Credit scoring: retrospection, implementation today & application of fundamental econometric models on German Credit dataset

Abstract

One of the most important tools of the evaluation process of a customer’ s capability to pay off a loan and classify the customer into “bad risks” or “good risks” is credit scoring. An additional role of credit scoring is to reduce the possibility of a customer to default, which predicts the borrower’s risk level. The basic idea is to compare the characteristics of a customer with the characteristics of other customers of previous periods. If the customer ‘s characteristics are similar to those who have been granted credit and paid off the application will be approved. There are two problems in using credit scoring models for larger enterprises. The first one includes the information and its quality. As the size of the companies we examine is getting bigger, financial information is getting more important. Since companies are not obliged to keep records of their clients, the most of the characteristics of datasets refer to the owners’ repayment history than their financial status on their working activities The second obstacle also has to do with the information. The base of the problem is structural. Generally, analysts consider positive to test large populations whose members’ characteristics have a satisfying degree of homogeneity. As the market is consisted of a variety of sectors, it becomes difficult for analysts and agencies to collect homogeneous data of their clients. Taking into consideration everything mentioned above, this paper is going to present and analyze the main methods used in the credit scoring processes.