CoCos – helpful to or a hindrance for banks?
The effect of contingent convertible bonds on
a bank’s stability
Jesse A. Ruitenbeek
2
Master thesis by:
Jesse A. Ruitenbeek
University of Groningen
Faculty of Economics and Business
MSc Finance 04 June 2020 Jesse A. Ruitenbeek H.W. Mesdagstraat 75a 9718 HE +31(0)6-36149963 j.a.ruitenbeek@studen.rug.nl S2748800 First supervisor: J.V. (Jules) Tinang Nzesseu
j.v.tinang@rug.nl
Second supervisor: Dr. G.T.J. (Gijsbert) Zwart
3
Abstract
During the financial crisis of 2007 governments across the world used bail-out capital to prevent banks from defaulting. This resulted in an increase in demand by authorities for quality, transparency and consistency of core capital of European banks. The Basel III international regulatory framework was implemented to achieve this goal. To meet the requirements set by Basel III, a new type of debt instruments with equity-like characteristics came to the European banking sector. These debt instruments are called contingent convertible bonds (CoCos) and are included in the core capital of the bank.
The CoCo has two essential features. First, it can be converted to equity when a certain trigger is hit. This conversion can be done by the bank itself as the issuer or by a financial authority. Secondly, the CoCo will be subject to a write-down of its principal value after the trigger is hit. These features are meant to ease the claim of debt in times of financial distress; therefore the CoCos are included in the core capital.
However, the high-risk high-reward CoCos also increase the volatility of the common equity, which is inconsistent with the demand for better quality core capital. This research addresses this inconsistency, by looking at whether the CoCos (as a percentage of the risk-weighted assets) increase the stability of the banks (measured as the probability of default of the banks). This probability of default is defined by the Black-Scholes-Merton formula for option pricing. The effect of the percentage of CoCos to risk-weighted assets on the probability of default by banks is calculated by performing a fixed effects model regression. After controlling for the bank size, the return on assets and the Basel III leverage ratio, the results show that there is a positive relation between the percentage of CoCos to risk-weighted assets and the probability of default by banks. This is opposed to what was hypothesized. In conclusion, the amount of CoCos outstanding decrease the stability of the banks. Therefore, they contradict the intentions of better quality core capital.
Keywords: Contingent Convertible bonds (CoCos), Additional Tier 1 capital, Basel III,
4
Table of content
1. Introduction ... 5
2. Literature review ... 7
2.1 Default risk ... 7
2.2 Additional Tier 1 capital ... 7
2.3 Contingent convertible bonds ... 8
2.4 Benefits of contingent convertible bonds ... 9
2.5 Disadvantages of contingent convertible bonds ... 10
2.6 Hypothesis ... 11 2.7 Control variables ... 12 3. Methodology ... 13 3.1 Sample selection ... 13 3.2 Dependent variable ... 13 3.3 Option-based approach ... 14
3.4 Estimation of the default probability ... 14
3.5 Assumptions ... 14
3.6 Application of the Merton model ... 15
3.7 Independent variable ... 17 3.8 Control variables ... 17 3.9 Statistical model ... 18 3.10 Data ... 19 4. Results ... 21 4.1 Descriptive statistics ... 21 4.2 Regression analysis ... 22 5. Discussion ... 25 5.1 Conclusion ... 25 5.2 Managerial implications ... 25
5.3 Limitations and future research ... 26
6. References ... 27
7. Appendix ... 31
7.1 Appendix A: Capital instruments main features template ... 31
7.2 Appendix B: List of abbreviations ... 33
7.3 Appendix C: Banks included in the sample ... 33
7.4 Appendix D: CoCos included in the sample... 35
5
1. Introduction
During the financial crisis which started in 2007, several banks in Europe and the United States had to be bailed out. Almost every household was affected by this bail-out policy, as the taxpayer ended up with the costs of this expensive policy (Shy & Stenbacka, 2017). According to Atkinson, Luttrell & Rosenblum (2013) the total costs to release the tensions from the financial institutions were between 40% and 90% of American national output. This equals $50,000 - $120,000 for every household in the United States. In the European Union, similar bail-outs were used. The United Kingdom supported its banks with an average of approximately £26 billion for the five largest banks. These numbers are troublesome when passed on to the taxpayers.
Banks are important to all citizens because they have direct influence on the amount of money circulating the economy. In addition, banks are to a large extend responsible for investment activities in the economy (Köffer, 2014). The bail-outs mentioned in the previous paragraph were needed, for these activities to continue. One of the problems in the pre-financial crisis era was that banks were considered ’’too-big-to-fail’’ (Strahan, 2013). This idea has proven to be extremely dangerous. It is fuelled by the expectation that the banks will be bailed out in times of financial distress (Strahan, 2013). Following this expectations the cost of debt will decrease, since the debtors get a perception of insurance. This causes a distortion, in advantage of debt financing over equity (Strahan, 2013). As a result, banks became highly leveraged, due to the decreased cost of debt.
In addition, as shown by the financial crisis, the recent level of contagion between financial institutions can have a severe negative impact (Helwege & Zhang, 2016). The bankruptcy of Lehman’s Brothers, where the government for once chose not to intervene, is seen to be one of the key causes of the financial crisis (Johnson & Mamun, 2012). Due to the high level of financial contagion there was a so-called domino effect amongst financial institutions. This domino effect lead to high instability in the banking sector (Radoslaw, 2019).
6 gives scores to asset types based on their riskiness. Safer assets will carry the lowest weight (e.g. government bonds have zero weight).
The common equity and retained earnings are classified as the highest quality equity. They can instantly be used as loss absorbing instruments. These instruments fall in the Common Equity Tier 1 category (CET1) (Conlon, Cotter & Molyneux, 2020). This is seen as the first line of defence when financial stability is compromised. Furthermore, banks hold so called additional core Tier 1 (AT1) capital1. This additional Tier 1 capital has to satisfy a set of principles2, in order to be counted as core capital (Köffer, 2014).
The core capital consists of multiple levels that need to be satisfied. The Common Equity Tier 1 ratio (1) has to be at least 4.5%, whereas the Tier 1 ratio needs to be increased to 6% by means of either additional Tier 1 or more common equity (2). Moreover, the total capital ratio (3) has to be at least 8%. Figure 1 graphically presents the layers of the core capital.
𝐶𝐸𝑇1 𝑅𝑎𝑡𝑖𝑜 = 𝐶𝐸𝑇1 𝑅𝑊𝐴 ≥ 4,5% (1) 𝑇𝑖𝑒𝑟 1 𝑅𝑎𝑡𝑖𝑜 = 𝐶𝐸𝑇1+𝐴𝑇 1 𝑅𝑊𝐴 ≥ 6% (2) 𝑇𝑜𝑡𝑎𝑙 𝐶𝑎𝑝𝑖𝑡𝑎𝑙 𝑅𝑎𝑡𝑖𝑜 = 𝐶𝐸𝑇1+𝐴𝑇1+ 𝑇𝑖𝑒𝑟2 𝑅𝑊𝐴 ≥ 8% (3)
The additional Tier 1 capital is of high importance, since it might help stabilize the capital ratios of the banks. Therefore, the main research question will be:
What is the effect of additional Tier 1 capital on the default risk?
Figure 1: Capital requirements under Basel II and Basel III. Source: Basel III – implications for banks’ capital structure (p. 6) by T. Köffer, 2014, Anchor Academic Publishing.
1 More details about the characteristics of the additional Tier 1 capital are given in the next section. 2 There are 14 criteria set by Basel III for capital instruments to count towards additional Tier 1 capital.
7
2. Literature review
2.1 Default risk
All banks encounter default risk. According to Cakici, Chatterjee & Chen (2019) default risk is measured as the probability that a firm will default. This means that the bank will not be able to pay off existing debt. Banks need to manage this risk appropriately, so they can have a creditworthy status. The importance of a creditworthy status lies within the fact that creditworthiness comes with a reduced set of compliance rules for a bank to issue new capital (Malz, 2011). In addition, banks with high creditworthiness are considered less risky and outside investors consequently require lower risk premia. This reduces the cost of debt for banks (Hsu, Lee, Liu & Zhang, 2015).
This means that in practice the banks are required to hold a certain amount of capital to guarantee with high probability that a bank will be able to pay off existing debt. As previously mentioned, following the credit crisis there was a lack of capital to prevent banks from defaulting. This lead to a series of reforms in capital requirements. An enhanced version of the Basel II framework came into existence, to increase stability and simultaneously decrease default probabilities (Swamy, 2018). The Basel III framework was composed of the same three pillars as its predecessor. The first pillar considers minimum capital requirements, the second pillar focusses on enhanced supervisory review and lastly, pillar three regards the risk disclosure (Köffer, 2014). This paper will focus on one of these reforms in capital requirements, namely the new form of additional Tier 1 capital that came to market.
2.2 Additional Tier 1 capital
8 AT1 capital primarily consists of two types of hybrid capital, being either share-based or bond-based (Liberadzki & Liberadzki 2019). The share-bond-based category consists of preferred shares and is typically used in the United States, opposed to the usage of the bond-like contingent convertible capital (CoCos) in Europe (Liberadzki & Liberadzki 2019). While both categories were classified by legislators on the same level of subordination, there is a preference for CoCos in Europe, due to the tax deductibility of interest payments in Europe when they came to market3
(Fiordelisi et al., 2019). This research will focus on European banks and therefore solely include CoCos as AT1 capital.
The initial goal of the additional Tier 1 is to enhance stability. However, this paper includes components of additional Tier 1 capital that can increase volatility (this will be elaborated in the following section) and in its turn decrease stability. I thus aim to show the link between regulatory funding requirements and the default probability, through the associated volatility of the additional Tier 1 CoCos.
2.3 Contingent convertible bonds
As previously mentioned, the first of the three pillars of Basel III is concerned with the minimum capital requirements. The minimum capital requirements focus amongst others on leverage ratios and countercyclical capital buffers (CCyB). One of these countercyclical capital instruments are contingent convertible bonds. These CoCos are the most popular form of additional Tier 1 capital; according to Beardsworth & Glover (2019), a combined amount of 160 billion euros have been issued in European CoCo bonds. They are hybrid debt instruments, which means they are classified as debt, equal to other bonds. CoCos also resemble other regular bonds in terms of the coupon payment and par value. Also, they show similarities in terms of time to maturity (Liao, Mehdian & Rezvanian, 2017). The difference however, when a prespecified trigger occurs, CoCos are either converted to a predetermined amount of common shares, or partially or entirely written-down (De Spiegeleer, Höcht, Marquet & Schoutens, 2017).
This is a distinctive type of uncertainty that the CoCos carry compared to typical bonds, called conversion risk (Liao et al., 2017). Investors need to be compensated for the additional risk. Therefore CoCos carry a higher yield than normal bonds.
3 It should be noted that in imitation of Sweden in 2017 and the Netherlands in 2019, more countries are
9
2.4 Benefits of contingent convertible bonds
Next to these basic characteristics of CoCos, there are also some significant implications. The first one is that since CoCos are initially debt-like instruments, their coupon payments can be tax deductible and consequently increase the tax shield of the company (Ammann, Blickle & Ehmann, 2017). This tax shield reduces the cost of debt and will subsequently also downwardly affect the cost of capital, leading to cheaper funding. Hence, the cost of capital for CoCos is lower than for common equity (Pratt, Grabowski, 2008).
In addition, CoCos might reduce the effect of debt overhang (Ammann et al., 2017). Debt overhang occurs when banks experience financial difficulties. Due to existing debt, they might be unable to attract new funds, even when these funds will be used to finance positive net present value projects (Ammann et al., 2017). The claim of existing debt on the new cash flows is too strong and the following incremental value of this project is too low. To reduce this claim of debt, a firm can issue new equity to increase cash flow. However, this is unfavourable in troublesome times. When a firm needs the additional equity the most, the equity investors will be least willing to provide this capital since most will be used to pay off the debt claim (Chen, Glasserman, Nouri & Pelger, 2017). When a firm has issued CoCos, a conversion trigger will be hit in troublesome times. This automatically causes the debt claim to loosen. Besides, the threat of conversion might incentivise equity holders to invest more for two reasons. First, fear of share dilution could instigate additional equity investment to prevent actual conversion (Chen et al., 2017). Besides, the tax shield created by the CoCo bonds adds value to a firm. Equity holders benefit from this tax shield and thus are incentivised to prevent the conversion to equity and consequently loss of the tax shield (Chen et al., 2017).
Furthermore, CoCos expose bondholders to the downside risk of a bank (Liberadzki & Liberadzki, 2019). When a trigger is hit, a CoCo bond will convert to equity or be written down. This means that CoCos can be classified as a so called bail-in tool4 (Liberadzki & Liberadzki, 2019). By definition, all secured liabilities are excluded from being a bail-in tool. This complies with the criterion for CoCos to be unsecured, meaning there are no assets imposed as collateral to the bondholders.
Bail-out capital instigates banks to take risks. Thus, in the previous decade, the financial industry shifted its reliance on bail-out capital to bail-in capital (Sanchez-Roger, Oliver-Alfonso & Sanchis-Pedregosa, 2018). Moral hazard, caused by the “too-big-to-fail” conception, is seen as the cause for risk taking under implicit bail-out expectation (Sanchez-Roger et al., 2018). Therefore, CoCos are constructed to shift the expensive bail-out costs for governments (and by extension taxpayers) to cheaper bail-in capital for investors (Liao et al., 2017).
4 Bail-in tools impose the costs of financial distress on the bondholders, by cancelling the claim of the
10
2.5 Disadvantages of contingent convertible bonds
While this might all sound beneficial, there are also studies that show the negative effect of CoCos on the stability of banks. De Spiegeleer et al. (2017) suggest that while the CET1 ratio has risen over the last couple of years, so has the volatility of this ratio. According to Koziol & Lawrenz (2012), banks with CoCo bonds have a higher incentive to choose rather large volatility projects. This might increase investors value, but will also increases the overall level of volatility. High volatility increases the likelihood for shocks to be more impactful, either disadvantageous or beneficial. The increase in volatility is driven by multiple explanations. One effect explained by Berg & Kaserer (2015) is that equity holders could be better off by letting the CoCos fall directly below the conversion trigger rather than just above this trigger. The losses in asset value will be spread out across a larger number of shares, when the CoCo bonds have been converted. So, part of the losses will be imposed on the CoCo holders. This effect will occur when the number of shares after conversion received by the former CoCo holders is too low (Berg & Kaserer, 2015). Furthermore, according to Koziol & Lawrenz (2012) large institutional fixed income investors may be prohibited to hold shares. When the CoCos in their portfolio hit the equity conversion, they now are holding shares and will be mandated to sell these shares. Selling the shares when the bank is already under pressure will decrease the share price even further and consequently have a negative impact on the stability of the bank (Koziol & Lawrenz, 2012).
11 A second risk embedded in CoCos is the trigger mechanism. There are two types of triggers used in the European Union, i.e. an accounting trigger and a regulatory trigger (Derksen, Spreij & Van Wijnbergen, 2018). The accounting trigger is based on a certain accountancy ratio, for example the CET1 ratio. However, since the accounting trigger is linked to the book value of the ratios, the speed at which the conversion will occur lessens (Flannery, 2005). The damage to the firm could already be done and therefore the protection mechanism of the CoCo will fail (Flannery, 2005). In addition, contingent convertible bonds conversion can also be triggered by the banks’ (supra-)national supervisor (Köffer, 2014). The supervisor assesses a point of non-viability (PONV). This happens at the discretion of the (supra-)national supervisor and might be based on private, non-public information, imposing extra uncertainty on when a regulatory conversion will be triggered (De Spiegeleer, Marquet & Schoutens, 2018). So far, only once an authority intervened, on 6 June 2017, when Banco Popular was found unable to satisfy its debt obligations (Liberadzki & Liberadzki, 2019). The PONV was established by the ECB and the shares of Banco Popular were sold to Banco Santander for a symbolic amount of €1,-. The CoCos outstanding by Banco Popular were fully converted to shares. However, due to the worthlessness of the shares of Banco Popular, the conversion effectively nullified the value of the CoCo bonds (De Spiegeleer et al., 2018). This type of trigger level risk is complex to model, since the regulatory conversion was non-transparent. The uncertainty caused by this unobservable trigger is not only imposed on CoCo bond holders, but also on shareholders, since their shares are subject to dilution when the trigger occurs. This translates into higher CET1 volatility.
2.6 hypothesis
As shown above, there are theories that deem CoCos to be beneficial, stating it would increase bank stability and reduce the default risk attached. The flexibility it provides to maintain the primary ratios is the leading argument. By contrast, other theories suggest that the CoCos are actually achieving the opposite. Downward shocks in share prices could be following institutional investors that have to divest their interest after conversion since they can be prohibited to hold shares. These theories suggest that the relation between the effect of CoCos and the default risk of a bank is fairly ambiguous. This research aims to expose the relationship between CoCos and default risk of banks. CoCos are structured to improve stability and are supposed to more than offset the additional equity volatility it might capture. Therefore, the following hypothesis will be tested:
12
2.7 Control variables
This paper will control for multiple variables. The first control variable of this research will be the leverage ratio. The tax shield created by financial leverage is an important aspect of a company, and can help increase resilience (Hossain, Khan & Sadique, 2018). However, as stated before, banks became highly leveraged as a result of the ‘’too-big-to-fail’’ expectations. This would cause banks to be more subject to financial distress in case of losses (Straham, 2013). The leverage ratio makes no distinction between different asset quality ratios and focusses solely on firm structure (Echevarria-Icaza & Sosvilla-Rivero, 2018).
In addition, this paper controls for the return on assets. The profitability is of utmost importance for the continuity of a bank (Duan, Sun & Wang, 2012). When a bank is hit with a prolonged period without profitability, financial distress may occur. Therefore, this relation is expected to be negatively related to default probability. Another rationale to include this ratio is that CoCos can create an incentive for firms to take on higher volatility projects (Koziol & Lawrenz, 2012). When the Maximum Distributable Amount floor becomes endangered, banks might become incentivised to take on high-risk high-reward projects hoping to stay above the MDA threshold. An increase in return on assets would help banks to stay above this threshold.
13
3. Methodology
3.1 Sample selection
This research will focus on banks within the European Union, including Switzerland and the United Kingdom. Basel III was implemented in the European Union law as of 2013. It was included in the fourth Capital Requirements Directive (CRD IV). The United Kingdom was a member of the European Union up until 31 January 2020. The banks in the United Kingdom had to comply to the CRD IV the same as other banks and will therefore be included. As for Switzerland, although not subject to the CRD IV, the adaption was directed through adequacy and liquidity ordinances, implemented by the Federal Department of Finance (FDF) of Switzerland. The economy of Switzerland is tied closely to the European Union. Swiss banks will therefore be included as well.
Not all banks within the European Union will be included in the sample. First, not all banks have issued additional Tier 1 CoCo bonds. In addition, banks that only have one CoCo bond outstanding are also excluded. Multiple CoCos are preferable, in order to capture distinctive variation in the data linked to new CoCo issuances. Secondly, I will exclude banks that are not listed on an exchange. In the default probability calculations, equity volatility will be derived through the dispersion of stock prices (this will be explained in further detail in the following section). Therefore, non-listed banks are excluded. The final sample consists of 16 banks, spread across 9 countries (see appendix C). Data becomes consistently available since 2015, this will be the starting year of the sample which lasts until the fourth quarter of 2019. Due to some exclusions (e.g. ABN Amro only became listed in the fourth quarter of 2015) there is a total of 253 observations.
3.2 Dependent variable
14
3.3 Option-based approach
The default probability estimation formula is transparent in what the default probabilities depend on. The main parameters of the formula capture the essence of the ambiguity associated with the CoCos (Bharath & Shumway, 2008). The capital structure will change with the issuance of CoCos and as previously mentioned, according to De Spiegeleer et al. (2017) so does the volatility. The volatility reflects current market information. The volatility also updates continuously when new information is available in the market (Mateev, 2019). The parameters of Mertons formula also reflect the underlying information about default risk of a bank, assuming their market prices reflect the risks appropriately (Jobert et al., 2004). This method could lead to more up-to-date default probabilities, due to the timely reflection of information (Hull, 2018). In addition, the parameters can easily be obtained by looking at readily available firm data. Therefore, it seems appropriate to estimate the default probability using Mertons model, opposed to less transparent estimators. For example, the CDS spreads or bond yield spreads can also be used to estimate the default probability, however, these estimations are not based on the observable capital structure and volatility.
3.4 Estimation of the default probability
The chosen merton model estimates default probabilities. It values the equity as a residual claim over the debt. This means that the value of equity can be calculated as the value of a firm, subtracted by the face value of its debt. The residual claim will be the equity; however the market value of the firm is subject to the volatility of the assets. Therefore, the estimated equity needs to be divided by the estimated firm volatility. This results in a z-score that calculates the ratio of the difference of the firm value over its debt, divided by the volatility of assets. This is also known as the distance to default (Bharath & Shumway, 2008). It then takes this z-score and substitutes this in a cumulative density function. Doing so enables one to calculate the probability that the banks total value will exceed the face value of its debt, forecasted over a certain period of time. This research will elaborate on the model in section 3.6.
3.5 Assumptions
15
𝑑𝑉 = µ𝑉𝑑𝑡 + 𝜎𝑉 𝑉𝑑𝑊 (4)
Where V is the total value of the firm and σv is the volatility of the firm. In addition, the μ is drift rate and dW is a standard wiener process (Bharath & Shumway, 2008).
The strong form of market efficiency supports the assumption of no predictive power across time intervals. It states that all information about the firm is reflected in the stock prices. So, the stock prices (and simultaneously the volatility) will be updated when CoCos are issued. The second assumption for the Merton model to work is that the firm only has one zero-coupon bond issued. This is considered the face value of the debt (or strike price in option theory). The debt D will mature at time T. The volatility will be adjusted to match this maturity (i.e. volatility over time period T). This is a highly stylized form, which is in practice unrealistic (Bingham & Kiesel, 1998). While the theoretical values inferred from the model might deviate from the market values, it does provide practical insight in estimating the probability of default from listed firms (Jovan & Ahčan, 2017).
In addition, the Merton model assumes constant volatility in the forecasting period (Mayo, 2003), while in reality the volatility is not constant. This research aims to reduce the effect of constant volatility by updating the probability of default every quarter, when new parameter values become available. The one-year ahead default probability estimation will thus be updated quarterly.
3.6 Application of the Merton model
16
Figure 2: Probability of default in the Merton model. Source: Credit Risk Modelling Using Excel and VBA (p. 28) by G. Löffler & P. Posch, 2011. John Wiley & Sons.
This model effectively requires two equations (5 & 8) to hold simultaneously. The first one (5) is the value of the equity, calculated as the expected value of the firm’s assets (V) minus the non-default discount value of debt (D). r Is the risk-free rate. Since the focus will be on European banks, the German government bonds will be considered the risk-free rate. Other parameters of the function are the N(dx) variables. N(d1) calculates the expected future firm value, whereas outcomes that are lower than the value of debt are counted as zero. This is based on a cumulative normal distribution function. N(d2) calculates the probability of the call option being exercised. These probabilities can be derived from formula (6) and (7). Time T is when the bond matures (i.e. the one-year ahead estimation of the default probability). At last, one needs to find the asset volatility of the bank. By Ito’s Lemma, the derivation of the firm’s equity can be inferred from formula (8). Equations (5) and (8) provide a pair of equations that can be solved simultaneously for the firm’s volatility and the total value of the firm.
17 To practically apply this formula, one should first look at the volatility of the equity. This research calculates the equity volatility as a banks’ stock variation around its mean. It does so by looking at adjusted closing share prices. The adjusted prices are used to eliminate the effect of corporate actions (e.g. dividend payments) which causes disturbance around the true share value (Whitworth & Zhang, 2010). Furthermore, the total market value of equity is also needed to calculate the total firm value. Hereafter, one should match T to the maturity. This research will be using quarterly data, but estimates one-year ahead probabilities. Therefore, T will be set equal to one year. As the risk-free rate I will use the German government bonds. As stated before, this research focusses on European banks, because they are primarily CoCo bond issuers. All parameters are obtainable via publicly traded data, except for the market value of the bank V and the asset volatility of the bank 𝜎𝑉. By simultaneously solving both equation (5) and (8), one can obtain these unobserved variables.
Now, all the variables can be obtained. The N(d2) probability function in option pricing is used to calculate the probability that a call option will be exercised (Delianedis & Geske, 2003). Equivalently, N(d2) will be the probability that the value of the assets is larger than the value of debt, so there is no default. The opposite is the probability of default, given by 1 – N(d2) = N(-d2).
3.7 Independent variable
The independent variable can be computed more straight forward than the dependent variable. I will use the total amount of CoCos outstanding against the risk-weighted assets of the bank. As stated before, CoCos were being issued following Basel III when there was a demand for extra stability. A CoCo is used in the calculations of the core capital ratios, where it was initially designed for. Therefore, I will use this component of the core capital ratio. The ratio will be calculated as following:
Percentage of additional Tier 1 CoCos to RWA = 𝑇𝑜𝑡𝑎𝑙 𝑎𝑚𝑜𝑢𝑛𝑡 𝑜𝑓 𝑎𝑑𝑑𝑖𝑡𝑖𝑜𝑛𝑎𝑙 𝑇𝑖𝑒𝑟 1 𝐶𝑜𝐶𝑜𝑠 𝑖𝑠𝑠𝑢𝑒𝑑
𝑇𝑜𝑡𝑎𝑙 𝑎𝑚𝑜𝑢𝑛𝑡 𝑜𝑓 𝑟𝑖𝑠𝑘−𝑤𝑒𝑖𝑔ℎ𝑡𝑒𝑑 𝑎𝑠𝑠𝑒𝑡𝑠
3.8 Control variables
The first control variable will be the leverage ratio. As previously mentioned, there has been found a negative relation in the literature between the leverage ratio and the default probability. The leverage ratio will be measured according to the Basel III standards:
Leverage ratio = 𝑇𝑖𝑒𝑟 1 𝑐𝑎𝑝𝑖𝑡𝑎𝑙
𝑇𝑜𝑡𝑎𝑙 𝑒𝑥𝑝𝑜𝑠𝑢𝑟𝑒𝑠
18 assets and total exposures comes from off-balance sheet adjustments (e.g. differences in accounting values and fair values of derivative contracts due to movements of the underlying asset) (EY, 2014).
Secondly, The return on assets (ROA) measures profitability. High profitability is beneficial to a bank. It measures how effectively the bank makes profit. Profit directly influences the retained earnings, which is part of the Common Equity Tier 1. Therefore, higher ROA will reduce the default probability, as it can be used to absorb losses. The return on assets will be calculated as:
Return on assets (ROA) = 𝑁𝑒𝑡 𝑖𝑛𝑐𝑜𝑚𝑒
𝑇𝑜𝑡𝑎𝑙 𝑎𝑠𝑠𝑒𝑡𝑠
The third control variable will be bank size. As stated in the literature review, there is a negative relation between the bank size and the default probability. According to Altunbas, Gambacorta & Marques-Ibanez (2010) the most common way to measure size is by taking the logarithm of total assets. By taking the natural logarithm of total assets, this research reduces the effect of outliers, by compressing the data to become more normalized. Moreover, it controls for heteroskedasticity.
3.9 Statistical model
With both the dependent variable and the independent variables quantified, I can perform a regression to test the hypothesis. I will use a fixed effects model, to control for the individual banks. This allows them to have their own intercept, which enables this research to control for unobserved (time-invariant) bank specific characteristics that might drive their default probability. The regression will have the following form, based on the information presented above:
Probability of default = β0 + β1 * CoCos + β2 * Leverage + β3 * ROA + β4 * Size +
β5 * Yeardummies + ε
19
3.10 Data
The data will come from different sources, due to the different nature of the data. The primary source of data used will be Thomson Reuters Eikon database. This database provides up to date as well as historic data on company specific variables. It provides information about its capital structure i.e. all capital instruments. In addition, one can search for additional Tier 1 capital. Here, all Tier 1 bonds are displayed. Information about whether the bond classifies as a CoCo bond is also presented. However, this classification is flawed, since it does not take into account all Basel III requirements. Therefore, when not complying to all requirements, it might not be recognised as a post-transitional Basel III CoCo bond. To ensure all the bonds in the sample are counted as post-transitional Basel III CoCo bonds, one should cross-check the CoCo bonds with the capital instruments main features disclosure provided by the banks. Looking at the corresponding ISIN codes in both sources simplifies the process. The capital instruments main features template is a template provided by the European Banking Authority (EBA) in line with regulation (EU) No 1423/2013 and requires all banks to disclose information on common equity, additional Tier 1 and Tier 2 capital (see appendix A). It is one of the results of the aforementioned strive to enhance transparency. One disclosure item (line 5, appendix A) provides the information on the post-transitional Basel III eligibility. Unfortunately, distinction between CoCo bonds and other hybrid instruments is not always clear. Therefore, these sources need to be combined, to ensure the classification as post-transitional Basel III additional Tier 1 CoCo bonds (see appendix D).
20
Table 1
Variables description.
All the variables that are used in this research are summarized in this overview.
Variable Name Type Measurement
Probability of default
PoD Dependent
variable
The probability of default is calculated using Black-Scholes-Merton option pricing formula. It is used to calculate the probability that the value of the assets does not exceed the value of the debt, subject to a set of simplifying assumptions.
Percentage of Tier 1 CoCos to RWA
P_COCO Variable of interest
The variable of interest measures whether the countercyclical instruments have the desirable effect. It is calculated as the amount of Tier 1 CoCos outstanding to the RWA of a bank.
Size Log_Assets Control variable The first control variable will be the bank size. To reduce the effect of outliers and skewness in the distribution, this research takes the natural log of the total assets. Return on Assets ROA Control variable Furthermore, the return on assets is
controlled for. It measures the profitability by calculating the ratio of net income over the total assets.
21
4. Results
4.1 Descriptive statistics
The final sample consists of 16 banks, observed over quarterly periods during a period of 5 years (i.e. 2015 until 2019). This results in 320 possible observations. However, not all data was available during this period. Dropping all missing data points resulted in an unbalanced panel dataset of a total of 253 observations.
Table 2 shows that the average percentage of CoCos outstanding is approximately 1.74% of the bank’s risk-weighted assets. However, only 1.5% can be included in the Tier 1 capital ratio. The excess cannot be included in the core capital. The probability of default has a relatively high standard deviation and is highly positively skewed. Therefore, the mean is subject to the effect of outliers on the right tail. In order to reduce the effect of outliers, the probability of default is winsorized (henceforth abbreviated by PoD_W). This is done at a 1% interval (i.e. approximately ±3σ). Since the effect of outliers is most drastic in the probability of default, only this variable will be winsorized.
Table 2
Summary statistics of variables.
The table presents the summary statistics. It shows the values of the variables. Starting with the winsorized probability of default (PoD_W) as the dependent variable. The second line presents the variable of interest; the percentage of CoCos to risk-weighted assets (P_CoCo). Below, the control variables are listed. The natural logarithm (Log_Assets), return on assets (ROA) and the Basel III leverage ratio (LEV_RAT) are displayed.
Observations Mean Std. Dev. Min Max
PoD_W 253 0.0020 0.0048 9.20e-12 0.0265
P_CoCo 253 0.0174 0.0076 0.0012 0.0406
Log_Assets 253 27.6308 0.5932 26.2541 28.5514
ROA 253 0.0035 0.0035 -0.0139 0.0106
LEV_RAT 253 0.0481 0.0080 0.0340 0.0690
22 calculate a variance inflation factor (VIF) to quantify the multicollinearity of the model. When the VIF exceeds a threshold of 10, there could be problems with multicollinearity (Hair, Black, Babin & Anderson 2014). The VIF value of the model show a mean value of 1.93 (a maximum value of 2.69). In conclusion, due to the low correlations and low VIF value, there are no problems with multicollinearity.
Table 3
Correlation Matrix.
The table presents the correlation matrix of all the variables in the model. The headers from left to right are: the winsorized probability of default (PoD_W) as the dependent variable, the percentage of CoCos to risk-weighted assets (P_CoCo), the natural logarithm (Log_Assets), return on assets (ROA) and the Basel III leverage ratio (LEV_RAT) as independent variables.
PoD_W P_CoCo Log_Assets ROA LEV_RAT
PoD_W 1.0000 P_CoCo -0.1583 1.0000 Log_Assets 0.1181 -0.2793 1.0000 ROA -0.4495 0.0082 -0.3619 1.0000 LEV_RAT -0.1328 0.1365 -0.3323 0.3845 1.0000 4.2 Regression Analysis
23 correlates to the error term, which can arise when variables are omitted. This leads to inconsistent and biased results (Roberts & Whited, 2013). As mentioned earlier, the fixed effects model accounts for unobserved (time-invariant) bank specific characteristics i.e. omitted variables. Thus, the fixed effects model will provide consistent results, assuming that the idiosyncratic errors and the explanatory variables are uncorrelated.
To test whether the hypothesis holds, this research performed a fixed effects regression. Doing so enables one to look at the direction of the coefficient, the strength of the coefficient as well as whether it is statistically different from zero. Table 4 displays three regression outputs. The first fixed effects model shows the effect on the dependent variable when only the variable of interest is included. Secondly, I include a pooled OLS regression to enable comparability. The final column presents the full model, a fixed effects regression with all variables included. This column is used to make the conclusions about the hypothesis and leads to the following results. The results from model 3 show a significant relation between the percentage of CoCos to risk-weighted assets and the probability of default by banks. It shows a coefficient of 0.2314 and the corresponding p-value is 0.014. Thus, it is statistically different from zero at the 5% level. This means that with a one percent increase in CoCo bonds to risk-weighted assets, the default probability increases with approximately 0.2314%. However, the direction is opposed to the hypothesis, therefore the results do not support the hypothesis. The hypothesis stated that there will be a negative relation between the percentage of CoCo bonds outstanding and the probability of default by European banks. The adjusted R-squared is 0.5041 which is a moderately strong explanation of the variation caused by the independent variables on the dependent variable. In addition, the return on assets and the leverage ratio are both significant at the 5% level. They show a coefficient of -0.6511 and -0.2475 respectively. Therefore, the two control variables are negatively related to the probability of default by banks. This is in line with the aforementioned theory on control variables. The logarithm of total assets shows a different sign as to what was expected by theory. However, this independent variable is insignificant and therefore there is insufficient evidence to conclude that this effect is statistically different form zero.
24
Table 4
Regression results.
The table presents the results of the three regression models. It displays the coefficients and robust standard errors of the independent variables on the dependent variable. The effect of the percentage of CoCos to risk-weighted assets (P_CoCo) as the variable of interest and the effect of the control variables: the natural logarithm (Log_Assets), return on assets (ROA) and the Basel III leverage ratio (LEV_RAT). Model (1) is a fixed effects model, which solely focusses on the percentage of CoCos to risk-weighted assets as an explanatory variable. Model (2) is a pooled OLS regression, included for comparability and displays all the variables. Model (3) displays the fixed effects model with all variables and is used to make conclusions. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 show the significance level, respectively.
(1) (2) (3)
VARIABLES Model 1 Model 2 Model 3
25
5. Discussion
5.1 Conclusion
The findings of this research show that there is a positive relation between the percentage of additional Tier 1 CoCos outstanding to risk-weighted assets and the probability of default. This is opposed to the hypothesis of this research. More important, it is opposed to why the CoCos were initially created. They were originally created to increase a bank’s resilience. However, the data shows that CoCo bonds actually decrease a bank’s resilience. This effect is troublesome since an increasing number of CoCo bonds have been issued by banks to count as Tier 1 capital. The mean of 1.74% CoCos outstanding show that banks typically rely more on CoCo bonds than Basel III even allows them to as regulatory capital.
As stated above, Basel III was implemented with the main goal to increase the quality, transparency and consistency of Tier 1 capital. Doing so would enhance a bank’s resilience. The CoCo bonds were thought of to achieve this goal. However, the results show that the opposite is true. As stated in the literature review, the issuance of CoCo bonds actually increases the CET1 volatility and therefore reduces its power to absorb losses. The CoCos raised the level of core capital, but simultaneously decreased the quality of it.
Furthermore, the additional Tier 1 bonds diverge from the goal of transparency. While their design might be communicated openly, there is one major flaw with regard to transparency. As previously stated, the CoCos are dependent on multiple triggers, such as the regulatory trigger. The aforementioned Banco Popular contingent convertible bonds were subject to a regulatory trigger being hit. This trigger was imposed by the ECB as the (supra-)national supervisor of the bank. The supervisor has the discretion to assess the PONV, which is based on private information. Thus, the regulatory trigger being hit is not discernible by the bondholders. This causes distorted expectations and uncertainty.
5.2 Managerial implications
26 Following Basel III, the CoCos were counted as the additional Tier 1 capital. Basel III was formed to strengthen the core capital. As the results from this research show, the opposite is achieved. CoCos do not strengthen the core capital and increase the probability of default by banks. Thus, from a regulating point of view, Basel V (as Basel IV is already accepted) should not use CoCo bonds to strengthen core capital.
5.3 Limitations and future research
The following section focusses on the limitations encountered when conducting this research. First, since CoCos are a relatively new instrument, their intentions of countercyclicality in economic downturn have been tested sporadically. The contingent convertibles market started to grow in 2014 (Echevarria-Icaza & Sosvilla-Rivero, 2018), which is followed by years of economic prosperity in the European Union & Switzerland. During this period only one bank wrote down its CoCos. This was reported when the EU-authorities declared the Banco Popular as failing (Liberadzki & Liberadzki 2019). However, this was not a test for countercyclicality as the overall market was still considered strong. It would therefore be interesting to see how banks with CoCos perform in troublesome times.
Furthermore, this research encountered some issues with the data collection. Basel III was introduced in 2013 and is scheduled to be fully phased in by 01 January 2022. This resulted in inconsistency in data availability. While gathering the data, gaps remained as banks did not provide information in a regular fashion. Datapoints had to be dropped, as they could not be reliably filled in. This resulted in an unbalanced panel dataset with fewer datapoints than was intended.
The focus of the research is solely on larger, listed banks. The reasoning is that these banks were amongst the first to have access to the CoCo market. Also, inference on the probability of default was made by looking at the share prices. This excludes the smaller banks or banks that are not listed. However, as the CoCo market grew bigger, the smaller and non-listed banks got access to the contingent convertible bonds market as well. This research has no external validity regarding the non-listed banks, as the probability of default calculation is completely inappropriate for the non-listed banks.
27
6.
References
Altunbas, Y., Gambacorta, L., & Marques-Ibanez, D. 2010. Bank risk and monetary policy. Journal of Financial Stability, 6(3), 121-129. doi:10.1016/j.jfs.2009.07.001 Ammann, M., Blickle, K., & Ehmann, C. 2017. Announcement effects of contingent
convertible securities: Evidence from the global banking industry. European Financial Management, 23(1), 127-152. doi:10.1111/eufm.12092
Atkinson, T., Luttrell, D., & Rosenblum, H. 2013 How bad was it? The costs and consequences of the 2007-09 financial crisis. Federal Reserve Bank of Dallas, Staff Paper no 20, july Beardsworth, T. & Glover, J. 2019, februari 18. Contingent convertibles. Retrieved from
https://www.bloomberg.com/quicktake/contingent-convertible-bonds
Berg, T., & Kaserer, C. 2015. Does contingent capital induce excessive risk-taking? Journal of Financial Intermediation, 24(3), 356-385. doi:10.1016/j.jfi.2014.11.002
Bharath, S., & Shumway, T. 2008. Forecasting default with the merton distance to default model. Review of Financial Studies, 21(3), 1339-1369. doi:10.1093/rfs/hhn044
Bingham, N. H., & Kiesel, R. 1998. Risk-neutral valuation : pricing and hedging of financial derivatives (Ser. Springer finance). Springer.
Cakici, N., Chatterjee, S., & Chen, R. 2019. Default risk and cross section of returns. Journal of Risk and Financial Management, 12(2). doi:10.3390/jrfm12020095
Calabrese, R., & Giudici, P. 2015. Estimating bank default with generalised extreme value regression models. Journal of the Operational Research Society, 66(11), 1783-1792. doi:10.1057/jors.2014.106
Chen, N., Glasserman, P., Nouri, B., & Pelger, M. 2017. Contingent capital, tail risk, and debt-induced collapse. The Review of Financial Studies, 30(11), 3921-3969. doi:10.1093/rfs/hhx067
Conlon, T., Cotter, J., & Molyneux, P. 2020. Beyond common equity: The influence of secondary capital on bank insolvency risk. Journal of Financial Stability. doi:10.1016/j.jfs.2020.100732
De Spiegeleer, J., Höcht, S., Marquet, I., & Schoutens, W. 2017. Coco bonds and implied cet1 volatility. Quantitative Finance, 17(6), 813-824. doi:10.1080/14697688.2016.1249019 De Spiegeleer, J., Marquet, I., & Schoutens, W. 2018. The risk management of contingent
convertible (coco) bonds (Ser. Springerbriefs in finance). Springer. https://doi.org/10.1007/978-3-030-01824-5
Delianedis, G., and Geske, R. 2003. Credit Risk and Risk Neutral Default Probabilities: Information about Rating Migrations and Defaults. Available online: https://papers.ssrn.com/sol3/papers.cfm?abstract_ id=424301&download=yes
Derksen, M., Spreij, P., Wijnbergen, S. van, & Centre for Economic Policy Research (Great Britain). 2018. Accounting noise and the pricing of cocos (Ser. Discussion paper series, financial economics, no. 12869). Centre for Economic Policy Research. http://www.cepr.org/active/publications/discussion_papers/dp.php?dpno=12869.
28 Echevarria-Icaza, V., & Sosvilla-Rivero, S. 2018. Systemic banks, capital composition, and coco bonds issuance: The effects on bank risk. International Journal of Finance & Economics, 23(2), 122-133. doi:10.1002/ijfe.1607
EY. 2014. Applying IFRS – Credit valuation adjustments for derivative contracts. Retrieved from https://www.ey.com/Publication/vwLUAssets/EY-credit-valuation-adjustments-for-derivative-contracts/$FILE/EY-Applying-FV-April-2014.pdf
Fiordelisi, F., Pennacchi, G., & Ricci, O. 2019. Are contingent convertibles going-concern capital? Journal of Financial Intermediation. doi:10.1016/j.jfi.2019.03.007
G. Kiewiet, I. van Lelyveld, and S. van Wijnbergen. 2017. Contingent convertibles: can the market handle them? Working paper DP12359, CEPR, 2017. URL www.cepr.org/active/publications/ discussion_papers/dp.php?dpno=12359.
Hamilton, B., & Nickerson, J. 2003. Correcting for endogeneity in strategic management research. Strategic Organization, 1(1), 51-78. Retrieved May 31, 2020, from www.jstor.org/stable/23727575
Helwege, J., & Zhang, G. 2016. Financial firm bankruptcy and contagion. Review of Finance, 20(4), 1321-1362. doi:10.1093/rof/rfv045
Hossain, M., Khan, M., & Sadique, M. 2018. Basel iii and perceived resilience of banks in the
brics economies. Applied Economics, 50(19), 2133-2146.
doi:10.1080/00036846.2017.1391999
Hsu, P., Lee, H., Liu, A., & Zhang, Z. 2015. Corporate innovation, default risk, and bond pricing. Journal of Corporate Finance, 35, 329-344. doi:10.1016/j.jcorpfin.2015.09.005 Hull, J. 2018. Risk Management and Financial Institutions (fifth edition). Hoboken, New
Jersey: John Wiley & Sons.
Jang, H., Na, Y., & Zheng, H. 2018. Contingent convertible bonds with the default risk premium. International Review of Financial Analysis, 59, 77-93. doi:10.1016/j.irfa.2018.07.003
Jobert, A., Kong, J., & Chan-Lau, J. A. 2004. An option-based approach to bank vulnerabilities
in emerging markets. Imf Working Papers, 04(33), 1–1.
https://doi.org/10.5089/9781451845211.001
Johnson, M., & Mamun, A. 2012. The failure of lehman brothers and its impact on other financial institutions. Applied Financial Economics, 22(5), 375-385. doi:10.1080/09603107.2011.613762
Joseph F. Hair Jr. William C. Black Barry J. Babin & Rolph E. Anderson. 2014. Multivariate data analysis. Retrieved from https://is.muni.cz/el/1423/podzim2017/ PSY028/um/_Hair_-_Multivariate_data_analysis_7th_revised.pdf
Jovan, M., & Ahčan, A. 2017. Default prediction with the merton-type structural model based on the nig lévy process. Journal of Computational and Applied Mathematics, 311, 414-422. doi:10.1016/j.cam.2016.08.007
Köffer, T. 2014. Basel iii, implications for banks' capital structure what happens with hybrid capital instruments? Hamburg, Germany: Anchor Academic Publishing. (2014). Retrieved February 16, 2020,
29 Liao, Q., Mehdian, S., & Rezvanian, R. 2017. An examination of investors’ reaction to the announcement of coco bonds issuance: A global outlook. Finance Research Letters, 22, 58-65. doi:10.1016/j.frl.2016.12.034
Liberadzki, M., & Liberadzki, K. 2019. Contingent convertible bonds, corporate hybrid securities and preferred shares : Instruments, regulation, management. Cham: Palgrave Macmillan. doi:10.1007/978-3-319-92501-1
Liebenberg, F., van, V. G., & Heymans, A. 2017. Contingent convertible bonds as countercyclical capital measures. South African Journal of Economic and Management Sciences, 20(1). https://doi.org/10.4102/sajems.v20i1.1600
Löffler, G. & Posch, P. N. 2011. Credit risk modeling using Excel and VBA (2nd ed., Ser. Wiley Finance). Wiley Online Library All Obooks.
M.J. Flannery. 2005. No pain, no gain: Effecting market discipline via ”reverse convertible debentures”. In Hal S. Scott, editor, Capital Adequacy beyond Basel: Banking, Securities, and Insurance, pages 171–196. Oxford University Press,
Malz, A. 2011. Financial risk management : Models, history, and institutions (Wiley finance series). Hoboken, N.J.: Wiley. (2011). Retrieved March 26, 2020, from Ebook Central Academix Complete
Mateev, M. 2019. Volatility relation between credit default swap and stock market: New empirical tests. Journal of Economics and Finance, 43(4), 681-712. doi:10.1007/s12197-018-9467-5
Mayo, A. 2003. Derivatives in financial markets with stochastic volatility (jean-pierre fouque, george papanicolaou, and k. ronnie sircar). Siam Review, 45, 612–613.
Merton, R. 1974. On the pricing of corporate debt: The risk structure of interest rates. The Journal of Finance, 29(2), 449-470
Pratt, S., & Grabowski, R. 2008. Cost of capital : Applications and examples (5th ed., [wiley finance]) [5th ed.]. Hoboken, New Jersey: Wiley. (2014). Retrieved April 3, 2020, Ebook Central Academix Complete
Radoslaw, M. 2019. Impact of the collapse of the lehman brothers bank and the 2008 financial crisis on global economic security. Scientific Journal of the Military University of Land Forces, 191(1), 149-158. doi:10.5604/01.3001.0013.2405
Ramirez, J. 2017. Handbook of basel iii capital : Enhancing bank capital in practice. Chichester, West Sussex, United Kingdom: John Wiley and Sons. (2017). Retrieved April 1, 2020, from Wiley Online Library All Obooks.
Roberts, M. R., & Whited, T. M. 2013. Endogeneity in empirical corporate finance. Handbook of the Economics of Finance, 2(Pa), 493–572. https://doi.org/10.1016/B978-0-44-453594-8.00007-0
Sanchez-Roger, M., Oliver-Alfonso, M. D., & Sanchis-Pedregosa, C. 2018. Bail-in: A sustainable mechanism for rescuing banks. Sustainability, 10(10). doi:10.3390/su10103789
Shy, O., & Stenbacka, R. 2017. An overlapping generations model of taxpayer bailouts of banks. Journal of Financial Stability, 33, 71-80. doi:10.1016/j.jfs.2017.10.003
30 Suganthi, K., Jayalalitha, G. 2019. Geometric brownian motion in stock prices. 1377(1).
doi:10.1088/1742-6596/1377/1/012016
Swamy, V. 2018. Basel iii capital regulations and bank profitability. Review of Financial Economics, 36(4), 307-320. doi:10.1002/rfe.1023
31
7.
Appendices
32
33
7.2 Appendix B: List of abbreviations
Appendix B
List of abbreviations.
All the abbreviations used in this research are displayed here. The left column shows the abbreviations and the right column presents the full name.
Abbreviation Fully written
AT1 Additional Tier 1
CCyB Countercyclical Capital Buffer
CDS Credit Default Swap
CET1 Common Equity Tier 1
CoCos Contingent convertibles
CRD IV Capital Requirements Directives IV
EBA European Banking Authority
FDF Federal Department of Finance
MDA Maximum Distributable Amount
PONV Point Of Non-Viability
ROA Return On Assets
RWA Risk-Weighted Assets
VIF Variance Inflation Factor
7.3 Appendix C: Banks included in the sample
Appendix C
List of banks in the sample.
All the banks included in the sample are displayed here. It shows the name of the bank, the corresponding country, the exchange where the bank is listed as well as the Reuters Instrument Code (RIC).
BANK COUNTRY EXCHANGE RIC
KBC Group Belgium Euronext Brussels KBC.BR
Danske Bank Denmark OMX Nordic
Exchange Copenhagen
DANSKE.CO
BNP Paribas France Euronext Paris BNPP.PA
Credit Agricole France Euronext Paris CAGR.SA
Société Générale France Euronext Paris SOGN.PA
Deutsche Bank Germany Deutsche Börse Xetra DBKGn.DE
UniCredit Italy Milan stock
exchange
CRDI.MI
ABN AMRO Group Netherlands Euronext
Amsterdam
34
BANK COUNTRY EXCHANGE RIC
ING Group Netherlands Euronext Amsterdam INGA.AS
Banco Bilbao Vizacaya Argentaria
Spain BME Spanish
Exchange
BBVA.MC
Banco Santander Spain BME Spanish
Exchange
SAN.MC Credit Suisse Group Switzerland SIX Swiss Exchange CSGN.S
Barclays United Kingdom London Stock
Exchange
BARC.L
HSBC Holdings United Kingdom London Stock
Exchange
HSBA.L Lloyds Banking Group United Kingdom London Stock
Exchange
LLOY.L Royal Bank of Scotland
Group
United Kingdom London Stock Exchange
35
Appendix D: CoCos included in the sample
Appendix D
List of CoCos in the sample.
All the CoCos included in the sample are presented here. It shows the International Securities Identification Number (ISIN) of the CoCo as well as the Reuters Instrument Code (RIC) of the issuing bank.
ISIN CoCo Bank RIC ISIN CoCo Bank RIC ISIN CoCo Bank RIC
BE0002463389 KBC.BR US251525AN16 DBKGn.DE CH0494734384 CSGN.S
BE0002592708 KBC.BR XS1539597499 CRDI.MI US06738EAA38 BARC.L
BE0002638196 KBC.BR XS1619015719 CRDI.MI XS1002801758 BARC.L
XS1044578273 DANSKE.CO XS1739839998 CRDI.MI XS1068561098 BARC.L
XS1190987427 DANSKE.CO XS1963834251 CRDI.MI US06738EAB11 BARC.L
DK0030386610 DANSKE.CO XS1278718686 ABNd.AS XS1068574828 BARC.L
XS1586367945 DANSKE.CO XS1693822634 ABNd.AS XS1274156097 BARC.L
XS1825417535 DANSKE.CO US456837AE3 INGA.AS XS1481041587 BARC.L
XS1247508903 BNPP.PA US456837AF0 INGA.AS XS1571333811 BARC.L
US05565AAN37 BNPP.PA XS1497755360 INGA.AS XS1658012023 BARC.L
US05565AAQ67 BNPP.PA XS1956051145 INGA.AS US06738EBA29 BARC.L
US05565ACA97 BNPP.PA US456837AR44 INGA.AS US404280AR04 HSBA.L
US05565ADW09 BNPP.PA XS0926832907 BBVA.MC US404280AS86 HSBA.L
US05565AGF49 BNPP.PA XS1033661866 BBVA.MC US404280AT69 HSBA.L
US05565AHN63 BNPP.PA XS1190663952 BBVA.MC US404280BC26 HSBA.L
FR0013433257 BNPP.PA XS1394911496 BBVA.MC US404280BL25 HSBA.L
US225313AD75 BNPP.PA XS1619422865 BBVA.MC US404280BN80 HSBA.L
XS1055037177 CAGR.SA US05946KAF84 BBVA.MC US404280BP39 HSBA.L
XS1055037920 CAGR.SA ES0813211002 BBVA.MC XS1111123987 HSBA.L
US225313AE58 CAGR.SA ES0813211010 BBVA.MC XS1298431104 HSBA.L
US225313AJ46 CAGR.SA US05946KAG67 BBVA.MC XS1624509300 HSBA.L
US225313AL91 CAGR.SA XS1043535092 SAN.MC XS1640903701 HSBA.L
XS0867614595 CAGR.SA XS1066553329 SAN.MC XS1882693036 HSBA.L
US83367TBF57 / USF8586CRW49
SOGN.PA XS1107291541 SAN.MC XS1884698256 HSBA.L
XS0867620725 SOGN.PA XS1692931121 SAN.MC XS1043545059 LLOY.L
US83367TBH14 / USF8586CXG25
SOGN.PA XS1602466424 SAN.MC XS1043552188 LLOY.L
US83368JFA34 / USF43628B413
SOGN.PA US22546DAB29 CSGN.S XS1043552261 LLOY.L
US83368JKG49 / USF43628C734
36
US83367TBU25 / SF8586CBQ45
SOGN.PA CH0352765157 CSGN.S US539439AG42 LLOY.L
US83367TBV08 / USF84914CU62
SOGN.PA CH0360172719 CSGN.S US539439AU36 LLOY.L
FR0013414810 SOGN.PA US225401AJ72 CSGN.S US53944YAJ29 LLOY.L
FR0013446424 SOGN.PA CH0428194226 CSGN.S XS2080995405 LLOY.L
DE000DB7XHP3 DBKGn.DE US225401AK46 CSGN.S US780099CK11 RBS.L
XS1071551474 DBKGn.DE CH0482172324 CSGN.S US780099CJ48 RBS.L
XS1071551391 DBKGn.DE US225401AL29 CSGN.S US780097BB64 RBS.L
7.5 APPENDIX E: Results from Hausman test and Breusch Pagan LM test
Appendix E.1
The Hausman test.
This table presents the result from the Hausman test. The test is conducted to evaluate whether this research should use the fixed effects model or the random effects model.
Coefficient (b) Fixed Coefficient (B) Random (b-B) difference Sqrt (diag (V_b -V_B)) S.E. P_CoCo 0.2314 0.0356 0.1958 .0441 Log_Assets 0.0048 -0.0002 .0050 .0037 ROA -0.6511 -0.5188 -.1323 .0742 LEV_RAT -0.2475 -0.0496 -.1979 .0705 Dum_16 0.0070 0.0072 -.0002 .0001 Dum_17 0.0011 0.0016 -.0005 .0004 Dum_18 -0.0008 0.0003 -.0011 .0004 Dum_19 -0.0005 0.0006 -.0012 .0004
b = consistent under H0 and Ha; obtained from xtreg
B = inconsistent under Ha, efficient under H0; obtained from xtreg
Test H0: difference in coefficient not systematic
Chi2(8) = (b-B)’[(V_b-V_B)^(-1)](b-B) = 26,52
Prob>Chi2 = 0.0009
37
Appendix E.2
The Breusch and Pagan Lagrange Multiplier test.
This table presents the result from the Breusch and Pagan Lagrange Multiplier test. The test is conducted to evaluate whether this research should use the pooled OLS model or the random effects model.
Var sd = sqrt(Var)
PoD_W 0.0000 0.0048
e 0.0000 0.0033
u 1.43e-06 0.0012
PoD_W[id,t] = Xb + u[id] + e[id,t]
Test: Var(u) = 0
Chibar2(01) = 6.65
Prob>Chibar2 = 0.0050
38
7.6 Appendix F: Stata do file code
//Import Data clear
cd "C:\Users\jesse\OneDrive\Bureaublad\studie\msc finance\Masterscriptie\Data\Statr sysuse CoCo_Stata
//Clean Data
rename LeverageratiobaselIII LEV_RAT rename Logtotalassets Log_Assets rename DP PoD
rename CoCo P_CoCo
drop if Log_Assets ==0 drop if P_CoCo ==0 drop if P_CoCo ==. rename ID rename Quarter t drop if PoD ==. drop if ROA ==. drop if LEV_RAT ==. //Winsorize Data
sktest PoD Log_Assets, noadjust
winsor PoD, generate (PoD_W) p(0.01) sktest PoD_W Log_Assets, noadjust
39 gen dum_17 = (Year==2017)
gen dum_18 = (Year==2018) gen dum_19 = (Year==2019)
//Define global variables global id
global t
global ylist PoD_W
global xlist P_CoCo Log_Assets ROA LEV_RAT global dumlist dum_16 dum_17 dum_18 dum_19
//Overview
describe $id $t $ylist $xlist summarize $ylist $xlist
//Set data as panel data sort $id $t
xtset $id $t xtdescribe
xtsum $id $t $ylist $xlist
//Check for multicollinearity corr $ylist $xlist $dumlist
//Estimate VIF values
reg $ylist $xlist $dumlist, robust estat vif
40 //Hausman test (fixed effects vs random effects)
xtreg $ylist $xlist $dumlist, fe estimates store fixed
xtreg $ylist $xlist $dumlist, re estimates store random
hausman fixed random
//Breusch-Pagan lm test (random effects vs pooled OLS) xtreg $ylist $xlist $dumlist, re
xttest0
//Fixed effects model with only P_CoCo and dummies (model 1) xtreg $ylist P_CoCo $dumlist, fe robust
ereturn list
//Pooled OLS test (model 2) reg $ylist $xlist $dumlist, robust ereturn list
//Fixed effects full model (model 3) xtreg $ylist $xlist $dumlist, fe robust ereturn list
//Table output
xtreg $ylist P_CoCo, fe robust
outreg2 using reg, dec (4) replace ctitle(test) word reg $ylist $xlist $dumlist, robust
41 outreg2 using reg, dec (4) ctitle(Full Model) word append
