نظم و انضباط بازار بانک: چرا میزان بازده در یادداشت های بانک صادر شده و اوراق قرضه به ریسک های بانک تابع حساس نیست ؟
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|24044||2011||15 صفحه PDF||سفارش دهید||13975 کلمه|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Journal of Banking & Finance, Volume 35, Issue 1, January 2011, Pages 21–35
The default risk sensitivity of yield spreads on bank-issued subordinated notes and debentures (SNDs) decreased after banks started issuing trust-preferred securities (TPS). The too-big-to-fail (TBTF) discount on yield spreads is absent prior to the LTCM bailout, but the size discount doubles after the LTCM bailout. Prior to TPS issuance and the LTCM bailout, SND yield spreads are sensitive to conventional firm-specific default risk measures, but not after the bailout. We find paradigm shift in determinants of yield spreads after the LTCM bailout. Yield spreads on TPS are sensitive to default risks and can provide an additional source of market discipline.
Our purpose in this study is to find reasons for the lack of default risk sensitivity in yield spreads on bank-issued subordinated notes and debentures (SNDs) documented in earlier studies. We also identify new determinants of yield spreads that are sensitive to default risks. We define yield spreads as the difference between the yield to maturity of risky debt and the yield to maturity of a risk-free debt of similar characteristics. Krishnan et al. (2005) point out that both yield spread levels and changes should reflect risk along the entire yield spread curve. When bank risks increase, we expect yield spreads on SNDs to increase and provide market discipline because SNDs are uninsured junior claims and are treated as Tier 2 capital for banks. Flannery and Sorescu (1996) and several later studies find that SND spreads are sensitive to bank risks, particularly after the enactment of the Federal Deposit Insurance Corporation Improvement Act (FDICIA) in 1991 because implicit guarantees, such as too-big-to-fail policy, were removed after FDICIA. Similarly, Imai (2007) shows that sensitivity of yield spreads increased after the removal of implicit guarantees in Japan. Studies prior to FDICIA find that yield spreads are not sensitive to bank risks (Avery et al., 1988 and Gorton and Santomero, 1990). A study by the Federal Reserve Board (FRB) concludes that SND spreads are less risk sensitive during the 1993–1997 period and that market discipline is weak (The Board of Governors, 1999). Krishnan et al. (2005) conclude that yield spreads do not provide useful signals for market discipline of banks because changes in yield spreads do not reflect changes in default risk during the 1994–1999 period, although yield spread levels do reflect firm-specific risks. Collin-Dufresne et al., 2001 and Krishnan et al., 2005 find that traditional default risk variables do not adequately explain yield spread changes on bonds also issued by non-financial firms. Driessen (2005) finds that the premium attributable to firm-specific default risk in a non-default situation is rather small even in yield spread levels. Hence, the bond market signals for banks through yield spreads on bank-issued SNDs in non-default situations is likely to be small for yield spread levels, and even smaller for yield spread changes due to changes in default risks. In contrast, Covitz and Downing (2007) find that default risk is a more important determinant of yield spreads than liquidity risk even in very short-term commercial paper that is inherently liquid, although Forte and Peña (2009) show that bond markets lag stock and credit default swap markets in price discovery. The findings of Covitz and Downing (2007) suggest omission of significant default risk factors in the case of long-term bonds. We identify three omitted factors that affect default risk: (i) The reduction in the default risk of SNDs since October 1996 after banks started issuing trust-preferred securities (TPS) (See Appendix A for details of TPS); (ii) the market’s enhanced perception that the government will bail out SND-issuing banks because they are “too-big-to-fail” (TBTF), particularly after the FRB brokered Long Term Capital Management (LTCM) bailout in September 1998; and (iii) ignoring idiosyncratic volatility as a proxy for firm-specific default risk. Our model includes all the variables of the Krishnan et al. (2005) model with the exception of bank-examination ratings. We show that the reduction in default risk due to these omitted factors could have been interpreted as a lack of risk sensitivity of yield spreads by The Board of Governors, 1999 and Krishnan et al., 2005. We also show that omitting the tax effects on yield spreads leads to errors in the decomposition of yield spreads. 1.1. Trust-preferred securities issuance TPS are junior capital that creates an additional buffer to SNDs from default risk and is treated on par with common stock. Both the ability to issue TPS and the actual issuance of TPS reduces the default risk of SNDs, and therefore should influence SND yield spreads. Banks started issuing TPS in October 1996 when the FRB approved TPS as Tier 1 capital. To provide a sense of the magnitude of additional default protection available to SND holders because of TPS issuance, we note that the outstanding SNDs issued by all banks at the end of 1998 was $102.8 billion, of which $100 billion was issued by the top fifty banks (The Board of Governors, 1999). In another study by The Federal Reserve System (2005) it is shown that SND-issuing banks also issued TPS totaling $28.1 billion between 1996 and 1999. In fact, over 800 banks issued TPS totaling $85 billion when including banks that had not issued SNDs. The results from the Federal Reserve studies indicate the importance and wide use of TPS because of its tax efficiency as the dividend paid on TPS is a tax deductible expense. We expect the default risk sensitivity of SNDs to decrease as the proportion of TPS increases in the bank’s capital structure. 1.2. Too-big-to-fail effects A second possible reason for the lack of sensitivity of SND yield spreads to bank risk is that large SND-issuing banks are often TBTF banks. On August 17, 1998, Russia declared a moratorium on its domestic and dollar debts, precipitating the near collapse of LTCM. The FRB intervened to bail out LTCM because the FRB feared financial contagion if LTCM were allowed to fail. Though the LTCM rescue efforts involved only the private sector, the FRB’s brokering of these efforts reinforced the TBTF policy, in spite of regulations such as FDICIA to improve the market discipline of banks, and sent signals to the markets that the FRB will bail out all systemically-important financial firms, whether they are FDIC-insured depository institutions or otherwise. We expect an increase in the size discount on the yield spreads after the LTCM bailout since larger institutions are more likely to be bailed out. Another indication of the importance of the TBTF effect is that in their assessments for rating bank-issued bonds, credit rating agencies such as Moody’s and Standard & Poor’s (S&P) consider the possibility of direct government support in rating banks believed too-big-to-fail. 1.3. Idiosyncratic volatility effects A third possible reason for the perceived lack of risk sensitivity of SND spreads is ignoring firm-level volatility. Idiosyncratic volatility increases the risk of default for bondholders. During the 1990s, yield spreads on bonds increased when stock returns were also increasing. Campbell et al. (2001) find that market volatility did not increase as much as idiosyncratic volatility. Campbell and Taksler (2003) argue that the increase in idiosyncratic volatility explains the divergence in stock and bond market performance. Campbell and Taksler (2003) find that equity volatility is an important determinant of yield spread on bonds and the effect is relatively higher for firms with higher long-term debt-to-total assets and total debt-to-equity ratios. Stiroh (2006) shows that volatility of bank stocks increases because risk has shifted to off-balance sheet activities and the income statement, such as non-interest income. Since banks are inherently highly levered, have several off-balance sheet risks, and the proportion of volatile non-interest income-to-total revenue is increasing over time, ignoring idiosyncratic volatility could lead to errors in estimation of the default risk component of yield spreads. Güntay and Hackbarth (2010) find that yield spreads are higher for firms with higher dispersion of analysts’ forecast, and changes in forecast dispersion predict the changes in yield spread. Analysts’ forecast dispersion appears to proxy for future cash flow uncertainty. Other studies in asset-pricing suggest that both systematic risks and idiosyncratic risks, as measured by the residual variance of an asset-pricing model, matter in the cross-section and time-series of stock returns (Malkiel and Xu, 1997, Malkiel and Xu, 2001 and Goyal and Santa-Clara, 2003). Many researchers including Krishnan et al. (2005) use the Chicago Board Options Exchange (CBOE) Volatility Index (VIX) as a proxy for firm volatility. Becker et al. (2009) find that VIX captures prior and future jump risk and has more information when compared to econometric model-based forecasts. However, Campbell and Taksler (2003) show that market volatility measures such as the standard deviation of index returns did not capture firm-specific default risks as well as total volatility measures such as the standard deviation of stock returns. Although the VIX index is highly correlated with the standard deviation of stock returns (sample correlation = 0.5025), we find that the standard deviation of stock return is a better proxy for volatility than the VIX index. 1.4. Tax effects Omission of the tax effect on yield spread can lead to errors in the decomposition of yield spreads attributable to factors such as liquidity, term structure, market risk, tax, and default risk. For bondholders, interest income on corporate bonds is taxable, but interest from Treasuries is not taxable. Elton et al. (2001) show that bondholders impound the tax cost in the yield spreads on corporate bonds. Driessen (2005) finds that the tax premium constitutes nearly one-third of the yield spread. We observe that the tax effect has been omitted in the study of Krishnan et al. (2005), and we add a tax factor to capture this effect found important in other studies of the decomposition of yield spreads. Following Campbell and Taksler (2003), we use coupon rates as a proxy for tax effects since higher coupons create a larger tax burden. 1.5. Hidden leverage We examine whether or not hidden leverage is a determinant of SNDs yield spreads. Bank leverage can increase or decrease through off-balance sheet activities depending on the contingent liabilities from changing market conditions. The LTCM crisis first highlighted the risks from derivative instruments for several financial institutions and for the entire financial system. Using on-balance sheet leverage alone may not be a good proxy for potential default risks. We find non-traditional risk measures, such as off-balance sheet risks, are one of the significant determinants of yield spreads, particularly after the LTCM bailout when traditional risk measures such as on-balance sheet leverage is not a determinant. We find that determinants of yield spreads are dynamic. Determinants of yield spreads seem to vary with additional information on sources of default risk, which poses additional challenges for modeling yield spreads. 1.6. Yield spread on trust-preferred securities Because TPS are junior claims to SNDs and are potentially more risk sensitive, we also investigate whether yield spreads on TPS provide useful market signals to monitor bank risks. We use TPS issuance in two ways: (1) to see if TPS could affect SND yield spreads, and (2) whether or not TPS yield spreads themselves are useful for market discipline. We find TPS yield spreads can be used as a market signal to identify potentially weak banks. 1.7. Contributions to the literature Following Krishnan et al. (2005), we extract the entire yield spread curve for each sample firm and for each transaction date for SNDs. We also examine the sensitivity of both the levels and the changes in yield spreads to firm-specific risk measures, after controlling for market, liquidity, and tax factors. We find that default risk sensitivity of yield spreads on bank-issued SNDs decreased after banks started issuing TPS. Prior to TPS issuance and the LTCM bailout, SND yield spreads levels are sensitive to conventional firm-specific default risk measures, but after the LTCM bailout even the levels of yield spreads are not sensitive to conventional default risk measures. We find no evidence of the TBTF discount on yield spreads prior to the LTCM bailout. However, the size discount on yield spreads doubles after the LTCM bailout indicating paradigm changes in the determinants of bank-issued SND yield spreads after the LTCM bailout. We find that the determinants of yield spread changes are jointly significant for the pre-TPS period. However, we also find that several firm-specific default risk proxies are not significant factors for yield spread changes even for the pre-TPS period. Our results support the findings of The Board of Governors (1999) since the levels of yield spreads are less sensitive to conventional measures of bank risks after TPS issuance. Our results are not only consistent with Krishnan et al. (2005), but are stronger since we find that even yield spread levels are not default risk sensitive after the ability to issue TPS and the LTCM bailout. We also directly study TPS yield spreads since they should be more sensitive to bank risks than SND. TPS are relatively new financial instruments and the secondary market for TPS is not deep for our sample period, thus we examine TPS yield spreads at the time of issue. We find that, in general, the yield spreads on primary issues of TPS are sensitive to conventional balance sheet risk measures after controlling for market and liquidity variables. In other words, yield spreads on TPS are a good candidate for monitoring banks. The rest of this paper is organized as follows. We briefly review related studies in Section 2. We develop our hypotheses in Section 3. We describe our sample in Section 4. We present our results in Section 5 and conclude in Section 6.
نتیجه گیری انگلیسی
We examine institutional investors’ secondary market transactions in bank-issued subordinated notes and debentures (SNDs) for the period 1994–1999. When we examine the full sample, we find that yield spread levels reflect firm-specific default risks. We find that idiosyncratic volatility is a better proxy for default risk compared to market volatility and omitting tax effects on yield spreads can lead to measurement errors. As the Federal Reserve Board (FRB) intervention in the LTCM bailout signaled the return of implicit government guarantees, we examine the bond transactions for the pre-LTCM bailout and post-LTCM bailout periods separately. We do not find evidence of a too-big-to-fail (TBTF) discount on yield spreads prior to the LTCM bailout. We find an order-of-magnitude reduction in the yield spreads on SNDs due to the size of the bank in general, without being identified as a TBTF bank, which may have been interpreted as lack of risk sensitivity in past studies. This finding implies that despite FDICIA, bond markets strongly believe in TBTF policy. After the LTCM bailout, even yield spread levels on SNDs do not reflect the levels of conventional measures of firm-specific risks. We also find paradigm changes in the determinants of yield spreads after the LTCM bailout. As bond market participants realized the risks and hidden leverage due to derivatives and off-balance sheet items during and after the LTCM crisis, off-balance sheet exposures became a significant determinant of yield spreads. Investors seemed to believe that a higher return on assets was possible only with increased risk, so the coefficient of return-on-assets (ROA) became significantly positive. Traditional risk measures such as leverage, and loan risks became irrelevant as determinants of yield spreads because markets seem to perceive that these measures do not reflect the true risk of the firm. Our findings have implications both for the models of bond yield spreads and models of credit risk. Banks started issuing trust-preferred securities (TPS) beginning in October, 1996. TPS offer an additional buffer to SNDs from default risk. To isolate the effects of TPS issuance and TBTF effects on SND yield spreads, we examine the bond transactions separately for the pre-TPS period, post-TPS period, post-TPS–pre-LTCM bailout period, and post-TPS–post-LTCM bailout period. We find that after banks started issuing TPS, firm-specific risk factors were less significant determinants of yield spreads on bank-issued SNDs. This finding supports our hypothesis that the default risk reduction on SNDs due to TPS might have been interpreted as lack of risk sensitivity of SND yield spreads. We also find that conventional firm-specific risks are significant factors in explaining SND yield spreads prior to TPS issuance by banks. We find that the determinants of bid-yield spreads and ask-yield spreads are different. Together, the levels of these two yield spreads could indicate current risks and potential future problems. We separately estimate TPS yield spreads to evaluate their usefulness for the market discipline of banks. Because TPS are junior claims to SNDs, yield spreads on TPS should be more risk sensitive. We find that the yield spreads on primary issues of TPS are sensitive to conventional balance sheet risk measures. Our findings imply that TPS yield spreads could be used as an effective market signal for monitoring banks. We conclude that debt markets are sensitive to changes in default risks and assess risks in the absence of full information. However, the determinants of yield spreads are time-varying, depending on market conditions and available information. The policy implications of our findings are several. Our results show that market discipline is less effective, particularly for large banks, after the FRB’s intervention in the LTCM bailout, which signaled the return of implicit guarantees in spite of the Federal Deposit Insurance Corporation Improvement Act (FDICIA) of 1991. Thus, the benefit of regulation to improve market discipline of banks is greatly diminished after the LTCM bailout in 1998, much earlier than the subprime crisis beginning in 2007. To restore market discipline, proposed financial regulation must enable orderly resolution of large bank failures without any intervention from the government. The disclosure requirements for off-balance sheet items need to be improved to enhance market discipline. The proposed transparency and standardization of derivatives instruments should help enhance estimation of risk by market participants. Although we believe we have extended the bank market discipline literature, our results indicate the difficulty in modeling bank risk using yield spreads and that there is much work left to do.