Nowcasting Ghana's Quarterly GDP using Hybrid Econometric and Machine Learning Algorithms

This study develops an explainable artificial intelligence (XAI)-driven hybrid econometric-machine learning framework for nowcasting and forecasting Ghana's quarterly GDP, addressing the absence of a robust real-time forecasting system for macroeconomic policymaking. Using quarterly data from 2000Q1-2024Q4, the study develops and evaluates 29 standalone and hybrid econometric and machine learning models. Feature engineering and selection were applied to 30 high-frequency domestic and international indicators, identifying the 10 most significant predictors of GDP growth. The proposed FAVAR-XGBoost model emerged as the best-performing algorithm, achieving a Root Mean Square Error (RMSE) of 0.845 and an " # of 0.932, outperforming all benchmark econometric and standalone machine learning models. To address the "black-box" challenge associated with artificial intelligence, the study employs SHapley Additive exPlanations (SHAP) to quantify the contribution of each predictor to GDP forecasts, thereby enhancing model transparency and supporting evidence-based policy decisions. The study contributes to the literature by demonstrating the superiority of hybrid econometric-machine learning models for GDP nowcasting in data-constrained economies, developing a scalable framework for real-time macroeconomic surveillance, and advancing the application of explainable AI in macroeconomic forecasting. The findings suggest that hybrid models can substantially improve the accuracy, interpretability, and policy relevance of GDP nowcasts. It is recommended that the Bank of Ghana, the Ministry of Finance, and other central banks adopt the proposed FAVAR-XGBoost framework to strengthen real-time GDP monitoring and support timely monetary and fiscal policy decisions.