Forecasting Disaggregated Producer Prices: A Fusion of Machine Learning and Econometric Techniques

ABSTRACT I propose a novel framework for forecasting euro‐area disaggregated producer prices that blends machine‐learning methods with traditional econometric models. Working at the sector level pays: Pricing behavior is heterogeneous, and no single model dominates across sectors or horizons. This highlights the necessity for a tailored approach that leverages the strengths of various forecasting methods to effectively capture the unique characteristics of each sector. My forecasting exercise has highlighted diverse pricing strategies linked to commodity prices, autoregressive behavior, or a mixture of both, with pipeline pressures being especially pertinent to final goods. Tree‐based methods—random forest and XGBoost—perform strongly, especially where pipeline pressures and input‐cost signals are informative. Newly proposed hybrid RF‐ARMAX and XGB‐ARMAX models are particularly effective for short‐term PPI inflation in commodity‐ and energy‐sensitive sectors.