Internationalization and earnings management: an artificial neural networks application

Abstract

The growing internationalization process experienced by enterprises, especially in Brazil, with weak investor protection, has provoked discussions about relevant issues such as the level of disclosure and transparency that the company has with its stakeholders. The objective is to analyze the prediction of earnings management from internationalization by means of Artificial Neural Networks (ANN). That is, in this study we used the ANN method as a complement to multiple linear regression, suppressing the problems that may have been caused by distortion, noise and irrelevant data in the database to analyze the impact of globalization on earnings management. Having determined the existence and the effect of the relationship between earnings management and internationalization, the input neurons were stipulated, as were the independent variables of the regression model, which are the proxies of internationalization (DOI, ADR, EXPVENDAS) as well as the output neuron variable depending on the model, which is a proxy for transparency - earnings management (TR4) .For all neural network configurations used, the correct percentage for the training data was 97% and 98% test, proving the effectiveness of Artificial Neural Networks results in management forecast.

Keywords

• Earnings management
• Internationalization
• Artificial Neural Network