Credit risk modelling using multi-state markov models

Abstract

This paper is devoted to credit risk modelling issues concerning mortgage commercial loans. Mortgage loans are one of the most popular type of loans provided by credit institutions. Like in the case of other loans, the main concern of institutions providing this type of product is a potential inability to recover the amount assigned to their clients (credit risk). In order to prevent possible losses for credit institutions resulting from clients entering in default, it is therefore crucial to study the behaviour of risky clients. This issue can be addressed through several models, namely through the multi-state Markov model, despite it constituting a more unusual approach in the context of dealing with credit risk modelling. The multi-state Markov model is a useful way of describing a process in which an individual moves through a series of states (finite number) in continuous time. By fitting this model to the loans of risky clients, it is possible to estimate the mean sojourn time in each state before a transition occurs, as well as the transition probabilities between the different states assumed by the contracts, therefore providing a relevant modelling framework for event history data. The present work relies upon 2008-13 databases from one of the biggest American companies that act in the secondary mortgage market, the Fannie Mae. Results show that with the application of the multi-state Markov model, contracts signed during 2013 are more propitious to a scenario of recovery when compared to those referring to the year 2008.