Information content in CDS spreads for equity returns

Int. Fin. Markets, Inst. and Money 30 (2014) 55–80

Contents lists available at ScienceDirect

Journal of International Financial Markets, Institutions & Money j o ur na l ho me pa ge : w w w . e l s e v i e r . c o m / l o c a t e / i n t f i n

Information content in CDS spreads for equity returns Peipei Wang a,∗, Ramaprasad Bhar b a b

School of Accounting Economics and Finance, Deakin University, Burwood, Victoria 3125, Australia The Australian School of Business, The University of New South Wales, Sydney 2052, Australia

a r t i c l e

i n f o

Article history: Received 28 June 2013 Accepted 20 January 2014 Available online 1 February 2014 JEL classification: G12 Keywords: Credit default swap Equity return Information spillover Financial crisis Price discovery

a b s t r a c t This study focuses on the information spillover between the credit protection returns and equity returns for 252 United States firms between 2004 and 2010. There is significant information flow from the equity market to the credit default swap (CDS) market under turmoil conditions for investment-grade firms and the reverse is true for non-investment-grade firms. There is also strong evidence of extra information contained in the positive credit protection return one day for the equity market the following day. The behaviour of the markets around credit announcement day leads us to believe that there may be informed trading in the CDS market for high-credit-rating firms. © 2014 Elsevier B.V. All rights reserved.

1. Introduction The no arbitrage argument suggests that, if information is revealed in either the equity or the credit default swap (CDS) market, the other market adjusts passively, either simultaneously or with a delay. Thus, the two markets achieve pool equilibrium. Moreover, hedge funds and private equity firms are active in a variety of trading strategies (e.g., capital structure arbitrage) that attempt to arbitrate across equity and CDS markets. However, short-horizon pricing discrepancies across a firm’s equity and credit markets are commonly documented. This indicates a lack of temporary integration between the two markets. Kapadia and Pu (2012) conclude that this discrepancy is related to impediments to arbitrage.

∗ Corresponding author. Tel.: +61 3 92446906. E-mail addresses: [email protected] (P. Wang), [email protected] (R. Bhar). 1042-4431/$ – see front matter © 2014 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.intfin.2014.01.005

56

P. Wang, R. Bhar / Int. Fin. Markets, Inst. and Money 30 (2014) 55–80

Hilscher et al. (2013) show that informed traders are primarily active in the equity market. However, there is a general absence of informed traders in the CDS market because the higher bid-ask spread in that market deters them. Despite the high percentage of uninformed traders in the CDS market, there is high sensitivity of credit protection returns1 to information. The existence of information spillover is an important topic in asset pricing. The information revelation, or price discovery and co-movement of the CDS and other financial markets, has been studied, along with the development of the CDS market (e.g., Hull et al., 2005; Zhang et al., 2005; Longstaff et al., 2005; Abid and Naifar, 2006; Dupuis et al., 2009). Recent research has examined and documented different lead – lag patterns of cross-market information flows; however, no consensus has yet been reached. For example, Acharya and Johnson (2007) find significant incremental information revelation in the CDS market, but only for negative credit news and for entities that subsequently experience adverse shocks. Norden and Weber (2009) study the co-movement of CDS, bond and stock markets between 2000 and 2002. They apply a VAR framework on CDS changes, bond-spread changes and stock returns and find that the stock market generally leads the CDS and bond markets. The strength of co-movement between CDS changes and stock returns increase as the credit quality decreases. In comparison to bond markets, the CDS market is more responsive to the stock market and plays a more important price discovery role, although there is no clear lead of the stock market with respect to the CDS market and vice versa. However, Hilscher et al. (2013) provide evidence that equity returns lead credit protection returns at daily and weekly intervals, while credit protection returns do not lead equity returns. In this study, we analyse the information flow between the equity and CDS markets, with particular focus on information content of credit protection returns using daily data. We explore both contemporaneous and lagged data. Our dataset covers 252 United States (US) non-financial and nonpublic administration firms from 1 July 2004 to 31 August 2010. At horizons of up to five trading days, credit protection returns respond negatively to lag equity returns, with the significance level tending to depend on the lag length. Nevertheless, the sensitivity of credit protection returns to equity returns is greater for investment-grade firms. For non-investment-grade firms, credit protection returns are more sensitive to equity returns in normal market conditions. The predictability of credit protection returns based on equity returns is low when the market is normal, particularly for issuers with a lower credit rating. This result suggests independence of the two markets, particularly when arbitrage opportunities are rare. The strategy is also not profitable, as indicated by the wide bid-ask spread when trading CDS contracts. It may be somewhat arbitrary to separate the sample into two sub-periods and label them as ‘normal market’ or ‘turmoil market’ conditions. To address this, we explore the response of credit protection returns to equity returns based on market performance (market level) or firm performance (firm level). We achieve this by including one lag credit protection return and one lag equity return to signal the market and checking if extra information can be obtained from the CDS market. Our empirical results show that credit protection returns are more sensitive to contemporaneous equity return if credit deterioration is detected in the CDS market on the previous day. If there is deterioration in the equity market on one trading day, there is an additional increase in the credit protection return for a typical investment-grade issuer the following day. There is no discernible response of the equity return to contemporaneous and lagged credit protection returns for a typical investment-grade issuer. However, there is a weak but statistically and economically meaningful response from the equity return to the one lag credit protection return for non-investment-grade firms. The behaviour of the equity return during the turmoil market is different. It is sensitive to contemporaneous credit protection returns for all firms, and this is statistically significant. Thus, the importance of lagged credit protection returns appears to be conditional on the state of the market. Next, we check the response between two markets around credit rating announcement day to explore the information content in the changing credit condition further. When the announced credit rating is lower (i.e., there is a downgrade), there is no significant response from credit protection

1

The credit protection return is approximated well by the percentage change in the credit spread (see Hilscher et al., 2013).

P. Wang, R. Bhar / Int. Fin. Markets, Inst. and Money 30 (2014) 55–80

57

returns to contemporaneous and lagged equity returns for investment-grade issuers. However, there is a significant negative response from one-day-ahead equity returns to credit protection returns on announcement day, particularly during the turmoil period. This provides evidence for some informed trading in the CDS market on investment-grade issuers. The credit rating upgrade announcement has an interesting effect. We find significant positive responses from one-day-ahead equity returns to credit protection returns on credit rating announcement day during normal market conditions. However, this may be due to an overreaction. If the credit rating announcement is a surprise and an upgrade, then one-day-ahead credit protection returns have a significant negative response to equity returns on announcement day. This implies that the equity market can catch the ‘up’ surprise more quickly. Under similar conditions, the oneday-ahead equity return responds positively to the credit protection return on announcement day. This indicates that the equity return may have overreacted earlier. From here, the paper is organised as follows. Section 2 provides a brief introduction to the CDS market and a description of our datasets. Section 3 presents the response from credit protection returns to equity returns and Section 4 documents the response in the other direction. Section 5 examines the response around credit announcement day and Section 6 concludes the paper. 2. CDS market and datasets description Introduced in 1997 by JP Morgan, CDS contracts have only been available for a large number of firms since the early 2000s. They have become the primary tool for managing default risks that exist in bank loans and corporate bonds. As the CDS market is an over-the-counter market and all participants are institutional, it is generally recognised that participants in the CDS market are more sophisticated and informed. Academic research indicates that CDS trading serves as a price discovery process and CDS spread is informative with respect to credit risks. For example, Blanco et al. (2005) find that the CDS market is potentially much more efficient than the bond market in signalling the creditworthiness of borrowers in the short term. Acharya and Johnson (2007) show that information flow from the CDS to the stock market becomes stronger when a CDS obligor’s credit condition deteriorates, and flow strength is related positively to the number of obligor’s relationship banks. Some government regulators have advocated a ban on so-called ‘naked’ long position in CDS contracts since the recent global financial crisis (GFC) because of the possibility that informed traders may trade private ‘credit deterioration’ information in the CDS market to maximise expected profits. The effectiveness of this regulation depends on whether the CDS market is a place of information revelation. Credit protection is conceptually similar to a put option on the underlying bond. Thus, holding credit protection is analogous to going long on a put option on the underlying asset. Based on BlackScholes’ formula for a European put option, the magnitude of the sensitivity of the put option return to return of the underlying asset becomes larger as the put option moves further out of the money. Thus, for firms with a higher credit rating, which tend to be far from default, such puts are further out of the money and, thus, are more sensitive to information. The percentage change in the credit spread approximates the return on holding credit protection well, as suggested by Berndt and Obreja (2010) and Hilscher et al. (2013). We obtain daily five-year US CDS levels on senior, unsecured debt from Bloomberg. We also collect daily equity market data, quarterly accounting data and monthly credit ratings for Standard & Poor (S&P) from the Center for Research in Security Prices (CRSP) and COMPUSTAT. A total of 252 US nonfinancial and non-public administration firms have available CDS and CRSP returns and monthly S&P rating information for the period 1 July 2004 to 31 August 2010. Table 1 reports summary statistics. The whole sample period is divided into two sub-periods (before and after the end of 2007) to enable us to determine the GFC effect and compare results with Hilscher et al. (2013). Across the study period, the average CDS level for sample firms is 209 basis points (bps) and the average credit rating is BBB. The large variance in credit quality is reflected in the interquartile range of 48–183 bps for credit spreads and ratings of BB+ to A−. In comparing the summary statistics for the separate sub-periods, it is not surprising that the average CDS level jumps from 97 bps before 2007 to 355 bps after 2007. The interquartile range of the average CDS level is much wider after 2007 (73–309 bps) than it is during

58

P. Wang, R. Bhar / Int. Fin. Markets, Inst. and Money 30 (2014) 55–80

Table 1 Summary statistics. CDS Level Panel A: (Period: 2004/07/01–2010/08/31) 209 Mean 48 25th Percentile 183 75th Percentile 201 Std. Dev. Panel B: (Period: 2004/07/01–2007/12/31) 97 Mean 29 25th Percentile 108 75th Percentile 40 Std. Dev. Panel C: (Period: 2008/01/01–2010/08/31) 355 Mean 73 25th Percentile 309 75th Percentile 218 Std. Dev.

Capitalization

Leverage

Rating

18,885 3,800 20,757 4,873

0.31 0.16 0.44 0.08

9.0 7.1 10.7 0.8

BBB A− BB+

19,981 4329 21,547 3787

0.28 0.14 0.38 0.05

8.8 7.0 10.3 0.5

BBB A− BBB−

17,442 3231 20,940 3583

0.37 0.19 0.51 0.06

9.4 7.3 11.0 0.5

BBB A− BB+

Table reports summary statistics for firms with available data from 1 Jul 2004 to 31 Aug 2010. We first calculate average and standard deviation for CDS level, Capitalization, Leverage and Rating across time for each firm. Then we take mean, 25th & 75th percentile across firms. S&P credit ratings are converted to numbers based on the rule as following: we assign 1 to AAA, 2 for AA+, and so on until 26 for D. We divide whole sample period as two sub-samples. Panel A, B and C represent summary statistics for whole sample period, sub-period before the end of 2007 and sub-period after 2007 respectively.

the prior sub-period (29–108 bps). Compared with CDS levels as a signal of firms’ credit health, the S&P rating is an inert measure. CDS contracts predominantly exist for large firms. Table 1 shows that, across the whole sample period, the average firm has a market capitalisation of US$19 billion, and 75 percent of firms have market capitalisation greater than US$2 billion. The leverage is computed as the ratio of the book value of debt to the sum of the book value of debt and market capitalisation. The book value of debt is defined as the sum of long-term debt and debt in current liabilities. The leverage has an interquartile range of 16 percent to 44 percent. With the equity value of most companies shrinking during the GFC period, the average market capitalisation declined and average leverage increased. Table 2 reports statistics of percentage bid-ask spreads for both CDS and stock markets across the whole period and sub-sample periods. The percentage bid-ask spread is defined as the ratio of bidask spread to the middle point of the bid and ask level. The average CDS percentage bid-ask spread is 1,189 bps and the average equity percentage bid-ask spread is 11 bps, which represents a ratio of 108. This is one possible reason for separated market equilibrium because the market maker sets CDS bid-ask spread sufficiently high to deter informed traders from trading in the CDS market. This ratio declines significantly from 160 (1436 vs. 9 bps) before 2007 to 67 (867 vs. 13 bps) after 2007. Table 2 also reports the percentage bid-ask spread for two groups of firms based on their S&P longterm credit rating at the beginning of each sample period.2 Firms with S&P long-term credit ratings of at least BBB− are classified as investment-grade firms, and those with ratings below BBB− are classified as non-investment-grade firms. The CDS percentage bid-ask spreads are considerably wider for investment-grade firms. As CDS contracts for investment-grade firms would be more sensitive to information (and in the interest of informed traders), the higher CDS percentage bid-ask spreads of those contracts support the argument that informed traders might be deterred from participating in the CDS market because of the high trading cost. The average CDS percentage bid-ask spread for investment firms after 2007 is only 57 percent of the previous level.3

2 We use both an averaged credit rating and credit rating at the beginning of the sample period to categorise firms as investment or non-investment grade. The results, based on these two classifications, generate more or less similar conclusions and do not change our key results. We present only results using credit rating at the beginning of the sample period to classify firms because of limited space. 3 For firms with an S&P long-term credit rating of A− and above, the average percentage bid–ask spread after 2007 was less than half of that in the prior sub-period.

P. Wang, R. Bhar / Int. Fin. Markets, Inst. and Money 30 (2014) 55–80

59

Table 2 Percentage bid-ask spread. Overall

Investment

252 No. firms Panel A: (Period: 2004/07/01–2010/08/31) CDS 1189 Mean 628 Std. Dev. Equity 11 Mean 19 Std. Dev. Panel B: (Period: 2004/07/01–2007/12/31) CDS 1436 Mean 638 Std. Dev. Equity 9 Mean 13 Std. Dev. Panel C: (Period: 2008/01/01–2010/08/31) CDS 865 Mean 280 Std. Dev. Equity 13 Mean 22 Std. Dev.

Non-investment

194

58

1312 699

779 389

9 16

16 26

1625 714

804 384

8 12

13 17

928 274

670 298

10 17

24 38

Table reports mean and standard deviation of both CDS percentage bid-ask spread and equity percentage bid-ask spread for firms with available data from 1 Jul 2004 to 31 Aug 2010. Firms are segregated into 2 credit categories: investment and Noninvestment grade based on S&P long-term credit rating at the beginning of each sample period. Firms with S&P long-term credit ratings of at least BBB− are classified as investment-grade firms, and those with ratings below BBB− are classified as non-investment-grade firms. The percentage bid-ask spread is defined as the ratio of bid-ask spread to the middle point of bid and ask price/level. We first calculate average and standard deviation for percentage bid-ask spread across time for each firm. Then we report cross-firm average for whole sample period and two sub-sample periods. Number of firms in each credit category is reported in row 2.

3. Response from credit protection returns to equity returns We begin examining the response from credit protection returns to equity returns of the same firm by running the following specification: Rcds,i,t = ˇ0 +

5  p=1

ˇ1,p Rcds,i,t−p +

10 

ˇ2,s Rstock,i,t−s + ei,t

(1)

s=0

where the dependent variable, Rcds,i,t , is the credit protection return for firm i on day t. The independent variables include Rstock,i,t−s , the equity return for firm i for day t − s where s takes from 0 to 10 days, and Rcds,i,t−p where p varies from one to five, to control firm fixed effect. The coefficient ˇ2,0 captures the contemporaneous response from credit protection returns to equity returns. As predicted by the model, we expect ˇ2,0 to be significantly negative if two markets co-move in the right direction. The coefficient ˇ2,s s, where s varies from one to 10, captures the response from credit protection returns to lags of equity returns. We expect that, if there is delayed information from the equity to the CDS market, ˇ2,s will be significant (negative), and, the older the return is, the less information it contains and the less significant the coefficient. For example, the coefficient ˇ2,1 will be significantly negative if there is delayed information from the previous equity market. We estimate the same model for two credit rating categories and three sample periods, respectively. The credit categories are investment and non-investment-grade firms. The sample periods are 1 July 2004 to 31 August 2010, 1 January 2001 to 31 December 2007 and 1 January 2008 to 31 August 2010. The regressions for these periods are reported in Tables 3A, 3B and 3C, respectively. All regressions presented in this study use panel data. It is well known that, if residuals are correlated across observations; for example, the residual of a given firm may be correlated across time

60

Lag

0

1

Panel A: investment −0.2937 −0.2165 (−9.62)*** (−12.95)*** −0.2937 −0.5102 Panel B: non-investment −0.1989 −0.1785 (−9.05)*** (−9.10)*** −0.3773 −0.1989

2

3

4

5

6

7

−0.0649 (−2.21)** −0.5751

−0.0144 (−0.83) −0.5895

−0.0209 (−1.02) −0.6104

−0.0124 (−0.72) −0.6228

0.0154 (1.00) −0.6074

0.0077 (0.44) −0.5997

−0.0809 (−3.33)*** −0.4582

−0.0359 (−0.57) −0.4941

−0.0877 (−2.04)** −0.5818

−0.0105 (−0.74) −0.5923

−0.0382 (−2.15)** −0.6306

−0.0075 (−0.23) −0.6381

9

10

R2

0.0139 (0.53) −0.5858

−0.0298 (−1.56) −0.6156

−0.0498 (−2.82)*** −0.6653

0.0125

−0.0206 (−1.13) −0.6586

−0.0114 (−0.78) −0.6700

0.0057 (0.26) −0.6643

0.0004

8

Table reports coefficients, cumulative coefficients and t-statistics from regression (Eq. (1)) of daily credit protection returns on its 5 lags (not reported), contemporaneous (lag = 0) and lagged (lag = 1 to lag = 10) daily equity returns between 1 Jul 2004 and 31 Aug 2010. The reported t-statistics in parentheses are based on adjusted standard error clustered by both date and firm. The regression is done separately on investment grade firms and non-investment grade firms, and results are reported in Panels A and B. Each firm is assigned to a particular credit rating category based on its S&P long-term credit rating at the beginning of each sample period. * , ** , and *** represent 10%, 5%, and 1% significance levels, respectively.

P. Wang, R. Bhar / Int. Fin. Markets, Inst. and Money 30 (2014) 55–80

Table 3A Response of the credit protection return to the equity return between 1 Jul 2004 and 31 Aug 2010.

Lag

0

1

Panel A: investment −0.2381 −0.2047 (−3.64)*** (−6.03)*** −0.2381 −0.4429 Panel B: non-investment −0.3210 −0.2289 (−2.95)*** (−5.25)*** −0.3210 −0.5499

2

3

4

5

6

7

8

9

10

R2

0.0000 (0.00) −0.4429

−0.0283 (−0.88) −0.4713

0.0028 (0.06) −0.4684

−0.0140 (−0.32) −0.4824

0.0268 (0.80) −0.4557

0.0652 (1.84)* −0.3904

0.0852 (0.83) −0.3053

−0.0312 (−0.65) −0.3364

−0.0598 (−1.10) −0.3962

0.0117

−0.1635 (−1.98)** −0.7134

−0.0679 (−0.26) −0.7812

−0.2947 (−1.55) −1.0759

0.0369 (0.92) −1.0389

−0.1222 (−1.87)* −1.1611

−0.0739 (−0.56) −1.2350

−0.0916 (−1.72)* −1.3266

0.0022 (0.09) −1.3244

0.0737 (0.93) −1.2507

0.0003

Table reports coefficients, cumulative coefficients and t-statistics from regression (Eq. (1)) of daily credit protection returns on its 5 lags (not reported), contemporaneous (lag = 0) and lagged (lag = 1 to lag = 10) daily equity returns between 1 Jul 2004 and 31 Dec 2007. The reported t-statistics in parentheses are based on adjusted standard error clustered by both date and firm. The regression is done separately on investment grade firms and non-investment grade firms, and results are reported in Panel A and Panel B. Each firm is assigned to a particular credit rating category based on it S&P long-term credit rating at the beginning of each sample period. * , ** , and *** represent 10%, 5%, and 1% significance levels, respectively.

P. Wang, R. Bhar / Int. Fin. Markets, Inst. and Money 30 (2014) 55–80

Table 3B Response of the credit protection return to the equity return between 1 Jul 2004 and 31 Dec 2007.0.

61

62

Lag

0

1

Panel A: investment −0.3455 −0.2346 (−9.21)*** (−11.38)*** −0.3455 −0.5800 Panel B: non-investment −0.1690 −0.1428 (−10.07)*** (−8.75)*** −0.1690 −0.3118

2

3

4

5

6

7

8

9

10

R2

−0.0895 (−3.55)*** −0.6696

−0.0184 (−0.77) −0.6879

−0.0530 (−2.08)** −0.7409

−0.0493 (−2.11)** −0.7902

−0.0043 (−0.19) −0.7945

−0.0159 (−0.70) −0.8104

−0.0102 (−0.49) −0.8206

−0.0294 (−1.23) −0.8500

−0.0550 (−2.60)*** −0.9050

0.0479

−0.0435 (−3.58)*** −0.3554

−0.0129 (−0.97) −0.3682

−0.0213 (−1.74)* −0.3896

−0.0258 (−2.33)** −0.4153

−0.0143 (−1.42) −0.4296

0.0013 (0.10) −0.4283

−0.0038 (−0.32) −0.4320

−0.0294 (−1.95)* −0.4614

−0.0216 (−1.83)* −0.4830

0.0697

Table reports coefficients, cumulative coefficients and t-statistics from regression (Eq. (1)) of daily credit protection returns on its 5 lags (not reported), contemporaneous (lag = 0) and lagged (lag = 1 to lag = 10) daily equity returns between 1 Jan 2008 and 31 Dec 2010. The reported t-statistics in parentheses are based on adjusted standard error clustered by both date and firm. The regression is done separately on investment grade firms and non-investment grade firms, and results are reported in Panel A and Panel B. Each firm is assigned to a particular credit rating category based on its S&P long-term credit rating at the beginning of each sample period. * , ** , and *** represent 10%, 5%, and 1% significance levels, respectively.

P. Wang, R. Bhar / Int. Fin. Markets, Inst. and Money 30 (2014) 55–80

Table 3C Response of the credit protection return to the equity return between 1 Jan 2008 and 31 Aug 2010.

P. Wang, R. Bhar / Int. Fin. Markets, Inst. and Money 30 (2014) 55–80

63

(firm effect) or the residual of a given time may be correlated across firms (time effect), OLS standard errors can be biased. To account for any such bias, we estimate standard errors clustered by both time and firm, and report t-statistics based on these two-way clustered standard errors, as suggested by Petersen (2009). Table 3A reports estimates of the coefficients of ˇ2,s , with cumulative coefficients shown in the last row in each panel. For different credit rating categories, credit protection returns are negatively contemporaneously related to equity returns. Decreasing the equity return by 1 percent is associated with a contemporaneous increase in the credit protection return of 29 bps for an investment-grade firm and 20 bps for a non-investment-grade firm. The greater sensitivity of credit protection returns for an investment-grade issuer is consistent with Merton’s (1974) insight that the out-of-money put option is more sensitive to the value of the underlying firm. Table 3A also reports the sensitivities of credit protection returns to lagged equity returns of one to 10 days. At horizons of up to five trading days, credit protection returns respond negatively to lagged equity returns; however, negative responses with a statistical significance above 1 percent are mainly found in lags of up to two days. The sensitivity of credit protection returns to equity returns is greater for investment-grade firms, which confirms the result from contemporaneous response. The market before the end of 2007 can be considered normal, while the market from 2008 to the end of our sample period is in turmoil. The results from the two sub-samples (Tables 3B and 3C) show the difference in linkage between equity markets and CDS markets under different macroeconomies. In the normal market, the sensitivities of credit protection returns to equity returns are higher for non-investment firms. For example, in the first sub-period (Table 3B) of up to five days, the cumulative response from credit protection returns to a 1 percent decrease in equity returns is 48 bps for investment-grade firms and 104 bps for non-investment-grade firms. The corresponding contemporaneous responses are 24 bps and 32 bps. The pattern is exactly opposite to that reported by Hilscher et al. (2013). Although equity returns can explain less of the variation in credit protection returns for non-investment-grade firms, credit protection returns are more sensitive to equity returns in the normal market. This is likely because, in a normal market, non-investment firms are more default sensitive than investment-grade firms are. The default probabilities of investment-grade firms are low and most CDS contract holders are liquidity holders, particularly during normal market periods. However, this is not the case during turmoil, when investment-grade firms are more likely to downgrade, making their CDS contracts very sensitive to value change of underlying firms. Conversely, non-investment firms, presumably with more default risk, tend to have more potential to upgrade than downgrade during turmoil.4 Therefore, during turmoil, credit protection returns of investment-grade firms are more sensitive to equity returns. In periods of up to five days, the cumulative response from credit protection returns to a 1 percent decrease in equity returns is 79 bps and 42 bps, respectively, for investment-grade and non-investment-grade firms, with 35 bps and 17 bps for the contemporaneous response. The above results are based on two sub-samples and clearly show that the responses from credit protection returns to equity returns could differ as a response to market performance (market level) or firm performance (firm level). We use both one lag credit protection returns and one lag equity returns to signal the market condition and examine whether extra information can be obtained from the CDS market by running the following model: Rcds,i,t = ˇ0 +

5 

ˇ1,p Rcds,i,t−p +

p=1

× I(R

5 

ˇ2,s Rstock,i,t−s + ˇ3,0 Rstock,i,t × I(R + ˇ3,1 Rstock,i,t−1 cds,i,t−1 >0)

s=0

) + ˇ4,0 Rstock,i,t × I(Rstock,i,t−1 <0) + ˇ4,1 Rstock,i,t−1 × I(Rstock,i,t−1 <0) + i,t

cds,i,t−1 >0

(2)

where I(Rcds,i,t−1 >0) is a dummy variable that takes 1 if the one lag credit protection return is positive and I(Rstock,i,t−1 <0) is a dummy variable that takes 1 if the one lag equity return is negative.

4 During the second sub-sample period, there are 171 firm-month downgrade events of which 121 are associated with investment-grade firms. Of the 71 firm-month upgrade events, 39 are associated with non-investment grade firms.

64

P. Wang, R. Bhar / Int. Fin. Markets, Inst. and Money 30 (2014) 55–80

Coefficient ˇ3,0 measures the sensitivity of credit protection returns to contemporaneous equity return conditions on the negative credit news in the previous trading day, represented by positive credit protection returns in the previous trading day. If there is extra information in the lagged credit protection return, ˇ3,0 should be significantly negative. The empirical results in Table 4A show this to be the case. If quoted CDS levels increased in the previous trading day, a 1 percent decrease in contemporaneous equity return is associated with an 11 bps (10 percent significance level) greater increase in the credit protection return for a typical investment-grade issuer and an 8 bps (5 percent significance level) greater increase for a typical non-investment-grade issuer. For the two different sub-sample periods (Tables 4B and 4C), a similar result is observed for non-investment-grade firms for the second sub-sample only; protection returns of non-investment-grade firms are more sensitive to equity returns when firms’ creditworthiness deteriorates during turmoil. Similarly, coefficient ˇ4,0 and ˇ4,1 measure the sensitivities of credit protection returns to contemporaneous and one lag equity return conditions on the negative credit news in the previous trading day, represented by positive credit protection return in the previous trading day. Table 4A shows that, if equity return on a given trading day is negative, the 1 percent decrease of equity return on that day is associated with an extra 12 bps (1 percent significance level) increase in the credit protection return the following day for a typical investment-grade issuer. This effect is observed for investment-grade firms for the whole sample period and for non-investment-grade firms during the turmoil period only. Overall, we find that there are significant information flows from the equity to the CDS market, particularly during market turmoil. The sensitivities of credit protection returns to equity returns are higher when firms’ creditworthiness deteriorates. 4. Response of equity returns to credit protection returns This section examines the information flow from the CDS to the equity market by checking the response from the equity returns to the credit protection returns of the same firms. We switch the position of credit protection and equity returns and run the following specification: Rstock,i,t = ˇ0 +

5  p=1

ˇ1,p Rstock,i,t−p +

10 

ˇ2,s Rcds,i,t−s + εt

(3)

s=0

The dependent variable, Rstock,i,t , is the equity return for firm i on day t. The independent variables include Rcds,i,t−s , the credit protection return for firm i for day t − s, where s takes from 0 to 10 days and Rstock,i,t−p where p varies from one to five, to control for firm fixed effect. We estimate the same model for two credit rating categories and three periods, as in Section 3, with results reported in Tables 5A, 5B and 5C, respectively, for different periods. Table 5A shows no statistically significant response from equity returns to contemporaneous and lagged credit protection returns for a typical investment-grade issuer. There is a weak statistically significant response from the equity return to the one lag credit protection return for non-investment firms. Surprisingly, when adjusted by series correlation, even equity returns do not co-move with contemporaneous protection returns. However, the relationship seems different under different market conditions. In the normal market (Table 5B), in a general sense, no significant information flow is detected from the CDS to the equity market other than that equity returns of non-investmentgrade firms have a statistically significant response to the one lag credit protection return. During market turmoil (Table 5C), the sensitivities of equity returns to contemporaneous credit protection returns are stronger (and statistically significant) for all firms. For non-investment-grade firms, there is extra information flow from the one lag credit protection return to the equity return. For example, a 1 percent increase in credit protection returns is associated with an 8.36 bps decrease for a typical investment-grade firm and a 19.76 bps decrease for a non-investment-grade firm. The Merton model of corporate debt predicts that equity returns should be less sensitive to credit protection returns for firms with high credit ratings. Table 5C provides some evidence in support of this. Next, we explore information flows from CDS to equity market conditions on both market performance and firm performance. As in Section 3, we use both the one lag credit protection return and the one lag equity return to signal the market and firm performance and determine whether extra

Lag

0

1

Panel A: investment −0.1601 −0.2311 (−6.10)*** (−8.60)*** Panel B: non-investment −0.1630 −0.1879 (−2.74)*** (−8.51)***

2

3

4

5

ˇ3,0

ˇ3,1

ˇ4,0

ˇ4,1

R2

−0.0641 (−2.18)**

−0.0124 (−0.72)

−0.0187 (−0.91)

−0.0112 (−0.65)

−0.1114 (−1.94)*

−0.0082 (−0.31)

−0.0241 (−0.70)

−0.1180 (−2.83)***

0.0126

−0.0817 (−3.33)***

−0.0374 (−0.60)

−0.0886 (−2.04)**

−0.0139 (−1.03)

−0.0804 (−2.29)**

0.0065 (0.13)

−0.0003 (−0.01)

0.0121 (0.13)

0.0004

Table reports coefficients of interest and t-statistics from regression (Eq. (2)) of daily credit protection returns on daily equity returns between 1 Jul 2004 and 31 Aug 2010. The reported t-statistics in parentheses are based on adjusted standard error clustered by both date and firm. The regression is done on investment grade and non-investment grade firms separately and results are reported in Panel A and Panel B. Each firm is assigned to a particular credit rating category based on its S&P long-term credit rating at the beginning of each sample period. * ** , , and *** represent 10%, 5%, and 1% significance levels, respectively.

P. Wang, R. Bhar / Int. Fin. Markets, Inst. and Money 30 (2014) 55–80

Table 4A Conditional response of the credit protection return to the equity return between 1 Jul 2004 and 31 Aug 2010.

65

66

Lag

0

1

Panel A: investment −0.0384 −0.1169 (−0.46) (−1.69)* Panel B: non-investment −0.2489 −0.4384 (−4.76)*** (−1.97)**

2

3

4

5

ˇ3,0

ˇ3,1

ˇ4,0

ˇ4,1

R2

0.0020 (0.02)

−0.0246 (−0.77)

0.0031 (0.06)

−0.0121 (−0.28)

−0.3322 (−1.45)

−0.0887 (−1.84)*

0.0226 (0.23)

−0.2717 (−2.89)***

0.0118

−0.1673 (−1.97)**

−0.0713 (−0.28)

−0.2968 (−1.56)

0.0311 (0.82)

0.0006 (0.00)

0.1981 (1.31)

−0.1455 (−1.21)

0.2630 (1.18)

0.0003

Table reports coefficients of interest and t-statistics from regression (Equation (2)) of daily credit protection returns on daily equity returns between 1 Jul 2004 and 31 Dec 2007. The reported t-statistics in parentheses are based on adjusted standard error clustered by both date and firm. The regression is done on investment grade and non-investment grade firms separately and results are reported in Panel A and Panel B. * , ** , and *** represent 10%, 5%, and 1% significance levels, respectively.

P. Wang, R. Bhar / Int. Fin. Markets, Inst. and Money 30 (2014) 55–80

Table 4B Conditional response of the credit protection return to the equity return between 1 Jul 2004 and 31 Dec 2007.

Lag

0

1

Panel A: investment −0.3014 −0.1854 (−5.75)*** (−8.46)*** Panel B: non-investment −0.1432 −0.0717 (−3.30)*** (−7.85)***

2

3

4

5

ˇ3,0

−0.0879 (−3.47)***

−0.0159 (−0.67)

−0.0513 (−2.04)**

−0.0482 (−2.07)**

−0.0519 (−1.29)

−0.0452 (−3.88)***

−0.0147 (−1.16)

−0.0204 (−1.70)*

−0.0254 (−2.40)**

−0.0588 (−2.33)**

ˇ4,0

ˇ4,1

R2

0.0542 (1.51)

−0.0367 (−1.07)

−0.1572 (−2.66)***

0.0479

−0.0461 (−1.47)

−0.0050 (−0.20)

−0.1267 (−3.92)***

0.0728

ˇ3,1

Table reports coefficients of interest and t-statistics from regression (Eq. (2)) of daily credit protection returns on daily equity returns between 1 Jan 2008 and 31 Aug 2010. The reported t-statistics in parentheses are based on adjusted standard error clustered by both date and firm. The regression is done on investment grade and non-investment grade firms separately and results are reported in Panel A and Panel B. * , ** , and *** represent 10%, 5%, and 1% significance levels, respectively.

P. Wang, R. Bhar / Int. Fin. Markets, Inst. and Money 30 (2014) 55–80

Table 4C Conditional response of the credit protection return to the equity return between 1 Jan 2008 and 31 Aug 2010.

67

68

Lag

0

1

Panel A: investment −0.0046 −0.0002 (−0.50) (−1.15) −0.0048 −0.0046 Panel B: non-investment −0.0007 −0.0001 (−1.32) (−1.91)* −0.0009 −0.0007

2

3

4

5

6

7

8

9

10

R2

0.0002 (0.63) −0.0047

0.0002 (0.65) −0.0044

−0.0002 (−0.62) −0.0046

0.0007 (1.11) −0.0039

0.0004 (1.00) −0.0035

0.0006 (1.03) −0.0029

0.0001 (0.31) −0.0028

−0.0003 (−0.57) −0.0031

−0.0001 (−0.17) −0.0031

0.0025

0.0001 (1.22) −0.0008

0.0000 (−0.02) −0.0008

0.0001 (1.16) −0.0007

0.0000 (0.45) −0.0007

0.0000 (0.41) −0.0006

0.0002 (1.03) −0.0004

0.0000 (−0.52) −0.0004

−0.0002 (−2.27)** −0.0007

−0.0002 (−1.50) −0.0008

0.0036

Table reports coefficients and t-statistics from regression (Eq. (3)) of daily equity returns on its 5 lags (not reported), contemporaneous (lag = 0) and lagged (lag = 1 to lag = 10) daily credit protection returns between 1 Jul 2004 and 31 Aug 2010. The reported t-statistics in parentheses are based on adjusted standard error clustered by both date and firm. The regression is done on investment grade and non-investment grade samples separately and results are reported in Panel A and Panel B. Each firm is assigned to a particular credit rating category based on its S&P long-term credit rating at the beginning of each sample period. * , ** , and *** represent 10%, 5%, and 1% significance levels, respectively.

P. Wang, R. Bhar / Int. Fin. Markets, Inst. and Money 30 (2014) 55–80

Table 5A Response of the equity return to the credit protection return between 1 Jul 2004 and 31 Aug 2010.

Lag

0

1

Panel A: investment −0.0009 0.0001 (0.68) (−1.13) −0.0009 −0.0008 Panel B: non-investment −0.0002 −0.0001 (−3.79)*** (−1.31) −0.0002 −0.0003

2

3

4

5

6

7

8

9

10

R2

0.0001 (0.67) −0.0007

0.0000 (0.38) −0.0006

−0.0001 (−0.82) −0.0007

0.0002 (0.95) −0.0006

0.0001 (0.95) −0.0005

0.0002 (1.16) −0.0002

0.0001 (0.93) −0.0001

0.0000 (−0.19) −0.0002

−0.0001 (−0.55) −0.0003

0.0016

0.0000 (1.35) −0.0003

0.0000 (−0.62) −0.0003

0.0001 (1.47) −0.0002

0.0000 (−0.02) −0.0002

0.0000 (−0.74) −0.0003

0.0000 (0.22) −0.0003

0.0000 (−1.31) −0.0003

−0.0002 (−3.11)*** −0.0005

−0.0002 (−1.67) −0.0006

0.0021

Table reports coefficients and t-statistics from regression (Eq. (3)) of daily equity returns on its 5 lags (not reported), contemporaneous (lag = 0) and lagged (lag = 1 to lag = 10) daily credit protection returns between 1 Jul 2004 and 31 Dec 2007. The reported t-statistics in parentheses are based on adjusted standard error clustered by both date and firm. The regression is done on investment grade and non-investment grade firms separately and results are reported in Panel A and Panel B. Each firm is assigned to a particular credit rating category based on its S&P long-term credit rating at the beginning of each sample period. * , ** , and *** represent 10%, 5%, and 1% significance levels, respectively.

P. Wang, R. Bhar / Int. Fin. Markets, Inst. and Money 30 (2014) 55–80

Table 5B Response of the equity return to the credit protection return between 1 Jul 2004 and 31 Dec 2007.

69

70

Lag

0

1

Panel A: investment −0.0836 −0.0013 (−0.19) (−2.58)*** −0.0849 −0.0836 Panel B: non-investment −0.1976 −0.0232 (−4.76)*** (−1.65)* −0.2208 −0.1976

2

3

4

5

6

7

8

9

10

R2

0.0040 (0.63) −0.0809

0.0030 (0.49) −0.0779

−0.0052 (−0.89) −0.0831

0.0058 (0.84) −0.0772

0.0044 (0.74) −0.0728

0.0040 (0.50) −0.0688

0.0010 (0.14) −0.0678

−0.0080 (−1.05) −0.0759

0.0019 (0.27) −0.0739

0.0326

0.0006 (0.05) −0.2203

0.0173 (1.25) −0.2030

−0.0102 (−0.83) −0.2133

0.0115 (0.83) −0.2018

0.0184 (1.27) −0.1834

0.0071 (0.53) −0.1762

−0.0080 (−0.64) −0.1842

−0.0150 (−1.11) −0.1993

0.0158 (1.32) −0.1834

0.0395

Table reports coefficients and t-statistics from regression (Eq. (3)) of daily equity returns on its 5 lags (not reported), contemporaneous (lag = 0) and lagged (lag = 1 to lag = 10) daily credit protection returns between 1 Jan 2008 and 31 Aug 2010. The reported t-statistics in parentheses are based on adjusted standard error clustered by both date and firm. The regression is done on investment grade and non-investment grade firms separately and results are reported in Panel A and Panel B. Each firm is assigned to a particular credit rating category based on its S &P long-term credit rating at the beginning of each sample period. * , ** , and *** represent 10%, 5%, and 1% significance levels, respectively.

P. Wang, R. Bhar / Int. Fin. Markets, Inst. and Money 30 (2014) 55–80

Table 5C Response of the equity return to the credit protection return between 1 Jan 2008 and 31 Aug 2010.

P. Wang, R. Bhar / Int. Fin. Markets, Inst. and Money 30 (2014) 55–80

71

information can be obtained from the CDS market by running the following model: 5 

Rstock,i,t = b0 +

b1,p Rstock,i,t−p +

5 

b2,s Rcds,i,t−s + b3,0 Rcds,i,t × I(Rstock,i,t−1 <0) + b3,1 Rcds,i,t−1

s=0

p=1

) + b4,0 Rcds,i,t × I(Rcds,i,t−1 >0) + b4,1 Rcds,i,t−1 × I(Rcds,i,t−1 >0) + i,t

× I(R

stock,i,t−1 <0

(4)

where I(Rstock,i,t−1 <0) is a dummy variable that takes 1 if the one lag equity return is negative and I(Rcds,i,t−1 >0) is a dummy variable that takes 1 if the one lag credit protection return is positive. The coefficients b4,0 and b4,1 measure the response from equity returns to credit protection return conditions on information from the previous trading day at the CDS market. If there is extra information represented in the positive credit protection return on the previous day, b4,0 and b4,1 , particularly b4,0 , should be significantly negative. Tables 6A, 6B and 6C report results based on three sample periods and, in each table, Panel A and Panel B report results based on two credit rating categories. The results provide strong evidence that there is extra information contained in credit protection returns on the previous day. The equity returns are more sensitive to credit protection returns if firms’ creditworthiness deteriorates, particularly during turmoil. 5. Response around date of credit announcement In the previous section, we find information flow from the CDS to the equity market; however, only conditions on market performance and firms’ credit deterioration is reflected in the credit protection return on the previous trading day. In this section, we examine the response between two markets around credit announcement day to explore the information content of the credit protection returns further. Once a credit rating is assigned to a debt obligation, a rating agency monitors the credit quality of the issuer and can assign it a different credit rating. An upgrade occurs when there is an improvement in the credit quality of an issuer, while a downgrade occurs when there is deterioration in the credit quality of an issuer. Typically, before an issuer’s rating is changed, the rating agency will announce in advance that it is reviewing the issuer with the potential for upgrade or downgrade. In such cases, the issuer is said to be on ‘rating watch’ or ‘credit watch’. In the announcement, the rating agency will state the direction of the potential change in rating – upgrade or downgrade. Typically, a decision will be made within three months. We first focus on the credit rating announcement and run the following three models: rx,i,t = a1 + b1 ry,i,t

(5)

rx,i,t+1 = a2 + c2 rx,i,t + b2 ry,i,t

(6)

rx,i,t = a3 +

5  s=1

c3s rx,i,t−s

5 

b3p ry,i,t−p

(7)

p=0

where subscript x, y represents the CDS or stock market. There are 144 downgrades during the first sub-sample period and 177 during the second. Fewer downgrades for non-investment-grade firms and more downgrades for investment-grade firms are observed. Upgrade events are rarer in our samples: 84 during the first sub-sample period and 92 during the second. Table 7A reports the response between the two markets around the credit rating downgrade announcements. There is neither significant response from credit returns to contemporaneous equity returns on the announcement day nor significant response from one-day-ahead credit protection returns to equity returns. However, credit protection returns on announcement day have a significant response to lagged equity returns for non-investment-grade firms. This result provides extra evidence that, for non-investment-grade issuers, CDS and equity markets are more closely linked, particularly when the whole market is in turmoil and the quality of firms’ credit deteriorates.

72

Lag

0

1

Panel A: investment 0.0012 −0.0001 (−0.12) (0.19) Panel B: non-investment −0.0161 0.0176 (2.04)** (−1.26)

2

3

4

5

b3,0

b3,1

b4,0

b4,1

R2

0.0003 (0.87)

0.0004 (1.03)

−0.0003 (−1.17)

0.0007 (1.22)

−0.0112 (−1.75)*

−0.0008 (−1.07)

−0.0092 (−1.50)

−0.0012 (−0.19)

0.0045

0.0000 (−0.58)

−0.0001 (−0.69)

0.0001 (1.10)

0.0000 (0.42)

−0.0181 (−2.10)**

−0.0002 (−1.90)*

−0.0218 (−2.43)**

0.0161 (1.25)

0.0044

Table reports coefficients of interest and t-statistics from regression (Eq. (4)) of daily equity returns on daily credit protection returns between 1 Jul 2004 and 31 Aug 2010. The reported t-statistics in parentheses are based on adjusted standard error clustered both by date and firms. The regression is done on two different credit rating samples and results are reported in Panel A and Panel B respectively. Each firm is assigned to a particular credit rating category based on its S&P long-term credit rating at the beginning of each sample period. * , ** , and *** represent 10%, 5%, and 1% significance levels, respectively.

P. Wang, R. Bhar / Int. Fin. Markets, Inst. and Money 30 (2014) 55–80

Table 6A Conditional response of the equity return to the credit protection return between 1 Jul 2004 and 31 Aug 2010.

Lag

0

1

Panel A: investment 0.0002 0.0026 (0.62) (1.50) Panel B: non-investment −0.0018 0.0081 (−0.22) (1.50)

2

3

4

5

b3,0

b3,1

b4,0

b4,1

R2

0.0002 (1.04)

0.0001 (0.69)

−0.0001 (−1.09)

0.0002 (1.00)

−0.0021 (−1.33)

−0.0004 (−1.41)

−0.0026 (−2.14)**

−0.0024 (−0.57)

0.0021

0.0000 (−0.87)

−0.0001 (−1.39)

0.0001 (1.43)

0.0000 (−0.04)

−0.0082 (−1.53)

−0.0002 (−4.09)***

−0.0094 (−1.78)*

0.0017 (0.22)

0.0023

Table reports coefficients of interest and t-statistics from regression (Eq. (4)) of daily equity returns on daily credit protection returns between 1 Jul 2004 and 31 Dec 2007. The reported t-statistics in parentheses are based on adjusted standard error clustered both by date and firms. The regression is done on two different credit rating samples and results are reported in Panel A and Panel B respectively. * , ** , and *** represent 10%, 5%, and 1% significance levels, respectively.

P. Wang, R. Bhar / Int. Fin. Markets, Inst. and Money 30 (2014) 55–80

Table 6B Conditional response of the equity return to the credit protection return between 1 Jul 2004 and 31 Dec 2007.

73

74

Lag

0

Panel A: investment −0.0682 (−2.75)*** Panel B: non-investment −0.1727 (−4.42)***

1

2

3

4

5

b3,0

b3,1

b4,0

b4,1

R2

0.0196 (1.09)

0.0050 (0.81)

0.0032 (0.53)

−0.0046 (−0.81)

0.0062 (0.88)

0.0219 (0.66)

0.0004 (0.03)

−0.1146 (−4.17)***

−0.0242 (−1.28)

0.0435

−0.0319 (−1.07)

0.0024 (0.20)

0.0164 (1.18)

−0.0090 (−0.74)

0.0123 (0.88)

0.0332 (0.69)

0.0090 (0.32)

−0.1262 (−2.99)***

0.0066 (0.20)

0.0421

Table reports coefficients of interest and t-statistics from regression (Eq. (4)) of daily equity returns on daily credit protection returns between 1 Jan 2008 and 31 Aug 2010. The reported t-statistics in parentheses are based on adjusted standard error clustered both by date and firms. The regression is done on two different credit rating samples and results are reported in Panel A and Panel B respectively. * , ** , and *** represent 10%, 5%, and 1% significance levels, respectively.

P. Wang, R. Bhar / Int. Fin. Markets, Inst. and Money 30 (2014) 55–80

Table 6C Conditional response of the equity return to the credit protection return between 1 Jan 2008 and 31 Aug 2010.

Table 7A Response between two markets around negative credit rating announcement. Panel A: Response of credit protection return to equity return Down

rcds,i,t : rstock,i,t

rcds,i,t+1 : rstock,i,t (rcds,i,t )

rcds,i,t : rstock,i,t , rstock,i,t−s (rcds,i,t−s , s = 1, 2, ...5)

b2

b30

b31

b32

b33

b34

b35

2004/07/01–2007/12/31

0.0299 (0.05) 0.0791 (0.19)

−0.4186 (−1.60) 0.1052 (1.55)

0.1859 (0.27) −0.0918 (−0.16)

0.2661 (0.69) 0.8132 (0.93)

0.6133 (1.06) −0.2810 (−0.62)

−0.3279 (−0.44) −1.6030 (−1.07)

0.8337 (1.41) 1.1713 (0.69)

−0.2348 (−0.50) −0.4595 (−0.50)

2008/01/01–2010/08/31

Investment No.: 91 Non-investment No.: 86

−0.1950 (−0.50) −0.2324 (−1.62)

0.0027 (0.02) −0.4210 (−1.53)

−0.0798 (−0.21) −0.0978 (−0.70)

0.1094 (0.61) −0.4942 (−2.60)***

0.0985 (0.75) −0.5724 (−2.05)**

−0.0972 (−0.58) 0.0038 (0.02)

−0.1320 (−0.79) −0.1319 (−1.00)

−0.1274 (−0.79) −0.7960 (−2.71)***

Panel B: Response of equity return to credit protection return Down

rstock,i,t : rcds,i,t

rstock,i,t+1 : rcds,i,t (rstock,i,t )

rstock,i,t : rcds,i,t , rcds,i,t−s (rstock,i,t−s , s = 1, 2, . . . 5)

2004/07/01–2007/12/31

Investment No.: 106 Non-investment No.: 38

0.0015 (0.05) 0.0079 0.22

−0.0281 (−2.21)** −0.0317 −1.21

0.0087 (0.29) −0.0075 (−0.16)

−0.0013 (−0.03) 0.0925 (0.60)

0.0179 (1.28) −0.2311 (−1.08)

0.1153 (1.84)* −0.1693 (−0.50)

−0.0871 (−2.11)** −0.5576 (−2.69)***

0.0527 (1.28) −0.0917 (−0.29)

2008/01/01–2010/08/31

Investment No.: 91 Non-investment No.: 86

−0.0749 (−0.49) −0.2443 (−2.00)**

0.0050 (0.06) 0.0059 (0.11)

−0.0378 (−0.20) −0.1259 (−0.78)

−0.1103 (−1.12) −0.5678 (−2.08)**

0.0507 (0.48) 0.2965 (1.17)

0.0861 (0.81) 0.7864 (2.38)**

−0.0386 (−0.28) 0.4183 (1.50)

−0.0660 (−0.53) 0.1549 (0.53)

P. Wang, R. Bhar / Int. Fin. Markets, Inst. and Money 30 (2014) 55–80

b1 Investment No.: 106 Non-investment No.: 38

Panel A reports response of credit protection returns to equity returns around the negative credit rating announcement and Table 7 Panel B reports response of equity returns to credit protection returns around the negative credit rating announcement. The whole samples are divided into two sub-periods and two credit categories. Firms with S&P long-term credit ratings of at least BBB− before announcement are classified as investment-grade firms, and those with ratings below BBB− before announcement are classified as non-investment-grade firms. Column 3 reports coefficient by of Equation (5), Column 4 reports coefficient of Equation (6) and Column 5 to Column 10 reports coefficients of Equation (7). The reported t-statistics in parentheses are based on adjusted standard error clustered by both date and firm. * , ** , and *** represent 10%, 5%, and 1% significance levels, respectively.

75

76

Table 7B Response between two markets around positive credit rating announcement. Panel A: Response of credit protection return to equity return Up

rcds,i,t : rstock,i,t

rcds,i,t+1 : rstock,i,t (rcds,i,t )

rcds,i,t : rstock,i,t , rstock,i,t−s (rcds,i,t−s , s = 1, 2, ...5)

b2

b30

b31

b32

b33

b34

b35

2004/07/01–2007/12/31

−0.4363 (−1.14) −0.3112 (−0.74)

−0.3583 (−1.01) 0.2486 (0.85)

−0.2659 (−0.40) −0.1062 (−0.27)

1.8360 (0.55) 0.3389 (1.46)

1.7533 (0.98) 0.0528 (0.09)

1.9797 (0.74) −0.7047 (−0.69)

2.4925 (1.47) 0.2786 (0.58)

0.9204 (1.24) −0.7720 (−1.99)**

2008/01/01–2010/08/31

Investment 47 Non-investment 45

−0.3465 (−1.22) 0.1108 (0.28)

−0.2144 (−0.52) −0.3030 (−3.01)***

−0.0191 (−0.06) 0.0677 (0.27)

0.4849 (1.28) 0.2186 (1.05)

−0.9477 (−2.12)** 0.9381 (2.01)**

−0.2066 (−0.44) 0.1715 (0.47)

−0.9214 (−1.43) −0.3886 (−1.10)

−0.7803 (−1.16) 0.6375 (1.61)

Panel B: Response of equity return to credit protection return Up

rstock,i,t : rcds,i,t

rstock,i,t+1 : rcds,i,t (rstock,i,t )

rstock,i,t : rcds,i,t , rcds,i,t−s (rstock,i,t−s , s = 1, 2, ...5)

2004/07/01–2007/12/31

Investment 57 Non-investment 27

−0.0020 (−0.66) −0.0324 (−0.68)

0.0062 (3.18)*** −0.0227 (−0.61)

−0.0009 (−0.38) −0.0097 (−0.26)

−0.0380 (−0.76) −0.1296 (−1.62)

−0.1407 (−1.30) −0.0521 (−0.64)

−0.0813 (−2.06)** 0.0045 (0.03)

0.0106 (0.17) 0.0043 (0.04)

0.1057 (1.77)* 0.1455 (2.17)**

2008/01/01–2010/08/31

Investment 47 Non-investment 45

−0.1538 (−1.39) 0.0641 (0.32)

0.0654 (0.41) 0.0269 (0.36)

−0.0088 (−0.06) 0.0388 (0.27)

−0.0696 (−0.91) 0.1128 (0.56)

0.3353 (1.51) −0.2822 (−0.59)

−0.1555 (−1.34) 0.2116 (1.12)

0.3555 (2.10)** −0.0172 (−0.06)

−0.0874 (−0.47) 0.1045 (0.38)

Panel A reports response of credit protection returns to equity returns around the positive credit rating announcement and Panel B reports response of equity returns to credit protection returns around the positive credit rating announcement. The whole samples are divided into two sub-periods and two credit categories. Firms with S&P long-term credit ratings of at least BBB− before announcement are classified as investment-grade firms, and those with ratings below BBB− before announcement are classified as non-investment-grade firms. Column 3 reports coefficient of Eq. (5), Column 4 reports coefficient of Eq. (6) and Column 5 to Column 10 reports coefficients of Eq. (7). The reported t-statistics in parentheses are based on adjusted standard error clustered by both date and firm. * , ** , and *** represent 10%, 5%, and 1% significance levels, respectively.

P. Wang, R. Bhar / Int. Fin. Markets, Inst. and Money 30 (2014) 55–80

b1 Investment 57 Non-investment 27

Panel A: Response of credit protection return to equity return Watch negative

Down No.: 226 Up No.: 88

rcds,i,t : rstock,i,t

rcds,i,t+1 : rstock,i,t (rcds,i,t )

rcds,i,t : rstock,i,t , rstock,i,t−s (rcds,i,t−s , s = 1, 2, ...5)

b1

b2

b30

b31

b32

b33

b34

b35

−0.1815 (−0.75) 0.1313 (0.25)

−0.0046 (−0.04) −0.2112 (−1.93)**

−0.1461 (−0.98) 0.1142 (0.45)

−0.6388 (−2.47)** 0.2032 (1.38)

−0.8415 (−2.88)*** 1.1517 (2.63)***

−0.0067 (−0.03) 0.1683 (0.53)

−0.2638 (−1.88)* −0.0819 (−0.28)

−1.2738 (−4.00)*** 1.0072 (2.90)***

Panel B: Response of equity return to credit protection return Watch negative

rstock,i,t : rcds,i,t

rstock,i,t+1 : rcds,i,t (rstock,i,t )

rstock,i,t : rcds,i,t , rcds,i,t−s (rstock,i,t−s , s = 1, 2, ...5)

Down No.: 226 Up No.: 88

−0.0719 (−0.94) 0.0596 (0.27)

−0.0086 (−0.31) 0.0680 (0.80)

−0.0658 (−0.97) 0.0460 (0.44)

−0.0097 (−0.13) 0.4151 (2.46)**

−0.1586 (−0.82) −0.6334 (−2.26)**

0.2213 (1.91)* 0.2453 (1.41)

−0.1659 (−0.78) 0.1081 (0.33)

−0.1304 (−0.97) 0.2855 (0.94)

Table reports responds between two markets around the credit rating announcement after negative credit watch. Row “Down” presents the case when there is a downgrade announcement following negative credit watch. Row “Up” presents the case when there is an upgrade announcement following negative watch. The reported t-statistics in parentheses are based on adjusted standard error clustered by both date and firm. * , ** , and *** represent 10%, 5%, and 1% significance levels, respectively.

P. Wang, R. Bhar / Int. Fin. Markets, Inst. and Money 30 (2014) 55–80

Table 8A Response between two markets around the credit rating announcement after negative credit watch.

77

78

Panel A: Response of credit protection return to equity return Watch positive

Up No.: 59 Down No.: 23

rcds,i,t : rstock,i,t

rcds,i,t+1 : rstock,i,t (rcds,i,t )

rcds,i,t : rstock,i,t , rstock,i,t−s (rcds,i,t−s , s = 1, 2, ...5)

b1

b2

b30

b31

b32

b33

b34

b35

0.6330 (0.88) 0.2324 (4.80)***

−0.2587 (−2.44)** 0.1172 (1.18)

−0.2555 (−0.56) 0.1147 (1.18)

0.9533 (1.97)** 0.7956 (4.00)***

1.0345 (2.15)** 0.4803 (1.12)

−0.1197 (−0.45) −0.5092 (−1.28)

−0.5758 (−1.18) 0.6355 (2.56)**

0.0352 (0.07) 0.0437 (0.15)

−0.0327 (−0.23) 0.1808 (0.22)

0.1105 (1.23) 1.2080 (1.35)

Panel B: Response of equity return to credit protection return Watch positive

rstock,i,t : rcds,i,t

rstock,i,t+1 : rcds,i,t (rstock,i,t )

rstock,i,t : rcds,i,t , rcds,i,t−s (rstock,i,t−s , s = 1, 2, . . . 5)

Up No.: 59 Down No.: 23

0.1826 (1.46) 1.6919 (1.72)*

0.1242 (2.47)** 0.1107 (0.79)

−0.0478 (−0.57) 0.8588 (1.10)

−0.0339 (−0.38) −0.8846 (−1.50)

−0.1610 (−0.87) −0.5920 (−0.56)

−0.0315 (−0.32) −0.0691 (−0.06)

Table reports responds between two markets around the credit rating announcement after positive credit watch. Row “Down” presents the case when there is a downgrade announcement following positive credit watch. Row “Up” presents the case when there is an upgrade announcement following positive watch. The reported t-statistics in parentheses are based on adjusted standard error clustered by both date and firm. * , ** , and *** represent 10%, 5%, and 1% significance levels, respectively.

P. Wang, R. Bhar / Int. Fin. Markets, Inst. and Money 30 (2014) 55–80

Table 8B Response between two markets around the credit rating announcement after positive credit watch.

P. Wang, R. Bhar / Int. Fin. Markets, Inst. and Money 30 (2014) 55–80

79

Of most interest is that there is both a statistically and an economically significant response from one-day-ahead equity return to credit protection returns on the credit rating announcement day during the normal market period (see Panel B of Table 7A). This result, at least to some extent, provides evidence that there may be informed trading in the CDS market for investment-grade firms. Table 7B reports the response between the two markets around the upgrade announcements. Again, we find significant response from one-day-ahead equity returns to credit protection returns on the credit rating announcement day during the normal market period; although, surprisingly, the coefficient is positive. By checking the results presented in Table 8B, we find that this positive response may be the result of overreaction to potentially positive news. Cantor and Hamilton (2007) find that, for corporate bonds, ratings combined with rating watches and the rating outlook status provide a better gauge for default risk than using the ratings alone. This is because, before an issuer’s rating is changed, the rating agency will announce in advance that it is reviewing the issuer with the potential for upgrade or downgrade. We combine the credit watch and credit rating information to examine the response between the two markets. Table 8A (Panel A) reports a response between the two markets when there is a downgrade or upgrade announcement following a negative credit watch. Table 8A (Panel B) reports a response between the two markets when there is a downgrade or upgrade announcement following a positive credit watch. In the case of a downgrade announcement following a negative credit watch, there is neither a significant response from credit protection returns to contemporaneous equity returns nor a response from one-day-ahead credit protection returns to equity returns on announcement day. The same results are observed in the opposite direction; that is, the credit protection return on announcement day can still be predicted by lagged equity returns, which is consistent with the general results. When there is a surprise on announcement day, that is, an upgrade announcement following negative credit watch, one-day-ahead credit protection returns have a significant negative response to equity returns on announcement day, rather than credit protection returns on the same day. This demonstrates that the equity market can catch the ‘up’ surprise more quickly. When there is an ‘up’ surprise, one-dayahead equity returns respond positively to credit protection returns on announcement day, which suggests that equity returns may overreact to positive news (although this response is not statistically significant in our sample). Table 8B reports a response between the two markets around announcement day following a positive credit watch. If there is an upgrade announcement, one-day-ahead equity returns respond positively to credit protection returns, which shows significant overreaction from equity returns to good news on firm’s credit quality. 6. Conclusion In this paper, we explore information flows between equity and CDS markets, focusing particularly on the information content of credit protection returns. We discover that equity returns can predict credit protection returns at a horizon of up to five days, particularly under turmoil market conditions. There does not appear to be any leading role of credit protection returns for investment-grade firms. We interpret this as independence of the two markets for high-credit-rating entities, particularly when arbitrage opportunities are rare. Such arbitrage opportunities are also not beneficial given the wide bid-ask spread when trading CDS contracts. It may appear arbitrary to separate the sample into two sub-periods and label them as normal conditions and turmoil conditions. To circumvent this criticism, we estimate our model conditioned on market performance (market level) or firm performance (firm level). This is achieved by introducing the one lag credit protection return and one lag equity return to signal the market performance and firm performance. Credit protection returns are more sensitive to lagged equity returns if credit deterioration is detected previously in the CDS market. This seems to suggest that, although the information contained in lagged credit protection returns cannot predict the magnitude of future equity returns, it provides the extra information for equity returns based on the condition of the credit market. We examine the responsiveness of each market around credit rating announcement day. We do not find any significant response from credit returns to contemporaneous equity returns on announcement day. In addition, there is no significant response of one-day-ahead credit protection returns to equity returns. However, credit protection returns on announcement day have a significant response to lagged

80

P. Wang, R. Bhar / Int. Fin. Markets, Inst. and Money 30 (2014) 55–80

equity returns for non-investment-grade firms. This result provides extra evidence that, for noninvestment-grade issuers, CDS markets and equity markets are more closely linked, particularly when the whole market is in turmoil and the quality of firms’ credit deteriorates. There is both a statistically and economically significant response from one-day-ahead equity returns to credit protection returns on credit rating announcement day during the normal market period. This provides some evidence that there may be informed trading in the CDS market for highcredit-rating firms. Therefore, it is obvious that, if there is informed trading in the CDS market, CDS contracts with high-credit-rating entities are the ones that would be exploited by informed traders. When there is a surprise on announcement day, one-day-ahead credit protection returns have a significant negative response to equity returns, rather than credit protection returns on that day, if an upgrade announcement follows a negative credit watch. This shows that the equity market can catch the ‘up’ surprise more quickly. If there is an upgrade announcement following a positive credit watch, one-day-ahead equity returns respond positively to the credit protection return. This indicates significant overreaction of equity returns to good news on a firm’s credit quality. References Abid, F., Naifar, N., 2006. Credit default swap rates and equity volatility: a nonlinear relationship. Journal of Risk Finance 7, 348–371. Acharya, V.V., Johnson, T.C., 2007. Insider trading in credit derivatives. Journal of Financial Economics 84, 110–141. Berndt, A., Obreja, J., 2010. Decomposing European CDS, returns. Review of Finance 14, 189–233. Blanco, R., Brennan, S., Marsh, I., 2005. An empirical analysis of the dynamic relationship between investment-grade bonds and credit default swaps. Journal of Finance 60, 2255–2281. Cantor, R., Hamilton, D.T., 2007. Adjusting corporate default rate for rating withdrawals. Journal of Credit Risk 3, 3–25. Dupuis, D., Jacquier, E., Papageorgiou, N., Rémillard, B., 2009. Empirical evidence on the dependence of credit default swaps and equity prices. Journal of Futures Market 29 (8), 695–712. Hilscher, J., Pollet, J.M., Wilson, M.I., 2013. Are credit default swaps a sideshow? Evidence that information flows from equity to CDS markets’, working paper. Social Science Research Network. Hull, J.C., Predescu, M., White, A., 2005. The relationship between credit default spreads, bond yields and credit rating announcements. Journal of Banking and Finance 28 (11), 2789–2811. Kapadia, N., Pu, X., 2012. Limited arbitrage between equity and credit markets. Journal of Financial Economics 105, 542–563. Longstaff, F.A., Neis, E., Mithal, S., 2005. Corporate yield spreads: default risk or liquidity? New evidence from the credit default swap market. Journal of Finance 60 (5), 2213–2253. Merton, R.C., 1974. On the pricing of corporate debt: the risk structure of interest rates. Journal of Finance 29, 449–470. Norden, L., Weber, M., 2009. The co-movement of credit default swap, bond and stock markets: an empirical analysis. European Financial Management 15, 529–562. Petersen, M.A., 2009. Estimating standard errors in finance panel data sets: comparing approaches. Review of Financial Studies 22, 432–480. Zhang, B.Y., Zhou, H., Zhu, H., 2005. Explaining credit default swap spreads with equity volatility and jump risks of individual firms. BIS working paper No. 181, Monetary and Economic Department.