A joint analysis of market indexes in credit default swap, volatility and stock markets

ABSTRACT

This paper analyses the joint dynamics of the CDS, volatility and stock markets using both VAR and Markov regime-switching VAR models with market index data. It shows that the joint behaviour of the three markets is better characterized by the Markov model with two regimes corresponding to low- and high-volatile market conditions. The relationship between changes in the market indexes under a regime is consistent with theory and persistent; the information transmission process of shocks to the markets is similar for the two regimes with a more important role for CDS shock; and the volatility in the money market is an important determinant of regime-switching. The findings have practical implications, particularly for hedging strategies with market indexes under different market conditions.

KEYWORDS:

• Credit default swap
• volatility
• market linkages
• Markov switching

JEL CLASSIFICATION:

• C10
• G12
• G13

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ There are also few other studies on joint analysis of two financial markets under bivariate setting. For instance, Caporin (2013) examines the dynamics for two pairs: (1) equity index and CDS index, and (2) equity index and VIX index for the purpose of hedging equity risk with the use of either the CDS index or the VIX index as instruments. Fenech, Vosgha, and Shafik (2014) use a copulas approach to analyse the joint dynamics of Australian CDS and equity market indexes to investigate the impact of the global financial crisis on the association between the two market indexes. In Wang and Bhar (2014), a joint analysis is performed for CDS spreads and equity returns for the US market using a VAR framework.

² MOVE stands for the Bank of America Merrill Lynch Treasury Option Volatility Estimate Index.

³ To be more precise in Bai and Collin-Dufresne (2011), the authors have two crisis periods: the first one from 1 July 2007 to 31 August 2008 and the second one from 1 September 2008 to 30 September 2009, this latter date being the end of their sample.

⁴ We omit the graphs to save space, which are available from the authors upon request.

⁵ The quarterly US real GDP and monthly US CPI data are downloaded from the website of the Federal Reserve Bank of St. Louis. We first obtain the residuals from the first-order autoregression of the quarterly changes in the US real GDP growth and the monthly changes in the US CPI separately, and we then calculate the sample correlation coefficients between these residuals and the residuals of the Markov-switching model that are next to the corresponding residuals from the autoregression in terms of time.

⁶ Christoffersen (1998) introduces the three hypotheses: conditional coverage, independence and unconditional coverage, for assessing the validity of an interval forecast obtained by using any type of models. If an interval forecast is efficient with respect to given information, then all the three hypotheses should hold.

⁷ Note that the widening of this spread led to rewrite the theory of interest rate derivatives, the so-called multi-curve approach, see among the works of Johannes and Sundaresan (2007), Fujii, Shimada, and Takahashi (2010) and Mercurio (2010).