Forecasting Future Corporate Value Including Information on Firm Environmental Activities: Using Machine Learning Techniques

Abstract

This study aims to develop a predictive model of future corporate value by incorporating environmental activity data disclosed in sustainability management reports. Specifically, it examines whether including environmental indicators such as carbon emissions, water usage, and waste recycling capacity enhances the accuracy of corporate value prediction compared to models that exclude such information. To this end, we applied three boosting-based machine learning classifiers (CatBoost, LGBM, and GradientBoost) using both classification and regression techniques. The key contribution of this study lies in integrating firm-level environmental activity information into machine learning-based prediction models an approach that has been rarely explored in accounting and finance research. The empirical results show that models incorporating corporate environmental data consistently outperform the baseline model across all metrics. First, adding carbon emissions data significantly improved prediction performance. Second, incorporating water usage data led to even greater accuracy. Third, the inclusion of waste recycling capacity further enhanced the predictive power of the model. These findings suggest that environmental performance indicators play a critical role in accurately forecasting long-Term firm value and offer meaningful insights for investors, managers, and policymakers.