基于Benford定律的公司年报数据质量评价研究

发布时间：2018-06-07 19:32

本文选题：Benford定律 + 年报数据质量　；参考：《北京交通大学》2017年硕士论文

【摘要】：在市场监管日益严格、上市公司内外部治理环境逐渐规范的背景下,如何快速而有效地识别和量化越来越隐秘的盈余管理行为,成为一个很重要的命题。本文创造性地构建了一个基于Benford定律的B指数综合评价模型,用以评价上市公司年报的数据质量,进而评估盈余管理的程度。B指数综合评价模型是一个相对性的评价方法,可以更好、更准确地评价来源相似、性质相近的数据集的相对数据质量。因为不同的数据集总体上符合Benford定律的程度必然有所差别,利用B指数就相当于将数据质量的评价标准“归一化”,从而使得同组内的数据集有了统一的数据质量评价参考点。本文运用B指数按财务指标、年份及股票市场板块三个方向进行了大样本实证研究,得出以下主要结论:(1)整体而言,我国上市公司历年披露的财务报表数据的首位有效数字(FSD)、第二位有效数字(SSD)以及第三位有效数字(TSD)频率分布均能较好地符合Benford定律,拟合程度均较高,除了管理费用、净利润等极少数科目的FSD的VN*(修正的K-S拟合优度检验)统计量较高之外;(2)历年来,我国上市公司年报数据的前三位有效数字与Benford定律的相关性呈现出明显的逐级下降的分布特征,数级之间的总体数据质量差异亦呈如是分布;FSD及SSD与Benford定律的相关系数平均都能达到0.95以上,但TSD与Benford定律的相关系数则普遍小于0.80,表明我国上市公司盈余管理有向后逐渐“隐秘化”的趋势;(3)我国上市公司披露的年报数据质量总体评价如下:1)就财务指标而言,无形资产和其他应收款的数据质量较好,管理费用和净利润的数据质量是这10项财务指标中最差的,资产负债表数据质量总体优于利润表;2)就年份而言,2015年、2009年、2010年以及2011年的数据质量较好,2012年和2013年的数据质量是近九年中最差的;3)就我国的股票市场板块而言,上证A股和深证主板A股的数据质量较高,创业板和深证A股的数据质量则较差。
[Abstract]:Under the background of increasingly strict market supervision and gradually standardized internal and external governance environment of listed companies, how to quickly and effectively identify and quantify more and more secret earnings management behavior has become a very important proposition. This paper creatively constructs a comprehensive evaluation model of B index based on Benford's law, which is used to evaluate the data quality of annual report of listed companies, and then to evaluate the degree of earnings management. B index comprehensive evaluation model is a relative evaluation method. It is better and more accurate to evaluate the relative data quality of similar and similar data sets. Because different data sets conform to Benford's law on the whole, the use of B index is equivalent to "normalizing" the evaluation standard of data quality, thus making the data sets in the same group have a unified data quality evaluation reference point. In this paper, a large sample of empirical research is carried out by using B index in three directions: financial index, year and stock market plate. The main conclusions are as follows: 1) as a whole, The frequency distribution of the first significant figure (FSDF), the second significant digit (SSDs) and the third digit significant figure (TSDD) of the financial statement data disclosed by listed companies in China over the years are all in good accordance with Benford's law, and the fitting degree is higher, except for the administrative expenses. The VNU (modified K-S goodness of fit Test) statistics of FSD in a few subjects such as net profit are higher than that of FSD over the years. The correlation between the first three digits of the annual report data of listed companies in China and Benford's law shows an obvious gradual decline in distribution. The difference of total data quality between several levels is the same. The correlation coefficients of FSD and SSD with Benford's law are above 0.95 on average. However, the correlation coefficient between TSD and Benford's law is generally less than 0.80, which indicates that earnings management of listed companies in China has a trend of backward "stealthiness" (3) the overall evaluation of the quality of annual report data disclosed by listed companies in China is as follows: 1) in terms of financial indicators, The quality of data on intangible assets and other receivables is better, and the data quality of management fees and net profits is the worst of the 10 financial indicators, Balance sheet data quality in general is better than income statement 2) in terms of year, data quality in 2015, 2009, 2010 and 2011 is better, and data quality in 2012 and 2013 is the worst in nearly nine years. The data quality of Shanghai A shares and Shenzhen main Board A shares is higher than that of growth Enterprise Market and Shenzhen Stock Exchange A shares.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F832.51;F275

【参考文献】