Firm-Specific Credit Risk Estimation in the Presence of Regimes and Noisy Prices

Abstract

Security prices are important inputs for estimating credit risk models. Yet, to obtain an accurate firm-specific credit risk assessment, one needs a reliable model and a methodology that filters all elements unrelated to the firm's fundamentals from observed market prices.Therefore, we first introduce a flexible hybrid credit risk model defined in a Markov-switching environment. It captures firm-specific changes in the leverage uncertainty during crises as well as the negative relationship between creditworthiness and recovery rates. Second, estimation is performed using maximum likelihood by accounting for latent regimes and unobserved noise included in security prices.Using CDS premiums for 225 firms of both CDX North American IG and HY indices, we perform two different empirical applications. The effects of stochastic recovery and the presence of regimes on theoretical credit spread curves are investigated. We also apply the model to corporate bond credit spreads to assess the importance of bond-specific liquidity in the bond-CDS basis.