| Financial fraud has always been the focus of widespread social attention.Frequent financial fraud incidents of listed companies in domestic and foreign capital markets have become a worldwide nuisance.Even in United States whose market is strict and mature,listed companies have frequently reported financial fraud scandals.Well-known giants,such as World Telecom and Global Airline,have fallen into a whirlpool of financial fraud,which has continuously refreshed people's awareness of financial fraud and enabled the world's capital markets be shrouded in the shadow of financial fraud.In recent years,financial fraud scandals of listed companies in China have been often seen,such as Kangmei Pharmaceutical,Xintai Electric,Great Wisdom,and Kanghua Agriculture and so on.They abandon the professional ethics of “good faith,ethics first,adherence to standards,and no false accounts”.The company's financial fraud reflects the lack of the overall social credit system.This action not only sharply reduces the reliability of financial report,but also damages the legitimate rights of external investors.Finally,it is not conducive to the healthy development of the capital market,and has caused serious wealth loss to the entire society.Therefore,the detection and prevention of corporate financial fraud is a matter of great concern to both the theoretical and practical circles.On October 16,2014,the China Securities Regulatory Commission issued “Several Opinions on Reform,Perfection and Strict Implementation of the Delisting System of Listed Companies”,which set out clear institutional arrangements for the delisting behavior of listed companies.Especially the “Implementation of a mandatory delisting system for major illegal companies” focuses on illegal acts such as fraudulent issuance and major information disclosure.On November 16,2018,China's Shanghai and Shenzhen Stock Exchange issued the “Implementation Measures for the Compulsory Delisting of Listed Companies”,which clearly proposed the serious treatment of listed companies' illegal behaviors,and also showed that China's capital market regulators are determined to stop the fraud of listed companies.Although the legal environment of China's capital market has been continuously optimized and the governance level of listed companies has been significantly improved,there are still many irregularities and cases of fraud.In the context of the era of big data,financial fraud is more hidden,and fraudsters have more ticks to manipulate financial indicators.Individual financial information is difficult to capture the motivation of financial fraud and prevent it.The text analysis of non-financial information is a hot topic in the fields of finance and accounting in recent years,and the value relevance of non-financial information is getting higher and higher.Non-financial information does not have a fixed disclosure format and is easy to be manipulated by management.So it is common for non-financial information to be manipulated to confuse or mislead users of information.The text manipulation of MD&A is matched with the company's negative information hiding behavior to achieve the purpose of covering up financial fraud.The value of non-financial information and its mechanism have not been fully explored.Text analysis methods based on deep machine learning can improve the under standability and credibility of non-financial information,and then have an overall utilization of non-financial information.The text analysis of this study based on deep machine learning can improve the intelligibility and credibility of non-financial information,and the overall utilization of non-financial information,so as to give full play to its value relevance.Specifically,this study uses deep machine learning methods or natural language processing techniques to perform text analysis on the MD&A part of the annual report to help information users better understand MD&A text information and help regulators monitor the quality of MD&A text disclosure at any time,thereby improving the value relevance of MD&A text information.On the basis of discussing the textual manipulation of the company's financial fraud and annual report disclosure,this study analyzes the time series of the text information of annual report before and after the company's financial fraud.First of all,this study uses the annual report text similarity,complexity,and intonation to measure the text form characteristics of annual report,and explores the temporal changes of the text form characteristics before and after financial fraud.Secondly,this study uses risk information,idiosyncratic risk,and non-idiosyncratic risk of annual report to measure the subject characteristics of the annual report,and explores the time series of the annual report before and after financial fraud.Thirdly,this study uses the short-term stock price change to capture the process of market identification about companies using text manipulation to cover up financial fraud.In the medium and long term,the market can gradually detect whether it reflects on the stock price and verify that the market is effective during the process of progressive identification of this behavior.Finally,this study attempts to analyze the text information of annual report to detect financial fraud,providing a new perspective for regulators to detect financial fraud in the context of artificial intelligence.The main research contents of this study are as follows:Firstly,this study focused on text manipulation of financial fraud and annual report information disclosure features,that is,the time series of regular changes in the information features of annual reports before and after financial fraud.First of all,when a company engages in financial fraud,it will strategically reduce the similarity of the text,provide more reasonable explanations,or disperse or overwhelm investors' attention to signs of financial fraud with massive information,rather than repeating important or non-important information.Increasing text similarity is more in line with the logic of concealing financial fraud.Secondly,lower readability will cause negative guessing,but higher readability will send a clear signal to the market and bring positive effects.Disclosing information can reasonably explain signs of financial fraud in a more readable way.Finally,market participants respond positively to optimistic language,emotions or intonation,and have a milder response to pessimistic intonation.Using optimistic intonation is more concealed to cover up financial fraud than pessimistic intonation.These analyses show that while companies are engaged in financial fraud,the similarity of non-financial information text is more likely to decrease,the complexity is more likely to increase,and the tone is more likely to be optimistic.This study found that: when companies are engaged in financial fraud,the similarity of MD&A texts is significantly reduced,the complexity is significantly increased,and the tone is significantly more optimistic.However,after the financial fraud is investigated,MD&A textual characteristics changed completely in the opposite direction,namely,the similarity increases significantly,the complexity increases significantly,and the tone is significantly less optimistic.This shows that financial fraud has caused regular changes in the characteristics of MD&A texts.The regular changes in the MD&A textual features before and after financial fraud are mainly reflected in companies with a strong fraud heart,more opportunities for fraud,a stronger demand for fraud,and a lower probability of fraud exposure.That is to say,the more serious the agency problem,the more fraudulent the company will be;the more internal control defects and the lower the corporate governance efficiency,the more the company will have a fraud motive;the stronger the stock issuance and salary motivation,the greater the external pressure,The company has a need for fraud;the worse the regional legal environment and the longer the interval between fraud investigations,the less likely the company's fraud will be exposed.Under the above circumstances,the company is more likely to conduct financial fraud and use MD&A text manipulation to conceal it carefully,so it also shows that the regular change of MD&A text characteristics before and after financial fraud is a strategic behavior of the company to use MD&A text manipulation to cover up financial fraud.Secondly,this study focused on the company's financial fraud and text manipulation of the disclosure of subject matter in annual report,that is to say,it focuses on the time series of the annual report text information before and after financial fraud.When a company engages in financial fraud,the management will increase the difficulty in interpreting the text information of the listed company's annual report,which will bring great challenges to investors in correctly interpreting the company information.Managers strategically disclose the content of the annual report text for their own profit.Specifically,text information disclosure is more controllable,in order to avoid being broken down,companies may reduce the information content of text information disclosure and increase non-specific information disclosure,such as policy risk,exchange rate risk,approval risk,tax risk,etc.At the same time,reduce the disclosure of idiosyncratic information,such as business risk,supply chain risk,sales risk,financial risk,accounting risk,audit risk,and reorganization risk,in order to cover up financial fraud.This study finds that: when a company engages in financial fraud,the company will strategically use the content of the annual report risk theme to conduct text manipulation.The specific manifestations are as follows: the annual report text risk theme content is significantly reduced,the idiosyncratic information is significantly reduced,and the nonidiomatic information is significantly increased.The company is more likely to conduct financial fraud and use MD&A text manipulation to cover it carefully,so it also shows that the regular change of the subject content of MD&A text before and after financial fraud is a strategic behavior of the company to use MD&A text manipulation to carefully cover up financial fraud.The company's strategic use of the annual report risk theme content for text manipulation is more significant in companies with small earnings management space,two-in-one integration and poor legalization.The internal and external governance of the company significantly affects the company's use of text.In addition,text subject content manipulation has a substitute effect on digital information manipulation,the company can use MD&A text manipulation to carefully cover up the financial fraud of strategic behavior,and managers can obtain higher excess compensation.Thirdly,this study focused on whether the market can effectively identify companies' using annual report text information to cover up financial fraud.The company carefully manipulates financial fraud by using text manipulation,which will temporarily confuse market participants' understanding of the information,increase the difficulty in identifying the behavior of the market,and delay the market's identification of financial fraud.The company's carefully concealed behavior may make the short-window stock price have no response to financial fraud and exacerbate the phenomenon of information drift.At the same time,the process of the company using text manipulation to cover up financial fraud is actually a process of accumulating bad news,and the market's recognition of this behavior is also a process of slowly releasing bad news.Therefore,the company's use of text manipulation to cover up financial fraud may cause bigger stock market crash risk.These show that it takes a long time for the market to identify a company's careful cover-up.This study tried to use five days,ninety days,and three windows of one year's share price changes to capture the process of market identification companies using text manipulation to cover up financial fraud,and finds that in the short term,the market could not identify companies that deliberately fraud and cover up behavior,but in the medium and long term,the market can gradually detect and reflect on the stock price,which shows that the company uses text manipulation to carefully cover up financial fraud to obtain short-term gains,and the market's gradual identification process of this behavior also reflects its partial effectiveness.The company uses MD&A text manipulation to carefully cover up financial fraud,which weaken the short-term market response,exacerbate the phenomenon of information drift,and increase the risk of stock price crash.This shows that the company's information strategy to cover up financial fraud has increased the difficulty of identifying the market.Fourthly,this study focused on the financial fraud of the annual report text detection company.Firstly,this study analyzes the shortcomings of existing financial fraud detection models.Secondly,it proposes the importance and feasibility of financial fraud detection in annual report texts,reconstructs text analysis and financial fraud detection-related models,and attempts to detect text data in recent years.Financial fraud has verified the feasibility of detecting financial fraud in the text of the annual report.Finally,the prediction results can be used by regulators to strengthen supervision of capital market information,found and investigated financial fraud problems in advance,and refined the significance of the research in this article for practical work.The possible innovations in this study are:Firstly,this article enriches the research on accounting information disclosure.Existing researches on information disclosure mainly focus on information disclosure quality evaluation,influencing factors and economic consequences.The forms of disclosure are management performance notices,annual reports,quarterly reports and other announcements.However,there are relatively few studies on information disclosure of annual reports.At the same time,foreign research on risk information disclosure focuses on the ability of capital market participants to recognize and perceive risk information.Few studies on the strategic disclosure of non-financial information by companies when they conduct financial fraud.This study also expands domestic risk information.Research in the field of disclosure.Secondly,this paper uses empirical research methods to verify the existing classic financial fraud theory,which is helpful to expand the research on financial fraud motives.The theory of financial fraud motivation has experienced a long period of development,including the iceberg theory,three-factor theory,GONE theory,factor theory,etc.,most of them use standardized research methods to explain the company's financial fraud motives,lacking strong evidence of research methods,this paper uses empirical research methods to verify The rationality of the classic GONE theory is effectively matched with the company's financial fraud motives,which complements the research of financial fraud motives.Thirdly,this paper expands the research on the identification of financial fraud from the perspective of non-financial information texts,and provides a new perspective and method for regulators to detect corporate financial fraud and prevent it from happening.In the context of the era of big data,the company's financial fraud is more hidden,so the predictive ability of financial indicators is limited.It is difficult for individual financial information to identify the motivation of financial fraud and prevent financial fraud.The characteristics of the text disclosed by the management foreshadowed the disclosure method of financial fraud.Compared with financial indicators,the text's role in predicting financial fraud is more effective than that of financial indicators.Existing researches are less focused on the important text disclosure form of annual report text information.MD & A text information provides a new perspective for identifying financial fraud.Therefore,regulators can investigate and punish financial fraud based on the company's MD & A text change rules and combining fraud formation elements. |
PDF Full Text Request