The Too-Big-to-Fail Premium in Tier-2 Capital Bonds and Additional Tier-1 Capital Bonds Primary Markets: Evidence from China
-
Le Yang
Abstract
This study examines the too-big-to-fail expectations in the primary market issuance spreads of commercial bank tier-2 capital bonds and additional tier-1 capital bonds in China. Using a sample of 574 issuances with total amount of 4.76 trillion RMB (749 billion USD), we conduct median regressions with the issuance spreads as the dependent variable. The coefficients of DSIB and GSIB are negative and statistically significant at 1% and 5% levels respectively in the full sample, and are negative and statistically significant at 1% level in the subsample after the guidance on asset management businesses was announced. Ceteris paribus, the issuance spreads of capital bonds of systemically important banks are 15.2 bps to 19.7 bps lower than those issued by other banks. The too-big-to-fail expectations and implicit guarantee of systemically important banks in the Chinese bond market, which narrow the primary market spreads of capital bonds issued by systemically important banks, may account for the results. To address the potential endogeneity issues, we use 90% quantile and 95% quantile of sample bank consolidated assets as proxies of systemic importance, and use China Development Bank bond yields to calculate issuance spreads, and the results show that the conclusion is robust.
1 Introduction
The central financial work conference in 2023 stressed the importance of promoting the high-quality development of the bond market, promoting the formation of longterm capital, and broadening the channels for bank capital replenishment. After the Global Financial Crisis and the Great Recession, authorities have made significant efforts in increasing the resolvability and resilience of financial institutions. In 2009 at the Pittsburg Summit, G20 leaders required the Financial Stability Board to identify measures to address the too-big-to-fail issue.
Systemically important financial institutions (SIFIs), especially global systemically important financial institutions (G-SIFIs), are large, complex and highly interconnected firms, the disorderly failures of which would cause significant disruptions to the financial system and the economy. Governments would be left with few options except to provide public funds for bailing out the failing financial firms (FSB, 2010). By admitting that certain financial institutions are too-big-to-fail, a series of issues would be incurred: First, moral hazards would be created by bailing out too-big- to-fail institutions. When big banks expect that they themselves are too-big-to-fail, banks would take excessive risks and increase the threats of their failures to financial stability. Financial institutions would have the incentives to expand their balance sheets, raise their risk appetites, and to be overloaded with debt financing. Systemically important financial institutions would seek for more risks than that are socially optimal if they do not fully bear the costs of their operating decisions, increasing the probability and costs of financial crisis (FSB 2021). Meanwhile, large banks may receive preferential treatments such as lower financing costs, if they are expected to be too-big-to-fail, granting large banks unfair competitive advantages. Secondly, bailing out financial institutions would create negative externalities and impose the losses of the shareholders and unsecured debt holders of the failing firms on taxpayers. Third, bailing out financial institutions would weaken market discipline. If depositors and debt holders regard certain financial institutions as too-big-to-fail, they may not analyze the risk profile of the deposits and debt instruments carefully, hampering the market monitoring process.
There are debates regarding whether market monitoring and market discipline on banks’ risk taking were effective. Bernanke (2007) and Niu (2008) argue that subordinated debt holders had strong incentives to monitor the banks’ risk taking, while Bliss and Flannery (2002) argue that the market could not influence the risk taking of banks. Belkhir (2013) argues that apart from lacking market discipline, market participants might not be able to restrict the excess risk taking behaviors of banks due to implicit guarantee, such as too-big-to-fail guarantee. Ashcraft (2008) argues that there were conditions for effective market discipline from subordinated investors.
The traditional subordinated debt could only absorb losses when the financial institutions failed, and during the Global Financial Crisis, some financial institutions were bailed out so their subordinated investors were left intact. In order to address the too-big-to-fail issue and distorted incentives of debt holders, regulatory authorities removed the subordinated bonds issued before the crisis from eligible regulatory capital. After the Global Financial Crisis, authorities required commercial banks to issue tier-2 capital bonds and additional tier-1 capital bonds, which were bail-in instruments. The tier-2 capital bonds and additional tier-1 capital bonds issued by global systemically important banks (G-SIBs) are total loss absorbing capacity (TLAC)instruments. Few strands of literature have investigated the factors influencing the credit spreads and the market discipline of tier-2 capital bonds and additional tier-1 capital bonds issued after the Global Financial Crisis.
The Chinese commercial banks issued first tier-2 capital bonds in July 2013. The outstanding amount of tier-2 capital bonds and additional tier-1 capital bonds was 6.53 trillion RMB (895 billion USD) by the end of June 2024, accounting for 49% of the total amount of the outstanding financial bonds, excluding policy bank bonds, in China. The average amount per issue was 7.95 billion RMB (1.25 billion USD), which was much higher than that of the non-financial corporate debt securities in China. Almost all the tier-2 capital bonds and additional tier-1 capital bonds issued by commercial banks are issued and traded in the Chinese interbank bond market.
Identifying the factors driving the primary market pricing of the capital bonds of Chinese commercial banks and whether there exists a “too-big-to-fail” premium in the issuance spreads of capital bonds of Chinese commercial banks are of great theoretical and practical importance. This paper examines the factors influencing the issuance spreads of tier-2 capital bonds and additional tier-1 capital bonds of Chinese commercial banks, and finds that the issuing spreads of tier-2 capital bonds and additional tier-1 capital bonds reflect the information contents in bond ratings and partial information contents in the banks’ risk indicators. This paper also finds evidence on the too-big-to-fail premium in the tier-2 capital bonds and additional tier-1 capital bonds primary markets. These results contribute to the strand of literature on the market discipline of banks’ risk taking, and the strand of literature on Chinese bond market, providing new evidence about the market discipline in the Chinese capital bond market. These results also provide new insights in the implicit guarantee of Chinese financial bonds.
The rest of the paper is organized as follows: Section 2 reviews the literature and develops the hypothesis. Section 3 describes the data sample and the construction of variables. Section 4 describes the empirical methodology. Section 5 presents the empirical findings on the too-big-to-fail premium in Chinese commercial bank capital bonds. Section 6 concludes the paper.
2 Literature Review and Hypothesis Development
Theoretically, when subordinated debt and senior debt were both the certain parts of an undifferentiated bond issue, subordinated debt would provide more expected value and protection for senior debt (Black and Cox, 1976). Subordinated bond spread is defined as the difference between the yield to maturity of the subordinated debenture and the yield to maturity of the risk free bond with similar characteristics (Balasubramnian and Cyree, 2011). Krishnan et al. (2005) argue that the level and spread changes of the entire yield curve should reflect the changes in risks. When therisks of banks increase, subordinated bond spreads are expected to widen, providing market discipline. Flannery and Sorescu (1996) find that subordinated bond credit spreads were sensitive to bank risks, especially after FDICIA was passed in 1991, since FDICIA restricted undifferentiated bailing out of failed banks. Similarly, Imai (2007) finds that the sensitivities of Japanese banks’ subordinated bonds credit spreads increased with the removals of implicit guarantees. Yet Avery et al. (1988) and Gorton and Santomero (1990), using data before the FDICIA, find that the credit spreads of bank subordinated bond yields to be not sensitive to bank risks. Krishnan et al. (2005) argue that credit spreads didn’t provide useful signaling for bank market discipline, and changes in credit spreads didn’t reflect the changes in default risks between 1994 and 1999, although the levels of credit spreads reflect the risks of individual institutions. Collin-Dufresne et al. (2001) and Krishnan et al. (2005) find that the traditional measures of default risks are not enough to explain the changes in non-financial corporate credit spreads. Driessen (2005) finds that the company specific default risk premium are small in the levels of credit spreads of solvent companies. Balasubramnian and Cyree (2011) infer that the signals of default implied by the changes in bank subordinated debt credit spreads have little impact on the levels of credit spreads. However, Covitz and Downing (2007) argue that default risk rather than liquidity risk is the more important determinant of credit spreads, even for short-term commercial papers. Balasubramnian and Cyree (2011) identify three omitted factors that might influence the subordinated bond default risks of US banks.
Bernanke (2007) argues that regulators should consider the incentives of different stakeholders: shareholders might gamble for resurrection via excess risk taking, while unsecured debt holders might not benefit from the successful excess risk taking of the banks. Bernanke (2007) argues that the incentives of subordinated debt holders to monitor risk taking are strong. Since the debt holders are sensitive to the probabilities of financial distress, the increase of bank risk taking would increase their debt financing costs. Bernanke (2007) also argues that the bond prices of banks provide useful information of banks’ risks. Counterparties and regulators could utilize this information and take measures to secure the safe operations of banks. Belkhir (2013) defines the market discipline in the context of subordinated debt as certain unsecured bank debt holders mitigating moral hazard behaviors of banks through certain mechanisms. Bliss and Flannery (2002) argue that there were two components of market monitoring: market monitoring and market influence. Market monitoring is the hypothesis that investors could monitor the changes in banks’ risk profiles and incorporate them into the securities prices. Market influence is the investors and regulators’ capacity to influence the banks’ risk taking behavior. Before the 2008 Global Financial Crisis, market discipline was regarded as the important aspect of banking supervision. Niu (2008) argue that the requirements of regular issuance of subordinated bonds increase the effectiveness of market discipline. However, the experience of the 2008 Global Financial Crisis indicated that the market mechanism might failed in preventing banks from taking risks excessively. As for the literature about market influence, Bliss and Flannery (2002) haven’t found evidence that managers responded by adjusting banks’ balance sheets to changes in credit spreads, arriving at the conclusions that market could not influence the risk takings of banks. Ashcraft (2008) tests the influence of the presence of subordinated debt on banks’ capital structures, and finds that as long as subordinated debt holders could restrain certain bank behaviors, there would be market discipline.
Belkhir (2013) argues that apart from lack of market discipline, the fact that market participants could not effectively restrain bank risk taking might be the results of implicit guarantee, such as too-big-to-fail guarantee. In the 2008 Global Financial Crisis, after the bankruptcy of Lehman Brothers, US government promised to avoid the failures of systemically important financial institutions and injected capital into major banks. Debt holders of systemically important financial institutions might lack the incentives to effective risk monitoring, as they assumed that the governments would not allow systemically important financial institutions to fail. It is necessary to test whether the disciplinary behaviors of subordinated debt holders are constrained in too- big-to-fail banks. Fajardo and Mendes (2020) finds that the banks that issue contingent convertible bonds (CoCos) are large, and in emerging economies, highly leveraged. Jaworski et al. (2017) argue that CoCos could potentially strengthen the resilience of the issuer as long as the probability of conversion trigger is higher than the VaR’s significance level. Bianchi et al. (2023) propose a model to jointly analyze the asset volatility implied in senior and subordinated bonds and credit default swap spreads, and core tier-1 capital volatility ratio extracted from additional tier-1 (AT1) bonds spreads. However, Allen and Golfari (2023) argue that CoCos are increasingly issued without punitive wealth transfer from shareholders to bondholders, thus eliminating incentives for banks to prevent excessive risk taking. Kund et al. (2023) find that banks issue additional tier-1 bonds to manage the Leverage Ratio, and the tier-2 capital bonds don’t influence the TLAC. Li et al. (2018) and Li et al. (2020) provide a comprehensive framework for understanding the relationship between Chinese bank’s capital structure and write-down bond issuance.
Traditional subordinated debt was designed to absorb losses before senior debt, yet the loss absorption was after the bankruptcy of the bank. In the Global Financial Crisis, when several banks were failing simultaneously, authorities injected capital into the banks and provided subordinated debt holders with de facto protection. To address the too-big-to-fail issue, financial reforms after the Global Financial Crisis treated traditional subordinated debt as not eligible for regulatory capital. Tier-2 capital bonds and additional tier-1 capital bonds with triggers of PONV (Point of Non-viable) are introduced and became important tools for commercial banks to raise capital and increase their loss-absorbing capacity. The first tier-2 capital bond was issued in July 2013 in China. In 2023, Chinese commercial banks issued 1.12 trillion RMB (153 billion USD) tier-2 capital bonds and additional tier-1 capital bonds.
Previous literature on the relationship between primary market issuance spreads of tier-2 capital bonds and additional tier-1 capital bonds and bank risks is relatively uncomprehensive. While Li et al. (2018) and Li et al. (2020) analyze the impact of write-down bond issuance on asset allocation, capital allocation and expected bankruptcy losses, this paper focuses more on the factors influencing the primary market issuance credit spreads of capital bonds and the implications. Kind et al. (2022) find that maximum distributable amount (MDA) has an economically significant impact on CoCo spreads of European banks. Since MDA is a unique regulatory concept introduced in the European Capital Requirement Directive, alternative variables need to be identified in the context of Chinese banking sector and capital bonds market. Bolton et al. (2023) use a case study approach to analyze the CoCo price behavior in the context of the collapse of Credit Suisse, yet few comprehensive analyses of the credit spreads of capital bonds issued by global systematically important banks have been done.
Theoretically, if variables indicating the bank risks were highly significant in determining the tier-2 capital bonds and additional tier-1 capital bonds issuance spreads, then it could be argued that capital bonds investors might restrain the excess risk taking of commercial banks. Otherwise the market discipline of capital bonds investors might be regarded as ineffective (Beyhaghi et al., 2014; Nguyen, 2013). Livingston et al. (2018) argue that there were significant information contents in the issuing spreads of Chinese corporate bonds.
Moreover, if variables indicating the systemic importance of banks are significantly negative in determining the capital bonds primary market credit spreads, there might be implicit guarantee or expectations of bailouts for too-big-to-fail banks, making the capital bonds spreads not sensitive to bank risks (Balasubramnian and Cyree, 2011). Acharya et al. (2016) examine too-big-to-fail premium in US financial bonds, yet the too-big-to-fail relationship was not observed in non-financial sector.
According to Acharya et al. (2016), bond investors might expect governments not to allow SIFIs to fail so as to protect the financial system and economic activities from disruptions. This is an implicit guarantee or the too-big-to-fail problem. If there are expectations of too-big-to-fail, the risk taking behaviors of banks might be insufficiently priced in by bond investors. The global systematically important bank list was first published in 2012 and the big four banks (ICBC, CCB, BOC, ABC) in China were all identified as global systematically important bank in 2015. The domestic systemically important banks (D-SIBs) list in China was published in 2021, yet the market might have long expected these banks to have systemic importance due to their size and the interconnectedness. With the above analysis, we have the following hypothesis:
Ceteris paribus, there exist implicit guarantee premium (too-big-to-fail premium) in tier-2 capital bonds and additional tier-1 capital bonds issued by Chinese systemically important banks, which lower their issuance spreads.
3 Data and Construction of Variables
This study uses the tier-2 capital bonds and additional tier-1 capital bonds issued from July 2013 to February 2022 as the sample. There are 574 sample issues and 268 issuers, with the total issuance amount of 4.76 trillion RMB (749 billion USD). Auction dates, security codes, security short-names, coupons, issuance dates, maturities, issuers, issuance amounts, ratings, total assets, nature of the firms, ROA, non-performing loan ratios, and shareholder equities to total assets ratios are obtained from the WIND database. 5-year and 10-year China Development Bank yields, FR007, yields to maturity of China Government Bonds of different maturities, secondary market yields of tier-2 capital bonds and additional tier-1 capital bonds are from WIND database and China Bond Financial Valuation Center. CPI, GDP expectations and total social finance figures are from the WIND database and the Bloomberg terminal.
AUCSPREAD is defined as the difference between the auction-date primary market yield of a tier-2 capital bond or an additional tier-1 capital bond and the yield of the China Government Bond with similar maturity assuming the excise of the issuer call option. The mean, standard deviation, minimum value and maximum value of AUCSPREAD are 179.95 bps, 57.17 bps, 56.29 bps and 381.55 bps respectively.
As for issuer characteristics variables (Issuer), this study constructs LNTA, ROA, NPL, LEVERAGE, VOL, and LnZ as independent variables. LNTA is the logarithm of total assets of the issuer bank in the latest financial year. ROA is the return on assets of the issuer bank in the latest financial year. NPL is the non-performing loan ratio of the issuer bank in the latest financial year. LEVERAGE is the ratio of shareholders’ equity and total assets of the issuer bank in the latest financial year. VOL is the standard deviation of the ROA of each commercial bank in the sample. Following Houston et al. (2010), LnZ is defined as
The bond characteristics variables (Facility) include DAAA, DAA, Rating, RM, DAT1, and LNAMOUNT. DAAA and DAA are dummy variables of issuance ratings. If the bond rating is AAA, the value of DAAA would be 1, and the value of DAAA would be 0 otherwise. If the bond rating is AA+, AA, or AA-, the value of DAA would be 1, and the value of DAA would be 0 otherwise. As in Livingston et al. (2018), we define Rating as a variable with discrete value from 1 to 5. If the bond rating is AAA, the value of Rating would be 5; if the bond rating is AA+, the value of Rating would be 4; if the bond rating is AA, the value of Rating would be 3; if the bond rating is AA-, the value of Rating would be 2; if the bond rating is A+, A, or A-, the value of Rating would be 1. RM is defined as the remaining maturity of the capital bond assuming that the issuer call option is exercised. DAT1 is a dummy variable with value of 1 if the bond is an additional tier-1 capital bond and with value of 0 if the bond is a tier-2 capital bond. LNAMOUNT is the logarithm of the amount of the bond issuance.
The macroeconomic and financial market variables (Macro) include GDP, CPI, TSF, AUC5YT, AUC10Y1Y, and FR007. GDP is the expected GDP growth rate in a given year. CPI is the expected CPI growth rate in a given year. GDP and CPI expectations are from the Bloomberg terminal. TSF is the annualized growth rate of total social finance lagged 6 months. AUC5YT is the 5-year China Government Bond yield to maturity at the auction date. AUC10Y1Y is the 10-year to 1-year spread of China Government Bonds yield to maturity at the auction date. FR007 is the seven-day repo rate at the auction date in the Chinese inter-bank bond market.
The bailout expectation variables (Bailout) include DSOE, DSIB and GSIB, which are all dummy variables. If a commercial bank is state-owned, the value of DSOE would be 1; otherwise the value of DSOE would be 0. If the issuer bank is one of the nineteen domestic systematically important banks identified by PBoC and CBIRC in 2021, the value of DSIB would be 1, otherwise the value of DSIB would be 0. If the issuer bank is one of ICBC, CCB, BOC or ABC, the value of GSIB would be 1, otherwise the value of GSIB would be 0.
4 Empirical Methodology
This study constructs the empirical model to regress the issuance spreads of tier-2 capital bonds and additional tier-1 bonds of commercial banks on variables of bailout expectations, issuer characteristics, bond characteristics, and macroeconomic and financial markets conditions (Balasubramnian and Cyree, 2013; Evanoff et al., 2011). In order to control possible paradigm shift, we also use subsamples before and after the announcement of the Instructive Guidance on Regulating the Asset Management Businesses of Financial Institutions in 2018. The regression model is as follows:
We label certain issuer bank as j, the capital bond issuance as i, and the auction date as t; (t-1) is the previous year end of the auction date. We use quantile (median) regression and calculate the heteroskedasticity-robust standard errors.
The endogeneity issue of using DSIB and GSIB as independent variables might not be of a big concern. PBoC and CBIRC used size, interconnectedness, substitution and complexity as indicators to assess the systemic importance of banks. Issuance spreads of capital bonds are not any indicator of systemic importance, and capital bonds only account for less than 2% of the consolidated balance sheets of global systematically important banks and domestic systematically important banks. So the influence of the issuance spreads of capital bonds of commercial banks on the systemic importance of banks might be small.
5 Empirical results
5.1 Regression Results of the Full Sample
In order to control the influence of extreme values, this study uses quantile (median) regression to estimate the model (1) and calculates heteroskedasticity-robust standard errors. To avoid multicollinearity, we calculate the VIF and drop the variables if the VIFs exceed 10.
The following table illustrates the median regression results. The dependent variables are the primary market issuance spreads of tier-2 capital bonds and additional tier-1 capital bonds of commercial banks. The independent variables include vectors of variables of issuer characteristics, bond characteristics, bailout expectations, and macroeconomic and financial market conditions. The sample is consisted of 574 observations. Pseudo R2 are between 0.325 and 0.458.
Systemically important banks are more complex and facing higher risks and uncertainties than ordinary banks. Ceteris paribus, the issuance spreads of systemically important banks’ capital bonds should not be lower than those of other banks. However, in the regression results, DSIB variable is negative and significant at 1% level. This result might indicate too-big-to-fail expectations for domestic systemically important banks in China, which lower the issuing spreads of domestic systematically important banks’ capital bonds. Regression results with all control variables show that the issuing spreads of domestic systematically important banks’ capital bonds are lower than those of other banks’ by an average of 17.5 bps, which were also economically significant. This result provides evidence for validating H1. However, DSOE is not statistically significant, which is interesting and indicates that the investors’ focus on the implicit guarantees of financial firms and non-financial firms are different.
The coefficients of Rating in column 1, 3, and 4 in Table 2 are negative and significant at 1% level. In column 2, the statistical significance and absolute value of ROA and VOL increase without Rating as the independent variable. This result indicates that capital bonds issuance spreads fully reflect information in credit rating. The higher the ratings from credit rating agencies (CRAs) are, the lower the issuance spreads of commercial bank capital bonds will be. Thus bond specific rating from CRAs may contain important information regarding the risk measures of banks. The positive sign of ROA indicates that bond investors are skeptical of extremely high return on assets, which might imply excess risk taking.
Summary Statistics
| Variables | Observations | Mean | Standard Deviation | Min | Max |
|---|---|---|---|---|---|
| AUCSPREAD (%) | 574 | 179.95 | 57.17 | 56.29 | 381.55 |
| LNTA | 574 | 3.46 | 0.99 | 1.84 | 5.52 |
| ROA (%) | 574 | 0.84 | 0.34 | -0.62 | 2.70 |
| NPL (%) | 574 | 1.66 | 0.57 | 0.63 | 5.35 |
| LEVERAGE (%) | 574 | 7.32 | 1.20 | 3.55 | 13.50 |
| VOL (%) | 574 | 0.29 | 0.19 | 0.04 | 1.64 |
| LnZ | 574 | 3.51 | 0.62 | 1.40 | 5.27 |
| DAAA | 574 | 0.24 | 0.43 | 0.00 | 1.00 |
| DAA | 574 | 0.52 | 0.50 | 0.00 | 1.00 |
| Rating | 574 | 3.03 | 1.48 | 1.00 | 5.00 |
| RM (Year) | 574 | 5.18 | 0.95 | 2.00 | 10.00 |
| DAT1 | 574 | 0.24 | 0.43 | 0.00 | 1.00 |
| LNAMOUNT | 574 | 1.29 | 0.76 | -0.40 | 2.93 |
| AMOUNT (hundreds of millions RMB) | 574 | 82.93 | 148.52 | 0.40 | 850 |
| GDP (%) | 574 | 6.00 | 2.12 | 2.20 | 8.10 |
| CPI (%) | 574 | 1.90 | 0.71 | 0.90 | 2.90 |
| TSF (%) | 574 | 12.53 | 1.67 | 10.26 | 20.89 |
| AUC5YT (%) | 574 | 3.07 | 0.39 | 1.79 | 4.01 |
| AUC10Y1Y (%) | 574 | 0.55 | 0.23 | -0.04 | 1.63 |
| FR007 (%) | 574 | 2.61 | 0.52 | 1.30 | 5.00 |
| DSOE | 574 | 0.70 | 0.46 | 0.00 | 1.00 |
| DSIB | 574 | 0.22 | 0.42 | 0.00 | 1.00 |
| GSIB | 574 | 0.11 | 0.31 | 0.00 | 1.00 |
| AFTERRULE | 574 | 0.72 | 0.45 | 0.00 | 1.00 |
The Too-Big-to-Fail Premium in Tier-2 Capital Bonds and Additional Tier-1 Capital Bonds Issuance Spreads
| (1) | (2) | (3) | (4) | |
|---|---|---|---|---|
| DSIB | –17.49*** | –81.81*** | –31.78*** | –34.74*** |
| (–3.47) | (–15.16) | (–5.49) | (–6.82) | |
| DSOE | –5.469 | –5.343 | –6.686 | –5.031 |
| (–1.47) | (–1.17) | (–1.62) | (–1.26) | |
| ROA | –2.533 | 21.13*** | 6.061 | |
| (–0.39) | (3.13) | (0.92) | ||
| NPL | 0.902 | 13.87*** | 7.912** | |
| (0.25) | (3.28) | (2.25) | ||
| LEVERAGE | –2.353 | –0.402 | –1.560 | |
| (–1.64) | (–0.22) | (–1.10) | ||
| VOL | 9.935 | 31.22** | 4.925 | |
| (0.74) | (2.49) | (0.37) | ||
| Rating | –29.08*** | –26.19*** | –24.44*** | |
| (–15.96) | (–16.50) | (–14.31) | ||
| RM | 2.042 | 1.590 | 1.280 | |
| (1.64) | (0.93) | (0.89) | ||
| DAT1 | 20.49*** | 15.62*** | 18.69*** | |
| (4.57) | (3.41) | (4.65) | ||
| Macro | No | Yes | Yes | Yes |
| Year Fixed Effects | Yes | No | No | No |
| N | 574 | 574 | 574 | 574 |
| Pseudo R2 | 0.4579 | 0.3245 | 0.4334 | 0.4375 |
Note: Quantile (median) regressions are applied. t statistics are in parentheses. *, **, and *** indicate that the coefficient is statistically significant at the 10%, 5% and 1% respectively. The dependent variable is AUCSPREAD. The same for the below.
Column 5 to 8 in Table 3 show the results with GSIB as the independent variable of bailout expectations. To avoid multicollinearity, we haven’t included year fixed effects and financial markets variables, but calculated heteroskedasticity-robust standard errors. The coefficient of GSIB is statistically significant and negative. Ceteris paribus, the primary market spreads of capital bonds issued by global systematically important banks are 15.2 bps to 19.7 bps lower than those issued by other banks. This result again provides evidence to validate H1. The coefficient of DSOE is still insignificant.
The Too-Big-to-Fail Premium in Tier-2 Capital Bonds and Additional Tier-1 Capital Bonds Issuance Spreads (Using Alternative Explanatory Yariables)
| (5) | (6) | (7) | (8) | |
|---|---|---|---|---|
| GSIB | –15.15*** | –78.58*** | –21.73*** | –19.73** |
| (–3.29) | (–16.60) | (–2.86) | (–2.45) | |
| DSOE | –0.895 | –6.349 | 2.207 | –2.859 |
| (–0.23) | (–1.34) | (0.58) | (–0.94) | |
| Issuer | Yes | Yes | No | Yes |
| Facility | Yes | No | Yes | Yes |
| Macro | No | Yes | Yes | Yes |
| Year Fixed Effect | Yes | No | No | No |
| N | 574 | 574 | 574 | 574 |
| Pseudo R2 | 0.4086 | 0.2353 | 0.3921 | 0.4031 |
5.2 Regression Results of Subsamples
In 2018, PBoC, CBIRC, CSRC, and SAFE jointly published the Instructive Guidance on Regulating the Asset Management Businesses of Financial Institutions (“the Guidance”), which promoted the marketization of the wealth management products issued by Chinese banks. This study divides the sample into two subsamples which are consisted of pre-guidance and post-guidance observations respectively. In Table 4, regression 1 and 2 use DSIB as the independent variable of bailout expectations. Regression 3 and 4 use GSIB as the independent variable of bailout expectations. Regression 1 and 3 use pre-guidance data which is consisted of 159 observations. Regression 2 and 4 use post-guidance data which is consisted of 415 observations. Pseudo R2 are between 0.441 and 0.521.
The Too-Big-to-Fail Premium in Tier-2 Capital Bonds and Additional Tier-1 Capital Bonds Issuance Before and After the Announcement of the Guidance
| (1) | (2) | (3) | (4) | |
|---|---|---|---|---|
| Pre_Guidance | Post_Guidance | Pre_Guidance | Post_Guidance | |
| DSIB | –0.841 | –28.53*** | ||
| (–0.06) | (–4.96) | |||
| GSIB | –44.06*** | –22.83*** | ||
| (–2.96) | (–3.50) | |||
| DSOE | –5.418 | –6.225 | –6.726 | –5.551 |
| (–1.17) | (–1.56) | (–1.22) | (–1.21) | |
| ROA | 16.48 | 5.908 | 21.67* | 8.138 |
| (1.45) | (1.15) | (1.74) | (1.45) | |
| NPL | 22.62*** | 8.620** | 21.94*** | 6.584* |
| (3.65) | (2.23) | (3.00) | (1.77) | |
| LEVERAGE | –5.542** | –0.652 | –2.606 | 0.619 |
| (–2.52) | (–0.40) | (–0.89) | (0.35) | |
| VOL | –3.858 | 8.205 | –7.971 | 26.48* |
| (–0.25) | (0.51) | (–0.40) | (1.66) | |
| Rating | –22.02*** | –29.88*** | –15.91*** | –33.13*** |
| (–6.05) | (–17.05) | (–6.43) | (–21.12) | |
| RM | 7.465 | 1.697 | 11.43 | 1.137 |
| (0.12) | (1.40) | (0.29) | (0.73) | |
| DAT1 | 17.43*** | 20.28*** | ||
| (3.99) | (4.30) | |||
| Macro | Yes | Yes | Yes | Yes |
| N | 159 | 415 | 159 | 415 |
| Pseudo R2 | 0.4410 | 0.5207 | 0.4572 | 0.5142 |
Following Balasubramnian and Cyree (2011), this study uses pre-guidance and post-guidance data to conduct regressions respectively, and the absolute value and statistical significance of all the coefficients are compared. The results of regression 3 and 4 show that the coefficients of GSIB are negative and statistically significant at 1% level in both subsamples, indicating that there might be too-big-to-fail premium for global systematically important banks both before and after the Guidance was published. Ceteris paribus, the issuance spreads of capital bonds of global systematically important banks are 22.8 bps lower than those of other banks, which are also economically significant.
The results of regression 1 and 2 show that the coefficient of DSIB is not statistically significant in the pre-guidance subsample, yet was statistically significant in the post-guidance subsample. Ceteris paribus, the issuance spreads of capital bonds of domestic systematically important banks are 28.5 bps lower than those of other banks, which are also economically significant. The too-big-to-fail premium might actually increase after the announcement of the Guidance. The coefficients of DSOE are insignificant in regressions 1 to 4.
5.3 Results of Regressions Using Alternative Variables to Address Endogeneity
In order to address the potential endogeneity problem, we follow the approach proposed by Acharya (2016) and use the 90% quantile and 95% quantile of sample bank consolidated assets as proxies for too-big-to-fail expectations. To further alleviate the endogeneity problem, we use the consolidated total assets of the financial year which is one year before the capital bond issuance. Since capital bonds only account for less than 2% of the consolidated balance sheet of banks, the issuance spreads might have relative small influence on the consolidated assets of commercial banks.
We define SIZE90 and SIZE95 as two dummy variables. The value of SIZE90 would be 1 if the issuer bank’s consolidated assets are higher than the 90% quantile of sample commercial banks’ assets, and the value of SIZE90 would be 0 otherwise. The value of SIZE95 would be 1 if the issuer bank’s consolidated assets are higher than the 95% quantile of sample commercial banks’ assets, and the value of SIZE95 would be 0 otherwise. The full sample is consisted of 574 observations from 2013 to 2022, and the subsamples are issuances before and after the announcement of the Guidance respectively.
The results in Table 5 and Table 6 show that the coefficients of SIZE90 and SIZE95 are statistically significant at 1% level and negative. Holding other factors, the issuance spreads of capital bonds issued by commercial banks with consolidated assets higher than the 90% quantile of sample bank assets are on average lower by 24.0 bps than those of capital bonds issued by other banks, and on average lower by 24.3 bps after the Guidance was announced. Holding other factors, the issuance spreads of capital bonds issued by commercial banks with consolidated assets higher than the 95% quantile of sample bank assets are on average lower by 23.5 bps than those of capital bonds issued by other banks, and on average lower by 20.0 bps in the post-guidance subsample. This indicates that the previous conclusion is still robust.
The Too-Big-to-Fail Premium in Tier-2 Capital Bonds and Additional Tier-1 Capital Bonds Issuance: Using SIZE90 as Independent Variable
| (1) | (2) | (3) | |
|---|---|---|---|
| Full | Pre_Guidance | Post_Guidance | |
| SIZE90 | –24.04*** | –5.917 | –24.34*** |
| (–3.96) | (–0.49) | (–4.85) | |
| DSOE | –7.404* | –7.396 | –6.130 |
| (–1.89) | (–1.52) | (–1.53) | |
| Issuer | Yes | Yes | Yes |
| Facility | Yes | Yes | Yes |
| Macro | Yes | Yes | Yes |
| N | 574 | 159 | 415 |
| Pseudo R2 | 0.4251 | 0.4416 | 0.5145 |
The Too-Big-to-Fail Premium in Tier-2 Capital Bonds and Additional Tier-1 Capital Bonds Issuance: Using SIZE95 as Independent Variable
| (1) | (2) | (3) | |
|---|---|---|---|
| Full | Pre_Guidance | Post_Guidance | |
| SIZE95 | –23.50*** | –9.149 | –19.97*** |
| (–3.49) | (–0.75) | (–3.31) | |
| DSOE | –5.314 | –6.016 | –7.220* |
| (–1.19) | (–1.33) | (–1.76) | |
| Issuer | Yes | Yes | Yes |
| Facility | Yes | Yes | Yes |
| Macro | Yes | Yes | Yes |
| N | 574 | 159 | 415 |
| Pseudo R2 | 0.4195 | 0.4419 | 0.5114 |
5.4 Results of Regressions Adding Other Bank Risk Variables
Another potential problem of the study is that it might omit some bank risk variables. We calculate the Z score following an approach similar to that of Houston et al. (2010), and add lnZ to the independent variables. We still use median regressions and calculate heteroskedasticity-robust standard errors. Full sample and the subsamples before and after the announcement of the Guidance are used.
Regression results in Table 7 and Table 8 show that the coefficients of lnZ are not significant in the full sample and the subsample after the Guidance is announced, and the increase in pseudo R2 is not significant. This result indicates that the information contents in Z score might be covered by other variables. The coefficients of DSIB and GSIB in the full sample and the subsample after the Guidance was announced are statistically significant at 1% and negative. The results in this sub-section show that, ceteris paribus, in the full sample period, the issuance spreads of capital bonds issued by global systematically important banks are 27.2 bps lower than those of capital bonds issued by other banks. Ceteris paribus, in the full sample period, the issuance spreads of capital bonds issued by domestic systematically important banks are 29.7 bps lower than those of capital bonds issued by other banks. The previous conclusion is still robust.
The Too-Big-to-Fail Premium in Tier-2 Capital Bonds and Additional Tier-1 Capital Bonds Issuance: Using SIZE95 as Independent Variable
| (1) | (2) | (3) | |
|---|---|---|---|
| Full | Pre_Guidance | Post_Guidance | |
| DSIB | –29.69*** | –17.66 | –27.60*** |
| (–4.65) | (–1.09) | (–4.62) | |
| DSOE | –4.824 | –7.187 | –6.446 |
| (–1.20) | (–1.30) | (–1.55) | |
| lnZ | 4.524 | 18.90* | –1.150 |
| (0.58) | (1.76) | (–0.15) | |
| Issuer | Yes | Yes | Yes |
| Facility | Yes | Yes | Yes |
| Macro | Yes | Yes | Yes |
| N | 574 | 159 | 415 |
| Pseudo R2 | 0.4286 | 0.4558 | 0.5207 |
The Too-Big-to-Fail Premium in Tier-2 Capital Bonds and Additional Tier-1 Capital Bonds Issuance: Using SIZE95 as Independent Variable
| (4) | (5) | (6) | |
|---|---|---|---|
| Full | Pre_Guidance | Post_Guidance | |
| GSIB | –27.18*** | –45.06*** | –21.35*** |
| (–3.20) | (–3.27) | (–3.02) | |
| DSOE | –3.004 | –5.647 | –7.916* |
| (–0.65) | (–1.37) | (–1.95) | |
| lnZ | 4.733 | 15.63** | –9.966 |
| (0.47) | (2.03) | (–1.09) | |
| Issuer | Yes | Yes | Yes |
| Facility | Yes | Yes | Yes |
| Macro | Yes | Yes | Yes |
| N | 574 | 159 | 415 |
| Pseudo R2 | 0.4209 | 0.4750 | 0.5152 |
5.5 Robustness Test
This sub-section uses alternative benchmark of China Development Bank bond yields in calculating AUCSPREAD. We still conduct quantile (median) regressions and calculate heteroskedasticity-robust standard errors. Independent variables with variance inflation factor (VIF) higher than 10 are dropped.
In Table 9, regression 1 to 3 use the full sample with 574 observations. Pseudo R2 are between 0.345 and 0.408. The coefficients of DSIB are statistically significant at 1% level and negative. The coefficient of DSOE is not significant. The previous results are still robust.
The Too-Big-to-Fail Premium in Tier-2 Capital Bonds and Additional Tier-1 Capital Bonds Issuance: Full Sample Using CDB Bond Yields as Benchmark
| (1) | (2) | (3) | |
|---|---|---|---|
| DSOE | –2.316 | 4.602 | 0.0833 |
| (–0.49) | (0.94) | (0.02) | |
| DSIB | –79.38*** | –40.83*** | –37.68*** |
| (–15.27) | (–4.22) | (–4.20) | |
| Issuer | Yes | No | Yes |
| Facility | No | Yes | Yes |
| Macro | Yes | Yes | Yes |
| N | 574 | 574 | 574 |
| Pseudo R2 | 0.3450 | 0.3991 | 0.4077 |
Note: The dependent variable was AUCSPREAD calculated with the China Development Bank bond yields as the benchmark.
In Table 10, regression 4 and 5 use the subsamples before and after the announcement of the Guidance, which are consisted of 159 observations and 415observations respectively. Median regressions are conducted and heteroskedasticity-robust standard errors are calculated. The results show that the coefficients of GSIB are statistically significant at 5% level and negative. The previous conclusion is still robust.
The Too-Big-to-Fail Premium in Tier-2 Capital Bonds and Additional Tier-1 Capital Bonds Issuance: Subsamples Using CDB Bond Yields as Benchmark
| (4) | (5) | |
|---|---|---|
| Pre_Guidance | Post_Guidance | |
| DSOE | –3.129 | –0.571 |
| (–0.51) | (–0.13) | |
| GSIB | –18.45 | –15.57** |
| (–1.43) | (–2.17) | |
| Issuer | Yes | Yes |
| Facility | Yes | Yes |
| Macro | Yes | Yes |
| N | 159 | 415 |
| Pseudo R2 | 0.4407 | 0.4610 |
Note: The dependent variable was AUCSPREAD calculated with the China Development Bank bond yields as the benchmark.
6 Conclusions
This study examines the too-big-to-fail expectations in the primary market issuance spreads of commercial bank tier-2 capital bonds and additional tier-1 bonds in China. Using a sample of 574 issuances with total amount of 4.76 trillion RMB (749 billion USD), we conduct quantile regressions with the issuance spreads as the independent variable. The coefficients of DSIB and GSIB are negative and statistically significant at 1% and 5% levels respectively in the full sample, and are negative and statistically significant at 1% level in the post-guidance subsample. Ceteris paribus, the issuance spreads of capital bonds of systemically important banks are 15.2 bps to 19.7 bps lower than those issued by other banks. This might be due to the too-big-to-fail expectations and implicit guarantee of systemically important banks in the Chinese bond market, which narrow the primary market spreads of capital bonds issued by systemically important banks. To address the potential endogeneity issues, we use 90% quantile and 95% quantile of sample bank consolidated assets as proxies of systemic importance, and use China Development Bank bond yields to calculate issuance spreads. The conclusion is still robust.
The policy implications from this study are that financial reforms should be further promoted so as to lower the too-big-to-fail premium. The too-big-to-fail premium shows that there still exists an implicit guarantee for global systematically important banks. This too-big-to-fail premium might distort the market incentives and erode market discipline. Further financial reforms to eradicate rigid payments could lower or eliminate the too- big-to-fail premium. Capital bonds investors should enhance the analysis of the credit profile of the issuer and the relevant covenants so as to enhance market discipline.
Global systematically important banks in China should prepare recovery and resolution plans (RRPs) and update the RRPs periodically in coordination with the crisis management group and resolution authorities in home and host jurisdictions, ensure the continuity of critical functions in resolution, have sufficient total loss absorbing capacity (TLAC), and share the information necessary for resolution with authorities in home and host jurisdictions. Resolution plans should be credible and feasible for implementation, so as to lower the possibility of a government bailout.
Financial authorities should enhance the effective resolution mechanism and prepare for the bail-in of the global systematically important banks. Financial authorities should also work on cross-border coordination and avoid the too-big-to-fail problem.
References
Allen, L., & Golfari, A. (2023). Do CoCos Serve the Goals of Macroprudential Supervisors or Bank Managers?. Journal of International Financial Markets, Institutions and Money, 84, 101761.10.1016/j.intfin.2023.101761Search in Google Scholar
Acharya, V., Anginer, D., & Warburton, A. J. (2016). The End of Market Discipline? Investor Expectations of Implicit Government Guarantees. Working Paper.Search in Google Scholar
Ashcraft, A. B. (2008). Does the Market Discipline Banks? New Evidence from Regulatory Capital Mix. Journal of Financial Intermediation, 17, 543–561.10.1016/j.jfi.2007.05.003Search in Google Scholar
Avery R., Belton, T., & Goldberg, M. (1988). Market Discipline in Regulating Bank Risk: New Evidence from Capital Markets. Journal of Money, Credit and Banking, 1988, 20, 597–610.10.2307/1992286Search in Google Scholar
Balasubramnian, B., & Cyree, K. B. (2011). Market Discipline of Banks: Why Are Yield Spreads on Bank-issued Subordinated Notes and Debentures Not Sensitive to Bank Risks? Journal of Banking and Finance, 35, 21–35.10.1016/j.jbankfin.2010.07.015Search in Google Scholar
Belkhir, M. (2013). Do Subordinated Debt Holders Discipline Bank Risk-taking? Evidence from Risk Management Decisions. Journal of Financial Stability, 9, 704–719.10.1016/j.jfs.2012.01.001Search in Google Scholar
Bernanke, B. S. (2007). Financial Regulation and the Invisible Hand Speech. New York, NY: New York City Law School.Search in Google Scholar
Beyhaghi, M., D’Souza, C., & Roberts, G. S. (2014). Funding Advantage and Market Discipline in the Canadian Banking Sector. Journal of Banking and Finance, 48, 396–410.10.1016/j.jbankfin.2013.08.006Search in Google Scholar
Bianchi, M. L., & Tassinari, G. L. (2023). Extracting Implied Volatilities from Bank Bonds. Quantitative Finance, 23(7–8), 1177–1197.10.1080/14697688.2023.2226370Search in Google Scholar
Black, F., & Cox, J. C. (1976). Valuing Corporate Securities: Some Effects of Bond Indenture Provisions. Journal of Finance, 31, 351–367.10.1111/j.1540-6261.1976.tb01891.xSearch in Google Scholar
Bliss, R., & Flannery, M. J. (2002). Market Discipline in the Governance of US Bank Holding Companies: Monitoring vs. Influencing. European Finance Review, 6, 361–395.10.1023/A:1022021430852Search in Google Scholar
Bolton, P., Jiang, W., & Kartasheva, A. (2023). The Credit Suisse CoCo wipeout: Facts, misperceptions, and lessons for financial regulation. Journal of Applied Corporate Finance, 35(2), 66–74.10.1111/jacf.12553Search in Google Scholar
Collin-Dufresne, P., Goldstein, R. S., & Martin, J. S. (2001). The Determinants of Credit Spread Changes. Journal of Finance, 56, 2177–2207.10.1111/0022-1082.00402Search in Google Scholar
Covitz, D., & Downing, C. (2007). Liquidity or Credit Risk? The Determinants of Very Short-term Corporate Yield Spreads. Journal of Finance, 62, 2303–2328.10.1111/j.1540-6261.2007.01276.xSearch in Google Scholar
Driessen, J. (2005). Is Default Event Risk Priced in Corporate Bonds? Review of Financial Studies, 18, 164–195.10.1093/rfs/hhi009Search in Google Scholar
Evanoff, D. D., Jagtiani, J. A., & Nakata, T. (2011). Enhancing Market Discipline in Banking: The Role of Subordinated Debt in Financial Regulatory Reform. Journal of Economics and Business, 63, 1–22.10.1016/j.jeconbus.2010.07.001Search in Google Scholar
Fajardo, J., & Mendes, L. (2020). On the Propensity to Issue Contingent Convertible (CoCo) Bonds. Quantitative Finance, 20(4), 691–707.10.1080/14697688.2019.1685124Search in Google Scholar
Financial Stability Board. (2021). Resolution Report: Glass Half-full or Still Half-Empty.Search in Google Scholar
Financial Stability Board. (2010). Policy Measures to Address Systemically Important Financial Institutions.Search in Google Scholar
Flannery, M. J., & Sorescu, S. M. (1996). Evidence of Bank Market Discipline in Subordinated Debenture Yields: 1983–1991. Journal of Finance, 51, 1347–1377.10.1111/j.1540-6261.1996.tb04072.xSearch in Google Scholar
Imai, M. (2007). The Emergence of Market Monitoring in Japanese Banks: Evidence from Subordinated Debt Market. Journal of Banking and Finance, 31, 1441–1460.10.1016/j.jbankfin.2006.07.007Search in Google Scholar
Jaworski, P., Liberadzki, K., & Liberadzki, M. (2017). How does Issuing Contingent Convertible Bonds Improve Bank’s Solvency? A Value-at-Risk and Expected Shortfall Approach. Economic Modelling, 60, 162–168.10.1016/j.econmod.2016.09.025Search in Google Scholar
Kind, A., Oster, P., & Peter, F. J. (2022). The Determinants of Banks’ AT1 CoCo Spreads. European Financial Management, 28(2), 567–604.10.1111/eufm.12314Search in Google Scholar
Krishnan, C. N. V., Ritchken, P. H., & Thomson, J. B. (2005). Monitoring and Controlling Bank Risk: Does Risky Debt Help? Journal of Finance, 60, 343–379.10.1111/j.1540-6261.2005.00732.xSearch in Google Scholar
Kund, A. G., Hertrampf, P., & Neitzert, F. (2023). Bail-in Requirements and CoCo Bond Issuance. Finance Research Letters, 53, 103569.10.1016/j.frl.2022.103569Search in Google Scholar
Gorton, G., & Santomero, A. (1990). Market Discipline and Subordinated Debt. Journal of Money. Credit and Banking, 22, 119–128.10.2307/1992132Search in Google Scholar
Houston, J. F., Lin, C., Lin, P., & Ma, Y. (2010). Creditor Rights, Information Sharing, and Bank Risk Taking. Journal of Financial Economics, 96(3), 485–512.10.1016/j.jfineco.2010.02.008Search in Google Scholar
Li, P., Han, Y., Lin, S., & Qiao, T. (2020). Chinese Write-Down Bonds: Issuance and Bank Capital Structure. Quantitative Finance, 20(12), 2055–2065.10.1080/14697688.2020.1814034Search in Google Scholar
Li, P., Meng, H., & Yu, F. (2018). Chinese Write-Down Bonds and Bank Capital Structure.Quantitative Finance, 18(9), 1543–1558.10.1080/14697688.2018.1444559Search in Google Scholar
Livingston, M., Poon, W., & Zhou, L. (2018). Are Chinese Credit Ratings Relevant? A Study of the Chinese Bond Market and the Credit Rating Industry. Journal of Banking and Finance, 87, 216–232.10.1016/j.jbankfin.2017.09.020Search in Google Scholar
Nguyen, T. (2013). The Disciplinary Effect of Subordinated Debt on Bank Risk Taking. Journal of Empirical Finance, 23, 117–141.10.1016/j.jempfin.2013.05.005Search in Google Scholar
Niu, J. (2008). Can Subordinated Debt Constrain Banks’ Risk Taking? Journal of Banking and Finance, 32, 1110–1119.10.1016/j.jbankfin.2007.09.020Search in Google Scholar
© 2024 Le Yang, published by De Gruyter
This work is licensed under the Creative Commons Attribution 4.0 International License.
Articles in the same Issue
- Frontmatter
- Cumulative Tariff Cost Rate and Structure in the Global Value Chain: A Theoretical and Empirical Study
- Transportation Network Density, Domestic Market Integration and Excess Sensitivity of Household Consumption
- The Too-Big-to-Fail Premium in Tier-2 Capital Bonds and Additional Tier-1 Capital Bonds Primary Markets: Evidence from China
- Drivers of Portfolio Flows into Chinese Debt Securities Amidst China’s Bond Market Development
- The Connections among the US, China and Russia through Safe Assets
- Research on How the Digital Economy Contributes to Achieving A High-Level and Balanced Development of the Elderly Care Service Industry
Articles in the same Issue
- Frontmatter
- Cumulative Tariff Cost Rate and Structure in the Global Value Chain: A Theoretical and Empirical Study
- Transportation Network Density, Domestic Market Integration and Excess Sensitivity of Household Consumption
- The Too-Big-to-Fail Premium in Tier-2 Capital Bonds and Additional Tier-1 Capital Bonds Primary Markets: Evidence from China
- Drivers of Portfolio Flows into Chinese Debt Securities Amidst China’s Bond Market Development
- The Connections among the US, China and Russia through Safe Assets
- Research on How the Digital Economy Contributes to Achieving A High-Level and Balanced Development of the Elderly Care Service Industry
