A Comparative Research Of The Performance Of Credit Risk Models

KMV model is a widely applied structural credit risk model,which utilizes stockprices information as major input.It is a forward-looking dynamic and timely, which canavoid the defect of reflecting the past status by accounting data. This model,with soundtheoretical support,many potential advantages and great popularity in practice,isnaturally very attractive for China. As an emerging market, Chinese market has manydifferences from foreign markets.If the KMV model can be applied to Chinese listedcompanies,it will be of greate significance to enhance our credit risk management.However without adequate objective validation processes,not only the benefits ofimplementing this sophisticated quantitative credit risk model cannot be fullyrealized,but also it is possible to make dangerous and wrong judgments. Thus,beforeusing it,it is necessary to test whether KMVmodel has a satisfactory discriminatingpower for Chinese listed companies.Z-score model, a multivariate linear model, withaccounting data as main input, has been well received in practice, but lack of convincingtheoretical basis. In this paper,the efficiency of KMV model is judged by comparingwith Z score model. Which model can evaluate the credit risk in Chinese market moreefficiently and accurately is the central topic of this reseach.In recent years, some scholars have already studied a lot on the KMV model andthese articles have great value for this paper, but subject to some objective conditions,they have some problems as follows: First, the sample is too small and the time span isshort; Second, the normal sample and the ST sample have the same size,withoutconsidering that bankrupt is a small probability event.In this paper,I choose950stocks from1459stocks listed on Shenzhen andShanghai Security Exchange on December31,2012under some standard.In the chosen950stocks，there are46*ST and ST stocks，which composite the ST sample，the other904stocks composite the normal sample.By analyzing the index got through Z-score model and the KMV model,wecompare their performance through empirical descriptive statistical analysis, regionalanalysis, ROC analysis and impact of asset value. The analysis result demonstrates thatthe KMV model has certain discriminating power but the power is limited and weaker than the Z-score model.Why the KMV model does not demonstrate its superiority in theory？This paperexplains it from the following perspective: First is environmental factors, China ’ssecurities market is weakly efficient, so stock prices may distort companies’ marketvalue;Second is the model itself, we uses the original KMV model which does not makeany changes; Third, because there are hardly any real Chinese listed companies inbankruptcy we have to define special treatment (ST)companies as defaulters.