Abstract: This study introduces a novel predictive framework for predicting South Africa's macroeconomic trends using national ESG in big data based on AI technology and deep learning. This study utilizes the GDELT database and AI-driven indicator construction methods to extract meaningful insights from 10.76 million news, generating ESG in big data at the national governance level. By combining traditional macroeconomic indicators with national ESG in big data, this study evaluates the predictive performance of econometric, machine learning, and deep learning models. The rolling out-of-sample prediction analysis shows that the LSTM model achieves the highest prediction accuracy. Subsequently, LSTM models with and without national ESG in big data were designed to evaluate the extent to which incorporating national ESG in big data improves prediction accuracy. This study demonstrates that national ESG in big data enhances the accuracy of macroeconomic forecasting, particularly improving the short-term forecasting performance of the models.

Keywords: macroeconomic forecasting; national ESG index; news data; AI; LSTM

JEL codes: C53; E17; O33; Q56

DOI: ...