FINANCIAL MARKET FORECASTING: A COMPARATIVE ANALYSIS OF ECONOMETRIC MODELS AND MACHINE LEARNING METHODS
FINANCIAL MARKET FORECASTING: A COMPARATIVE ANALYSIS OF ECONOMETRIC MODELS AND MACHINE LEARNING METHODS

Summary/Abstract: This study examines the relative performance of traditional econometric models and machine learning methods in forecasting stock market indices. Using a panel dataset covering ten developed economies over the period 2000–2024, the analysis compares Fixed Effects, Random Effects, and Dynamic Panel models with machine learning algorithms including XGBoost, Random Forest, and Gradient Boosting. The empirical results reveal that the Dynamic Panel model outperforms all alternative approaches, achieving the highest predictive accuracy (R² = 0.8556; MAPE = 6.5%). Among machine learning models, XGBoost provides the best performance but remains inferior to the dynamic specification. These findings highlight the critical role of temporal dependence in financial market forecasting. The study contributes to the literature by providing a unified and systematic comparison of econometric and machine learning approaches within a cross-country panel framework. The results suggest that, despite the growing popularity of machine learning techniques, well-specified econometric models that explicitly incorporate dynamic structures can offer superior predictive performance.

• Details
• Contents