Dynamic modelling of corporate Defaults

Geir Drage Berentsen


In this talk I will present a new model for count time series useful for modelling  corporate defaults. The model is an extension of a traditional autoregressive count time series model, where the parameters are allowed to depend on the state of an unobserved Markov chain. The estimated model for US monthly corporate defaults indicates (at least) two regimes and that the so-called contagion effect, that current defaults affect the probability of other firms defaulting in the future, is more present in one of these regimes, even after controlling for financial and economic covariates. The effects of financial and economic covariates are also significantly different in each of the regimes. These results imply that the notion of contagion in the default count process is time-dependent, and thus more dynamic than previously believed.