Research On The Influencing Factors Of Financial Fraud In Listed Companies

Financial fraud is a significant and developmental area of capital market research in recent years,which also has been a hot issue both at home and abroad.Foreign research on financial fraud has a long history,which also has had a relatively mature results on the identification method of financial fraud.The main methods used to obtain these results are:Logistic regression model,Bayesian discriminant model,Decision tree,Neural network model,etc.The domestic research results on financial fraud compared with the foreign research results appear to be thin and simple.But to be sure,in the process of the establishment on the identification model,foreign and domestic financial indicators of the definition and the composition of financial index system has a certain difference,domestic scholars are generally according to China's specific national conditions and the development of the laws and regulations,and refer to the specific financial environment is to build for our country's financial fraud identification modelIt is not difficult to find that the commonly used statistical methods in the process of establishing financial fraud identification model are Logistic regression model and decision tree model.Therefore,this paper combines the financial data of the domestic securities market in the past ten years,and uses Logistic regression model to build a financial fraud identification model.In this paper,by querying the announcements issued by SRC and by comparing the public information collected from the CNINF network,the A-share listed companies with financial fraud were identified and the samples after January 1,2008 were selected for analysis.Since there is a lag in the announcement of penalties for financial fraud,all samples from the last year,namely 2019,are deleted from this paper.According to statistics,from 2008 to 2018,108listed companies were involved in the disclosure of financial fraud,and a total of 312 financial reports were involved.In the sample selection,the samples are from the corresponding annual reports of Chinese A-share listed companies with financial irregularities disclosed by SRC from January 1,2008 to December 31,2018.Some listed companies also falsify temporary reports or quarterly reports,but such reports are inconsistent with some financial indicators in the annual report,so they are not included in the sample.Moreover,financial data of fraud incidents before IPO are not included in the sample selection.Our indicators are divided into continuous variables and dummy variables.For all indicators of non-dummy variables,the sub-value of the industry was calculated by the classification of CITIC First-level industries,that is,x=(x-min)/(max-min),so as to eliminate the influence of the industry on the relevant indicators,so that the range of the index value was limited to[0,1],which was more comparable between industries.Samples with missing indicators were directly deleted,resulting in 278 fake samples and 26024 controlled samples.Finally,a total of 53 characteristic indicators were screened out through the descriptive statistics and significance test of the samples.Then,factors were gradually added into the model through the stepwise regression method,and a financial fraud prediction model composed of 9 factors was obtained.