by robustness checks in section eight. The conclusion and policy implications are given in section nine. Lastly, section ten discusses the limitations of this study and recommendations for further research.
2. STRESS TESTS
This section focuses on stress tests. First, the need for stress test is explained. Furthermore, the general approaches and designs are described. Lastly, descriptions of the 2010 and 2011 EU stress test are given, as those are the tests this thesis focuses on.
2.1 The need for stress tests
Since the early 1990s financial institutions have conducted stress tests. In 2001, Peria et al. emphasized the importance of regulatory stress tests in order to assess the financial stability in relation to a changing macroeconomic environment. With the emergence of cross-‐country capital flows together with the innovation and diversification of financial products, the financial sector has become more opaque. This makes supervision and regulation difficult tasks to perform, as the exposure to risks cannot be determined. Performing stress tests on banks exposes this vulnerability of the financial system. By evaluating losses under different scenarios, transparency on risk exposure is increased.
The need for stress testing was underlined during the recent financial crisis, as regulators needed an instrument that could credibly evaluate the capital adequacy (Schuermann, 2014). A stress test is a simulation of the evolvement of a financial institution’s balance sheet under pre-‐specified conditions. The time span covered is usually two years. By evaluating the pro forma balance sheet at the end of the covered period, regulators can assess whether a bank has ‘’passed’’ or ‘’failed’’ the test, hence whether a bank has sufficient capital to survive even in dire periods. In essence, the solvency of a financial institution is being put to the test (Flannery, 2013).
2.2 Stress test design
Stress tests can be of a microprudential approach, a macroprudential approach or a combination of the two. The microprudential approach focuses on the health of financial institutions in isolation. The assets held by financial institutions pose risk. By evaluating these risky assets at different values relative to the capital held, the loss-‐bearing capacity can be established. The outcome of this ratio, i.e. the result of the stress test, provides a guideline for regulators whether to step in or not. The macroprudential approach is based on the role that financial institutions have in the real economy, as the connecting force between providers and borrowers of credit. Therefore, actions such as deleveraging can have negative effects on the real economy. The goal set out by the macroprudential approach is to reduce these adverse
effects. Unlike the microprudential approach, the macroprudential approach looks at the financial system as a whole. Since the macroprudential test is performed on the entire financial system, even well-‐capitalized financial institutions can be intervened on if on the aggregate level capital is insufficient. (Greenlaw et al., 2012). Besides establishing the capital shortfall, the approach should also provide a credible plan to increase capital if needed. This would increase the credibility of the results (Schuermann, 2014, Candelon and Sy, 2015).
In addition to the approach, a choice must be made on which scenarios and risks to include. Financial institutions are exposed to several risks, such as credit risk, liquidity risk and equity risk. Scenarios regarding these risks must be established. These scenarios can be based on historical events, or can be of a hypothetical nature. Historical data ensures that the scenario used in the stress test is plausible, but a drawback is that this approach may be irrelevant if an environment is evolving quickly. When the scenario is purely hypothetical, the possible circumstances included are of a more flexible nature. However, the probability that these circumstances are actually realized is complex to determine. (Peria et al., 2001).
In order to model the movement of balance sheet items, assumptions have to be made about the activities undertaken by the tested institutions (Hirtle and Lehnert, 2014). It is of crucial importance that the model of the stress test is specified accurately, in order to assess and evaluate the risks appropriately. If the model is misspecified, the outcome of a stress test can provide a false, lower risk exposure that can have dangerous consequences (Peria et al., 2001). In the use of consolidated data, for instance, lies the implicit assumption that capital and liquidity can freely move (internationally) within a financial institution. This is the opposite of ‘ring-‐fencing’, where there is no capital movement between subsidiaries and its parent. It has been shown in the past that ring-‐fencing does occur by, for example, blocking the sale of assets from more profitable subsidiaries. The use of unconsolidated data that takes ring-‐fencing into account leads to outcomes of EU stress tests that deviate from the official results (Cerutti and Schmieder, 2014).
Lastly, the regulators must decide how much, if any, to disclose about the results. Disclosing the stress test results to the public decreases opacity. One current wide-‐held view is that the opacity in the financial sector caused excessive risk taking. This in turn explains why the financial crisis was so severe. However, there are other benefits and costs associated with disclosing. On the one hand, more information is always better, as investors can then make more informed decisions. Hence, investors engage in market discipline (further explained in
MARKET DISCIPLINE & DEPOSITOR DISCIPLINE). If investors perceive a bank as being too risky, the
bank becomes unattractive and the price of its debt and equity claims decrease. Raising funds would then be much more difficult for the bank, which mitigates the harm the bank’s actions cause. This provides an incentive for the bank not to engage in risky behavior. On the other
hand, it provides an incentive to engage in suboptimal behavior by ‘dressing-‐up‘ the balance sheet in order to pass the test, which is stimulated by disclosing the methodology used for the stress tests. This suboptimal behavior can harm financial stability. Disclosure of results can also undermine risk sharing between banks. When the consequences of a shock are known before the actual shock has occurred, risk sharing and insurance are not possible. Goldstein and Leitner (2013) provide a model on risk sharing. In short, only when the overall economy is in a bad state, partial disclosure can help start risk sharing activities among banks. Furthermore, based on Morris and Shin (2002), if market participants base their actions largely on what others do, disclosing information can lead to over-‐reaction. Instead of relying on their private and valuable information, market participants base their actions solely on public information. This overreaction reduces the efficiency of market discipline (Goldstein and Sapra, 2013). Regulators therefore must take into account that disclosing bad results may lead to bank runs (Hirtle and Lehnert, 2014).
2.3 The European Union stress tests of 2010 and 2011
The Committee of European Banking Supervision (CEBS) carried out a stress test on banks in the European Union (EU), which started in March 2010, with a main focus on capital adequacy. Although stress tests were regularly conducted at a national level by national authorities, the CEBS considered the 2010 stress test to be useful as it was conducted at a European and bank-‐ specific level, with all institutions being subject to the same methodology and scenarios (CEBS, 2010a). Likewise, a stress test was conducted by the European Banking Authority (EBA), the successor of the CEBS, in 2011. In both stress tests the resilience of banks was tested using a baseline and adverse scenario(s), as measured by the Tier 1 ratio in 2010 and the Core Tier 1 (CT1) ratio in 2011, which are both ratios of capital over risk weighted assets. The difference stems from the capital included in the ratio; CT1 only includes high quality capital (CEBS 2010b, EBA, 2011a). The approach of the 2010 and 2011 EU stress tests was both micro-‐ and macro-‐ prudential; tests were performed at bank level, but these tests were performed to increase confidence and discipline in the market as a whole, via transparency (Acharya et al., 2014).
Per test, the sample of banks included was chosen on the basis of the amount of assets held. In descending order, banks combined per country must hold at least 50% of total assets in that country. In both tests, 91 banks were included, covering a large part of the total EU banking sector (in terms of assets). However, the results of the 2011 stress test were only disclosed for 90 banks. (EBA, 2011b, CEBS, 2010b).
All scenarios were based on forecasts by the European Commission and incorporated changes in macro-‐economic variables such as GDP, unemployment and inflation, as well as changes in the financial market. The tests were conducted under the assumption of static
balance sheets; implying no growth of exposure to risks. However, specifically for the 2011 stress test, banks had the opportunity to raise additional capital in January through April of 2011. In total, €50bn of capital was raised by for instance issuing common equity, government support or restructurings (CEBS, 2010b, EBA, 2011a).
The 2010 test included an extra adverse scenario to stress banks for a sovereign shock, which was added in the light of the sovereign debt crisis in 2010 (Schuermann, 2014). The 2011 stress test also incorporated sovereign risk, in addition to credit and market risk. The risk was modeled by applying haircuts to sovereign bond holdings in the trading book. Given that the possibility of default was not explicitly modeled and most of the sovereign exposure was in the banking book, which was not stressed for this specific risk, the assessment of sovereign risk was found to be too lenient (Ong and Pazarbasioglu, 2014).
The disclosed results of the 2010 test contained exposure to risky assets in the banking book, trading book and sovereigns per bank, in contrast to the 2009 stress test where only aggregate results were disclosed (CEBS, 2010c). Information disclosed in the 2011 results was much more elaborate than in the 2010 results. Schuermann (2014) contributes this to the stress tests conducted in 2010 by independent parties on Irish banks. Whereas these banks ‘passed’ the test conducted by the CEBS, the results of the second test called for an additional raise of €24bn of capital. The test results disclosed by the independent parties were far more elaborate than the results disclosed by the CEBS and helped tighten the sovereign credit spread (Schuermann, 2014). Following this example, the bank-‐specific results of the stress test of 2011 were more detailed. The EBA acknowledges this increase in detailed reporting: ‘’The 2011 EU wide stress test contains an unprecedented level of transparency on banks’ exposures and capital composition to allow investors, analysts and other market participants to develop an informed view on the resilience of the EU banking sector’’ (EBA, 2011a, p3).
For both tests, the aggregate (Core) Tier 1 ratio decreased under the adverse scenario, but was still well above the benchmark (see APPENDIX A, figure A1 and A2). The decrease is
mostly due to an increase in risk-‐weighted assets. By taking into account the restructuring and government support, the decline of the CT1 ratio under the 2011 stress test was tempered. In the 2010 test, seven banks ‘failed’, i.e. had a Tier 1 ratio below 6% in the adverse scenario (including sovereign shock), and had to raise additional capital (CEBS, 2010a). Eight banks did not pass the 2011 stress tests, as their CT1 ratio (including the additional capital raising) fell below 5% (EBA, 2011a). For both tests, a contingency plan for failing banks did not exist. Solely a recommendation was given to national supervisors to act upon the results (Candelon and Sy, 2015).
The approach was subject to criticism. As mentioned before, sovereign risk was deemed not to be captured appropriately. Markets felt that the probability of sovereign debt
restructuring or default was much larger than modeled and would therefore impose more losses (Candelon and Sy, 2015). Overall, the scenarios were (in hindsight) mild (IMF, 2013): banks that passed the test requested recapitalization or failed not long after disclosure (Candelon and Sy, 2015). Combining this with the lack of back-‐up plans if banks did fail the tests, credibility was impaired. Overall, the general consensus is that the goal of restoring market confidence was not met by either of the two tests (Ong and Pazarbasioglu, 2014, IMF, 2013). This is especially visible in the reaction of the stock market to the 2011 results: all EU banks experienced a significant drop in stock prices after disclosure. However, the stock returns for banks that were tested in 2010 did increase upon disclosure, indicating the 2010 stress test being somewhat of a success (Candelon and Sy, 2015).
3. MARKET DISCIPLINE & DEPOSITOR DISCIPLINE
Both market discipline and depositor discipline are described in this section. The emphasis is on depositor discipline, which is explained using a theoretical model. Furthermore, empirical work regarding depositor discipline and factors affecting it are discussed.
3.1 Market Discipline
Financial institutions deal with many parties such as depositors, stockholders and creditors. These parties can all exert discipline, in which they penalize a financial institution if costs increase due to excessive risk taking on the financial institution’s side. This concept is known as market discipline (Martinez Peria and Schmukler, 2001). Market discipline consists of two elements: market monitoring and market influence. Market monitoring entails that the changes in risk profiles of a financial institution are directly observed and taken into account, for example by a change in the institution’s stock price. This in turn signals the supervisory organ about the institution’s condition. Market influence prevails when the behavior of counterparties influences the behavior of the financial institution (Flannery, 2001). In order for market discipline to be effective, information must be publicly available. Involved parties can then assess how they are affected by the behavior of financial institutions. However, actions are valued differently because of different interests of claimants (Bliss and Flannery, 2002).
Given its potential to act as a additional tool to “common” regulatory instruments and its benefits in terms of signaling the supervisory organs, market discipline has not gone unnoticed by the regulators and supervisors. The third pillar of the Basel II accord is targeted at market discipline. The Committee aims at increasing effective market discipline, as it incentivizes financial institutions to ‘’conduct their business in a safe, sound and efficient manner’’ (Basel Committee on Banking Supervision, 2001, p5). As such, it complements the monitoring and disciplinary actions of supervisory organs (Basel Committee on Banking Supervision, 2001). By setting rules on the information that financial institutions must disclose, transparency is increased. An increase in information available to investors enhances market discipline and provides useful signals to supervisors. Institutions must disclose both qualitative and quantitative information on their position regarding corporate structure, capital structure and risk (Lopez, 2003).
Concerning the European Union, the European Banking Authority also recognizes the importance of market discipline. The Basel framework is imposed on EU banks by the EBA, and as such the above-‐explained rationale for increased disclosure is also applicable for the EU. The EBA has undertaken several actions with the aim of enhancing market discipline in the EU. An
example is that the EBA has provided guidelines on the disclosure of information, with regard to substance and frequency (EBA, 2014a). Additionally, in 2013 the EBA conducted a transparency exercise, in which information on capital composition and risk exposure was disclosed (EBA, 2015). Furthermore, the stress tests conducted by the EBA and the subsequent disclosure of the results also were intended to facilitate market discipline. The EBA aims at increasing transparency that induces market discipline, by subjecting all financial intermediaries included to the same test and providing results that are comparable among the sample (EBA, 2014b).
3.2 Depositor Discipline: a model
Depositors of financial intermediaries can exert discipline by the following: (i) altering the quantity of deposits (i.e. withdrawal), or (ii) requiring another price on deposits (i.e. interest rate change) (Martinez Peria and Schmukler, 2001).
The supply and demand can be modeled using the quantity of deposits and its interest rate as the price. The model, established by Klein (1971) and extended by Billett et al. (1998) and Jordan (2000), assumes a financial intermediary wants to maximize profits. Its core business is lending and its ability to extend loans is finite. This finiteness results in a downward sloping marginal revenue curve (in figure 1: curve mr1). The marginal cost curve (in figure 1: curve mci) is upward sloping, as it represents insured deposits. The market for insured deposits is finite. Attracting more insured funding would require higher search costs for the financial intermediary, hence the upward slope. Uninsured deposits are an additional source of funding. The marginal cost curve for uninsured deposits (in figure 1: curve mc!",!") is flat, implying that the financial intermediary can attract a limitless amount of uninsured deposits, since competition between banks is assumed. The marginal cost of uninsured deposits is dependent on the risk the financial intermediary takes, depicted by 𝜎 in figure 1. The profit maximizing bank will attract funding via deposits up to the point where marginal revenue equals the lowest marginal cost possible, which in this case means that it attracts D1 depositors, consisting of
Source: Jordan (2000), p19
Figure 1: Deposit model
source: Jordan (2000), pp 19
insured depositors I1 and uninsured depositors (D1-‐I1). When bank risk-‐taking, i.e. sigma, increases, uninsured depositors will demand a higher interest rate on their deposits, shifting the horizontal marginal cost curve upwards. This decreases the overall level of deposits. Hence, the model shows that an increase in bank risk-‐taking will lead to both a withdrawal of (uninsured) deposits and an increase in the price of deposits, i.e. the interest rate. There is ample empirical literature related to depositor discipline and the above-‐described model. A selective literature review is given below.
3.3 Depositor Discipline: selective literature review
This section will discuss relevant literature on depositor discipline. First, a small overview is given of the more general depositor discipline literature, comparing different studies. Thereafter, empirical work on the effect of transparency, Too-‐Big-‐Too-‐Fail/Too-‐Protected-‐To-‐ Fail, and deposit insurance is discussed, as these factors influence the effectiveness deposit discipline in theory.
The main approach of existing literature is to assess whether bank risk-‐taking has an influence on the price and/or quantity of deposits. Park (1995) looks at the relationship between the probability of failure for US banks and the movement of deposits. The probability of bank failure is estimated using bank fundamentals, which are available from bank balance sheets. Park’s main finding is that the probability of failure has a significant negative effect on the growth rate of deposits, and a positive effect on the interest rate on deposits. Furthermore, evidence suggests that rather than demanding a higher interest rate, depositors withdraw their (uninsured) funding. Park and Peristiani (1998) apply the same methodology as Park for a sample of US banks and their thrift deposits. Overall, the same conclusion is found as in the study of Park; market discipline is present in the form of higher interest rates and lower deposit growth rates for riskier banks. Jordan (2000) investigates depositor discipline prior to bank failure in the US and finds that in the eight quarters preceding bank failure, the level of total deposits decreases by 11.10%. As for the interest rate on insured deposits, evidence suggests that the failing banks were offering higher interest rates on deposits than their competitors. Wu and Bowe (2012) examine the existence of depositor discipline in China. The authors find evidence for the general notion that depositors react to bank risk-‐taking. Martinez Peria and Schmukler (2001) look at depositor discipline in relation bank risk-‐taking for Argentina, Chile and Mexico individually. The CAMEL ratios (which are extensively explained in the section
VARIABLES DESCRIPTION) are used to measure risk-‐taking. In general, the authors find that for all
countries an increase in bank risk-‐taking leads to a negative growth of deposits and an increase in the interest rate. Barajas and Steiner (2001) find similar results for Colombia.
3.3.1 Transparency
As noted before, an increase in information available to investors enhances market discipline. Enhanced information provision to depositors can have a positive effect on depositor discipline, as it allows depositors to make more well-‐informed decisions (Wu and Bowe, 2012, Nier and Baumann, 2006). There is positive evidence for this theory throughout the literature.
Wu and Bowe (2012) construct a disclosure index, where banks are awarded points for specific data being disclosed. This index-‐variable is significant and positive in their study, implying that an increase in information has a positive effect on the deposit growth rate. This relationship is attributed to increased depositor confidence when more information is disclosed. Hamid (n.d.) find the same results for countries in East Asia. While controlling for the potential endogeneity between disclosure and deposit growth rate, he finds that depositors not only react to the risks a bank takes, but also to the amount of information that is disclosed about those risks. Spiegel and Yamori (2004) model transparency differently. The authors make a distinction on whether Japanese banks use marking-‐to-‐market or book values concerning their assets. The authors conclude that deposit growth of banks that report market values is more sensitive to bank fundamentals than deposit growth of banks that use book values, suggesting that transparency increases market discipline. Both Nier and Baumann (2006) and Demirgüç-‐ Kunt
et al. (2008) examine the effect of transparency on bank risk-‐taking, albeit indirectly. Nier and Baumann use a cross-‐country sample and model market discipline as a function of depositor insurance systems, the ratio of uninsured deposits relative to total deposits and disclosure. The authors conclude that an increase in disclosure is related to higher capital buffers and lower realized asset risk, via market discipline. Demirgüç-‐Kunt
et al. investigate if compliance with the Core Principles for Effective Bank Supervision (BCPs) is related to the financial soundness of banks. A high level of transparency is one of the principles. The authors find robust evidence that Moody’s ratings are positively related to accurate information provision towards both the regulator and the market. Berger and Turk-‐Ariss’ (2014) evidence on the effect of transparency on depositor discipline is somewhat mixed. The authors find that depositors at listed, small banks discipline bank risk-‐taking more than depositors at unlisted, small banks do. Like other literature, it provides evidence of transparency enhancing market discipline. However, in the same paper results show that depositors at listed, large banks exert less discipline than depositor at unlisted, large banks.
3.3.2 Too Big To Fail and Too Protected To Fail
When a financial institution is substantially large and is very interconnected with other large banks, its failure may pose systemic risk. Governments are therefore often reluctant to let such a large financial institution default. By supplying a struggling institution with funds to
counteract the impending failure, banks are being treated as “Too Big To Fail” (TBTF) (Wheelock, 2012). Financial institutions will be kept afloat, thereby reducing bank panic and potential bank runs by depositors (Berger and Turk Ariss, 2014). However, this support system may induce a moral hazard problem; if financial institutions know that whatever happens, they will not fail, it encourages excessive risk-‐taking (Wheelock, 2012). If there is reluctance to let a state-‐owned banks fail, the bank is considered “Too Protected To Fail” (TPTF) and the same moral hazard problem can arise. In theory the TBTF-‐ and the TPTF-‐label may impair monitoring and disciplining by depositors (Oliveira et al., 2011). Empirical evidence, however, is mixed.
Park’s (1995) results are inconclusive. The TBTF hypothesis suggests that large banks would attract more depositors and can offer a lower interest rate, resulting in a positive and negative coefficient, respectively. The coefficient on size, however, is negative for both dependent variables. Therefore, the effect of bank size on depositor behavior is unclear. Different results are found by Berger and Turk-‐Ariss (2014), who compare depositor discipline across differently sized banks in the EU and US. Depositor discipline is modeled by the changes in the growth rate of deposits and the interest rate on deposits. Empirical results show that depositors at small banks exert more discipline than depositor at large banks. The authors therefore find evidence of the TBTF-‐hypothesis. As for a difference in discipline between the EU and US, the finding is that there was a higher level of depositor discipline in US. The authors suggest that this difference stems from the perception of a high bail-‐out probability for EU banks. Barajas and Steiner (2001) find that size significantly matters for depositor discipline in only a few of their specifications. The effect is positive, suggesting a TBTF belief. Furthermore, Oliveira et al. (2011) examine if the TBTF hypothesis holds for Brazilian banks during the crisis of 2008. Several estimations are performed to find the effect of the TBTF perception on the deposit growth rate, while controlling for other reasons why large banks may be attractive for depositors, such as diversification or superior risk management techniques. They find that specifically during a time of crisis, there is a large transfer of deposits from small to large banks. Hence, the results show evidence of depositors having the perception that large banks are TBTF. Additionally, the authors look at the effect of government ownership on depositor discipline. Government ownership does have a weak significant positive effect on the deposit growth rate for Brazilian banks in crisis times, meaning that depositors perceive banks as TPTF. In normal times, however, the effect is insignificant. Moreover, Wu and Bowe (2012) find that the deposit growth rate is positively related to state-‐ownership of banks. The authors explain this by stating that state-‐ownership functions as an ‘implicit guarantee’. This causes depositors to ignore risk taking, as they regard their deposits as safe. Similar results are found by Barajas and Steiner (2001) regarding state-‐ownership of Colombian banks.
3.3.3 Deposit insurance
Similar to TBTF and TPTF, deposit insurance schemes are used to protect the economy from systemic risk. The argument is that, in the case of (looming) bank failure, insured deposits will not be withdrawn, as they are perceived as safe. Hence, bank runs are avoided. However, deposit insurance schemes have the same downside as TBTF and TPTF; (credible) insurance, whether implicitly or explicitly, can have a weakening effect on depositor discipline. Additionally, there is a moral hazard problem as insurance can lead to excessive risk-‐taking by the bank (Martinez Peria and Schmukler, 2001, Demirgüç-‐Kunt
and Huizinga, 1999).
Demirgüç-‐Kunt
and Huizinga (1999) use a cross-‐country sample to assess the effect of deposit insurance on market discipline and find evidence that both the deposit growth rate and the deposit interest rate are sensitive to bank fundamentals, a sign of market discipline, despite the existence of depositor guarantees. However, the more explicit deposit insurance, the less the growth-‐ and interest rate respond to changing risk profiles. Similar results are found by Jordan (2000) who distinguishes between insured and uninsured CDs. Jordan finds that the decrease in level of deposits prior to bank failure is mainly due to a substantial decrease of the uninsured deposits. In terms of the interest rate on deposits, results show that banks with the largest uninsured depositor base have had the highest increase in interest rate. The overall conclusion of Jordan is that uninsured depositors react the most to the looming bank failure. In line with the previous mentioned studies, Maechler and McDill (2006) find that for US banks, bank fundamentals have a significant effect on the fraction of uninsured savings deposits to total deposits. They conclude that there is market discipline exerted by uninsured depositors. A high interest rate has a negative effect on uninsured deposits, but a positive effect on insured deposits; suggesting that deposit insurance weakens discipline. The conclusion of Martinez Peria and Schmukler (2001) is somewhat different. Their previously described study is extended by making a distinction between insured and uninsured deposits. They find that there is no significant difference in the discipline exerted by insured and uninsured depositors; both parties discipline risky banks. Even under implicit insurance for all depositors, discipline is exerted. An explanation given by the authors is that the insured depositors still face costs when banks fail, due to for instance the time it takes to retrieve the deposits or an underfunded insurance system. Ghosh and Das (2006) find that in the Indian banking sector insurance weakens the depositor discipline exerted in terms of deposit growth, but not in terms of interest rate. The interest rate on deposits is still sensitive to changes in bank fundamentals, suggesting a form of discipline exists.
4. METHODOLOGY
In this section, the empirical approach for testing the relation between stress test disclosure and depositor discipline is explained in detail. Using panel data, two methods are utilized to answer the research question: an event study and a regression-‐model respectively.
4.1 Event study
Following Black and Hazelwood (2012), who study the effect of participation in Troubled Asset Relief Program on bank risk-‐taking, an event study will be conducted. This will provide a graphical representation of the in-‐ and/or outflow of deposits around the time that the results of stress tests are disclosed by the CEBS/EBA and allows a comparison in depositor behavior on multiple levels. In order to somewhat answer the research question stated in this thesis, the following distinctions are analyzed graphically:
1. Overall depositor behavior in relation to stress tests results. Tested and non-‐tested banks are compared for both stress tests.
2. Specific depositor behavior in relation to stress test results. The flow of retail and corporate deposits are compared for both stress tests.
3. Depositor behavior in relation to specific stress test results. Financial institutions with different outcomes are compared for both stress tests.
The event study will analyze the behavior of depositors by interpreting the seasonally adjusted ratio of deposits to assets (hereafter: the deposit ratio). X-‐12 Arima is used to seasonally adjust the data. The ratio will be normalized to zero at the ‘event date’, that is at the time results are disclosed. The time frame for the event study includes three quarters before and after the quarter in which disclosing happened. This specific time frame refrains from having two disclosure moments in one event study. In the preceding and following quarters relative to the quarter where results are disclosed, no disclosures in relation to other stress test results prevailed. This is done in order to limit the bias. The following timeline represents the set-‐up of the event study:
q-‐3 q-‐2 q-‐1 q (=0) q+1 q+2 q+3 Where: q = quarter