Modeling Credit Spreads through Regime Switching with Gradual Transition

Credit spreads behavior displays a range of features that are challenging to model—strongly fat-tailed distribution of changes, periods of relative stability interrupted by prolonged violent shifts, lack of symmetry in spread rises vs. falls, etc. This article proposes a new model for spread behavior that incorporates these peculiarities without bringing excessive mathematical complexity. At the core of our approach is a Hidden Markov Model (HMM) that assumes that spreads follow a 2-state stochastic process. In a key departure from traditional HMM, the authors introduce explicit auto-regression into their formulation. The assumption behind that innovation is that while regime switches may be instantaneous and regimes may be characterized by different spread “fair” levels, the transition between such levels is gradual as opposed to instantaneous. As they illustrate, this assumption is critical for proper description of the spreads dynamic. The model lends itself well to tactical asset allocation involving high-yield credit assets and in a broad, multi-asset class setting.