The Study of Factors Affecting on Credit Risk of Bank Customers Using Non-Parametric and Semi-Parametric Survival Analysis Models

اقتصاد پولی مالی

Article 4, Volume 24, Issue 14, October 2017, Pages 88-123   

Document Type: پژوهشی

DOI: 10.22067/pm.v24i13.52294

Authors

nahid rajabzadeh moghani; mohamadreza lotfalipour^* ; Ahmad Seifi; mostafa razmkhah

ferdowsi

Abstract

Credit risk is one of the important risks in banking industry. So, identification of effective factors has been of substantial interest to banks. The aim of this paper is to identify factors that impact on credit risk with approach to time to approach.
In this study, a random sample of 5316 customers that borrowed from Maskan Bank during 1388-1393, is used. This paper, using common Survival Analysis models including Kaplan-Meier non-parametric model and proportional hazard Cox semi-parametric model, aims to identify effective factors affecting the default risk of costumers.
The results of non-parametric models, after categorizing variables based on frequency and Stepanova and Thomas (2002) technique, indicate that variables such as loan amount, number of installments, number of children, education, age,job type and job title have effect on survival probability and hazard rate curves. Entering variables simultaneously in Cox regression suggests that loan amount, number of installments, marital status, education, age and job type have effect on default risk of costumers. Harel and Lee test indicates that proportional hazard assumption for all variables is satisfied; hence theInterpretation of results is reliable. In addition diagnostics methods such as Cox-Snell, Martingle and Schoenfeld residuals all confirmed that the model fits well. So banks should pay attention to these variables when they want to decide about granting loans to costumers.

Keywords

Credit Risk; time to default; survival analysis; Cox model; Kaplan-Meier model