Study On Credit Rating Of Listed Companies Based On Ordered Probit Model

Influenced by the international economic situation, market economy of our country has also been developing and changing constantly, followed by the credit ratingplaying a more and more important role in the development of the economy. While the existing credit rating systems and methods are mostly built on financial index data, investment industry and banking industry, even others, also put forward higher requirements to estimate the credit rating of enterprises accurately. By learning research achievements of credit rating in the domestic and overseas, we find that non-financial indicators also have significant influence on credit rating of enterprises. Comprehensive consideration of financial indicators and non-financial indicators can estimate credit rating of enterprises more accurately. And because credit rating of enterprises is polytomous ordinal response variable, we choose the ordered probit model in this paper. This model is one of the most commonly used models to estimate ordered discrete variables.In this paper we collect financial indicators and non-financial indicators of38listed companies as sample data, making use of statistical software R language for descriptive statistical analysis, correlation analysis and regression analysis of the data. Then use step wise regression method to simplifying the model. Estimation result shows that the non-financial indicators really have significant effects on credit rating of enterprises. It also shows that the stepwise selection model achieves the best AIC value and has better goodness of fit. Each of the variables left in the stepwise selection model has significant impact on the credit rating of enterprises. The ordered probit model obtains the estimation result for having the assumption that the random disturbance term has the standard normal distribution So we use Lagrange multiplier test on the assumption in this paper. The result shows that we can’t refuse that the random disturbance term in the model has the standard normal distribution.