Pioneer Jones vs the modifiers: case of detecting accrual-based earnings management using advanced machine learning classifiers in an emerging economy

Abstract

Purpose The purpose of this study is to investigate the potential improvement in the intelligent classification of opportunistic accrual-based earnings management (AEM) using advanced machine learning (ML) algorithms by incorporating different Jones-based measures for the target variable, including the standard Jones (1991) model and its most-cited modifications; Dechow et al. (1995) and Kothari et al. (2005). Design/methodology/approach Using the design science research framework, the study examined the performance of ML algorithms in classifying the AEM in a sample of non-financial firms listed on the Egyptian Stock Exchange from 2016–2022. The classification models were developed with a set of financial features and three different sets of target variables. The paper uses three advanced ML classifiers, Extreme Gradient Boosting (XGBoost), Gradient Boosting and Random Forest to classify multi-class AEM. Findings Acknowledging a significant improvement in the detection accuracy, Kothari et al. (2005) demonstrated superior performance compared to earlier Jones-based models when used as an AEM target variable proxy in developing all ML classifiers, especially the XGBoost classifier. Originality/value This study significantly contributes to the field by establishing pioneering evidence for the most accurate measure of the AEM target variable among the most cited Jones-based models. This allows for the effective classification and prediction of AEM with higher precision using advanced ML classifiers.