Incorporating heterogeneity and macroeconomic variables into multi-state delinquency models for credit cards

Multistate delinquency models model the probability that an credit account transits from one state of delinquency to another between any two points in the life of the account. Using a large sample of credit card accounts we parametrise such models with flexible baselines defined in terms of splines, and investigate whether predictive accuracy is enhanced by the incorporation of account specific random effects as well as the incorporation of macroeconomic variables. We conclude that macroeconomic variables are statistically significant in such models, that the inclusion of random effects renders some fixed effects less statistically significant but does not enhance predictive accuracy