Rating-based CDS curves

ABSTRACT

This paper explores the extent to which term structure of individual credit default swap (CDS) spreads can be explained by the firm's rating. Using the Nelson–Siegel model, we construct, for each day, CDS curves from a cross-section of CDS spreads for each rating class. We find that individual CDS deviations from the curve tend to diminish over time and CDS spreads converge towards the fitted curves. The likelihood of convergence increases with the absolute size of the deviation. The convergence is especially stable if CDS spreads are lower relative to the rating-based curve. Trading strategies exploiting the convergence generate an average return of 3.7% (5-day holding period) and 9% (20-day holding period).

KEYWORDS:

• Credit default swap
• trading strategy
• CDS term structure
• credit rating

JEL:

• G11
• G17
• G24

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 ‘CDSs activity heats up’, Financial Times,https://www.ft.com/content/c47dce8e-ca9f-11e5-be0b-b7ece4e953a0.

2 The accuracy of the default probability estimate is important here. The potential models span the classical structural models such as Merton (1974), and its extensions including, a flexible corporate debt structure and default date in Leland and Toft (1996), or creditor-shareholder bargaining at firm's default (Fan and Sundaresan 2000; Ericsson and Renault 2006), as well as the reduced-form models including, for example, Altman (1968), Jarrow, Lando, and Turnbull (1997), Duffie and Singleton (1999), and Campbell, Hilscher, and Szilagyi (2008) among others.

3 For example, firms' ratings still can be used to determine the capital requirements in banks under the Basel III framework https://www.bis.org/bcbs/publ/d424.htm.

4 For instance, GFI/FENICS constructs single-name CDS spreads using Hull–White methodology; Markit also provides various smoothed credit curves (such as single-name CDS curves and sector credit curves) by pair-wise interpolating individual CDS spreads. See the Markit (2012, June) user manual for more information. In practice, credit curves are often used by clients to analyze the delta risk of the CDS spreads (CV01) or to assess the CDS spreads for other tenors. Credit curve providers might not consider the term structure of the CDS spreads, or provide the accuracy test for these curves.

5 Empirical literature suggests multiple individual CDS factors that impact the spreads. Das, Hanouna, and Sarin (2009) find that both accounting-based and market-based credit information are important drivers of CDS spreads. Several studies find that CDS illiquidity increases CDS spreads. See, for example, Tang and Yan (2007), Corò, Dufour, and Varotto (2013), and Das and Hanouna (2009) among others.

6 See Longstaff, Mithal, and Neis (2005).

7 Markit requires the data providers to report the quote for CDS spreads and the corresponding recovery rate at the individual entity level.

8 One alternative to our implied hazard rate would be the forward hazard rate, such that each CDS of the same underlying is priced using different forward hazard rates at different periods. The empirical challenge is that such forward hazard rates are inherently ‘unsmooth’, subject to a very high level of noise and estimation errors.

9 In a later section, we show that our approach is superior to the median in prediction for CDS movement.

10 The times to maturity of the CDSs in Markit are 6 months, 1, 2, 3, 4, 5, 7, 10, 15, 20, and 30 years. We select the CDSs with time to maturity 10 years or less, because these CDSs are traded more frequently.

11 See Markit (2012) for more details on the data cleaning process.

12 See Hull, Predescu, and White (2004) for further explanation.

13 See, e.g. Moody's Investor Service, Annual Default Study: Corporate Default and Recovery Rates, 1920–2015 https://www.moodys.com/researchdocumentcontentpage.aspx?docid=PBC_1018455.

14 Further looking at the time-series dynamics of CDS spreads for different maturities, we find that the pattern is almost identical for all maturities, with the only difference that spreads of shorter-term CDS contracts are generally smaller than those of longer-term contracts.

15 Note that since there is no active secondary market for CDS contracts, the composite spreads reported by Markit represent the spreads of newly issued contracts with set maturities from 6 months to 10 years.

16 Note that the reported minimum for 6-months fitted values is negative of −2.49 bps. This is a single negative observation in our sample, obtained for an AAA curve on 25 July 2006. The reason for such result is poor quality of calibration, as on that date there are only two observations to calibrate four parameters. There are no other instances of negative fitted values in our sample.

17 Further looking at the median residuals, we find that they are closer to zero in absolute values, although remain negative.

18 Further inspection of the annual variation of the convergence speed for this rating class finds convergence in most of years, except during financial crisis in 2008 and 2009.

19 We apply the transaction costs only to portfolios 1 and 5, since the trading direction in these portfolios is clearly determined, and we attempt to verify if the performance of the long-short portfolio remains positive after inclusion of transaction costs.