ANK BAILOUTS : H OW DO DEPOSITORS REACT ? B

(1)

B

ANK BAILOUTS

:

H

OW DO DEPOSITORS REACT

?

By

Elmer Peter Alfred Pals* June 2016

Supervisor: Dr. M.A. Lamers

ABSTRACT

This thesis addresses depositor discipline in the face of one of the largest capital injection programmes for banks ever performed: TARP. Banks holding TARP funds experience lower deposit inflow than non-TARP banks, which can be explained by a negative stigma. A wake-up call for depositors could be a the root of increased depositor discipline on real estate exposure for banks holding TARP funds. The additional depositor discipline does not disappear when funds are repaid. When banks repay TARP, extra depositor discipline is exerted on capitalisation: highly capitalised repaying banks experience relatively higher deposit inflow.

Keywords: Banking ∙ Bank bailout ∙ Depositor discipline ∙ TARP JEL classification: G01 ∙ G21 ∙ G28

(2)

1. INTRODUCTION

Is a bank bailout a wake-up call for depositors, or do they exhibit moral hazard and reduce their discipline on the involved bank? Both theories have their merits and in this thesis I attempt to find a satisfying answer to this question.

Depositor discipline, a component of market discipline, should punish banks that take on too much risk and reward safe banks. If deposits at bailed out banks are deemed safe due to an implicit government guarantee, discipline will decrease and banks may be tempted into excessive risk taking. On the other hand, a bank bailout may serve as a wake-up call for depositors that the involved bank is not necessarily safe. The resulting deposit outflow can be detrimental to the bailed out bank that might need to turn to the government for additional support.

Evidence of the existence of depositor discipline is given in many papers (see, e.g. Goldberg and Hudgins, 1996; Baer and Brewer, 1986; Cook and Spellman, 1994; Park and Peristiani, 1998; Berger and Turk-Ariss, 2014). The record on whether moral hazard or the wake-up call effect is the dominant force for depositors is less clear. Previous research has found evidence of both the wake-up call effect and of moral hazard among depositors. Most literature regarding the wake-up call effect researches a market-wide effect of either a bailout programme or bank failures (see e.g. Martinez Peria and Schmukler, 2001; Karas, Pyle, and Schoors, 2013; Lamers, 2015). Hasan et al. (2013) find evidence of bank-specific wake-up call effects. The moral hazard argument has been made for bond spreads by Acharya, Anginer, and Warburton (2013). Berger and Turk-Ariss (2014) argue that government interventions in the financial sector have decreased depositor discipline.

(3)

The dataset used in this thesis is a collection of all U.S. commercial banks during the 2005Q1 until 2013Q4 period. The bank bailout researched in this thesis is a subprogram of the Troubled Asset Relieve Program (TARP), the Capital Purchase Program (CPP). Around 700 banks received bailout capital through the CPP programme during the financial crisis. Data on both the transfer of government capital to banks as well as the repayment of the money capital by banks to the government are used.

I find that significant depositor discipline existed during the period studied. Furthermore, I find that banks that hold bailout funds experience a deposit outflow, consistent with a negative stigma effect. This effect disappears when the bailout funds are repaid. Depositor discipline on holding real estate assets increases for banks that hold bailout funds and does not disappear when the funds are repaid. This is consistent with a signal given to depositors that the real estate assets held by banks that needed bailout funds were troubled. In addition to extra discipline on real estate, banks that repaid bailout funds also experience stronger depositor discipline on their capital ratio.

Section 2 presents the theoretical framework for this paper, section 3 elaborates on the institutional background. Section 4 discusses the used data. Section 5 develops the empirical methodology employed in this thesis. Section 6 outlines the results of this thesis, section 7 concludes.

2. THEORETICAL FRAMEWORK

This section starts with an introduction of depositor discipline. Next, the theoretical explanations for depositor reactions to a bailout are discussed. The section is concluded with the development of hypotheses based on the theory.

2.1. Depositor discipline

The third pillar of Basel III, disclosure, intends to enhance market discipline. Higher quality information for stockholders, bondholders, and depositors increases their ability to monitor banks on their risk profile. Depositors could move their deposits elsewhere when a bank does not offer adequate compensation to compensate for its riskiness. By offering incentives for banks to decrease their risk profile depositors can discipline their bank into more prudent behaviour.

(4)

significant depositor discipline on uninsured deposits over the 1986-1989 period. For a small sample of U.S. banks Baer and Brewer (1986) find the existence of depositor discipline in uninsured deposits and stress its importance for maintaining a stable financial sector. Cook and Spellman (1994) research a sample of banks in California, Arizona, and Nevada from January 1987 until August 1988 and find signs of depositor discipline even in insured deposits. Park and Peristiani (1998) study depositor discipline in U.S. thrifts over the period from 1987 to 1991. They find signs of depositor discipline both for uninsured as well as for insured depositors. In a recent study Berger and Turk-Ariss (2014) research depositor discipline during the financial crisis. The sample used consists of a little over 2000 banks from both the U.S. as well as the EU. They conclude that depositor discipline decreased because of government actions taken, but that depositor discipline still exists for most banks.

2.2. Wake-up call

A wake-up call for depositors occurs when an event transpires that makes them reassess the safety of their deposits. Martinez Peria and Schmukler (2001) show that depositor discipline increased in Argentina, Chile and Mexico after the banking crises in the 1980s and 1990s. Karas, Pyle, and Schoors (2013) employ a dataset of Russian banks over the period from 1995 until 2007 and find evidence of a wake-up call for uninsured depositors after the 1998 and 2004 banking crises. Thus, depositors can become more alert after a shock in the financial system has shown them that holding deposits is not necessarily safe. The wake-up call effect could also be targeted at specific banks. Hasan et al. (2013) use data of commercial banks in 11 Central European countries and their (mostly foreign) parent banks. The authors find that government aid received by parent companies negatively influences deposit growth in their Central European subsidiaries. This can be attributed to the negative stigma depositors attach to a bank. Hoshi and Kashyap (2010) argue that the fear of a negative stigma prevented Japanese banks from applying for the amount of bailout funding they needed during the Japanese financial crisis of the 1990s. Thus, a wake-up call can affect the whole financial sector, but can also affect single banks. It is this last property that is researched in this thesis.

2.3. Moral hazard

(5)

There is some evidence of bank risk-taking behaviour because of moral hazard. Duchin and Sosyura (2014) employ a dataset that includes banks’ applications and approvals by the authorities for the TARP programme. They find, for the 2006 until 2010 period, that banks approved for the TARP programme improve their balance sheet risk ratios, but give more weight to risky assets within the same asset class. This supports the findings of Black and Hazelwood (2013) who employ a dataset containing around 60 U.S. commercial banks and their branches over the November 2007 till August 2010 period. They find that small banks receiving TARP decreased risky lending compared to their peers, whereas large banks increased their risky lending compared with other large banks. For Germany, over the period from 1995 to 2006, Dam and Koetter (2012) show that banks which have an implicit government guarantee that they will be rescued in case of failure take more risk. Thus, bank risk-taking could increase if banks have a certain expectation of being bailed out in case of trouble.

Evidence of moral hazard behaviour of bank depositors is still limited. But, by examining the Greek banking sector from 2002 till 2010, Vasileiou (2014) indicates that TBTF banks have larger deposit inflows than non-TBTF banks. More conclusive results are found by Acharya, Anginer, and Warburton (2013) who discover that an (implicit) TBTF guarantee exists for the largest U.S. financial institutions during the period of 1990 until 2012. This guarantee allows these large banks to borrow at lower rates. If depositors see a bank bailout as evidence of an implicit government guarantee, depositor discipline should decrease at bailed out banks. Berger and Turk-Ariss (2014) describe that depositor discipline for large U.S. banks decreased during the 2008 financial crisis due to measures taken by the government. Moral hazard from depositors trusting that their bank will be bailed out in case of failure can thus reduce depositor discipline.

2.4. Hypotheses

(6)

HYPOTHESIS I: Banks experience lower deposit inflows over the period that they hold TARP funds compared to banks that do not hold TARP funds.

Conversely, when banks repay the bailout funds, this should send a positive message to depositors as the bank apparently does not need the bailout funds anymore. This should result in a larger deposit inflow than banks that still hold bailout funds and bring them up to equal footing with non-bailout banks in terms of deposit inflow.

HYPOTHESIS II: Repayment of TARP funding results in a larger inflow of deposits compared to banks that hold TARP-funds and similar deposit inflow compared to banks that have not received TARP funds.

Further, when depositors find that a bank holds bailout funds, they could exert stronger discipline on the bank. This wake-up call effect should result in a higher deposit inflow for banks that have healthy financials and a lower deposit inflow for banks that have weaker financials. This is stated in Hypothesis III.

HYPOTHESIS III: Banks that hold TARP funds face stronger depositor discipline than banks that do not hold TARP funds.

As the bailed out bank repays the bailout funds it should be able to reduce the extra depositor discipline. Hypothesis IV states that depositor discipline should reduce back to a level comparable to banks that have not received bailout funds.

HYPOTHESIS IV: Repayment of TARP funds decreases the depositor discipline exerted on these banks as compared to the situation prior to repayment.

3. INSTITUTIONAL BACKGROUND

(7)

3.1. The Troubled Asset Relief Program

To counter the problems in the financial sector, the U.S. government introduced TARP. Initially this programme was aimed at removing troubled assets from banks’ balance sheets to restore trust in the financial sector, yet this approach did not yield the quick results the U.S. government aimed for. Hence, a new program within TARP was introduced: the Capital Purchase Program (CPP).

Under the CPP program, the U.S. Department of the Treasury injected preferred equity into banks directly. To achieve a flying start to the programme nine of the largest banks of the U.S. strongly encouraged accept TARP funds. After these first nine banks, the programme expanded rapidly. In total, around 700 banks have received funding through the CPP programme. The CPP programme greatly resembled an earlier capital infusion by Warren Buffet’s Berkshire Hathaway in Goldman Sachs. In that investment Warren Buffet acquired five billion dollars in preferred shares in Goldman Sachs and received warrants to purchase common shares at a prearranged price. The preferred equity had coupons of 10% annually, and was callable by Goldman Sachs at any time at a 10% premium. To achieve fast implementation of the CPP programme, its terms were more attractive for banks than the Buffett deal. Veronesi and Zingales (2010) argue that a programme akin to the deal Buffett made with Goldman Sachs might not have been accepted by four of the first participants. The preferred equity had a 5% annual dividend, which would rise to 9% after five years, and was callable at any moment at face value. The rise in the dividend after five years shows the intent of the Treasury to make this programme transitory and encourage banks to raise private capital as soon as market circumstances permitted.

Banks participating in the TARP programme faced restrictions on executive compensation. Banks were generally viewed as the perpetrators of the financial crisis. This resulted in a public outcry against the high salaries, and bonuses in particular, still handed out in the financial sector. To accommodate this anger, financial compensation for high-ranking bank employees in banks that received TARP funds was curtailed. Limitations on executive compensation were regarded as problematic by banks, as they would have trouble attracting talented employees, but also offered an incentive for bank managers to pay back TARP funds as soon as possible.

(8)

opted out, they had strong capital ratios, stable funding profiles, high asset quality and operated in better performing regions. Not all banks that applied for the CPP programme were approved by the Treasury to receive funds. As non-approval by the regulators or Treasury could constitute a negative signal towards depositors and bondholders about the financial stability of the bank in question, applications (and thus the banks that did not get approval) were not made public. Approved banks were larger and had higher asset quality. This conclusion is largely shared by Cornett, Li, and Tehranian (2013), who find that TARP funds were mostly directed at banks with temporary problems such as liquidity shortage, but otherwise strong characteristics. Ng, Vasvari, and Wittenberg Moerman (2010) have similar findings for stock-listed banks, and show that TARP banks outperformed non-TARP banks. However, their research suggests that there is a negative signal sent to investors from participating in the programme. The initial underperformance of TARP stocks is later reversed when the market is less stressed and more informed about the fundamentals of the TARP participants.

(9)

early. Repayment might therefore be also just because the bank did not need the funds as badly as they did when the bank applied.

3.2. Deposit insurance

Deposit insurance can weaken the incentive for depositors to monitor banks. During the 2008 financial crisis, the amount to which deposits were insured went up from $100,000 to $250,000 in the United States. In the light of this increase, one might wonder whether insured depositors still impose discipline on banks. It turns out that even depositors who are insured participate in bank runs. As Martinez Peria and Schmukler (2001) mention, depositors care about the safety and liquidity of their deposits. An insurance deposit scheme can provide the safety aspect, given that the insuring party can credibly guarantee all deposits (Cook and Spellman, 1994). The insuring party in the United States, the FDIC, is backed by the U.S. government, which makes the guarantee credible. However, the liquidity of deposits may be in jeopardy during the period after default of a bank, a situation described by Kane (1989) which happened in Ohio in 1985. This second argument can explain why, even with deposits guaranteed by a credible authority, there were bank runs in the recent crisis. For example, the heavily market funded Northern Rock became the subject of a bank run. While Northern Rock was an exception in terms of its funding structure, as is pointed out by Shin (2009), the image of people queuing to retrieve their deposits shows that depositors are concerned about their money when a bank failure looms. The Federal Deposit Insurance Corporation (2010) states that a check is sent within a few days to insured depositors for them to access their deposits in the case that there is no party willing to take over the failed bank. This course of action keeps inconvenience to a minimum for depositors, but the uncertainty and hassle involved still gives an incentive for insured depositors to put their money in a safe bank. Depositor discipline might have decreased over time due to more comprehensive deposit insurance, but the incentive for depositors to safeguard their money is still present. 4. DATA

For the composition of the dataset, multiple sources are used. Bank-level balance sheet and income statement data are collected from Reports of Condition and Income (Call Reports). Data on the TARP CPP1 programme is gathered from transaction reports made public by the U.S. Department of the Treasury. These reports contain all CPP transactions made by the Treasury. Data on prices are collected from the Federal Reserve Bank of St.

(10)

Louis. The data used in this thesis are quarterly and runs from 2005Q1 to 2013Q4. By using this this time period, more than three years before and more than four years after the first TARP payment are included. This ensures that the effects of the TARP programme can be accurately measured. The next sections discuss the structure of the dataset and the used variables.

4.1. Structure of the data

Call report data contain data for multiple types of banks. Only data on commercial banks are used for the analysis, all other bank data are removed from the sample. This is standard procedure in banking literature as the set-up of other types of banks differs substantially from the set-up of commercial banks (Berger and Roman, 2015; Cornett, Li and Tehranian, 2013). For instance, Savings & Loans corporations are forced by law to invest a large portion of their assets in mortgage-related products, a requirement that does not hold for commercial banks. After removing entities that are not commercial banks, all financial data of banks that are part of a Bank Holding Company (BHC) are consolidated to the BHC level. Banks that are not part of a BHC are consolidated to the highest bank entity level. The entity created using the data will be referred to in this dissertation as a bank, while indeed it is often a consolidation of multiple commercial banks.

(11)

funds from their parent banks. Decisions within a BHC are made jointly and TARP funds can be redistributed within the BHC. The existence of within-BHC capital markets and the possibility of TARP funds redistribution make it pertinent to treat banks within a Bank Holding Company as a group rather than as individual banks.

To control for imperfect data and extreme observations all data is truncated at the 1st and 99th percentile. Moreover, Mergers and Acquisitions (M&A) have occurred frequently in the period studied. As these events disrupt the data and might influence the results it is important to control for M&A activity. Following Gatev and Strahan (2006), I delete all bank-quarter observations for which total assets grew by more than 10%. This transformation should suffice to filter out most effects of M&As in the sample period.

Following Duchin and Sosyura (2014), I exclude 17 banks which were subject to the Capital Assessment Plan (CAP). The banks excluded are: American Express, Bank of America (including Merrill Lynch), Bank of New York Mellon, BB&T, Capital One, Citigroup, Fifth Third Bancorp, Goldman Sachs, JP Morgan, KeyCorp, Morgan Stanley, PNC Financial Services, Regions Corp., State Street, SunTrust, U.S. Bancorp, and Wells Fargo (including Wachovia). These banks were part of further tests during the CAP, and got an implicit TBTF guarantee.

4.2. TARP variables

Of all TARP capital injections, most receivers were BHCs. As stated by Berger and Roman (2015), 572 receivers were BHCs against 87 commercial banks and 50 Savings and Loans or Thrift institutions. The CPP recipients specified in the U.S. Treasury's transaction report (of date) does not include RSSD numbers or FDIC certificate numbers. Therefore, the banks have to be matched manually with RSSD numbers using data from the National Information Center (NIC), the FDIC, and WRDS. Nine institutions remain unidentified in the data used; the other institutions were matched sing the name, city and state of the bank. Consolidated banks in the dataset are considered to have received TARP when one of the entities within the consolidated bank is included in the CPP transaction sheet.

(12)

held by natural persons Cornett, Li, and Tehranian (2013). For the sake of simplicity, Treasury investments in preferred equity and subordinated debt are considered similar and no distinction is made in the data. As can be seen in Fig. 1, which shows the participation in the CPP programme, CPP investments using subordinated debt represent only a small minority of investments.

Fig. 1 Cumulative CPP participation

This figure shows TARP-participation over time. The dark shaded area represents preferred equity investments, whereas the light shaded area shows subordinated debt investments.

The U.S. Treasury has reduced its exposure to banks after 2008. Its shares or debentures were either bought back by the institution, or the Treasury has sold its stake to outside parties. For the transformation of data, only the full repayment of the Government’s stake by the bank is considered as full repayment. Some banks have repurchased the Government’s shares or debentures in multiple tranches. In this case, the last payment made is considered as the moment on which the bank is off TARP.

(13)

To clarify the exact definition of the CPP Received and CPP Repaid variables, I take the case of an actual bank: Haviland Bancshares, Inc. This bank from Haviland, Kansas, has received a CPP investment of US$ 425,000 on 13 March 2009. The invested amount plus dividends and proceeds from warrants was paid back on 29 December 2010. Fig. 2 is a graphical representation of the timeline for Haviland Bancshares, Inc. The figure shows both the time of entry in the CPP programme, as well as the moment of exit. In this case, the TARP funds were returned to the treasury, which does not hold for all banks. In the case that a bank does not pay back the TARP funds, the CPP Received variable will keep the value of 1 until the end of the period studied. Noteworthy is that there is no possibility of the CPP Repaid variable to take the value of one when the CPP Received variable has not had a value of 1.

Fig. 2 TARP variable timeline for Haviland Bancshares, Inc.

This figure shows the timeline for Haviland Bancshares, Inc. It shows when CPP Received and CPP Repaid take values of either 1 or 0.

4.3. Further variables

(14)

Regulators nowadays assess the stability of a banking institution based on the CAMEL ratings system. This system assesses the Capital Adequacy, Assets Quality, Management Capability, Earnings, and Liquidity. This thesis follows their assessment of the stability of banking institutions and uses variables, which are discussed in the data section, to proxy the components of the CAMEL rating system.

The CAMEL rating system is proxied in this dissertation using six variables2: Capital Adequacy, for Asset Quality Total Loans over Total Assets, Real Estate Loans over Total Loans, and Nonperforming Loans (NPL) over Total Assets are used. For Earnings, Return on Assets is used and a measure for Liquidity is included. Capital Adequacy refers to Equity over Total Assets, as equity capital is junior to deposits it serves as a buffer to absorb losses. Therefore, I conjecture that a higher value of the Capital Adequacy variable results in a higher deposit inflow. Total Loans over Total Assets captures part of the business model of a bank, the fraction of assets that a bank has invested in loans, which are relatively illiquid. From the view of liquidity, bank riskiness should increase with a higher value of the Total Loans over Total Assets variable. Depositors might on the other hand assess banks with a more “traditional” business model positively. Real Estate Loans over Total Loans captures another aspect of the bank business model: exposure to real estate. As real estate took a great hit during the crisis, banks with a large fraction of these assets on their balance sheet should be losing deposits compared to banks with less exposure. NPL over Total Assets measures the fraction of bad quality loans in the portfolio of a bank. This fraction can be influenced both by bad luck, a bank might have extended loans in a region hit by adverse conditions, or by deliberate past risk taking behaviour. In both cases, a larger fraction of NPL over Total Assets can be a negative signal to depositors. Return on Assets is a very direct measure of the earnings of a bank, and the profitability of a bank’s assets. A higher value should indicate stronger performance of a bank and therefore attract deposits. Liquidity measures the amount of liquid assets a bank has at its disposal over the total assets of a bank. More liquid banks are less vulnerable to acute shocks making depositors withdraw funds. On the other hand, liquid assets yield a very low return and a very liquid bank could therefore run into issues with profitability.

2

(15)

4.4. Descriptive statistics

Table 1 presents the summary statistics for the sample used over the 2005Q1 to 2013Q4 period. The data presented are the transformed data. Notable is for instance the change in deposits, where the mean change is only 0.8% per quarter, but has a standard deviation of 4%. Liquidity is another figure that draws attention, with a mean value of 9.45%, but a large standard deviation of 7.3%. In this sample, 3.3% of bank-quarter observations are a bank that holds TARP funds and 1.24% of bank-quarter observations are of banks that have repaid TARP.

Table 1 Summary statistics

(1) (2) (3) (4) (5)

VARIABLES N mean sd min max

Change in Deposits (%) 190,741 0.818 4.043 -13.109 11.596

CPP Received 195,422 0.033 0.179 0 1

CPP Repaid 195,422 0.012 0.111 0 1

Capital Adequacy 191,514 0.107 0.032 0.042 0.312

Total Loans / Total Assets 191,514 0.636 0.150 0.104 0.911

Real Estate / Total Loans 190,155 0.687 0.174 0.099 0.974

NPL / Total Assets 193,468 0.012 0.015 0 0.100

Return on Assets 191,514 0.002 0.003 -0.018 0.009

Liquidity 191,514 0.095 0.073 0.011 0.445

Implicit Deposit Rate (%) 190,742 0.479 0.285 0.046 1.265

Ln(Total Assets) 195,422 11.893 1.304 4.087 20.248

This table presents the summary statistics of the variables used in this thesis. Change in Deposits denotes the percentage logarithmic quarterly change in deposits. CPP Received and CPP Repaid are indicator variables for banks that received CPP funds as shown in Fig. 2. Capital Adequacy is defined as Equity over Total Assets, Return on Assets is Net Income over Total Assets, and Liquidity is Liquid Assets over Total Assets. The Implicit Deposit Rate is interest expense over total deposits and is expressed as a percentage.

4.5. Difference between TARP banks and non-TARP banks

(16)

Table 2 Difference between Non-TARP and TARP banks

TARP banks Non-TARP banks Difference

VARIABLES p5 mean p95 p5 mean p95 mean

Change in Deposits (%) -6.285 1.320 8.488 -7.137 0.179 7.879 -1.140***

Capital Adequacy 0.070 0.095 0.129 0.071 0.108 0.176 0.013***

Total Loans / Total Assets 0.573 0.755 0.885 0.379 0.666 0.865 -0.089***

Real Estate / Total Loans 0.504 0.750 0.914 0.328 0.675 0.912 -0.075***

NPL / Total Assets 0.001 0.013 0.035 0.000 0.012 0.044 0.000

Return on Assets -0.004 0.001 0.003 -0.006 0.001 0.005 0.001***

Liquidity 0.015 0.045 0.104 0.018 0.077 0.210 0.032***

Implicit Deposit Rate (%) 0.454 0.745 1.032 0.348 0.657 0.997 -0.088***

Ln(Total Assets) 11.271 13.310 16.091 9.914 11.746 13.761 -1.564***

Comparison of means between Non-TARP (N=5,029) and TARP (N=463) banks. A TARP bank as defined as a bank that has received TARP funds at some point over the duration of the programme. Change in Deposits denotes the percentage logarithmic quarterly change in deposits. Capital Adequacy is defined as Equity over Total Assets, Return on Assets is Net Income over Total Assets, and Liquidity is Liquid Assets over Total Assets. The Implicit Deposit Rate is interest expense over total deposits and is expressed as a percentage. Data on banks are as of 1 July 2008. Banks subject to CAP are not included in this comparison. In the last column, the difference between the means of both groups is shown. A t-test is performed to compare the means of both groups. *** p<0.01, ** p<0.05, * p<0.1

Table 2 shows that TARP banks are on average larger than non-TARP banks. Furthermore, TARP banks experience higher deposit growth in the second quarter of 2008 than non-TARP banks, which could be explained by the higher rate they offer on their deposits. TARP banks have lower capital ratios, return on assets and liquidity. Total Loans over Total Assets is larger for TARP banks, as is Real Estate over Total Loans. There is no significant difference for Non-performing Loans over Total Assets between TARP and non-TARP banks.

5. METHODOLOGY

This section contains an elaboration on the empirical methods used to test the stated hypotheses. First, I show the estimation method used for testing whether or not there was evidence of depositor discipline during the period studied and whether there was a negative stigma to TARP that might decrease deposit inflow for participating banks. Further, I present the model for testing whether there was a wake-up call effect for depositors at banks that received TARP.

5.1. Estimation of effects TARP on deposit inflow

To test Hypotheses I and II, Eq.(2) is estimated.

∆ 𝐷𝑒𝑝𝑜𝑠𝑖𝑡𝑠𝑖,𝑡 = 𝜇𝑖 + 𝑑𝑡+ 𝛽1𝐶𝑃𝑃 𝑅𝑒𝑐𝑒𝑖𝑣𝑒𝑑𝑖,𝑡−1+ 𝛽2 𝐶𝑃𝑃 𝑅𝑒𝑝𝑎𝑖𝑑𝑖,𝑡−1+

𝛽₃𝑅𝑖𝑠𝑘_{𝑖,𝑡−1}+ 𝛽₄𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠_{𝑖,𝑡−1}+ 𝜀_𝑖,𝑡

(17)

ΔDepositsi,t represents the logarithmic change in deposits for bank i, at time t

compared to time t-1. CPP Receivedi,t—1 indicates whether bank i holds TARP funds at time

t-1. A negative value for β1 indicates a smaller deposit inflow for banks that hold TARP

funds compared to banks that do not hold TARP funds. This would indicate a negative stigma for banks holding TARP funds. CPP Repaidi,t—1 shows whether bank i has repaid TARP

funds at time t-1. When β2 is positive, banks that have repaid TARP funds experience a

higher inflow of deposits than banks that have not repaid TARP funds. Riski,t-1 is a vector of

CAMEL variables as discussed in the data section for bank i at t-1. Coefficient β3 indicates

the relation between the bank risk variables and the change in deposits for bank i, an indication of depositor discipline. The vector Controlsi,t-1 consists of the implicit deposit rate

and the natural logarithm of total assets for bank i at t-1. To investigate whether depositors reward banks that repay TARP funds, the difference between β1 and β2 is tested. When β2 is

larger than β1, this indicates that deposit inflow for banks that repaid TARP is larger than

deposit inflow for banks that hold TARP funds.

The process of depositors choosing which bank they put their money in and banks determining what interest rate to offer. This process implicates that there exists simultaneity between prices and quantities in a specification using interest rates as an explanatory variable for deposit inflows. To overcome this endogeneity problem, I use a framework in which depositors are influenced by both interest rates as well as fundamental characteristics of banks in the previous quarter. This framework is also employed by Martinez Peria and Schmukler (2001), and has the attractive property that the explanatory variables represent the most recent set of information available to depositors. As deposit inflow in the current quarter cannot reasonably influence interest rate setting behaviour of banks in the previous quarter. Thus, the used framework should negate the simultaneity problem.

The model employed for researching the effects of the TARP programme on deposit inflow resembles the model used by Berger and Turk-Ariss (2014). The country dimension present in their model is omitted in this research as all banks in this paper are from the United States. As in the Berger and Turk-Ariss model, I include time fixed effects (μi) to control for

changes in the overall macroeconomic environment. To control for possible heteroskedasticity, White robust standard errors are used in Eq.(2). All standard errors are clustered at the bank level. Moreover, I include bank fixed effects (dt) to control for

(18)

5.2. TARP and depositor discipline

The model in Eq.(2) can be used to detect whether there was depositor discipline over the studied period and to examine whether banks there was a negative stigma to TARP. The next econometric model is used to examine whether depositor discipline becomes more stringent for banks that hold or paid back TARP funds. The model presented in Eq.(3) is an extension of the model in Eq.(2). In this case, the CPP variables are interacted with the risk proxies. From the regression estimates, a conclusion can be drawn on whether TARP funding increases or decreases discipline at the concerned banks relative to non-TARP banks.

∆ 𝐷𝑒𝑝𝑜𝑠𝑖𝑡𝑠_𝑖,𝑡 = 𝜇_𝑖 + 𝑑_𝑡+ 𝛽₁𝐶𝑃𝑃 𝑅𝑒𝑐𝑒𝑖𝑣𝑒𝑑_{𝑖,𝑡−1}+ 𝛽₂𝐶𝑃𝑃 𝑅𝑒𝑝𝑎𝑖𝑑_{𝑖,𝑡−1}+ 𝛽₃𝑅𝑖𝑠𝑘_{𝑖,𝑡−1}+ 𝛽₄𝐶𝑃𝑃 𝑅𝑒𝑐𝑒𝑖𝑣𝑒𝑑_{𝑖,𝑡−1}∗ 𝑅𝑖𝑠𝑘_{𝑖,𝑡−1}+ 𝛽₅𝐶𝑃𝑃 𝑅𝑒𝑝𝑎𝑖𝑑_{𝑖,𝑡−1}∗ 𝑅𝑖𝑠𝑘_{𝑖,𝑡−1}+ 𝜀_𝑖,𝑡

( (3)

In this model, β4 captures the difference in depositor discipline between banks that

hold TARP funds and those that do not. Further, coefficient β5 show the difference in

depositor discipline between bank that have repaid TARP funds and banks that have not (yet) repaid or have not participated in the programme. To examine the difference in depositor discipline between banks that hold or have repaid TARP β4 and β5 are compared. The results

of these tests are shown in the next section. 6. RESULTS

This section shows the results of the empirical analysis. First, the presence of general depositor discipline is examined. Second, the effects of accepting or repaying TARP funds on deposit inflow are tested. Third, the existence of additional depositor discipline on banks that participated in TARP is shown.

6.1. Depositor discipline

(19)

Capitalisation, and Loans over Total Assets for Central European countries. The negative sign for liquidity, implying that more liquid banks experience a lower deposit inflow, is unexpected but similar to the result found in Lamers (2015). Overall, the results indicate that depositors discipline banks on their fundamentals and that banks showing more solid financials enjoy a higher deposit inflow than banks with weaker fundamentals.

6.2. Deposit outflow for TARP-banks

Column 2 in Panel A of Table 3 shows a regression that includes controls for bank size and the implicit deposit rate, but does not include CAMELs. Both CPP variables are significant at the 1% level. Banks holding TARP funds corresponds to a 0.69% lower quarterly deposit inflow than banks that do not hold TARP funds. On the other hand, banks that have repaid TARP funds enjoy a 0.33% higher quarterly inflow of deposits than banks that have either not repaid or not received TARP funds. As Panel B shows, banks that have repaid TARP funds have a higher deposit inflow than banks that still hold TARP funds.

(20)

Table 3 Effect of TARP and CAMELs on deposit inflow

Panel A: Regression results

(1) (2) (3)

VARIABLES Only CAMELs Only CPP Full regression

Capital Adequacy 26.725*** 26.912***

(1.243) (1.248)

Total Loans / Total Assets 8.419*** 8.390***

(0.267) (0.266)

Real Estate / Total Loans -2.474*** -2.473***

(0.298) (0.298) NPL / Total Assets -28.042*** -27.486*** (1.216) (1.225) Return on Assets 56.051*** 55.411*** (5.763) (5.761) Liquidity -11.758*** -11.786*** (0.366) (0.366) CPP Received -0.692*** -0.392*** (0.098) (0.093) CPP Repaid 0.328*** -0.134 (0.122) (0.119) Observations 163,626 174,180 163,626 R-squared 0.123 0.057 0.123 Number of Banks 6,794 6,942 6,794

Controls YES YES YES

Bank Fixed Effects YES YES YES

Time Fixed Effects YES YES YES

Panel B:test for equality of CPP indicators

β CPP Received - β CPP Repaid -1.020*** -0.258**

(71.78) (5.08)

(21)

6.3. The wake-up call effect of TARP

Table 4 shows the results of the empirical test of Eq. (3). The inclusion of the interaction terms of the CAMELs with CPP Received and CPP Repaid does not influence the coefficients for the CAMELs or control variables compared to the estimates shown in Table 3 column 3. Of the interaction effects with CPP Received, only Real Estate over Total Loans is statistically significant. Thus, depositor discipline for banks holding TARP funds compared to banks that do not hold TARP funds is only stronger for the ratio of Real Estate over Total Loans. The coefficient for this interaction effect is negative and significant at the 1% level. A two standard deviation (34.8%) difference in the Real Estate over Total Loans ratio corresponds to a 0.67% lower deposit inflow for banks holding TARP funds compared to banks that do not hold TARP funds. The difference between the coefficients of the interaction effects of CPP Received and CPP Repaid with Real Estate over Total Loans is tested in column 2 of Table 4. These two coefficients do not differ significantly, which suggests that increased depositor discipline on real estate assets continues after the TARP funds are repaid. The reputation damage incurred by accepting TARP funds seems to be long lasting and is not mended by the repayment of TARP.

Uncertainty about the quality of real estate assets during the 2008 financial crisis might lie at the root of the long-lasting increased depositor discipline on real estate assets of TARP-banks. Accepting TARP funds could signal future losses on the asset portfolio of the bank, analogous to the fear of Japanese banks described in Hoshi and Kashyap (2010). They describe a reluctance of Japanese banks to participate in a bank recapitalisation programme, as it might create doubts about the quality of the bank’s assets. This wake-up call effect seems to be present for depositors at banks that accept TARP.

(22)

Table 4 Effect of TARP and CAMELs on deposit inflow, including interaction terms

(1) (2)

VARIABLES x CPP Received x CPP Repaid Difference

CPP Received 0.398 (0.894) CPP Repaid -0.147 (1.039) Capital Adequacy 26.574*** 4.189 19.583*** 15.394** (1.270) (4.203) (6.081) (6.04)

Total Loans / Total Assets 8.391*** 0.226 -0.594 -0.82

(0.269) (0.784) (1.031) (0.61)

Real Estate / Total Loans -2.406*** -1.926*** -2.081*** -0.155

(0.299) (0.696) (0.773) (0.03) NPL / Total Assets -27.620*** 2.259 13.211 10.952 (1.268) (4.154) (11.541) (0.87) Return on Assets 54.543*** 13.065 -60.889 -73.954 (5.986) (20.630) (61.696) (1.40) Liquidity -11.782*** 0.876 -2.755 -3.631 (0.369) (1.873) (1.931) (2.56) Observations 163,626 Number of Banks 6,794 R-squared 0.123 Controls YES

Bank Fixed Effects YES

Time Fixed Effects YES

(23)

The remaining interaction effects of CAMELs with CPP Received and CPP Repaid do not show signs of increased depositor discipline when holding or repaying TARP funds. Apart from Capital Adequacy, there is no statistically significant difference between the coefficients for the CAMELs when interacted with CPP Received or CPP Repaid. Regarding Hypothesis III, holding TARP funds does increase depositor discipline, but only for real estate assets. This can be attributed to uncertainty about the quality of these assets, where accepting TARP signals problems in the real estate asset portfolio. With respect to Hypothesis IV, the repayment of TARP does not decrease depositor discipline. Depositor discipline on real estate holdings of a bank remains equal to the effect prior to repayment and there exists additional discipline on capitalisation of the banks compared to banks that did not receive or repay TARP funds.

6.4. Robustness tests

To verify that the results are valid, this section describes the robustness tests performed. As a test whether the results are affected by the capital injections in the form of subordinated debt for S-corps, I change the definition of the CPP Received and CPP Repaid variables to only include Treasury purchases of preferred equity. These variables are zero at all times for the banks that have participated in the programme, but only sold subordinated debt to the Treasury. The result of the full regression as in Eq.(2) using the newly defined TARP variables can be found in column 1 of Appendix C. All signs and significance levels are equal to the results shown in Table 3.

Announcement of approval for the CPP programme was followed quickly by the disbursement of funds by treasure to participating banks. There might however be some anticipation present from depositors who have indications that banks will receive TARP funds over the next period. To test for the possibility that the anticipated participation in the TARP programme biases the results obtained I include the CPP variables without the one period lag used in Table 3. The results of this regression are presented in Appendix C, column 2. While the sign and significance level of coefficients for the CAMELs and the CPP variables are similar, the difference in the CPP variables is not statistically significant in this case.

(24)

first repayment changes the results. The results of this regression are shown in Appendix C, column 3. The results show little change for the estimation coefficients.

The same robustness tests are also performed for the specification including interaction effects. These robustness tests are located in Appendices D-F. The robustness test for which only preferred equity investments are considered CPP is included in Appendix D. Coefficients show equal signs, but in this case the interaction effect of CPP Repaid with Liquidity and the difference between the interaction effects of liquidity with CPP Received and CPP Repaid are statistically significant, but only at the 10% level. Appendix E considers the anticipation of TARP. In this table, the interaction effect of Real Estate over Total Loans is smaller and only significant at the 5% level. In Appendix F, the first repayment of TARP is taken as full repayment. In this case, the interaction term between Real Estate over Total Loans and CPP Repaid is a little smaller and only significant at the 5% level as compared to significance at the 1% level in the baseline specification.

6.5. Limitations

As is mentioned in the theoretical framework, asset-side risk taking behaviour by banks can increase due to an implicit or explicit bailout guarantee by the government. Duchin and Sosyura (2014) and Black and Hazelwood (2013) show that TARP-banks increase within asset class risk-taking. Depositors may be aware of or anticipate moral hazard behaviour of banks receiving TARP. Since I do not have access to loan-level data for bank assets, I cannot control for this problem.

The lack of loan-level data, or at least indicators of asset quality, hinders the

interpretation of the negative coefficients for the interaction effects between Real Estate over Total Loans. There could be larger problems in the real estate portfolio of banks that accepted TARP compared to banks that did not participate. Without the data, it is impossible to state whether the negative coefficient for the interaction effects of Real Estate over Total Loans is due to information depositors have on the quality of a bank’s assets or due to a negative signal because of accepting TARP funds.

(25)

A general wake-up effect for depositors due to the financial crisis might bias the results found in this thesis. By interacting the CPP indicator values with bank fundamentals, I compare depositor discipline with both non-TARP banks, but also with the whole sample of banks prior to the financial crisis. If there is a general increase of depositor discipline during the period that the CPP indicator variables take positive values, a bias towards the existence of a bank-specific wake-up call could exist.

7. CONCLUSION

In this section, I conclude this thesis, list the most important outcomes, give policy recommendations, and outline the scope for further research associated with this thesis.

Even in an era with deposit insurance and bank bailouts, depositor discipline exists. Banks with stronger fundamentals experienced on average larger deposit inflows. Insured depositors could prefer stronger banks for liquidity or uncertainty reasons. Using data on U.S. commercial banks in the 2005Q1 until 2013Q4 period, I find indications that banks receiving TARP experience lower deposit growth. This effect can be traced back to the signal banks give when accepting TARP funds. When a bank accepts TARP funds, it signals that the quality of its assets might be worse than they look at first sight. This is particularly true for the real estate asset portfolio of banks during and after the financial crisis. As the value of the real estate loans portfolio was uncertain in this period, the negative signal that accepting TARP gives, results in a deposit outflow.

The TARP capital injections did not cause a decrease in depositor discipline for the participating banks. There might be an overall decrease in depositor discipline due to the bailout package, but that is beyond the scope of this thesis. A bank choosing to participate in the TARP programme experiences a lower deposit inflow that is likely to be the result of uncertainty about the quality of its real estate assets. Resolving this uncertainty was ironically the initial purpose of the TARP programme. An earlier restructuring of bank real estate portfolios, forcing (participating) banks to realise their losses might decrease the uncertainty.

(26)

(27)

APPENDIX A.LIST OF VARIABLES

Variable Description Source

Dependent Variable:

Δ Deposits Natural logarithm of change in Total Deposits (ln(rcfd2200t /

rcfd2200t-1))*100

WRDS Call Reports

TARP Variables:

CPP Received 1 from the moment that CPP funds have been received by the bank. 0 After the funds

have been fully paid back. Always 0 if the bank has not participated in the CPP programme.

U.S. Treasury

CPP Repaid 1 from the moment onwards that CPP funds have been fully repaid. 0 otherwise. U.S. Treasury

CAMEL Proxies:

Capital Adequacy Equity/Total Assets (rcfd3210/rcfd2170) Call Reports

Asset Quality: Total Loans / Total Assets Total Loans/Total Assets (rcfd2122/rcfd2170) Call Reports

Asset Quality: Real Estate / Total Loans Real Estate Loans/Total Loans (rcfd1410/rcfd2122) Call Reports

Asset Quality: Nonperforming Loans (NPL) / Total Assets

Nonperforming Loans/Total Assets ((rcfd1403+rcfd1407)/rcfd2170) Call Reports

Management Quality: Cost / Income Noninterest Expense/Total Income(riad4093/(riad4074+riad4079)) Call Reports

Earnings: Return on Assets (ROA) Net Income/Total Assets (riad4340/rcfd2170) Call Reports

Liquidity Liquid Assets/Total Assets ((rcfd0010+rcfdb989+rcfdb987)/rcfd2170) Call Reports

Control Variables:

Implicit Deposit Rate Interest Expense/Total Deposits (riad4073/rcfd220) Call Reports

Ln(Total Assets) Natural Logarithm of Inflation-Adjusted Total Assets (ln(rcfd2170t/GDP Deflatort*100)) Call Reports (rcfd2170)

(28)

APPENDIX B:CORRELATION MATRIX Implicit Deposit Rate Capital Adequacy Total Loans / Total Assets Real Estate / Total Loans NPL / Total

Assets Cost / Income

Return on

Assets Liquidity

Ln(Total Assets)

Implicit Deposit Rate 1

Capital Adequacy -0.116 1

Total Loans / Total

Assets 0.318 -0.245 1

Real Estate / Total Loans 0.019 -0.135 0.255 1

NPL / Total Assets -0.088 -0.084 0.179 0.206 1 Cost / Income -0.081 -0.091 -0.067 0.120 0.304 1 Return on Assets -0.002 0.112 -0.037 -0.157 -0.421 -0.698 1 Liquidity _-0.321 _0.117 _-0.303 _-0.132 _0.018 _0.193 _-0.095 ₁ Ln(Total Assets) 0.050 -0.181 0.207 0.315 0.111 -0.192 0.011 -0.298 1

(29)

APPENDIX C.ROBUSTNESS TESTS: NO INTERACTION EFFECTS

(1) (2) (3)

VARIABLES Only Preferred Equity Anticipated TARP First Repayment as Full Repayment

Capital Adequacy 26.887*** 26.945*** 26.903***

(1.247) (1.249) (1.247)

Total Loans / Total Assets 8.395*** 8.396*** 8.389***

(0.266) (0.266) (0.266)

Real Estate / Total Loans -2.472*** -2.476*** -2.472***

(0.298) (0.298) (0.298) NPL / Total Assets -27.557*** -27.539*** -27.474*** (1.224) (1.226) (1.225) Return on Assets 55.566*** 55.268*** 55.376*** (5.760) (5.765) (5.761) Liquidity -11.780*** -11.783*** -11.784*** (0.366) (0.366) (0.366) CPP Received -0.367*** -0.351*** -0.404*** (0.098) (0.094) (0.093) CPP Repaid -0.129 -0.171 -0.116 (0.124) (0.118) (0.118) Observations 163,626 163,626 163,626 R-squared 0.123 0.123 0.123 Number of Banks 6,794 6,794 6,794

Controls YES YES YES

Bank Fixed Effects YES YES YES

Time Fixed Effects YES YES YES

Panel B:test for equality of CPP indicators

β CPP Received - β CPP Repaid -0.238** -0.180 -0.288**

(3.96) (2.57) (6.52)

(30)

APPENDIX D.ROBUSTNESS TEST: INTERACTION EFFECTS, ONLY PREFERRED EQUITY INVESTMENTS

(1) (2)

(0.268) (0.864) (1.071) (0.34)

Real Estate / Total Loans -2.403*** -2.154*** -2.190*** -0.036

(0.299) (0.726) (0.821) (0.00) NPL / Total Assets -27.536*** -0.176 14.266 14.442 (1.263) (4.310) (11.974) (1.41) Return on Assets 55.151*** 7.221 -63.426 -70.647 (5.972) (21.549) (68.076) (1.06) Liquidity -11.775*** 0.951 -3.153* -4.104* (0.369) (2.048) (1.888) (2.95) Observations 163,626 Number of Banks 6,794 R-squared 0.123 Controls YES

(31)

APPENDIX E.ROBUSTNESS TEST: INTERACTION EFFECTS, ANTICIPATED TARP

(1) (2)

CPP Received 0.309 (0.908) CPP Repaid 0.058 (0.998) Capital Adequacy 26.545*** 5.178 18.670*** 13.492** (1.274) (4.197) (5.808) (5.10)

(0.269) (0.794) (0.983) (0.91)

Real Estate / Total Loans -2.408*** -1.772** -2.146*** -0.374

In this table, the results of a robustness test are presented where the CPP variables are not lagged. Column 1 shows the regression results including the interaction effects with the CPP variables. Capital adequacy denotes Equity over Total Assets, Return on Assets is Net Income over Total Assets, Liquidity is Liquid Assets over Total Assets. Controls are the Implicit Deposit Rate and the natural logarithm of Total Assets in 2005 U.S. dollars. Column 2 shows the difference between the interaction terms of CPP Received and CPP Repaid. CAMELs and Controls are lagged one period. Standard errors are clustered at consolidated bank level. Constant not reported. Robust standard errors in parentheses in column 1, F-values in parentheses in column 2. *** p<0.01, ** p<0.05, * p<0.1

(32)

APPENDIX F.ROBUSTNESS TEST: INTERACTION EFFECTS, FIRST REPAYMENT AS TARP REPAID

(1) (2)

(0.269) (0.787) (1.029) (0.39)

Real Estate / Total Loans -2.406*** -1.972*** -1.967** 0.005

In this table, the results of a robustness test are presented. At the first repayment CPP Received=0 and CPP Repaid=1. Column 1 shows the regression results including the interaction effects with the CPP variables. Capital adequacy denotes Equity over Total Assets, Return on Assets is Net Income over Total Assets, Liquidity is Liquid Assets over Total Assets. Controls are the Implicit Deposit Rate and the natural logarithm of Total Assets in 2005 U.S. dollars. Column 2 shows the difference between the interaction terms of CPP Received and CPP Repaid. All regressors are lagged one period. Standard errors are clustered at consolidated bank level. Constant not reported. Robust standard errors in parentheses in column 1, F-values in parentheses in column 2. *** p<0.01, ** p<0.05, * p<0.1

(33)

LIST OF REFERENCES

Acharya, V. V, Anginer, D., Warburton, A.J., 2013. The end of market discipline? Investor expectations of implicit state guarantees. Unpublished working paper. New York University Stern School of Business.

Baer, H., Brewer, E., 1986. Uninsured deposits as a source of market discipline: Some new evidence. Economic Perspectives 10, 23–31.

Bayazitova, D., Shivdasani, A., 2012. Assessing TARP. Review of Financial Studies 25, 377–407.

Berger, A.N., Roman, R.A., 2015. Did TARP banks get competitive advantages? Journal of Financial and Quantitative Analysis 50, 1199–1236.

Berger, A.N., Turk-Ariss, R., 2014. Do Depositors Discipline Banks and Did Government Actions During the Recent Crisis Reduce this Discipline? An International Perspective. Journal of Financial Services Research 48, 1–46.

Black, L.K., Hazelwood, L.N., 2013. The effect of TARP on bank risk-taking. Journal of Financial Stability 9, 790–803.

Cook, D.O., Spellman, L.J., 1994. Repudiation Risk and Restitution Costs: Toward Understanding Premiums on Insured Deposits. Journal of Money, Credit and Banking 26, 439-459.

Cornett, M.M., Li, L., Tehranian, H., 2013. The performance of banks around the receipt and repayment of TARP funds: Over-achievers versus under-achievers. Journal of Banking and Finance 37, 730–746.

Dam, L., Koetter, M., 2012. Bank bailouts and moral hazard: Evidence from Germany. Review of Financial Studies 25, 2343–2380.

De Haas, R., Van Lelyveld, I., 2010. Internal capital markets and lending by multinational bank subsidiaries. Journal of Financial Intermediation 19, 1–25.

(34)

Federal Deposit Insurance Corporation, 2010. When a Bank Fails - Facts for Depositors, Creditors, and Borrowers [WWW Document]. URL:

https://www.fdic.gov/consumers/banking/facts/payment.html (accessed 3.21.16). Gatev, E., Strahan, P.E., 2006. Banks’ advantage in hedging liquidity risk: Theory and

evidence from the commercial paper market. Journal of Finance 61, 867–892.

Gilje, E., Loutskina, E., Strahan, P., 2016. Exporting Liquidity: Branch Banking and Financial Integration. The Journal of Finance 71, 1159–1184.

Goldberg, L.G., Hudgins, S.C., 1996. Response of uninsured depositors to impending S&L failures: Evidence of depositor discipline. Quarterly Review of Economics and Finance 36, 311–325.

Hasan, I., Jackowicz, K., Kowalewski, O., Kozłowski, Ł., 2013. Market discipline during crisis: Evidence from bank depositors in transition countries. Journal of Banking and Finance 37, 5436–5451.

Hoshi, T., Kashyap, A.K., 2010. Will the U.S. bank recapitalization succeed? Eight lessons from Japan. Journal of Financial Economics 97, 398–417.

Houston, J., James, C., Marcus, D., 1997. Capital market frictions and the role of internal capital markets in banking. Journal of Financial Economics 46, 135–164.

Kane, E.J., 1989. How incentive incompatible deposit insurance funds fail. Unpublished Working Paper. National Bureau of Economic Research.

Karas, A., Pyle, W., Schoors, K., 2013. Deposit Insurance, Banking Crises, and Market Discipline: Evidence from a Natural Experiment on Deposit Flows and Rates. Journal of Money, Credit and Banking 45, 179-200.

Lamers, M.A., 2015. Depositor Discipline and Bank Failures in Local Markets During the Financial Crisis. SOM Research Reports 15007-EEF.

(35)

Ng, J., Vasvari, F.P., Wittenberg Moerman, R., 2010. The Participants in the TARP Capital Purchase Program: Failing or Healthy Banks? Unpublished Working Paper. Massachusetts Institute of Technology, Sloan School of Business.

Park, S., Peristiani, S., 1998. Market Discipline by Thrift Depositors. Journal of Money, Credit and Banking 30, 347–364.

Shin, H.S., 2009. Reflections on Northern Rock: The Bank Run that Heralded the Global Financial Crisis. Journal of Economic Perspectives 23, 101–119.

Vasileiou, E., 2014. The New Bail-in Regime and the Need for Stronger Market Discipline: What Can We Learn From the Greek Case? International Journal of Finance & Banking Studies 3, 85–113.