欧洲央行-欧元区能源通胀预测的新模型（英）-2025_58页_3mb

Summary of the Working Paper: A New Model to Forecast Energy Inflation in the Euro Area

Core Content

This working paper introduces a new forecasting framework for energy inflation in the euro area, specifically focusing on the Harmonized Index of Consumer Prices (HICP) energy component. The proposed model, called the Short-Term Inflation Projection (STIP) model suite, uses a granular, bottom-up approach by modeling each of the seven subcomponents of HICP energy separately: petrol, diesel and liquid fuels, gas, electricity, heat energy, and solid fuels. The framework is based on Bayesian Vector Autoregression (BVAR) models and incorporates several key features to improve forecast accuracy and robustness.

Main Features of the Model

• Granular, Bottom-Up Approach: Energy prices are modeled separately for each subcomponent, reflecting their unique characteristics in terms of frequency, taxation, regulation, and drivers.
• High-Frequency Indicators: Weekly data for car fuels and liquid fuels is used to provide more timely signals for monthly inflation.
• Pre-Tax Price Modeling: The models use pre-tax price data to better capture the relationship between energy prices and their drivers, as excise duties and VAT significantly influence consumer energy prices.
• Stochastic Volatility: The residual variance-covariance matrix is allowed to vary over time, introducing heteroskedasticity.
• Outlier Correction: Extreme observations are handled using a discrete mixture representation to distinguish between regular and outlier data points.
• Ragged Edge Handling: The framework accounts for publication delays of different indicators by using a state-space representation and simulation smoother.
• Seasonal Adjustments: Seasonal terms are included in certain energy components to improve model fit.

Key Components of the Model

The following are the main subcomponents and their specifications:

Forecast Evaluation

• Data Period: Forecasts are evaluated using real-time data vintages from March 2014 to June 2023.
• Forecast Frequency: Forecasts are produced quarterly, aligning with the Eurosystem/ECB staff macroeconomic projections (B)MPE.
• Benchmark Comparison: The STIP models outperform the Narrow Inflation Projection Exercise (NIPE) forecasts for all horizons except the very short term. This performance is especially notable after the pandemic.
• Conditional vs. Unconditional Forecasts: Unconditional forecasts are better than conditional ones for horizons longer than three months, emphasizing the importance of accurate assumptions for future commodity price paths.

Implications of the Model

• Policy and Practice Relevance: The model is useful for both practitioners and policymakers, especially in assessing the inflationary impact of energy-related policies such as those under the Fit-for-55 package.
• Climate Policy Analysis: The framework allows for the evaluation of the inflation effects of climate change mitigation measures, including carbon taxes under the EU Emissions Trading System (ETS) 2.
• Structural Changes: The model is adaptable to ongoing structural changes in energy markets, such as the decoupling of electricity prices from gas prices and the increasing role of renewable energy sources.
• Model Robustness: The model performs well in capturing the transmission of energy price shocks, with immediate and strong responses to oil price shocks, and more delayed responses to natural gas price shocks. The strength of these responses is level-dependent, with higher pass-through at higher commodity price levels.

Conclusion

The proposed STIP model suite provides a more accurate and detailed approach to forecasting energy inflation in the euro area. By incorporating a granular, bottom-up structure and accounting for key features such as stochastic volatility, outlier correction, and high-frequency data, the models achieve better forecast performance compared to existing benchmarks. The model is also useful for scenario and sensitivity analysis, especially in the context of evolving energy markets and climate policies.