Positive Research For Credit Risk In Listed Companies--The Comparative Analysis On The Basis Of Logistic Model

Modern market economy is the economy based on credit, market economy is the credit economy in a sense. Under the background of fact that global credit is inflating constantly, the credit risks expose more and more serious. The credit risks have already become the key risk that financial systems of various countries have faced. How to measure the credit risks accurately become the focus that financial institution , investor , government pay close attention to. The measure of the credit risks becomes one of the subjects of the risk research field with most challenging.This paper carried on theoretical research to the credit risks prediction of listed company and regarded listed company of our country as samples to carry on positive research.In chapter one, we analyzed the existing credit risk model and corresponding positive research, and choose Logistic regression model on the basis of comparing the several kinds of models .At present, there are four kinds of credit risks model. It is the traditional methods of analyzing credit risks, statistics model based on accounting data, neural network model and modern credit risks models based on market value. We compared the theories of all kinds of models and find: The traditional methods is simpler, but lack quantitative analytical capacity, and this method can not measure credit risk accurately; The commonly most used one is the linear discrimination analysis and probability model in the statistics models. Linear discrimination analysis is strict to distribution of variables, however it is difficult to meet the above request in practice, so the rationality of the model have to suspect; The neural network model does not have the data distribution problems, but the all course in the middle of it is simulated by the computer, so we are unable to receive the rational economics meaning; The credit risks models based on market value use the complicated mathematics method and the database, so it can measure enterprise's credit risks accurately and scientific. But in our country the field of credit risk prediction is in the starting stage, there are not intact historical data and lack the theoretical foundation of developing this kind of model, so this credit risk model is unable to receive application in our country at present.We have chosen Logistic regression model as the theory models of credit risks prediction of listed company. Logistic regression model is a kind of non-linear probability model, it is not strict to distribution of the data, have not used the complicated mathematics methods either, the conclusion that draw from Logistic model show that Logistic model can predict credit risk correctly. In chapter two, this paper provides the course of Logistic regression model and introduces the characteristic that Logistic regression model is suitable for measuring the credit risks. In essence, Logistic regression model is a non-linear probability model, it can provide the default probability. Because Logistic regression model is non-linear, it can control the probability within 0 and 1.Though Logistic model is non-linear, we can change it into linear form through Logit. In the course of estimating of parameter, we can utilize nature of linear regression. Utilizing these characteristics of Logistic regression model, we can discriminate credit risks .In chapter three, we carry on the positive research. In positive research we have chosen 186 companies as developing samples. On this basis, we develop a Logistic regression model which contains 9 independent variables. In addition, It is closing date with April 27 , 2004, we obtain 80 companies as test samples, by which we examined the prediction ability of the model and found more than 86% of the prediction accuracy .The fourth chapter provide the conclusion and policy recommendations. Through the positive research we find that Logistic model have 9 independent variables finally, including 3 debt paying ability variables, 3 profit ability variables, 2 development ability variables and a assets managerial a...