财务报告舞弊可能性识别的实证研究

Empirical Research on Identifying the Possibility of Fraudulent Financial Statements

本文根据注册会计师发表的审计意见类型和证监会对上市公司的处罚结果选择研究样本,选取23个财务指标作为研究变量,对财务报告舞弊可能性企业和非财务报告舞弊可能性企业进行实证分析。非参数Wilcoxon Mann Whitney检验法用来做差异性检验,Logistic分类法和BP神经网络用来对两类样本进行分类。实验结果表明,两类企业在负债水平、偿债能力、资产规模、资产结构、营运资金周转、盈利能力以及公司增长等方面有显著差异,并且,两类公司的分类准确率较高。本文的目的是给注册会计师提供关于某公司存在财务报告舞弊可能性的警示性信息。

We chosen sample under examination according to CPA＇s auditing opinion and punishment results of China Securities Regulatory Commission. Financial indicators are selected as research variable for empirical analysis on possibility of fraudulent financial statements and non-fraudulent financial statements. In the experiment, Wilcoxon Mann Whitney test is used for significance test, and. Logistic regression and back-propagation neural network are used for classification. The experiment result showed that two kinds of financial statements have significant differences in the debt levels, solvency, asset size, asset structure, operating fund turnover, profitability and company growth. Moreover, two kinds of financial statements can be classified with high accuracy using the proposed classification method. The purpose of this paper is to provide warning information to CPA on possibility of fraudulent financial statements.