Investigating the Impact of Real Earnings Management on Stock Returns Using a Hybrid Approach of Factor Analysis and Artificial Neural Networks

This study examines the relationship between real earnings management (REM) and stock returns in the Tehran Stock Exchange using a hybrid methodology combining exploratory factor analysis (EFA) and artificial neural networks (ANN). Financial data from 150 listed companies (2018–2023) were analyzed based on Roychowdhury’s (2006) model. Findings revealed that Iranian investors react negatively to downward REM (reducing real activities) as a signal of future risk, while upward REM is associated with lower returns. The hybrid EFA-ANN approach improved stock return prediction accuracy by identifying hidden patterns and nonlinear relationships, outperforming traditional models. Results confirmed the moderating role of firm size, book-to-market (B/M) ratio, and price momentum: small firms and high B/M firms were more susceptible to REM effects, whereas larger firms exhibited resilience due to stronger governance. The innovation of this study lies in integrating econometric methods with neural network technologies to analyze emerging markets, enabling the identification of complex mechanisms through which REM affects prices. These insights emphasize the need for regulators to enhance disclosure requirements for abnormal operational costs and overproduction. Investors can leverage REM indicators and moderating factors to optimize decision-making.
Keywords: Real Earnings Management, Stock Returns, Factor Analysis, Artificial Neural Networks, Tehran Stock Exchange.