CDS Momentum: Slow-Moving Credit Ratings and Cross-Market Spillovers
CDS Momentum: Slow-Moving Credit Ratings and Cross-Market Spillovers
Lee, Jongsub; Naranjo, Andy; Sirmans, Stace
2021-05-21 00:00:00
Abstract This paper highlights the adverse consequences of sluggish credit rating updates in creating information efficiency distortions and investment anomalies. We first document significant credit default swap (CDS) return momentum yielding 7.1% per year. We further show that cross-market momentum strategies based on information in past CDS performance generates an alpha of 10.3% per year in stocks and 7.3% per year in bonds. These CDS momentum and cross-market effects are stronger among more liquid, informationally rich CDS contracts whose CDS spreads move in anticipation of important, yet slow-moving, credit rating changes. (JEL G12, G14) Received February 19, 2020; editorial decision July 10, 2020 by Editor Jeffrey Pontiff. Authors have furnished an Internet Appendix, which is available on the Oxford University Press Web site next to the link to the final published paper online. Related markets often reveal information nonsynchronously, with different frequencies, speeds, and content. An important question is to what extent does a more informationally efficient market trading alongside a related sluggish, but material, information market generate market anomalies? A case in point is the credit default swap (CDS) market and credit ratings, where quick-moving CDS at-market spreads have an informational advantage and often predict credit ratings (Flannery, Houston, and Partnoy 2010; Hull, Predescu, and White 2004; Lee, Naranjo, and Sirmans 2016). For example, Figure 1 displays the time series of S&P credit ratings and 5-year CDS at-market spreads for four well-known companies (i.e., Ford Motor, Citigroup, Marriott International, Avon Products). This figure illustrates that CDS spreads move continuously in anticipation of upcoming discrete rating changes. Figure 2 further illustrates the relations between CDS spreads and credit rating changes centered on downgrades and upgrades, respectively. For both directional changes, CDS spreads move several months ahead of future rating changes. CDS spreads gradually run up (or run down) 90 days prior to an upcoming rating downgrade (or upgrade). Over this 90-day credit rating change cycle, CDS spreads show continuously positive autocorrelations until and upon the eventual rating change. This anticipatory pattern of CDS spreads foreshadows potential CDS return momentum effects arising from sluggish and discrete credit rating updates.1 Figure 1 Open in new tabDownload slide Examples of CDS spreads and S&P credit ratings This figure displays time series of firm-level CDS spreads and S&P credit rating for four companies in various industries: Ford Motor Company (panel A), Citigroup Incorporated (panel B), Marriott International (panel C), and Avon Products Incorporated (panel D). A +1 notch change in S&P credit ratings represents a downgrade based on the rating scale that includes subratings at each level between AA and D (e.g., AA+, AA, and AA-). CDS data come from Markit. The time period is January 2003 to December 2015. Figure 1 Open in new tabDownload slide Examples of CDS spreads and S&P credit ratings This figure displays time series of firm-level CDS spreads and S&P credit rating for four companies in various industries: Ford Motor Company (panel A), Citigroup Incorporated (panel B), Marriott International (panel C), and Avon Products Incorporated (panel D). A +1 notch change in S&P credit ratings represents a downgrade based on the rating scale that includes subratings at each level between AA and D (e.g., AA+, AA, and AA-). CDS data come from Markit. The time period is January 2003 to December 2015. Figure 2 Open in new tabDownload slide CDS spreads and S&P credit rating changes This figure displays a plot of the cumulative percentage change in 5-year CDS spread over the interval [−90,90] days centered on rating downgrade events (top figure) and upgrade events (bottom figure). The figure includes 1,342 rating downgrade events averaging a 1.36 notch deterioration in credit quality and 922 rating upgrade events averaging a 1.28 notch improvement in quality. A +1 notch change in S&P credit ratings represents a downgrade based on the rating scale that includes subratings at each level between AA and D (e.g., AA+, AA, and AA-). CDS data come from Markit. The time period is January 2003 to December 2015. Figure 2 Open in new tabDownload slide CDS spreads and S&P credit rating changes This figure displays a plot of the cumulative percentage change in 5-year CDS spread over the interval [−90,90] days centered on rating downgrade events (top figure) and upgrade events (bottom figure). The figure includes 1,342 rating downgrade events averaging a 1.36 notch deterioration in credit quality and 922 rating upgrade events averaging a 1.28 notch improvement in quality. A +1 notch change in S&P credit ratings represents a downgrade based on the rating scale that includes subratings at each level between AA and D (e.g., AA+, AA, and AA-). CDS data come from Markit. The time period is January 2003 to December 2015. Researchers generally attribute the CDS market’s informational advantage to its unique structure in which trades are brokered by bulge bracket investment banks who are well-informed about the overall capital markets (Acharya and Johnson 2007; Qiu and Yu 2012). Branching further into the information content of CDS contracts, others have also recently examined cross-market informational linkages from CDS to equity markets (Lee, Naranjo, and Velioglu 2018). These studies find significant firm-specific information flows from CDS to stock markets, particularly for entities with negative news, more informed insiders, such as relationship banks, high CDS contract liquidity, and private companies. Several studies using CDS market data (Friewald, Wagner, and Zechner 2014) are also able to reconcile well-known asset pricing puzzles, such as a distress puzzle (Campbell, Hilscher, and Szilagyi 2008), that arise when using traditional default risk measures or corporate bond yields. These findings suggest that CDS markets play important information signaling functions and could be helpful in addressing various market inefficiencies. In this paper, we provide evidence on return anomalies and cross-market spillover effects in the CDS market stemming from the influence of credit ratings. We focus our tests on the return momentum anomaly and address two important questions. First, does momentum exist in CDS returns? Second, are there momentum spillovers from the CDS market to the bond and stock markets? Given the sophisticated participants in the CDS market, one might expect that CDS momentum profits would not exist.2 However, while changes to ratings move slowly, they materially affect firm fundamentals (Dichev and Piotroski 2001; Kisgen 2006,2007; Kliger and Sarig 2000), and such changes could create common market frictions in related firm securities, including stocks and bonds belonging to the same firm (Manso 2013). Through the anticipation of the upcoming real consequences, past CDS performance could signal future return momentum in related securities’ markets. Using 5-year CDS contracts on 871 U.S. firms from January 2003 to December 2015, we document several novel findings. First, we find significant CDS return momentum. A 3-month formation and 1-month holding period CDS return momentum strategy yields 0.59% per month. The performance is better for entities with lower credit ratings (1.10% per month for junk-grade entities) and high CDS trading liquidity (1.05% per month in the highest depth tercile). A $ 100 investment in this strategy starting in January 2003 grows to approximately $ 235 by December 2015 with an annualized Sharpe ratio of 0.99 (see panel A of Figure 3). Given annualized Sharpe ratios of 0.29 and 0.45, respectively, for U.S. equity value and momentum strategies in the post-1972 period (Asness, Moskowitz and Pedersen 2013), these CDS momentum profits are substantial even on a simple risk-adjusted basis. The strategy also tends to perform better during the crisis period (0.91% per month for the period of January 2008 to December 2011). We further show that CDS momentum returns are robust to using various formation and holding periods, controlling for fundamental risk factors from both equity and bond markets, and are also profitable on a transaction-cost-adjusted basis. Finally, we show that CDS momentum profits are generated from incremental information in past CDS returns above and beyond information possessed in past stock and bond returns. Figure 3 Open in new tabDownload slide CDS momentum This figure displays various representations of the performance of our CDS momentum strategy using a formation period J=3m and rebalanced monthly. Panel A shows the cumulative wealth of a $100 investment, and panel B shows the 24-month moving average monthly return. The long-short CDS momentum strategy is constructed by purchasing (selling short) CDS contracts in the highest (lowest) quintile of past 3-month CDS performance. Portfolio constituents are equally weighted. CDS data are from Markit. The time period spans January 2003 to December 2015. Figure 3 Open in new tabDownload slide CDS momentum This figure displays various representations of the performance of our CDS momentum strategy using a formation period J=3m and rebalanced monthly. Panel A shows the cumulative wealth of a $100 investment, and panel B shows the 24-month moving average monthly return. The long-short CDS momentum strategy is constructed by purchasing (selling short) CDS contracts in the highest (lowest) quintile of past 3-month CDS performance. Portfolio constituents are equally weighted. CDS data are from Markit. The time period spans January 2003 to December 2015. What are the sources of these CDS momentum profits? As conjectured earlier, we find that the profits are almost entirely a result of correct anticipation of future credit rating changes by the CDS market. That is, the winner (loser) portfolio is driven by the positive (negative) returns of firms that undergo rating upgrades (downgrades) in the 6 months following portfolio formation. We find that CDSs generally experience a significant cumulative return run-up of 2.53% (return run-down of -3.48%) in the 6 months leading up to a rating upgrade (downgrade), which is further followed by a return of 0.78% (-2.39%) during the month of the announcement.3 This “anticipated” credit rating change channel works best for low-grade firms and CDS contracts with high depth, which are arguably the most demanded by informed traders in the credit market. This rating change channel also distinguishes the sources of CDS momentum from those of corporate bond momentum as similar tests performed by Jostova et al. (2013) show no relation between bond momentum and credit rating changes. Is CDS momentum simply a manifestation of bond momentum? No. We provide evidence supporting recent research highlighting CDS as a nonredundant security. In particular, the CDS market is argued to provide an important economic function through the standardized nature of the CDS contract, which leads the CDS to have lower trading costs and greater liquidity than the cash bond and makes the CDS market the preferred trading venue for speculators and the leader in price discovery (Das, Kalimipalli and Nayak 2014; Oehmke and Zawadowski 2015, 2016). Consistent with this view, we show that the information contained in past performance of the CDS is superior to that of the bond. Not only do we find that CDS momentum is robust to past bond performance but also we find that past bond performance, after accounting for past CDS performance, has no ability to predict future CDS returns. This finding is consistent with the recent findings in Lee, Naranjo and Velioglu (2018). Despite the distinct mechanisms underlying the momentum effects in bond and CDS markets, the two related securities are still linked through arbitrage activities. Therefore, one would expect a higher degree of similarity between CDS and bond momentum for firms with CDS contracts. In this regard, we find that the rating change channel of CDS momentum, in fact, can be applied to bond momentum as well, but only for firms with CDS. That is, removing firms that undergo future rating changes entirely eliminates (the already much weaker) bond momentum profits for CDS-matched firms but minimally affects bond momentum profits for non-CDS-matched firms. Furthermore, even within the cross-section of CDS-matched firms, we show that firm-level impediments to arbitrage, such as fragmentation of bond issues (Oehmke and Zawadowski 2016), weaken the link and exacerbate the disconnect between the momentum effects in the two markets. Second, we find significant cross-market momentum spillover from the CDS market to both bond and stock markets. A long-short strategy based simply on past CDS performance generates up to 0.53% in monthly bond returns and 0.62% in monthly stock returns. To address correlation in the cross-sectional ranking of performance across markets, we construct strategies using conditional portfolio sorts on past CDS performance while controlling for past bond or stock performance. We find that the conditional momentum spillover strategies perform significantly better than the unconditional strategies, providing statistically significant returns of up to 0.59% (Sharpe ratio of 2.14) for CDS-to-bond strategies and 0.79% (Sharpe ratio of 1.00) for CDS-to-stock strategies. This CDS momentum spillover is unexplained by common stock and bond risk factors and is unrelated to traditional stock momentum crash risk (Barroso and Santa-Clara 2015; Daniel, Jagannathan, and Kim; Daniel and Moskowitz 2016). In fact, the conditional CDS-to-stock spillover strategy experienced a positive return of 6.71% during the worst month of traditional stock momentum during our sample period in April 2009 when the Fama-French UMD factor returned a detrimental −34.58%. We also find this cross-market spillover to be stronger among junk-grade firms as well as firms with high CDS depth, suggesting that substantial price discovery occurs in the CDS market relative to the stock and bond markets for entities that attract the highest hedging demand for underlying credit risk as well as contracts with more trade quotes. Moreover, our cross-market spillover stock trading strategy further minimizes transaction cost adjusted return effects (Frazzini, Israel, and Moskowitz 2012) as we trade stocks with CDS, which tend to be the stocks of larger firms in the NYSE/AMEX-listed universe. We also confirm a close relation between the performance of our cross-market spillover strategies and the significant predictability of the incremental information in CDS returns on future credit rating changes in “anticipated” directions. In these spillover strategies, CDS returns assist in maintaining an ideal exposure to future credit rating change events: a net positive exposure to upgrades and a net negative exposure to downgrades. Until the anticipated events occur, the long-short strategies benefit from the realized price divergence between winner firms that experience an upgrade and loser firms that experience a downgrade. Overall, our results suggest that greater information signaling in the CDS market, together with sluggish updates on corporate credit ratings assigned by major rating agencies, creates anomalies, such as return momentum within the CDS market as well as momentum spillover from the CDS market to the related stock and bond markets.4 We make several important contributions to the literature on return momentum and capital market efficiency, credit risk, and cross-market linkages. In particular, we document profitable CDS return momentum investment strategies, which extends the literature on the existence of return momentum across various asset classes to an economically important and growing market (Asness, Moskowitz and Pedersen 2013; Beyhaghi and Ehsani 2017; Menkhoff et al. 2012; Miffre and Rallis 2007; Okunev and White 2003; Pirrong 2005). Importantly, we show that the underlying mechanisms of CDS and corporate bond return momentum (Jostova et al. 2013) are different and provide evidence of a link between CDS return momentum and the predictability of past CDS returns on future rating changes in “anticipated” directions. In this regard, we contribute to the literature on the informational relation between CDS markets and credit rating agencies, which documents the effectiveness of market CDS rates as a potential alternative credit risk benchmark to credit ratings (Acharya and Johnson, 2007; Flannery, Houston, and Partnoy 2010; Hull, Predescu, and White 2004; Lee, Naranjo, and Sirmans 2016; Norden and Weber 2004; Qiu and Yu 2012). Relatedly, our CDS return momentum channel highlights how sluggish moves by major ratings agencies in assigning corporate credit ratings influences capital market efficiency (Manso 2013). We are also the first to document significant cross-market return momentum spillovers from CDS to stock and bond markets, extending the literature on related asset cross-market interaction effects in momentum returns (Gebhardt, Hvidkjaer and Swaminathan 2005) and recent research on the important economic role of the CDS market that makes it a “preferred trading venue” for speculators and a leader in price discovery (Das, Kalimipalli, and Nayak. 2014 2014; Oehmke and Zawadowski 2015, 2016). Finally, we highlight the role of CDS depth in explaining significant predictability of CDS returns on future rating changes and also the incremental information content in CDS returns over stock and bond returns. This depth effect is consistent with the recent findings on endogenous liquidity in CDS markets, whereby more information flows from the CDS to the stock when the liquidity of a CDS contract is high (Qiu and Yu 2012). 1. Data and CDS Return Computation We obtain data from a variety of sources. The sample covers 871 publicly held U.S. companies from January 2003 to December 2015 for which we obtain CDS, bond, and equity price data. CDS data are acquired from the Markit Group, a leading financial information services company. All of our selected CDS contracts have 5-year maturities since these corporate CDS contracts are the most liquidly traded, and they are denominated in U.S. Dollars. The “Big Bang” protocol of April 2009 changed the standard for CDS contracts on a number of dimensions, including a move from Modified Restructuring (MR) to No Restructuring (XR) for North American corporate CDS contracts.5 As such, our database consists of MR contracts prior to the “Big Bang” and XR contracts afterward. Markit constructs a composite CDS spread using input from a variety of market makers and ensures each daily observation passes a rigorous cleaning test to ensure accuracy and reliability.6Table 1 provides summary statistics and a correlation matrix of variables used in this study. The mean spread of CDS contracts in our sample is approximately 185 basis points. As a measure of liquidity, Markit reports on a daily basis each firm’s CDS market “depth,” or the number of distinct contributors providing quotes used to construct the composite spread. Markit requires a minimum of two contributors. The mean depth of our sample is 6.78. Table 1 Sample statistics A. Summary statistics . . Mean . SD . Min . Med . Max . N . Market cap ($ bil) 20.52 33.80 0.29 8.87 203.67 69,018 NYSE/AMEX decile 8.51 1.71 2 9 10 69,018 Stock return (%) 0.99 10.49 −99.68 1.09 259.66 69,018 Bond return (%) 0.42 2.59 −89.74 0.14 29.75 69,018 CDS spread (bps) 184.99 354.35 2.67 88.68 9739.77 69,018 CDS return (%) 0.02 2.49 −55.67 0.02 110.72 69,018 CDS depth 6.78 4.51 2.00 6.00 31.00 69,018 S&P rating 8.93 3.05 1.00 9.00 22.00 68,202 A. Summary statistics . . Mean . SD . Min . Med . Max . N . Market cap ($ bil) 20.52 33.80 0.29 8.87 203.67 69,018 NYSE/AMEX decile 8.51 1.71 2 9 10 69,018 Stock return (%) 0.99 10.49 −99.68 1.09 259.66 69,018 Bond return (%) 0.42 2.59 −89.74 0.14 29.75 69,018 CDS spread (bps) 184.99 354.35 2.67 88.68 9739.77 69,018 CDS return (%) 0.02 2.49 −55.67 0.02 110.72 69,018 CDS depth 6.78 4.51 2.00 6.00 31.00 69,018 S&P rating 8.93 3.05 1.00 9.00 22.00 68,202 Open in new tab Table 1 Sample statistics A. Summary statistics . . Mean . SD . Min . Med . Max . N . Market cap ($ bil) 20.52 33.80 0.29 8.87 203.67 69,018 NYSE/AMEX decile 8.51 1.71 2 9 10 69,018 Stock return (%) 0.99 10.49 −99.68 1.09 259.66 69,018 Bond return (%) 0.42 2.59 −89.74 0.14 29.75 69,018 CDS spread (bps) 184.99 354.35 2.67 88.68 9739.77 69,018 CDS return (%) 0.02 2.49 −55.67 0.02 110.72 69,018 CDS depth 6.78 4.51 2.00 6.00 31.00 69,018 S&P rating 8.93 3.05 1.00 9.00 22.00 68,202 A. Summary statistics . . Mean . SD . Min . Med . Max . N . Market cap ($ bil) 20.52 33.80 0.29 8.87 203.67 69,018 NYSE/AMEX decile 8.51 1.71 2 9 10 69,018 Stock return (%) 0.99 10.49 −99.68 1.09 259.66 69,018 Bond return (%) 0.42 2.59 −89.74 0.14 29.75 69,018 CDS spread (bps) 184.99 354.35 2.67 88.68 9739.77 69,018 CDS return (%) 0.02 2.49 −55.67 0.02 110.72 69,018 CDS depth 6.78 4.51 2.00 6.00 31.00 69,018 S&P rating 8.93 3.05 1.00 9.00 22.00 68,202 Open in new tab B. Correlation matrix . . . (1) . (2) . (3) . (4) . (5) . (6) . (7) . (1) Market cap 1.000 (2) Stock return 0.005 1.000 (3) Bond return −0.005 0.275 1.000 (4) CDS spread −0.187 −0.019 −0.086 1.000 (5) CDS return −0.001 0.409 0.363 −0.108 1.000 (6) CDS depth 0.106 −0.021 −0.023 −0.127 −0.029 1.000 (7) S&P rating −0.553 0.031 0.014 0.510 0.031 −0.171 1.000 B. Correlation matrix . . . (1) . (2) . (3) . (4) . (5) . (6) . (7) . (1) Market cap 1.000 (2) Stock return 0.005 1.000 (3) Bond return −0.005 0.275 1.000 (4) CDS spread −0.187 −0.019 −0.086 1.000 (5) CDS return −0.001 0.409 0.363 −0.108 1.000 (6) CDS depth 0.106 −0.021 −0.023 −0.127 −0.029 1.000 (7) S&P rating −0.553 0.031 0.014 0.510 0.031 −0.171 1.000 Panel A presents summary statistics, and panel B presents a correlation matrix of variables used in this study. Data are monthly from January 2003 to December 2015. Equity data are obtained from CRSP, CDS data come from Markit, and S&P Ratings come from Compustat. Stock, bond, and CDS returns are reported on a monthly basis. S&P rating is numbered sequentially from 1=AAA to 22=D. N refers to the number of firm-month observations. Open in new tab B. Correlation matrix . . . (1) . (2) . (3) . (4) . (5) . (6) . (7) . (1) Market cap 1.000 (2) Stock return 0.005 1.000 (3) Bond return −0.005 0.275 1.000 (4) CDS spread −0.187 −0.019 −0.086 1.000 (5) CDS return −0.001 0.409 0.363 −0.108 1.000 (6) CDS depth 0.106 −0.021 −0.023 −0.127 −0.029 1.000 (7) S&P rating −0.553 0.031 0.014 0.510 0.031 −0.171 1.000 B. Correlation matrix . . . (1) . (2) . (3) . (4) . (5) . (6) . (7) . (1) Market cap 1.000 (2) Stock return 0.005 1.000 (3) Bond return −0.005 0.275 1.000 (4) CDS spread −0.187 −0.019 −0.086 1.000 (5) CDS return −0.001 0.409 0.363 −0.108 1.000 (6) CDS depth 0.106 −0.021 −0.023 −0.127 −0.029 1.000 (7) S&P rating −0.553 0.031 0.014 0.510 0.031 −0.171 1.000 Panel A presents summary statistics, and panel B presents a correlation matrix of variables used in this study. Data are monthly from January 2003 to December 2015. Equity data are obtained from CRSP, CDS data come from Markit, and S&P Ratings come from Compustat. Stock, bond, and CDS returns are reported on a monthly basis. S&P rating is numbered sequentially from 1=AAA to 22=D. N refers to the number of firm-month observations. Open in new tab We acquire our bond data set from TRACE, which includes secondary bond market transactions covering both investment-grade and high-yield bonds. We follow the data cleaning process described by Jostova et al. (2013) and Bessembinder et al. (2009). Our bond return computation follows Jostova et al. (2013) using the clean prices provided by TRACE. The firm-level bond return represent an issue-weighted average of all of the firm’s individual bond issues. We gather equity data from the Center for Research in Security Prices (CRSP), and we require the firm’s equity to ordinary common shares traded on the NYSE, AMEX or Nasdaq. We also require each firm to have a market capitalization of at least $ 100 million and a stock price at the time of portfolio formation of at least $ 1. Delisting returns from CRSP are used in the event that a stock is delisted. The mean monthly stock return of our sample is 0.99%. Firm size is measured through equity market capitalization and is computed as the number of common shares outstanding multiplied by the price of the firm’s stock. The mean market capitalization of our sample is $20.52 billion. S&P credit ratings are available through Compustat. We convert the rating into a numerical score in which 1 represents “AAA,” 2 represents “AA+,” …, and 22 represents “D.” The mean S&P rating of our sample is 8.93, which is roughly a “BBB” rating. For comparison to the larger equity universe, we measure the relative size of our firms by computing decile ranges of equity market capitalization for all stocks in the NYSE/AMEX universe and assigning a corresponding size decile to each firm in our sample (e.g., a size decile value of 10 indicates the firm falls in the highest size decile of firms in the NYSE/AMEX universe). Our sample represents larger firms in the universe of stocks and is fairly normally distributed across credit ratings. Figure 4 portrays the size decile and credit rating distribution of firms in our sample. Panel A is a histogram of the size decile distribution. The mean (median) NYSE/AMEX size decile is 8.51 (9). More than 60% of our firms fall in the two highest size deciles, while less than 10% are in the smallest five deciles, suggesting that CDS contracts generally trade on large firms and that our results are unlikely to be driven by small firms. Panel B of Figure 4 provides a credit rating histogram and shows the rating distribution to be roughly normal. A little more than one-third of the firms are junk grade (i.e., below BBB-). Furthermore, panel C of Figure 4 shows that our sample of firms has reasonable representation across industries. A histogram of the CDS depth distribution is also provided in panel D of the same figure. Internet Appendix A defines the variables. Fig. 4 Open in new tabDownload slide Sample composition: Histograms This figure presents four histograms illustrating the composition of the sample used in this study. Panel A presents a histogram of the percentage of the sample in each decile of all NYSE/AMEX-listed firms. For example, decile 10 refers to the biggest 10% of firms in the NYSE/AMEX universe. Panel B displays a histogram of the percentage of the sample in each S&P credit rating category. Panel C presents an industry histogram. Panel D offers a histogram of CDS contract depth (i.e., number of contributors to the quoted spread). CDS data come from Markit. The time period is January 2003 to December 2015. Fig. 4 Open in new tabDownload slide Sample composition: Histograms This figure presents four histograms illustrating the composition of the sample used in this study. Panel A presents a histogram of the percentage of the sample in each decile of all NYSE/AMEX-listed firms. For example, decile 10 refers to the biggest 10% of firms in the NYSE/AMEX universe. Panel B displays a histogram of the percentage of the sample in each S&P credit rating category. Panel C presents an industry histogram. Panel D offers a histogram of CDS contract depth (i.e., number of contributors to the quoted spread). CDS data come from Markit. The time period is January 2003 to December 2015. 1.1 CDS return computation To compute the CDS holding period (excess) return, we need to compute the profits/losses (P&L) of a CDS over a given holding interval. The P&L of a CDS trading with a unit $ 1-notional is what we term the CDS holding period excess return. This notion of a CDS holding period excess return is consistent with Berndt and Obreja (2010). They view the protection seller’s position in a CDS as a long asset swap position in the risky par-bonds issued by the same reference entity. Hence, the protection seller’s position in a CDS could be viewed as a 100% levered risky par-bond position that is financed at the default-free riskless rate, which serves as the basis for the notion of “excess” return.7 We interchangeably use the terms CDS holding period excess return, CDS holding period return, and CDS return throughout this manuscript. We illustrate the computation steps of the P&L of a CDS as follows: First, we provide a standard CDS pricing model as in O’Kane (2008).8 Then, through this pricing framework, we define the mark-to-market value of a CDS with a unit $1-notional using at-market spread quotes from Markit. The change of these mark-to-market values over a given holding period determines the CDS holding period return. 1.1.1 CDS return: Pricing framework and mark-to-market We split the pricing of a CDS contract with a unit $1-notional into two legs, the premium leg and protection leg. To simplify our illustration, we assume that we are on the inception date of a 5-year CDS. This fresh 5-year contract matures on the first IMM date 5 years after the trade date. For example, a 5-year CDS contract trading on December 20, 2013, matures on March 20, 2018.9 The premium leg has two components. First, there are 21 scheduled premium payments on a quarterly cycle with the CDS IMM dates—the 20th of March, June, September, and December—until the maturity date as long as the reference entity survives. When there is a credit event, there is a payment of the premium that has accrued since the last quarterly premium payment date. This is the second component of the premium leg. Let us denote the quarterly premium payment dates over a 5-year horizon by ti, i=1,2,…,21 , and let t0 denote our valuation date. Given the quoted spread of S0 at time-t0, the present value of the first component of the premium leg becomes S0∑n=1n=21Δ(tn−1,tn)Q(t0,tn)Z(t0,tn),(1) where Δ(tn−1,tn) denotes the accrual factor for the time period, [tn−1,tn] , and Q(t0,tn) and Z(t0,tn) , respectively, denote the survival probability of the reference entity and default-free discounting factor for the time period, [t0,tn] . Now, we consider the premium accrued at default for the nth premium period, [tn−1,tn] . Over an infinitesimal time interval, [s,s+ds] for s∈[tn−1,tn] , the expected present value of the premium accrued upon default is given by S0Δ(tn−1,s)(−dQ(t0,s))Z(t0,s).(2) Then, the value of the premium accrued upon default for all 21 premium periods is given by S0∑n=1n=21∫tn−1tnΔ(tn−1,s)Z(t0,s)(−dQ(t0,s)).(3) By summing Equations (1) and (3), the present value of the premium leg becomes Premium Leg PV=S0·RPV01(t0,t21),(4) where RPV01(t0,t21) is given by RPV01(t0,t21)=∑n=1n=21Δ(tn−1,tn)Q(t0,tn)Z(t0,tn)+∑n=1n=21∫tn−1tnΔ(tn−1,s)Z(t0,s)(−dQ(t0,s)).(5) The integration in the second term in Equation (5) can be approximated as ∫tn−1tnΔ(tn−1,s)Z(t0,s)(−dQ(t0,s))≃12Δ(tn−1,tn)Z(t0,tn)(Q(t0,tn−1)−Q(t0,tn)).(6) Thus, we have RPV01(t0,t21)=∑n=1n=21Δ(tn−1,tn)Z(t0,tn)Q(t0,tn)+∑n=1n=2112Δ(tn−1,tn)Z(t0,tn)(Q(t0,tn−1)−Q(t0,tn)).(7) Assuming a constant loss given default, (1−R) , together with the standard assumption of independence of interest rate and the default time, we can write the present value of the protection leg as Protection Leg PV=(1−R)∫t0t21Z(t0,s)(−dQ(t0,s))≃(1−R)∑n=1n=21Z(t0,tn)(Q(t0,tn−1)−Q(t0,tn)).(8) The second line shows that the integration in the first line is performed by discretizing the 5-year horizon by 21 intervals with each coupon payment date.10 Combining the present values of the premium and the protection legs gives the mark-to-market value of a 5-year short protection position of a CDS with a unit $1-notional as V(t0)=S0·RPV01(t0,t21)−(1−R)∑n=1n=21Z(t0,tn)(Q(t0,tn−1)−Q(t0,tn)),(9) where RPV01(t0,t21) is as in Equation (7).11 With the quoted spread, S0, V(t0)=0 as required.12 However, soon after the inception of trading, this requirement is no longer true since the market spread of the CDS reference entity moves from the spread that the protection seller/buyer are locked into. Finally, with this pricing framework, we can easily define the P&L of a CDS with a unit $1-notional over a holding period, [t0,t′] . For simplicity, we assume for a moment that this interval is short enough so that we can ignore any coupon flows and also potential credit event during this holding period. If we entered as a seller of a protection at time-t0 and unwind the position at time-t’ by buying a protection on the same reference entity and the same maturity date, then the CDS holding period excess return is given as CDS return(t0,t′)=−(S(t′)−S(t0))·RPV01(t′,t21).(10) S(t0)·RPV01(t′,t21) in the above Equation (10) denotes the time-t’ value of the protection we sold at time-t0, and −S(t′)·RPV01(t′,t21) the time-t’ cost to purchase the protection on the same reference entity with the same maturity date. If there is a credit event over our holding period, then the realized return will be equal to −(1−R), where R is a realized recovery rate upon the credit event. Equation (10) does not take into account coupon flows during our holding period and the accrued premium that should be exchanged at each selling and buying transaction of the default protection. We carefully incorporate these factors when we implement this CDS return framework using the quoted spreads from Markit. The U.S. $ Libor curve retrieved from DataStream is calibrated to fit the Nelson-Siegel-Svensson (NSS) curve (Nelson and Siegel 1987; Svensson 1994), and we construct the default-free discounting factors using the fitted values of the NSS curve. After all these considerations, we compute CDS returns based on “clean” P&L’s. 1.1.2 Implementation of the CDS returns using Markit data As illustrated above, we approach CDS returns from the perspective of the protection seller (i.e., a negative return corresponds to an increase in credit risk). We do monthly trading of 5-year contracts. To compute the CDS return, we need to construct the survival probability curve on a given valuation date- t′, Q(t′,ti) , for i=1,2,…,21 . Instead of bootstrapping this survival probability curve using the quoted spreads of CDS contracts across entire maturity groups, we assume a flat hazard rate, h(t′,ti)≡−1Q(t′,ti)∂Q(t′,ti)∂ti=h for ∀i=1,2,…,21 , and calibrate the hazard from the quoted spread of a 5-year CDS contract.13 Our monthly trading of a CDS starts at midnight of the 19th of a given month and ends at midnight of the 19th of the next month, and therefore this trading timeline ensures us to trade only 5-year contracts with a fixed maturity date in and out of each trade.14 With this trading timeline, we do not have to be concerned about the violation of no arbitrage in contract prices across different maturity groups that could exist due in part to our flat hazard rate assumption. We trade only 5-year contracts, and therefore our trading strategy returns are immune from this potential cross-sectional pricing inconsistency in CDS prices. By trading only fresh 5-year contracts, we can further ensure that the liquidity concerns of our trading returns are minimal, which is another potential benefit of our trading timeline. There is no intermediate coupon income for this protection seller during a 1-month holding period. All the coupon values are fully embedded in RPV01(t′,t21) . Credit events possibly occur during our monthly holding period. If a credit event occurs, we need to assign the realized loss given default to the CDS return for that holding period. We find that 40 firms in our sample experience credit events during our sample period. We use the realized recovery rate information that we compiled from the Creditex Group and make appropriate adjustments in our holding period excess returns.15 In untabulated robustness checks, we further confirm the robustness of our results to these credit event concerns by exclusively focusing on the firms that survive throughout our whole sample period. These results are available on request from the authors. As explained in Section 1.1.1, an accrued premium should be adjusted when the CDS trade occurs in between the quarterly coupon payment interval. The accrued premium payment is handled two different ways during our sample period. Before the 2009 “Big Bang” protocol, the first premium accrued since the trade date was paid either on the next immediate coupon date (i.e., short-stub) or the following coupon date (i.e., long-stub), depending on the trade date. If the trade date fell within a 30-day window prior to the first upcoming coupon date, it would follow the long-stub rule and the short-stub otherwise. However, post-“Big Bang,” these complicated accrued premium payment rules disappeared, and now the single-name CDS contracts trade just like the Markit CDS indices where the new protection seller will receive the full quarterly coupon on each coupon payment date. Any “over”paid premium to this seller by the protection buyer is rebated up-front.16 We follow this post-“Big Bang” coupon convention when we adjust the accrued premium to get the clean P&L of our CDS trading. We compute the P&L’s on the running spread basis, instead of using fixed 100 bps/500 bps coupons with up-front adjustments. Since the default-free rate during our sample period is relatively low, the potential errors in our treatment of the stub algorithms should be minimal for CDS returns in the pre-“Big Bang” period.17 Table 1 provides summary statistics on CDS returns for the firms in our sample over the years 2003–2015. The average CDS return over our time period is 0.02% with a standard deviation of 2.49%. Figure 5 displays the equally weighted average CDS spread (panel A) as well as the time series of equally weighted CDS and stock returns (panel B). Panel B shows that CDS returns are highly correlated with stock returns; however, the CDS returns could potentially lead the stock returns, particularly during the recent financial crisis period. This pattern would suggest an informational advantage embedded in CDS returns under certain conditions. Fig. 5 Open in new tabDownload slide Time series of CDS spreads, CDS returns, and stock returns This figure displays the time series of CDS spreads (panel A) and CDS returns and stock returns (panel B). All series are monthly and equally weight the constituents. The CDS return series is scaled by a constant σstock/σcds such that both stock and CDS return series have the same standard deviation. The sample consists of 881 firms and spans the time period January 2003 to December 2015. CDS data come from Markit, and equity data come from CRSP. Fig. 5 Open in new tabDownload slide Time series of CDS spreads, CDS returns, and stock returns This figure displays the time series of CDS spreads (panel A) and CDS returns and stock returns (panel B). All series are monthly and equally weight the constituents. The CDS return series is scaled by a constant σstock/σcds such that both stock and CDS return series have the same standard deviation. The sample consists of 881 firms and spans the time period January 2003 to December 2015. CDS data come from Markit, and equity data come from CRSP. 2. CDS Momentum Our CDS momentum trading strategy is constructed in the spirit of Jegadeesh and Titman (1993). To create CDS momentum portfolios, firms are sorted at the end of each month into five equally sized groups, P1 to P5, based on their CDS performance over the past J months (referred to as the formation period). We write default protection on firms in P5 that have the highest returns (winners) and acquire default protection on firms in P1 that have the lowest returns (losers). All positions are unwound and rebalanced after K months (referred to as the holding period) and equally weighted returns are computed for each performance quintile. Similar to previous studies (Jegadeesh and Titman 1993, among others), holding period returns greater than 1 month are constructed in a manner that avoids overlapping returns.18 Table 2 summarizes momentum profits in the CDS market over 2003–2015. Panel A reports the raw returns of the momentum portfolios. We first document the profits of the CDS momentum strategy using a formation and holding period of 1 month ( J=1m,K=1m ). The long-short return using this formation period is 0.48% per month with a t-statistic of 3.61, which implies short-term momentum rather than reversal. Therefore, unlike Jegadeesh and Titman (1993), who adjust the formation period to avoid short-term reversals in the stock market, we rely on a CDS momentum strategy that can be implemented without a 1-month gap between the formation and holding periods. For these reasons, we use contiguous formation and holding periods for the remaining CDS momentum strategies implemented in this paper. Table 2 CDS momentum A. CDS momentum returns (% per month) . . CDS performance quintiles . Long-short strategy P5–P1 . . P1 . P2 . P3 . P4 . P5 . K=1m . K=3m . K=6m . J=1m −0.16 −0.10 −0.01 0.07 0.33 0.48*** 0.35*** 0.22*** (−0.59) (−0.91) (−0.14) (0.94) (1.41) (3.61) (5.10) (3.58) 0.96 1.03 0.81 J=3m −0.25 −0.07 −0.00 0.08 0.34 0.59*** 0.48*** 0.25*** (−0.99) (−0.76) (−0.04) (0.85) (1.43) (4.61) (3.70) (2.79) 0.99 1.00 0.64 J=6m −0.14 −0.04 0.00 0.04 0.26 0.41*** 0.31** 0.20 (−0.59) (−0.48) (0.02) (0.39) (1.06) (3.47) (2.23) (1.41) 0.67 0.57 0.39 J=12m −0.09 −0.06 −0.02 0.04 0.25 0.33* 0.30* 0.19 (−0.30) (−0.53) (−0.32) (0.48) (1.31) (1.87) (1.72) (1.02) 0.49 0.47 0.30 Precrisis period (Jan. 2003–Dec. 2007) J=3m −0.03 0.03 0.02 0.12 0.36 0.40** 0.31 0.16 (−0.18) (0.61) (0.32) (0.98) (1.08) (2.22) (1.34) (1.35) 1.08 0.93 0.68 Crisis period (Jan. 2008–Dec. 2011) J=3m −0.69 −0.30 −0.07 0.01 0.22 0.91*** 0.77*** 0.25 (-1.00) (-1.24) (−0.37) (0.04) (0.37) (2.99) (3.30) (1.27) 0.95 1.02 0.41 Postcrisis period (Jan. 2012–Dec. 2015) J=3m −0.07 0.04 0.04 0.11 0.45** 0.51*** 0.42*** 0.35*** (−0.31) (1.33) (0.75) (1.50) (2.06) (5.50) (2.95) (2.77) 1.92 1.76 1.63 Investment-grade firms (BBB- and above) J=3m −0.06 −0.07 −0.01 0.02 0.09 0.15** 0.13** 0.03 (−0.43) (-1.05) (−0.12) (0.38) (0.69) (2.53) (2.13) (0.52) 0.46 0.48 0.13 Junk-grade firms (BB+ and below) J=3m −0.44 −0.05 0.10 0.21 0.66* 1.10*** 1.00*** 0.58*** (-1.07) (−0.32) (0.60) (1.06) (1.84) (4.65) (3.94) (3.73) 1.17 1.20 0.85 A. CDS momentum returns (% per month) . . CDS performance quintiles . Long-short strategy P5–P1 . . P1 . P2 . P3 . P4 . P5 . K=1m . K=3m . K=6m . J=1m −0.16 −0.10 −0.01 0.07 0.33 0.48*** 0.35*** 0.22*** (−0.59) (−0.91) (−0.14) (0.94) (1.41) (3.61) (5.10) (3.58) 0.96 1.03 0.81 J=3m −0.25 −0.07 −0.00 0.08 0.34 0.59*** 0.48*** 0.25*** (−0.99) (−0.76) (−0.04) (0.85) (1.43) (4.61) (3.70) (2.79) 0.99 1.00 0.64 J=6m −0.14 −0.04 0.00 0.04 0.26 0.41*** 0.31** 0.20 (−0.59) (−0.48) (0.02) (0.39) (1.06) (3.47) (2.23) (1.41) 0.67 0.57 0.39 J=12m −0.09 −0.06 −0.02 0.04 0.25 0.33* 0.30* 0.19 (−0.30) (−0.53) (−0.32) (0.48) (1.31) (1.87) (1.72) (1.02) 0.49 0.47 0.30 Precrisis period (Jan. 2003–Dec. 2007) J=3m −0.03 0.03 0.02 0.12 0.36 0.40** 0.31 0.16 (−0.18) (0.61) (0.32) (0.98) (1.08) (2.22) (1.34) (1.35) 1.08 0.93 0.68 Crisis period (Jan. 2008–Dec. 2011) J=3m −0.69 −0.30 −0.07 0.01 0.22 0.91*** 0.77*** 0.25 (-1.00) (-1.24) (−0.37) (0.04) (0.37) (2.99) (3.30) (1.27) 0.95 1.02 0.41 Postcrisis period (Jan. 2012–Dec. 2015) J=3m −0.07 0.04 0.04 0.11 0.45** 0.51*** 0.42*** 0.35*** (−0.31) (1.33) (0.75) (1.50) (2.06) (5.50) (2.95) (2.77) 1.92 1.76 1.63 Investment-grade firms (BBB- and above) J=3m −0.06 −0.07 −0.01 0.02 0.09 0.15** 0.13** 0.03 (−0.43) (-1.05) (−0.12) (0.38) (0.69) (2.53) (2.13) (0.52) 0.46 0.48 0.13 Junk-grade firms (BB+ and below) J=3m −0.44 −0.05 0.10 0.21 0.66* 1.10*** 1.00*** 0.58*** (-1.07) (−0.32) (0.60) (1.06) (1.84) (4.65) (3.94) (3.73) 1.17 1.20 0.85 Open in new tab Table 2 CDS momentum A. CDS momentum returns (% per month) . . CDS performance quintiles . Long-short strategy P5–P1 . . P1 . P2 . P3 . P4 . P5 . K=1m . K=3m . K=6m . J=1m −0.16 −0.10 −0.01 0.07 0.33 0.48*** 0.35*** 0.22*** (−0.59) (−0.91) (−0.14) (0.94) (1.41) (3.61) (5.10) (3.58) 0.96 1.03 0.81 J=3m −0.25 −0.07 −0.00 0.08 0.34 0.59*** 0.48*** 0.25*** (−0.99) (−0.76) (−0.04) (0.85) (1.43) (4.61) (3.70) (2.79) 0.99 1.00 0.64 J=6m −0.14 −0.04 0.00 0.04 0.26 0.41*** 0.31** 0.20 (−0.59) (−0.48) (0.02) (0.39) (1.06) (3.47) (2.23) (1.41) 0.67 0.57 0.39 J=12m −0.09 −0.06 −0.02 0.04 0.25 0.33* 0.30* 0.19 (−0.30) (−0.53) (−0.32) (0.48) (1.31) (1.87) (1.72) (1.02) 0.49 0.47 0.30 Precrisis period (Jan. 2003–Dec. 2007) J=3m −0.03 0.03 0.02 0.12 0.36 0.40** 0.31 0.16 (−0.18) (0.61) (0.32) (0.98) (1.08) (2.22) (1.34) (1.35) 1.08 0.93 0.68 Crisis period (Jan. 2008–Dec. 2011) J=3m −0.69 −0.30 −0.07 0.01 0.22 0.91*** 0.77*** 0.25 (-1.00) (-1.24) (−0.37) (0.04) (0.37) (2.99) (3.30) (1.27) 0.95 1.02 0.41 Postcrisis period (Jan. 2012–Dec. 2015) J=3m −0.07 0.04 0.04 0.11 0.45** 0.51*** 0.42*** 0.35*** (−0.31) (1.33) (0.75) (1.50) (2.06) (5.50) (2.95) (2.77) 1.92 1.76 1.63 Investment-grade firms (BBB- and above) J=3m −0.06 −0.07 −0.01 0.02 0.09 0.15** 0.13** 0.03 (−0.43) (-1.05) (−0.12) (0.38) (0.69) (2.53) (2.13) (0.52) 0.46 0.48 0.13 Junk-grade firms (BB+ and below) J=3m −0.44 −0.05 0.10 0.21 0.66* 1.10*** 1.00*** 0.58*** (-1.07) (−0.32) (0.60) (1.06) (1.84) (4.65) (3.94) (3.73) 1.17 1.20 0.85 A. CDS momentum returns (% per month) . . CDS performance quintiles . Long-short strategy P5–P1 . . P1 . P2 . P3 . P4 . P5 . K=1m . K=3m . K=6m . J=1m −0.16 −0.10 −0.01 0.07 0.33 0.48*** 0.35*** 0.22*** (−0.59) (−0.91) (−0.14) (0.94) (1.41) (3.61) (5.10) (3.58) 0.96 1.03 0.81 J=3m −0.25 −0.07 −0.00 0.08 0.34 0.59*** 0.48*** 0.25*** (−0.99) (−0.76) (−0.04) (0.85) (1.43) (4.61) (3.70) (2.79) 0.99 1.00 0.64 J=6m −0.14 −0.04 0.00 0.04 0.26 0.41*** 0.31** 0.20 (−0.59) (−0.48) (0.02) (0.39) (1.06) (3.47) (2.23) (1.41) 0.67 0.57 0.39 J=12m −0.09 −0.06 −0.02 0.04 0.25 0.33* 0.30* 0.19 (−0.30) (−0.53) (−0.32) (0.48) (1.31) (1.87) (1.72) (1.02) 0.49 0.47 0.30 Precrisis period (Jan. 2003–Dec. 2007) J=3m −0.03 0.03 0.02 0.12 0.36 0.40** 0.31 0.16 (−0.18) (0.61) (0.32) (0.98) (1.08) (2.22) (1.34) (1.35) 1.08 0.93 0.68 Crisis period (Jan. 2008–Dec. 2011) J=3m −0.69 −0.30 −0.07 0.01 0.22 0.91*** 0.77*** 0.25 (-1.00) (-1.24) (−0.37) (0.04) (0.37) (2.99) (3.30) (1.27) 0.95 1.02 0.41 Postcrisis period (Jan. 2012–Dec. 2015) J=3m −0.07 0.04 0.04 0.11 0.45** 0.51*** 0.42*** 0.35*** (−0.31) (1.33) (0.75) (1.50) (2.06) (5.50) (2.95) (2.77) 1.92 1.76 1.63 Investment-grade firms (BBB- and above) J=3m −0.06 −0.07 −0.01 0.02 0.09 0.15** 0.13** 0.03 (−0.43) (-1.05) (−0.12) (0.38) (0.69) (2.53) (2.13) (0.52) 0.46 0.48 0.13 Junk-grade firms (BB+ and below) J=3m −0.44 −0.05 0.10 0.21 0.66* 1.10*** 1.00*** 0.58*** (-1.07) (−0.32) (0.60) (1.06) (1.84) (4.65) (3.94) (3.73) 1.17 1.20 0.85 Open in new tab B. Risk-adjusted performance . . Long-short CDS momentum returns ( J=3m, K=1m ) . . (1) . (2) . (3) . (4) . (5) . α 0.67*** 0.73*** 0.70*** 0.74*** 0.73*** (4.14) (4.50) (4.26) (4.23) (4.17) βDEF −0.45** −0.30** −0.27 (-2.44) (-2.15) (-1.49) βTERM −0.10 −0.07 −0.06 (-0.91) (-0.77) (-0.69) βMKT −0.15** −0.13** −0.10* −0.09* (-2.19) (-1.98) (-1.86) (-1.80) βSMB −0.09 −0.09 −0.08 −0.09 (-1.08) (-1.21) (-1.03) (-1.11) βHML 0.15** 0.19** 0.14** 0.16* (2.32) (2.22) (2.04) (1.75) βUMD 0.06 0.03 (1.05) (0.44) N 156 156 156 156 156 Adj. R2 .079 .091 .096 .112 .108 B. Risk-adjusted performance . . Long-short CDS momentum returns ( J=3m, K=1m ) . . (1) . (2) . (3) . (4) . (5) . α 0.67*** 0.73*** 0.70*** 0.74*** 0.73*** (4.14) (4.50) (4.26) (4.23) (4.17) βDEF −0.45** −0.30** −0.27 (-2.44) (-2.15) (-1.49) βTERM −0.10 −0.07 −0.06 (-0.91) (-0.77) (-0.69) βMKT −0.15** −0.13** −0.10* −0.09* (-2.19) (-1.98) (-1.86) (-1.80) βSMB −0.09 −0.09 −0.08 −0.09 (-1.08) (-1.21) (-1.03) (-1.11) βHML 0.15** 0.19** 0.14** 0.16* (2.32) (2.22) (2.04) (1.75) βUMD 0.06 0.03 (1.05) (0.44) N 156 156 156 156 156 Adj. R2 .079 .091 .096 .112 .108 Open in new tab B. Risk-adjusted performance . . Long-short CDS momentum returns ( J=3m, K=1m ) . . (1) . (2) . (3) . (4) . (5) . α 0.67*** 0.73*** 0.70*** 0.74*** 0.73*** (4.14) (4.50) (4.26) (4.23) (4.17) βDEF −0.45** −0.30** −0.27 (-2.44) (-2.15) (-1.49) βTERM −0.10 −0.07 −0.06 (-0.91) (-0.77) (-0.69) βMKT −0.15** −0.13** −0.10* −0.09* (-2.19) (-1.98) (-1.86) (-1.80) βSMB −0.09 −0.09 −0.08 −0.09 (-1.08) (-1.21) (-1.03) (-1.11) βHML 0.15** 0.19** 0.14** 0.16* (2.32) (2.22) (2.04) (1.75) βUMD 0.06 0.03 (1.05) (0.44) N 156 156 156 156 156 Adj. R2 .079 .091 .096 .112 .108 B. Risk-adjusted performance . . Long-short CDS momentum returns ( J=3m, K=1m ) . . (1) . (2) . (3) . (4) . (5) . α 0.67*** 0.73*** 0.70*** 0.74*** 0.73*** (4.14) (4.50) (4.26) (4.23) (4.17) βDEF −0.45** −0.30** −0.27 (-2.44) (-2.15) (-1.49) βTERM −0.10 −0.07 −0.06 (-0.91) (-0.77) (-0.69) βMKT −0.15** −0.13** −0.10* −0.09* (-2.19) (-1.98) (-1.86) (-1.80) βSMB −0.09 −0.09 −0.08 −0.09 (-1.08) (-1.21) (-1.03) (-1.11) βHML 0.15** 0.19** 0.14** 0.16* (2.32) (2.22) (2.04) (1.75) βUMD 0.06 0.03 (1.05) (0.44) N 156 156 156 156 156 Adj. R2 .079 .091 .096 .112 .108 Open in new tab C. Robustness to alternative momentum strategies . . Long-short CDS momentum returns ( J=3m, K=1m ) . . (1) . (2) . (3) . (4) . (5) . (6) . (7) . (8) . α0 0.59*** 0.55*** 0.52*** 0.54*** 0.46*** 0.44*** 0.42*** 0.47*** (4.17) (4.29) (4.39) (4.08) (4.43) (3.82) (3.69) (4.66) αIC 0.53 0.42 0.35 0.51 (1.33) (1.20) (1.08) (1.21) βUMD 0.06 −0.03 (1.20) (-0.69) βIMOM3 0.20*** 0.05 (3.63) (1.31) βFMOM3 0.34*** 0.29*** (3.21) (3.24) βPEAD 0.22 −0.04 (1.59) (-0.64) βUMD×IC 0.14** (2.03) βIMOM3×IC 0.21*** (3.10) βFMOM3×IC 0.07 (0.32) βPEAD×IC 0.51*** (2.63) N 156 156 156 156 156 156 156 156 Adj. R2 .013 .105 .054 .037 .036 .131 .049 .097 C. Robustness to alternative momentum strategies . . Long-short CDS momentum returns ( J=3m, K=1m ) . . (1) . (2) . (3) . (4) . (5) . (6) . (7) . (8) . α0 0.59*** 0.55*** 0.52*** 0.54*** 0.46*** 0.44*** 0.42*** 0.47*** (4.17) (4.29) (4.39) (4.08) (4.43) (3.82) (3.69) (4.66) αIC 0.53 0.42 0.35 0.51 (1.33) (1.20) (1.08) (1.21) βUMD 0.06 −0.03 (1.20) (-0.69) βIMOM3 0.20*** 0.05 (3.63) (1.31) βFMOM3 0.34*** 0.29*** (3.21) (3.24) βPEAD 0.22 −0.04 (1.59) (-0.64) βUMD×IC 0.14** (2.03) βIMOM3×IC 0.21*** (3.10) βFMOM3×IC 0.07 (0.32) βPEAD×IC 0.51*** (2.63) N 156 156 156 156 156 156 156 156 Adj. R2 .013 .105 .054 .037 .036 .131 .049 .097 This table presents our CDS momentum results. Firms are sorted on past CDS return into five equally sized portfolios in which portfolio P1 (P5) is the group with the lowest (highest) past CDS return using a formation period of J months. The long-short strategy is constructed by purchasing CDS contracts of firms in portfolio P5 (High) and selling CDS contracts in portfolio P1 (Low) and holding for period K months. Momentum portfolios equally weight the constituents. The annualized Sharpe ratio (presented in italics) is equal to 12 multiplied by the mean monthly return divided by the standard deviation of monthly returns. The 3- and 6-month holding periods are reported as monthly returns and are computed so as to avoid overlapping returns. Panel A reports returns with notation for each portfolio return following the pattern: mean, (t-statistic), and annualized Sharpe ratio. Panel B presents the risk-adjusted performance of CDS momentum using formation period of 3 months and holding period of 1 month ( MOMJ=3mcds ). Bond market factors include default (DEF) and term (TERM) factors (acquired from Ibbotson, Citigroup, and CRSP). Stock market factors include the market excess return (MKT), size (SMB), value (HML), and momentum (UMD) factors (all from Kenneth French’s website). Panel C presents regressions explaining CDS momentum with various alternative momentum strategies, including J=3m industry momentum (IMOM3), J=3m factor momentum, and post-earnings-announcement drift (PEAD) from Daniel, Hirshleifer, and Sun (2020). Newey-West t-statistics (12 lags) are provided in parentheses. * p <.1; ** p <.05; *** p <.01. Open in new tab C. Robustness to alternative momentum strategies . . Long-short CDS momentum returns ( J=3m, K=1m ) . . (1) . (2) . (3) . (4) . (5) . (6) . (7) . (8) . α0 0.59*** 0.55*** 0.52*** 0.54*** 0.46*** 0.44*** 0.42*** 0.47*** (4.17) (4.29) (4.39) (4.08) (4.43) (3.82) (3.69) (4.66) αIC 0.53 0.42 0.35 0.51 (1.33) (1.20) (1.08) (1.21) βUMD 0.06 −0.03 (1.20) (-0.69) βIMOM3 0.20*** 0.05 (3.63) (1.31) βFMOM3 0.34*** 0.29*** (3.21) (3.24) βPEAD 0.22 −0.04 (1.59) (-0.64) βUMD×IC 0.14** (2.03) βIMOM3×IC 0.21*** (3.10) βFMOM3×IC 0.07 (0.32) βPEAD×IC 0.51*** (2.63) N 156 156 156 156 156 156 156 156 Adj. R2 .013 .105 .054 .037 .036 .131 .049 .097 C. Robustness to alternative momentum strategies . . Long-short CDS momentum returns ( J=3m, K=1m ) . . (1) . (2) . (3) . (4) . (5) . (6) . (7) . (8) . α0 0.59*** 0.55*** 0.52*** 0.54*** 0.46*** 0.44*** 0.42*** 0.47*** (4.17) (4.29) (4.39) (4.08) (4.43) (3.82) (3.69) (4.66) αIC 0.53 0.42 0.35 0.51 (1.33) (1.20) (1.08) (1.21) βUMD 0.06 −0.03 (1.20) (-0.69) βIMOM3 0.20*** 0.05 (3.63) (1.31) βFMOM3 0.34*** 0.29*** (3.21) (3.24) βPEAD 0.22 −0.04 (1.59) (-0.64) βUMD×IC 0.14** (2.03) βIMOM3×IC 0.21*** (3.10) βFMOM3×IC 0.07 (0.32) βPEAD×IC 0.51*** (2.63) N 156 156 156 156 156 156 156 156 Adj. R2 .013 .105 .054 .037 .036 .131 .049 .097 This table presents our CDS momentum results. Firms are sorted on past CDS return into five equally sized portfolios in which portfolio P1 (P5) is the group with the lowest (highest) past CDS return using a formation period of J months. The long-short strategy is constructed by purchasing CDS contracts of firms in portfolio P5 (High) and selling CDS contracts in portfolio P1 (Low) and holding for period K months. Momentum portfolios equally weight the constituents. The annualized Sharpe ratio (presented in italics) is equal to 12 multiplied by the mean monthly return divided by the standard deviation of monthly returns. The 3- and 6-month holding periods are reported as monthly returns and are computed so as to avoid overlapping returns. Panel A reports returns with notation for each portfolio return following the pattern: mean, (t-statistic), and annualized Sharpe ratio. Panel B presents the risk-adjusted performance of CDS momentum using formation period of 3 months and holding period of 1 month ( MOMJ=3mcds ). Bond market factors include default (DEF) and term (TERM) factors (acquired from Ibbotson, Citigroup, and CRSP). Stock market factors include the market excess return (MKT), size (SMB), value (HML), and momentum (UMD) factors (all from Kenneth French’s website). Panel C presents regressions explaining CDS momentum with various alternative momentum strategies, including J=3m industry momentum (IMOM3), J=3m factor momentum, and post-earnings-announcement drift (PEAD) from Daniel, Hirshleifer, and Sun (2020). Newey-West t-statistics (12 lags) are provided in parentheses. * p <.1; ** p <.05; *** p <.01. Open in new tab Panel A of Table 2 also presents CDS momentum returns using various longer formation periods: 3 ( J=3m ), 6 ( J=6m ), and 12 months ( J=12m ). The 3-month formation period provides the largest momentum profits, producing a 1-month return of 0.59% (or 7.08% on an annual basis) with a t-statistic of 4.61 and an annualized Sharpe ratio of 0.99.19 The loser portfolio P1 produces an equally weighted monthly return of -0.25%, and the winner portfolio P5 produces a return of 0.34%. A $100 investment in this strategy and rolling 24-month average returns are depicted in Figure 3. Monthly returns using longer formation periods of 6 ( J=6m ) and 12 months ( J=12m ) are 0.41% (t-statistic of 3.47) and 0.33% (t-statistic of 1.87), respectively. CDS momentum profits are also significant for holding periods K longer than 1 month. For example, the strategy using a 3-month formation period ( J=3m ) generates future 3- ( K=3m ) and 6-month ( K=6m ) average monthly returns of 0.48% and 0.25%, respectively, both of which are statistically significant at the 1% level. The magnitude of the average monthly return of the long-short strategy is reduced as the holding period K is extended. Panel A of Table 2 also presents the performance of the 3-month momentum strategy ( J=3m ) over different time periods. The precrisis period spans January 2003 to December 2007, the crisis period spans January 2008 to December 2011, and the postcrisis period spans January 2012 to December 2015. One-month holding period ( K=1m ) long-short strategy returns are positive and statistically significant at at least the 5% level in all three subperiods. The magnitude is greatest during the crisis period with average returns of 0.91% bps (t-statistic of 2.99) but also significant during the precrisis and postcrisis periods with returns of 0.40% (t-statistic of 2.22) and 0.51% (t-statistic of 5.50), respectively. We find similar patterns in the magnitude of CDS momentum returns for holding periods longer than 1 month.20 Panel A of Table 2 also shows CDS momentum strategy returns separately for investment-grade (BBB- and above) and junk-grade entities (BB+ and below). Consistent with both stock and bond momentum evidence (Avramov et al. 2007; Jostova et al. 2013), the CDS momentum long-short strategy is substantially higher for junk-grade firms than investment-grade firms. Junk-grade firms produce a return of 1.10% with a t-statistic of 4.65 and an annualized Sharpe ratio of 1.17. Although smaller in magnitude, the investment-grade firms still produce a statistically significant momentum return of 0.15% with a t-statistic of 2.53 and a Sharpe ratio of 0.46. This result contrasts with earlier research findings that corporate bond momentum does not exist among investment-grade entities in the post-1991 time period as documented by Jostova et al. (2013). Next, in panel B of Table 2, we test the exposure of CDS momentum to common stock and bond risk factors. The momentum strategy analyzed in this subsection uses formation period J=3m and holding period K=1m , which was shown to yield the greatest momentum profits as illustrated in panel A of Table 2. Using ordinary least squares (OLS) with Newey-West standard errors and a lag length of 12 months, we estimate rPt=αP+β′PFt+ePt,(11) where rPt is the CDS return for momentum portfolio P, αP is the portfolio’s abnormal return, and Ft is a vector of common risk factors from bond and stock markets.21 The first bond-based factor, DEF, captures the difference in returns between BBB-rated and AAA-rated bonds. The second bond-based factor, TERM, is the difference in returns between long-term and short-term bonds. The equity factors—MKT, SMB, HML, and UMD—are the monthly return series from Kenneth French’s website.22 In Internet Appendix G, we also show robustness to a variety of factors, including a “localized” stock momentum factor using our sample of firms and the strategy J=12m and K=1m , skipping a month between formation and holding periods, which is standard in the stock momentum literature (see Asness, Moskowitz, and Pedersen 2013 (2013), among others). The monthly long-short strategy alpha coefficients shown in panel B range from 0.67% to 0.74% and are similar in magnitude to the raw CDS momentum returns of 0.59% per month as previously reported in panel A. Across the five different factor models, the alpha coefficients prove to all be statistically significant at the 1% level.23 CDS momentum exhibits a generally negative relation to both default risk and stock market factors, suggesting that CDS losers may be more sensitive to fluctuating market conditions than CDS winners. In panel C of Table 2, we further investigate the relation of CDS momentum to alternative constructions of stock-based momentum, including factors capturing industry momentum (Moskowitz and Grinblatt 1999), factor momentum (Ehsani and Linnainmaa 2020), and post-earnings announcement drift (Daniel, Hirshleifer, and Sun 2020). Internet Appendix B defines the variables in detail. In columns 1 to 4, we perform univariate regressions explaining stock momentum with each alternative momentum strategy. Of these, CDS momentum exhibits the strongest overall relation to industry momentum (IMOM3) with an adjusted R2 of 0.105. The CDS momentum alpha coefficient is the lowest against factor momentum (FMOM3), which parallels the finding of Ehsani and Linnainmaa (2020) that momentum in common stock market factors has significant explanatory power over traditional stock momentum. We further interact these alternative momentum factors with our crisis period dummy (January 2008 to December 2011). For UMD, IMOM3, and PEAD, the relation to CDS momentum is entirely concentrated in the crisis period. FMOM3 displays a relatively stable relationship to CDS momentum over time. 2.1 CDS momentum and firm characteristics In this subsection, we investigate the relation between CDS momentum and firm characteristics. Our objective is to both test the robustness of our CDS momentum result and explore potential mechanisms underlying it. We first examine variation in average firm characteristics across the momentum portfolios but do not find evidence that it explains the positive long-short returns. Then, we compare CDS momentum performance across subsamples and find that the CDS momentum strategy is stronger among small firms, those with poor credit ratings, and CDS contracts with greater liquidity. The momentum strategy analyzed in this subsection uses formation J=3m and holding K=1m , which was shown to yield the greatest momentum profits in our baseline results in panel A of Table 2. First, in panel A of Table 3, we examine whether the difference in average firm size (equity market capitalization), credit rating, or CDS liquidity (as measured by CDS contract depth) across the CDS momentum portfolios explains the positive returns of our long-short momentum strategy. For each portfolio, we compute the average characteristic value and the characteristic-adjusted return. We find that, relative to the CDS loser portfolio, the CDS winner portfolio consistently contains smaller, riskier firms with slightly lower CDS depth. To examine whether the potential risks associated with these characteristics drive the CDS momentum profits, we report characteristic-adjusted returns in Table 3. The characteristic adjustment is made by removing the mean return at each decile or group from the actual return (see Internet Appendix C). Even after adjusting for these characteristics, the magnitude of CDS momentum remains large and all long-short returns are statistically significant at the 1% level. The size-adjusted, rating-adjusted, and depth-adjusted returns produce monthly long-short returns of 0.64%, 0.57%, and 0.63%, respectively. Adjusting for all three characteristics produces long-short returns of 0.60% per month. Table 3 CDS momentum and firm characteristics A. Average firm characteristics of CDS momentum portfolios . . . CDS performance quintiles . . . . P1 . P2 . P3 . P4 . P5 . P5–P1 . (t-statistic) . . Size decile (1-10) Average size decile 5.18 6.52 6.33 5.56 4.05 −1.12*** (-4.87) Size-adjusted return −0.27 −0.08 −0.04 0.03 0.37 0.64*** (4.49) S&P credit rating (1=AAA, 2=AA+, etc.) Average rating 9.16 7.56 7.81 8.75 11.20 2.04*** (6.23) Rating-adjusted return −0.28 −0.06 0.00 0.04 0.30 0.57*** (4.49) CDS depth Average depth 6.95 6.94 7.02 6.94 6.16 −0.79*** (-3.56) Depth-adjusted return −0.29 −0.10 −0.03 0.08 0.34 0.63*** (4.71) Size, S&P credit rating, and CDS depth Adjusted return −0.27 −0.08 −0.01 0.02 0.33 0.60*** (4.57) A. Average firm characteristics of CDS momentum portfolios . . . CDS performance quintiles . . . . P1 . P2 . P3 . P4 . P5 . P5–P1 . (t-statistic) . . Size decile (1-10) Average size decile 5.18 6.52 6.33 5.56 4.05 −1.12*** (-4.87) Size-adjusted return −0.27 −0.08 −0.04 0.03 0.37 0.64*** (4.49) S&P credit rating (1=AAA, 2=AA+, etc.) Average rating 9.16 7.56 7.81 8.75 11.20 2.04*** (6.23) Rating-adjusted return −0.28 −0.06 0.00 0.04 0.30 0.57*** (4.49) CDS depth Average depth 6.95 6.94 7.02 6.94 6.16 −0.79*** (-3.56) Depth-adjusted return −0.29 −0.10 −0.03 0.08 0.34 0.63*** (4.71) Size, S&P credit rating, and CDS depth Adjusted return −0.27 −0.08 −0.01 0.02 0.33 0.60*** (4.57) Open in new tab Table 3 CDS momentum and firm characteristics A. Average firm characteristics of CDS momentum portfolios . . . CDS performance quintiles . . . . P1 . P2 . P3 . P4 . P5 . P5–P1 . (t-statistic) . . Size decile (1-10) Average size decile 5.18 6.52 6.33 5.56 4.05 −1.12*** (-4.87) Size-adjusted return −0.27 −0.08 −0.04 0.03 0.37 0.64*** (4.49) S&P credit rating (1=AAA, 2=AA+, etc.) Average rating 9.16 7.56 7.81 8.75 11.20 2.04*** (6.23) Rating-adjusted return −0.28 −0.06 0.00 0.04 0.30 0.57*** (4.49) CDS depth Average depth 6.95 6.94 7.02 6.94 6.16 −0.79*** (-3.56) Depth-adjusted return −0.29 −0.10 −0.03 0.08 0.34 0.63*** (4.71) Size, S&P credit rating, and CDS depth Adjusted return −0.27 −0.08 −0.01 0.02 0.33 0.60*** (4.57) A. Average firm characteristics of CDS momentum portfolios . . . CDS performance quintiles . . . . P1 . P2 . P3 . P4 . P5 . P5–P1 . (t-statistic) . . Size decile (1-10) Average size decile 5.18 6.52 6.33 5.56 4.05 −1.12*** (-4.87) Size-adjusted return −0.27 −0.08 −0.04 0.03 0.37 0.64*** (4.49) S&P credit rating (1=AAA, 2=AA+, etc.) Average rating 9.16 7.56 7.81 8.75 11.20 2.04*** (6.23) Rating-adjusted return −0.28 −0.06 0.00 0.04 0.30 0.57*** (4.49) CDS depth Average depth 6.95 6.94 7.02 6.94 6.16 −0.79*** (-3.56) Depth-adjusted return −0.29 −0.10 −0.03 0.08 0.34 0.63*** (4.71) Size, S&P credit rating, and CDS depth Adjusted return −0.27 −0.08 −0.01 0.02 0.33 0.60*** (4.57) Open in new tab B. CDS momentum returns by size, rating, and CDS depth . . . Characteristic-based terciles (T) . . . T1 . T2 . T3 . T3–T1 . . Small Large Size 1.06*** 0.35*** 0.21*** −0.85*** (5.19) (2.72) (2.77) (-5.10) 1.03 0.72 0.73 −0.92 Size (orth.) 0.77*** 0.43*** 0.50*** −0.27 (3.93) (4.02) (3.17) (-1.62) 0.82 0.78 0.97 −0.35 Safe Risky S&P rating 0.07 0.25*** 1.34*** 1.27*** (1.28) (2.84) (4.62) (4.63) 0.23 0.61 1.05 1.11 S&P rating (orth.) 0.18* 0.60*** 0.83*** 0.65*** (1.73) (3.29) (4.83) (3.67) 0.29 0.91 1.08 0.92 Low High CDS depth −0.06** 0.16 1.05*** 1.11*** (-2.13) (1.08) (5.91) (6.07) −0.57 0.39 1.09 1.16 CDS depth (orth.) 0.05 0.66*** 0.79*** 0.74*** (0.50) (3.89) (4.57) (3.59) 0.17 1.20 0.85 0.78 B. CDS momentum returns by size, rating, and CDS depth . . . Characteristic-based terciles (T) . . . T1 . T2 . T3 . T3–T1 . . Small Large Size 1.06*** 0.35*** 0.21*** −0.85*** (5.19) (2.72) (2.77) (-5.10) 1.03 0.72 0.73 −0.92 Size (orth.) 0.77*** 0.43*** 0.50*** −0.27 (3.93) (4.02) (3.17) (-1.62) 0.82 0.78 0.97 −0.35 Safe Risky S&P rating 0.07 0.25*** 1.34*** 1.27*** (1.28) (2.84) (4.62) (4.63) 0.23 0.61 1.05 1.11 S&P rating (orth.) 0.18* 0.60*** 0.83*** 0.65*** (1.73) (3.29) (4.83) (3.67) 0.29 0.91 1.08 0.92 Low High CDS depth −0.06** 0.16 1.05*** 1.11*** (-2.13) (1.08) (5.91) (6.07) −0.57 0.39 1.09 1.16 CDS depth (orth.) 0.05 0.66*** 0.79*** 0.74*** (0.50) (3.89) (4.57) (3.59) 0.17 1.20 0.85 0.78 This table shows the relation between CDS momentum and firm characteristics, namely, size (equity market capitalization), S&P credit rating, and CDS depth. Panel A reports average characteristics and characteristic-adjusted returns of our CDS momentum strategy ( MOMJ=3mcds ). Characteristic-adjusted CDS returns are computed by first creating quintiles of the characteristic and then subtracting the quintile’s average CDS return from the raw CDS return. Panel B reports momentum returns within characteristic-based terciles formed each month. The CDS momentum strategy is based on quintiles of past 3-month CDS return (P5 are CDS winners, P1 are CDS losers). Momentum portfolios equally weight the constituents and are rebalanced at the end of each month. Newey-West t-statistics (12 lags) are provided in parentheses. Notation for each portfolio return follows the pattern: mean, (t-statistic), and annualized Sharpe ratio. * p <.1; ** p <.05; *** p <.01. Open in new tab B. CDS momentum returns by size, rating, and CDS depth . . . Characteristic-based terciles (T) . . . T1 . T2 . T3 . T3–T1 . . Small Large Size 1.06*** 0.35*** 0.21*** −0.85*** (5.19) (2.72) (2.77) (-5.10) 1.03 0.72 0.73 −0.92 Size (orth.) 0.77*** 0.43*** 0.50*** −0.27 (3.93) (4.02) (3.17) (-1.62) 0.82 0.78 0.97 −0.35 Safe Risky S&P rating 0.07 0.25*** 1.34*** 1.27*** (1.28) (2.84) (4.62) (4.63) 0.23 0.61 1.05 1.11 S&P rating (orth.) 0.18* 0.60*** 0.83*** 0.65*** (1.73) (3.29) (4.83) (3.67) 0.29 0.91 1.08 0.92 Low High CDS depth −0.06** 0.16 1.05*** 1.11*** (-2.13) (1.08) (5.91) (6.07) −0.57 0.39 1.09 1.16 CDS depth (orth.) 0.05 0.66*** 0.79*** 0.74*** (0.50) (3.89) (4.57) (3.59) 0.17 1.20 0.85 0.78 B. CDS momentum returns by size, rating, and CDS depth . . . Characteristic-based terciles (T) . . . T1 . T2 . T3 . T3–T1 . . Small Large Size 1.06*** 0.35*** 0.21*** −0.85*** (5.19) (2.72) (2.77) (-5.10) 1.03 0.72 0.73 −0.92 Size (orth.) 0.77*** 0.43*** 0.50*** −0.27 (3.93) (4.02) (3.17) (-1.62) 0.82 0.78 0.97 −0.35 Safe Risky S&P rating 0.07 0.25*** 1.34*** 1.27*** (1.28) (2.84) (4.62) (4.63) 0.23 0.61 1.05 1.11 S&P rating (orth.) 0.18* 0.60*** 0.83*** 0.65*** (1.73) (3.29) (4.83) (3.67) 0.29 0.91 1.08 0.92 Low High CDS depth −0.06** 0.16 1.05*** 1.11*** (-2.13) (1.08) (5.91) (6.07) −0.57 0.39 1.09 1.16 CDS depth (orth.) 0.05 0.66*** 0.79*** 0.74*** (0.50) (3.89) (4.57) (3.59) 0.17 1.20 0.85 0.78 This table shows the relation between CDS momentum and firm characteristics, namely, size (equity market capitalization), S&P credit rating, and CDS depth. Panel A reports average characteristics and characteristic-adjusted returns of our CDS momentum strategy ( MOMJ=3mcds ). Characteristic-adjusted CDS returns are computed by first creating quintiles of the characteristic and then subtracting the quintile’s average CDS return from the raw CDS return. Panel B reports momentum returns within characteristic-based terciles formed each month. The CDS momentum strategy is based on quintiles of past 3-month CDS return (P5 are CDS winners, P1 are CDS losers). Momentum portfolios equally weight the constituents and are rebalanced at the end of each month. Newey-West t-statistics (12 lags) are provided in parentheses. Notation for each portfolio return follows the pattern: mean, (t-statistic), and annualized Sharpe ratio. * p <.1; ** p <.05; *** p <.01. Open in new tab Although we do not find that systematic exposure to these firm and contract characteristics are key drivers of momentum profits in CDS returns, their relation to the performance of CDS momentum is still an open and interesting question. An analysis of this relation could provide some insights on pinning down potential mechanisms that drive CDS momentum profits. In panel B of Table 3, we present long-short strategy returns within terciles (T1 to T3) of firm size, S&P rating, and CDS depth, respectively. We first create terciles based on each raw variable, and then to better isolate each variable’s own effect we orthogonalize it to the other two variables.24 We find that small firms produce greater momentum returns compared with large firms. The smallest size tercile (T1) produces a long-short return that is 0.85% greater than the largest tercile (T3), a result that is statistically significant at the 1% level. However, this size effect is confounded with both S&P rating and CDS depth. After orthogonalizing to rating and depth, the size effect is significantly diminished to a return differential of 0.27% with a t-statistic of 1.62.25 Confirming the result in Table 2 that junk-grade firms generate larger momentum returns than investment-grade firms, we also find that the riskiest rating tercile produces a long-short return that is 1.27% greater than the safest tercile (see S&P Rating results). This result is statistically significant at the 1% level. Even after orthogonalizing rating to size and depth, the difference is 0.65% and still statistically significant at the 1% level (see S&P Rating (Orthogonalized)), indicating a close relation between CDS momentum profits and the reference entities’ ratings. These CDS momentum results with corporate credit ratings are consistent with those of Avramov et al. (2007), who document that stock momentum exists only within firms with poor credit ratings. Related to this point, Avramov et al. (2013) further show that momentum returns are generated primarily in high credit risk firms that experience continuing financial distress. We find that rating changes, both upgrades and downgrades, are much more frequent in junk-grade entities than investment-grade entities; there are 7.2% (3.6%) more upgrades (downgrades) among junk-grade entities in our sample. Later in Section 2.2, we explore the role of ratings upgrades (downgrades) following substantial run-ups (rundowns) in CDS returns in driving CDS momentum profits. Using our third characteristic, we analyze the performance of CDS momentum across different contract depths. The relationship between depth and momentum can a priori take several forms. If CDS momentum is driven by illiquidity, then we might expect momentum to be concentrated in the least liquid CDS contracts (i.e., lowest CDS depth). Relatedly, if depth is an indicator of the information environment of the firm (similar to analyst coverage), we might expect momentum to be concentrated in the most opaque firms, particularly if it is driven by slow information diffusion and underreaction (Hong et al 2000; Hong and Stein 1999). In contrast to these first two hypotheses, we might expect momentum to be greatest among firms with relatively high depth if depth captures a glamour effect. Lee and Swaminathan (2000) find that stock return momentum is more pronounced among high volume stocks, an effect that is primarily driven by a quicker and strong rebound of low volume losers relative to high volume losers. They explain this quicker reversal of the low volume losers through a phenomenon they term as the momentum life cycle (MLC). A fourth hypothesis we consider, which places an emphasis on the endogenous aspect of CDS liquidity (Qiu and Yu 2012), is that CDS depth is positively correlated with the level of informed trading/quote updating in the CDS market. In this case, the more informed high depth CDS contracts could have richer information on firm default risk, thereby more accurately predicting future credit rating changes. With this endogenous liquidity, more upgrades (downgrades) could exist in the winner (loser) CDS momentum portfolio. This rating change channel is also documented as a main driver of equity momentum returns (Avramov et al. 2013). The third and fourth hypotheses, which are described above, predict greater momentum profits in more actively trading entities, that is, entities with high CDS depth. As shown in panel B of Table 3, we find support for the latter two hypotheses that CDS contracts with high depth generate higher momentum returns than those with low depth. The most liquid tercile (T3) based on CDS depth produces a long-short return that is 1.11% greater than the least liquid tercile (T1), a result that is statistically significant at the 1% level. The depth effect is strong even after orthogonalization to both size and rating; the difference between T3 and T1 remains large at 0.74% per month and is statistically significant at the 1% level. Moreover, the CDS momentum returns in the highest depth tercile (1.05%) are also robust to transaction cost adjustments. Following the method of Bongaerts, De Jong, and Driessen (2011) and using Datastream and Capital IQ bid-ask spread data,26 we find the average transaction cost of CDS winners (losers) in the highest CDS depth tercile to be 32.2 bps (34.2 bps). With a realized month-to-month turnover rate of 42.0% (43.9%) for the winner (loser) portfolio, our 3-month formation and 1-month holding period CDS momentum strategy yields 0.765% per month, net of transaction costs, for the high depth group.27 2.2 Credit rating changes Motivated by our findings on the effects of S&P rating and CDS depth on the profitability of the CDS momentum strategy, we examine in this subsection whether CDS momentum profits are generated from “anticipated” rating changes, that is, upgrades following run-ups and downgrades following rundowns of cumulative CDS returns. If this is the channel through which CDS momentum profits are generated, the MLC hypothesis (Lee and Swaminathan, 2000), which also predicts greater momentum profits for high-depth CDS contracts, would be effectively ruled out. That is, we aim to differentiate between two competing hypotheses, namely, the MLC hypothesis of Lee and Swaminathan (2000) and the endogenous liquidity hypothesis proposing that the informed trading implied by high CDS depth leads to anticipated credit rating changes in the future (Lee, Naranjo, and Velioglu 2018). Confirming prior studies on the relationship between CDS and credit ratings, panel A of Table 4 shows that a firm has positive (negative) CDS returns in the months prior to a rating upgrade (downgrade) and a large positive (negative) return during the month of the announcement. In the month prior to the new rating announcement, upgraded (downgraded) firms already experience a 6-month cumulative CDS return of 2.53% (-3.48%). During the month of the announcement, upgrades (downgrades) further experience a return of 0.78% (-2.39%). To the extent that our momentum portfolios effectively place firms to be upgraded (downgraded) in the winner (loser) portfolio, the long-short return will benefit from the divergence in spread movements between anticipated upgrades and downgrades. These rating upgrades (downgrades) following past cumulative CDS return run-up (run-down) are referred to in the table as “anticipated events.” Table 4 CDS momentum and credit rating changes A. Cumulative CDS return prior to credit ratings change event . . Cumulative CDS return (beginning at month t – 6) . CDS return in month t . . t – 6 . t – 5 . t – 4 . t – 3 . t – 2 . t – 1 . . Upgrade 0.42 1.26*** 1.61*** 1.61*** 1.76** 2.53*** 0.78*** (1.13) (2.73) (3.31) (2.82) (2.54) (4.30) (3.50) Downgrade −0.98*** −1.65*** −2.31*** −2.76*** −2.88*** −3.48*** −2.39*** (-4.87) (-6.69) (-7.51) (-8.09) (-7.71) (-6.71) (-5.45) A. Cumulative CDS return prior to credit ratings change event . . Cumulative CDS return (beginning at month t – 6) . CDS return in month t . . t – 6 . t – 5 . t – 4 . t – 3 . t – 2 . t – 1 . . Upgrade 0.42 1.26*** 1.61*** 1.61*** 1.76** 2.53*** 0.78*** (1.13) (2.73) (3.31) (2.82) (2.54) (4.30) (3.50) Downgrade −0.98*** −1.65*** −2.31*** −2.76*** −2.88*** −3.48*** −2.39*** (-4.87) (-6.69) (-7.51) (-8.09) (-7.71) (-6.71) (-5.45) Open in new tab Table 4 CDS momentum and credit rating changes A. Cumulative CDS return prior to credit ratings change event . . Cumulative CDS return (beginning at month t – 6) . CDS return in month t . . t – 6 . t – 5 . t – 4 . t – 3 . t – 2 . t – 1 . . Upgrade 0.42 1.26*** 1.61*** 1.61*** 1.76** 2.53*** 0.78*** (1.13) (2.73) (3.31) (2.82) (2.54) (4.30) (3.50) Downgrade −0.98*** −1.65*** −2.31*** −2.76*** −2.88*** −3.48*** −2.39*** (-4.87) (-6.69) (-7.51) (-8.09) (-7.71) (-6.71) (-5.45) A. Cumulative CDS return prior to credit ratings change event . . Cumulative CDS return (beginning at month t – 6) . CDS return in month t . . t – 6 . t – 5 . t – 4 . t – 3 . t – 2 . t – 1 . . Upgrade 0.42 1.26*** 1.61*** 1.61*** 1.76** 2.53*** 0.78*** (1.13) (2.73) (3.31) (2.82) (2.54) (4.30) (3.50) Downgrade −0.98*** −1.65*** −2.31*** −2.76*** −2.88*** −3.48*** −2.39*** (-4.87) (-6.69) (-7.51) (-8.09) (-7.71) (-6.71) (-5.45) Open in new tab B. CDS momentum returns and credit rating change events . . CDS quintiles . Long-short strategy (= P5 – P1) . . P1 (loser) . P5 (winner) . Full . Precrisis . Crisis . Postcrisis . Frequency of future rating change events Net exposure Upgrades ( nup/N ) 3.7% 8.6% 4.9%*** 4.2%*** 6.8%*** 3.9%*** Downgrades ( ndown/N ) 14.4% 7.2% −7.2%*** −8.8%*** −8.3%** −4.2%** Average CDS return within rating change subgroups Firms with Downgrades Upgrades anticipated events −1.41*** 1.39** 2.81*** 1.87*** 4.48*** 2.31*** (-3.29) (2.57) (5.01) (5.24) (3.28) (4.79) All other firms 0.08 0.27 0.19 0.13 0.16 0.30*** (0.49) (1.15) (1.46) (0.53) (0.61) (2.77) B. CDS momentum returns and credit rating change events . . CDS quintiles . Long-short strategy (= P5 – P1) . . P1 (loser) . P5 (winner) . Full . Precrisis . Crisis . Postcrisis . Frequency of future rating change events Net exposure Upgrades ( nup/N ) 3.7% 8.6% 4.9%*** 4.2%*** 6.8%*** 3.9%*** Downgrades ( ndown/N ) 14.4% 7.2% −7.2%*** −8.8%*** −8.3%** −4.2%** Average CDS return within rating change subgroups Firms with Downgrades Upgrades anticipated events −1.41*** 1.39** 2.81*** 1.87*** 4.48*** 2.31*** (-3.29) (2.57) (5.01) (5.24) (3.28) (4.79) All other firms 0.08 0.27 0.19 0.13 0.16 0.30*** (0.49) (1.15) (1.46) (0.53) (0.61) (2.77) Open in new tab B. CDS momentum returns and credit rating change events . . CDS quintiles . Long-short strategy (= P5 – P1) . . P1 (loser) . P5 (winner) . Full . Precrisis . Crisis . Postcrisis . Frequency of future rating change events Net exposure Upgrades ( nup/N ) 3.7% 8.6% 4.9%*** 4.2%*** 6.8%*** 3.9%*** Downgrades ( ndown/N ) 14.4% 7.2% −7.2%*** −8.8%*** −8.3%** −4.2%** Average CDS return within rating change subgroups Firms with Downgrades Upgrades anticipated events −1.41*** 1.39** 2.81*** 1.87*** 4.48*** 2.31*** (-3.29) (2.57) (5.01) (5.24) (3.28) (4.79) All other firms 0.08 0.27 0.19 0.13 0.16 0.30*** (0.49) (1.15) (1.46) (0.53) (0.61) (2.77) B. CDS momentum returns and credit rating change events . . CDS quintiles . Long-short strategy (= P5 – P1) . . P1 (loser) . P5 (winner) . Full . Precrisis . Crisis . Postcrisis . Frequency of future rating change events Net exposure Upgrades ( nup/N ) 3.7% 8.6% 4.9%*** 4.2%*** 6.8%*** 3.9%*** Downgrades ( ndown/N ) 14.4% 7.2% −7.2%*** −8.8%*** −8.3%** −4.2%** Average CDS return within rating change subgroups Firms with Downgrades Upgrades anticipated events −1.41*** 1.39** 2.81*** 1.87*** 4.48*** 2.31*** (-3.29) (2.57) (5.01) (5.24) (3.28) (4.79) All other firms 0.08 0.27 0.19 0.13 0.16 0.30*** (0.49) (1.15) (1.46) (0.53) (0.61) (2.77) Open in new tab C. CDS momentum and credit rating change events: Investment-/junk- grade . . CDS quintiles . Long-short strategy (= P5 – P1) . . P1 (loser) . P5 (winner) . Full . Precrisis . Crisis . Postcrisis . Investment-grade firms Frequency of future rating change events Net exposure Upgrades ( nup/N ) 2.3% 2.5% 0.2% −0.5% 0.8% 0.4% Downgrades ( ndown/N ) 12.4% 6.9% −5.4%*** −5.9%*** −5.8%** −4.5%*** Average CDS return within rating change subgroups Firms with Downgrades Upgrades anticipated events −0.50** −0.03 0.30* 0.17 0.73* 0.03 (-2.08) (-0.31) (1.69) (1.58) (1.73) (0.12) All other firms 0.05 0.09 0.04 0.11 −0.08 0.08* (0.42) (0.66) (0.56) (1.22) (-0.42) (1.90) Junk-grade firms Frequency of future rating change events Net exposure Upgrades ( nup/N ) 6.2% 11.9% 5.7%*** 3.8%** 10.1%*** 3.8%*** Downgrades ( ndown/N ) 18.5% 8.2% −10.3%*** −13.9%*** −11.0%** −5.1%* Average CDS return within rating change subgroups Firms with Downgrades Upgrades anticipated events −2.29*** 1.68*** 3.93*** 2.35*** 5.60*** 4.23*** (-3.49) (3.01) (5.69) (4.33) (3.55) (4.98) All other firms 0.23 0.52 0.30 0.14 0.20 0.58** (0.85) (1.44) (1.29) (0.33) (0.50) (2.12) Return difference (junk - investment): 3.63*** Firms with anticipated events (5.58) Return difference (junk - investment): 0.26 All other firms (1.31) C. CDS momentum and credit rating change events: Investment-/junk- grade . . CDS quintiles . Long-short strategy (= P5 – P1) . . P1 (loser) . P5 (winner) . Full . Precrisis . Crisis . Postcrisis . Investment-grade firms Frequency of future rating change events Net exposure Upgrades ( nup/N ) 2.3% 2.5% 0.2% −0.5% 0.8% 0.4% Downgrades ( ndown/N ) 12.4% 6.9% −5.4%*** −5.9%*** −5.8%** −4.5%*** Average CDS return within rating change subgroups Firms with Downgrades Upgrades anticipated events −0.50** −0.03 0.30* 0.17 0.73* 0.03 (-2.08) (-0.31) (1.69) (1.58) (1.73) (0.12) All other firms 0.05 0.09 0.04 0.11 −0.08 0.08* (0.42) (0.66) (0.56) (1.22) (-0.42) (1.90) Junk-grade firms Frequency of future rating change events Net exposure Upgrades ( nup/N ) 6.2% 11.9% 5.7%*** 3.8%** 10.1%*** 3.8%*** Downgrades ( ndown/N ) 18.5% 8.2% −10.3%*** −13.9%*** −11.0%** −5.1%* Average CDS return within rating change subgroups Firms with Downgrades Upgrades anticipated events −2.29*** 1.68*** 3.93*** 2.35*** 5.60*** 4.23*** (-3.49) (3.01) (5.69) (4.33) (3.55) (4.98) All other firms 0.23 0.52 0.30 0.14 0.20 0.58** (0.85) (1.44) (1.29) (0.33) (0.50) (2.12) Return difference (junk - investment): 3.63*** Firms with anticipated events (5.58) Return difference (junk - investment): 0.26 All other firms (1.31) Open in new tab C. CDS momentum and credit rating change events: Investment-/junk- grade . . CDS quintiles . Long-short strategy (= P5 – P1) . . P1 (loser) . P5 (winner) . Full . Precrisis . Crisis . Postcrisis . Investment-grade firms Frequency of future rating change events Net exposure Upgrades ( nup/N ) 2.3% 2.5% 0.2% −0.5% 0.8% 0.4% Downgrades ( ndown/N ) 12.4% 6.9% −5.4%*** −5.9%*** −5.8%** −4.5%*** Average CDS return within rating change subgroups Firms with Downgrades Upgrades anticipated events −0.50** −0.03 0.30* 0.17 0.73* 0.03 (-2.08) (-0.31) (1.69) (1.58) (1.73) (0.12) All other firms 0.05 0.09 0.04 0.11 −0.08 0.08* (0.42) (0.66) (0.56) (1.22) (-0.42) (1.90) Junk-grade firms Frequency of future rating change events Net exposure Upgrades ( nup/N ) 6.2% 11.9% 5.7%*** 3.8%** 10.1%*** 3.8%*** Downgrades ( ndown/N ) 18.5% 8.2% −10.3%*** −13.9%*** −11.0%** −5.1%* Average CDS return within rating change subgroups Firms with Downgrades Upgrades anticipated events −2.29*** 1.68*** 3.93*** 2.35*** 5.60*** 4.23*** (-3.49) (3.01) (5.69) (4.33) (3.55) (4.98) All other firms 0.23 0.52 0.30 0.14 0.20 0.58** (0.85) (1.44) (1.29) (0.33) (0.50) (2.12) Return difference (junk - investment): 3.63*** Firms with anticipated events (5.58) Return difference (junk - investment): 0.26 All other firms (1.31) C. CDS momentum and credit rating change events: Investment-/junk- grade . . CDS quintiles . Long-short strategy (= P5 – P1) . . P1 (loser) . P5 (winner) . Full . Precrisis . Crisis . Postcrisis . Investment-grade firms Frequency of future rating change events Net exposure Upgrades ( nup/N ) 2.3% 2.5% 0.2% −0.5% 0.8% 0.4% Downgrades ( ndown/N ) 12.4% 6.9% −5.4%*** −5.9%*** −5.8%** −4.5%*** Average CDS return within rating change subgroups Firms with Downgrades Upgrades anticipated events −0.50** −0.03 0.30* 0.17 0.73* 0.03 (-2.08) (-0.31) (1.69) (1.58) (1.73) (0.12) All other firms 0.05 0.09 0.04 0.11 −0.08 0.08* (0.42) (0.66) (0.56) (1.22) (-0.42) (1.90) Junk-grade firms Frequency of future rating change events Net exposure Upgrades ( nup/N ) 6.2% 11.9% 5.7%*** 3.8%** 10.1%*** 3.8%*** Downgrades ( ndown/N ) 18.5% 8.2% −10.3%*** −13.9%*** −11.0%** −5.1%* Average CDS return within rating change subgroups Firms with Downgrades Upgrades anticipated events −2.29*** 1.68*** 3.93*** 2.35*** 5.60*** 4.23*** (-3.49) (3.01) (5.69) (4.33) (3.55) (4.98) All other firms 0.23 0.52 0.30 0.14 0.20 0.58** (0.85) (1.44) (1.29) (0.33) (0.50) (2.12) Return difference (junk - investment): 3.63*** Firms with anticipated events (5.58) Return difference (junk - investment): 0.26 All other firms (1.31) Open in new tab D. CDS momentum and credit rating change events: CDS depth . . CDS quintiles . Long-short strategy (= P5 – P1) . . P1 (loser) . P5 (winner) . Full . Precrisis . Crisis . Postcrisis . High-depth firms Frequency of future rating change events Net exposure Upgrades ( nup/N ) 3.5% 8.2% 4.7%*** 3.4%*** 7.1%*** 3.9%*** Downgrades ( ndown/N ) 15.7% 7.4% −8.3%*** −9.1%*** −9.7%*** −6.0%*** Average CDS return within rating change subgroups Firms with Downgrades Upgrades anticipated events −1.62*** 1.54*** 3.04*** 1.99*** 4.58*** 2.81*** (-3.42) (3.07) (5.87) (4.53) (3.96) (5.68) All other firms 0.07 0.22 0.15 0.16 0.02 0.26 (0.40) (1.02) (1.14) (0.75) (0.07) (1.58) Low-depth firms Frequency of future rating change events Net exposure Upgrades ( nup/N ) 4.1% 10.7% 6.7%*** 8.5%*** 7.8%*** 3.2%*** Downgrades ( ndown/N ) 11.2% 7.1% −4.2%*** −5.2%** −4.2% −2.8%*** Average CDS return within rating change subgroups Firms with Downgrades Upgrades anticipated events −1.05** 0.48** 0.86** 0.74 1.15 0.71* (-2.00) (2.13) (2.40) (1.33) (1.61) (1.68) All other firms 0.16 0.30 0.14 0.05 0.02 0.37 (0.84) (0.78) (0.47) (0.10) (0.03) (1.30) Return difference (high depth - low depth): 2.18*** Firms with anticipated events (3.55) Return difference (high depth - low depth): 0.01 All other firms (0.04) D. CDS momentum and credit rating change events: CDS depth . . CDS quintiles . Long-short strategy (= P5 – P1) . . P1 (loser) . P5 (winner) . Full . Precrisis . Crisis . Postcrisis . High-depth firms Frequency of future rating change events Net exposure Upgrades ( nup/N ) 3.5% 8.2% 4.7%*** 3.4%*** 7.1%*** 3.9%*** Downgrades ( ndown/N ) 15.7% 7.4% −8.3%*** −9.1%*** −9.7%*** −6.0%*** Average CDS return within rating change subgroups Firms with Downgrades Upgrades anticipated events −1.62*** 1.54*** 3.04*** 1.99*** 4.58*** 2.81*** (-3.42) (3.07) (5.87) (4.53) (3.96) (5.68) All other firms 0.07 0.22 0.15 0.16 0.02 0.26 (0.40) (1.02) (1.14) (0.75) (0.07) (1.58) Low-depth firms Frequency of future rating change events Net exposure Upgrades ( nup/N ) 4.1% 10.7% 6.7%*** 8.5%*** 7.8%*** 3.2%*** Downgrades ( ndown/N ) 11.2% 7.1% −4.2%*** −5.2%** −4.2% −2.8%*** Average CDS return within rating change subgroups Firms with Downgrades Upgrades anticipated events −1.05** 0.48** 0.86** 0.74 1.15 0.71* (-2.00) (2.13) (2.40) (1.33) (1.61) (1.68) All other firms 0.16 0.30 0.14 0.05 0.02 0.37 (0.84) (0.78) (0.47) (0.10) (0.03) (1.30) Return difference (high depth - low depth): 2.18*** Firms with anticipated events (3.55) Return difference (high depth - low depth): 0.01 All other firms (0.04) This table provides results on the relationship between CDS momentum and future S&P credit rating changes. Panel A presents the cumulative CDS return (%) leading up to a credit rating change event in month t. Panel B examines CDS momentum returns and credit rating changes. After forming the momentum portfolios, we track the returns of firms that have a rating change in the next 6 months and those that do not. Frequency of ratings change events (percentage of firms in the portfolio labeled nup/N for upgrades and ndown/N for downgrades) are shown for winners (P5) and losers portfolios (P1). Returns are reported separately for firms that do and do not undergo a rating change in the “anticipated” direction, that is, an upgrade for winners (P5) and a downgrade for losers (P1). Panel C divides firms by investment/junk grade and repeats the analysis. Panel D divides firms by high/low depth and repeats the analysis. Newey-West t-statistics (12 lags) are provided in parentheses. * p <.1; ** p <.05; *** p <.01. Open in new tab D. CDS momentum and credit rating change events: CDS depth . . CDS quintiles . Long-short strategy (= P5 – P1) . . P1 (loser) . P5 (winner) . Full . Precrisis . Crisis . Postcrisis . High-depth firms Frequency of future rating change events Net exposure Upgrades ( nup/N ) 3.5% 8.2% 4.7%*** 3.4%*** 7.1%*** 3.9%*** Downgrades ( ndown/N ) 15.7% 7.4% −8.3%*** −9.1%*** −9.7%*** −6.0%*** Average CDS return within rating change subgroups Firms with Downgrades Upgrades anticipated events −1.62*** 1.54*** 3.04*** 1.99*** 4.58*** 2.81*** (-3.42) (3.07) (5.87) (4.53) (3.96) (5.68) All other firms 0.07 0.22 0.15 0.16 0.02 0.26 (0.40) (1.02) (1.14) (0.75) (0.07) (1.58) Low-depth firms Frequency of future rating change events Net exposure Upgrades ( nup/N ) 4.1% 10.7% 6.7%*** 8.5%*** 7.8%*** 3.2%*** Downgrades ( ndown/N ) 11.2% 7.1% −4.2%*** −5.2%** −4.2% −2.8%*** Average CDS return within rating change subgroups Firms with Downgrades Upgrades anticipated events −1.05** 0.48** 0.86** 0.74 1.15 0.71* (-2.00) (2.13) (2.40) (1.33) (1.61) (1.68) All other firms 0.16 0.30 0.14 0.05 0.02 0.37 (0.84) (0.78) (0.47) (0.10) (0.03) (1.30) Return difference (high depth - low depth): 2.18*** Firms with anticipated events (3.55) Return difference (high depth - low depth): 0.01 All other firms (0.04) D. CDS momentum and credit rating change events: CDS depth . . CDS quintiles . Long-short strategy (= P5 – P1) . . P1 (loser) . P5 (winner) . Full . Precrisis . Crisis . Postcrisis . High-depth firms Frequency of future rating change events Net exposure Upgrades ( nup/N ) 3.5% 8.2% 4.7%*** 3.4%*** 7.1%*** 3.9%*** Downgrades ( ndown/N ) 15.7% 7.4% −8.3%*** −9.1%*** −9.7%*** −6.0%*** Average CDS return within rating change subgroups Firms with Downgrades Upgrades anticipated events −1.62*** 1.54*** 3.04*** 1.99*** 4.58*** 2.81*** (-3.42) (3.07) (5.87) (4.53) (3.96) (5.68) All other firms 0.07 0.22 0.15 0.16 0.02 0.26 (0.40) (1.02) (1.14) (0.75) (0.07) (1.58) Low-depth firms Frequency of future rating change events Net exposure Upgrades ( nup/N ) 4.1% 10.7% 6.7%*** 8.5%*** 7.8%*** 3.2%*** Downgrades ( ndown/N ) 11.2% 7.1% −4.2%*** −5.2%** −4.2% −2.8%*** Average CDS return within rating change subgroups Firms with Downgrades Upgrades anticipated events −1.05** 0.48** 0.86** 0.74 1.15 0.71* (-2.00) (2.13) (2.40) (1.33) (1.61) (1.68) All other firms 0.16 0.30 0.14 0.05 0.02 0.37 (0.84) (0.78) (0.47) (0.10) (0.03) (1.30) Return difference (high depth - low depth): 2.18*** Firms with anticipated events (3.55) Return difference (high depth - low depth): 0.01 All other firms (0.04) This table provides results on the relationship between CDS momentum and future S&P credit rating changes. Panel A presents the cumulative CDS return (%) leading up to a credit rating change event in month t. Panel B examines CDS momentum returns and credit rating changes. After forming the momentum portfolios, we track the returns of firms that have a rating change in the next 6 months and those that do not. Frequency of ratings change events (percentage of firms in the portfolio labeled nup/N for upgrades and ndown/N for downgrades) are shown for winners (P5) and losers portfolios (P1). Returns are reported separately for firms that do and do not undergo a rating change in the “anticipated” direction, that is, an upgrade for winners (P5) and a downgrade for losers (P1). Panel C divides firms by investment/junk grade and repeats the analysis. Panel D divides firms by high/low depth and repeats the analysis. Newey-West t-statistics (12 lags) are provided in parentheses. * p <.1; ** p <.05; *** p <.01. Open in new tab To analyze the relationship between CDS momentum and credit rating changes, we analyze the frequency of rating upgrades and downgrades within each of the winner and loser portfolios and report separately the returns of firms that do and do not experience an anticipated rating change within 6 months of portfolio formation. In a similar analysis conducted by Jostova et al. (2013), no relation is found between the momentum profits in corporate bond returns and the future rating change channels. Contrary to their findings, our results in Table 4 show that CDS momentum and ratings changes are closely related, suggesting distinct mechanisms generating momentum effects within each market. In Table 4, we focus on the most profitable CDS momentum strategy that uses formation period J=3m and holding period K=1m as we did in Section 2.1. First, we find that the long-short CDS momentum strategy has a net positive exposure to rating upgrades and net negative exposure to rating downgrades. Panel B of Table 4 shows the frequency of rating upgrade and downgrade events (denoted nup/N and ndown/N , respectively) within the CDS winner and loser portfolios. We find that 8.6% of CDS winners (P5) experience a rating upgrade, compared to only 3.7% for CDS losers (P1). On the other hand, we find that 14.4% of CDS losers and only 7.2% of CDS winners get downgraded. The long-short strategy is favorably exposed to future credit rating changes with a net positive exposure of 4.9% to upgrades and a net negative exposure of -7.2% to downgrades, both of which are statistically significant at the 1% level. Next, as more direct evidence of our proposed mechanism for CDS momentum profits, we report separately the returns of firms that do and do not experience an anticipated rating change event. In panel B of Table 4, we show that CDS momentum profits are exclusively generated in firms with anticipated credit rating changes over the 6-month horizon following the portfolio formation date; these firms generate 2.81% per month with a t-statistic of 5.01. By contrast, firms that do not undergo an anticipated rating change event contribute very little to the long-short CDS momentum strategy, generating only 0.19% per month with a t-statistic of 1.46.28 Note also that the total rating change frequency differential between CDS winners and CDS losers is widest during the crisis period (15.1% = 6.8% − (−8.3%)). Consistent with this result, the return magnitude of the overall long-short strategy also was shown to be maximized during the crisis period, generating 0.91% per month (see Table 2), which appears to be primarily driven by firms that experience anticipated ratings changes (see panel B of Table 4). How are these findings related to the S&P rating and CDS depth characteristics? In panel C of Table 4, we first examine the sensitivity of the relationship between CDS momentum and ratings changes to the rating level. We divide firms into investment- and junk-grade firms and redo the analysis. Again, for the long-short portfolio, we find a net positive exposure to upgrades and net negative exposure to downgrades in both investment-grade (5.6% = 0.2%- (-5.4%)) and junk-grade groups (16.0% = 5.7%- (-10.3%)). We also observe that removing firms with “anticipated” credit rating changes entirely eliminates momentum returns in both investment- and junk-grade groups; the long-short strategy among these firms produces 0.04% and 0.30% per month for investment- and junk-grade firms, respectively, neither of which is statistically significant. Absent firms with anticipated rating changes, there is no significant difference in momentum returns between investment- and junk-grade firms (0.26% with a t-statistic of 1.31). The better performance of the CDS momentum strategy among junk-grade firms clearly comes from the difference in future rating changes in anticipated directions as the monthly performance differential between junk- and investment-grade firms amounts to 3.63% per month with a t-statistic of 5.58. What is the role of CDS depth in the relationship between CDS momentum and ratings changes? In panel D of Table 4, we divide firms into high depth and low depth and repeat the previous analysis. The evidence supports the hypothesis that CDS momentum profits are greater among high-depth firms relative to low-depth firms because of their superior ability to predict ratings changes in the anticipated directions. That is, the combined net exposure of the long-short momentum strategy to anticipated rating changes—positive net exposure to upgrades and negative net exposure to downgrades—is greater among high-depth firms ( 13.0%=4.7%−(−8.3% )) compared to low-depth firms ( 10.9%=6.7%−(−4.2%) ). Furthermore, firms that experience anticipated rating events generate a higher long-short return for the high-depth group (3.04% with a t-statistic of 5.87) compared to the low depth group (0.86% with a t-statistic of 2.40). The difference in these returns is substantial at 2.18% ( =3.04%−0.86% ) and statistically significant at the 1% level. Our findings regarding the CDS depth effect are consistent with Qiu and Yu (2012) who recently document that CDS depth is a proxy for informed trading in the CDS market. The at-market spreads of CDS contracts with high endogenous liquidity are also shown to reveal information earlier than stock returns, particularly when the credit quality of reference entities deteriorates. The ability of past CDS returns with high depth contracts to predict rating downgrades better than rating upgrades is also in line with their findings. These results are sharply confirmed in a panel vector autoregression (VAR) model by Lee, Naranjo, and Velioglu (2018). The analysis in panel D of Table 4 allows for the possibility that past CDS returns have an informational advantage over stock and bond returns in predicting future ratings changes. 3. Bond Momentum versus CDS Momentum One reasonable hypothesis is that our CDS momentum is no more than a reflection of the momentum exhibited by the underlying bond market as documented by Jostova et al. (2013), hereafter referred to as JNPS. As an explanation for their findings, JNPS conclude that slow information diffusion likely accounts for the persistence of bond momentum. However, unlike the results of similar tests conducted by JNPS, we find a significant role of credit rating changes in generating CDS momentum (see Section 2.2). To reconcile these differences and shed light on the subject, we now more directly investigate the relationship between CDS momentum and bond momentum. We start by examining corporate bond return momentum using the full TRACE data set. The bond momentum strategy is computed at the firm-level by value-weighting individual bond issues by their issue size. Bond-based momentum portfolios (labeled PB) are constructed from quintiles of past bond performance using formation periods (J) ranging from 1 to 12 months. We equally weight firms in each portfolio and examine average portfolio returns over a 1-month holding period (K), using end-of-month rebalancing. Following JNPS, formation periods between 3 and 12 months skip the most recent month.29 Panel A of Table 5 reports average monthly returns of the bond momentum portfolios and the long-short strategy that buys bond winners (PB5) and sells short bond losers (PB1). Also reported are the long-short strategy’s annualized Sharpe ratio and the six-factor alpha coefficient based on a model with the stock factors MKT, SMB, HML, and UMD and the bond factors TERM and DEF. Similar to JNPS, we find that bond momentum strategies generate significant profits when using formation periods ranging from 3 to 12 months (under the header “Full TRACE”). The long-short strategy based on a 3-month formation period generates robust returns of 0.54% per month with a t-statistic of 3.63 and a Sharpe ratio of 1.03. The 6- and 12-month formation periods also produce significant returns of 0.52% and 0.53% per month, respectfully. The long-short strategy appears to gradually become statistically weaker as the formation period is lengthened from 3 to 12 months. Table 5 Bond momentum versus CDS momentum A. Bond momentum . . Bond performance quintiles . Long-short strategy . . PB1 . PB2 . PB3 . PB4 . PB5 . PB5–PB1 . 6-factor α . Sharpe . Average bond return (% per month), K=1m Full TRACE J=1m 0.55 0.49*** 0.44*** 0.43*** 0.41* −0.14 −0.14 −0.27 (1.57) (3.87) (3.81) (3.27) (1.73) (-0.66) (-0.54) J=3m† 0.16 0.43*** 0.50*** 0.52*** 0.70*** 0.54*** 0.57*** 1.03 (0.50) (3.54) (4.23) (3.87) (2.75) (3.63) (3.74) J=6m† 0.14 0.43*** 0.51*** 0.53*** 0.66*** 0.52*** 0.47*** 0.87 (0.41) (3.34) (4.15) (4.09) (2.73) (3.20) (2.74) J=12m† 0.11 0.46*** 0.46*** 0.48*** 0.64*** 0.53* 0.54** 0.76 (0.27) (2.80) (3.76) (4.27) (3.93) (1.87) (2.07) CDS-matched firms J=1m 0.71*** 0.52*** 0.44*** 0.44*** 0.40** −0.30*** −0.27*** −1.00 (3.73) (3.76) (4.34) (3.23) (2.13) (-3.50) (-5.21) J=3m† 0.41* 0.40*** 0.52*** 0.51*** 0.62*** 0.21** 0.23** 0.54 (1.93) (2.88) (4.47) (4.18) (3.24) (2.12) (2.07) J=6m† 0.41* 0.48*** 0.49*** 0.51*** 0.57*** 0.16 0.18 0.37 (1.87) (3.85) (4.15) (4.50) (3.16) (1.58) (1.63) J=12m† 0.40 0.46*** 0.41*** 0.46*** 0.54*** 0.14 0.18 0.30 (1.58) (3.26) (3.69) (4.01) (4.23) (0.88) (1.35) Non-CDS-matched firms J=1m 0.52 0.48*** 0.44*** 0.43*** 0.40 −0.13 −0.13 −0.23 (1.36) (3.72) (3.74) (3.22) (1.58) (-0.54) (-0.58) J=3m† 0.09 0.42*** 0.50*** 0.51*** 0.73*** 0.64*** 0.66*** 1.12 (0.26) (3.40) (4.13) (3.64) (2.69) (3.62) (3.93) J=6m† 0.08 0.40*** 0.50*** 0.55*** 0.69*** 0.61*** 0.52*** 0.94 (0.21) (3.05) (4.01) (3.98) (2.63) (3.39) (2.89) J=12m† 0.05 0.45** 0.45*** 0.48*** 0.67*** 0.62* 0.60** 0.83 (0.10) (2.60) (3.65) (4.26) (3.82) (1.97) (2.11) A. Bond momentum . . Bond performance quintiles . Long-short strategy . . PB1 . PB2 . PB3 . PB4 . PB5 . PB5–PB1 . 6-factor α . Sharpe . Average bond return (% per month), K=1m Full TRACE J=1m 0.55 0.49*** 0.44*** 0.43*** 0.41* −0.14 −0.14 −0.27 (1.57) (3.87) (3.81) (3.27) (1.73) (-0.66) (-0.54) J=3m† 0.16 0.43*** 0.50*** 0.52*** 0.70*** 0.54*** 0.57*** 1.03 (0.50) (3.54) (4.23) (3.87) (2.75) (3.63) (3.74) J=6m† 0.14 0.43*** 0.51*** 0.53*** 0.66*** 0.52*** 0.47*** 0.87 (0.41) (3.34) (4.15) (4.09) (2.73) (3.20) (2.74) J=12m† 0.11 0.46*** 0.46*** 0.48*** 0.64*** 0.53* 0.54** 0.76 (0.27) (2.80) (3.76) (4.27) (3.93) (1.87) (2.07) CDS-matched firms J=1m 0.71*** 0.52*** 0.44*** 0.44*** 0.40** −0.30*** −0.27*** −1.00 (3.73) (3.76) (4.34) (3.23) (2.13) (-3.50) (-5.21) J=3m† 0.41* 0.40*** 0.52*** 0.51*** 0.62*** 0.21** 0.23** 0.54 (1.93) (2.88) (4.47) (4.18) (3.24) (2.12) (2.07) J=6m† 0.41* 0.48*** 0.49*** 0.51*** 0.57*** 0.16 0.18 0.37 (1.87) (3.85) (4.15) (4.50) (3.16) (1.58) (1.63) J=12m† 0.40 0.46*** 0.41*** 0.46*** 0.54*** 0.14 0.18 0.30 (1.58) (3.26) (3.69) (4.01) (4.23) (0.88) (1.35) Non-CDS-matched firms J=1m 0.52 0.48*** 0.44*** 0.43*** 0.40 −0.13 −0.13 −0.23 (1.36) (3.72) (3.74) (3.22) (1.58) (-0.54) (-0.58) J=3m† 0.09 0.42*** 0.50*** 0.51*** 0.73*** 0.64*** 0.66*** 1.12 (0.26) (3.40) (4.13) (3.64) (2.69) (3.62) (3.93) J=6m† 0.08 0.40*** 0.50*** 0.55*** 0.69*** 0.61*** 0.52*** 0.94 (0.21) (3.05) (4.01) (3.98) (2.63) (3.39) (2.89) J=12m† 0.05 0.45** 0.45*** 0.48*** 0.67*** 0.62* 0.60** 0.83 (0.10) (2.60) (3.65) (4.26) (3.82) (1.97) (2.11) Open in new tab Table 5 Bond momentum versus CDS momentum A. Bond momentum . . Bond performance quintiles . Long-short strategy . . PB1 . PB2 . PB3 . PB4 . PB5 . PB5–PB1 . 6-factor α . Sharpe . Average bond return (% per month), K=1m Full TRACE J=1m 0.55 0.49*** 0.44*** 0.43*** 0.41* −0.14 −0.14 −0.27 (1.57) (3.87) (3.81) (3.27) (1.73) (-0.66) (-0.54) J=3m† 0.16 0.43*** 0.50*** 0.52*** 0.70*** 0.54*** 0.57*** 1.03 (0.50) (3.54) (4.23) (3.87) (2.75) (3.63) (3.74) J=6m† 0.14 0.43*** 0.51*** 0.53*** 0.66*** 0.52*** 0.47*** 0.87 (0.41) (3.34) (4.15) (4.09) (2.73) (3.20) (2.74) J=12m† 0.11 0.46*** 0.46*** 0.48*** 0.64*** 0.53* 0.54** 0.76 (0.27) (2.80) (3.76) (4.27) (3.93) (1.87) (2.07) CDS-matched firms J=1m 0.71*** 0.52*** 0.44*** 0.44*** 0.40** −0.30*** −0.27*** −1.00 (3.73) (3.76) (4.34) (3.23) (2.13) (-3.50) (-5.21) J=3m† 0.41* 0.40*** 0.52*** 0.51*** 0.62*** 0.21** 0.23** 0.54 (1.93) (2.88) (4.47) (4.18) (3.24) (2.12) (2.07) J=6m† 0.41* 0.48*** 0.49*** 0.51*** 0.57*** 0.16 0.18 0.37 (1.87) (3.85) (4.15) (4.50) (3.16) (1.58) (1.63) J=12m† 0.40 0.46*** 0.41*** 0.46*** 0.54*** 0.14 0.18 0.30 (1.58) (3.26) (3.69) (4.01) (4.23) (0.88) (1.35) Non-CDS-matched firms J=1m 0.52 0.48*** 0.44*** 0.43*** 0.40 −0.13 −0.13 −0.23 (1.36) (3.72) (3.74) (3.22) (1.58) (-0.54) (-0.58) J=3m† 0.09 0.42*** 0.50*** 0.51*** 0.73*** 0.64*** 0.66*** 1.12 (0.26) (3.40) (4.13) (3.64) (2.69) (3.62) (3.93) J=6m† 0.08 0.40*** 0.50*** 0.55*** 0.69*** 0.61*** 0.52*** 0.94 (0.21) (3.05) (4.01) (3.98) (2.63) (3.39) (2.89) J=12m† 0.05 0.45** 0.45*** 0.48*** 0.67*** 0.62* 0.60** 0.83 (0.10) (2.60) (3.65) (4.26) (3.82) (1.97) (2.11) A. Bond momentum . . Bond performance quintiles . Long-short strategy . . PB1 . PB2 . PB3 . PB4 . PB5 . PB5–PB1 . 6-factor α . Sharpe . Average bond return (% per month), K=1m Full TRACE J=1m 0.55 0.49*** 0.44*** 0.43*** 0.41* −0.14 −0.14 −0.27 (1.57) (3.87) (3.81) (3.27) (1.73) (-0.66) (-0.54) J=3m† 0.16 0.43*** 0.50*** 0.52*** 0.70*** 0.54*** 0.57*** 1.03 (0.50) (3.54) (4.23) (3.87) (2.75) (3.63) (3.74) J=6m† 0.14 0.43*** 0.51*** 0.53*** 0.66*** 0.52*** 0.47*** 0.87 (0.41) (3.34) (4.15) (4.09) (2.73) (3.20) (2.74) J=12m† 0.11 0.46*** 0.46*** 0.48*** 0.64*** 0.53* 0.54** 0.76 (0.27) (2.80) (3.76) (4.27) (3.93) (1.87) (2.07) CDS-matched firms J=1m 0.71*** 0.52*** 0.44*** 0.44*** 0.40** −0.30*** −0.27*** −1.00 (3.73) (3.76) (4.34) (3.23) (2.13) (-3.50) (-5.21) J=3m† 0.41* 0.40*** 0.52*** 0.51*** 0.62*** 0.21** 0.23** 0.54 (1.93) (2.88) (4.47) (4.18) (3.24) (2.12) (2.07) J=6m† 0.41* 0.48*** 0.49*** 0.51*** 0.57*** 0.16 0.18 0.37 (1.87) (3.85) (4.15) (4.50) (3.16) (1.58) (1.63) J=12m† 0.40 0.46*** 0.41*** 0.46*** 0.54*** 0.14 0.18 0.30 (1.58) (3.26) (3.69) (4.01) (4.23) (0.88) (1.35) Non-CDS-matched firms J=1m 0.52 0.48*** 0.44*** 0.43*** 0.40 −0.13 −0.13 −0.23 (1.36) (3.72) (3.74) (3.22) (1.58) (-0.54) (-0.58) J=3m† 0.09 0.42*** 0.50*** 0.51*** 0.73*** 0.64*** 0.66*** 1.12 (0.26) (3.40) (4.13) (3.64) (2.69) (3.62) (3.93) J=6m† 0.08 0.40*** 0.50*** 0.55*** 0.69*** 0.61*** 0.52*** 0.94 (0.21) (3.05) (4.01) (3.98) (2.63) (3.39) (2.89) J=12m† 0.05 0.45** 0.45*** 0.48*** 0.67*** 0.62* 0.60** 0.83 (0.10) (2.60) (3.65) (4.26) (3.82) (1.97) (2.11) Open in new tab B. Average CDS returns of bond-based and CDS-based performance quintiles . . Conditional bond momentum . Conditional CDS momentum . . ( MOMJbond | MOMJcds ) . ( MOMJcds | MOMJbond ) . . PB5–PB1 . 6-factor α . Sharpe . P5–P1 . 6-factor α . Sharpe . Average CDS return (% per month), K=1m J=1m 0.07 0.10 0.23 0.39*** 0.42*** 0.93 (0.73) (0.97) (3.24) (2.96) J=3m 0.09 0.14 0.30 0.47*** 0.56*** 0.90 (1.06) (1.59) (4.82) (4.20) J=6m −0.00 0.04 −0.01 0.35*** 0.38*** 0.70 (-0.05) (0.32) (3.91) (3.46) J=12m 0.03 0.05 0.09 0.29* 0.37** 0.50 (0.26) (0.69) (1.92) (2.14) B. Average CDS returns of bond-based and CDS-based performance quintiles . . Conditional bond momentum . Conditional CDS momentum . . ( MOMJbond | MOMJcds ) . ( MOMJcds | MOMJbond ) . . PB5–PB1 . 6-factor α . Sharpe . P5–P1 . 6-factor α . Sharpe . Average CDS return (% per month), K=1m J=1m 0.07 0.10 0.23 0.39*** 0.42*** 0.93 (0.73) (0.97) (3.24) (2.96) J=3m 0.09 0.14 0.30 0.47*** 0.56*** 0.90 (1.06) (1.59) (4.82) (4.20) J=6m −0.00 0.04 −0.01 0.35*** 0.38*** 0.70 (-0.05) (0.32) (3.91) (3.46) J=12m 0.03 0.05 0.09 0.29* 0.37** 0.50 (0.26) (0.69) (1.92) (2.14) Open in new tab B. Average CDS returns of bond-based and CDS-based performance quintiles . . Conditional bond momentum . Conditional CDS momentum . . ( MOMJbond | MOMJcds ) . ( MOMJcds | MOMJbond ) . . PB5–PB1 . 6-factor α . Sharpe . P5–P1 . 6-factor α . Sharpe . Average CDS return (% per month), K=1m J=1m 0.07 0.10 0.23 0.39*** 0.42*** 0.93 (0.73) (0.97) (3.24) (2.96) J=3m 0.09 0.14 0.30 0.47*** 0.56*** 0.90 (1.06) (1.59) (4.82) (4.20) J=6m −0.00 0.04 −0.01 0.35*** 0.38*** 0.70 (-0.05) (0.32) (3.91) (3.46) J=12m 0.03 0.05 0.09 0.29* 0.37** 0.50 (0.26) (0.69) (1.92) (2.14) B. Average CDS returns of bond-based and CDS-based performance quintiles . . Conditional bond momentum . Conditional CDS momentum . . ( MOMJbond | MOMJcds ) . ( MOMJcds | MOMJbond ) . . PB5–PB1 . 6-factor α . Sharpe . P5–P1 . 6-factor α . Sharpe . Average CDS return (% per month), K=1m J=1m 0.07 0.10 0.23 0.39*** 0.42*** 0.93 (0.73) (0.97) (3.24) (2.96) J=3m 0.09 0.14 0.30 0.47*** 0.56*** 0.90 (1.06) (1.59) (4.82) (4.20) J=6m −0.00 0.04 −0.01 0.35*** 0.38*** 0.70 (-0.05) (0.32) (3.91) (3.46) J=12m 0.03 0.05 0.09 0.29* 0.37** 0.50 (0.26) (0.69) (1.92) (2.14) Open in new tab C. Bond momentum: Excluding firms with future rating change events . . Bond momentum (baseline) . Excluding future rating events . . PB5–PB1 . 6-factor α . Sharpe . PB5–PB1 . 6-factor α . Sharpe . Average bond return (% per month), K=1m J=3m , All ratings Full TRACE 0.54*** 0.52*** 1.03 0.45*** 0.43*** 0.94 (3.63) (3.58) (3.02) (3.02) CDS-matched 0.21** 0.25** 0.54 0.12 0.13 0.30 (2.12) (2.15) (1.29) (1.11) Non-CDS-matched 0.64*** 0.66*** 1.12 0.56*** 0.58*** 1.06 (3.62) (3.90) (2.98) (3.48) J=3m , junk-grade firms (BB+ and below) Full TRACE 1.01*** 1.06*** 1.32 0.87*** 1.02*** 1.19 (3.37) (4.08) (2.87) (3.87) CDS-matched 0.52** 0.68*** 0.73 0.12 0.24 0.18 (2.30) (2.61) (0.58) (1.00) Non-CDS-matched 1.13*** 1.21*** 1.35 0.97*** 1.10*** 1.22 (3.42) (4.32) (2.94) (4.03) C. Bond momentum: Excluding firms with future rating change events . . Bond momentum (baseline) . Excluding future rating events . . PB5–PB1 . 6-factor α . Sharpe . PB5–PB1 . 6-factor α . Sharpe . Average bond return (% per month), K=1m J=3m , All ratings Full TRACE 0.54*** 0.52*** 1.03 0.45*** 0.43*** 0.94 (3.63) (3.58) (3.02) (3.02) CDS-matched 0.21** 0.25** 0.54 0.12 0.13 0.30 (2.12) (2.15) (1.29) (1.11) Non-CDS-matched 0.64*** 0.66*** 1.12 0.56*** 0.58*** 1.06 (3.62) (3.90) (2.98) (3.48) J=3m , junk-grade firms (BB+ and below) Full TRACE 1.01*** 1.06*** 1.32 0.87*** 1.02*** 1.19 (3.37) (4.08) (2.87) (3.87) CDS-matched 0.52** 0.68*** 0.73 0.12 0.24 0.18 (2.30) (2.61) (0.58) (1.00) Non-CDS-matched 1.13*** 1.21*** 1.35 0.97*** 1.10*** 1.22 (3.42) (4.32) (2.94) (4.03) Open in new tab C. Bond momentum: Excluding firms with future rating change events . . Bond momentum (baseline) . Excluding future rating events . . PB5–PB1 . 6-factor α . Sharpe . PB5–PB1 . 6-factor α . Sharpe . Average bond return (% per month), K=1m J=3m , All ratings Full TRACE 0.54*** 0.52*** 1.03 0.45*** 0.43*** 0.94 (3.63) (3.58) (3.02) (3.02) CDS-matched 0.21** 0.25** 0.54 0.12 0.13 0.30 (2.12) (2.15) (1.29) (1.11) Non-CDS-matched 0.64*** 0.66*** 1.12 0.56*** 0.58*** 1.06 (3.62) (3.90) (2.98) (3.48) J=3m , junk-grade firms (BB+ and below) Full TRACE 1.01*** 1.06*** 1.32 0.87*** 1.02*** 1.19 (3.37) (4.08) (2.87) (3.87) CDS-matched 0.52** 0.68*** 0.73 0.12 0.24 0.18 (2.30) (2.61) (0.58) (1.00) Non-CDS-matched 1.13*** 1.21*** 1.35 0.97*** 1.10*** 1.22 (3.42) (4.32) (2.94) (4.03) C. Bond momentum: Excluding firms with future rating change events . . Bond momentum (baseline) . Excluding future rating events . . PB5–PB1 . 6-factor α . Sharpe . PB5–PB1 . 6-factor α . Sharpe . Average bond return (% per month), K=1m J=3m , All ratings Full TRACE 0.54*** 0.52*** 1.03 0.45*** 0.43*** 0.94 (3.63) (3.58) (3.02) (3.02) CDS-matched 0.21** 0.25** 0.54 0.12 0.13 0.30 (2.12) (2.15) (1.29) (1.11) Non-CDS-matched 0.64*** 0.66*** 1.12 0.56*** 0.58*** 1.06 (3.62) (3.90) (2.98) (3.48) J=3m , junk-grade firms (BB+ and below) Full TRACE 1.01*** 1.06*** 1.32 0.87*** 1.02*** 1.19 (3.37) (4.08) (2.87) (3.87) CDS-matched 0.52** 0.68*** 0.73 0.12 0.24 0.18 (2.30) (2.61) (0.58) (1.00) Non-CDS-matched 1.13*** 1.21*** 1.35 0.97*** 1.10*** 1.22 (3.42) (4.32) (2.94) (4.03) Open in new tab D. Bond momentum of CDS-matched firms and bond fragmentation . . Low fragmentation . High fragmentation . . PB5–PB1 . 6-factor α . Sharpe . PB5–PB1 . 6-factor α . Sharpe . Average bond return (% per month), K=1m J=3m , CDS-matched All ratings 0.32*** 0.34*** 0.65 0.12 0.16 0.31 (2.62) (2.76) (1.36) (1.28) Junk grade 0.84** 1.03*** 0.86 −0.00 0.19 0.00 (2.52) (3.08) (-0.01) (1.02) D. Bond momentum of CDS-matched firms and bond fragmentation . . Low fragmentation . High fragmentation . . PB5–PB1 . 6-factor α . Sharpe . PB5–PB1 . 6-factor α . Sharpe . Average bond return (% per month), K=1m J=3m , CDS-matched All ratings 0.32*** 0.34*** 0.65 0.12 0.16 0.31 (2.62) (2.76) (1.36) (1.28) Junk grade 0.84** 1.03*** 0.86 −0.00 0.19 0.00 (2.52) (3.08) (-0.01) (1.02) This table presents evidence on the relation between bond momentum and CDS momentum. Panel A reports bond momentum returns the full TRACE sample as well as the CDS-matched and non-CDS-matched subsamples. To construct the bond momentum strategy, we sort firms on past bond return into five equally sized portfolios in which portfolio PB1 (PB5) is the group with the lowest (highest) past bond return using a formation period J months. The long-short strategy purchases bonds in PB5 and sells short bonds in PB1, rebalancing at the end of each month ( K=1m ). To avoid the 1-month reversal in bonds, formation periods longer than 1 month skip the most recent month. Panel B examines future CDS returns of conditional bond-based and CDS-based performance quintiles. Conditional bond (CDS) quintiles are constructed by firms creating quintiles of past CDS (bond) returns and then creating conditional quintiles of past bond (CDS) return within the CDS (bond) return quintiles. For example, firms in the conditional CDS winner and loser portfolios are equally drawn from bond return quintiles. Panel C examines bond momentum before (“Baseline”) and after excluding firms with future credit rating changes. Panel D divides the CDS-matched subsample into low and high bond issue fragmentation subgroups and recomputes bond momentum. Fragmentation is based on a Herfindahl index on the number of bond issues and scaled by issue size. The six-factor α is based on a factor model with the DEF, TERM, MKT, SMB, HML, and UMD factors. Newey-West t-statistics (12 lags) are provided in parentheses. Notation for each portfolio return follows the pattern: mean, (t-statistic), and annualized Sharpe ratio. * p <.1; ** p <.05; *** p <.01. Open in new tab D. Bond momentum of CDS-matched firms and bond fragmentation . . Low fragmentation . High fragmentation . . PB5–PB1 . 6-factor α . Sharpe . PB5–PB1 . 6-factor α . Sharpe . Average bond return (% per month), K=1m J=3m , CDS-matched All ratings 0.32*** 0.34*** 0.65 0.12 0.16 0.31 (2.62) (2.76) (1.36) (1.28) Junk grade 0.84** 1.03*** 0.86 −0.00 0.19 0.00 (2.52) (3.08) (-0.01) (1.02) D. Bond momentum of CDS-matched firms and bond fragmentation . . Low fragmentation . High fragmentation . . PB5–PB1 . 6-factor α . Sharpe . PB5–PB1 . 6-factor α . Sharpe . Average bond return (% per month), K=1m J=3m , CDS-matched All ratings 0.32*** 0.34*** 0.65 0.12 0.16 0.31 (2.62) (2.76) (1.36) (1.28) Junk grade 0.84** 1.03*** 0.86 −0.00 0.19 0.00 (2.52) (3.08) (-0.01) (1.02) This table presents evidence on the relation between bond momentum and CDS momentum. Panel A reports bond momentum returns the full TRACE sample as well as the CDS-matched and non-CDS-matched subsamples. To construct the bond momentum strategy, we sort firms on past bond return into five equally sized portfolios in which portfolio PB1 (PB5) is the group with the lowest (highest) past bond return using a formation period J months. The long-short strategy purchases bonds in PB5 and sells short bonds in PB1, rebalancing at the end of each month ( K=1m ). To avoid the 1-month reversal in bonds, formation periods longer than 1 month skip the most recent month. Panel B examines future CDS returns of conditional bond-based and CDS-based performance quintiles. Conditional bond (CDS) quintiles are constructed by firms creating quintiles of past CDS (bond) returns and then creating conditional quintiles of past bond (CDS) return within the CDS (bond) return quintiles. For example, firms in the conditional CDS winner and loser portfolios are equally drawn from bond return quintiles. Panel C examines bond momentum before (“Baseline”) and after excluding firms with future credit rating changes. Panel D divides the CDS-matched subsample into low and high bond issue fragmentation subgroups and recomputes bond momentum. Fragmentation is based on a Herfindahl index on the number of bond issues and scaled by issue size. The six-factor α is based on a factor model with the DEF, TERM, MKT, SMB, HML, and UMD factors. Newey-West t-statistics (12 lags) are provided in parentheses. Notation for each portfolio return follows the pattern: mean, (t-statistic), and annualized Sharpe ratio. * p <.1; ** p <.05; *** p <.01. Open in new tab Next, we divide the full TRACE bond data set into CDS-matched and non-CDS-matched subgroups and reconstruct the bond momentum portfolios. In terms of both return magnitude and statistical significance, we find that bond momentum is much stronger in non-CDS-matched firms than CDS-matched firms. The strategy using a 3-month formation period generates 0.21% per month (t-statistic of 2.12) in CDS-matched firms and 0.64% per month (t-statistic of 3.62) in non-CDS-matched firms. The strategy’s Sharpe ratio for the non-CDS-matched sample is more than twice that of the CDS-matched sample (1.12 versus 0.54). This is similar to the finding of JNPS that bond momentum is significantly stronger for smaller/opaque firms, which is used to support their conclusion that the existence of bond momentum is likely a result of slow diffusion of information into bond prices. What is particularly striking is the relative weakness of bond momentum compared to CDS momentum using the same sample of firms (CDS-Matched). The J=3m CDS momentum strategy generates a robust 0.59% per month (see panel A of Table 2), while the equivalent bond momentum strategy yields a much weaker 0.21% per month (see panel A of Table 5).30 If CDS and bond momentum are both driven by slow diffusion of information, then the relative strength of CDS momentum might imply that (1) information diffusion occurs much more slowly for the CDS than the bond and (2) there exists significant information spillover from the bond to the CDS. However, we show that neither of these appears to be true. For one, unlike bond momentum, our findings on CDS momentum show greater profitability when more information is compounded into CDS prices regarding upcoming ratings changes (see Section 2.2). Furthermore, Figure 6 shows that CDS performance signals more strongly anticipate upcoming rating changes than those of the bond. One can see that the CDS winner (loser) portfolio has a much greater net exposure to future upgrades (downgrades) than the bond winner (loser) portfolio, which refutes the notion that the CDS experiences a relatively slower diffusion of information into prices. Fig. 6 Open in new tabDownload slide Average net exposure to future credit rating changes This figure displays the average net exposure to future credit rating changes in the CDS and bond momentum winner and loser portfolios. The average net exposure is computed as the percentage of firms in the portfolio that experience an upgrade during the 6 months following portfolio formation minus the percentage of firms that experience a downgrade. A positive value indicates that portfolio constituents experience more upgrades than downgrades. CDS (bond) momentum portfolios are quintiles of the past 3-month CDS (bond) return. CDS data are from Markit, and bond data are from TRACE. Time period spans 2003–2015. Fig. 6 Open in new tabDownload slide Average net exposure to future credit rating changes This figure displays the average net exposure to future credit rating changes in the CDS and bond momentum winner and loser portfolios. The average net exposure is computed as the percentage of firms in the portfolio that experience an upgrade during the 6 months following portfolio formation minus the percentage of firms that experience a downgrade. A positive value indicates that portfolio constituents experience more upgrades than downgrades. CDS (bond) momentum portfolios are quintiles of the past 3-month CDS (bond) return. CDS data are from Markit, and bond data are from TRACE. Time period spans 2003–2015. In additional tests, we find that past bond performance signals, after controlling for past CDS performance signals, do not serve as strong indicators of future CDS returns. We further find that CDS momentum remains strong even after controlling for past bond performance. For these tests, we examine CDS trading strategies using conditional portfolio sorts on the past performance of one market while controlling for past performance in the other. For instance, we perform conditional portfolio sorts on past bond performance (labeled MOMJbond | MOMJcds ) by first creating quintiles of past CDS performance and then creating conditional quintiles of past bond performance within the CDS-based quintiles. Under this setup, variation in CDS performance across the conditional bond performance quintiles is minimal and hence does not explain the results. As can be seen in panel B of Table 5, we find no evidence that the marginal information in past bond performance can be used to predict future CDS returns. Long-short CDS trading strategies based on conditional bond performance quintiles ( MOMJbond | MOMJcds ) generate statistically insignificant raw returns ranging from 0.00% to 0.09% using formation periods J=1m to J=12m . Furthermore, the J=3m conditional CDS momentum strategy that controls for past bond performance ( MOMJcds | MOMJbond ) still generates a raw return of 0.47% per month (t-statistic of 4.82), an alpha coefficient of 0.56% (t-statistic of 4.20), and a Sharpe ratio of 0.90. Recent research on the important economic role of the CDS market also helps guide the interpretation of these findings. The standardized nature of the CDS contract, as opposed to the fragmentation of individual issues in the bond market, leads to lower trading costs and enhanced liquidity. For these reasons, the CDS market is the preferred trading venue of speculative traders and investors with short-term liquidity needs (Oehmke and Zawadowski 2015, 2016). Indeed, Oehmke and Zawadowski (2016) show that lower trading costs drive speculative trading to take place in the CDS market. Similarly, Das, Kalimipalli, and Nayak (2014) find that bond liquidity and efficiency fall as institutional investors migrate to the CDS market. Because price discovery largely takes place in the CDS market, past performance signals of the CDS are more rich with information than those of the bond.31 Our results in panel B of Table 5 support this notion. Despite this documented difference in the information dynamics between bond momentum and CDS momentum, one might still expect the properties of bond momentum to exhibit a greater degree of similarity for CDS-matched firms than non-CDS-matched firms, because for CDS-matched firms, cross-market arbitrageurs could keep the two markets continually linked in their prices to a greater extent. We thus hypothesize that bond momentum profits for the CDS-matched group are more closely related to the CDS momentum in its underlying mechanism, that is, the future credit rating changes in the anticipated directions. In panel C of Table 5, we compare the performance of the J=3m bond momentum strategy (under the header “Baseline”) to the same strategy after excluding firms that undergo credit rating changes during the 6 months after portfolio formation (under the header “Excluding Future Rating Events”). We do this comparison for three different groups of firms: (1) the full TRACE universe, (2) CDS-matched firms, and (3) non-CDS-matched firms. Using the full TRACE sample of bonds, we find that bond momentum remains profitable and statistically significant at the 1% level even after removing firms that undergo future rating events (confirming the JNPS result). The long-short monthly raw return (Sharpe ratio) falls from 0.54% (1.03) to 0.45% (0.94). Importantly, we uncover a dramatic difference between the CDS-matched and non-CDS-matched groups in the importance of credit rating change events for bond momentum. Removing firms with upcoming ratings changes reduces the J = 3m bond momentum return by 43% ( =0.12/0.21−1 ) for CDS-matched firms, while it decreases by only 13% ( =0.56/0.64−1 ) for non-CDS-matched firms. We repeat the analysis for junk-grade firms and find similar tendencies: a return reduction of 77% ( =0.12/0.52−1 ) for CDS-matched firms and only 14% ( =0.97/1.13−1 ) for non-CDS-matched firms. The statistical significance of bond momentum for CDS-matched firms disappears after removing firms with upcoming credit rating changes. Impediments to arbitrage and security-specific valuation differences could exacerbate the disconnect between the CDS and the bond. Recent evidence on the CDS-bond basis, that is, the contemporaneous price differential between the cash bond and its CDS counterpart, shows that prices do in fact evolve differently in the two markets. Greater fragmentation of corporate bond issues, in particular, makes CDS-bond arbitrage more difficult (Oehmke and Zawadowski 2016). Similarities in the properties of bond and CDS momentum, therefore, could vary in the cross-section of the CDS-matched subgroup depending on firm-level conditions for the ease of arbitrage. Given the strong performance of CDS momentum, we hypothesize that bond momentum is likely to be also strong when conditions are most favorable for arbitrage. For CDS-matched firms, we explicitly examine this cross-sectional heterogeneity in bond momentum. We follow Oehmke and Zawadowski (2016) using the degree of fragmentation of a firm’s bond issues as a proxy for the impediments to arbitrage. Higher bond fragmentation inhibits arbitrage by impeding bond liquidity, contributing to higher round trip trading costs and lower turnover. Therefore, the greater is the bond fragmentation, the more difficult the bond is to arbitrage with the CDS. Following Oehmke and Zawadowski (2016), we calculate a monthly Herfindahl index of bond issues for every firm in the sample (adjusted for total bond issuance). One advantage of this measure, as discussed by Oehmke and Zawadowski (2016), is that it is determined relatively exogenously to the trading environment as compared to more direct measures of bond trading costs, which are usually highly endogenous. Panel D of Table 5 presents bond momentum profits for CDS-matched firms by this degree of bond issue fragmentation (above or below the month’s median value). We find that bond momentum is much stronger and only statistically significant for low fragmentation firms, where CDS-bond arbitrage conditions are most favorable. This group of firms exhibit bond momentum return of 0.32% per month with a t-statistic of 2.62. Similarly, among junk-grade firms, we find that bond momentum is profitable only in the low fragmentation group, generating 0.84% per month with a t-statistic of 2.52. By contrast, bond momentum generates zero profits in the high fragmentation the junk-grade group.32 This evidence supports the notion that when impediments to arbitrage are low, related markets are more integrated and return phenomena, such as return momentum, display greater similarity in their underlying momentum generating mechanisms. In summary, bond return momentum exists for CDS-matched firms for the anticipated ratings change channel, just like the same mechanism that drives the CDS momentum. However, given much weaker predictability of past bond returns on future rating changes in anticipated directions, the magnitude of bond return momentum is significantly lower than its CDS momentum counterpart. For the CDS-matched firms, we also find that the less disconnection between CDS and bond prices due in part to less impediment to arbitrage, the stronger bond momentum effect as CDS’s stronger predictability on future rating changes in anticipated directions could assist bond return momentum more effectively. 3.1 CDS momentum spillover to the bond market In this subsection, we test whether CDS performance signals spill over to the bond market. To do this, we start by examining the future bond returns of portfolios based on quintiles of past CDS performance. The long-short strategies in this case purchase the bonds of “past CDS winners” and sell short the bonds of “past CDS losers.” We present results based on univariate sorts on the past CDS performance signal (labeled MOMJcds ) as well as conditional portfolio sorts that aim to distinguish past CDS performance from past bond performance (labeled as MOMJcds | MOMJbond ). We examine formation periods ranging from J=1m to J=12m , using contiguous formation and holding periods and rebalancing at the end of each month. Panel A of Table 6 provides strong evidence of spillover from CDS to bonds. At all formation periods considered, the long-short strategies based on univariate portfolio sorts (labeled MOMJcds , shown in the top half of panel A) generate positive and statistically significant raw returns ranging from 0.36% to 0.53% per month (maximized at J=3m ). The spillover is robust to common risk factors, including the bond market DEF and TERM factors and the stock market MKT, SMB, HML, and UMD factors. The six-factor alpha coefficients range from 0.35% to 0.52% per month, all of which are statistically significant at the 1% level. Table 6 Momentum spillover from CDS to bonds A. Average bond returns within past CDS performance quintiles . . CDS performance quintiles . Long-short strategy . . P1 . P2 . P3 . P4 . P5 . P5–P1 . 6-factor α . Sharpe . Average bond return (% per month), K=1m Univariate portfolio sorts ( MOMJcds ) J=1m 0.25 0.41*** 0.50*** 0.61*** 0.71*** 0.46*** 0.43*** 1.42 (1.19) (3.30) (4.56) (4.27) (3.33) (2.98) (4.24) J=3m 0.22 0.44*** 0.47*** 0.61*** 0.75*** 0.53*** 0.52*** 1.62 (1.08) (3.75) (4.18) (4.25) (3.60) (4.76) (5.61) J=6m 0.28 0.48*** 0.50*** 0.54*** 0.70*** 0.42*** 0.37*** 1.14 (1.35) (4.28) (4.25) (3.96) (3.49) (4.10) (4.06) J=12m 0.28 0.43*** 0.44*** 0.52*** 0.64*** 0.36*** 0.35*** 0.89 (1.19) (3.27) (4.12) (4.62) (3.70) (2.66) (2.92) Conditional portfolio sorts ( MOMJcds | MOMJbond ) J=1m 0.19 0.42*** 0.52*** 0.62*** 0.76*** 0.57*** 0.54*** 2.11 (0.94) (3.29) (4.25) (4.40) (3.81) (4.11) (5.57) J=3m 0.18 0.40*** 0.50*** 0.65*** 0.77*** 0.59*** 0.61*** 2.14 (0.97) (3.08) (4.00) (4.45) (4.03) (5.41) (7.06) J=6m 0.23 0.42*** 0.53*** 0.58*** 0.74*** 0.51*** 0.52*** 1.63 (1.21) (3.38) (4.26) (4.20) (3.94) (4.92) (6.14) J=12m 0.23 0.45*** 0.47*** 0.53*** 0.66*** 0.44*** 0.48*** 1.36 (1.08) (3.15) (3.81) (4.48) (4.04) (3.50) (4.59) by CREDIT RATING ( J=3m ) Inv. grade 0.30** 0.46*** 0.47*** 0.53*** 0.59*** 0.29*** 0.28*** 1.35 (2.09) (4.29) (3.95) (4.59) (4.09) (3.61) (4.97) Junk 0.05 0.46** 0.67*** 0.73*** 0.99*** 0.94*** 1.02*** 2.06 (0.18) (2.32) (3.44) (3.59) (3.67) (5.98) (6.47) by CDS DEPTH ( J=3m ) High 0.15 0.43*** 0.49*** 0.62*** 0.79*** 0.64*** 0.69*** 2.18 (0.77) (3.37) (4.00) (4.53) (4.25) (5.84) (7.68) Low 0.20 0.37*** 0.64*** 0.67*** 0.79*** 0.60*** 0.62*** 1.41 (1.00) (2.71) (4.54) (4.06) (3.28) (3.81) (4.52) A. Average bond returns within past CDS performance quintiles . . CDS performance quintiles . Long-short strategy . . P1 . P2 . P3 . P4 . P5 . P5–P1 . 6-factor α . Sharpe . Average bond return (% per month), K=1m Univariate portfolio sorts ( MOMJcds ) J=1m 0.25 0.41*** 0.50*** 0.61*** 0.71*** 0.46*** 0.43*** 1.42 (1.19) (3.30) (4.56) (4.27) (3.33) (2.98) (4.24) J=3m 0.22 0.44*** 0.47*** 0.61*** 0.75*** 0.53*** 0.52*** 1.62 (1.08) (3.75) (4.18) (4.25) (3.60) (4.76) (5.61) J=6m 0.28 0.48*** 0.50*** 0.54*** 0.70*** 0.42*** 0.37*** 1.14 (1.35) (4.28) (4.25) (3.96) (3.49) (4.10) (4.06) J=12m 0.28 0.43*** 0.44*** 0.52*** 0.64*** 0.36*** 0.35*** 0.89 (1.19) (3.27) (4.12) (4.62) (3.70) (2.66) (2.92) Conditional portfolio sorts ( MOMJcds | MOMJbond ) J=1m 0.19 0.42*** 0.52*** 0.62*** 0.76*** 0.57*** 0.54*** 2.11 (0.94) (3.29) (4.25) (4.40) (3.81) (4.11) (5.57) J=3m 0.18 0.40*** 0.50*** 0.65*** 0.77*** 0.59*** 0.61*** 2.14 (0.97) (3.08) (4.00) (4.45) (4.03) (5.41) (7.06) J=6m 0.23 0.42*** 0.53*** 0.58*** 0.74*** 0.51*** 0.52*** 1.63 (1.21) (3.38) (4.26) (4.20) (3.94) (4.92) (6.14) J=12m 0.23 0.45*** 0.47*** 0.53*** 0.66*** 0.44*** 0.48*** 1.36 (1.08) (3.15) (3.81) (4.48) (4.04) (3.50) (4.59) by CREDIT RATING ( J=3m ) Inv. grade 0.30** 0.46*** 0.47*** 0.53*** 0.59*** 0.29*** 0.28*** 1.35 (2.09) (4.29) (3.95) (4.59) (4.09) (3.61) (4.97) Junk 0.05 0.46** 0.67*** 0.73*** 0.99*** 0.94*** 1.02*** 2.06 (0.18) (2.32) (3.44) (3.59) (3.67) (5.98) (6.47) by CDS DEPTH ( J=3m ) High 0.15 0.43*** 0.49*** 0.62*** 0.79*** 0.64*** 0.69*** 2.18 (0.77) (3.37) (4.00) (4.53) (4.25) (5.84) (7.68) Low 0.20 0.37*** 0.64*** 0.67*** 0.79*** 0.60*** 0.62*** 1.41 (1.00) (2.71) (4.54) (4.06) (3.28) (3.81) (4.52) Open in new tab Table 6 Momentum spillover from CDS to bonds A. Average bond returns within past CDS performance quintiles . . CDS performance quintiles . Long-short strategy . . P1 . P2 . P3 . P4 . P5 . P5–P1 . 6-factor α . Sharpe . Average bond return (% per month), K=1m Univariate portfolio sorts ( MOMJcds ) J=1m 0.25 0.41*** 0.50*** 0.61*** 0.71*** 0.46*** 0.43*** 1.42 (1.19) (3.30) (4.56) (4.27) (3.33) (2.98) (4.24) J=3m 0.22 0.44*** 0.47*** 0.61*** 0.75*** 0.53*** 0.52*** 1.62 (1.08) (3.75) (4.18) (4.25) (3.60) (4.76) (5.61) J=6m 0.28 0.48*** 0.50*** 0.54*** 0.70*** 0.42*** 0.37*** 1.14 (1.35) (4.28) (4.25) (3.96) (3.49) (4.10) (4.06) J=12m 0.28 0.43*** 0.44*** 0.52*** 0.64*** 0.36*** 0.35*** 0.89 (1.19) (3.27) (4.12) (4.62) (3.70) (2.66) (2.92) Conditional portfolio sorts ( MOMJcds | MOMJbond ) J=1m 0.19 0.42*** 0.52*** 0.62*** 0.76*** 0.57*** 0.54*** 2.11 (0.94) (3.29) (4.25) (4.40) (3.81) (4.11) (5.57) J=3m 0.18 0.40*** 0.50*** 0.65*** 0.77*** 0.59*** 0.61*** 2.14 (0.97) (3.08) (4.00) (4.45) (4.03) (5.41) (7.06) J=6m 0.23 0.42*** 0.53*** 0.58*** 0.74*** 0.51*** 0.52*** 1.63 (1.21) (3.38) (4.26) (4.20) (3.94) (4.92) (6.14) J=12m 0.23 0.45*** 0.47*** 0.53*** 0.66*** 0.44*** 0.48*** 1.36 (1.08) (3.15) (3.81) (4.48) (4.04) (3.50) (4.59) by CREDIT RATING ( J=3m ) Inv. grade 0.30** 0.46*** 0.47*** 0.53*** 0.59*** 0.29*** 0.28*** 1.35 (2.09) (4.29) (3.95) (4.59) (4.09) (3.61) (4.97) Junk 0.05 0.46** 0.67*** 0.73*** 0.99*** 0.94*** 1.02*** 2.06 (0.18) (2.32) (3.44) (3.59) (3.67) (5.98) (6.47) by CDS DEPTH ( J=3m ) High 0.15 0.43*** 0.49*** 0.62*** 0.79*** 0.64*** 0.69*** 2.18 (0.77) (3.37) (4.00) (4.53) (4.25) (5.84) (7.68) Low 0.20 0.37*** 0.64*** 0.67*** 0.79*** 0.60*** 0.62*** 1.41 (1.00) (2.71) (4.54) (4.06) (3.28) (3.81) (4.52) A. Average bond returns within past CDS performance quintiles . . CDS performance quintiles . Long-short strategy . . P1 . P2 . P3 . P4 . P5 . P5–P1 . 6-factor α . Sharpe . Average bond return (% per month), K=1m Univariate portfolio sorts ( MOMJcds ) J=1m 0.25 0.41*** 0.50*** 0.61*** 0.71*** 0.46*** 0.43*** 1.42 (1.19) (3.30) (4.56) (4.27) (3.33) (2.98) (4.24) J=3m 0.22 0.44*** 0.47*** 0.61*** 0.75*** 0.53*** 0.52*** 1.62 (1.08) (3.75) (4.18) (4.25) (3.60) (4.76) (5.61) J=6m 0.28 0.48*** 0.50*** 0.54*** 0.70*** 0.42*** 0.37*** 1.14 (1.35) (4.28) (4.25) (3.96) (3.49) (4.10) (4.06) J=12m 0.28 0.43*** 0.44*** 0.52*** 0.64*** 0.36*** 0.35*** 0.89 (1.19) (3.27) (4.12) (4.62) (3.70) (2.66) (2.92) Conditional portfolio sorts ( MOMJcds | MOMJbond ) J=1m 0.19 0.42*** 0.52*** 0.62*** 0.76*** 0.57*** 0.54*** 2.11 (0.94) (3.29) (4.25) (4.40) (3.81) (4.11) (5.57) J=3m 0.18 0.40*** 0.50*** 0.65*** 0.77*** 0.59*** 0.61*** 2.14 (0.97) (3.08) (4.00) (4.45) (4.03) (5.41) (7.06) J=6m 0.23 0.42*** 0.53*** 0.58*** 0.74*** 0.51*** 0.52*** 1.63 (1.21) (3.38) (4.26) (4.20) (3.94) (4.92) (6.14) J=12m 0.23 0.45*** 0.47*** 0.53*** 0.66*** 0.44*** 0.48*** 1.36 (1.08) (3.15) (3.81) (4.48) (4.04) (3.50) (4.59) by CREDIT RATING ( J=3m ) Inv. grade 0.30** 0.46*** 0.47*** 0.53*** 0.59*** 0.29*** 0.28*** 1.35 (2.09) (4.29) (3.95) (4.59) (4.09) (3.61) (4.97) Junk 0.05 0.46** 0.67*** 0.73*** 0.99*** 0.94*** 1.02*** 2.06 (0.18) (2.32) (3.44) (3.59) (3.67) (5.98) (6.47) by CDS DEPTH ( J=3m ) High 0.15 0.43*** 0.49*** 0.62*** 0.79*** 0.64*** 0.69*** 2.18 (0.77) (3.37) (4.00) (4.53) (4.25) (5.84) (7.68) Low 0.20 0.37*** 0.64*** 0.67*** 0.79*** 0.60*** 0.62*** 1.41 (1.00) (2.71) (4.54) (4.06) (3.28) (3.81) (4.52) Open in new tab B. CDS-to-bond momentum spillover through anticipated rating change events . . Conditional CDS quintiles . . . . ( MOMJ=3mcds | MOMJ=3mbond ) . Long-short strategy . . P1 (loser) . P5 (winner) . P5–P1 . 6-factor α . Frequency of future rating change events Net exposure Upgrades ( nup/N ) 3.6% 6.8% 3.1%*** Downgrades ( ndown/N ) 12.7% 6.9% −5.8%*** Average bond return within rating change subgroups Firms with Downgrades Upgrades anticipated events −0.56* 1.38*** 1.90*** 2.03*** (-1.76) (3.72) (5.53) (6.41) All other firms 0.34** 0.73*** 0.39*** 0.37*** (2.17) (3.99) (4.88) (4.98) B. CDS-to-bond momentum spillover through anticipated rating change events . . Conditional CDS quintiles . . . . ( MOMJ=3mcds | MOMJ=3mbond ) . Long-short strategy . . P1 (loser) . P5 (winner) . P5–P1 . 6-factor α . Frequency of future rating change events Net exposure Upgrades ( nup/N ) 3.6% 6.8% 3.1%*** Downgrades ( ndown/N ) 12.7% 6.9% −5.8%*** Average bond return within rating change subgroups Firms with Downgrades Upgrades anticipated events −0.56* 1.38*** 1.90*** 2.03*** (-1.76) (3.72) (5.53) (6.41) All other firms 0.34** 0.73*** 0.39*** 0.37*** (2.17) (3.99) (4.88) (4.98) Open in new tab B. CDS-to-bond momentum spillover through anticipated rating change events . . Conditional CDS quintiles . . . . ( MOMJ=3mcds | MOMJ=3mbond ) . Long-short strategy . . P1 (loser) . P5 (winner) . P5–P1 . 6-factor α . Frequency of future rating change events Net exposure Upgrades ( nup/N ) 3.6% 6.8% 3.1%*** Downgrades ( ndown/N ) 12.7% 6.9% −5.8%*** Average bond return within rating change subgroups Firms with Downgrades Upgrades anticipated events −0.56* 1.38*** 1.90*** 2.03*** (-1.76) (3.72) (5.53) (6.41) All other firms 0.34** 0.73*** 0.39*** 0.37*** (2.17) (3.99) (4.88) (4.98) B. CDS-to-bond momentum spillover through anticipated rating change events . . Conditional CDS quintiles . . . . ( MOMJ=3mcds | MOMJ=3mbond ) . Long-short strategy . . P1 (loser) . P5 (winner) . P5–P1 . 6-factor α . Frequency of future rating change events Net exposure Upgrades ( nup/N ) 3.6% 6.8% 3.1%*** Downgrades ( ndown/N ) 12.7% 6.9% −5.8%*** Average bond return within rating change subgroups Firms with Downgrades Upgrades anticipated events −0.56* 1.38*** 1.90*** 2.03*** (-1.76) (3.72) (5.53) (6.41) All other firms 0.34** 0.73*** 0.39*** 0.37*** (2.17) (3.99) (4.88) (4.98) Open in new tab C. CDS-to-bond momentum spillover beyond stock-to-bond momentum spillover . . Conditional CDS quintiles . . . . . ( MOMJcds | MOMJstock ) . Long-short strategy . . P1 . P2 . P3 . P4 . P5 . P5–P1 . 6-factor α . Sharpe . Average bond return (% per month), K=1m J=1m 0.35* 0.42*** 0.47*** 0.57*** 0.68*** 0.33*** 0.32*** 1.29 (1.85) (3.39) (4.07) (4.08) (3.46) (3.09) (4.37) J=3m 0.28 0.40*** 0.50*** 0.58*** 0.74*** 0.46*** 0.49*** 1.61 (1.53) (3.31) (4.01) (4.18) (3.62) (4.19) (5.51) J=6m 0.33* 0.47*** 0.50*** 0.51*** 0.69*** 0.36*** 0.34*** 1.17 (1.80) (3.91) (4.13) (3.61) (3.50) (4.46) (4.92) J=12m 0.40** 0.41*** 0.50*** 0.52*** 0.69*** 0.30*** 0.29*** 1.05 (2.09) (3.07) (3.97) (3.98) (3.95) (3.69) (3.78) C. CDS-to-bond momentum spillover beyond stock-to-bond momentum spillover . . Conditional CDS quintiles . . . . . ( MOMJcds | MOMJstock ) . Long-short strategy . . P1 . P2 . P3 . P4 . P5 . P5–P1 . 6-factor α . Sharpe . Average bond return (% per month), K=1m J=1m 0.35* 0.42*** 0.47*** 0.57*** 0.68*** 0.33*** 0.32*** 1.29 (1.85) (3.39) (4.07) (4.08) (3.46) (3.09) (4.37) J=3m 0.28 0.40*** 0.50*** 0.58*** 0.74*** 0.46*** 0.49*** 1.61 (1.53) (3.31) (4.01) (4.18) (3.62) (4.19) (5.51) J=6m 0.33* 0.47*** 0.50*** 0.51*** 0.69*** 0.36*** 0.34*** 1.17 (1.80) (3.91) (4.13) (3.61) (3.50) (4.46) (4.92) J=12m 0.40** 0.41*** 0.50*** 0.52*** 0.69*** 0.30*** 0.29*** 1.05 (2.09) (3.07) (3.97) (3.98) (3.95) (3.69) (3.78) This table reports future bond returns within quintiles based on past J-month CDS performance. In panel A, we report both simple univariate portfolio sorts on past CDS performance ( MOMJcds ) and conditional portfolio sorts on past CDS performance that control for past bond performance ( MOMJcds | MOMJbond ). The conditional portfolio sorts create CDS-based performance quintiles within bond-based performance quintiles, ensuring minimal variation in past bond performance across the CDS performance quintiles. Panel B reports the frequency of rating change events (labeled nup/N for upgrades and ndown/N for downgrades) within the winner and loser conditional CDS performance quintiles. Also reported are bond returns of those that do and do not undergo “anticipated” rating change events. Panel C reports future bond returns within conditional CDS performance quintiles that control for past stock performance. The six-factor α is based on a factor model with the DEF, TERM, MKT, SMB, HML, and UMD factors. CDS data are from Markit, and bond data are from TRACE. The sample period spans January 2003 to December 2015. Newey-West t-statistics (12 lags) are provided in parentheses. * p <.1; ** p <.05; *** p <.01. Open in new tab C. CDS-to-bond momentum spillover beyond stock-to-bond momentum spillover . . Conditional CDS quintiles . . . . . ( MOMJcds | MOMJstock ) . Long-short strategy . . P1 . P2 . P3 . P4 . P5 . P5–P1 . 6-factor α . Sharpe . Average bond return (% per month), K=1m J=1m 0.35* 0.42*** 0.47*** 0.57*** 0.68*** 0.33*** 0.32*** 1.29 (1.85) (3.39) (4.07) (4.08) (3.46) (3.09) (4.37) J=3m 0.28 0.40*** 0.50*** 0.58*** 0.74*** 0.46*** 0.49*** 1.61 (1.53) (3.31) (4.01) (4.18) (3.62) (4.19) (5.51) J=6m 0.33* 0.47*** 0.50*** 0.51*** 0.69*** 0.36*** 0.34*** 1.17 (1.80) (3.91) (4.13) (3.61) (3.50) (4.46) (4.92) J=12m 0.40** 0.41*** 0.50*** 0.52*** 0.69*** 0.30*** 0.29*** 1.05 (2.09) (3.07) (3.97) (3.98) (3.95) (3.69) (3.78) C. CDS-to-bond momentum spillover beyond stock-to-bond momentum spillover . . Conditional CDS quintiles . . . . . ( MOMJcds | MOMJstock ) . Long-short strategy . . P1 . P2 . P3 . P4 . P5 . P5–P1 . 6-factor α . Sharpe . Average bond return (% per month), K=1m J=1m 0.35* 0.42*** 0.47*** 0.57*** 0.68*** 0.33*** 0.32*** 1.29 (1.85) (3.39) (4.07) (4.08) (3.46) (3.09) (4.37) J=3m 0.28 0.40*** 0.50*** 0.58*** 0.74*** 0.46*** 0.49*** 1.61 (1.53) (3.31) (4.01) (4.18) (3.62) (4.19) (5.51) J=6m 0.33* 0.47*** 0.50*** 0.51*** 0.69*** 0.36*** 0.34*** 1.17 (1.80) (3.91) (4.13) (3.61) (3.50) (4.46) (4.92) J=12m 0.40** 0.41*** 0.50*** 0.52*** 0.69*** 0.30*** 0.29*** 1.05 (2.09) (3.07) (3.97) (3.98) (3.95) (3.69) (3.78) This table reports future bond returns within quintiles based on past J-month CDS performance. In panel A, we report both simple univariate portfolio sorts on past CDS performance ( MOMJcds ) and conditional portfolio sorts on past CDS performance that control for past bond performance ( MOMJcds | MOMJbond ). The conditional portfolio sorts create CDS-based performance quintiles within bond-based performance quintiles, ensuring minimal variation in past bond performance across the CDS performance quintiles. Panel B reports the frequency of rating change events (labeled nup/N for upgrades and ndown/N for downgrades) within the winner and loser conditional CDS performance quintiles. Also reported are bond returns of those that do and do not undergo “anticipated” rating change events. Panel C reports future bond returns within conditional CDS performance quintiles that control for past stock performance. The six-factor α is based on a factor model with the DEF, TERM, MKT, SMB, HML, and UMD factors. CDS data are from Markit, and bond data are from TRACE. The sample period spans January 2003 to December 2015. Newey-West t-statistics (12 lags) are provided in parentheses. * p <.1; ** p <.05; *** p <.01. Open in new tab Next, to further distinguish past CDS performance signals from those in the bond market, we condition CDS-to-bond momentum spillover on past bond performance (labeled MOMJcds | MOMJbond , shown in the bottom half of panel A). To do this, we first create quintiles of past bond performance, and then create conditional quintiles of past CDS performance within the bond-based quintiles. CDS winners and losers are drawn equally from all bond-based quintiles, ensuring that past bond performance will be roughly the same across the conditional CDS performance quintiles. For all formation periods considered, we find that the long-short return (P5-P1) is larger and more statistically significant after conditioning on past bond performance, for example, the J=3m CDS-to-bond momentum spillover strategy’s monthly raw return improves from 0.53% (t-statistic of 4.76) to 0.59% (t-statistic of 5.41) and the annualized Sharpe ratio improves from 1.62 to 2.14. Again, the CDS-to-bond spillover is robust to common stock and bond factors, generating monthly alpha coefficients ranging from 0.48% to 0.61% that are all statistically significant at the 1% level.33Figure 7 provides a visualization of the conditional CDS-to-bond spillover strategy. Panel A of Figure 7 shows that a $100 investment in the long-short conditional CDS-to-bond momentum spillover strategy grows to nearly $250 from 2003 to 2015, and panel B of Figure 7 shows that, like CDS momentum, its performance magnitude is maximized during periods of heightened market-wide distress risk. Fig. 7 Open in new tabDownload slide CDS momentum spillover to the bond market This figure highlights CDS-to-bond momentum spillover. Panel A shows the growth of a $100 investment in a J=3m conditional CDS-to-bond long-short momentum spillover strategy. Panel B divides our monthly time series into three equally sized groups based on the market-wide value-weighted average CDS spread at the beginning of the holding period and displays the average monthly return for the conditional CDS-to-bond momentum spillover strategy. The conditional CDS-to-bond momentum spillover strategy is based on CDS-based performance quintiles formed within bond-based performance quintiles and buys (sells short) bonds of firms in the highest (lowest) quintile. CDS data are from Markit, and bond data are from TRACE. The time period spans January 2003 to December 2015. Fig. 7 Open in new tabDownload slide CDS momentum spillover to the bond market This figure highlights CDS-to-bond momentum spillover. Panel A shows the growth of a $100 investment in a J=3m conditional CDS-to-bond long-short momentum spillover strategy. Panel B divides our monthly time series into three equally sized groups based on the market-wide value-weighted average CDS spread at the beginning of the holding period and displays the average monthly return for the conditional CDS-to-bond momentum spillover strategy. The conditional CDS-to-bond momentum spillover strategy is based on CDS-based performance quintiles formed within bond-based performance quintiles and buys (sells short) bonds of firms in the highest (lowest) quintile. CDS data are from Markit, and bond data are from TRACE. The time period spans January 2003 to December 2015. To further shed light on CDS-to-bond spillover effects, we analyze cross-sectional trends and divide our sample of firms into investment/junk grade groups and recompute the J=3m conditional CDS-to-bond momentum returns (shown at the bottom of panel A). We find that the spillover effect is greatest among junk-grade firms in terms of both the magnitude of long-short returns and statistical significance. The long-short return is 0.94% (t-statistic of 5.98 and Sharpe ratio of 2.06) for junk-grade firms versus 0.29% (t-statistic of 3.61 and Sharpe ratio of 1.35) for investment-grade firms. We additionally find stronger CDS-to-bond spillover when CDS depth is high. We divide our sample by high/low CDS contract depth and recompute the J=3m conditional CDS-to-bond momentum returns (shown at the bottom of panel A). The long-short strategy’s raw (annualized Sharpe ratio) is 0.64% per month (2.18) for high-depth firms while it is only 0.60% per month (1.41) for low-depth firms. This is consistent with the notion that more information is generated in the CDS market when CDS depth is high. Would anticipated future credit rating changes again play a primary role in the spillover from CDS to bonds? With the past CDS performance acting as a guide, the long-short strategy is positioned to have an ideal exposure to future credit rating change events—a net positive exposure to upgrades and a net negative exposure to downgrades—and benefits from the realized price divergence of firms that undergo an anticipated rating change. For this exercise, to focus on the marginal value of the CDS performance signal over that of the bond, we use the conditional CDS momentum strategy and choose formation period J=3m as it yields the strongest result in panel A. In panel B of Table 6, we first examine the frequency of rating change events within each of the winner and loser portfolios and confirm that the long-short portfolio does in fact have positive exposure to upgrades and negative exposure to downgrades. The conditional CDS winners experience 3.1% more upgrades and 5.8% fewer downgrades relative to conditional CDS losers. Secondly, within the winner and loser portfolios, it is the firms that experience rating changes in the anticipated directions that largely drive the long-short returns (generating 1.90% per month with a t-statistic of 5.53). By contrast, all other firms collectively exhibit weaker spillover effects but still generate a statistically significant 0.39% per month. Both JNPS and Gebhardt, Hvidkjaer, and Swaminathan (2005) document momentum spillover from the stock market to the bond market. Are our CDS-to-bond spillover results subsumed by the stock-to-bond spillover found in these previous studies? To address this, we perform conditional portfolio sorts on past CDS performance that control for past stock performance (labeled MOMJcds | MOMJstock ). We find that conditioning on past stock performance (shown in panel C of Table 6) produces very similar results to the univariate CDS-to-bond spillover results (shown in the top half of panel A). At formation horizon J=3m , we find that the overall long-short return falls from 0.53% to 0.46% per month and the annualized Sharpe ratio decreases from 1.62 to 1.61. Because we observe only a marginal deterioration in performance, we conclude that the vast majority of information in past CDS performance relevant for future bond returns is not captured in past stock performance. This inspires an additional question: to what extent does momentum spill over from the CDS market to the stock market? 3.2 CDS momentum spillover to the stock market In this subsection, we explore momentum spillover from CDS to stocks by analyzing the future stock returns within past CDS performance quintiles. Given that we trade stocks instead of CDS contracts, transaction cost concerns in this momentum spillover strategy are minimal (Frazzini, Israel, and Moskowitz, 2012).34 Table 7 presents the results. Each stock trading strategy weights portfolio stocks by market capitalization. For the univariate portfolio sorts on past CDS performance (labeled MOMJcds ), we find positive and statistically significant long-short stock returns of 0.52% and 0.46% per month for formation periods J=1m and J=3m , respectively. Importantly, controlling for common stock and bond market factors, namely, MKT, SMB, HML, UMD, DEF, and TERM, leads to improvements in performance for all CDS-to-stock momentum spillover strategies considered, generating statistically significant positive six-factor alpha coefficients ranging from 0.50% to 0.63% per month. It should be noted that this strong risk-adjusted performance accounts for exposure to traditional stock momentum, suggesting that past CDS performance signals provide value above and beyond past stock performance signals. Table 7 Momentum spillover from CDS to stocks A. Average stock returns within past CDS performance quintiles . . CDS performance quintiles . Long-short strategy . . P1 . P2 . P3 . P4 . P5 . P5–P1 . 6-factor α . Sharpe . Average stock return (% per month), K=1m J=1m 0.45 0.74* 0.90*** 0.96*** 0.96* 0.52* 0.63*** 0.50 (0.81) (1.88) (3.02) (2.95) (1.73) (1.89) (2.66) J=3m 0.50 0.54 0.87** 0.99*** 0.95* 0.46** 0.52** 0.46 (0.86) (1.38) (2.48) (3.06) (1.95) (2.01) (2.19) J=6m 0.54 0.76* 0.84** 0.96*** 1.07** 0.54 0.51** 0.45 (0.89) (1.82) (2.37) (2.84) (2.50) (1.54) (2.03) J=12m 0.40 0.72 0.81** 0.91*** 1.02*** 0.62 0.50** 0.48 (0.58) (1.60) (2.15) (2.78) (2.61) (1.51) (2.01) Conditional portfolio sorts ( MOMJcds | MOMJstock ) J=1m 0.60 0.72** 0.78** 1.04*** 1.00* 0.39 0.41 0.49 (1.27) (2.07) (2.43) (2.88) (1.80) (1.38) (1.53) J=3m 0.46 0.69* 0.93** 1.00*** 0.95* 0.49** 0.54** 0.61 (0.90) (1.86) (2.56) (3.39) (1.77) (2.21) (2.57) J=6m 0.41 0.82** 0.83** 0.84** 1.20*** 0.79*** 0.86*** 1.00 (0.79) (2.38) (2.28) (2.47) (2.81) (3.16) (3.45) J=12m 0.52 0.72* 0.87** 0.87** 1.07*** 0.55** 0.60** 0.74 (1.07) (1.74) (2.48) (2.40) (3.02) (2.34) (2.53) by CREDIT RATING ( J=6m ) Inv. grade 0.36 0.79** 0.81** 1.02*** 0.78 0.41* 0.48** 0.49 (0.79) (2.09) (2.14) (3.48) (1.64) (1.78) (2.02) Junk 0.64 0.89 0.93* 1.18** 1.42** 0.78*** 0.81*** 0.62 (1.00) (1.55) (1.67) (1.98) (2.51) (3.42) (3.57) by CDS DEPTH ( J=6m ) High 0.34 0.80** 0.81** 0.85** 1.11*** 0.77*** 0.87*** 0.89 (0.67) (2.21) (2.37) (2.37) (2.78) (3.25) (3.76) Low 1.28*** 0.87 1.03** 0.82 1.09** −0.15 −0.14 −0.10 (3.26) (1.46) (2.39) (1.21) (2.26) (-0.43) (-0.37) A. Average stock returns within past CDS performance quintiles . . CDS performance quintiles . Long-short strategy . . P1 . P2 . P3 . P4 . P5 . P5–P1 . 6-factor α . Sharpe . Average stock return (% per month), K=1m J=1m 0.45 0.74* 0.90*** 0.96*** 0.96* 0.52* 0.63*** 0.50 (0.81) (1.88) (3.02) (2.95) (1.73) (1.89) (2.66) J=3m 0.50 0.54 0.87** 0.99*** 0.95* 0.46** 0.52** 0.46 (0.86) (1.38) (2.48) (3.06) (1.95) (2.01) (2.19) J=6m 0.54 0.76* 0.84** 0.96*** 1.07** 0.54 0.51** 0.45 (0.89) (1.82) (2.37) (2.84) (2.50) (1.54) (2.03) J=12m 0.40 0.72 0.81** 0.91*** 1.02*** 0.62 0.50** 0.48 (0.58) (1.60) (2.15) (2.78) (2.61) (1.51) (2.01) Conditional portfolio sorts ( MOMJcds | MOMJstock ) J=1m 0.60 0.72** 0.78** 1.04*** 1.00* 0.39 0.41 0.49 (1.27) (2.07) (2.43) (2.88) (1.80) (1.38) (1.53) J=3m 0.46 0.69* 0.93** 1.00*** 0.95* 0.49** 0.54** 0.61 (0.90) (1.86) (2.56) (3.39) (1.77) (2.21) (2.57) J=6m 0.41 0.82** 0.83** 0.84** 1.20*** 0.79*** 0.86*** 1.00 (0.79) (2.38) (2.28) (2.47) (2.81) (3.16) (3.45) J=12m 0.52 0.72* 0.87** 0.87** 1.07*** 0.55** 0.60** 0.74 (1.07) (1.74) (2.48) (2.40) (3.02) (2.34) (2.53) by CREDIT RATING ( J=6m ) Inv. grade 0.36 0.79** 0.81** 1.02*** 0.78 0.41* 0.48** 0.49 (0.79) (2.09) (2.14) (3.48) (1.64) (1.78) (2.02) Junk 0.64 0.89 0.93* 1.18** 1.42** 0.78*** 0.81*** 0.62 (1.00) (1.55) (1.67) (1.98) (2.51) (3.42) (3.57) by CDS DEPTH ( J=6m ) High 0.34 0.80** 0.81** 0.85** 1.11*** 0.77*** 0.87*** 0.89 (0.67) (2.21) (2.37) (2.37) (2.78) (3.25) (3.76) Low 1.28*** 0.87 1.03** 0.82 1.09** −0.15 −0.14 −0.10 (3.26) (1.46) (2.39) (1.21) (2.26) (-0.43) (-0.37) Open in new tab Table 7 Momentum spillover from CDS to stocks A. Average stock returns within past CDS performance quintiles . . CDS performance quintiles . Long-short strategy . . P1 . P2 . P3 . P4 . P5 . P5–P1 . 6-factor α . Sharpe . Average stock return (% per month), K=1m J=1m 0.45 0.74* 0.90*** 0.96*** 0.96* 0.52* 0.63*** 0.50 (0.81) (1.88) (3.02) (2.95) (1.73) (1.89) (2.66) J=3m 0.50 0.54 0.87** 0.99*** 0.95* 0.46** 0.52** 0.46 (0.86) (1.38) (2.48) (3.06) (1.95) (2.01) (2.19) J=6m 0.54 0.76* 0.84** 0.96*** 1.07** 0.54 0.51** 0.45 (0.89) (1.82) (2.37) (2.84) (2.50) (1.54) (2.03) J=12m 0.40 0.72 0.81** 0.91*** 1.02*** 0.62 0.50** 0.48 (0.58) (1.60) (2.15) (2.78) (2.61) (1.51) (2.01) Conditional portfolio sorts ( MOMJcds | MOMJstock ) J=1m 0.60 0.72** 0.78** 1.04*** 1.00* 0.39 0.41 0.49 (1.27) (2.07) (2.43) (2.88) (1.80) (1.38) (1.53) J=3m 0.46 0.69* 0.93** 1.00*** 0.95* 0.49** 0.54** 0.61 (0.90) (1.86) (2.56) (3.39) (1.77) (2.21) (2.57) J=6m 0.41 0.82** 0.83** 0.84** 1.20*** 0.79*** 0.86*** 1.00 (0.79) (2.38) (2.28) (2.47) (2.81) (3.16) (3.45) J=12m 0.52 0.72* 0.87** 0.87** 1.07*** 0.55** 0.60** 0.74 (1.07) (1.74) (2.48) (2.40) (3.02) (2.34) (2.53) by CREDIT RATING ( J=6m ) Inv. grade 0.36 0.79** 0.81** 1.02*** 0.78 0.41* 0.48** 0.49 (0.79) (2.09) (2.14) (3.48) (1.64) (1.78) (2.02) Junk 0.64 0.89 0.93* 1.18** 1.42** 0.78*** 0.81*** 0.62 (1.00) (1.55) (1.67) (1.98) (2.51) (3.42) (3.57) by CDS DEPTH ( J=6m ) High 0.34 0.80** 0.81** 0.85** 1.11*** 0.77*** 0.87*** 0.89 (0.67) (2.21) (2.37) (2.37) (2.78) (3.25) (3.76) Low 1.28*** 0.87 1.03** 0.82 1.09** −0.15 −0.14 −0.10 (3.26) (1.46) (2.39) (1.21) (2.26) (-0.43) (-0.37) A. Average stock returns within past CDS performance quintiles . . CDS performance quintiles . Long-short strategy . . P1 . P2 . P3 . P4 . P5 . P5–P1 . 6-factor α . Sharpe . Average stock return (% per month), K=1m J=1m 0.45 0.74* 0.90*** 0.96*** 0.96* 0.52* 0.63*** 0.50 (0.81) (1.88) (3.02) (2.95) (1.73) (1.89) (2.66) J=3m 0.50 0.54 0.87** 0.99*** 0.95* 0.46** 0.52** 0.46 (0.86) (1.38) (2.48) (3.06) (1.95) (2.01) (2.19) J=6m 0.54 0.76* 0.84** 0.96*** 1.07** 0.54 0.51** 0.45 (0.89) (1.82) (2.37) (2.84) (2.50) (1.54) (2.03) J=12m 0.40 0.72 0.81** 0.91*** 1.02*** 0.62 0.50** 0.48 (0.58) (1.60) (2.15) (2.78) (2.61) (1.51) (2.01) Conditional portfolio sorts ( MOMJcds | MOMJstock ) J=1m 0.60 0.72** 0.78** 1.04*** 1.00* 0.39 0.41 0.49 (1.27) (2.07) (2.43) (2.88) (1.80) (1.38) (1.53) J=3m 0.46 0.69* 0.93** 1.00*** 0.95* 0.49** 0.54** 0.61 (0.90) (1.86) (2.56) (3.39) (1.77) (2.21) (2.57) J=6m 0.41 0.82** 0.83** 0.84** 1.20*** 0.79*** 0.86*** 1.00 (0.79) (2.38) (2.28) (2.47) (2.81) (3.16) (3.45) J=12m 0.52 0.72* 0.87** 0.87** 1.07*** 0.55** 0.60** 0.74 (1.07) (1.74) (2.48) (2.40) (3.02) (2.34) (2.53) by CREDIT RATING ( J=6m ) Inv. grade 0.36 0.79** 0.81** 1.02*** 0.78 0.41* 0.48** 0.49 (0.79) (2.09) (2.14) (3.48) (1.64) (1.78) (2.02) Junk 0.64 0.89 0.93* 1.18** 1.42** 0.78*** 0.81*** 0.62 (1.00) (1.55) (1.67) (1.98) (2.51) (3.42) (3.57) by CDS DEPTH ( J=6m ) High 0.34 0.80** 0.81** 0.85** 1.11*** 0.77*** 0.87*** 0.89 (0.67) (2.21) (2.37) (2.37) (2.78) (3.25) (3.76) Low 1.28*** 0.87 1.03** 0.82 1.09** −0.15 −0.14 −0.10 (3.26) (1.46) (2.39) (1.21) (2.26) (-0.43) (-0.37) Open in new tab B. CDS-to-stock momentum spillover through anticipated rating change events . . Conditional CDS quintiles . . . . ( MOMJ=6mcds | MOMJ=6mstock ) . Long-short strategy . . P1 (loser) . P5 (winner) . P5–P1 . 6-factor α . Frequency of future rating change events Net exposure Upgrades ( nup/N ) 4.0% 7.6% 3.6%*** Downgrades ( ndown/N ) 13.1% 6.2% −7.0%*** Average stock return within rating change subgroups Firms with Downgrades Upgrades anticipated events −1.38* 2.72*** 4.15*** 4.10*** (-1.88) (4.00) (9.68) (12.02) All other firms 0.63 1.07** 0.43* 0.47** (1.29) (2.53) (1.88) (1.99) B. CDS-to-stock momentum spillover through anticipated rating change events . . Conditional CDS quintiles . . . . ( MOMJ=6mcds | MOMJ=6mstock ) . Long-short strategy . . P1 (loser) . P5 (winner) . P5–P1 . 6-factor α . Frequency of future rating change events Net exposure Upgrades ( nup/N ) 4.0% 7.6% 3.6%*** Downgrades ( ndown/N ) 13.1% 6.2% −7.0%*** Average stock return within rating change subgroups Firms with Downgrades Upgrades anticipated events −1.38* 2.72*** 4.15*** 4.10*** (-1.88) (4.00) (9.68) (12.02) All other firms 0.63 1.07** 0.43* 0.47** (1.29) (2.53) (1.88) (1.99) This table reports future stock returns within quintiles based on past J-month CDS performance. In panel A, we report both simple univariate portfolio sorts on past CDS performance ( MOMJcds ) and conditional portfolio sorts on past CDS performance that control for past stock performance ( MOMJcds | MOMJstock ). The conditional portfolio sorts create CDS-based performance quintiles within stock-based performance quintiles, ensuring minimal variation in past stock performance across the CDS performance quintiles. Panel B reports the frequency of rating change events (labeled nup/N for upgrades and ndown/N for downgrades) within the winner and loser conditional CDS performance quintiles. Also reported are stock returns of those that do and do not undergo “anticipated” rating change events. The six-factor α is based on a factor model using bond market factors DEF and TERM and stock market factors MKT, SMB, HML, and UMD. CDS data are from Markit, and stock data are from CRSP. The sample period spans January 2003 to December 2015. Newey-West t-statistics (12 lags) are provided in parentheses. * p <.1; ** p <.05; *** p <.01. Open in new tab B. CDS-to-stock momentum spillover through anticipated rating change events . . Conditional CDS quintiles . . . . ( MOMJ=6mcds | MOMJ=6mstock ) . Long-short strategy . . P1 (loser) . P5 (winner) . P5–P1 . 6-factor α . Frequency of future rating change events Net exposure Upgrades ( nup/N ) 4.0% 7.6% 3.6%*** Downgrades ( ndown/N ) 13.1% 6.2% −7.0%*** Average stock return within rating change subgroups Firms with Downgrades Upgrades anticipated events −1.38* 2.72*** 4.15*** 4.10*** (-1.88) (4.00) (9.68) (12.02) All other firms 0.63 1.07** 0.43* 0.47** (1.29) (2.53) (1.88) (1.99) B. CDS-to-stock momentum spillover through anticipated rating change events . . Conditional CDS quintiles . . . . ( MOMJ=6mcds | MOMJ=6mstock ) . Long-short strategy . . P1 (loser) . P5 (winner) . P5–P1 . 6-factor α . Frequency of future rating change events Net exposure Upgrades ( nup/N ) 4.0% 7.6% 3.6%*** Downgrades ( ndown/N ) 13.1% 6.2% −7.0%*** Average stock return within rating change subgroups Firms with Downgrades Upgrades anticipated events −1.38* 2.72*** 4.15*** 4.10*** (-1.88) (4.00) (9.68) (12.02) All other firms 0.63 1.07** 0.43* 0.47** (1.29) (2.53) (1.88) (1.99) This table reports future stock returns within quintiles based on past J-month CDS performance. In panel A, we report both simple univariate portfolio sorts on past CDS performance ( MOMJcds ) and conditional portfolio sorts on past CDS performance that control for past stock performance ( MOMJcds | MOMJstock ). The conditional portfolio sorts create CDS-based performance quintiles within stock-based performance quintiles, ensuring minimal variation in past stock performance across the CDS performance quintiles. Panel B reports the frequency of rating change events (labeled nup/N for upgrades and ndown/N for downgrades) within the winner and loser conditional CDS performance quintiles. Also reported are stock returns of those that do and do not undergo “anticipated” rating change events. The six-factor α is based on a factor model using bond market factors DEF and TERM and stock market factors MKT, SMB, HML, and UMD. CDS data are from Markit, and stock data are from CRSP. The sample period spans January 2003 to December 2015. Newey-West t-statistics (12 lags) are provided in parentheses. * p <.1; ** p <.05; *** p <.01. Open in new tab To address correlation in the cross-sectional ranking of stock and CDS returns, we examine the performance of conditional sorts that isolate past CDS performance from past stock performance (labeled MOMJcds | MOMJstock ). For formation horizons ranging from J=3m to J=12m , we find that the conditional strategy performs significantly better than the equivalent univariate strategy. We observe a dramatic improvement in raw performance, risk-adjusted performance, and Sharpe ratio of CDS-to-stock momentum spillover strategies after purging past stock performance information. Performance is maximized at J=6m , which generates a raw stock return of 0.79% per month (t-statistic of 3.16), a four-factor alpha coefficient of 0.86% (t-statistic of 3.45), and an annualized Sharpe ratio of 1.00. These results clearly point to past CDS performance being able to reveal important information about future stock returns.35 Importantly, the conditional CDS-to-stock spillover does not exhibit the same severe drawdowns as traditional stock momentum during bear market rebounds (Barroso and Santa-Clara 2015; Daniel, Jagannathan, and Kim 2019; Daniel and Moskowitz 2016). In the 3 months spanning March 2009 to May 2009, when the Fama-French UMD factor experienced a cumulative return of -49.3%, the cumulative return of the conditional CDS-to-stock strategy was +6.0%. This is clear in Figure 8, which shows that the conditional CDS-to-stock momentum spillover strategy grows to roughly $325 (panel A), avoiding the crash risk of traditional stock momentum and averaging its highest returns during bear markets (panel B). Fig. 8 Open in new tabDownload slide CDS momentum spillover to the stock market This figure highlights CDS-to-stock momentum spillover. Panel A shows the growth of a $100 investment in a J=6m conditional CDS-to-stock long-short momentum spillover strategy. Panel B divides our monthly time series into three equally sized groups based on the cumulative 2-year stock market return at the beginning of the holding period and displays the average monthly return for the conditional CDS-to-stock momentum spillover strategy. The conditional CDS-to-stock momentum spillover strategy is based on CDS-based performance quintiles formed within stock-based performance quintiles and buys (sells short) the stock of firms in the highest (lowest) quintile. CDS data are from Markit, and stock data are from CRSP. The time period spans January 2003 to December 2015. Fig. 8 Open in new tabDownload slide CDS momentum spillover to the stock market This figure highlights CDS-to-stock momentum spillover. Panel A shows the growth of a $100 investment in a J=6m conditional CDS-to-stock long-short momentum spillover strategy. Panel B divides our monthly time series into three equally sized groups based on the cumulative 2-year stock market return at the beginning of the holding period and displays the average monthly return for the conditional CDS-to-stock momentum spillover strategy. The conditional CDS-to-stock momentum spillover strategy is based on CDS-based performance quintiles formed within stock-based performance quintiles and buys (sells short) the stock of firms in the highest (lowest) quintile. CDS data are from Markit, and stock data are from CRSP. The time period spans January 2003 to December 2015. Next, continuing in panel A of Table 7, we divide the sample into investment-grade/junk-grade subgroups and recompute the conditional CDS-to-stock momentum returns using the best-performing conditional CDS-to-stock momentum strategy with formation period J=6m . We find spillover from CDS to stocks for both investment- and junk-grade groups with long-short returns of 0.41% (t-statistic of 1.78) and 0.78% (t-statistic of 3.42), respectively. Our finding that CDS momentum spillover is slightly stronger for the junk-grade group is consistent with the finding of Acharya and Johnson (2007), who document that the CDS market reveals information earlier than the stock market for poorly rated firms. Lastly, in panel A of Table 7, we divide our sample of firms into high and low CDS depth groups and recompute the conditional CDS momentum spillover returns. Consistent with our earlier results for single-market CDS momentum and CDS-to-bond spillover, we find strong CDS-to-stock spillover for high-depth firms and no spillover for low-depth firms. The high depth group performs significantly well, generating a long-short strategy of 0.77% per month (t-statistic of 3.25) and a four-factor alpha of 0.87% per month (t-statistic of 3.76). By contrast, the low depth group generates a statistically insignificant long-short return of -0.15% per month, providing more evidence that endogenous liquidity in the CDS market is associated with distinct information being generated in the CDS market (Qiu and Yu, 2012) and supporting the notion of segmentation with respect to information content in CDS and stock markets.36 Next, in panel B of Table 7 we explore the relation between CDS-to-stock momentum spillover and credit rating changes, focusing on the conditional CDS-to-stock spillover strategy using a formation period of J=6m (the best-performing strategy in panel A based on conditional CDS performance signals). We find that, even after purging past stock performance information, the CDS performance signal continues to provide direction on upcoming credit rating changes; the long-short strategy is ideally positioned to have a net positive exposure to rating upgrades (+3.6%) and a net negative exposure to rating downgrades (-7.0%). Firms that experience a rating change in the anticipated direction drive the CDS-to-stock spillover with a long-short return of 4.15% per month (t-statistic of 9.68) compared to the 0.43% per month (t-statistic of 1.88) generated by all other firms. Overall, Table 7 shows clear evidence of CDS momentum spillover to the stock market, which tends to be stronger for firms of lower credit quality and higher CDS depth. Investors that utilize past CDS performance information in their decision-making processes benefit from positioning their portfolios to have a favorable exposure to upcoming credit rating changes. 4. Conclusion We examine the extent to which momentum exists in the CDS market and whether it spills over to other closely related asset markets sharing a common fundamental link, namely, the bond and stock markets. Despite the CDS market’s relative information efficiency, we document both economically and statistically significant CDS return momentum profits. We further show that the underlying mechanisms driving CDS return momentum are distinct from those associated with corporate bond return momentum (Jostova et al. 2013) and highlight the role of CDS return information in predicting future credit rating changes in “anticipated” directions, with past CDS winners (losers) undergoing more frequent rating upgrades (downgrades) in subsequent periods. We also document the ability of a future rating change channel to drive return momentum in the CDS market, particularly for firms with poor credit ratings (Acharya and Johnson 2007) and/or high CDS contract depth (Qiu and Yu 2012). Importantly, we also document that CDS returns contain useful incremental information beyond the information in bond and stock returns. We show significant cross-market CDS-to-bond and CDS-to-stock momentum spillover effects, both of which are closely related to future rating changes in momentum portfolios that occur in “anticipated” directions. We also find these cross-market information dynamics are greater among entities with poor credit ratings and/or greater CDS market liquidity. In addition, our novel conditional CDS-to-stock momentum spillover strategy that isolates the CDS-market-specific performance component does not exhibit the well-documented crash during the recent crisis period (Barroso and Santa-Clara 2015; Daniel, Jagannathan, and Kim 2019; Daniel and Moskowitz 2016). This new stock trading strategy, which generates 0.79% per month with an annualized Sharpe ratio of 1.00, is unexplained by a broad range of stock and bond market risk factors. Overall, our results provide further evidence regarding a number of important empirical regularities documented in the recent momentum literature; in particular, momentum exists and is interrelated across asset classes and markets (Asness, Moskowitz, and Pedersen 2013). We uncover an important source underlying momentum in a single market and potential spillovers to other related securities: common corporate credit rating changes generate a CDS return momentum and its spillover to bonds and stocks of the same firm through more timely captured credit risk information in CDS than sluggish ratings responses. We provide evidence consistent with the existence of such a mechanism by focusing on closely related asset classes—CDS, bonds, and stocks—that are all linked through underlying firm fundamentals. At the same time, our study also suggests that bond and stock investors can substantially improve their investment returns by paying attention to the incremental information originating in highly liquid CDS contracts. Our results suggest that further investigating the efficacy of CDS market information on investment strategies or in the design of other financial claim contracts could be a useful avenue for future research. We thank Viral Acharya, Mark Flannery, Jason Karceski, Francis Longstaff, Stas Nikolova, Jay Ritter, Marco Rossi, Mike Ryngaert, and Marti Subrahmanyam for their helpful comments and suggestions. We are also grateful for helpful comments from our discussants and participants at the 2014 Financial Management Association Annual Meeting, the 2015 American Economic Association Annual Meeting, the 2017 Southern Finance Association Annual Meeting, and the 2017 Southwestern Finance Association Annual Meeting. Jongsub Lee gratefully acknowledges the financial support of the Institute of Finance and Banking, and the Institute of Management Research at Seoul National University. Footnotes 1 It is important to note that the rating agencies’ practice of weighing accuracy and stability (rating through the cycle) creates, in part, less informationally efficient and slower ratings adjustments. Ratings agencies argue that an increase in ratings accuracy by increasing their responsiveness to more timely market information has a trade-off of reducing ratings stability. They argue that this trade-off is also important to market participants given that rating changes have real consequences (through ratings-based portfolio governance rules and rating triggers) that are costly to reverse (see Moody’s, September 2006, “Analyzing the Tradeoff Between Ratings Accuracy and Stability”). 2 Jostova et al. (2013) attribute the existence of bond momentum to the slow dissemination of information into prices. 3 Our CDS return definition is based on the marking-to-market profit (or loss) from CDS trading. To clarify the notation, it should be noted that CDS returns (CDS spread change) are positive (negative) for rating upgrades (downgrades). 4 It is important to note that while past CDS returns could be very useful in predicting future rating changes that materially affect firm fundamentals, this may not produce CDS momentum if the market is fully efficient. Our results suggest that while CDS prices underreact to information about prospective rating changes, past CDS returns are an important proxy of such upcoming rating changes that are materially relevant, although sluggish. Past CDS returns better capture the predictive information compared with past bond and stock returns. We show that this temporal mechanism generates CDS momentum and its spillover to other related securities markets. 5 For more details, visit http://www.stacesirmans.com/markit-cds-big-bang-2009.pdf. 6 We further impose a CDS activity filter by eliminating CDS contracts with a significant amount of missing and stagnant observations. Specifically, we eliminate contracts that have missing spreads at least 10% of days or stagnant spreads at least 90% of days over the prior 6 months. 7 This comparison between the par asset swap package and the credit default swap is accurate for a relatively short holding period. For a long holding period, the value of the interest swap component embedded in the asset swap package could diverge from zero, making the comparison inexact. 8 As we will explain later, we use a flat hazard rate assumption, which essentially makes O’Kane’s (2008) pricing model an ISDA CDS Standard model used by Markit. This model is also called the JP Morgan model by practitioners. 9 Since 2003, at any moment in time, the most liquid T-year CDS contract is the one that matures on the first IMM date T years after the trade date. See O’Kane (2008, section 5.2.1) for market convention details on determining the maturity date of a standard CDS contract. 10 Here, we directly use the lower bound of the discretized integration. See O’Kane (2008, section 6.6) for more details about this bound discussion. 11 For a protection buyer, the mark-to-market value would be just the opposite, −V(t0) . 12 If our valuation falls between two consecutive coupon payment dates, the quoted spread will make the clean mark-to-market value zero. We need to adjust for accrued premium since the last coupon payment date when we compute the clean mark-to-market. 13 The flat hazard rate, h, is calibrated such that the clean mark-to-market of the 5-year CDS contract is zero for a given quoted spread, S(t′) , on a valuation date-t’. 14 Here is an example of our trading timeline: for the trade ending on March 19, 2013, we sold a 5-year protection at the beginning of February 20, 2013, with the at-market spread on that date and unwound this position by buying a protection at the end of March 19, 2013, with the at-market spread on this unwinding date. Clearly, we are selling and buying a 5-year CDS contract with the same maturity date of March 20, 2018, in and out of this trade. The realized P&L for this holding period is then marked on our trading book at midnight of March 19, 2013. 15 http://www.creditfixings.com/CreditEventAuctions/fixings.jsp is the web address of the Creditex Group. 16 For more details on this discussion, see the section entitled “Trading with a Full Coupon” on p. 17 in http://www.stacesirmans.com/markit-cds-big-bang-2009.pdf. 17 The annualized 90-day Treasury-bill rate during our sample period was 1.91% on average. 18 For instance, the 3-month holding period return is computed by equally averaging the 1-month returns of three strategies using momentum portfolios formed in the current month, 1 month prior, and 2 months prior. 19 Panel A of Internet Appendix I confirms that past CDS performance has predictive power for future CDS returns using Fama-Macbeth cross-sectional regressions. 20 A potential validity concern is the seemingly large premium for hedging against default risk. In particular, P1 is the short leg of the portfolios, so the momentum trader is a P1 protection buyer who earns a seemingly nontrivial 0.25% every month by hedging against default risk as shown in panel A of Table 2 for J=3m . We thank an anonymous referee for raising this point and providing some guidance in potentially address it. Importantly, this validity concern is alleviated by examining the temporal source of this effect and transaction costs. In particular, in the middle of panel A of Table 2, we observe that the negative returns of P1 are concentrated in the “Crisis Period” (an average of −0.69% is earned from January 2008 to December 2011). Importantly, much of this result is driven by a few select months in the sample period. The worst two monthly returns, for instance, are 5.5 and 4.9 standard deviations below the mean. Driving these negative returns were two systemic credit events in which a sharp rise in CDS spreads proved favorable to those who purchased default protections on CDS losers (September, October, and November around the Lehman Brothers’ collapse in 2008 and August and September 2011 at the core of the 2011 European sovereign credit risk crisis). When we exclude these months, we find that the average return on the CDS loser portfolio rises from -0.25% to +0.06% (t-statistic of 0.30). In other words, purchasing default protection on CDS losers is usually a costly endeavor but pays off during extreme crises. Transaction costs (see Internet Appendix F) further reduce the seemingly large premium for hedging against default risk. We find the average estimated portfolio transaction cost for P1 to be 17.3 bps over our sample period. We replicate Bongaerts, De Jong, and Driessen (2011) for their sample period of April 2004 to June 2008 and then extend the data to our full sample period to obtain the new transaction cost of 17.3 bps. Internet Appendix F reports more details of the transaction cost estimates. 21 Monthly factor returns in Table 2 are compiled from daily factor returns to match the specific month-to-month CDS return timing (see Section 1.1.2). 22 http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html is the address of Kenneth French’s webpage. Internet Appendix B provides further details about our factor definitions. 23 In Internet Appendix E, we provide factor model results for individual CDS momentum portfolios P1 to P5. 24 To accommodate the nonlinear hump-shaped relationship between rating and depth (Qiu and Yu 2012), we add a squared rating variable to each depth regression (and vice versa). For example, orthogonalized depth is computed as the residual from a regression of depth on size, rating, and rating squared. 25 Recall that our sample consists of firms that are relatively large. The smallest size tercile still has an average equity market capitalization of $2.5 billion. Therefore, the insignificant size effect is somewhat expected from this sample property. 26 We retrieve the CDS bid-ask spread information from Datastream and Capital IQ since Markit does not provide microstructure measures, except for CDS contract depth. The CDS bid-ask spread information covers the 2004–2015 time period. 27 Looking more directly into the effect of liquidity on CDS spreads, Tang and Yan (2007) find that the liquidity impact on the CDS spreads is insignificant for high depth contracts. 28 In Internet Appendix J, we perform additional tests to shed light on the link between CDS momentum and ratings changes. First, we find that more recent CDS returns are stronger predictors of upcoming ratings changes. This predictive power is gradually phased out, and by the fourth or fifth lag the monthly CDS return is statistically insignificant. Second, CDS momentum returns tend to be higher during months in which past CDS returns display greater cross-sectional ability to predict upcoming ratings changes. 29 We skip the recent month as it follows JNPS and avoids the 1-month bond return reversal. However, regarding the comparison of CDS momentum to bond momentum, we reach similar conclusions when including the recent month in the formation period of the bond momentum portfolios. 30 Internet Appendix D examines the overlap between the bond and CDS momentum portfolios; the examination reveals that the winner and loser portfolios share roughly one-third of constituents across the bond and CDS. 31 Blanco, Brennan, and Marsh (2005) analyze the bond and CDS of 27 firms and find that roughly 80% of price discovery occurs in the CDS. See also the recent additional findings on the same information advantage of CDS over bond around credit rating and notch changes by Lee, Naranjo, and Velioglu (2018). 32 Much of the performance differential is due to the loser portfolio: low fragmentation bond losers perform much worse than high fragmentation bond losers. 33 Panel B of Internet Appendix I further confirms the predictive ability of past CDS performance for future bond returns. 34 Frazzini, Israel, and Moskowitz (2012) find that the main return anomalies to standard asset pricing models, such as value and momentum, are robust to transaction costs in stock markets and are therefore implementable and sizeable. 35 Panel C of Internet Appendix G shows that the conditional CDS-to-stock momentum strategy is robust to a broader set of stock market risk factors. Internet Appendix H shows that the results hold when stocks are equally weighted. Panel C of Internet Appendix I confirms the information content in past CDS performance for future stock returns through Fama-Macbeth cross-sectional regressions. 36 Kapadia and Pu (2012) document information segmentation between these two markets. References Acharya V. V. , Johnson T. C. ( 2007 ). Insider trading in credit derivatives . Journal of Financial Economics 84 : 110 – 41 . Google Scholar Crossref Search ADS WorldCat Asness C. S. , Moskowitz T. J. , Pedersen L. H. . 2013 . Value and momentum everywhere . Journal of Finance 68 : 929 – 85 . Google Scholar Crossref Search ADS WorldCat Avramov D. , Chordia T. , Jostova G. , Philipov A. . 2007 . Momentum and credit rating . Journal of Finance 62 : 2503 – 20 . Google Scholar Crossref Search ADS WorldCat Avramov D. , Chordia T. , Jostova G. , Philipov A. . 2013 . Anomalies and financial distress . Journal of Financial Economics 108 : 139 – 59 . Google Scholar Crossref Search ADS WorldCat Barroso P. , Santa-Clara P. . 2015 . Momentum has its moments . Journal of Financial Economics 116 : 111 – 20 . Google Scholar Crossref Search ADS WorldCat Berndt A. , Obreja I. . 2010 . Decomposing European CDS returns . Review of Finance 14 : 189 – 233 . Google Scholar Crossref Search ADS WorldCat Bessembinder H. , Kahle K. M. , Maxwell W. F. , Xu D. . 2009 . Measuring abnormal bond performance . Review of Financial Studies 22 : 4219 – 58 . Google Scholar Crossref Search ADS WorldCat Beyhaghi M. , Ehsani S. . 2017 . The cross-section of expected returns in the secondary corporate loan market . Review of Asset Pricing Studies 7 : 243 – 77 . Google Scholar Crossref Search ADS WorldCat Blanco Brennan , Marsh. 2005 . An empirical analysis of the dynamic relation between investment-grade bonds and credit default swaps . Journal of Finance 60 : 2255 – 81 . Google Scholar Crossref Search ADS WorldCat Bongaerts De Jong , Driessen. 2011 . Derivative pricing with liquidity risk: Theory and evidence from the credit default swap market . Journal of Finance 66 : 203 – 40 . Google Scholar Crossref Search ADS WorldCat Campbell Hilscher , Szilagyi. 2008 . In search of distress risk . Journal of Finance 63 : 2899 – 39 . Google Scholar Crossref Search ADS WorldCat Daniel Hirshleifer , Sun. 2020 . Short- and long-horizon behavioral factors . Review of Financial Studies 33 : 1673 – 36 . Google Scholar Crossref Search ADS WorldCat Daniel Jagannathan , Kim . 2019 . A hidden Markov model of momentum. Working Paper, Columbia University, Northwestern University, and Georgia Institute of Technology . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Daniel K. , Moskowitz T. . 2016 . Momentum crashes . Journal of Financial Economics 122 : 221 – 47 . Google Scholar Crossref Search ADS WorldCat Das Kalimipalli , Nayak. 2014 . Did CDS trading improve the market for corporate bonds ? Journal of Financial Economics 111 : 495 – 525 . Google Scholar Crossref Search ADS WorldCat Dichev I. D. , Piotroski J. D. . 2001 . The long–run stock returns following bond ratings changes . Journal of Finance 56 : 173 – 203 . Google Scholar Crossref Search ADS WorldCat Ehsani S. , Linnainmaa J. . 2020 . Factor momentum and the momentum factor. Working Paper, Dartmouth College and Northern Illinois University . Flannery Houston , Partnoy . 2010 . Credit default swap spreads as viable substitutes for credit ratings . University of Pennsylvania Law Review 158 : 10 – 031 . Google Scholar OpenURL Placeholder Text WorldCat Frazzini Israel , Moskowitz . 2012 . Trading costs of asset pricing anomalies . Working Paper, AQR Capital Management and University of Chicago . Google Scholar Crossref Search ADS Google Preview WorldCat COPAC Friewald N. , Wagner C. , Zechner J. . 2014 . The cross-section of credit risk premia and equity returns . Journal of Finance 69 : 2419 – 69 . Google Scholar Crossref Search ADS WorldCat Gebhardt W. R. , Hvidkjaer S. , Swaminathan B. 2005 . Stock and bond market interaction: Does momentum spill over ? Journal of Financial Economics 75 : 651 – 90 . Google Scholar Crossref Search ADS WorldCat Hong H. , Lim T. , Stein J. C. . 2000 . Bad news travels slowly: Size, analyst coverage, and the profitability of momentum strategies . The Journal of Finance 55 : 265 – 95 . Google Scholar Crossref Search ADS WorldCat Hong H. , Stein J. C. . 1999 . A unified theory of underreaction, momentum trading, and overreaction in asset markets . Journal of Finance 54 : 2143 – 84 . Google Scholar Crossref Search ADS WorldCat Hull J. , Predescu M. , White A. . 2004 . The relationship between credit default swap spreads, bond yields, and credit rating announcements . Journal of Banking and Finance 28 : 2789 – 811 . Google Scholar Crossref Search ADS WorldCat Jegadeesh N. , Titman S. . 1993 . Returns to buying winners and selling losers: Implications for stock market efficiency . Journal of Finance 48 : 65 – 91 . Google Scholar Crossref Search ADS WorldCat Jostova G. , Nikolova S. , Philipov A. , Stahel C. W. . 2013 . Momentum in corporate bond returns . Review of Financial Studies 26 : 1649 – 93 . Google Scholar Crossref Search ADS WorldCat Kapadia N. , Pu X. . 2012 . Limited arbitrage between equity and credit markets . Journal of Financial Economics 105 : 542 – 64 . Google Scholar Crossref Search ADS WorldCat Kisgen D. J. 2006 . Credit ratings and capital structure . Journal of Finance 61 : 1035 – 72 . Google Scholar Crossref Search ADS WorldCat Kisgen D. J. 2007 . The influence of credit ratings on corporate capital structure decisions . Journal of Applied Corporate Finance 19 : 65 – 73 . Google Scholar Crossref Search ADS WorldCat Kliger D. , Sarig O. . 2000 . The information value of bond ratings . Journal of Finance 55 : 2879 – 902 . Google Scholar Crossref Search ADS WorldCat Lee C. , Swaminathan B. . 2000 . Price momentum and trading volume . Journal of Finance 55 : 2017 – 69 . Google Scholar Crossref Search ADS WorldCat Lee J. , Naranjo A. , Sirmans S. . 2016 . Exodus from sovereign risk: Global asset and information networks in the pricing of corporate credit risk . Journal of Finance 71 : 1813 – 56 . Google Scholar Crossref Search ADS WorldCat Lee J. , Naranjo A. , Velioglu G. . 2018 . When do CDS spreads lead? rating events, private entities, and firm-specific information flows . Journal of Financial Economics 130 : 556 – 78 . Google Scholar Crossref Search ADS WorldCat Manso G. 2013 . Feedback effects of credit ratings . Journal of Financial Economics 109 : 535 – 48 . Google Scholar Crossref Search ADS WorldCat Menkhoff L. , Sarno L. , Schmeling M. , Schrimpf A. . 2012 . Currency momentum strategies . Journal of Financial Economics 106 : 660 – 84 . Google Scholar Crossref Search ADS WorldCat Miffre J. , Rallis G. . 2007 . Momentum strategies in commodity futures markets . Journal of Banking and Finance 31 : 1863 – 86 . Google Scholar Crossref Search ADS WorldCat Moskowitz T. J. , Grinblatt M. . 1999 . Do industries explain momentum ? Journal of Finance 54 : 1249 – 90 . Google Scholar Crossref Search ADS WorldCat Nelson C. R. , Siegel A. F. . 1987 . Parsimonious modeling of yield curves . Journal of Business 60 : 473 – 89 . Google Scholar Crossref Search ADS WorldCat Norden L. , Weber M. 2004 . Informational efficiency of credit default swap and stock markets: The impact of credit rating announcements . Journal of Banking and Finance 28 : 2813 – 43 . Google Scholar Crossref Search ADS WorldCat Oehmke M. , Zawadowski A. . 2015 . Synthetic or real? The equilibrium effects of credit default swaps on bond markets . Review of Financial Studies 28 : 3303 – 37 . Google Scholar Crossref Search ADS WorldCat Oehmke M. , Zawadowski A. . 2016 . The anatomy of the CDS market . Review of Financial Studies 30 : 80 – 119 . Google Scholar Crossref Search ADS WorldCat O’Kane D. 2008 . Modelling single-name and multi-name credit derivatives . Hoboken, NJ : John Wiley & Sons . Google Scholar Google Preview OpenURL Placeholder Text WorldCat COPAC Okunev J. , White D. . 2003 . Do momentum-based strategies still work in foreign currency markets? Journal of Financial and Quantitative Analysis 38 : 425 – 48 . Google Scholar Crossref Search ADS WorldCat Pirrong C. 2005 . Momentum in futures markets. Working Paper, University of Houston . Qiu J. , Yu F. . 2012 . Endogenous liquidity in credit derivatives . Journal of Financial Economics 1