Title: Network-based PARX models to measure contagion in credit default counts
Authors: Paolo Giudici - University of Pavia (Italy) [presenting]
Arianna Agosto - University of Pavia (Italy)
Abstract: PARX models (Poisson Autoregression with Exogenous Covariates) are based on the assumption that a count time series, conditional on its past, follows a Poisson distribution with a time-varying autoregressive intensity whose formulation includes a set of exogenous covariates. The model was applied previously to the time series of monthly defaults among Moody's rated industrial companies in the period 1982-2012. The weakness of the approach, when applied to credit risk modelling, is the way it defines contagion: the autoregressive component may be influenced by misspecification or latent effects. The aim is to overcome this problem by explicitly defining a contagion component, based on a network model formulation, considering several time series corresponding to different economic sector and/or geographical areas.