Model for the estimation of impairment due to credit risk

Abstract

The article develops an estimation model for expected loss as support to the Credit Risk Management System for a corporation of social economy. it describes the setting of a logistic binary regression model to estimate the probability the non-compliance of its affiliates and, it evaluates the model’s capacity to discriminate between compliance and non-compliance using the Hold Out recurrent method. Once the model is fixed and the probability is estimated, the recovery rate is determinate according to the guarantee offer by each credit. so, is possible the estimation of expected loss as well as the portfolio provisions required by each fund.