Abstract: The CreditMetrics model is a quantified credit risk management model. The determination of credit transition probability matrix is a key problem for this model. This paper presents a method validated by stochastic simulation for estimating credit transition probability matrix, and determines appropriate sample size by error analysis. It improves the loan cash flow calculation in the original model that is all possible transition situations within n years for a given type of loan customers to be considered for the load cash flow conversion. Finally we calculate the loan value at risk (VaR) with a kernel estimate method, and compare with that of the original model. The result shows the effectiveness of our method.

Key words: credit risk, CreditMetrics model, stochastic simulation, value at risk (VaR)