Raising the bar in commodity price forecasting: Evidence from iterated forecast combinations

Forecasting commodity prices is important for investment and risk management, yet developing a unified framework that performs well across commodities remains challenging. Existing studies show that forecast combination methods can be effective in high-dimensional settings, but they focus on one-step combinations and provide limited guidance on balancing the resulting forecast against a stable benchmark. Building on this insight, we introduce an iterated combination framework that adaptively combines traditional forecast combinations with a naive no-change benchmark forecast, integrating macroeconomic, market-level, and technical signals. Using data on eight commodities from 1982 to 2023, we find that the iterated approach consistently outperforms benchmark forecasts, traditional combination methods, and advanced single-model techniques, and delivers meaningful economic gains for investors. We further provide evidence that its superior performance is linked to better capturing energy-market supply–demand dynamics. These findings have important implications for asset allocation and risk management in commodity markets.