A comparison of log-logistic survival model to logistic model in credit risk modelling
Credit scoring systems are based on statistical and operations research models that seek to establish the likelihood of future credit behaviour based on information from past borrowers. Given that the cost of credit risk involves exposures and credit lifetimes, and that these exposures are amortized through this lifetime, it is important to understand not just if, but also when credit events are expected. This paper focuses on the lifetimes of credit exposures relative to voluntary account closure by borrowers. It also seeks to determine whether a lifetime model can compete with the logistic model in modeling account closures. I have shown how the classic Pearson correlation and the Gini coefficient could be applied to assess model fit, and that the log-logistic survival model performs better than the logistic model based on the available data.
CitationM.Sc (Mathematical Statistics)
SponsorhipUniversity of Nairobi
School of Mathematics, University of Nairobi