基于贝叶斯判别的中国上市公司利润操纵识别模型研究  

Bayes Classifying Analysis of Earnings Manipulation in Listed Companies in China

作　　者：李延喜[1] 高锐[1] 包世泽[1] 姚宏[1] 

机构地区：[1]大连理工大学管理学院,大连辽宁116024

出　　处：《预测》2007年第3期56-60,共5页Forecasting

基　　金：国家自然科学基金资助项目(70302008);霍英东教育基金资助项目(101084);教育部博士点基金资助项目(20040141031);辽宁省基金资助项目(1050354)

摘　　要：准确识别上市公司利润操纵行为对于规范证券市场、提高证券市场资源配置的效率、防范投资风险具有重要意义。本文从分析上市公司利润操纵的动机、手段及其表现入手,初步设计14个有一定配比关系和内在联系的财务指标,并经过独立样本T检验进一步筛选出其中的5个。选取证监会以利润操纵名义查处的35家上市公司为样本组,并以样本组公司为参照,选取同行业资产规模差距不超过5%且经营状况稳定、无重大违规行为的35家公司作为控制样本组,运用贝叶斯判别分析进行横向对比与判别,构建利润操纵分类识别模型,实证结果显示识别率达到了70%。Correctly recognizing the earnings manipulating behaviors of listed companies is very important for the healthy development of stock market in China. Based on the analyses of motivations, main tricks and appearances of aggressive earnings manipulation, we initially design 14 financial indexes which hold certain matching relationships and interrela- tionships among each others. Then, after further screening, we finally select 5 indexes from the initial 14 indexes. We select 35 listed companies which have been punished by China Securities Regulatory Commission as the earnings manipu- lation samples; and also select 35 healthy listed companies with the same firm size and same industry as the 35 punished companies as the control samples. Through using Bayes classifying analysis to compare the two groups of samples, we es- tablish a classification model of earnings manipulation. The empirical results indicate that the model can obtain good classification effects.

关 键 词：利润操纵 信息不对称 贝叶斯判别分析 识别模型 

分 类 号：F23[经济管理—会计学]