The Impact of S&P 500 Index Revisions on Credit Default Swap Market

Transcription

1 The Impact of S&P 500 Index Revisions on Credit Default Swap Market By Lindsay Baran Department of Finance Kent State University Ying Li School of Business University of Washington Bothell Chang Liu Department of Finance Kent State University Zilong Liu Department of Finance Kent State University And Xiaoling Pu* Department of Finance Kent State University JEL code: G12, G14 Key words: S&P 500 Index revision, credit default swap (CDS), information, certification *Corresponding author: Department of Finance, College of Business Administration, Kent State University, Kent, OH 44240; Phone: (330) ; We would like to thank Jim Miller for his insights, the helpful comments from participants in the research seminar at Kent State University and the 2015 Eastern Finance Association (EFA) annual meeting in New Orleans. 1

2 The Impact of S&P 500 Index Revisions on Credit Default Swap Market May 2015 Abstract We investigate the impact of S&P 500 Index revisions on the credit default swap (CDS) market using the abnormal CDS spread changes of event firms over a sample period of Our results show that only addition announcements significantly negatively impact CDS spreads over both short- and long-term windows. The negative abnormal CDS spread change is more pronounced (1) during the financial crisis period, (2) for speculative grade firms over the short-term and (3) for investment grade firms over the long-term. Our findings in the CDS market provide new evidence for the certification hypothesis in S&P 500 Index revisions literature. 2

3 1. Introduction Ever since the first documentation of price effects of Standard and Poor s (S&P) Index revisions more than two decades ago, researchers have proposed multiple hypotheses to explain the stock price increases (decreases) associated with addition to (deletion from) the S&P 500 Index: the stock price reaction to these announcements should either be information free, resulting from the downward-sloping demand curves for index stocks; or it should involve information relevant in pricing the newly added or removed stocks conveyed in the revision decision by S&P. Even though S&P repeatedly claims that index revisions do not in any way reflect an opinion on the investment merits of the company, past studies examine information related to changes in liquidity, changes in investor awareness, certification of the performance of the newly added or removed firms, or certification of the industry of the newly added firms (for example, Jain, 1987; Dhillon and Johnson, 1991; Chen, Noronha and Singal, 2004; and Cai, 2007). We examine whether S&P 500 Index revision involves information by exploring the abnormal spread changes in credit default swap (CDS hereafter) market for the event firms. If S&P 500 Index revisions do involve information relevant to firm value, we may be able to observe not only price effects in the stock market, but price effects in other related markets that reflect fundamentals of the firms, for example, credit market. The structural model of Merton (1974) states that equity and debt are both contingent claims on the underlying firm value, which suggests that factors that affect firm value would influence both equity and bond markets. Previous literature, such as Dhillon and Johnson (1991), documents that S&P 500 Index addition brings information not only to the equity market but also to the options and bond markets. If S&P 500 Index revisions involve information on equity value, then the same information might also impact debt value and 3

4 the CDS market, which is a debt derivatives market in which investors trade credit risk of individual firms. Whereas past studies have used bond markets to detect the existence of information from the S&P 500 Index revisions (Dhillon and Jonson, 1991), our study looks into the CDS market which provides a more liquid setting (e.g., Longstaff, Mithal, and Neis, 2005) to examine the reaction of credit market to equity index revision, especially over short-term windows. We find that when measured as the difference in average cumulative change of CDS spreads between event firms and the overall CDS market, the abnormal change of CDS spreads for addition firms is negatively significant over both short- and long-term windows. The difference in cumulative abnormal changes between event firms and their industry and size matched peers remains significant, with a confidence level of 5% or better. Our findings are potentially consistent with the existence of information in S&P 500 Index revisions. Whereas prior studies establish that information may stem from improved liquidity, investor awareness, or certification of future performance, we narrow down the source of such information with the help of the uniqueness of the CDS market. As participants in the CDS market are usually large institutions that have access to information (Acharya and Johnson, 2007), it is likely that these institutions are already aware of the existence of firms that have a presence in the CDS market prior to index inclusion. We therefore argue that increased investor awareness may not explain our findings of information captured by the CDS market. Since the institutional investors who participate in the CDS market specialize in evaluating credit risk and are sensitive to deteriorating credit conditions at firms with 4

5 CDS contracts, deletion from the S&P 500 Index is unlikely to carry much additional information to these institutions regarding the CDS spreads. We suggest that this may explain why there is no significant difference in abnormal CDS spreads upon deletion announcements. Prior studies (Hegde and McDermott, 2003; Chen, Noronha, and Singal, 2004; Becker-Blease and Paul, 2006) find that the positive information conveyed in S&P 500 Index inclusions pertains to sustained stock liquidity improvements. Given that decreased CDS spreads are also associated with improved CDS liquidity (Bongaerts, de Jong, and Driessen, 2011; Tang and Yan, 2014), we investigate the liquidity hypothesis by (1) examining the effect of improved stock liquidity on the abnormal CDS spread change following index additions and (2) exploring the liquidity change in the CDS market. We do not find that the improvement in stock liquidity of addition firms to be associated with the observed cumulative abnormal CDS spread changes. Neither do we find significant liquidity improvement in the CDS market after a stock is added to the S&P 500 Index. Thus, the liquidity hypothesis is unable to explain our findings. We suggest that the certification hypothesis could explain our results, as certification hypothesis argues that S&P 500 Index additions involve information on an added firm s future operating performance, potential longevity, or representativeness in that firm s industry (Dhillon and Johnson 1991; Denis, McConnell, Ovtchinnikov, and Yu, 2003). Further subsample analyses show that the magnitude of the CDS spread reaction varies with conditions that influence the correlation between markets. For example, the cumulative abnormal CDS spread is more pronounced during the financial crisis period when the correlation between equity and credit markets is higher. 5

6 The difference is also more pronounced for addition firms with speculative grade rating over short-term windows, even though it is more pronounced for addition firms with investment grade rating over long-term windows. As correlation influences information transmission and certification effect is stronger for lower rated firms, we maintain that our findings provide new evidence that supports the certification hypothesis. Our study contributes to several strands of literature. First, we find persistent and significantly negative CDS cumulative abnormal changes for addition firms into the S&P 500 Index over both short- and long-term windows. Few studies explored the index revision effect on credit markets (Dhillon and Johnson, 1991). Our study expands this earlier work using the more liquid CDS market which facilitates quicker incorporation of information than the corporate bond market (Blanco, Brennan, and Marsh, 2005). Second, whereas it is challenging to distinguish empirically various hypotheses that attempt to explain the S&P 500 Index addition effect (Shleifer, 1986), we narrow down to the certification hypothesis as the most likely explanation for our findings in the CDS market through various tests. We show that not only does credit quality play a role in the magnitude of abnormal CDS spread changes upon S&P 500 Index revision, but the timing of addition announcements matters. Consistent with the certification hypothesis, CDS spreads respond more favorably during financial crisis period when overall market credit risk is high. The remainder of the paper is organized as follows. Section 2 reviews literature. Section 3 describes the sample and provides descriptive statistics. Sections 4 present the main findings. Section 5 conducts additional tests, and Section 6 concludes. 6

7 2. Literature review and motivation Prior studies document a stock price reaction to inclusion in or removal from the S&P 500 Index. These studies find permanent stock price increases after S&P 500 Index inclusion and temporary stock price declines upon S&P 500 Index removal. One prior study (Dhillon and Johnson, 1991) explored the impact of S&P 500 Index revisions on the credit market using bonds, but there is little knowledge of the impact on the credit derivatives market. To explain the stock price reaction after S&P 500 Index revision announcements, researchers have proposed several explanations which can be categorized into five competing hypotheses. 1 The first hypothesis, downward-sloping demand curve hypothesis, argues that there is no information conveyed in S&P 500 Index revisions and the price effect arises because non-index stocks are imperfect substitutes for index stocks (Scholes, 1972). This hypothesis is supported by some previous work (Shleifer, 1986; Lynch and Mendenhall, 1997; Kaul, Mehrotra and Morck, 2000; Greenwood, 2005). A second explanation is that temporary price pressure from index fund rebalancing drives these price changes (supported by Harris and Gurel, 1986; Elliott and Warr, 2003; Shankar and Miller, 2006; Hrazdil, 2009). Given that these two hypotheses are primarily related to the supply and demand for stocks, our exploration of the credit market in this study cannot lend additional support nor rule out these hypotheses. The remaining three hypotheses concur that index revisions are not informationfree events but each proposes a different form of information transmission. The certification hypothesis deals with whether S&P 500 Index inclusion or removal conveys 1 A more detailed description of the literature about these hypotheses can be found in Kappou, Brooks, and Ward (2008) and Baran and King (2012). 7

8 unknown information about future performance to explain the price response to announcements. Jain (1987) provides empirical evidence that S&P 500 Index addition conveys information to investors, which might change their perceptions of the stocks. Dhillon and Johnson (1991) investigate stock, bond, and option prices around the announcements of the listings in the S&P 500 Index and find bond and option prices to move with stock prices from 1978 to 1988, suggesting that there is information involved in these announcements that impacts the equity, debt, and option markets. Denis, McConnell, Ovtchinnikov and Yu (2003) and Platikanova (2008) also argue that addition to the S&P 500 Index is not information free by discovering earnings improvement around the announcement date. In addition, Cai (2007) finds the positive addition information may spread to the industry of the company. On the other hand, Shleifer (1986) confirms the positive price effects after the S&P 500 Index inclusion but argues that the inclusion does not mean that the firm has good quality since the abnormal returns are not related with the bond ratings. Harris and Gurel (1986) state that the addition events to the S&P 500 Index do not carry specific information since the stock prices reverse back over the 30 days following the announcement in their sample. Beneish and Gardner (1995) support this argument by investigating the price and the trade volume of newly listed Dow Jones Industrial Average firms. Hrazdil and Scott (2009) find that improved earnings following index inclusions are due to these firms larger discretionary accruals, not because of the information effect from the additions. The investor awareness hypothesis draws on the observation that the price responses to inclusion and removal are asymmetric. Chen, Noronha and Singal (2004) 8

9 show that the shadow cost (Merton, 1987) declines upon inclusion due to increased investor awareness but does not decline on index removal since investors do not become unaware of removed stocks. Finally, the liquidity hypothesis proposes that inclusion to the index brings sustained improvements in stock liquidity and reductions in bid-ask spread, consistent with an increase in price. Amihud and Mendelson (1986) propose that firms included into the S&P 500 Index tend to attract more institutional holdings and larger trading volume. This trend leads to less information asymmetry and the stock becomes more liquid, which moves the stock price upward. Following this explanation, Beneish and Gardner (1995), Chung and Kryzanowski (1998), Hegde and McDermott (2003), Chen, Noronha and Singal (2004), and Becker-Blease and Paul (2006) find consistent empirical evidence from the S&P 500 Index and other U.S. market indices. Dhillon and Johnson (1991) demonstrate that information is involved in S&P 500 Index additions by showing price effects in bonds and options markets. The CDS market is more liquid than the bond market as new information is impounded into CDS spreads more rapidly than into corporate bond prices (Blanco, Brennan, and Marsh, 2005). This feature of the CDS market adds to the benefits of investigating the credit market reaction using the CDS spreads. If S&P 500 Index revision announcements carry information, inclusion into the index should be associated with abnormal CDS spread decrease and exclusion from the index should be associated with abnormal CDS spread increase. By examining the market reaction in the CDS market which trades on information that is less driven by changing demand for index stocks, we can focus on whether information in S&P 500 Index revision announcements affects the credit market and how such information is impounded. 9

10 Certain features of the CDS market also help us differentiate the channels through which the information involved in the announcements impounds into the credit market. For example, since participants of the CDS market are usually institutional investors who are well-informed (Acharya and Johnson, 2007), we do not expect the announcements to change the awareness of the CDS reference entity. 2 If information is involved in the S&P 500 Index revision announcements, the reaction in CDS market would vary with the level of integration between equity and credit markets. Previous literature has documented a low correlation between stock returns and credit spread changes (Collin-Dufresne, Goldstein, and Martin, 2001; Blanco, Brennan, and Marsh, 2005). However, correlations between equity and credit markets are also higher in the crisis (Kapadia and Pu, 2012) as market integration is higher in financial crises (Bekaert, Harvey, and Ng, 2005). Thus it is possible that the S&P 500 Index revision information would have more pronounced effect on the CDS market in the recent crisis. Previous studies (e.g., Kapadia and Pu, 2012) also document that firms with low credit ratings usually have larger correlations between stock returns and credit spread changes than investment grade firms. This suggests that the information from the equity market may have more influence on speculative grade firms since correlations between equity returns and credit spread changes are larger in these firms. In particular, if the S&P 500 Index addition announcements contain positive information about the firm, we expect that firms with higher default risk will benefit more. 2 CDS contract provides insurance against a default by a firm, which is known as the reference entity (Hull, Predescu, and Whilt, 2004) 10

11 3. Data and sample statistics 3.1. Sample construction We analyze the abnormal changes in the CDS spreads whose underlying reference firms were added into or removed from the S&P 500 Index from January 2001 to December The S&P 500 Index revision events are hand-collected from changes in the monthly lists of index constituents in Compustat (Baran and King, 2012). We then use the Lexis-Nexis news retrieval system to verify the announcement day of the index revisions. The sample excludes stocks which were added due to a merger or takeover, spinoff, or change in share and contains a total of 248 addition and 241 deletion firms. We obtain information on stock returns and number of shares outstanding from the Center for Research in Security Prices (CRSP), and relevant firm-level financial data from the quarterly updated Compustat database. We collect information on CDS spreads, depth, and rating from Markit, a leading provider of CDS data for price discovery, risk and valuations, and end-of-day price updates. There are 989 North American firms in our CDS data from Markit in the period from 2001 to We use the five-year CDS spreads since they are the most liquid among different maturities. The CDS spreads represent the premium that the insurance buyer pays to exchange for the residual value of the debt from the insurance protector in case default occurs. We use the composite credit rating, which is the average rating of Fitch Ratings, Moody s Investors Service, and S&P s rating services, provided in Markit as our measure of credit rating. 3 Our sample of S&P 500 Index changes includes a longer time period, however we limit the time period given the data availability on CDS spreads in the Markit database. 11

12 Even though Markit has quite thorough coverage of the CDS markets, not every firm has CDS contracts traded on them. After merging with the Markit database, we find 128 addition firms and 152 deletion firms with available CDS data. Further, we require event firms to have CDS spread observations around the announcement dates. Our final sample contains 63 newly added firms and 42 firms that are dropped from the S&P 500 Index. Each year the number of firms which are added into or removed from the index varies. In our sample, more firms are added into the index in 2006 and Among the addition firms, 42 firms have investment grade ratings (ratings of BBB or above) and the remaining firms have speculative ratings (ratings that are below BBB) in the Markit database. To ensure that credit rating changes do not confound our findings, we examine the credit rating change history for each event firm around the index revision announcements. There are no credit rating changes for all short-term windows up to 15 days after the announcements. We find multiple credit rating changes for 13 firms over long-term windows, but the changes are usually in nearby rating categories, for example, from AA to A and back to AA. We believe that these rating changes do not confound our longterm results either Summary statistics Panels A and B of Table 1 present the summary statistics for 63 addition firms and 42 deletion firms, respectively. We compute the statistics across all the observations in the whole sample period. CDS spread is the daily composite five-year CDS spread in basis points. Market capitalization is calculated by the daily stock price times the number of shares. Equity volatility is the annualized standard deviation calculated based on daily 12

13 returns. Leverage is equal to the ratio of book value of debt to the sum of book value of debt and market capitalization. The book value of debt is computed as the sum of longterm debt and current liabilities in debt. Return on assets (ROA) is defined as net income divided by total assets. Market-to-book ratio is defined as the ratio of market value to book value of equity. We notice major differences between addition and deletion firms in terms of their average size: the market capitalization is $11.33 and $6.13 billion, and total assets are $33.60 and $10.33 billion, respectively. Even though both groups have similar average leverage, the profitability and valuation ratios are higher for the addition firms, with ROA of 1.01% for newly included firms, compared to 0.41% for the deletion firms. Despite similar leverage levels, we observe that the average CDS spread for the addition firms is basis points (bps), much lower than the average spread for deletion firms of bps. Furthermore, in untabulated results, we compare firm characteristics between those with and without CDS contracts for all firms that have had S&P 500 Index revisions. There is no significant difference in size, equity volatility, leverage, ROA, total assets, and market-to-book ratio between our sample firms and the group without CDS. This suggests that our sample is not contaminated by selection bias. [Table 1 about here] 4. Effect of S&P 500 Index Revision on CDS Spreads Addition to and deletion from the S&P 500 Index involve asymmetric stock price responses, as documented in Chen, Noronha, and Singal (2004). In this section, we 13

14 investigate the effects on abnormal CDS spreads from the S&P 500 Index revisions in both the short- and long-term Short-term effects on abnormal changes of CDS spreads Table 2 reports the market adjusted cumulative abnormal changes of CDS spreads for event firms in various short-term windows. We use the cross-sectional CDS spread average from the whole Markit CDS sample as the benchmark index to calculate market adjusted abnormal CDS spread changes. The abnormal change for an event firm is computed as the difference between the firm s CDS spread changes ((CDS it - CDS it-1 )/ CDS it-1 ) and the market CDS spread changes, i.e., the average change of all available CDS spreads in our CDS Markit sample, computed as ((CDS mt - CDS mt-1 )/ CDS mt-1 ) where CDS mt = N j=1 CDS jt N. 4 We apply cross-sectional t-tests, sign tests, and signed rank tests to investigate whether the cumulative abnormal CDS changes are significantly different from zero. The sign and signed rank tests are non-parametric tests, which do not require the data to be normally distributed. Although the signed rank test has more statistical power than the sign test, the latter would have more statistical power for data with outliers or a heavy-tailed distribution. These two tests are similar to the t-test and serve as additional robustness checks. We report the results from both addition and deletion firms. 5 For the addition sample, we find significant reductions in abnormal CDS spreads upon index inclusion. The CDS spreads react strongly to the inclusion event in the short windows around the 4 We perform the robust tests using (CDS it - CDS it-1 ) as the spread changes, and the results are similar. These results are available upon request. 5 In a robustness test, we winsorize the abnormal CDS spread changes at the top and bottom one percentile levels to mitigate the influence from outliers and compute the cumulative abnormal changes in each event window. The results are similar as those reported in the paper. The results are not reported here to save space, but available upon request. 14

15 announcement date. There are strong negative cumulative abnormal changes two days before the announcement date. For example, in the window [AD-2, AD+1] (AD is the abbreviation for announcement date), we observe significant cumulative abnormal change of percent. We also observe further negative movement in the credit market in the windows [AD-2, AD+n], in which n is 15 days or fewer. Our evidence is consistent with the findings in the equity market, 6 which document short-term stock price surge around the announcement date that a firm is added to the S&P 500 Index (Harris and Gurel, 1986; Lynch and Mendenhall, 1997). We do not find significant reactions in the CDS market upon announcement of the firms deletion from the S&P 500 Index in all event windows in similar tests, where the abnormal CDS spread change is similarly measured. This finding is contrary to those in the equity market, which document a stock price decline immediately after the announcement but a reversal after 60 days (e.g., Chen, Noronha and Singal, 2004). As the CDS market captures the downside risk, CDS market participants probably possess similar information on the reference entity firm that is deleted from the S&P 500 Index. Thus unlike its price effect in the equity market, the information of the S&P 500 Index removal has little or no impact on the CDS market. Due to the insignificant effect of deletion news on CDS spreads, we focus on addition announcements in our subsequent analysis. 7 [Table 2 about here] 6 We also check the stock abnormal returns around the S&P 500 Index revisions [AD-1, AD+1] in our sample, and find both addition and deletion have strong impact on the stock market, which is consistent with the previous literature. The results are available upon request. 7 We conduct all the empirical tests described in this paper for both addition and deletion firms and report only results for addition firms as the results for deletion firms are not significant. These results are available upon request. 15

16 4.2.Effects on abnormal changes of CDS spreads in the matching sample We next construct a matching sample by industry and size and compare the difference in abnormal CDS spread change between the event firms and the matches. As Spiegel points out in his 2015 EFA Annual Meeting keynote speech (Spiegel and Tookes, 2015), a broad match based on two-digit SIC code may not correct for industry specific changes among firms as the two firms can be very faraway in the nature of their businesses. Since the CDS market participants are well-informed, the information that causes the abnormal CDS spread change upon addition announcements is likely to be S&P s confirmation of the reference entity firm s representativeness in its industry. Therefore a closer match is desirable to identify the existence of such information. We identify industry matches based on the four-digit SIC codes of the event firms. 8 Then we choose firms within the same industry and with similar sizes as the event firms to construct the matching sample. 9 Since not every traded firm has CDS data and S&P 500 Index addition firms are usually large, we find matching firms for about two thirds (2/3) of the event firms. This further reduces our sample for this analysis to 42 index inclusions. We compare the firm characteristics (size, equity volatility, leverage, market-to-book ratio) of the matching firms and the event firms. Overall, our matching sample mimics the characteristics of the event firms quite well, as the t-test statistics of the differences on market capitalization, total assets, ROA and stock volatility range between 0.10 and The leverage of the matching firms is slightly higher than the event firms but the difference remains statistically insignificant (t-stat=1.60). The 8 We also use the three-digit and two-digit SIC code to match for industry, and find significant differences between the newly added firms and SIC matches. 9 We chose firms in the SIC code matching industry, and then select the firms that are closest in size to the event firms. 16

17 matches also tend to have slightly higher market-to-book ratio (t-stat=1.76). We report the above in Panel A of Table 3. The results on the abnormal CDS spread change of the event firms, the matches, and their differences are presented in Panel B of Table 3. We find that only event firms have a significantly negative abnormal CDS spread change while their close matches do not respond to the addition announcements in all event windows. With the close industry matches based on four-digit SIC code, the difference in cumulative abnormal CDS spread change between the event firms and the matches is statistically significant in all the shortterm windows, with a confidence level of 5% or better. To further investigate the effect of addition announcement on the industry, we investigate whether the cumulative abnormal CDS spread changes vary with event firms industry weight. This test is based on arguments in Cai (2007), who shows that as S&P 500 Index addition brings new information to the industry, the stock price reaction of the matches is smaller in magnitude when the event firm s industry weight is high. Our untabulated results from the test find no statistically significant difference for the abnormal CDS spread changes of the matches in multiple event windows. Therefore, despite the findings documented in the equity market, we find little evidence that S&P 500 Index addition announcements involve information for matching firms in the CDS market. Next, we conduct more empirical tests in an attempt to narrow down the possible explanations for our findings. [Table 3 about here] 4.3. Liquidity effect 17

18 One possible reason for strong reduction of abnormal CDS spread changes in both short- and long-term is liquidity improvement of the event firms CDS contracts after the addition which reduces information asymmetry and leads to the decrease of credit risk (Bongaerts, de Jong, and Driessen, 2011; Tang and Yan, 2007). Hegde and McDermott (2003) make such arguments for the added stocks after S&P 500 Index announcements by showing a sustained liquidity increase in the added stocks. Many other studies also document lower bid-ask spreads due to the index addition, such as Shleifer (1986), and Beneish and Whaley (1996). To explore this possibility, we investigate the liquidity changes in the CDS market after the S&P 500 Index addition announcement. We use market depth to measure CDS liquidity, which computes the number of the contributors for the price quote on each day. We do not have access to the CDS trading volume data from the Depository Trust & Clearing Corporation. The bid or ask prices are not available through the Markit database either, so we do not have these alternative liquidity measures for the CDS market. Instead, we use market depth, which is widely used in CDS studies (Kapadia and Pu, 2012; Loon and Zhong, 2014) and properly captures the liquidity level in the CDS market. Table 4 presents the addition announcement effect on CDS market liquidity in the full sample, a subsample with investment grade firms only, and a subsample with speculative grade firms only, respectively. We report the results on liquidity change over a short-term window in Panel A. Across all the short-term windows around the announcement date, the average market depth is stable with about six contributors for each price quote. We do not observe substantial improvement or deterioration of the credit market liquidity before or after the addition. In several t-tests for the difference two 18

19 days before (AD-2) announcement and n days after announcement (AD+n, n=1, 2, 3, 5, 10, and 15), the t-statistics are all insignificant. The robust results across all the subsamples suggest that the CDS market liquidity over short-term does not change because of the addition events. Panel B of Table 4 reports the results on liquidity change for the three groups of firms over long-term windows. Not surprisingly, we do not find significant depth changes for the event firms in various window lengths. All the t-test statistics in the event windows are insignificant. The long-term depth does not experience substantial change due to the addition news, consistent with the previous findings in short-term windows. Therefore, abnormal CDS spread declines following inclusion cannot be explained by improvement in CDS liquidity. To see whether improvements in stock liquidity are related to CDS spread declines, we estimate the impact of the changes in stock liquidity on the cumulative CDS spread change in a multiple regression setting and find that the improved stock market liquidity cannot explain our findings. 10 In summary, the insignificant liquidity changes in market depth around the addition event suggest that liquidity is not the main driver for the substantial abnormal changes in the CDS spreads. This is not surprising, as the CDS market is a contract-based market where liquidity is less influential (Longstaff, Mithal, and Neis, 2005). Since stock liquidity does not explain the abnormal changes in the CDS spreads, we conclude that liquidity hypothesis is not a good explanation for the significant reduction in cumulative abnormal CDS spread change of the addition firms. [Table 4 about here] 4.4 Effects on abnormal changes of CDS spreads in the long-term 10 We follow Amihud (2002) to measure the stock market liquidity and report these results in Table 8. 19

20 While many studies (Harris and Gurel, 1986; Hegde and McDermott, 2003; Becker-Blease and Paul, 2006) examine the short-term reaction of stock prices to S&P 500 Index addition, Chan, Kot, and Tang (2013) investigate long-term effects of S&P 500 Index revisions and document significant price effects. We carry our investigation of S&P 500 Index addition effect in the credit market to time periods up to three years to examine whether the addition events convey information to the credit market in the longterm. 11 Specifically, we examine the long-term impact of S&P 500 addition on abnormal CDS spread changes. If S&P 500 Index addition carries long-term positive information of the firm, the credit spreads should decrease and should not reverse in the long-term. Table 5 presents the long-term cumulative abnormal CDS spread changes in various event windows across the whole sample. We compare the firm characteristics, such as size, leverage, and market-to-book ratio, at the beginning and end of the event windows, and do not find significant difference. However, in one, two, and three years 12 after the announcement date, the CDS cumulative abnormal changes are negative and significant at one percent level. Our findings suggest that firms added to the S&P 500 Index have shrinking CDS spreads in the long term. One possible reason could be that these added firms may be subject to a higher level of scrutiny by investors and analysts which helps to reduce information asymmetry (Denis, McConnell, Ovtchinnikov and Yu, 2003) and improves firm performance. The strong CDS price reaction in the long-term suggests that the favorable information from S&P 500 Index addition is persistent, and the short-term 11 We check the long-term reaction in the abnormal CDS changes of the deletion firms and find that deletion from the S&P 500 has little impact in the CDS market. Results are available upon request. 12 We assume that there are about 252 trading days in a year so that one-, two-, and three-year windows correspond to 252, 504, and 756 trading days. 20

21 decrease in abnormal CDS spreads does not reverse in the long-term. We also checked the long-term reaction for deletion firms, but did not find significant results. [Table 5 about here] 5. Additional Tests 5.1 Subsample Analysis Market integration is stronger in the financial crisis (Bekaert, Harvey, and Ng, 2005; Allen and Gale, 2000). If S&P 500 Index addition involves information flow from the equity market to the CDS market, we expect to observe a more pronounced effect of the S&P 500 Index additions on the CDS market during crisis periods. Table 6 reports the results of abnormal CDS changes in various short-term windows in the crisis and noncrisis periods with the financial crisis period being defined as from 2008 to Out of the 63 addition firms, there are 13 additions during the financial crisis period and 50 out of the crisis period. We find that the cumulative abnormal CDS spread changes are significant in all the short-term windows when additions occurred during the crisis, but firms with additions during the non-crisis period experience CDS spread declines that are largely statistically insignificant. Specifically, in the window [AD-2, AD+1], the average CDS spread decrease for additions during crisis period is about 2.57 percent, which suggests a strong reaction in the CDS market to the index inclusion announcement. However, for additions occurring during non-crisis periods, only in one post-announcement window ([AD-2, AD+10]) is the cumulative abnormal CDS change significantly different from zero, although all windows show CDS spread declines. 13 We do not examine the events in the dot-com crash of 2002 due to the small sample size. 21

22 Our observations in Table 6 are consistent with the various studies that report the low correlations between the equity and credit markets (e.g., Collin-Dufrense, Goldstein and Martin, 2001; Blanco, Brennan and Marsh, 2005) in the tranquil time period. Thus even if the S&P 500 Index inclusion announcement involves relevant information on firm value, it may not be reflected in the CDS spreads during non-crisis periods. Therefore, our finding that the impact of S&P 500 Index addition is more pronounced during the crisis could be due to the higher correlation between the stock and credit markets during this period of time (Kapadia and Pu, 2012) and the contagion effect in the financial crisis (Allen and Gale, 2000). [Table 6 about here] Next, we investigate how firms with different credit ratings react to the S&P 500 Index additions in both short- and long-term and report the results in Table 7. Addition to the S&P 500 Index could be a certification of good quality (Chen, Noronha, and Singal, 2006; Baran and King, 2012), and there will be a differential effect on addition firms with varying credit quality. Thus, our expectation is that the addition of speculative grade firms would send stronger positive signals to the credit market than those investment grade firms, leading to greater reductions in the abnormal CDS spreads. Out of the 63 addition firms, there are 42 additions with investment grade ratings and 21 with speculative grade ratings. Consistent with our conjecture, we observe the average magnitude of cumulative abnormal CDS spread changes for speculative grade firms is larger than that of the investment grade firms, by about one to five percentage points in the short-term. For example, in the window [AD-2, AD+5], speculative grade 22

23 firms have an average of 2.62 percent CDS spread reduction while investment grade firms do not significantly respond to the index addition news. Panel B of Table 7 shows an interesting fact that the reaction of speculative grade firms is not as large as that of investment grade firms in longer windows. In the [AD-2, AD+252] window, the reaction of the speculative rated firms is stronger than investment grade firms, while in the ([AD-2, AD+504] window the average cumulative abnormal CDS spread change of the investment grade firms is more than 10 percentage points lower than that of the speculative grade firms. This contrasts with the earlier observation in the short-term windows. Therefore, as addition news provides certification from S&P, which is both a rating agency and the index provider, speculative grade firms benefit more in the short-term. For lower rated firms, CDS investors may pay more attention to the short-term changes in default risk, since they are closely related to the debt valuation of these firms. Thus, the positive news may have more immediate effect for speculative grade firms in the short-term. Even though the positive impact does not last over the long-term, it does not reverse either. In two of the long-term after-announcement windows, the long-term CDS cumulative abnormal changes for speculative grade firms are negative and marginally significant. The cumulative abnormal changes for the investment grade firms are significant for all the long-term windows. These findings are also consistent with past literature on certification effect, which suggests that it is more pronounced with smaller and less prestigious firms (Megginson and Weiss, 1991). [Table 7 about here] Crises are hard to predict and represent a relatively exogenous shock (Lemmon and Lins, 2003; Chen, Ma, Malatesta and Xuan, 2011). Furthermore, we do not expect 23

24 demand for index stocks to be stronger during financial crises periods or for stocks with speculative rating. Even though it is hard to differentiate the demand-related hypotheses and the certification hypothesis using CDS market responses to S&P 500 Index addition announcements, we believe that results from our subsample analysis are more consistent with the certification hypothesis that suggests information is involved in S&P 500 Index revisions. 5.2 Multivariate Regression Analysis To explore whether our subsample results remain robust after controlling for other firm characteristics that influence CDS spreads, we conduct a regression analysis and present the results in Table 8. The dependent variable is the cumulative abnormal CDS spread changes in the window [AD-2, AD+15]. Specifically, we control for firm size, leverage, market-to-book ratio, and ROA, as well as stock liquidity, denoted as Liquidity and measured following Amihud (2002). We also include two constructed dummy variables: Crisis that takes value 1 for additions during the crisis period and 0 otherwise; InvestGrade that takes value 1 for reference entity firms with investment grade and 0 otherwise. Similar to the results in Table 6, the coefficient on Crisis is negative and significant at the 5% level. In addition, supporting the findings in Table 7, the coefficient on InvestGrade is positive and significant. The coefficient on Liquidity is not significant. Since Liquidity is correlated with Crisis (the correlation coefficient is 0.51) and the estimation may be subject to multicollinearity concerns, we report the results from a regression without including Liquidity in a separate column (2). The coefficients on Crisis and InvestGrade remain significant with a slightly better confidence level. [Table 8 about here] 24

25 6. Conclusions To the best of our knowledge, this study is the first to provide a comprehensive analysis of the CDS market reaction to the firms added to or deleted from the S&P 500 Index. During the period from 2001 to 2012, we find that the CDS market reacts significantly to the favorable addition information in both short- and long-term windows. Deletion from the index has little impact on the CDS market. Different from the findings in the equity market, the addition information does not have impact on the industry peers of the event firms. Liquidity changes in the CDS market are insignificant and do not explain the cumulative abnormal CDS spread changes. Neither does stock liquidity improvement. Our findings show that the CDS market absorbs the S&P 500 Index revision information in a timely manner during the financial crisis while the effect is weaker in the non-crisis period. In addition, our results indicate that S&P 500 Index additions convey more positive information to firms with speculative credit rating in the short run. In the long run, the investment grade firms have persistent negative abnormal CDS changes while the results of the speculative grade firms are weaker but do not reverse. These findings support the hypothesis that addition to the S&P 500 Index provides a quality certification of the firm. Future research could use an expanded sample of revision firms to explore other factors that may influence how information flows between the equity and credit markets. 25

26 References Acharya, V., and T. Johnson, Insider trading in credit derivatives. Journal of Financial Economics 84 (1), Allen, F. and D. Gale, Financial contagion. Journal of Political Economy 108, Amihud, Y., 2002, Illiquidity and stock returns: Cross-section and time series effects. Journal of Financial Markets 5, Amihud, Y. and H. Mendelson, Asset pricing and the bid-ask spread. Journal of Financial Economics 17, Baran, L. and T.H.D. King, Cost of equity and S&P 500 Index revisions. Financial Management 41, Becker-Blease, J.R. and D.L. Paul, Stock liquidity and investment opportunities: Evidence from index additions. Financial Management 35, Beneish, M.D. and J.C. Gardner, Information costs and liquidity effects from changes in the Dow Jones Industrial Average list. Journal of Financial and Quantitative Analysis 30, Beneish, M.D. and R.E. Whaley, An anatomy of the S&P Game : The effects of changing the rules. Journal of Finance 51, Bekaert, G., C. Harvey, and A. Ng, Market integration and contagion. Journal of Business 78, Blanco, R., S. Brennan, and I.W. Marsh, An empirical analysis of the dynamic relation between investment-grade bonds and credit default swaps. Journal of Finance 60, Bongaerts, D., F. de Jong, and J. Driessen, Derivatives pricing with liquidity risk: Theory and evidence from the credit default swap market. Journal of Finance 66, Cai, J., What s in the news? Information content of S&P 500 additions. Financial Management 36, Chan, K., H.W. Kot, and G.Y.N. Tang, A comprehensive long-term analysis of S&P 500 Index additions and deletions. Journal of Banking & Finance 37, Chen, L., Y. Ma, P. Malatesta, and Y. Xuan, Ownership structure and the cost of corporate borrowing, Journal of Financial Economics 100, Chen, H., G. Noronha, and V. Singal, The price response to S&P 500 Index additions and deletions: Evidence of asymmetry and a new explanation. Journal of Finance 59, Chen, H., G. Noronha, and V. Singal, Index changes and losses to index fund investors. Financial Analysts Journal 62,

27 Chung, R. and L. Kryzanowski, Are the market effects associated with revisions to the TSE300 Index robust? Multinational Finance Journal 2, Collin-Dufresne, P., R.S. Goldstein, and J.S. Martin, The determinants of credit spread changes. Journal of Finance 56, Denis, D.K., J.J. McConnell, A.V. Ovtchinnikov, and Y. Yu, S&P 500 Index additions and earnings expectations. Journal of Finance 58, Dhillon, U. and H. Johnson, Changes in the Standard and Poor s 500 list. Journal of Business 64, Elliott, W.B. and R.S. Warr, Price pressure on the NYSE and Nasdaq: Evidence from S&P 500 Index changes. Financial Management 32, Greenwood, R., Short- and long-term demand curves for stocks: theory and evidence on the dynamics of arbitrage. Journal of Financial Economics 75, Harris, L. and E. Gurel, Price and volume effects associated with changes in the S&P 500 list: New evidence for the existence of price pressures. Journal of Finance 41, Hegde, S.P. and J.B. McDermott, The liquidity effects of revisions to the S&P 500 Index: An empirical analysis. Journal of Financial Markets 6, Hrazdil, K., The effect of demand on stock prices: New evidence from S&P 500 weight adjustments. Managerial Finance 35, Hrazdil, K. and T. Scott, S&P 500 Index revisited: Do index inclusion announcements convey information about firms future performance? Quarterly Journal of Finance and Accounting 48, Hull, J., M. Predescu, and A.White, The relationship between credit default swap spreads, bond yields, and credit rating announcements. Journal of Banking & Finance 28, Jain, P.C., The effect on stock price from inclusion in or exclusion from the S&P 500. Financial Analysts Journal 43, Kapadia N. and X.L. Pu, Limited arbitrage between equity and credit markets. Journal of Financial Economics 105, Kappou, K., C. Brooks, and C.W.R. Ward, A re-examination of the index effect: Gambling on additions to and deletions from the S&P 500 s gold seal. Research in International Business and Finance, Elsevier, vol. 22(3), Kaul, A., V. Mehrotra, and R. Morck, Demand curves for stock do slope down: New evidence from an index weights adjustment. Journal of Finance 55, Lemmon, M.L. and K. Lins, Ownership structure, corporate governance, and firm value: evidence from the East Asian financial crisis, Journal of Finance 58,

28 Longstaff, F., S. Mithal, and E. Neis, Corporate yield spreads: Default risk or liquidity? New evidence from the credit default swap market. Journal of Finance 60, Loon, Y.C., and Z.D.K. Zhong, The impact of central clearing on counterparty risk, liquidity, and trading: Evidence from the credit default swap market. Journal of Financial Economics 112, Lynch, A.W. and R.R. Mendenhall, New evidence on stock price effects associated with changes in the S&P 500 Index. Journal of Business 70, Megginson, W. and K. Weiss, Venture capitalist certification in initial public offerings. Journal of Finance 46, Merton, R.C., On the pricing of corporate debt: the risk structure of interest rates. Journal of Finance 29, Merton, R.C., Presidential address: A simple model of capital market equilibrium with incomplete information. Journal of Finance 42, Platikanova, P., The long-term price effect of S&P 500 Index addition and earnings quality. Financial Analysts Journal 64, Scholes, M.S., The market for corporate securities: Substitution versus price pressure and the effects of information on share price. Journal of Business 45, Shankar, S.G. and J.M. Miller, Market reaction to changes in the S&P SmallCap 600 Index. Financial Review 41, Shleifer, A., Do demand curves for stocks slope down. Journal of Finance 41, Spiegel, M. and H. Tookes, An IPO s impact on rival firms. Working paper, Yale University. Tang, D. Y. and H. Yan, Liquidity and credit default swap spreads. Working paper, University of Hong Kong. 28