The Generation of New Models

Abstract

Early versions of the rating models were designed to introduce in the credit process the calculation of the probability of default (PD), in other words to group customers into classes based on their creditworthiness. The aim was to equip banks with a model to be used both in the acceptance/revision of a loan and in the monitoring of customer’s behavior. Based on this, the cornerstones of the architecture of credit rating were built: customer segmentation; the use of different profiles of investigation (quantitative, qualitative and sectorial) and their weighing according to the customer segment; the introduction of correction factors in case of prejudicial elements; and the determination of PD through a mapping process. The end result was to determine the theoretical pricing of a loan based on the creditworthiness of the borrower (risk-adjusted pricing).