CDS spreads determinants and COVID-19 pandemic: A Bayesian Markov-switching model

Abstract

A deep understanding of the CDS spreads determinants is crucial for both policy makers interested in preserving the stability of the financial system and of financial insiders interested in managing credit and financial risks. The literature is mainly focused on the pre-subprime crisis, and either consider linear models with a large number of covariates or nonlinear models, such as regime Markov switching models, with a small number of explanatory variables and two regimes only. The aim of this thesis is to investigate the determinants of the European iTraxx corporate index considering a large set of explanatory variables within a Markov switching model framework. The focus is on the post 2007-2009 crisis and more precisely on the period from October 2011 to April 2020 which includes the recent COVID-19 pandemic events. The dataset includes financial and economic variables usually employed in CDS spreads analysis and some new explanatory variables such as the Baltic Dry Index as a proxy for the economic activity and lagged values of the iTraxx index. The analysis is conducted in two steps. First a multivariate regression model is estimated via OLS method on a rolling window to provide some evidence of variation in the parameters. Second, stability tests are also used to detect structural breaks in the linear relationship and to motivate the use of nonlinear models. Finally, the in-sample and out-of-sample analysis of the forecasting performances of different Markov switching models has been performed. The empirical results suggest that: more than 2 regimes should be used after the COVID-19 pandemic to model CDS spreads; the impact of the covariates varies across regimes; and that the economic activity index has some predictive power for changes in the iTraxx index.