Application Research Of LOGIT And KMV Credit Method

Credit is an essential part of our lives. As one of the main risks that financial institutions especially banks face—credit risk, it not only has influences on all aspects of modern social and economic life, also affects a country’s macroeconomic policy and economic development, even affects the stability and development of the global economy. Along with the financial crisis occurs frequently and increasingly widespread, large-scale corporate events of default emerge, many people has been aware of the increasingly important position of risk measurement methods in asset pricing and risk management of credit. A series of effects coming from the subprime crisis which originated in the United States have take strong measure against global finance and economy, which make people pay more attention to the risks, in particular credit risk management and the predict effects of model. At the same time, along with the reform of Chinese financial system and the openness of the domestic financial market continue to deepen, world economy has an increasing influence on domestic economy, so we need the higher level of credit risk management. But compared with developed countries, Chinese credit risk management technology is still in its fancy, so the credit risk measurement and management reform will be bound to one of the priorities of the next step of our financial institutions.This paper introduces the related theories of the credit risk and the present situation of study at home and broad at first, we can see the direction of the development of the credit risk measurement models. In this paper, the credit risk measurement methods are divided into the traditional credit metrics, multivariate statistical metrics and modern credit risk measurement, we introduce their main content, feature and advantages and disadvantages in turn. Secondly, we focuses on the mathematical principles, parameter selections and usable methods of the Logit model and KMV model, Thirdly, we introduce the CAP curve and the ROC curve which effects are obvious in the comparative analysis of the model’s discrimination ability.In the empirical part of the model, we select the listed companies of the chemical industry in China as a sample, and make the credit risk prediction using regression fit of Logit model and KMV model based on the theory of option pricing.85listed companies and23financial data in2011are selected as the estimated sample, we extract seven common factors using principal component analysis in order to reduce the influences of overfit and multicollinearity.In the end we select55listed companies in2012as a test sample to monitor model using the Logit model, to KMV model, we select stock market price,the value of liabilities, the risk-free rate and the growth rate indicators of the same55listed companies in2012to forecast their default probability. To the Logit model and the KMV model’s predictive power analysis, we further apply to the CAP curve. ROC curve, and Brier Score and compare their prediction accuracy and conclude that KMV model prediction accuracy is significantly higher than Logit model from both the graphs and indicators. The reasons are mainly that Logit model is mainly limited to the assessment of the company’s financial performance, while ignoring the impact of non-financial factors such as macroeconomic, industry cycles and so on for the credit risk. The Logit model has a probability of default prediction divergence and higher Blair fraction with dichotomous variable model, so the forecast is not very ideal out of the sample; financial data reflects the company’s history and lack of timeliness. From the prediction results of the models, error probability of the first type is significantly greater than the second type for both Logit model KMV model, so we can indicate that forecast accuracy of defaulting company is much greater than the non-defaulting companyFinally, we give a summary, including the conclusions of the empirical analysis, the limitations of the model and suggestions for credit risk management. We compare credit risk measurement models which currently are more popular using a combination of qualitative and quantitative methods from the perspective of comparative analysis, strive to elaborate comprehensively and analyze the problem in-depth.