Automated credit decision process – an insight into developing a credit-scoring model within the Nepalese banking sector

Abstract

There has been significant growth post-2000 in consumer credit within transition economies. Credit scoring, established within Western institutions, has the potential to be used to assess consumer creditworthiness here. This paper presents challenges and complexities relating to credit-scoring model development within the Nepalese banking sector. The research incorporates a mixed methods approach, involving model development using secondary data, supported by five in-depth interviews involving lending managers. A model was developed deploying binary logistic regression comprising six customer characteristics. Its overall ability to predict known outcome was high, particularly repayment success, although challenges remain in terms of predicting failure, pointing to a relative absence of current data on such customers. Implementation challenges also exist, reliance on traditional judgement prevails, together with ignorance of possible approaches to modelling. Decision overrides occur due to conflict between restricting defaulting customers and growth targets, traditional practice retention and desire to demonstrate expertise amongst lending managers.