Characteristics And Detection For Accounting Fraud Pattern Of Listed Companies In China

Accounting fraud (AF) is a deliberate misstatement or amount skipping of deceiving the users of accounting information or financial report, especially the investors and the creditors; or a disclosure of financial reporting of deceiving the users, especially the investors and the creditors. During the past few years, AF made the inventors'loss exceed 10 billion dollars. To a large extent, the robust operation and efficient performance of capital market depends on the quality of the accounting information received from the market. The unreliability of this information can directly cause the inefficacy and morbidity of the capital market. Therefore, AF pattern recognition is extraordinary key for economic growth and national prosperity and breaking out this issue is significantly important.Current studies on AF detection, mainly, adopted the supervised learning by contrasting the fraud vs. non-fraud companies that were selected from the fraud companies and non-fraud companies in the announcement and followed choice based and matched sample research designs. However, decision-making for non-fraud companies is irrational and under future uncertainty on the premise of the nonstandard stock market in China. The reason is as follows: related systems (IPO systems, shares-rationing polices, delisting systems) in stock market has driven enterprise's fraud desire of accounting fraud to be strong and prevalent; social and legal environment, administrative supervision systems, corporate governance structure and accounting standards and systems exit limitation makes enterprise fraud possible; some grave cases reveal fraud disclosure time always drop behind that frauds happen. Positive audit statesment are rectified as denial or comments-reserved after intervention by China Securities Regulatory Commission. Many companies falled into this category like Yinguangxia, Lantian stock, Jinzhou port and Zhengbaiwen. Therefore, the training data would directly cause the unreliability and infidelity of fraud model from supervised learning for classification.The essential of accounting fraud is to disobey accounting law. Obviously, Chinese law is different from the west countries and this will fundamentally cause the discrepancy of fraud types, which further confine fraud data to country– bounded or indigenization, and also raise a new question on AF pattern and detection of Chinese enterprises. Considering the limitation of the fallibility and indigenization of non-fraud data from companies, this research starts from two perspectives: one is to apply economics principles such as contracts and principal-agent to analysis the fraud motivation and fraud regulation about China listed companies; the other is an empirical study on fraud pattern and detecting methods based on fraud data.This dissertation, which chose punishment announcement released by China Securities Regulatory Commission between 2002 to 2006 as the original references and selects ninety fraud companies as samples, makes frequency analysis and statistic test on the fraud types, purpose, means and performance of listed companies and the people involved in the case and penalties situation through the methods of descriptive statistics and the proportion test. Next AF detection-oriented data mining framework was proposed based on the literature review. Under such framework, a taxonomy for detection methods was offered through multi-dimensional analysis of learning approach, detection method, mining algorithms and related technique. Then we proposed the data mining model with a consideration the unreliable characteristics of non-fraud data in China listed companies. On the basis of this model, association mining was used to discover the associated pattern. The Apriori and Apriori PT algorithms were adopted to find out the strong association rules of fraud listed companies from the unsupervised perspective. Above important detection method was given after the explanation on these association rules. Finally, this paper presented the fraud goverance model and designed the corresponding anti-fraud countermeasure from the macroscopical levels and microcopical levels. The research has the following main conclusions:Firstly, the occurrence of accounting fraud behavior is the result of individual behavior tending towards benefit and these behaviors are the regulation context-specific and the behavior of each unit is the results of specific institutional arrangements. The actual environment of Chinese transforming economy and emerging securities markets determines accounting fraud pervasive.Secondly, the two main types of accounting fraud in China listed companies are the violation disclosure of accounting information and fraudulent financial reporting, and the listed companies have no preference between these two types. In the violation disclosure of accounting information, non-integrity disclosure is the main stream; and the omission of the guarantee information and related transactions information is main content in the non-integrity disclosure. The listed companies show their obvious preference at inflated profit compared with the inflated net assets in fraudulent financial reporting (FFR). In inflated profit, fictitious income is the most comment technique and obtained by the inflated sales'income. Among the staff involved fraud in the listed companies, the members in the company's board of directors and top managers are always involve fraud; and it is obviously too light from the punishment intensity.Thirdly, there are strong associations among fraud types. Understated liabilities and inflated net assets are always used simultaneously. Moreover, under the premise of understated liabilities, the probability of inflated net assets would increase nearly four times. And strong association rules also happen among fictitious incomes, understated liabilities and inflated net assets.