Using an innovative random regime shift detection methodology, we identify and confirm two distinct regime types in the dynamics of credit spreads: a level regime and a volatility regime. The level regime is long lived and shown to be linked to Federal Reserve policy and credit market conditions, whereas the volatility regime is short lived and, apart from recessionary periods, detected during major financial crises. Our methodology provides an independent way of supporting structural equilibrium models and points toward monetary and credit supply effects to account for the persistence of credit spreads and their predictive power over the business cycle.

Empirical studies on credit spread determinants are predicated on the presence of a single-regime over the entire sample period and thus find limited explanatory power. A single-regime model hides the fact that explanatory variables take on different loadings across changing patterns in credit spreads. In a model with endogenous regimes for credit spreads or with monetary regimes, we find that market, default, and liquidity factors have superior explanatory power because of their interaction with the regime. Lower improvements are found when the regime is defined according to the credit supply regime or the NBER regimes (announced and official).