What is it about?
We use machine-learning and deep-learning models to forecast euro area inflation (HICP) and 12 spending categories using monthly Eurostat data (2000–2023). We show which parts of inflation are easier or harder to predict, and how accuracy changes in calm periods versus crisis and shock years—useful for a clearer inflation “weather report.”
Featured Image
Why is it important?
Inflation forecasting matters most when conditions shift fast. Instead of only predicting headline HICP, we benchmark several ML/DL methods across 12 subcomponents and compare performance in different regimes (post-crisis, stable, and recent shock). We also add simple diagnostics (Forecastability Index and error spread) to flag categories that are reliable versus inherently uncertain helping analysts and policymakers focus attention where forecasts are trustworthy.
Perspectives
What surprised me most was how uneven “predictability” is across the inflation basket. Some components behave like steady clocks, while others swing wildly and punish any single model choice especially in 2023-type shocks. Working at the component level made the policy relevance clearer: better forecasts aren’t just about smaller errors, but about knowing where uncertainty is structural and communicating that uncertainty explicitly.
Tibor Bareith
ELTE Centre for Economic and Regional Studies Hungary
Read the Original
This page is a summary of: Enhancing Policy Insights: Machine Learning-Based Forecasting of Euro Area Inflation HICP and Subcomponents, Forecasting, October 2025, MDPI AG,
DOI: 10.3390/forecast7040063.
You can read the full text:
Contributors
The following have contributed to this page
