Spatial Dependence in Credit Scoring of Group Lending : Evidence from China

This study aims to extend the standard credit scoring model and identify the credit risk of group-loan-taking farmers. For do this, we apply spatial random effects to the logistic model, and construct three different credit scoring models for group-loan-taking farmers: the logistic-INLA model without random effects, the logistic-Inde-INLA model with independent random effects and the logistic-CAR-INLA model with Conditional Autoregressive (CAR) random effects. Using data on 5,462 group-loan-taking farmers from a microfinance institution in China, we verify the spatial dependence of in credit risk of group lending. The results show that the models with spatial random effects significantly improve the prediction performance of credit risk for group-loan-taking farmers. Furthermore, we identify the key factors that affect the credit risk of group-loan-taking farmers. Our study offers new valuable insights for credit risk assessment, and the proposed models can be directly applied in the practice of credit risk assessment of financial institutions