Evaluasi Keakuratan Model Beneish M-Score Sebagai Alat Deteksi Kecurangan Laporan Keuangan (Kasus Perusahaan Pada Otoritas Jasa Keuangan di Indonesia)

Abstract

This research has been conducted aiming to see in more detail whether the fraud detection model that has been used so far, the Beneish M-Score, is capable of being one of the tools to see financial report fraud occurring in the business world. This is interesting to study considering that many companies in the Financial Services Authority (OJK) in Indonesia receive warnings and even fines for the delay in submitting financial reports to Capital Market Authority. To carry out the analysis process as in the objectives in the first paragraph, the research team took a sample of 23 companies on the Indonesia Stock Exchange, where the companies were in the list of OJK. The 23 companies that were sanctioned by the OJK compared to 23 not sanctioned companies. In sample of companies that were sanctioned by the OJK, the number of non-manipulator companies according to the Beneish M-Score calculation was 62% and for companies included in the manipulator classification only 38%. Whereas in the sample of companies not subject to sanctions from the OJK, the number of companies included in the non-manipulator category is actually smaller, 52%, calculated using M-Score. This is the main basis for further research. In this study, the analysis process is carried out by quantitative explanatory analysis using probit regression models (probit models), on financial statement data which are categorized into two, the financial ratio with original data from OJK (audited) and the financial ratio with data modification (advanced business analysis). The results show that Beneish M-Score Model could not be implemented effectively to detect the fraud in the companies under control by OJK because only 2 (two) variables influence the existence of fraudulent, are Asset Quality Index (AQI) and Total Accrual To Total Assets (TATA). Thus, it is appropriate and important for the Beneish M-Score modeling to be equipped with other models that are more able to explain.