The Role of Financial Indicators in the Prediction of Voluntary Carbon Disclosure: A Comparative Analysis with Machine Learning Methods

Since the Industrial Revolution, carbon dioxide emissions and deforestation have been considered the primary causes of climate change. Many countries are developing policies to reduce greenhouse gas emissions and are encouraging firms to disclose and reduce their carbon emissions. This study aims to identify the potential financial determinants of carbon risk awareness, as measured by the willingness to respond to the CDP (Carbon Disclosure Project) survey, among firms listed on the Borsa Istanbul between 2016 and 2023, using machine learning methods. The findings reveal that whether firms will make voluntary carbon disclosures can be predicted with an accuracy rate exceeding 92% using nonlinear, ensemble learning-based Random Forest and XGBoost algorithms in models based on financial indicators. Furthermore, analyses conducted with explainable artificial intelligence tools indicate that specific financial ratios, such as the ratio of equity to total debt, the ratio of fixed assets to equity, and the ratio of long-term debt to total debt, significantly enhance the model's explainability within the XGBoost algorithm. Finally, the study highlights the potential of machine learning algorithms to improve investors' risk analysis in predicting corporate carbon emissions and demonstrates that this finding contributes to both the theoretical and practical development of sustainable investment strategies.