آیا تجارت CDS بهبود بازار اوراق قرضه شرکت ها شده است؟
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|15066||2014||31 صفحه PDF||سفارش دهید||محاسبه نشده|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Journal of Financial Economics, Volume 111, Issue 2, February 2014, Pages 495–525
Financial innovation through the creation of new markets and securities impacts related markets as well, changing their efficiency, quality (pricing error), and liquidity. The credit default swap (CDS) market was undoubtedly one of the salient new markets of the past decade. In this paper we examine whether the advent of CDS trading was beneficial to the underlying secondary market for corporate bonds. We employ econometric specifications that account for information across CDS, bond, equity, and volatility markets. We also develop a novel methodology to utilize all observations in our data set even when continuous daily trading is not evidenced, because bonds trade much less frequently than equities. Using an extensive sample of CDS and bond trades over 2002–2008, we find that the advent of CDS was largely detrimental. Bond markets became less efficient, evidenced no reduction in pricing errors, and experienced no improvement in liquidity. These findings are robust to various slices of the data set and specifications of our tests.
A major innovation in the fixed-income and credit markets since the turn of the century is the introduction of the credit default swap (CDS), a credit insurance contract with a payoff linked to that of the default or change in credit characteristics of an underlying reference bond or issuer. Innovation, however, is a double-edged sword with likely mixed positive and negative outcomes. The creation of new securities could complete markets and favorably impact information generation and dissemination, as well as liquidity, yet, such innovations could also have negative externalities if the gains accrue to only a few market participants and cause an adverse impact on the rest of the market. In this paper we examine whether the advent of the CDS market improved the secondary corporate bond market in terms of its underlying efficiency, market quality, and liquidity.1 Taking a time series perspective, we examine the following question: did an issuer's bonds become more efficient and liquid after CDS trading was instituted on the reference instruments of the issuer? From a cross sectional perspective, we query: Are bonds of firms with traded CDS contracts more efficient and liquid than bonds of firms without any CDS contracts? Did corporate bond trading decline after the introduction of CDSs because traders were able to implement a credit view better and more cheaply in the CDS markets? Fig. 1 shows the mean size of bond trades relative to the date of inception of CDS trading for our sample of firms with traded CDS contracts benchmarked to a control sample of firms with no CDS introduction. The mean trade size falls in the two years following CDS introduction, indicating an evident decline in secondary bond market activity. Similarly, Fig. 2 depicts a likely drop in mean turnover of bonds of issuers with CDS contracts once CDS trading begins, with no appreciable change for control sample bonds. Full-size image (48 K) Fig. 1. Mean size of bond trades before and after introduction of credit default swaps (CDS). The upper plot shows the average size of each bond transaction (in millions of dollars) on a daily basis over periods of 500 trading days (two years) before and after the introduction of CDSs for the sample of CDS issuers, and the lower plot depicts the same for a pooled control sample of CDS nonissuers. The control sample includes all bond issues by firms that meet the selection criteria outlined in Appendix A but did not issue any CDSs until the end of 2009. The plots are based on data organized as continuous time series in which zero trade days are included. Panel A of Table 7 reports trade volume based on discrete panel data that exclude zero trade days. Figure options Full-size image (43 K) Fig. 2. Mean bond turnover before and after introduction of credit default swaps (CDS). The upper plot shows the average turnover for each bond transaction (volume as a percent of total amount outstanding) on a daily basis over periods of 500 trading days (two years) before and after the introduction of CDSs for the sample of CDS issuers, and the lower plot depicts the same for a pooled control sample of CDS nonissuers. The control sample includes all bond issues by firms that meet the selection criteria outlined in Appendix A but did not issue any CDSs until the end of 2009. The plots are based on data organized as continuous time series in which zero trade days are included. Panel A of Table 7 reports turnover based on discrete panel data that exclude zero trade days. Figure options Figs. 1 and 2 indicate that bond trading could have declined, but it is likely that bond market efficiency improved if the CDS market generated useful information that was quickly reflected in bond prices. As our empirics show, relative to other asset classes the informational efficiency of corporate bonds is poor both before and after the advent of CDS trading, and interestingly, bonds become more inefficient after CDS trading commences. This suggests that the CDS markets had a detrimental effect on bond market efficiency. Bond market quality showed no signs of improvement after CDS introduction. Also, using various measures of liquidity we find that post-CDS, on a relative basis, more liquidity attributes deteriorated than improved. The prior literature on bond market efficiency examines lead–lag relations between corporate bonds and equity markets as a way of assessing the relative efficiency of bonds to equity (e.g., Kwan, 1996 and Hotchkiss and Ronen, 2002; Downing, Underwood, and Xing, 2009; Ronen and Zhou, 2013). The findings are mixed. Our goal in this paper is different from that of the prior literature. Whereas we do revisit the issue of bond market efficiency, our goal is to assess what role the CDS markets played vis-á-vis the bond markets and to determine whether CDS trading was beneficial or detrimental to the underlying bond markets on criteria such as efficiency, quality, and liquidity. We examine these criteria before and after the inception of CDS trading when benchmarked against a control sample of firms with no CDS introduction. Our econometric specifications extend earlier work, necessitated by the increasing complexity of the fixed-income markets. Corporate bonds contain call and amortization features, various default triggers, and conversion and put options. Therefore, in this paper we consider multivariate lead–lag relations of corporate bond returns to returns on various other securities that would also be incorporating issuer-specific and systematic market-wide information. The issuer-specific information includes equity return of the issuer [as equity value is an input to deriving a firm's credit spread in structural models pioneered by Merton, 1974] and CDS spreads of the issuer (to measure the underlying credit risk of the firm). We consider aggregate systematic variables such as implied volatility embedded in equity index options (to capture information about market-wide business and credit risk) and return on interest rate swaps (which are increasingly used as benchmark interest rate instruments). Hence, a wide range of factors are used to assess the efficiency of bonds relative to other securities. In addition, lagged corporate bond returns could explain current returns if bonds are weak-form inefficient. To judge whether or not the introduction of CDS s was beneficial in enhancing bond market efficiency, we run relative efficiency tests for periods prior to the commencement of CDS trading for a firm, and we compare these results to the period after CDS trading commences. By regressing bond returns on contemporaneous and lagged values of these issuer-specific and aggregate variables, and testing for the joint significance of the lagged variables, we determine whether bonds are relatively inefficient compared with other securities that are also impounding information about the firm. Using this approach, we find that corporate bonds became increasingly inefficient as CDS markets matured. Our empirical analysis of bond market efficiency is robust and valid for a wide range of econometric tests and data configurations. We analyze bonds individually and also jointly in panel data analyses. We control for endogeneity in the decision to introduce CDS and confirm that there is nothing unique about the firms for which CDS trading commences (such as firms that are expected to become more illiquid or decline in credit quality) that drives our results.2 To complement the lagged regression analyses, we undertake difference-in-differences (DID) tests in which we augment our sample of pre- and post-CDS bond transactions for CDS issuers with control samples of bond transactions by CDS nonissuers (firms with no CDS introduction). This analysis also provides definite evidence of declining bond market efficiency after CDS introduction. We also construct different subsamples (eliminating the financial crisis period, removing periods of nascency in the CDS market, examining subperiods of maturity in the CDS trading of a reference issuer, examining if the results differ for liquid versus illiquid bonds, vary by firm size, and by ratings and maturity) and find that the results hold for these subsamples as well. In short, the post-CDS decline in efficiency of bonds relative to other asset classes persists across various subsamples and for alternate robust specifications. Did CDS trading improve the accuracy of bond prices? Following Hotchkiss and Ronen (2002) we implement the market quality measure (q) of Hasbrouck (1993). Hasbrouck's measure examines the discrepancy between efficient prices and transaction prices to assess the extent of pricing error. The inverse of the variance of pricing error is a metric of market quality. Whereas this metric is related to market efficiency, its focus is on whether prices accurately impound information. We compute q measure for bonds before and after the advent of CDS trading. The measure does not improve after CDS trading begins, suggesting that CDS markets did not enhance bond market quality. Did bond market liquidity respond favorably to the inception of CDSs? We compute several proxies for liquidity before and after the introduction of CDS trading. Our metrics include number and dollar volume of bond trades, turnover, the LOT illiquidity measure of Lesmond, Ogden, and Trzcinka (1999), the illiquidity metric of Amihud (2002) and the related Amivest liquidity measure, the spread illiquidity measure of Roll (1984), the covariance illiquidity gamma of Bao, Pan, and Wang (2011), and the zeros impact and Roll impact illiquidity metrics based on Goyenko, Holden, and Trzcinka (2009). Many trading and price impact measures remain unaffected by inception of CDS markets. Amongst metrics that changed, more liquidity attributes deteriorated after the introduction of CDSs than those that improved. Overall, no evidence shows that CDS introduction improved the liquidity of underlying bonds. Unlike equity markets that have much higher trading frequencies, examining these properties of bond markets is complicated by the fact that bonds are thinly traded, and consecutive days of trading might not always exist to compute returns for our tests. To ensure that available data are used to the best extent possible, we develop alternative approaches to augmenting the data, thereby resulting in larger data sets. The procedures are described in Section 3 and Appendix C. Re-running our analysis on an augmented data set confirms the robustness of our empirical analyses. Taken together, the results suggest that CDS introduction did not improve secondary bond market efficiency, quality, or liquidity. What explains our results? One possible explanation is that price discovery mainly occurs in the CDS market because of microstructure factors that make it the most convenient location for the trading of credit risk. Another is that different participants in the cash and derivative markets trade for different reasons (e.g., Blanco, Brennan, and Marsh, 2005). The CDS market involves very active trading and is mostly dominated by institutional players and, hence, constitutes a highly likely venue for all informed trading.3 At the same time, the corporate bond market is significantly less liquid. Bonds are traditionally held by buy-and-hold investors. Further, with the proliferation of the collateralized debt obligation securitization market, corporate bonds were increasingly parked inside pools and were not actively traded. For these reasons, as institutional investors migrated to the CDS market over time, corporate bond markets became less liquid and active [though Trade Reporting and Compliance Engine (TRACE) mandates did improve bond market liquidity; see, Edwards, Harris, and Piwowar, 2007; Bessembinder, Maxwell, and Venkataraman, 2006]. When we track the extent of trading by institutions before and after CDS introduction, we find evidence of a likely demographic shift by large institutional traders from trading bonds to trading CDSs to implement their credit views, resulting in declining efficiency and quality in bond markets. In addition, an analysis of the signed bond transactions by insurance companies reveals higher trade execution costs in the post-CDS period. Our findings echo earlier results found in option markets, where the price discovery role of options is more pronounced when the liquidity of the option market is higher compared with that of the stock market, when options provide higher leverage, and when the probability of informed trading is high (Easley, O'Hara, and Srinivas, 1998). Informed trading and price discovery in credit markets now also occurs in CDS markets, in addition to bond markets. Credit auctions also enhance the information in bond markets as clarity about recovery values increases (Gupta and Sundaram, 2012). Recent global over-the-counter (OTC) derivative market reform, in particular the regulatory efforts in the US spearheaded by the Commodity Futures Trading Commission and the Securities and Exchange Commission, recognize the important role of CDS markets vis-á-vis the bond markets, and it remains to be seen how new regulations will impact the bond markets.4 CDS trading is moving to centralized clearing counterparties (CCCs), in which the techniques in this paper could be applied in future work to assess whether the opening of a CCC has a beneficial impact on bond markets. The paper proceeds as follows. In Section 2 we review related work and distinguish our goals and methodology from earlier research. Section 3 describes the data set we employ. This section is complemented by Appendix C, which explains our new approach to creating nonoverlapping returns with a view to utilizing the entire data set for analysis, particularly for robustness tests. Section 4 presents tests of bond market efficiency and the finding that CDS markets could have been detrimental to bond market efficiency. We explore alternative cuts of the data set and variations of specifications (such as difference-in-differences tests, controls for endogeneity and fixed effects, and tests across subsamples) as robustness tests and show that the main findings about decline in efficiency are preserved. Section 5 explores the impact of CDS trading on bonds through the lens of the Hasbrouck (1993)q-measure and finds no improvement in market quality. Section 6 examines how CDS trading impacted the liquidity of underlying bonds using several metrics. There is no evidence of liquidity enhancement in bond markets. In Section 7 we consider one likely mechanism by which CDS introduction could hurt the efficiency of bond markets: the demographic shift by large institutional traders from trading bonds to trading CDSs. Conclusions and discussion are offered in Section 8.
نتیجه گیری انگلیسی
The credit default swap market was one of the salient new markets of the past decade. Trading in CDS has been blamed for the speculative frenzy leading to the beginning of the financial crisis in 2008, though Stulz (2010) concludes that credit default swaps were not responsible for causing or worsening the crisis. Nobelist Joseph Stiglitz went so far as to suggest that CDS trading by large banks should be banned.23 Still, the creation of new markets could have beneficial information and liquidity effects on underlying markets. Conrad (1989) and Skinner (1989) show that options trading reduced volatility in underlying equity markets. In sovereign bond markets, Ismailescu and Phillips (2011) provide evidence that the introduction of credit default swaps improved efficiency in the underlying sovereign bonds. We examine whether CDS trading was beneficial to bonds in reference names by looking at whether informational efficiency, market quality, and liquidity improved once CDS trading commenced. Our econometric specification accounts for information across CDS, bond, equity, and volatility markets. We also develop a novel methodology to utilize all observations in our data set even when continuous daily trading is not evidenced, because bonds trade much less frequently than equities. The empirical evidence suggests that the advent of CDS was largely detrimental to secondary bond markets. Bond markets became less efficient relative to other securities and evidenced greater pricing errors and lower liquidity. These findings are robust to various slices of the data set and specifications of our tests. Our findings have bearings on the recent CDS market regulatory reform proposals and the debate surrounding the impact and usefulness of CDS markets. Whereas we examine bond market efficiency, quality, and liquidity, this research did not examine the effect on credit, i.e., the impact on the quality of firms that experienced CDS introduction. Our endogeneity corrections did note that bond returns are negatively related to the implicit probability of CDS introduction, complementing the comprehensive analysis of this issue in Subrahmanyam, Tang, and Wang (2011). Other open questions remain that are not considered in this paper. Does CDS trading make forecasting default easier for reference names than for firms on which no CDS trades? How do capital structures change for firms that have CDSs traded versus firms with no CDSs? How does ratings volatility change when CDSs are introduced? Are firms that have CDSs traded more likely or less likely to have securitized debt? And, eventually, how does the trading of CDSs on centralized exchanges change the information environment for CDSs and bonds? These issues and questions are left for future research.