Factor Models and the Credit Risk of a Loan Portfolio
by Edgardo Palombini of Fondo Interbancario di Tutela dei Depositi (FITD)
Abstract: Factor models for portfolio credit risk assume that defaults are independent conditional on a small number of systematic factors. This paper shows that the conditional independence assumption may be violated in one-factor models with constant default thresholds, as conditional defaults become independent only including a set of observable (time-lagged) risk factors. This result is confirmed both when we consider semi-annual default rates and if we focus on small firms. Maximum likelihood estimates for the sensitivity of default rates to systematic risk factors are obtained, showing how they may substantially vary across industry sectors. Finally, individual risk contributions are derived through Monte Carlo simulation.
Keywords: Asset correlation, factor models, loss distribution, portfolio credit risk, risk contributions.