The Estimation of Probability of Default Using Logistic Regression

The aim of this work is to develop a suitable model that estimates a probability of default of client's loan. As estimation method was used a logistic regression and a probit regression and two definitions of default, 60 and 90 days overdue. The work describes the method of construction, estimation and testing of scoring models and a structure of dataset, which was used in the practical part. Firstly, it was created a theoretical model that was later confronted with estimates. Estimated models were compared by described statistics as McFadden R^2, the ability to diversify was investigated by the Lorenz curve and by the Gini coefficient. It was found that the logistic and the probit regressions have almost the same results, and that 90 days is preferable definition of default than 60 days