Earnings Management Of Smes During Initial Public Ofifngs And Its Detection

SMEs board of Shenzhen Stock Exchange has been regarded as a reliable platform for company financing since May2004. Compared with large sophisticated enterprises, SMEs(small-and medium-sized enterprises) with short time establishment and less reputable records, capital shortage plagued its development. Initial Pubic Offings (IPO) can enable companies to raise funds in the capital market, improving their visibility, assisting enterprises to gain greater development opportunities. However, Under CSTS administrative control of SMEs board quantity and size, once companies successfully obtained IPO permission, they can also gain scarcely external resources. Therefore, managers have a strong incentive to engage in earnings management behavior and meet the listed requirements. IPO earnings management leads to excessive polishing finical report, covering corporate financial problem which generating performance decline in post listing period. This will also harm the interests of external investors and company’s long-term development. In this paper, we analysis SMEs IPO earnings management behavior, evaluate key factors impact on earnings management, use support vector machine learning methods to study IPO earnings management, and build discriminate model to detect the effect of earnings management and its mechanisms.The study basically includes three parts:Firstly, we analysis the existence of listed SMEs IPO earnings management and its degree, the specific performance and available tool. We use case study and statistics methods, select sample of236Shenzhen SMEs board listed Companies from January1,2004to December31,2010, the company amount meet the requirement of frequency distribution analysis. Chose same period and industry of main board companies as matched sample, with comprehensive considering high growth of listed SMEs companies, high ownership concentration, as well as low level of IPO regulation, we use game model to analysis the formation mechanism of SMEs IPO earnings management. We found that the size of the issuer’s equity, corporate growth and earnings management costs would lead to earnings management behavior. In addition, considering other factors, such as external government circumstance, internal corporate governance and IPO market conditions and so on. We measure IPO earnings management factors from multiple perspectives, summarize earnings management characteristics, and build earnings management index for further applicably detection.Secondly, the paper introduced support vector machine (SVM model) which is machine learning methods to discriminate SMEs IPO earnings management. We treat company’s earnings management as a classification problem, based on non-earnings management and earnings management IPO companies as sample data for training SVM model by constructing optimal hyperplane Classification of earnings management. When training by choosing a different set of features, selected the optimal set characteristics of earnings management and classification model for identification and the prediction of unknown samples.To test SVM discrimination function, we chose10SMEs earnings management company boarded on Shenzhen stock market as test sample who displayed serious declination in2011. Combining traditional methods to comparatively analysis earnings management discrimination, the results shows that compared with the traditional method of Jones model detection, SVM model can effectively detect the consequence of earnings management.,The results also proved earnings managements multiple reason:enterprise listing year’s cash flow, accruals, corporate governance, earnings management economic consequences and IPO stock market temperature. Regardless of their compliance costs and financial intermediaries’ regulation, it also reveals that China’s securities regulatory authorities for IPO earnings management supervision and punishment is insufficient, financial audit verification function of financial intermediation in the IPO process needs to be improved.Based on comprehensive and multi-angle analysis of earnings management behavior and influencing factors, we scientifically determine the extent of the earnings management. Our studies are not only a supplement for domestic empirical studies on earnings management, but also provide a useful tool for regulation and audit authority. We suggest that regulators should strengthen the extent of the audit for SMEs board, increase earnings management cost, corporate financial penalties for illegal operations. Meanwhile, the paper uses discriminate method of machine learning, artificial intelligence technology, which can provide effective supervision of earnings management tool for regulatory and related financial audit department.