The European Central Bank published an Economic Bulletin article describing enhancements to its short-term forecasting models for euro area real GDP growth, aimed at improving performance in an environment where recent shocks have made traditional approaches less reliable. The update centres on a revised “workhorse” bridge-equation toolbox, complemented by experimental machine learning methods to help capture heightened volatility, changing relationships in the data and forecast uncertainty. The revised framework moves from six bridge equations to two supply-side bridge equations that forecast GDP growth using value added in industry, services and construction, with the quarterly predictors generated by updated dynamic factor models and vector autoregressive models that combine monthly and quarterly indicators and incorporate stochastic volatility. The predictor set is rebalanced to include newly available hard data such as services production and to reduce reliance on survey indicators whose relationship with activity has weakened, while the models continue to produce both point forecasts and density forecasts and provide a “news” decomposition of forecast revisions by indicator group. In a real-time evaluation for the first quarter of 2022 to the second quarter of 2025, the new ECB models showed bias closer to zero and lower mean absolute forecast errors than the previous framework, and their density forecasts were more accurate as measured by the continuous ranked probability score, although Eurosystem and ECB staff macroeconomic projections were more accurate overall for point forecasts. As a cross-check, a quantile regression forest model delivered comparable post-pandemic performance, improving relative to the workhorse models as more data arrive and providing an interpretable contribution analysis via Shapley values. The ECB indicates it will monitor forecast performance regularly, revise the short-term GDP forecasting models as needed, and continue exploring new data sources and advanced machine learning methods.
European Central Bank 2026-01-15
European Central Bank updates its short-term euro area GDP forecasting framework and benchmarks revised models against machine learning alternatives
The European Central Bank has enhanced its short-term forecasting models for euro area real GDP growth by revising its bridge-equation toolbox and incorporating experimental machine learning methods to better capture volatility and forecast uncertainty. The updated models, which showed improved accuracy in real-time evaluations, will be regularly monitored and revised as needed, with exploration of new data sources and advanced techniques.