A data science framework for profit health assessment: development and validation

Abstract

This study introduces a cutting-edge profit health assessment framework that merges signaling theory and agency theory within a data science context, leveraging both financial and nonfinancial indicators to provide a comprehensive, multidimensional evaluation of earnings quality in publicly traded companies. This model transcends conventional earnings management frameworks by focusing on the holistic earnings condition of firms and addressing challenges related to information asymmetry and managerial motives. The empirical analysis utilizes data from small and medium-sized private enterprises listed on China’s SME Board between 2018 and 2022, employing machine learning algorithms to rigorously validate the model’s effectiveness. The results reveal that 85.7% of the penalized firms by the China Securities Regulatory Commission from 2018 to 2023 exhibited low levels of profit health. This data science driven analysis deepens the understanding of corporate earnings health for regulators and investors. The proposed framework not only expands the theoretical underpinnings of earnings management but also serves as a novel evaluative tool for stakeholders, paving the way for broader application across diverse markets and sectors and rethinking earnings quality assessment methodologies through a data science lens.