Financial Analysts Journal Volume 67 Number 4 ©2011 CFA Institute Pricing Credit Default Swaps with Option-Implied Volatility Charles Cao, Fan Yu, and Zhaodong Zhong Using the industry benchmark CreditGrades model to analyze credit default swap (CDS) spreads across a large number of companies during the 2007–09 credit crisis, the authors demonstrate that the performance of the model can be significantly improved by calibrating it with option-implied volatility rather than with historical volatility. Moreover, the advantage of using option-implied volatility is greater among companies with more volatile CDS spreads, more actively traded options, and lower credit ratings. he credit derivatives market has grown expo- The link between CDSs and the options market nentially over the past decade, especially for can arise in several contexts. From a theoretical credit default swaps (CDSs). This develop- option-pricing perspective, the option-implied vol- T ment has brought with it the need to under- atility reflects the expected future volatility and the stand the pricing of CDSs. Financial institutions often volatility risk premium, both of which have been use CDS contracts to hedge against the credit risk in shown to explain CDS valuation in a regression- their loan portfolios. More recently, CDS contracts based framework (Cao, Yu, and Zhong 2010). From have become popular in relative value trading strat- a market microstructure perspective, recent evi- egies, such as capital structure arbitrage (Currie and dence points to the presence of informed trading in Morris 2002). Consequently, a suitable pricing model both the options market (Cao, Chen, and Griffin must reproduce both accurate CDS spreads and the 2005; Pan and Poteshman 2006) and the CDS mar- relationship between CDS spreads and the pricing of ket (Acharya and Johnson 2007). Theoretically, other corporate securities, such as common stocks, whether informed traders will exploit their infor- stock options, and corporate bonds. mation by using derivatives is likely to be a func- Using an industry benchmark model called tion of the leverage and liquidity of the derivatives CreditGrades, we conducted an empirical study of market and the overall presence of informational CDS pricing. As explained in Finger (2002), the asymmetry (Black 1975; Back 1993; Easley, O’Hara, CreditGrades model was jointly developed by and Srinivas 1998). Thus, one would expect the Deutsche Bank, Goldman Sachs, J.P. Morgan, and information content of option-implied volatility for the RiskMetrics Group as a standard of transpar- CDS valuation to exhibit company-level variations ency in the credit market. Based mostly on the sem- consistent with these predictions. inal Black and Cox (1976) model and extended to account for uncertain default thresholds, the Credit- The literature is replete with research on the Grades model provides simple, closed-form formu- relationship between CDS pricing and equity vola- las that relate CDS pricing to the equity price and tility. For example, several researchers have ana- equity volatility. We examined the performance of lyzed the connection between CDS spreads and the model across a large number of companies. historical equity volatilities (see Campbell and Tak- More importantly, using data from both equity and sler 2003; Ericsson, Jacobs, and Oviedo 2009; Zhang, options markets, we estimated the parameters of the Zhou, and Zhu 2009). Our study’s focus on option- model by incorporating the option-implied volatil- implied volatility differs from the focus of those ity into the calibration procedure. studies. Cremers, Driessen, Maenhout, and Wein- baum (2008) estimated a panel regression of corpo- Charles Cao is professor of finance at Pennsylvania State rate bond yield spreads and options market University, University Park. Fan Yu is associate profes- variables. Cao, Yu, and Zhong (2010) estimated sor of financial economics at Claremont McKenna Col- company-level time-series regressions of credit lege, Claremont, California. Zhaodong Zhong is spreads and focused on the role of the volatility risk assistant professor of finance at Rutgers University, premium in explaining CDS pricing. In our study, Piscataway, New Jersey. we addressed the inherently nonlinear relationship July/August 2011 www.cfapubs.org 67 Financial Analysts Journal between CDS spreads and equity volatility by fit- exchange-listed equity options in the United ting a structural credit risk model. Moreover, we States. We did not use the standardized concentrated on the cross-sectional interpretation of implied volatility provided by OptionMetrics company-level CDS pricing errors. Although our because that measure can be noisy owing to the study is similar in spirit to Stamicar and Finger small number of contracts used in OptionMet- (2006), who used case studies to illustrate the cali- rics’ interpolation process. Instead, we used bration of the CreditGrades model with options the binomial model for U.S. options with dis- data, our analysis is both deeper and broader in crete dividend adjustments to estimate the scope, with a significantly larger sample of compa- level of implied volatility that would minimize nies and a longer sample period that includes the 1 the sum of squared pricing errors across all put recent credit crisis. A supportive study by Luo options with nonzero open interest. (2008) showed that an extension of Yu’s analysis • Other variables. From CRSP we collected data (2006) of capital structure arbitrage to incorporate on daily stock returns, equity prices, and com- options market information significantly increases mon shares outstanding. We obtained data on the Sharpe ratio of this popular hedge fund strategy. the book value of total liabilities and total assets from Compustat. Using stock returns, Data and Summary Statistics we calculated historical volatility measures We obtained the data on the variables that we used with estimation horizons of 22, 63, 126, 252, in our study from various sources. and 1,000 trading days. We defined the lever- • Credit default swaps. We collected data on age ratio as total liabilities divided by the sum single-name CDS spreads from the Markit of total liabilities and market capitalization. Group. According to Markit, market makers We excluded companies in the financial, util- send it CDS data based on their official books ity, and government sectors. We required that each and records. The data then undergo a rigorous company have at least 377 observations (about 18 cleaning process to test for staleness, outliers, months of daily observations) of the CDS spread, and inconsistency. Any submitted data that fail the implied volatility, the 252-day historical vola- any one of these tests are rejected. The full-term tility, and the leverage ratio and that each company structures of CDS spreads and recovery rates have no more than 5 percent of missing observa- are available by entity, tier, currency, and tions between the first and last dates of its coverage. restructuring clause. In our study, we used the composite spreads of U.S.-dollar-denominated Our final sample consisted of 332 companies over five-year CDS contracts written on senior January 2007–October 2009. Table 1 reports the unsecured debt of North American obligors. cross-sectional summary statistics of the time- Furthermore, we limited our sample to CDS series means of the variables. The mean CDS spread contracts that allow for so-called modified was 198.30 bps, and the cross-sectional standard restructuring, which restricts the range of deviation was 243.94 bps. The average company maturities of debt instruments that can be had an implied volatility of 44.37 percent, a 252-day delivered during a credit event. historical volatility of 43.38 percent, and a leverage • Equity options. We obtained options data from ratio of 43.57 percent. Finally, the average company OptionMetrics, which provides daily closing had a market capitalization of $20.61 billion, about prices, open interest, and trading volume on the size of a typical S&P 500 Index company. Table 1. Summary Statistics, January 2007–October 2009 Standard Mean Q1 Median Q3 Deviation CDS spread (bps) 198.30 57.02 110.52 225.22 243.94 Historical volatility (%) 43.38 33.47 40.47 51.55 14.14 Implied volatility (%) 44.37 35.28 42.32 49.48 12.81 Market cap ($ billions) 20.61 3.50 8.78 20.84 39.68 Leverage ratio (%) 43.57 30.78 44.10 54.41 16.49 Notes: This table presents cross-sectional summary statistics of the time-series means for 332 sample companies. CDS spread is the daily five-year composite credit default swap spread. Historical volatility is for 252 trading days. Implied volatility is the volatility inferred from put options with nonzero open interest. Market capitalization is the product of the stock price and shares outstanding. Leverage ratio is total liabilities divided by the sum of total liabilities and market capitalization. 68 www.cfapubs.org ©2011 CFA Institute Pricing Credit Default Swaps with Option-Implied Volatility The CreditGrades Model V λ2 d = 0 e To address the nonlinear dependence of the CDS LD spread on its determinants, we conducted a pricing A = σλ22+ analysis by using the CreditGrades model, a struc- t t tural credit risk model in which equity volatility is With constant interest rate r, bond recovery calculated with information from either the options rate R, and survival probability function q(t), we market or the stock market. Although a full menu can show that the CDS spread for maturity T is of structural models has been developed following − ()1− R∫T ers dq () s the seminal work of Merton (1974), we chose the =− 0 c − . (4) ∫T rs () CreditGrades model for three reasons. First, it 0eqsds appears to be widely used by practitioners (Currie Substituting q(t) into Equation 4, the CDS and Morris 2002). Second, it contains an element of spread for maturity T is given by uncertain recovery rates, which helps generate real- 10− qHT()+ () istic short-term credit spreads.