The impact of global integration on crude oil returns forecasting: Introducing a global risk factor

This study introduces a global risk factor as a predictive variable for crude oil returns, assessing its effectiveness relative to a benchmark historical average return model. Three types of forecasting models are employed: autoregressive models, financial variable models, and multivariate forecasting models. These models utilise techniques such as Lasso regression, Complete Subset Regression (CSR), and the Three-Pass Regression Filter (3PRF). Incorporating the global risk factor consistently improves prediction accuracy across all model specifications. The CSR specification achieves the highest directional accuracy and delivers a substantial reduction in mean squared prediction error. The global risk factor captures significant events in international financial markets and enhances forecast stability during periods of high volatility, thereby addressing a key limitation of earlier research that did not adequately account for unexpected shocks. The analysis underscores the importance of U.S.-specific attributes in evaluating crude oil returns within a globally integrated framework.