Abstract:

The main objective of this study is to determine whether companies engage in financial manipulation. For this purpose, 16 companies operating in Borsa Istanbul between 2018 and 2021 were selected. Eight of these companies consist of those that received administrative fines. The companies' receipt of administrative fines was obtained by examining the Capital Market Board bulletins. The other eight companies were randomly selected from similar sectors and sub-sectors that operated in Borsa Istanbul during the same period. With this information, for each company, it was determined whether they engaged in financial manipulation for the operating year. The company is assumed to be classified as a manipulator in the year it receives an administrative fine and the year preceding the administrative fine while being classified as a non-manipulator in other years. Initially, a comparison was made between the Beneish M-Score model, which is one of the models used to identify financial manipulation, and the classes obtained from the Capital Markets Board. Due to the differences identified from this comparison, two models were created using the Random Forest machine learning method to predict these classifications. In one model, only the financial indicators from the Beneish M-Score were included, while in the other, along with these variables, GDP growth, inflation, and exchange rate were also incorporated. The empirical findings derived from the study suggest that the model enriched with economic variables outperforms the traditional Beneish M-Score model in detecting financial statement manipulations.

Keywords: