Title: Reducing model risk using Bayesian approach: Application to PD modelling of mortgage loans
Authors: Zheqi Wang - University of Edinburgh (United Kingdom) [presenting]
Jonathan Crook - University of Edinburgh (United Kingdom)
Galina Andreeva - University of Edinburgh (United Kingdom)
Abstract: A new Bayesian informative prior selection method is proposed to reduce model risk of ignored information and improve model performances. We use logistic regression to model the probability of default of mortgage loans using both Bayesian approach with various priors and frequentist approach. In the Bayesian informative prior selection method we propose, we treat coefficients in the PD model as time series variables. We build ARIMA models to forecast the coefficient values in future time periods and use these ARIMA forecasts as priors. We find that the informative Bayesian models using this prior selection method outperform both frequentist models and Bayesian models with other priors in terms of model performances.