The CDS market has existed for a long time but the LCDS market was launched in 2006 in both US and Europe. We obtain our CDS and LCDS data from Markit who collects the quotes on LCDS spreads from large financial institutions and other high quality data sources and produces the LCDS spread database on a daily basis starting from April 11, 2008. Our sample is from April 11th, 2008 to March 30th, 2012, which
encompasses the credit crisis and the Great Recession. Since the LCDS contracts can be divided into US LCDS and Euro LCDS based on the embedded cancellable feature, we only use US LCDS to construct the portfolio in order to keep our analysis model free.
In the CDS market, the contracts on senior unsecured debts are selected since this type of contract is the most liquid and is used frequently in the literature. In the LCDS market, the contracts on the first-lien syndicated loans are selected since the claims on collateral for the first-lien loans are senior to those of the second-lien loans, which indicate more reliable estimated recovery rates. In addition, the LCDS contracts on first- lien loans are the majority and more liquid compared to those on the second-lien loans. We restrict our CDS and LCDS contracts to those in the United States and denominated in US dollars. To ensure that the first-passage default and survival probabilities of the
172
CDS contracts are exactly the same as those of the corresponding LCDS, we match the daily LCDS and CDS data based on company name, denominated currency, restructure clauses and time to maturity. We only study the contracts with a 5-year maturity since they are the most liquid contracts and the most studied in the previous literature.94 Markit
also reports the estimated recovery rates obtained from their clients. These recovery rate expectations at time of issue may differ from subsequent recovery-rate expectations and actual recovery rates, especially during bad economic times.95 Nevertheless, these data
represent the only available proxy for the real recovery rates96 (especially for LCDS
contracts) and have been used in previous studies.97 Table IV-1 reports the summary
statistics for our full sample and sub-samples. We eliminate the observations whose CDS spreads (or LCDS spreads) are greater than 1 and the single name contracts which have less than 120 consecutive daily observations. In addition, we obtain the accounting variables from COMPUSTAT, economic macro variables from Federal Reserve H.15 database and equity information from Bloomberg. After merging all these datasets and removing the missing observations and private firms, the full sample contains 68,147 firm-clause-daily cross-sectional observations for 120 single names during the sample period from April 11, 2008 to March 30, 2012.
[Insert Table IV-1 about Here]
94 See Jorion and Zhang (2007), Cao, Yu and Zhong (2010, 2011), Schweikhard and Tsesmelidakis (2011),
Qiu and Yu (2012) and Zhang, Zhou and Zhu (2009).
95 Jokivuolle and Peura (2003), Altman, Brady, Resti and Sironi (2005), Hu and Perraudin (2002) and
Chava, Stefanescu and Turnbull (2006) report that the recovery and default rates are negatively correlated.
96 The real recovery rates are collected from Moody’s Default and Recovery Database and discussed in
section 4.
173
In the full sample, the mean LCDS and CDS spreads are around 3.7% and 4.6%, respectively. Both medians are smaller than their corresponding means which indicate asymmetric distributions and fat tails, especially on the right side. These style factors are also verified by positive skewness and high kurtosis for the CDS and LCDS spreads. The distributions of recovery rates for the LCDS and CDS contracts are close to a Gaussian distribution with slightly negative skewness. Both the mean and median of the LCDS recovery rates, around 65% and 70% respectively, are greater than the corresponding statistics for the CDS contracts, around 38% and 40% respectively. Intuitively, the syndicated secured loans which are the underlying asset of LCDS are usually backed up with collateral and have claim priority compared to the senior unsecured debts which are the underlying asset that backs the CDS once the default event occurs. The sub-sample of investment grades (includes firms rated greater than or equal to BBB), accounts for more than 60% of the total observations, while junk rated contracts and not rated contracts share almost equally the rest of the observations, approximately 20% each. As expected, both the mean and median of the CDS and LCDS spreads in the investment grade sub- sample are relatively lower, while the mean and median of the recovery rates are relatively higher compared to the junk subsample. In terms of the accounting variables, the average values in the full sample are around 25 million for total assets, 61% leverage ratio, 54% tangible asset ratio and 1.54 current ratio. Compared to the statistics of the investment grade firms, the means for the junk firms are higher for the leverage ratio and lower for total assets, tangible asset ratios and current ratios. The not rated firms are mostly relatively small firms in terms of their total assets. Their leverage ratios, tangible asset ratios and current ratios are diverse.
174
The daily idiosyncratic volatilities98 of the full sample have a mean around 2.4%
with positive skewness and extremely high kurtosis. As expected, both the mean and median of daily idiosyncratic volatilities of the investment grade firms are relatively lower than those of junk rated firms. For the not rated firms, the daily idiosyncratic volatilities are more volatile compared to the other sub-samples.