재무지표에 기초한 횡령 및 배임 위험평가

초록

The purpose of this paper is (1) to identify financial indicators that can help assess the risk for misappropriation and the breach of trust using the unexpected values of accounting information either in a cross section or in a time series, and (2) to compare and evaluate the performance of logistic regression model and support vector machine in the fraud prediction. The fraud prediction using financial indicators that are readily obtained on the financial statements is expected to help satisfy the information demand of various interested parties. The results show that (1) the unexpected values of key financial indicators in four decision areas (i.e., operating management, investment management, financial strategy, and dividend policies) in the prior period are mostly statistically associated with the fraud detected in the current period, (2) the analysis based on both the probability change of logistic regression model and the hyper-plane weights of support vector machine shows that the profitability-related indicators (e.g., gross profit margins and operating cash flows) are found to be more useful for fraud prediction, and (3) support vector machine significantly outperforms logistic regression model in the prediction accuracy. The results of this paper imply that the risk for misappropriation and breach of trust can be assessed and predicted just simply by observing the accounting information on the financial statements. In addition, the results of this paper also imply the possibility of applying support vector machine to the accounting issues. Support vector machine is believed to be one of classification algorithms with superior prediction accuracy in such areas as business, medicine, and others. However, little research in accounting field applied support vector machine to accounting issues. From the methodological perspective, the discussion of applying support vector machine to accounting issues is able to contribute to accounting literature.

키워드