Effectiveness of the Basel III Bail-In Framework: Evidence from the Hybrid Security Market

Transcription

1 Effectiveness of the Basel III Bail-In Framework: Evidence from the Hybrid Security Market James Cummings and Yilian Guo Macquarie University September 2018 Abstract As part of the Basel III capital reforms, a new Additional Tier 1 (AT1) capital is introduced with the intention to provide greater loss-absorbing capacity which is the contingent convertible capital instruments (Cocos) widely discussed in the literature. Eligible Cocos that included as Additional Tier 1 capital need to be issued with both a capital ratio trigger and a mandatory point of non-viability trigger. With Basel III capital reforms started to implement in Australian banking system from 1 January 2013, which is three years ahead of Basel Committee s phasein deadline. We examine the extent to which a risk premium is embedded in the credit spreads of newly issued Basel III AT1 Cocos and whether there is an impact on the secondary market liquidity of Cocos. Our estimations suggest that (i) the Basel III AT1 Cocos realise a risk premium of around 20 basis points compared with hybrid securities issued under the Basel I and Basel II regimes; (ii) Coco investors are more sensitive to equity-based measures of banking risk: systematic risk, such as hybrid beta (iii) Pricing of hybrid securities has become more sensitive to a bank s common equity position, which provide evidence that the CET1 ratio based trigger matters for hybrid investors; (iv)the secondary market liquidity of Cocos is generally more expensive since the implementation of the Basel III capital reforms. JEL classification: G21, G28 Keywords: Cocos, Market discipline, Basel III capital reforms, Contingent convertible

2 1 Introduction Market discipline refers to the process that investors correct incentive to understand and price bank risks when provide funding(flannery, 2001). Intuitively, if investors understand bank risk correctly, they would ask for additional compensation that reflecting in declined security price and raised yield spread if bank risk increase. If a bank fails, all uninsured debt holders will bear the loss of a bank run, which stimulate their incentives to monitor banks risk-taking activities (Balasubramnian and Cyree, 2014). Past literature mostly propose to use uninsured wholesale debt as the risk monitoring tool as well as a supplement of capital requirements for maintaining a sound and resilient banking system (Benston and Kaufman, 1994; Wall, 1989; Evanoff and Wall, 2000; Evanoff and Jagtiani, 2004). Contrary to equity investors, unsecured debt investors do not realise the upward gains from bank s risky activities. When a bank is on its way towards insolvency and common equity tier 1 (CET1) instruments, i.e. ordinary shares, have been exhausted, investors of hybrid securities are the second in turn to absorb bank losses. Theoretically, this default process supposes to give them a strong incentive to monitor the risk-taking activities of the bank and a preference for low-risk investments. However, the financial crisis made evidence of severe moral hazard problem and reduced market discipline due to the banking safety net. Public rescue of failed systemically important financial institutions and massive central bank lending have worked as a last resort, which distorts bank investors sensitivity to the risk profile and weakens market discipline(acharya et al., 2016). On the other hand, due to the highly Interconnectedness inside the banking system, as well as between the financial sector and the general economy, if there is a credible threat to wholesale debt holders, it will be at the expense of financial stability. Because of the moral hazard problem and systemic risk concern, even though many empirical papers have found evidences that wholesale debt spreads, especially subordinated debt spreads, are informative about the issuing firms risk position and financial conditions (Flannery and Sorescu, 1996; Jordan et al., 2000; Morgan and Stiroh, 2001; Sironi, 2003), academics and practitioners point out the significance in urging additional oversight forms to help mitigate the perceptions of government support and restore the safety and soundness of banking system. One potential solution that is widely discussed by academics and regulators is to make wholesale debt a bit more like equities. Therefore, wholesale debt holders might become more 1

3 sensitive to bank risk and function as the monitoring tool for banks risk taking activities. Coco, a hybrid debt instrument that converts into common equity when a bank is over-levered with low core capital but still has the capacity to recapitalise itself, is proposed to respond to the recent financial crisis with the incentive to reduce potential government interventions and bailouts. Cocos are unsecured hybrid securities that designed to improve the absorption of any bank s unexpected future losses through automatic recapitalisation when a pre-defined trigger event is reached(avdjiev et al., 2013). Since Flannery (2002), many studies have discussed the proper design of Cocos to be included as a regulatory capital. During the crisis, many banks had insufficient capital to support the risks they had taken. To prevent further spread of crisis, costly bank recapitalisation, government intervention and central bank lending have been relied upon. When it is difficult for a going concern bank to issue new shares in a bad state, Cocos convert into ordinary shares to recapitalise the bank and replace bailout with bail in. Cocos not only automatically convert into common equity to inject capital into the banking sector and help mitigate potential government bailout, but also stimulate investors sensitivity to monitor bank default risks due to the potential losses associated with the conversion event gets automatically triggered. Coco investors will ask for a higher compensation when they realise a bank is taking excess risks, which expose them to the event of conversion or principal write-down. It is expected that introduction of Cocos will help improve market discipline by encouraging investors to better monitor banks risk-taking activities, which would contribute to improved market discipline in the banking system. The designs and concerns of different triggers have been discussed in past papers, such as accounting ratios, bank management s option and market stock price (Squam Lake Working Group and others, 2009; Flannery, 2009; Duffie and Lando, 2001; Birchler et al., 2006; Berg and Kaserer, 2015; Sundaresan and Wang, 2015). Many academics and regulators support the idea of using informative market prices as the mandatory-conversion trigger for Cocos. However, the multiple equilibrium and incentive problem would potentially lead to regulators reluctance of including Cocos with a market price trigger as regulatory capital. Also, market-price-based triggers may exacerbate financial stabilities. The role of Cocos in the new Basel III capital reforms lies in a newly defined Additional Tier 1 (AT1) category that can replace up to 1.5% of the common equity in Tier 1(Basel Committee on Banking Supervision, 2010). As defined by Basel Committee on Banking 2

4 Supervision (2010), hybrid securities that eligible to be included as AT1 capital need to be issued with the loss-absorbing contingent convertible feature, which involves a capital ratio trigger and a discretionary point of non-viability (PONV) trigger controlled by the national regulator. That is, if the issuing bank breach either a per cent CET1 capital ratio or a discretionary PONV trigger, the AT1 hybrid securities will be written down or converted into ordinary shares. The contingent convertible feature would potentially turn hybrid security investors into shareholders to fulfil the gap in core tier 1 capital when a bank is close to failure, which exposes them to higher risk. Therefore, a critical intention of this contingent convertible mechanism is to increase the risk-monitoring activities of hybrid security investors and thereby improve market discipline in the banking system. The Basel III capital reforms started to be implemented in Australia from 1 January 2013, which is three years ahead of the Basel Committee phase-in deadline 1. This provides us with an appropriate experimental environment to test the effectiveness of the contingent convertible feature on the pricing and liquidity of bank capital instruments. In addition, all Cocos issued by Australian banks are classified as liabilities under accounting standards, which means they have to issue with both a capital ratio trigger and a point of non-viability trigger to be eligible AT1. The United States has effectively eliminated the use of coco bonds by requiring that any AT1 instrument must be accounted for and treated as equity when it is first issued. This treatment reduces the tax advantage of coco bonds over common equity(flannery, 2014). Cocos issued by the European banks can have a single capital trigger or dual triggers to be eligible AT1. We investigate to what extent pricing and liquidity of bank capital instruments in banking system changed after the implementation of the Basel III capital reforms, in particular, the contingent convertible mechanism applied to Additional Tier 1 (AT1) hybrid securities. Using secondary market hybrid security transactions, we explore whether or not credit spreads increase for hybrid securities issued with the contingent convertible feature, i.e. the contingent convertible capital instruments (Cocos) that widely discussed in the literature, and whether there is a more significant relationship between bank default risk and credit spreads of Cocos. When a bank exposes to higher default risks, we expect credit spread on Cocos to increase in greater magnitude compared with conventional hybrid securities and to potentially facilitate stronger market discipline. In addition, we investigate whether or not the secondary market 1 Even the Basel III capital reforms started to implement from 1 January 2013, 4 Cocos in our sample are issued before that date. 3

5 liquidity of hybrid securities is impacted by the Basel III contingent convertible feature. With the risk profiles of Cocos are much more complex than conventional hybrid securities, we expect that investors would ask for higher compensation to hold Cocos. This will potentially lead to a higher transaction costs for investors, which may impede the ability of Basel III AT1 hybrid securities to promote market discipline in the banking system. Different from past research mainly focuses on the risk monitoring effects of wholesale debt or deposit, and theoretical design or pricing of Cocos, our study contribute to the literature of both market discipline and Cocos by documenting the extent to which the credit spreads of Cocos are sensitive to bank default risk and firm-specific financial conditions. More importantly, using secondary market transactions of hybrid securities, we provide early evidence for the pricing of a risk premium of Cocos due to the contingent convertible mechanism introduced since the Basel III. A wider spread of Cocos relative to conventional hybrid securities indicates that investors anticipate the risk of the potential and uncertainty that Cocos get converted into common equity or have their principal written-down if a bank experiences substantial losses. That is, it suggests Coco investors are demanding an additional risk premium because they are more exposed to bank risk. Our results suggest an approximately 89 to 116 basis points difference in credit spreads between the conventional hybrid securities and Cocos. We further investigate whether the contingent convertible feature would have the effect of producing a stronger relation between credit spreads and bank risk. As a result, we find little evidence for improved sensitivity of Cocos credit spreads to debt-based bank default risk. This result might due to the fact that investors attention is more focused on the contingent convertible feature and the opaque of national regulators measurement of a bank s non-viable. In addition, there is a strong argument that the Cocos have a high risk profile only due to the uncertainty about the future potential conversion (Australian Securities & Investments Commission, 2013). However, we find evidence of a stronger relationship between equity-based measures of systematic risk, i.e., Hybrid beta, and hybrid securities credit spreads. While prudential regulatory frameworks, including the Dodd-Frank Act and the Basel III capital reforms, put in efforts to build a robust financial system that has the ability to absorb unpredictable shocks, it is invariably beneficial to maintain market-wise liquidity. As generally defined, market liquidity is the ability to execute transactions immediately with limited price impact and low transaction costs. By efficiently allocating capital and risk, and monetary policy 4

6 and financial stability, market liquidity functions as the facility for the efficient allocation of economic resources. Even in normal times, market liquidity can differ from asset class to asset class. Assets with worse liquidity tend to have for higher liquidity risk premium, which associated with higher transaction costs faced by market participants. As a potential tradeoff of having enhanced capital and loss-absorbing capacity, there might be less liquidity during normal times, which leads to the concern regarding market liquidity in potential future turbulence(adrian et al., 2017). Narrow down to the loss-absorbency mechanism in the Basel III capital reforms applied to AT1 hybrid securities, previous literature have addressed an incentive issue associated with the use of Cocos, which is the concern regarding not only the issuers incentive to issue coco bonds, but also the cost for investors to participant in the trading of Cocos(Flannery, 2014; Chen et al., 2013; Albul et al., 2015). Cocos might become very illiquid in stressed market conditions, due to the opaque of information about the national regulators measure of a bank s non-viable and the timing of future potential conversion or write down. Then some investors will be reluctant to hold these securities or demand a large liquidity premium for doing so. Therefore, we investigate the impact of the contingent convertible feature on the market liquidity of hybrid securities. Using the time-weighted bid-ask spread as a measure of market illiquidity, we find Cocos are traded with approximately 4 basis points higher transaction cost compared with conventional hybrid securities. With higher transaction cost during normal times, it is worth investigating whether the liquidity of Cocos will get worse when a bank exposes to high risk or in systemic turmoil. To address this concern, we further examine whether the market liquidity of coco bonds is more sensitive to bank default risks, however, find little evidence for an increased risk sensitivity of the market liquidity of Cocos. That is, transaction costs in the hybrid securities market are resilient to changes in banks risk profile and financial conditions. 5

7 2 Literature review Since the 1988 Basel Accord, capital adequacy has been the main focus of prudential regulation in the banking system. Basel committee designs the mechanism to tie bank risk to the bank s capital level. Under the second pillar of the Basel Accord, national supervisors need to make sure that banks in each jurisdiction maintain an adequate level of capital. However, what we observe from the crisis is that capital requirements only has not contributed to a low-risk banking system because the opaque in banks accounting books lead to regulatory capital ratios overstate banks real capability of absorbing losses. As a good example, when the market value of Citigroup s Tier 1 capital at 1% of total assets, its book value still account at 11.8% in the end of 2008 (Duffie, 2009). Moreover, US firms that bankrupted or get acquired during the latest great recession still stood 12.3% to 16.1% Tier 1 capital ratios at the quarter-end before they closed (Kuritzkes and Scott, 2009). The market turmoil provides real evidence that capital adequacy itself might not be enough to ensure banks take less risk and maintain adequate lossabsorbing ability. The calls for alternative oversight forces attract a growing awareness among academics, practitioners and regulators as the banking system has become much complex than before. For this reason, Basel committee introduced market discipline as the second pillar of Basel Capital Accord(Basel II) framework, which would work as an alternative supplement to support the resilience and soundness of the banking system. The concept of market discipline, by definition among financial institutions and regulators, is the financial institutions security investors acting as monitoring agents to ensure the discipline of the firm Berger (1991). As defined by Bliss and Flannery (2002), two steps are involved in the market discipline process, which are monitoring and influence. Monitoring refers to the process that bank security holders and other investors monitor the risk activities the bank involves and require compensation for bearing bank risk. For marketable securities, the risk profile of the issuing bank should be reflected in the prices of the securities. Meanwhile, the second step, influence, is the process by which bank s security prices change stimulate bank responses to reduce their involvement in risk-taking activities. Among all the proposals, the most popular one is that subordinated debt could potentially be an additional layer of regulatory capital. In the second pillar of Basel II framework, the proposal of market discipline also argued the vital role of subordinated debt in establishing more effective market discipline in 6

8 the banking sector. A vast number of studies focus on the disciplining effect of subordinated debt in the last two decades (Flannery and Sorescu, 1996; Demirgüç-Kunt and Huizinga, 2004; Krishnan et al., 2005). Indeed, there is evidence that bank information and risks are embedded in the subordinated debt spreads prior to the market turmoil in However, the global financial crisis has revealed the fact of insufficient market discipline in the banking system and financial institutions anticipations of government bailouts (Flannery, 1998; Sironi, 2003; Gropp et al., 2004; Dam and Koetter, 2012). For the market discipline to be effective in the banking system, the price of banks debt instrument needs to be risksensitive, which require investors pricing banks risk profiles objectively when provide funding. Also, investors must perceive no government rescue if a bank becomes insolvent. Following the subordination, investors of hybrid securities such as preference shares and capital notes should be the second in turn to bear bank losses, followed by subordinated debt. However, farreaching public rescues and central bank lending in the United States and European countries have encouraged perceptions of government support as the last resort when bank fail. The expectation of potential government bailout weakens market discipline by reducing bank investors risk monitoring incentives, which reflected in a lower cost of bank funding, especially of systemically important financial institutions (Noss and Sowerbutts, 2012). Investors lack of risk sensitivity stimulates banks to take excessive risks and cause moral hazard problem in the financial sector. The banking safety net also leaves the expensive financial burden on taxpayers. In response to the questioned market discipline and the dysfunction of subordinated debt mechanism during the financial crisis, Basel III capital reforms require include not only subordinated debts but also hybrid securities as loss-absorption capital instruments. Different from eligible Tier 2 subordinated debts working as a gone concern regulatory capital that bail in a bank after it becomes insolvent, hybrid securities acting as going concern contingent convertible (Coco) Additional Tier 1 capital that recapitalising the bank before failure. When the bank s loss-absorbing capacity falls to a certain level, Cocos will be forced to convert into ordinary shares or have their principal written down. Either a conversion or principal write-down will restore an adequate level of a bank s capital and allow the bank to continue operating with sufficient loss-absorbing ability. Before the introduction of Basel III, many researchers have designed and discussed the use of contingent convertible instruments as an alternative solution to recapitalise going concern financial institutions. In the design of Cocos, it is argued that this new type of instrument may help in mitigating investors perception of 7

9 government support and reduce systemic risk in the financial sector. Treasury (2009) suggest that Cocos would not dilute earnings belong to existing shareholders as it is issued and functions as debt. Thus, Cocos may solve the problem of banks reluctance of raising equity when they are in good condition. Moreover, if a bank becomes going concern, conversion of a sufficient amount of Cocos would return the level of loss-absorbing capital and recapitalise the bank (Squam Lake Working Group and others, 2009). Lastly, the potential for a punitive conversion of Coco bonds may restore bank managers incentives to lessen risk-taking and maintain sufficient level of capital (Himmelberg and Tsyplakov, 2012). Because of those benefits that Cocos may bring in pursuing prudential capital structures of banks and reducing financial distress, some academics have urged prudential regulators to require banks use Coco bonds as regulatory capital. The proper design of Cocos has been debated by academics and prudential regulators, including the trigger event of Cocos, as well as their conversion rate. Many papers discuss the designs of the trigger event, including the initial design of market capital ratio (Flannery, 2002, 2005), accounting ratios (Squam Lake Working Group and others, 2009), and management option (Glasserman and Wang, 2011). Flannery (2002) first explore the potential to incorporate reverse convertible debentures (RCD) in banking firms capital structure, which can convert into common equities if the firm s capital ratio declines below a pre-specified level. Unlike conventional convertible bonds, the new instrument, RCD, would convert at the banking firm s current stock market price, which meets the purpose of forcing existing shareholders to bear the full cost of their decisions of taking excessive risks. However, each trigger design would have their limitations. For instance, triggers based on accounting measures would have the backward-looking issue. Also, accounting ratios can be manipulated by managers. On the other hand, allow bank managers to have the authority to decide conversion might lead to their incentive of late conversion or no conversion. From bank managers perspective, they would prefer a bailout rather than having banks Coco bonds convert to common equities that dilute existing shareholding. Because of those concerns, many academics follow the original design of Flannery (2002), pay most attention to trigger event placed on market prices. In the original design of Coco bonds, the trigger event is a bank s market capital ratio, while the Coco bonds would convert at the stock s current market price. However, this mechanism would provide unusual opportunities to short sellers, who would try to trigger conversion by bidding down the stock price. Flannery (2009) then propose an alternative measure for conversion rate, 8

10 which converts coco bonds into a certain number of common shares based on the trigger price, regardless of the current share price. However, this arrangement exposes bondholders to value loss if the trigger event has been breached Flannery (2014). Among all discussions in the literature, the problem of multiple equilibrium prices is one of the major issues of market price trigger. When conversion of Coco bonds gets triggered, the conversion price will reflect both the value of bank assets and the value redistributed between existing shareholders and Coco investors (Prescott, 2012; Berg and Kaserer, 2015; Glasserman and Nouri, 2016). This point is initially identified by Birchler et al. (2006) that supervisor might use market prices as the basis for policy actions. Berg and Kaserer (2015) apply a model initially proposed by Duffie and Lando (2001), which implies the uncertainty regards a bank s true asset value. Based on this model, Berg and Kaserer (2015) contend that investors uncertainty may mitigate the problem of multiple equilibrium prices. However, their paper does not indicate the exact mechanism of the trigger event. Sundaresan and Wang (2015) examines the characteristics of Coco bonds that using stock market price as the trigger event, and treat Coco bonds as American options. They propose a discrete time model and try to solve the equilibrium problem. They document that for a unique competitive equilibrium to exist, a mandatory conversion must not transfer value between equity holders and coco investors. In the Basel III capital reforms, a newly defined Additional Tier 1(AT1) capital instruments provide an extra level of the loss-absorption capital buffer. As a going-concern capital instrument, AT1 instruments would convert into equity or write down principal before bank failure. According to Flannery (2014), AT1 hybrid securities are Basel Committee s only treatment of Cocos. Different from proposals in the majority of past studies that using market price as the trigger of Cocos, eligible Basel III AT1 hybrid securities are issued with not only a book value per cent CET1 ratio, but also a regulatory discretionary PONV trigger. Ideally, Cocos with such dual-triggers would overcome the problems with market price trigger and pure accounting ratio trigger discussed in the literature. Moreover, what expected by the regulators is the implementation of contingent convertible mechanism would help restore investors right incentives to monitor bank risk. However, few studies have looked at the effect of this new regulatory instrument on improving market discipline. Spiegeleer and Schoutens (2012) apply conventional asset pricing techniques to actual Cocos issued by Lloyds Bank. However, not only those Cocos were issued with only a CET1 ratio trigger, but also their 9

11 study focuses on the modelling of these securities. Contributing to the existing literature, our study attempt to explore the regulatory effectiveness of the contingent convertible feature in promoting risk monitoring. Examining the correlation between credit spread of hybrid securities and various variables, this paper aims to investigate the following hypotheses: H1: Investors require a risk premium when they invest in the Basel III Cocos due to the contingent convertible feature. H2: Cocos investors are more sensitive to the banks risk-taking activities compared with conventional hybrid security investors. In addition, the second major issue of contingent convertible capital is the incentive effects, which would further pose complexion on market supply and demand. From the issuers perspective, one concern needs to be addressed is whether bank shareholders would choose to issue contingent convertible capital instruments instead of junior debt voluntarily. Albul et al. (2015) assume firms only would issue perpetual debt and study a model proposed initially by Leland and Toft (1996). Their model suggests that because the initial infinite-lived bondholders would realise all reduction in bankruptcy costs, firm shareholders would never voluntarily issue Coco bonds. On the contrary, Chen et al. (2013) study a similar model, in which assume an endogenous and finite debt maturity and also based on Leland and Toft (1996). With the intuition that reduction in bankruptcy costs will lead to a decline in interest rate required by subsequent debt holders, shareholders would eventually realise the benefit and willing to issue coco bonds. From investors perspective, the complexion of valuation in Cocos would raise the question of whether investors would understand the risks involved in the contingent convertible capital and whether the uncertainty in valuation would have a negative impact on investors demand. Glasserman and Nouri (2016) look at hypothetical Cocos which issue with a regulatory capital ratio trigger based on stock price. They document the correlation between the fair yield of contingent capital in their model and some of the input variables, such as the volatility of the firm s assets, the size of the Cocos, and the recovery rates when firm breaches the capital ratio trigger of Cocos. However, they also mention the limitation in their design, which is the incapacity to observe and estimate some of these inputs and produce obstacles for encouraging market demand. Therefore, we then look at the impact of the contingent convertible feature on 10

12 hybrid market liquidity and try to address the following hypothesis: H3: The introduction of the contingent convertible feature weakens the market liquidity of hybrid securities issued by banks. H4: Market liquidity of Coco instruments is more sensitive to bank default risk due to the contingent convertible feature. 11

13 3 Data and Methodology 3.1 Data and sample This study focuses on 56 hybrid securities issued by 9 Australian banks and traded from January 2003 to Dec The sample included in the study comprises 26 conventional hybrid securities (i.e. hybrid securities issued before the Basel III regime) and 30 Cocos (i.e. Basel III AT1 hybrid securities) issued since 1 Jan 2013, that is, the start date of Basel III implementation in Australia. Even hybrid securities are classified as shareholders equity on banks balance sheet; they also have the elements of debt instruments. As suggested by the Australian Stock Exchange (ASX), hybrid securities issue with a promise to compensate investors with interest until a pre-specified maturity date, which can be either fixed or floating rate (Australian Stock Exchange, 2017). However, as the most junior debt instrument and only senior to ordinary shares, hybrid securities generally have a higher rate of return compared with subordinated debts and senior bonds. The higher yields that investors realise is due to the corresponding higher risk of hybrid securities, which include not only the uncertainty of timing and amount of dividend firms pay to their hybrid investors but also the potential of conversion into common equity and early termination. As the newly issued hybrid securities are mainly preference shares and capital notes, also due to these two types of hybrid instruments have no derivative features, we only include those two types in our sample. Because of the combined features of both debt and equities, the pricing of hybrid securities is more complex than the other asset classes. To correctly calculate credit spread of hybrid securities, we collect issue level information of hybrid securities from Bloomberg, including their issue date, maturity type, effective maturity date, dividend rate, dividend type, dividend frequency, par value, issue amount, call feature and convertible feature. We also hand collect information of whether a hybrid security is issued with a contingent convertible feature (i.e. the Basel III triggers) or not from hybrid security prospectus published by banks. In our sample, all Cocos are floating-rate perpetual securities with no maturity, while the conventional hybrids have a combination of fixed-rate, floating-rate and variable dividend types. Moreover, to determine the effective maturity date, we consider both the maturity type and call 12

14 feature. To start, for a hybrid security has maturity type at maturity, the primary maturity date is used as the effective maturity date. Second, if a hybrid security has perpetual maturity type but has a mandatory conversion date, we use the mandatory conversion date as the effective maturity date. Third, consider a hybrid security has perpetual maturity type, if it is callable and has been called by the bank, we apply the called date as the effective maturity date. If it is callable with a price step-up feature, we use the first call date with price step-up as the effective maturity date due to the potential incentive of redemption provided to hybrids issuers. Otherwise, we either use the mandatory conversion date as the effective maturity date or treat it as perpetuity. Last, if the dividend type of a hybrid security is variable, which suggest a pre-scheduled change of dividend type on a specified future date, we employ the pre-specified change date as the effective maturity date. Also, We apply the dividend type of the first stage as the effective dividend type. Further, hybrid securities issued by Australian banks are all listed and traded on the ASX. We collect end of day close prices of hybrid securities from Bloomberg and abstract calendar month-end prices for calculating credit spread. Regarding the benchmark rate, we obtain government bond information and monthly yield from Bloomberg as benchmark risk-free rate for fixed-rate hybrids. We obtain bank bill swap rate from Australian Financial Markets Association as benchmark reference rate for floating-rate hybrids. For testing the impact of the contingent convertible feature on the market liquidity of hybrid securities, we obtain intraday time and sales data from Thomson Reuters Tick History (TRTH) to calculate time-weighted bid-ask spread across the day. With regard to control variables, to disentangle the effects of the contingent convertible feature from different bond characteristics, variation of risk, bank-specific characteristics and credit market conditions on banks borrowing costs, we collect daily data on the bank equity returns and market capitalisation from Thomson Reuters Datastream. Semi-annual data are obtained from Worldscope on the total assets, total liabilities, shareholders equity, net income, nonperforming loans, total loans, trading assets, cash and liquid assets, and short-term debt for the sample banks. In addition, for two variables controlling for credit market conditions, interest rates on bank accepted bills and the generic Australian Government bond of different maturities are collected from the Reserve Bank of Australia. Average redemption rate of S&P/ASX Australian Corporate Bond Index is obtained from Thomson Reuters Datastream. 13

15 3.2 Contingent convertible feature, bank risk and control variables To find out whether a hybrid security issued by Australian banks has the contingent convertible feature or not, we look into the issue prospectus of each hybrid security and search for keywords in Regulatory treatment. On the prospectus, if the security is convertible or/and has the writedown with capital and non-viability trigger, the security is identified as a regulatory Coco instrument. In addition, the prospectus will also specify in Post-transitional Basel III rules that this security is Additional Tier 1 for hybrid securities issued after the implementation of Basel III capital reforms. Otherwise, Post-transitional Basel III rules is Ineligible refers to that hybrid is old style, i.e. issued before the Basel III capital reforms. Table 1 lists the risk, bond characteristics, firm-specific and macro variables and the respective effect of credit spread levels. To measure the default risk of individual banks, seven different risk proxies are used in this study. To start, following numbers of previous studies Gropp et al. (2004, 2006); Hillegeist et al. (2004); Duffie et al. (2007); Bharath and Shumway (2008); Acharya et al. (2016), we employ Merton (1974) structural default model to calculate the first risk measure distance-to-default (MertonDD). As suggested by Bharath and Shumway (2008), MertonDD is not only popular among researchers and practitioners, Basel Committee on Banking Supervision (1999) also consider exploring the model for credit risk modelling. This model treats a firm s equity as a European call option that the underlying asset of the option is the firm s assets, while the firm s liabilities is the strike price. MertonDD measures the number of standard deviations away from the default point, where the default point is defined as the point at which the assets of the bank are just equal to its liabilities. If the market value of the firm s total assets drops below its book value of total liabilities, the call option will be left unexercised and the default firm will be passed onto its debt holders Hillegeist et al. (2004). Also, as shown by Duffie et al. (2007), MertonDD have significant power in generating a term structure of default probabilities over time. For enhancing the readability of the regression results, we follow Gropp et al. (2006) to use a MertonDD to indicate a greater risk using an increase in this variable. Besides MertonDD, we also use the volatility of market value of bank assets (AssetVolatility) as the second risk measure, which is a crucial variable in predicting default as it measures the likelihood of a firm s asset value can go down to a certain degree that incapable to repay its debt holders. 14

16 Table 1: List of risk, bond characteristics, firm-specific and macro variables used in regression Expected effect Expected effect Proxy for on credit spread on illiquidity Panel A: bank risk variables Negative Merton distance to default (- MertonDD) Equity value bank risk Positive Positive Asset volatility (Asset Volatility) Equity value bank risk Positive Positive Negative Z-score (- Zscore) On-balance sheet bank risk Positive Positive Idiosyncratic equity volatility (Idiosyncratic volatility) Equity value bank risk Positive Positive Non-performing loan/toal loan (NPL) On-balance sheet bank risk Positive Positive Beta value of hybrid securities (Hybrid beta Hybrid value bank risk Positive Positive residual term of hybrid securities (Hybrid residual Hybrid value bank risk Positive Positive 15 Panel B: bond characteristic variables Logarithm of issue size (Log issue size) Liquidity Negative Negative Remaining maturity (Tenor) Potential illiquidity Positive Negative Callable feature (Callable) Bond call option for issuers Positive Negative Convertible feature (Convertible) Bond call option for investors Negative Negative Panel C: firm-level variables for banks Return on assets (ROA) Profitability/Operational efficiency Negative Negative Total equity to total assets ratio (Equity ratio) Leverage Negative Negative Maturity mismatch (Maturity mismatch) Liquidity Negative Panel D: Macro variables 10 year - 3 month Treasury rate (Term premium) Slope of the yield curve Positive S&P/ASX corporate bond index spread (Default premium) Systematic risk Positive S&P/ASX 200 VIX (VIX) Equity market volatility Positive Positive

17 Campbell et al. (2008) criticise the forecasting power of Merton distance-to-default when put conditions on other variables. Bharath and Shumway (2008) also find the functional form in the Merton model is the main contributor to the forecasting power of Merton distance-todefault, rather than the procedure to solve default probability. To address these concerns, we apply an alternative measure of bank risk, idiosyncratic equity volatility, which is documented by Campbell and Taksler (2003) to have a direct relationship with the cost of borrowing for corporate issuers. Beltratti and Stulz (2012) also suggest that a higher proportion of loans and other earning assets in banks total assets is associated with higher idiosyncratic volatility, while a bank s Tier 1 capital ratio is negatively correlated with idiosyncratic volatility. To compute idiosyncratic volatility (Idiosyncratic volatility), for every trading day, we estimate a rolling market model using the standard Capital Asset Pricing Model (CAPM) over past 180 calendar days from 2002 to The market portfolio is the S&P ASX 200 index and the risk-free rate is the three-month BBSW rate. Moreover, since hybrid market return and volatility are expected to be highly correlated with the market returns and volatility, we further apply the standard CPAM model to the returns of hybrid securities, to calculate the beta value (Hybrid beta) and residual term (Hybrid residual) of hybrid returns over market returns as our last two marketvalue based risk measures. Same as idiosyncratic equity volatility (Idiosyncratic volatility), we use the S&P ASX 200 index as the market portfolio and the three-month BBSW rate as the risk-free asset. The above five risk measures may have the potential shortcoming that not all publicly-available information is efficiently reflected in stock market prices. The stock market may not accurately reveal all the information as those reflected in the financial statements (Sloan, 1996). Therefore, we employ two accounting-based risk measures to test the robustness of our results from regressions using market-based risk measures. The first one is Zscore, which is an alternative measure of the distance from a bank s insolvency(roy, 1952; De Nicolo, 2000; Bertay et al., 2013; Laeven and Levine, 2009). Specifically, Zscore = ROA + Equity Ratio δroa where ROA is the mean return on assets over the past four years, while δroa is the standard deviation of the return on assets over the past four years (Roy, 1952). De Nicolo (2000) suggest a strictly monotonic and negative relationship between the Z-score and the probability 16

18 of insolvency. As a higher Z-score indicates the bank is more stable, we take the negative of the Z-score for better readability of the results table. In addition, as suggested by Flannery and Sorescu (1996), banks with a higher proportion of non-performing loans are more likely to become insolvent because it reveals the quality of banks loan portfolio and a higher portion indicates lower asset quality. We apply a ratio of non-performing loans to total loans (NPL) as an alternative risk measure, which is widely applied by previous studies (Sironi, 2003; Demirgüç-Kunt and Huizinga, 2004; Elyasiani and Jia, 2008; Brewer and Jagtiani, 2013; Balasubramnian and Cyree, 2014). Our analysis also considers other factors that potentially would have an impact on hybrid security credit spreads. Following vast previous banking and bond pricing literature, we employ several control variables related to bond characteristics, strength and profitability of the bank, and credit market conditions to disentangle the differential of spreads between Cocos and the old-style hybrids from any other factors might contribute to the spread difference. First, as hybrid security credit spreads might sensitive to the endogenous effects of debt instruments themselves, we control for the determinants of hybrid securities characteristics, such as the logarithm of issue amount (Log issue size) 2 and term to maturity (Logtenor) to control for different bond characteristics. We also add dummy variables Callable and Convertible to control for the callable feature and convertible feature of hybrid securities. Then, we consider three firm-specific variables, namely, return on assets (ROA), equity ratio (Equity Ratio) and maturity mismatch (Mismatch). Further, we also include three macroeconomic control variables in our model, which are Term Spread and De f ault Premium. Term Spread is the spread between generic 10-year Australian Government bond interest rates and 90-day bank bill rates. De f ault Premium is the S&P/ASX corporate bond index, which is an index comprised of bonds issued by Australian Companies as a proxy for systematic default premium. The last one is the S &P/ASX 200 VIX index (V IX) for measuring the aggregated volatility in the share market, which is calculated using the 30-day implied volatility of the S&P/ASX 200 put and call option mid prices. As S&P/ASX 200 is only available since January 2008, we use the average of the implied volatilities on closest maturity S&P/ASX 200 put and call options as V IX. 2 Issue sizes of Hybrid securities are adjusted by inflation using consumer price index (CPI) from the Reserve Bank of Australia 17

19 4 Results and Discussions 4.1 Descriptive statistics Table 2 presents descriptive statistics for the observations of bank hybrid securities and key variables involved in the analysis. The average credit spread is basis points from January 2003 to Dec Figure 1 provide a yearly average of credit spreads on hybrid securities issued by Australian banks over the sample period. The figure is separated into 3 parts based on the timeline of Basel regimes. The dark grey bars represent the average credit spread of hybrid securities issued under Basel I and Basel II regimes, while the light grey bars illustrates those of hybrids issued with the contingent convertible feature. We can observe the credit spreads of Cocos are higher than those of the conventional hybrid securities. The differential in credit spread indicates a risk premium required by investors due to the additional risk post to investors. However, detailed analyses are needed to prove the empirically significant correlation between the contingent convertible and the spread difference. Table 3 presents the sample Australian banks involved in the analysis. As risk variables such as Merton distance to default, Asset volatility and Idiosyncratic volatility need to use equity price data, we could only include exchange listed banks in our sample. As defined by Australian Prudential Regulation Authority (2013), based on the size, interconnectedness, substitutability and complexity of banks, four of the total ten banks are identified as domestic systemically important banks. Table 4 reports the Pearson correlation coefficients between the key variables included in our analysis. As illustrated by the correlation table, the credit spreads of hybrid securities are significantly correlated with bond characteristics, bank default risk, bank profitability and financial positions, and credit market condition, which provide some evidence for the information embedded in the credit spreads of hybrid securities. Noticeably, hybrid credit spreads are negatively correlated with the log of total assets, which indicates the fact that larger banks realise a lower funding cost when issuing hybrid securities. 18

20 Table 2: Descriptive Statistics for bank hybrid security observations This table presents summary statistics for The sample period is from January 2003 to December Merton distance-to-default is distance to default value calculated using the Merton model. Asset Volatility % pa is the asset volatility calculated using the Merton model. Z-score is the accounting z-score. Non-performing loans % is non-performing loans divided by total loans. Idiosyncratic volatility is residual calculated from a rolling market model using the standard Capital Asset Pricing Model (CAPM) over past 180 calender days, which has bank stock return as the individual stock return. Hybrid beta is the beta term regressing from the CAPM using hybrid security return as the individual stock return. Hybrid residual is the residual regressing from the CAPM using hybrid security return as the individual stock return. Tenor year is the the term to maturity of hybrid securities. Issue size $mil is the natural logarithm of the issue size of hybrid securities, adjusted by inflation. Credit Spread bps is the difference between hybrid security monthly yields and maturity-matched government bond. While for floating-rate hybrid securities, Credit Spread bps is the discount margin, which is the difference between the internal rate of return on the hybrid cash flows and the reference bank bill swap rate, assuming that the reference rate does not change over the life of the hybrid security. -Merton is descending ranking of distance to default calculated using the Merton model. Quoted spread bps is the time-weighted bid ask spread averaged over the month. Total Assets $bil is the book value of total assets. Equity Ratio % is the book value of equity divided by book value of assets. Return on assets % is the return on assets, computed by net income divided by total assets. Maturity Mismatch % represents maturity mismatch and is calculated as short-term debt minus cash divided by total liabilities. Term spread bps is the term structure premium, measured by the yield spread between 10-year Australian Government bonds and 90-day bank accepted bills. Default premium bps is the default risk premium, measured by the yield spread between bonds in the S&P/ASX corporate bond index and Australian Government bonds with nearest average time-to-maturity. S&P ASX 200 VIX is the S&P/ASX 200 VIX index. Standard Lower Upper Data item Mean deviation quartile Median quartile Bank risk variables Merton distance-to-default Asset volatility % pa Z-score Non-performing loans % Idiosyncratic volatility Hybrid beta Hybrid residual Bond characteristic Issue size $mil Tenor years Credit spread bps Quoted spread bps Firm-level variables Total Assets $bil Equity ratio % Maturity mismatch % Return on assets % pa Macro variables Term spread bps Default premium bps S&P ASX 200 VIX

21 Table 3: Sample banks with subordinated debts observations This table presents the sample Australian domestic banks included in the analysis. The sample period is January 2003 to December All of the sample banks are listed on the Australian Securities Exchange. Four of the sample banks are designated by the Australian Prudential Regulation Authority as domestic systemically important banks (D-SIBs). Bank name Australia and New Zealand Banking Group Limited Commonwealth Bank of Australia National Australia Bank Limited Westpac Banking Corporation Bank of Queensland Limited Bendigo and Adelaide Bank Limited Macquarie Bank Limited St.George Bank Limited Suncorp-Metway Limited Bank type D-SIB D-SIB D-SIB D-SIB Non-D-SIB Non-D-SIB Non-D-SIB Non-D-SIB Non-D-SIB Figure 1: Yearly average of credit spreads between old-style hybrid securities and Cocos This figure shows the yearly average spreads of hybrid securities issued by sample Australian banks Credit Spread (y-axis) is in basis points. For fixed-rate hybrid securities, Credit Spread is the difference between hybrid security monthly yields and maturity-matched government bond. While for floating-rate hybrid securities, Credit Spread is the discount margin, which is the difference between the internal rate of return on the hybrid cash flows and the reference bank bill swap rate, assuming that the reference rate does not change over the life of the hybrid security. The time period (x-axis) is from January 2003 to December The light grey bars represent credit spreads of Cocos, which are issued under the Basel III regime. The dark grey bars represent the credit spreads of conventional hybrid securities, which are issued under the Basel I and II regimes. 20