Research And Improvement Of Financial Fraud Identification Model Of A-share Listed Companies Based On M-score

Since the establishment of Chinese securities market nearly 30 years ago,its scale has been growing.However,the accompanying financial fraud has always been hindering the development of the market.While disturbing the market order and attacking the confidence of investors,it also wastes a lot of national resources.Financial fraud cases emerge in endlessly and repeatedly.How to effectively identify financial fraud has become a common topic in practice and academia.Domestic research on financial fraud identification is limited,the characteristic variables used in each research are different,the methods to evaluate the accuracy of the model are not detailed enough,and the discussion on threshold setting is less.Investors lack a simple and effective fraud identification method.In this paper,through combing the research results of financial fraud at home and abroad,M-score model which is widely used abroad is selected for research.However,indexes of M-score are selected according to the market environment and the characteristics of corporate fraud in the United States,which is quite different from China in both accounting system and construction background.This paper uses the data of Chinese A-share listed companies to test Mscore,and finds that the model has poor applicability in the Chinese market,with Type I error as high as 70.37%.In order to build a financial fraud identification model suitable for Chinese market,this paper starts from the background of M-score construction,and takes the places that are not suitable for Chinese market as the basis for index addition and the direction of model optimization.On the basis of literature review and statistical results induction,7 financial indicators and 3 non-financial indicators are added.Through the Mann-Whitney U test,it is proved that these indicators are significantly different between the two types of samples,which can represent the characteristics of fraud of Listed Companies in China.Then,through Wald's stepwise backward regression method,a logistic model is constructed,which includes 9 indexes: gross profit index,sales growth index,depreciation index,leverage index,accrual coefficient,fixed asset ratio,current ratio,equity concentration and audit opinion type.Indexes can explain financial fraud from three dimensions: motivation,behavior,and consequence.The optimal threshold is selected by the Beneish Expected Cost Method.Finally,the test results of the modified model show that the accuracy rate of fraud identification is significantly improved compared with the original M-score model,and the Type I error is reduced from 70.37% to 19.75%.The ROC curve test also proves the significant identification effect of the modified model again.At the end of this paper,fraud detection methods are proposed based on the constructed model,and specific governance measures are given in combination with the three-factor fraud theory.This paper is different from the existing methods that use a large number of variables to screen and regression to select indicators.Instead,it starts with the background of the model construction and specifically adds indicators that can characterize the fraud characteristics of Chinese listed companies.Moreover,the research sample was subdivided into the specific fiscal year in which the fraud occurred,making the modified M-score model more suitable for the Chinese market.