Banca de DEFESA: NOEL KENTINUS OLOKODANA

Uma banca de DEFESA de MESTRADO foi cadastrada pelo programa.
STUDENT : NOEL KENTINUS OLOKODANA
DATE: 30/01/2023
TIME: 08:00
LOCAL: Plataforma teams
TITLE:

Fraud in financial reporting: effects on stock prices and forecasting through the support vector machine

KEY WORDS:

financial statement frauds; event study; market reaction; support vector machine; fraud detection.

PAGES: 108
BIG AREA: Ciências Sociais Aplicadas
AREA: Administração
SUBÁREA: Ciências Contábeis
SUMMARY:

Abstract Fraud occurrences have increased significantly in recent years in many countries (PricewaterhouseCoopers, 2020). Probably, this has negatively affected the efficiency of capital markets, turning them into an environment of distrust, causing investment flight and hesitation in the entry of new investors. According to Hung et al. (2015), corporate fraud does not only affect creditors or shareholders of defrauding firms, but also it affects customers, suppliers, the financial market as a whole, the government; that is, the entire economy. According to Dyck et al. (2010), financial fraud in accounting reports is usually detected much later, after suspicions of the impacts of the crime. It turns out that post factum action (fraud detection) is not efficient in reducing or eliminating occurrences; in this sense, preventive action is relevant and necessary. There are several methods for preventing fraud in financial reporting, and this research aims to analyze the market reaction to fraud without and with its knowledge. For this, the study compared stock price behavior around the date the fraud was committed and around the date it was discovered. Secondly, the study analyzed the accuracy of the support vector machine (SVM) in predicting fraud in financial reports, considering financial indices and nonfinancial indices. The research used event studies methodology to achieve its first objective and SVM to achieve its second objective. Positive abnormal returns were found one day after and from the fourth to the tenth day after financial reporting fraud. Without knowledge of the fraud, the market reacts positively to the performance of fraud in financial reports, generating abnormal gains for the fraudsters. Probably, due to the time it takes for the suspected fraud to be judged by the Securities and Exchange Commission of Brazil (CVM), the market ignores the regulatory agency’s decision when officially recognizing the suspected practice or maneuver as fraudulent, for breaking the law. Sensitive variables to detect fraud risks were identified and the SVM polynomial model was capable to predict fraud risks in at least 64% of fraudulent financial reports. The findings contribute to the academy about the consequences of fraud in financial reports in the market and also encourage other research. The study also contributes to the work of auditors and regulators, recommending the SVM polynomial model to predict fraud in financial reports.

BANKING MEMBERS:
Interno - 2726278 - BRUNO VINICIUS RAMOS FERNANDES
Externo à Instituição - MICHELE RILANY RODRIGUES MACHADO - UFG
Presidente - 1554042 - RODRIGO DE SOUZA GONCALVES