An Estimation Method For Default Probability Based On Sample Matching And An Application

In the credit risks management of the Commercial bank, probability of defaultrefers to the possibility that the borrowers who are not able to repay the principal andinterest of the bank loan or fulfill the related obligations in a future certain timeaccording to the contract requirement. The probability of default measure to theborrower which has been listed as the key element in the New Basel Capital Accord’sIRB, is an important aspect in the credit risk management of modern commercialbank.In this paper, we firstly discuss the definition of the default probability, becausethe accurate definition of defaults and default probability is the basis and premise ofthe research of default probability. And then we summarizes the domestic and foreignresearch about default probability on the basis of related research. Estimating thedefault probability is an elementary work in the risk management of commercialbanks. Many quantitative models are applied in the area. But size of “bad” samples ismuch smaller than the size of “good” samples. Biases arise if any quantitative modelapplied directly on the whole sample, which definitely leads to underestimation ofdefault risk. This paper proposes a method which searches for a good match between“good” and bad samples, then employs logistic model. The method is applied to aprovincial bank for its credit asset in the manufacturing industry. The empirical resultsshow that the method produces a balance prediction with a high accuracy, and a gooddiscriminating capacity.Credit rating comprehensively assesses the borrower’s capacity and intention ofon time repay the principal and interest of their loan in some way. As a elementarytool of credit risk management of the Commercial bank, borrower’s rating isespecially important, because the ratings is the bank’s important basis of approvingand pricing loan. This paper tries to use developed model of a default probability toestimate the debtor’s default probability, and then maps default probability into creditrating in the clustering way, finally determines the debtor’s default probability per credit rating according to the default frequency per rating, the result shows that therating method is feasible. we can obtain every borrower’s credit rating in2007,2008and2009samples, and rating migrate in this periods.