The impact of Basel III on bank performance:
Evidence from the securitization market
Leonie van der Laan, 11145684
July, 2016
University of Amsterdam, Amsterdam Business School
MSc Business Economics, Finance track
Master Thesis
Statement of Originality
This document is written by Student Leonie van der Laan 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 Economic and Business is responsible solely for the supervision of completion of the work, not for the contents.
Abstract
This study examines the impact of the Basel III act on the performance of US banks. A difference-in- differences method is conducted, assigning banks to treatment and control groups according to their pre-Basel III exposure to securitization. The sample constitutes of a panel dataset of 12 US banks during 2009 until 2015. Results show that, after the enactment of the Basel III act, bank performance decreased more for banks with greater reliance on securitization. It can be concluded that improving liquidity might be preferable relative to making the banking system safer. In order to test the robustness of the model, the same regressions were conducted as if the announcements and enactment took place on different dates. The results imply that the robustness of the model is questionable.
Table of Contents 1. Introduction ... 5 2. Literature review ... 6 2.1. Securitization ... 7 2.2. Bank run ... 8 2.3. Basel ... 9 3. Methodology ... 11 4. Data ... 13 4.1. CDS spreads ... 13 4.2. Funding cost ... 13 4.3. Control variables ... 13 4.3.1. Treasury rates ... 13 4.3.2. LIBOR... 14 4.3.3. VIX index ... 14
4.3.4. S&P Total Return ... 14
4.4. Descriptive statistics ... 14
5. Results ... 16
5.1. Determination treatment and control group. ... 16
5.2. Regression analysis ... 17
5.2.1. Announcement on the revision of the Basel II act ... 17
5.2.2. Announcement of the Basel III act ... 17
5.2.3. Publication of the Basel III act ... 18
5.2.4. First phase of the enactment of the Basel III act ... 18
6. Robustness checks ... 19
Conclusion ... 19
References ... 21
Appendix ………. 23
1. Introduction
Shadow banking was first introduced by the economist Paul McCulley in 2007, as a banking system in which non-depositary banks could keep complex securitized loans from their balance sheet, because they were not subject to the regulations posed to depositary banks (McCulley, 2009). These short-term securitized loans were often transformed and packaged from long-term and illiquid assets so it would improve the liquidity of these securities. Examples of these securitized loans are asset backed securities (ABS), repurchase agreements (repo) and collateralized debt obligations (CDOs). These loans were in turn insured against default by credit default swaps (CDS) acquired in the shadow banking market. Lack of confidence and a run on the short-term loan market were factors that led to the global financial crisis. In the aftermath of the financial crisis in 2007, improved regulation on financial banks and regulatory frameworks were developed and implemented in order to protect against future failures of the financial market. The Basel III act and the Dodd Frank Street Reform and Consumer Protection in the United States are examples of reforms with the aim to prevent excessive risk taking and to end cases in which institutions are ‘too big to fail’. Also the shadow banking system became subject to tightened regulations. For instance, restrictions on proprietary trading and the short-selling of CDS were imposed and higher capital requirements to compensate for CDS were implemented.
This paper will study the bank performance of 12 US banks after the implementation of the Basel III act and its corresponding capital requirements. The Basel III act made improvements on the Basel II act by, inter alia, increasing capital requirements to mitigate the risks of the securitized banking system, but with the aim not to harm the securitization of low quality securitization (Bank of International Settlements, 2014). The research question studied in this thesis will be: To what extent have the capital requirements of the Basel III act an impact
on the bank performance of US banks by focusing on the exposure to securitization? This study
is relevant as the securitization of assets prior to the crisis led to increased liquidity, an important determinant of bank performance. Higher capital requirements potentially result in decreased liquidity, which in turn could affect economic growth (IMF, 2015). It is therefore of interest to study the impact of the higher capital requirements of the Basel III act.
Prior literature on capital requirements posed to the securitization market is rather limited. The Basel III accord is enacted by the Basel committee with the aim to improve the banking system, to improve risk management and governance and to strengthen the bank’s transparency (Bank of International Settlements, n.a.). However, criticism towards this act exists, it has been argued that regulations to strengthen the banking system may instead have a
negative impact on banks. For instance, Han et al. (2015) state that higher capital requirements lead to an increase in loan sales which in turn leads to the worsening of screening and monitoring loans. Furthermore, tightened capital requirements would translate into a growing shadow market (Han, Park, & Pennacchi, 2015). Allen et al. (2012) note that the implementation of the Basel III act itself is more challenging than the capital and liquidity requirements (Allen, Chan, & Milne, 2012). This study will fill the gap left in prior literature by examining the impact of the higher capital requirements posed to, in particular, the securitization market.
The study will focus on US banks after the enactment of the Basel III act. A difference-in-differences (DiD) method is conducted with banks that are more subject to the securitization banking system as the treatment group and banks that are less affected by the securitization banking system as the control group. A sensitivity analysis on the exposure to pre-Basel III securitization is performed for each bank in order to determine the treatment group and the control group by means of the median. Thus, among the treatment group and the control group, there will be six banks each. Subsequently, a DiD-method is performed for several announcement dates and the enactment date of the Basel III act and for several time frames. The average treatment effect is obtained from this study, which will give insights on the impact of higher capital requirements on bank performance. This study finds that the average treatment effect is equal to an increase of 0.041% in the change in daily CDS spreads after the enactment.
The paper consists of 7 sections including the introduction. The second section is a literature review which consists of the main theories in already existing literature on this topic and how this relates to this study. The third section will cover the methodology including the econometric model that will be conducted in this study. The fourth section will consist of the data and its descriptive statistics. The fifth section will cover the main results of the analysis conducted. The sixth section will cover robustness checks in order to examine the robustness of the study. And lastly, the seventh section will be concluding the results of this study.
2. Literature review
In order to elaborate on previous literature, one should understand the main theories that deal with securitization and bank runs. Literature on the securitized banking system and its role in the global financial crisis will be assessed further in depth. Furthermore, literature on the enactment of the Basel III act will be provided and the influence of regulations to enhance financial health after the crisis will be elaborated on.
2.1. Securitization
Securitization is the process of transferring an illiquid asset into a liquid security, by combining and pooling together several financial products of different credit qualities. By means of this asset combination, one can eliminate some of the interest payments and principal payments by shifting the credit risk to outside investors (Jobst, 2008). Accordingly, securitization has become a common approach for banks to reduce its capital requirements.
The securitization market originates from the 1970s in the US with the securitization of guaranteed residential home mortgages of Ginnie Mae traded as mortgage backed securities (MBS) in the secondary market for mortgages. Prior to the development of MBS, the secondary market for loans was illiquid and trading in these loans was relatively costly. The invention of MBS made it easy to pass these loans through to outside investors by pooling loans together and transforming them into liquid products allowing the elimination of interest rate risk from the balance sheet by the lender (Cowan, 2003). The quality of these products is relatively hard to measure, as the lower the quality is of the collateral, the harder it is to structure the security (Ghent, Torous, & Valkanov, 2016). Shortly after the introduction of MBS, the ABS was introduced in the 1980s. These loans involve several asset types, such as credit card loans, auto loans, home equity loans and student loans. During the 2000s a security was introduced which is somewhat comparable to ABS and MBS but consists of less and larger products from different asset classes, referred to as collateralized debt obligations (CDOs). The cash flows from these loans come from the leaseholder of the contract and are sold to special purpose vehicles (SPVs) (Gorton & Souleles, 2005). SPVs are entities that are created by the company that will be transferring the assets to the SPV in order to eliminate the risk of bankruptcy. This is also the only feature the SPV has and it has no employees nor a physical location.
The funding of these assets is generally executed via repos or ABCP programmes in the shadow banking market. In a repurchase agreement, the seller will sell an asset, often in government securities, to the buyer which will be repurchased at a later predetermined time in collateralized with securities (Krishnamurthy, Nagel, & Orlov, 2014). The repo rate is the rate at which commercial banks can borrow from the national central bank, whereas the haircut is the difference in the rate at which the security can be bought and sold, given as 1 minus the notional amount over the collateral. Similarly, ABCP programs offer short-term loans, but via SPVs, and are regularly used to finance securities as ABS and MBS in the run up to the financial crisis (Covitz, Liang, & Suarez, 2013). The study by Acharya et al. (2013) find that SPVs for ABCP were used by commercial banks in order to evade regulations so it could reduce their effective capital requirements. Furthermore, their results showed that not the investors who
heavily invested in ABCP incurred large losses due to the crisis, but instead the commercial banks faced large losses due to their insurances with outside investors (Acharya, Schnabl, & Suarez, 2013). Generally, the market for repos is larger than the market for ABCP, however the study by Krishnamurthy, Nagel and Orlov (2014) finds that the market for ABCP and direct investments is larger than the repo market in funding securitized assets (Krishnamurthy, Nagel, & Orlov, 2014). In order to control for the volume invested in repos and ABCP programmes, this study will take both variables into account in the assessment of a bank’s exposure to securitization.
The banks trading in these securities will most likely trade in credit default swaps which are used as an insurance in order to transfer the credit risk on the underlying assets. This process is also called synthetic securitization by selling the credit risk instead of the security itself (Jobst, 2008). CDS spreads have therefore become an indicator of a firm’s credit risk. Gorton and Metrick (2011) use CDS spreads as a measure of the risk premium of the company buying the protection. The CDS spread would therefore be a well indicator for the performance of a bank, that is, if the chance of default becomes larger then the spread is also expected to increase. However this is not the only determinant of a spread as it also depends on the structure of the market, investors and the state of the economy as a whole (Chiaramonte & Casu, 2012). The magnitude of the CDS market depends on whether the underlying credit is risky or not. If the underlying credit is risky the bank would rather sell off the loans than setting up a CDS contract. However, stricter regulation on banks can lead to the expansion of the CDS market (Parlour & Winton, 2013).
Gennaioli et al. (2013) introduced a newly developed model of shadow banking in which the outside investor is the driving force behind the demand short term debt, in particular by the securitization of these loans. The continuing role of securitization, among other things, eventually harms the banking system when tail risks are disregarded. In the proposed model they emphasize that basing capital requirements on risk weights is not enough to eliminate risks due to credit agencies but instead the focus should lay on the overall leverage of a bank (Gennaioli, Schleifer, & Vishny, 2013).
2.2. Bank run
The collapse of Lehman Brothers in 2008 led to a panic in the banking industry and to a subsequent run on short-term debt which has been examined by Ivashina and Scharfstein (2010). By the end of 2007, the volume in short-term loans including syndicated loans started
(Ivashina & Scharfstein, 2010). During a bank run, depositors withdraw their deposits from the bank when there is for instance lack of confidence. Such bank runs eventually result in the failure of banks or even to disruptions in the monetary system as a whole. Diamond and Dybvig (1983) developed a model which suggested that even healthy banks could suffer from bank runs and that bank runs eventually could lead to serious economic issues in the banking system (Diamond & Dybvig, 1983). The worldwide bank run in 2007-2008 was partially caused by some elements of the securitization market, an example of such an element is the run on the repo market. This run is different than the traditional bank run explained before, but instead this run is driven by a run on the securitized banking system (Gorton & Metrick, 2011). The run on the securitized banking system was mainly driven by the run on repo, however the market for ABCP suffered from runs as well. Furthermore, Gorton and Metrick (2009) argue that in order to have a liquid market asymmetric information should be in place. This can be explained by increased repo haircuts when the securitized banking system has to sell its assets off (Gorton & Metrick, 2009). Due to the lack of confidence, the repo haircut increased which caused the withdrawal of deposits from banks and led to a subsequent freeze of the repo market (Gorton & Metrick, 2011). Unlike the traditional banking system securitized loans do not occur on the bank’s balance sheet, however some of the underlying securitized bonds could be used as collateral and will appear on the balance sheet (Gorton & Metrick, 2011). These events and the unregulated shadow banking system ensured the development of tightened regulation. Basel III in particular provided stricter rules concerning the capital adequacy of banks.
2.3.Basel
The Basel III accord is enacted by the Basel Committee on Banking Supervisions. This committee was established by the G10 in the 1970s when international financial markets faced foreign exchange losses due to the failure of the Bretton Woods system (Bank of International Settlements, 2015). The aim of the committee is to enhance financial stability and to provide supervision on the banking system, with capital adequacy as the main point of focus. Differences in the national capital standards of banks led to inequality between international banks in the 1980s. As a consequence, the Basel Committee aimed to enhance international financial stability by equalizing these capital standards for all participating members by means of the Basel Capital Accord, or the Basel I act, in 1988. By this agreement, banks have to hold on to a minimum capital requirement of 8%, measured as capital over risk-weighted assets, by 1992. In 2004, the enactment of the Basel II accord took place, which aimed to further develop the standards of the Basel I accord and to improve capital regulation. The preceding financial
crisis in 2008 learned that the requirements of the act were insufficient to prevent a crisis with the magnitude of the global financial crisis. In response to the financial crisis, the committee announced that it will improve the Basel II act, including improved regulation on the complex securitization system and the corresponding off-balance sheet funding. In December 2010, the committee published the official rules on the new and higher capital and liquidity requirements, as referred to as the Basel III accord. The Basel III accord included improvements on the Basel II accord and made additional amendments. For instance, a bank has to hold a minimum amount of capital that would compensate for the potential losses on all the bank’s assets and off balance sheet vulnerabilities measured as a leverage ratio. The Basel III accord will be phased in over a period of five years and should be in full progress by 2019.
Allen et al. (2012) argue that the problem of Basel III does not concern capital and liquidity requirements, but instead the consequent adjustments of financial regulation in the banking industry has a negative impact on the market for credit (Allen, Chan, & Milne, 2012). The implementation of the Basel III act is phased in in several stages, whereas Allen et al. (2012) argue that a credit crisis is expected before the accord is in full service. The Bank of International Settlements (BIS) (2010) studied the long-term impact of stricter requirements on liquidity and capital on the economy and state that, due to the higher capital requirements, banks will become safer. Safer banks result in lower costs of capital and will lead to a decreased impact on lending spreads (Bank for International Settlements, 2010). Berger and Bouwman (2013) confirm that capital adequacy helped to enhance financial health during financial crises for as well small banks and medium to large banks (Berger & Bouwman, 2013). The study by Ayadi et al. (2016) find that regulation has no impact on bank performance if a bank complies to all regulations and standards. This study is basing its results on international capital requirements and the Basel Core Principles for Effective Bank Supervision Furthermore, regulation has little impact on bank efficiency, however higher standards could affect bank efficiency in some cases (Ayadi, Naceur, Casu, & Quinn, 2016). Whereas Cihak et al. (2013) argue that simple regulations but well enforced are less likely to result in crises as opposed to more complex regulations (Cihak, Demirguc, & Soledad Martinez Peria, 2013). The Basel III accord is complex in a way as the rule is not implemented at once but gradually per phase. It is evident that dissension exists on the impact of tightened regulation on bank performance and it is therefore of interest to study the impact of the Basel III accord.
3. Methodology
This study will use panel data in order to examine the bank performance of US banks after the implementation of the Basel III act. The hypothesis that will be tested in this study is the following:
𝐻": The enactment of the Basel III act has a negative impact on the bank performance of US
banks due to tightened regulation on the securitized banking system.
The first phase-in arrangement of the Basel III act is in force by the 1st of January 2013, and will be gradually implemented until it is in full service by 2019. In addition, announcement dates of the Basel III act will be examined in order to indicate the impact of Basel III on bank performance. On the 12th of September 2010, the G10 Group of Governors and Heads of Supervision announced that it agreed on strengthening the capital requirements by means of Basel III. The announcement followed after the review of the capital and liquidity measures of Basel II by the Basel committee on the 26th of July that year. On the 16th of December 2010, the committee published the official text on the capital and liquidity requirements of Basel III.
A difference-in-differences (DiD) method will be conducted by means of a treatment group and a control group. This study will focus on a bank’s exposure to securitization, therefore a distinction will be made between banks that are highly operative in the securitization market and banks that are less subject to securitization, translating into the treatment group and the control group. The DiD method is a common approach to investigate the impact of, for instance, policy measures by measuring the difference between the group that is affected and the group not affected. The essential regression model that will be used in this study is the following:
∆𝐶𝐷𝑆 𝑠𝑝𝑟𝑒𝑎𝑑/0 = 𝛼 + 𝛽𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡 𝑔𝑟𝑜𝑢𝑝/ + 𝛾𝑃𝑜𝑠𝑡 𝐵𝑎𝑠𝑒𝑙 𝐼𝐼𝐼 𝑎𝑐𝑡0+
𝛿𝑇𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡 𝑔𝑟𝑜𝑢𝑝/ ∗ 𝑃𝑜𝑠𝑡 𝐵𝑎𝑠𝑒𝑙 𝐼𝐼𝐼 𝑎𝑐𝑡0+ 𝑐𝑜𝑛𝑡𝑟𝑜𝑙 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠 + 𝛼/+ 𝜀/0 (1) The log transformed daily CDS spreads will be regressed on the dummy variable treatment group, the dummy variable post announcement date or enactment date and the interaction term of both dummy variables, which is the variable of interest, controlled for macroeconomic variables. The macroeconomic variables include treasure rates, as the 10 year US bond rate, the 3-month T-bill rate and the term spread, the 3-month LIBOR rate, the change in the volatility index (VIX) and the change in S&P 500 total return. The dummy variable treatment group will be equal to 0 or 1 depending on the bank’s exposure to securitization. In order to determine which bank belongs to which group, the following regression model will be conducted:
∆𝐶𝐷𝑆 𝑠𝑝𝑟𝑒𝑎𝑑/0 = 𝛼 + 𝛽 𝑓𝑢𝑛𝑑𝑖𝑛𝑔 𝑐𝑜𝑠𝑡𝑠0+ 𝑐𝑜𝑛𝑡𝑟𝑜𝑙 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠0+ 𝜀/0 (2) Among funding costs, the overnight repo rate, the interest rate on 30-day ABCP, and the interaction term of both variables will be included, in addition to macroeconomic control variables. We will run the regression for each bank and it is of importance that the measure that will point out the exposure to securitization before Basel III is significant for each bank in order to appropriately assign banks to treatment and control. This measure can either be the coefficient on the repo rate, the ABCP rate, the interaction term or a measurement of some of them. The median from the measure will be obtained, as such that the banks with a sensitivity coefficient above the median are relatively more exposed to securitization and will therefore serve as the treatment group. The banks below this measure will serve as the control group.
Four dates, the three announcement dates and the enactment date, will be taken into account in the difference-in differences analysis and each date will be tested for several timeframes. Per announcement date or enactment date, the regression analysis will consist of three regressions, for +/- one week, +/- two weeks and +/- four weeks. The dummy variable for post announcement date or post enactment date will be equal to 1 from the day on that the announcement is made and equal to 0 for the days before the announcement. Furthermore, regression will be conducted as if the announcement or enactment took place one day before or one day after in order to assess whether private information has been available to the public before the announcement or enactment.
The fundamental regression model will control for fixed entity effects using the STATA command xtreg, fe, which is given by α in formula (1). A fixed effects regression model is used when the sample contains two or more observations per bank. Controlling for fixed entity effects implies that a dummy variable for eleven of the twelve US banks, otherwise the regressors would be perfectly multicollinear, will be included in the regression model in order to control for variation in different banks. The variable treatment is omitted from the regression due to collinearity as it does not vary with time.
Placebo tests will be conducted as in the study by Campello et al. (2016) in order to test the robustness of the study, that is, regressions will be conducted as if the announcement dates and the enactment date are on different dates (Campello, Ladika, & Matta, 2016). The study will run the regression in (1) for an event window of 8 weeks in which after 4 weeks the placebo announcement or enactment takes place. The study should find insignificant results on the average treatment effect in the regressions on the placebo events in order to emphasize on the robustness of the study.
4. Data
The sample consists of 12 US banks which have been chosen dependent on their availability in the Thomson Reuters CDS database. Data will be acquired from 2009 until 2015, so it will take both the time into account of before the announcements of the Basel III act and after the announcements until to the end of 2015 when the phase-in arrangements of the Basel III act were in progress.
4.1. CDS spreads
Data on CDS spreads will be acquired from the Thomson Reuters CDS database for US banks. This study will solely focus on 5-year CDS spreads in order to eliminate differences in CDS contracts. Furthermore, it will focus on senior debt and all spreads are of the type ‘XR’ (no restructuring), so restructuring events will be excluded from the sample. The logarithm will be taken from CDS spreads in order to adjust for the large values CDS spreads have, as spreads are given in basis points.
4.2. Funding costs
The data on funding costs, the repo rate and the rate on asset backed commercial paper (ABCP), is acquired from Thomson Reuters and the Federal Reserve respectively. The overnight repo rate will be taken into account, as well as the 30-day ABCP interest rate. CDS spreads per bank will be regressed on the repo rate, the ABCP interest rate and the interaction term of the repo rate and the ABCP interest rate to figure out which banks are more subject to securitization before Basel III. The repo rate, as well as the ABCP rate, are expected to be positively correlated with CDS spreads, as both rates tend to increase during periods of economic uncertainty and illiquidity.
4.3. Control variables
This study will in particular control for macroeconomic variables for which data is available on a daily basis. US treasury rates, LIBOR, the change in VIX rate and the change in the total return on the S&P 500 will be among the control variables.
4.3.1. Treasury rates
The 10-year US government bond rate, as well as the 2-year US government bond rate, is acquired from the Thomson Reuters database. The term spread, given as the difference between
the long-term and short-term bond rate, either the difference between the 10-year bond rate and the 2-year bond rate or the difference between the 10-year bond rate and the 3-month T-bill rate, will be added. Moreover, the 3-month Treasure bill rate will be added as a control variable. The 10-year and 2-year US bond rates and the 3-month US T-bill rate are expected to be positively correlated with CDS spreads.
4.3.2. LIBOR
The 3-month London Interbank Offered Rate (LIBOR) will be taken into account as a benchmark for short-term interest rates. The daily data on LIBOR rates is acquired from the ICE Benchmark Administration Limited via the Federal Reserve Bank of St. Louis. The LIBOR rate is expected to be positively correlated with CDS spreads, as for instance a weakened banking system translates into higher interest rates and higher spreads.
4.3.3. VIX index
The volatility index (VIX) will be taken into account as a measure of volatility. The VIX index is a measure of the implied volatility of S&P 500 index options and gives a 30-day expectation of the market volatility. The data is acquired from the Chicago Board Options Exchange (CBOE). The VIX index is converted into logarithms in order to adjust for the large values as the VIX index is given in percentage points. The VIX index is expected to be positively correlated to CDS spreads as volatility tends to widen spreads.
4.3.4. S&P Total Return
Data on the total return of the S&P 500 will also be acquired from the CBOE. The logarithm of the S&P total return will be taken in order to adjust for the large values. The S&P total return is expected to be negatively correlated with CDS spreads, as higher total returns translate into smaller spreads.
4.4. Descriptive statistics
Figure 1 shows the plotted daily CDS spreads of the sample banks from the 1st of January 2009 to the 31st of December 2015. From the graph, it can be argued that CDS spreads experienced a significant increase in 2009 in the aftermath of the financial crisis. The summary statistics in table 1 on the CDS spreads of all banks in the sample confirm this. In 2009, the maximum spread was equal to approximately 1336 basis points (HSBC) and, as can be observed from the
graph, represents a large outlier in comparison to the other banks. Nevertheless, the other banks experienced widened spreads as well, resulting in a mean of 196 basis points and a standard deviation of 157 basis points. In 2010, banks recovered from the financial crisis and spreads stabilized, which resulted in CDS spreads varying between 39 basis points and 299 basis points. In 2010 as well, on the 27th of July 2010, the Basel Committee first announced the revision on the Basel II act and agreed on the strengthening of capital and liquidity requirements in response to the financial crisis. Likewise, on the 12th of September 2010, the committee announced that the strengthened capital and liquidity requirements were published under Basel III. Thereafter, on the 16th of December 2010, the official text of the Basel III act has been published. A small peak in spreads in 2010 appear in the graph, empirical evidence will have to prove whether the announcements on the Basel III act did have an impact on bank performance considering the spike in CDS spreads. By mid 2011 a significant peak appears in the graph as CDS spreads widened due to the European sovereign debt crisis. The possible default of the Greek government and staggering economies as Spain and Italy instigating uncertainty in the US banking industry. The maximum CDS spread at that point was approximately equal to 610 basis points (Morgan Stanley) and the mean and standard deviation were equal to 137 and 89 basis points respectively. Likewise, in 2012, spreads remained widened due to the uncertainty on the European sovereign debt market and spreads varied between 50 and 453 basis points. It was not until 2013 that the market became more stable and that the first phase-in arrangements made their appearance. However, from the graph it is unclear whether the enactment of Basel III, the 1st of January 2013, had an impact on bank performance and should indeed be examined.
Table 2 contains the summary statistics of all variables that will be included in the regression analyses. Whereby, the log transformed CDS spreads of the sample and the log transformed CDS spreads per bank will act as independent variable in the main regression analysis and the determination of the control and treatment group respectively. From the table it can be derived that banks as the Bank of NY Mellon, PNC Financial and Charles Schwab tend to have the narrowest spreads, whereas HSBC has the widest spreads in the sample. Funding costs are given in panel B in which is shown that the overnight repo rate varies between -0.01% and 6.40% and the ABCP rate fluctuates between 0.12% and 0.95% in the sample period. Negative repo rates occur in periods of financial distress, which implies that demand for short-term loans is high. Summary statistics for the macroeconomic control variables are given in Panel C. The long-term Treasury bond rate ranges from 1.4% to 3.99%, whereas the short-term Treasury bill rate varies between 0 and 0.3%. The term spread, the gap between short-term and long-term interest rates, fluctuates between 0.15% and 1.07%, with a mean
difference of 0.68%. Furthermore, the LIBOR rate varies between 0.22% and 1.42%, and is on average equal to 0.37% with a standard deviation of 0.21%, which indicates that the LIBOR rate somewhat fluctuates.
5. Results
In this section the main results will be presented after investigating each bank’s exposure to securitization in order to draw conclusions on what specification of treatment and control to choose. The main results should give insights on the impact of the Basel III in the particular context.
5.1. Determination treatment and control group.
Per bank the change in CDS spread will be regressed on funding costs and control variables. In table 3, the regression results of the regression of each bank’s CDS spread on the repo rate, the ABCP rate, the interaction term, the 10-year US bond rate and the term spread, the difference between the 10-year bond rate and the rate on the 3-month T-bill rate. This specific model gives for at least one of the funding costs a significant result, the repo rate. The median of the coefficients on the repo rate is equal to 1.489, which should draw the line between treatment and control. The mean of daily CDS spreads per bank is plotted against the bank’s sensitivity to securitization in figure 2. The banks on the right of the median will serve as the treatment group. Notable is the positive sensitivity coefficient for Charles Schwab as opposed to the negative values of the other banks. Overall, the negative values would indicate a negative correlation between CDS spreads and the overnight repo rate, which is different than expected. The mean of the CDS spreads of the treatment group derived from the regression, Charles Schwab, JPMorgan, Citigroup, Morgan Stanley, Wells Fargo and HSBC, and the mean of the control group, containing the residual banks, are plotted in figure 3. From this figure, it can be argued that the mean of the treatment group is for the vast majority larger than the mean of the control group until the beginning of 2013. This is interesting in terms of this study as the first phase of Basel III took place. In this study, it is assumed that the size of spreads indicates the financial health of a bank. From this figure, it can be concluded that bank performance improved after stricter regulation posed to the shadow banking system by means of higher capital requirements. The regression analysis should explain the impact of Basel III on the securitization market.
5.2. Regression analysis
For several dates a regression analysis will be conducted in order to examine whether the announcements on the Basel III act and the enactment itself had a significant impact on the performance of US banks by means of securitization.
5.2.1. Announcement on the revision of the Basel II act
The announcement on the 26th of July will be examined in order to indicate whether this announcement on Basel III could give a causal interpretation to the impact of Basel III on bank performance. In the regression analysis, for which the results are given in panel A of table 4, the variable treatment is omitted from the regression due to multicollinearity. The interaction term of treatment and post announcement is the variable of interest in this study as it explains the impact of the rule by assessing the difference in mean outcome of treatment and control. However, the results on the average treatment effect are not significant and no conclusions can be drawn from the regression analysis that would support this study.
In panel B of table 4, the results are given for the 8-week, 4-week and the 2-week event window but as if the announcement date was 1 day before or 1 day after the actual date, this should indicate whether there has already been information available before the announcement. However, the coefficient on the average treatment effect is insignificant as the same applies to the results in panel A. From this it can be concluded that the announcement on the 26th of July
could not give a causal interpretation on the impact of Basel III on bank performance given the exposure to securitization, nor did information leakage apply that had a significant impact on bank performance.
5.2.2. Announcement of the Basel III act
The results of the regression model with the 12th of September 2010 as a benchmark are given in panel A of table 5. The conclusion that can be drawn from the results is that the difference in the mean outcome of the treatment group and the mean outcome of the control group within the timeframe of 8 weeks is negative, whereas the results on the timeframe of 4 weeks and 2 weeks are insignificant. This would, thus, suggest that bank performance improved after the second announcement, concluding from the regression results of the first regression. The regression results demonstrate a decrease of 0.026% in spreads after the announcement and the results are explained by 49.3% by the model. Furthermore, the first two regression results in panel B of table 5 imply that there was already some noise around the announcement date as the results
are significant for the timeframe of 8 weeks. Above all, the results for +/- 2 weeks and +/- 1 week are insignificant for panel A and panel B. Overall, the outcome is not convincing enough to speculate on the impact of the accord on bank performance.
5.2.3. Publication of the Basel III act
In panel A of table 6, the results of the regression model with the 16th of December 2010 as benchmark are given. All the results on the average treatment effect are significant at the 1% significance level and imply an increase in spreads after the publication of Basel III. The variation in results is explained by the model for 26.7% in the timeframe of +/- 1 week to 60.8% in the timeframe of +/- 4 weeks. The publication of the rules on strengthening regulation on the capital adequacy of banks, by for instance the higher capital requirements to compensate for the exposure to securitization, led to a short disturbance in the banking system. It would suggest that bank performance decreased more for banks with a greater reliance on securitization after the announcement, hence the null hypothesis cannot be rejected. The result would confirm the results of the study by Allen et al. (2012) that the continuing adjustments in regulatory frameworks on capital adequacy would have a negative impact on the credit market (Allen, Chan, & Milne, 2012). In panel B, a significant as well positive result is found for the average treatment effect in all regressions, therefore it can be concluded that some private information has already been available to the public before the rules on the act was officially published.
5.2.4. First phase of the enactment of the Basel III act
The results of the regression with the enactment date as a benchmark are given in panel A of table 7. Above all, the coefficients on the average treatment effect are positive and significant at the 1% significance level and the 5% significance level. Similarly, these results confirm a decrease in bank performance after the enactment of the Basel III accord considering exposure to securitization. The result of the coefficient on the average treatment effect in the last regression indicates that the difference between mean outcome of the treatment group and control group within the timeframe of two weeks is equal to an increase of 0.041% in spreads, explained by 50.8% by the model. The null hypothesis can therefore not be rejected as for both the third announcement date and the enactment date increased spreads are obtained, whereas the results on the first and second announcement date are not significant or convincing enough to draw conclusions from. The results are therefore not in line with the results of BIS (2010) that cost of capital and spreads will decrease after the strengthening of capital requirements
(Bank for International Settlements, 2010). Above all, the decrease in cost of capital does not outweigh the decrease in liquidity due to tightened regulation. Furthermore, the results in panel B are somewhat equal to the results obtained in panel A. Likewise, the coefficient on the average treatment effect is significant and it can be concluded that the days around the enactment date are relevant in the examination of the impact of Basel III on bank performance.
6. Robustness checks
Placebo tests are done in order to test for robustness by assuming that the announcement dates and the enactment date were on different dates. The dataset is divided into timeframes of two months and it is assumed that after one month the announcement or the enactment takes place. The regression can be compared to the first regression in table 4, 5, 6 and 7, the coefficient on the average treatment effect should thus be equal to the regression analysis for the actual date. This subsequently results in 11 regressions per announcement date or enactment date, of which 5 regressions in the period before the actual date and 5 regressions in the period after the actual date. As can be concluded from the main results, the coefficients on the average treatment effect taking 4 weeks before and 4 weeks after into account are significant for the regression analyses on the second announcement, the third announcement and the enactment date, therefore the robustness checks will solely include these three regression models. The results are given in figure 4 and are all arranged from large to small coefficients. The first figure shows the regression results of the second announcement, from which it can be seen that the coefficient is a negative value, but does not appear to be statistically significant at the 95% confidence interval, which is illustrated by the two lines in the figure. The same applies to regression analysis on the third announcement date and the enactment date including the placebo events. However, the average treatment effect of one of the placebo events lays outside the confidence interval and would therefore be interesting for future studies. The robustness of the model is questionable and future research should perform other econometric models in order to reveal the impact of Basel III.
Conclusion
This study examined the impact of higher capital requirements on the performance of US banks after the implementation of the first phase of the Basel III act. Higher capital requirements were implemented in order to compensate for the off-balance sheet funding of securitized loans in the shadow banking market and this study has examined whether these requirements harm
bank performance. Results show that bank performance weakened after the enactment of the Basel III act as the change in spreads for instance increased with 0.041% within a 2-week timeframe. The results indicate that stricter capital requirements posed to banks in order to mitigate the risks of the shadow banking market did not improve the performance of banks. The performance of banks with a greater reliance on securitization did worse than banks with less reliance on securitization. It can be concluded that improving liquidity might be preferred to making the banking system safer. These results are not in line with a majority of prior literature on the impact of capital regulation on bank performance. The results do not support the results of the BIS that strengthened regulation will lead to lower costs of capital as banks become safer (Bank for International Settlements, 2010). However, as argued by Allen et al. (2012), the rule might be too complex due to the phase-in process and the continuing changes in regulation in order to be effective (Allen, Chan, & Milne, 2012). The robustness of the model is questionable following the results of the robustness checks and future research should perform other econometric models in order to mitigate misleading results and to confirm the results.
The small sample of 12 banks is one of the limitations of this study. If the sample would contain more US banks the results would be more reliable and an enhanced distinction can be made between treatment and control. The average treatment effect is taken from the difference in the mean outcome of the treatment group and control group after the enactment, however banks in the control group also participate in the shadow banking system only to a smaller extent. A larger sample would improve the determination of treatment and control. In addition, the Basel III act is not in full service yet, therefore in order to examine the impact on bank performance, it would be preferred to study the impact after the act is fully implemented. An alternative interpretation of the results could be that banks more exposed to securitization perform less well than the control group apart from taking the Basel III act into account. Future research should take this into account, as well as different econometric models to examine the impact.
References
Acharya, V., Schnabl, P., & Suarez, G. (2013). Securitization without risk transfer. Journal of
Financial Economics, 515-536.
Allen, B., Chan, K., & Milne, A. (2012). Basel III: Is the cure worse than the disease?
International Review of Financial Analysis.
Ayadi, R., Naceur, S., Casu, B., & Quinn, B. (2016). Does Basel compliance matter for bank performance. Journal of Financial Stability.
Bank for International Settlements. (2010). An assessment of the long-term economic impact
of stronger capital and liquidity requirements. Bank for International Settlements.
Bank of International Settlements. (2014). Basel III Document: Revisions to the securitisation
framework.
Bank of International Settlements. (2015, October). A brief history of the Basel committee. Retrieved from http://www.bis.org/bcbs/history.pdf
Bank of International Settlements. (n.a.). International regulatory framework for banks (Basel
III). Retrieved from Bank for International Settlements:
http://www.bis.org/bcbs/basel3.htm
Berger, A. N., & Bouwman, C. H. (2013). How does capital affect bank performance during financial crises? Journal of Financial Economics, 146-176.
Campello, M., Ladika, T., & Matta, R. (2016). Debt restructuring costs and corporate bankruptcy.
Chiaramonte, L., & Casu, B. (2012). The determinants of bank CDS spreads: evidence from the financial crisis. The European Journal of Finance, 861-887.
Cihak, M., Demirguc, A., & Soledad Martinez Peria, M. (2013). Bank regulation and supervision in the context of the global crisis. Journal of Financial Stability.
Covitz, D., Liang, N., & Suarez, G. A. (2013). The Evolution of a Financial Crisis: Collapse of the Asset-Backed Commercial Paper Market. Journal of Finance, 815-848. Cowan, C. (2003). Hearing on Protecting Homeowners: Preventing Abusive Lending While
Preserving Acces to Credit. American Securitization Forum . New York.
Diamond, D. W., & Dybvig, P. H. (1983). Bank Runs, Deposit Insurance, and Liquidity.
Journal of Politcal Economy, 401-419.
Gennaioli, N., Schleifer, A., & Vishny, R. W. (2013). A model of shadow banking. Journal of
Finance, 1331-1363.
Ghent, A., Torous, W., & Valkanov, R. (2016). Complexity in Structure Finance. Gorton, G., & Metrick, A. (2009). Haircuts. NBER working paper series.
Gorton, G., & Metrick, A. (2011). Securitized banking and the run on repo. Journal of
Financial Economics.
Gorton, G., & Souleles, N. S. (2005). Special purpose vehicles and securitization. NBER
Working Paper 11190.
Han, J., Park, K., & Pennacchi, G. (2015). Corporate taxes and securitization. Journal of Finance.
IMF. (2015). Global Financial Stability Report: Risks Rotating to Emerging Markets. Ivashina, V., & Scharfstein, D. (2010). Bank lending during the financial crisis of 2008.
Journal of Financial Economics, 319-338.
Jobst, A. (2008). Back to basics: What is securitization? Finance and Development. Krishnamurthy, A., Nagel, S., & Orlov, D. (2014). Sizing up repo. Journal of Finance. McCulley, P. (2009, May). The Shadow Banking System and Hyman Minsky's Economic
Journey. Global Central Bank Focus.
Parlour, C., & Winton, A. (2013). Laying off credit risk: Loan sales versus credit default
swaps. 25-45: Journal of Financial Economics.
Appendix
Figure 1. CDS spreads per US Bank.
This figure shows the CDS spreads of twelve banks in the United States in the period of the 1st of January 2009 to the 31st of December 2015. All CDS spreads are given in basis points.
Table 1. Summary statistics of CDS spreads per year.
This table shows the summary statistics of the CDS spreads of the sample of US banks per year from 2009 to 2015. The mean, standard deviation, minimum and maximum are given in basis points.
Year Obs Mean Std. Dev. Min Max
2009 2895 196.1551 156.9069 38.9200 1335.5780 2010 3132 112.0688 46.6417 38.9200 298.9099 2011 3120 137.4334 88.9676 48.8100 609.8130 2012 3132 146.7426 82.7473 50.2300 453.0229 2013 3132 84.7144 30.6422 25.8500 190.4300 2014 3132 61.5525 22.0560 26.5800 108.8300 2015 3132 63.9958 23.0338 27.0900 104.1800 CDS spreads
Table 2. Summary statistics of the CDS spread of the sample of US banks and funding costs.
This table shows the summary statistics of the CDS spreads of all US banks in basis points, funding costs in percentage points and control variables in the percentage points from the 1st of January 2009 to the 31st of December 2015.
Variable Obs Mean Std. Dev. Min Max
Panel A: CDS spreads (in bp)
CDS spreads (total) 21675 113.7625 89.6161 25.8500 1335.5780 CDS spreads (total) 21675 4.5359 0.5946 3.2523 7.1971 Bank of America (BAC) 1826 4.8718 0.5181 4.0647 6.1894 Bank of NY Mellon (BK) 1826 4.4455 0.2840 3.8879 4.6445 Capital One (COF) 1826 4.1787 0.5325 3.2992 6.0580 Citigroup (C) 1826 4.9384 0.5415 4.0697 6.4922 American Express (AXP) 1826 4.3309 0.5862 3.5407 6.4968 Goldman Sachs (GS) 1826 4.9038 0.4290 4.1559 6.0486 Morgan Stanley (MS) 1826 5.0391 0.5498 4.0826 6.4132 PNC Financial (PNC) 1754 4.4648 0.2664 4.0863 5.8033 Charles Schwab (SCHW) 1661 3.9205 0.2217 3.4834 4.2509 HSBC 1826 4.4766 0.9030 3.2523 7.1971 JP Morgan Chase (JPM) 1826 4.4394 0.3069 3.7880 5.4524 Wells Fargo (WFC) 1826 4.3628 0.4326 3.5664 5.7088
Panel B: Funding costs
Overnight repo rate 21912 0.1512 0.0789 -0.0100 0.0640 ABCP rate 30-day 21912 0.2541 0.1173 0.1200 0.9500 Repo rate * ABCP rate 21912 0.0427 0.0387 -0.0019 0.3008
Panel C: Macroeconomic variables
US bond 10-year 21912 2.5688 0.6326 1.404 3.9894 US T-bill 3-month 21912 0.0815 0.0602 0.0000 0.3000
Termspread 21912 0.6799 0.2449 0.1520 1.0672
S&P Total Return 21144 7.8169 0.3129 6.9985 8.2788
LIBOR 21222 0.3733 0.2166 0.2229 1.4213
Table 3. Regression of the change in CDS spreads per bank on funding costs and control variables.
AXP BAC BK C COF GS
Repo rate -1.577*** -1.641*** -2.060*** -0.743** 1.665*** -1.796*** (0.30) (0.29) (0.22) (0.29) (0.25) (0.24) ABCP rate 2.024*** 1.405*** -1.103*** 1.330*** 1.535*** 0.714*** (0.22) (0.21) (0.16) (0.21) (0.18) (0.17) Repo*ABCP rate 3.252*** 0.448 2.959*** 1.049 3.386*** 3.058*** (0.93) (0.91) (0.68) (0.90) (0.78) (0.73) US bond 10 year -0.418*** -0.276*** -0.441*** -0.501*** -0.200*** -0.503*** (0.03) (0.02) (0.02) (0.02) (0.02) (0.02) Termspread 1.437*** -0.273*** 0.446*** 1.306*** 0.494*** 0.205*** (0.07) (0.07) (0.05) (0.07) (0.06) (0.06) Constant 4.615*** 5.922*** 5.825*** 5.721*** 4.471*** 6.336*** (0.09) (0.08) (0.06) (0.08) (0.07) (0.07) R2 0.729 0.556 0.599 0.573 0.631 0.732 N 1043 1043 1043 1043 1043 1043
Standard errors in parentheses * p < 0.1, ** p < 0.05, *** p < 0.01
Table 3. Regression of the change in CDS spreads per bank on funding costs and control variables (Continued).
This table shows the regression of the change in CDS spreads on the overnight repo rate, the 30-day ABCP interest rate, the interaction term of the overnight repo rate and the 30-30-day ABCP rate, the 10 year US bond rate, the term spread measured as the difference of the 10 year US bond rate and the 3-month T-bill rate for each bank.
HSBC JPM MS PNC SCHW WFC Repo rate -1.401*** -0.595*** -1.208*** -2.175*** 0.430* -1.203*** (0.40) (0.21) (0.25) (0.37) (0.25) (0.23) ABCP rate 2.702*** 1.252*** 1.272*** -0.094 0.572*** 1.461*** (0.29) (0.15) (0.18) (0.24) (0.15) (0.17) Repo*ABCP rate 2.999** 0.873 1.488* 8.069*** -2.292** 1.332*** (1.23) (0.63) (0.76) (1.28) (0.91) (0.71) US bond 10 year -0.671*** -0.306*** -0.554*** -0.322*** 0.128*** -0.099*** (0.03) (0.02) (0.02) (0.02) (0.01) (0.02) Termspread 1.619*** -0.022*** 0.368*** 0.927*** -1.008*** 0.091 (0.10) (0.05) (0.06) (0.05) (0.04) (0.06) Constant 5.415*** 5.084*** 6.465*** 5.056*** 3.951*** 4.565*** (0.12) (0.06) (0.07) (0.08) (0.05) (0.07) R2 0.686 0.692 0.736 0.451 0.708 0.490 N 1043 1043 1043 971 878 1043
Standard errors in parentheses * p < 0.1, ** p < 0.05, *** p < 0.01
Figure 2. Sensitivity to repo rate per US bank.
This figure shows the sensitivity to the repo rate per bank, where sensitivity is given on the x-axis and the mean of CDS spreads per bank on the y-axis in the years 2009 until 2012.
Figure 3. Scatter plot of the treatment group and control group.
This figure shows the plot of the mean of CDS spreads of the treatment group and the control group defined by the exposure to the repo rate. Among the treatment group are the following banks: JPMorgan, Citigroup, Wells Fargo, Morgan Stanley, HSBC and Charles Schwab. The control group constitutes of the Bank of America, American Express, Bank of New York Mellon, PNF Financial, Capital One and Goldman Sachs.
Table 4. Difference-in-differences with the first announcement date, 26th of July 2010, as benchmark. Panel A Post announcement Treatment * Post announcement US Bond 10 yr US T-Bill 3 month Termspread LIBOR VIX SPTR Constant
Entity fixed effects Time fixed effects R-squared N +/- 1week -0.958** (0.01) 0.017 (0.02) -0.018 2.678** 0.646 (2.58) 2.783 (0.45) 0.019 (0.02) -0.040** +/- 2 weeks 132 0.645 No Yes (22.93) 38.377* (2.85) -4.083 (0.70) -1.086 (1.07) (0.60) (0.77) 0.693 (0.18) -0.224 (0.01) +/- 4 weeks -0.063*** 252 0.576 No Yes (9.94) -8.359 (1.26) 1.723 (0.24) 0.294 (0.47) 0.758 (0.26) -0.273 (0.16) (0.01) 0.411** 0.004 (0.02) 1.101*** (0.18) -1.331*** (0.32) 0.038 CDS spread 492 0.585 No Yes (3.61) 11.392*** (0.44) -0.663 (0.13) (0.19) -0.092
Table 4. Difference-in-differences with the first announcement date, 26th of July 2010, as benchmark (Continued).
This table shows the regressions of the change in CDS spreads per bank on post announcement (26th of July 2010), the interaction term treatment * post announcement, the 10-year US bond rate,
the 3-month T-bill rate, the term spread as measured as the difference between the 10-year US bond and the 2-year US bond rate, the LIBOR rate, the change in volatility measured as VIX, and the change in total return of the S&P 500. The time frame of the regression analysis takes +/- 4 weeks, +/- 2 weeks, and +/- 1 week into account. In panel B, the regression analysis is conducted as if the announcement date was 1 day before and 1 day after the actual announcement date within the event window of +/- 4 weeks, +/- 2 weeks and +/- 1 week.
* p < 0.10, ** p < 0.05, *** p < 0.01
Panel B
-1 day +1 day -1 day +1 day -1 day +1 day Post announcement -0.054*** -0.052*** -0.025 -0.042** -0.001 0.019 (0.02) (0.02) (0.02) (0.02) (0.05) (0.03) Treatment 0.006 0.002 0.021 0.015 0.014 0.016 * Post announcement (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) US Bond 10 yr 0.366** 0.366** -0.287 -0.220 -1.007* -1.368** (0.16) (0.16) (0.17) (0.18) (0.53) (0.54) US T-Bill 3 month 0.075 -0.018 0.914 0.251 1.931 3.394 (0.32) (0.32) (0.82) (0.80) (6.14) (2.69) Termspread -1.298*** -1.299*** -0.229 0.254 0.763 1.263* (0.18) (0.18) (0.27) (0.26) (0.59) (0.72) LIBOR 1.192*** 1.212 1.177*** 0.500 3.005*** 4.416** (0.20) (0.19) (0.41) (0.56) (0.76) (1.73) VIX -0.017 -0.169 0.206 0.319 -1.069 -1.717* (0.13) (0.13) (0.24) (0.24) (1.16) (0.91) SPTR -0.400 -0.811* 1.125 2.158 -4.111 -7.163* (0.43) (0.45) (1.23) (1.33) (4.60) (3.98) Constant 9.188** 12.758*** -3.749 -11.559 38.369 62.311* (3.58) (3.75) (9.72) (10.49) (37.37) (31.60)
Entity fixed effects Yes Yes Yes Yes Yes Yes
Time fixed effects No No No No No No
R-squared 0.580 0.580 0.571 0.574 0.643 0.647
N 492 492 252 252 132 132
Standard errors in parentheses * p < 0.10, ** p < 0.05, *** p < 0.01
+/- 4 weeks +/- 2 weeks +/- 1 week
Table 5. Difference-in-differences with the second announcement date, 12th of September 2010, as benchmark. Panel A Post announcement Treatment * Post announcement US Bond 10 yr US T-Bill 3 month Termspread LIBOR VIX SPTR Constant
Entity fixed effects Time fixed effects R-squared N (0.01) 0.011 (0.02) -0.077*** +/- 1week 0.480*** (2.84) -0.318 (0.09) -0.519*** (0.01) -0.038*** +/- 2 weeks 132 No 0.595 Yes (17.31) -13.712 (2.00) 1.823 (0.32) 0.073 (6.58) 16.513** (0.15) -0.188 (0.13) -0.417*** (0.01) -0.012 +/- 4 weeks 240 0.561 No Yes (6.47) 0.861 (0.74) 0.061 (0.11) -0.019 (3.05) 13.530*** (0.18) 0.300 (0.51) 0.135 (0.01) -0.026*** (0.01) -0.014 -2.396*** (0.17) -0.278 (0.39) (0.12) 0.283 CDS spread 480 0.493 No Yes (3.35) 28.027*** (0.42) -2.820*** (0.07) -0.350*** (0.32)
Table 5. Difference-in-differences with the second announcement date, 12th of September 2010, as benchmark (continued).
This table shows, in panel A, the regressions of the change in CDS spreads per bank on post announcement (12th of September 2010), the interaction term treatment * post announcement, the 10-year US bond rate, the 3-month T-bill rate, the term spread as measured as the difference between the 10-year US bond and the 2-year US bond rate, the LIBOR rate, the change in volatility measured as VIX, and the change in total return of the S&P 500. The time frame of the regression analysis takes +/- 4 weeks, +/- 2 weeks, and +/- 1 week into account. In panel B, the regression analysis is conducted as if the announcement date was 1 day before and 1 day after the actual announcement date within the event window of +/- 4 weeks, +/- 2 weeks and +/- 1 week.
Panel B
-1 day +1 day -1 day +1 day -1 day +1 day Post announcement -0.027** -0.003 -0.024*** -0.020 -0.039*** -0.044*** (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Treatment -0.025*** -0.027*** -0.013 -0.011 0.007 0.013* * Post announcement (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) US Bond 10 yr 0.138 0.120 -0.266** -0.419*** -0.399*** -0.692*** (0.11) (0.13) (0.12) (0.16) (0.09) (0.11) US T-Bill 3 month 0.354 0.313 0.202 -0.143 -1.489 -10.943*** (0.39) (0.40) (0.52) (0.52) (2.77) (2.84) Termspread -0.246 -0.273 0.149 0.277 0.420*** 0.421*** (0.16) (0.19) (0.17) (0.22) (0.15) (0.16) LIBOR -2.298*** -2.558*** 8.442*** 9.818*** -0.147 3.854 (0.31) (0.30) (2.96) (.99) (6.36) (6.34) VIX -0.326*** -0.343*** -0.051 -0.090 -0.101 -1.203*** (0.07) (0.07) (0.11) (0.11) (0.31) (0.30) SPTR -2.515*** -3.119*** -0.933 -1.282* -2.134 -6.983*** (0.41) (0.38) (0.65) (0.65) (1.45) (1.31) Constant 25.536*** 30.332*** 9.830* 12.359** 21.583 62.619*** (3.29) (3.09) (5.70) (5.76) (12.60) (11.72)
Entity fixed effects Yes Yes Yes Yes Yes Yes
Time fixed effects No No No No No No
R-squared 0.505 0.491 0.559 0.540 0.589 0.577
N 480 480 240 240 132 132
Standard errors in parentheses * p < 0.10, ** p < 0.05, *** p < 0.01
CDS spread
Table 6. Difference-in-differences with the third announcement date, 16th of December 2010, as benchmark. Panel A Post announcement Treatment * Post announcement US Bond 10 yr US T-Bill 3 month Termspread LIBOR VIX SPTR Constant
Entity fixed effects Time fixed effects R-squared N (0.00) (0.01) 0.018*** 0.030***
+/- 4 weeks +/- 2 weeks +/- 1 week
(0.43) -0.243* 0.177 (0.10) 0.186* (4.62) 18.370*** (0.07) 132 0.267 No Yes (7.38) 23.290*** (0.81) -1.760** (0.10) -0.186* (5.42) 15.512*** (0.13) -1.187 (0.45) 0.054 (0.05) 0.093* 228 0.545 No Yes (4.50) -0.056 (0.01) 0.022*** (0.01) -0.024** 3.977*** -0.073 (0.37) -0.542 (0.05) (0.21) -1.076*** (0.04) -0.299*** (0.85) 0.608 No Yes (1.61) 13.178*** CDS spread (0.07) -0.179** (0.44) -0.102 (0.05) -0.060 (0.01) 0.027*** (0.01) -0.004 456
Table 6. Difference-in-differences with the third announcement date, 16th of December 2010, as benchmark (continued).
This table shows, in panel A, the regressions of the change in CDS spreads per bank on post announcement (16th of December 2010), the interaction term treatment * post announcement, the 10-year US bond rate, the 3-month T-bill rate, the term spread as measured as the difference between the 10-year US bond and the 2-year US bond rate, the LIBOR rate, the change in volatility measured as VIX, and the change in total return of the S&P 500. The time frame of the regression analysis takes +/- 4 weeks, +/- 2 weeks, and +/- 1 week into account. In panel B, the regression analysis is conducted as if the announcement date was 1 day before and 1 day after the actual announcement date within the event window of +/- 4 weeks, +/- 2 weeks and +/- 1 week.
Panel B
-1 day +1 day -1 day +1 day -1 day +1 day
Post announcement 0.011 -0.012 0.002 -0.014* 0.004 -0.006 (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) Treatment 0.028*** 0.028*** 0.025*** 0.020*** 0.024*** 0.016*** * Post announcement (0.01) (0.01) (0.01) (0.01) (0.01) (0.01) US Bond 10 yr -0.059 -0.057 -0.064 -0.070 0.257** 0.236* (0.05) (0.05) (0.05) (0.05) (0.10 (0.12) US T-Bill 3 month 0.129 -0.292 -0.533 -0.500 -0.540 -0.103 (0.42) (0.44) (0.36) (0.37) (0.47) (0.53) Termspread -0.199*** -0.187*** -0.110 -0.067 -0.395*** -0.295* (0.07) (0.07) (0.07) (0.07) (0.14) (0.16) LIBOR 3.948*** 4.176*** 8.094** 13.854*** -2.151 1.292 (0.82) (0.84) (3.50) (2.75) (2.70) (2.50) VIX -0.295*** -0.298*** 0.056 0.084 -0.167 -0.074 (0.04) (0.04) (0.05) (0.05) (0.10) (0.13) SPTR -1.235*** -0.964*** -0.807** -0.199 -1.140 -0.024 (0.20) (0.24) (0.35) (0.46) (0.81) (1.25) Constant 14.400*** 12.298*** 8.747*** 2.168 14.728** 4.602 (1.48) (1.81) (3.31) (3.92) (6.38) (10.13)
Entity fixed effects Yes Yes Yes Yes Yes Yes
Time fixed effects No No No No No No
R-squared 0.619 0.607 0.560 0.536 0.272 0.162
N 456 456 228 228 132 132
Standard errors in parentheses * p < 0.10, ** p < 0.05, *** p < 0.01
CDS spread
Table 7. Difference-in-differences with the enactment date, 1st of January 2013, as benchmark.
Panel A Post enactment Treatment * Post enactment US Bond 10 yr US T-Bill 3 month Termspread LIBOR VIX SPTR Constant
Entity fixed effects Time fixed effects R-squared N 0.433 216 (14.03) 25.014* (1.65) Yes No (0.97) -0.616 -0.721 (0.46) -2.108 -0.176 (0.12) (3.98) -9.058** (0.53) -0.060** 0.047*** (0.01) (0.02) 0.250 46.674 (29.91) No Yes 96 0.508 (19.33) -0.476** (0.22) -4.874 (3.25) 0.553 456
+/- 4 weeks +/- 2 weeks +/- 1week
0.006 (0.06) (0.02) 0.041** -1.090 (3.90) -0.458 (1.78) 0.576 (4.33) -4.308 (0.53) 20.456*** (4.97) Yes No -5.679** (2.82) 0.131*** -1.685*** (0.04) CDS spread -0.056*** (0.01) 0.043*** (0.01) 0.012 (0.28) -0.354 (0.27) -0.314 (0.33)
Table 7. Difference-in-differences with the enactment date, 1st of January 2013, as benchmark (continued).
This table shows the regressions of the change in CDS spreads per bank on post enactment (1st of January 2013), the interaction term treatment * post enactment, the 10-year US bond rate, the 3- month T-bill rate, the term spread as measured as the difference between the 10-year US bond and the 2-year US bond rate, the LIBOR rate, the change in volatility measured as VIX, and the change in total return of the S&P 500. The time frame of the regression analysis takes +/- 4 weeks, +/- 2 weeks, and +/- 1 week into account. In panel B, the regression analysis is conducted as if the enactment date was 1 day before and 1 day after the actual enactment date within the event window of +/- 4 weeks, +/- 2 weeks and +/- 1 week.
Panel B
-1 day +1 day -1 day +1 day -1 day +1 day
Post enactment -0.041*** -0.056*** -0.040* -0.060** 0.003 0.006 (0.02) (0.01) (0.02) (0.02) (0.06) (0.06) Treatment 0.041*** 0.043*** 0.046*** 0.047*** 0.042** 0.041** * Post enactment (0.01) (0.01) (0.01) (0.01) (0.02) (0.02) US Bond 10 yr -0.022 0.012 0.041 0.250 2.035 -1.090 (0.29) (0.28) (0.49) (0.46) (5.02) (3.90) US T-Bill 3 month -0.472 -0.354 -0.302 -0.721 0.166 -0.458 (0.31) (0.27) (0.94) (0.97) (2.60) (1.78) Termspread -0.323 -0.314 -0.548 -0.616 -3.026 0.576 (0.35) (0.33) (0.59) (0.53) (6.14) (4.33) LIBOR -4.334 -5.679** -10.564** -9.058** -15.701 -4.308 (3.19) (2.82) (4.92) (3.98) (20.05) (19.33) VIX -0.141*** -0.131*** -0.290*** -0.176 -0.477** -0.476** (0.05) (0.04) (0.10) (0.12) (0.22) (0.22) SPTR -1.570*** -1.685*** -3.594** -2.108 5.999 -4.874 (0.57) (0.53) (1.41) (1.65) (4.05) (3.25) Constant 19.243*** 20.456*** 37.727*** 25.014* 58.944 46.674 (5.41) (4.97) (12.28) (14.03) (38.00) (29.91)
Entity fixed effects Yes Yes Yes Yes Yes Yes
Time fixed effects No No No No No No
R-squared 0.546 0.553 0.423 0.433 0.502 0.508
N 456 456 216 216 96 96
Standard errors in parentheses * p < 0.10, ** p < 0.05, *** p < 0.01
CDS spread
Figure 4. Plot of DiD estimator Basel III against placebo events.
The first figure shows the DiD estimator coefficient results of the regressions of the second announcement date including placebo events, whereas the second figure shows the coefficient results of the third announcement and placebo events. The last figure gives the coefficient results on the enactment date and the placebo events. The results are arranged from large to small coefficients.
