Although the lending business is of great importance to the earnings and risk situation of banks, there are few empirical studies on the lending and credit surveillance process. The research project "Loan management" launched by the Center for Financial Studies (CFS), Frankfurt/Main, attempts to help to close this gap in cooperation with six universal banks in Germany. Thematically, the CFS project focuses on setting the terms for the loan, the market assessment of credit risks, and on renegotiations in situations of distress. The objective of the partial project "Default probability and rating in the lending business", conducted jointly by CFS and ZEW, is to analyse existing rating models for credit risk assessment. To avoid possible losses in the lending business, credit worthiness checks are carried out in banks by determining the default probability prior to granting the loan or prior to extending loan agreements falling due. The classification of the client in a credit rating scheme should reflect the default risk as accurately as possible. Yet, the first studies conducted on the basis of the CFS data set indicate that the connection between credit rating and the probability of the loan turning into a bad loan, is not as conspicuous as one would have expected. On the basis of these findings, we devised an estimated model with which it is possible to estimate the probability of a distressed loan occurring in the lending business. Another focus is on examining the influence the terms and collateralisation of the loan commitment have on the default probability. In this context, the method of statistical neural networks developed by ZEW permits a precise modelling of the functional links.