Evaluating the Impact of ESG and Decarbonization Metrics on Stock Price Prediction

Abstract

This study investigates the effectiveness of incorporating sustainability metrics, specifically ESG scores and corporate decarbonization targets, into stock price prediction models using LSTM neural networks. The aim is to assess whether these non-financial indicators enhance predictive accuracy across different sectors. The analysis focuses on five BIST 100 companies from diverse industries, using data spanning from 1 January 2020 to 19 March 2025. Three model configurations were tested: one based solely on historical stock prices, one with added ESG scores, and another with decarbonization data. The data were preprocessed using normalization techniques and split into training and testing sets to ensure robust model performance. Results were evaluated using MAE, MSE, and RMSE. Findings reveal that sustainability metrics improved prediction accuracy primarily in emission-intensive sectors like aviation (THYAO) and oil refining (TUPRS), while offering limited or even negative impact in others, such as defense (ASELS). Surprisingly, steel producer EREGL showed only modest gains despite expectations of higher sensitivity. Overall, the study shows that the influence of sustainability metrics on financial forecasting varies by sector. It underscores the importance of tailoring input features to fit the unique dynamics of each industry.