Credit Rating Of Small And Medium-sized Enterprises Based On Bayesian Model Averaging

Small and medium-sized enterprises play an increasingly important role in the market economy.As the most active force in the national economy,they not only provide a large number of jobs for people,but also increase the proportion of the national output value.However,due to the weak foundation of SMEs,small scale of enterprises,high operational risk,low financial transparency and non-standard operation,the financing environment of SMEs is not optimistic.This also leads to their unsatisfactory living and development environment.In addition,because of the serious information asymmetry between banks and credit guarantee institutions and small and medium-sized enterprises,financial institutions are also facing great credit risks,which aggravates the difficulty of financing.The credit rating of SMEs has always been the focus of academic and practical circles.It has great practical significance to improve the accuracy of credit rating prediction of SMEs.However,the research of scholars in China and abroad usually relies on a single model,which is easily disturbed by model setting and variable selection,and it has ignored the uncertainty of the model.This also makes it impossible for us to identify the importance of many factors affecting the credit rating of SMEs in China accurately and effectively.At present,the model uncertainty in the credit rating research of small and medium-sized enterprises in China has not been paid attention enough.So this paper tries to make some contributions in this respect.The objective of this paper is to research the credit rating of SMEs from the perspective of model uncertainty.In order to has a better understanding of the selection of variables and measurement methods in this empirical study,this paper firstly summarizes the influencing factors and statistical analysis models of SMEs’ credit rating,and at the same time combs the relevant literature of Bayesian Model Average in the past.Then,numerical simulation is used to analyze the ability of Bayesian model averaging method in variable selection and model building,compare to Lasso method and stepwise regression method.Then this paper considers the uncertainty of models,identifies and tests the importance of many factors that may affect the credit rating of small and medium-sized enterprises by BMA method,and then tests the effectiveness of BMA method in solving the uncertainty of models from two aspects.This paper provides possible explanations for the result.In the empirical study,this paper is based on the non-listed SMEs that issued and traded corporate bonds on the Shanghai Stock Exchange and the Shenzhen Stock Exchange in 2012 and 2017.The result shows that among the indicators that may affect the credit rating of SMEs in China are the scale of assets,inventory turnover,operating cash flow ratio.Also,the sales cost ratio and operating profit margin have strong explanations for SMEs’ credit rating,and some other factors have been analyzed accordingly.Finally,the out-of-sample prediction is made and compared with Lasso method and stepwise regression method.The final empirical results show that Bayesian model averaging method not only solves the problem of model uncertainty,but also improves the prediction accuracy of the established model.