The Application Of KMV Model In Prediction Of Credit Risk

With financial system construction and economic development in our country, financial system risk prevention is more and more emphasized, the outbreak of subprime crisis rings the alarm again for whole banking and market investors. How to discriminate and forecast financial risk, especially credit risk, is very important to decrease non-performing loan and promote the resistance of banking and investors, so more and more scholars input energy in this research field. The research in our country include two stages, in the first stage, scholars mainly use multiple discriminant analysis and traditional financial indicators to forecast and discriminate financial risk, in the second stage, more and more credit risk model, such as KMV, CMAC, Artificial Neural Network, Credit Metrics, are used. In this article, I try to combine traditional financial indicators with KMV model to build a new early warning system in order to promote the forecast precision of credit risk.This paper includes two models, the first is multivariable analysis model which includes traditional financial indicators, the second is a new model invented, with default distance(DD) and expected default frequency(EDF) which are the core factors in KMV model. I use Matlab software to calculate DD and EDF and need PCA discriminate model to extract five principle components from twenty-one financial indicators, then take these variables as the independent variables in new model and use Logit model to take regression testing on the two models. The result show that both the prediction accuracies are over 90%, however, the prediction accuracy of the new model is higher than the former model's prediction accuracy. Therefore, this try is effective and promote a new thought to research on banking credit risk early warning system in the future. In this paper I select 114 listed companies as samples. The sample size is large and samples are representative and the variables in KMV model have also been modified to fit Chinese market, which make the conclusions of this paper are even more value for Chinese market.