20160126-USDA-How_USDA_Forecasts_Retail_Food_Price_Inflation_43页_2mb

Summary of USDA Retail Food Price Inflation Forecasting

Core Content

This technical bulletin outlines the methodology used by the U.S. Department of Agriculture's Economic Research Service (ERS) to forecast retail food price inflation. It details the evolution of ERS's forecasting techniques, the factors influencing food prices, and the performance of the updated methods compared to previous ones.

Main Points

• Purpose of Forecasts: Retail and wholesale food price forecasts are crucial for farmers, processors, consumers, and policymakers to understand future price trends and make informed decisions.
• Forecasting Timeframe: ERS provides 12- to 18-month forecasts for 7 farm, 6 wholesale, and 19 retail food categories.
• Methodology Update: In 2011, ERS updated its forecasting method to incorporate more rigorous statistical techniques and account for the multistage U.S. food supply system.
• Forecasting Approaches:
  • Vertical Price Transmission (VPT): Uses input price forecasts from earlier stages of production to predict retail prices.
  • Autoregressive Moving Average (ARMA): Relies on historical CPI data and a time trend when VPT is not feasible due to data limitations.
• Performance Evaluation: The new methodology shows modest improvements in forecast accuracy and precision, with fewer and smaller revisions needed compared to the previous method.
• Structural Breaks: These are considered in time series analysis to improve forecast accuracy. Tests are used to detect structural breaks, which may occur due to policy changes, trade dynamics, or external shocks.
• Data Sources: ERS uses farm, wholesale, and input price data, along with the Bureau of Labor Statistics (BLS) relative importance shares to weight CPI subcategories in aggregate forecasts.

Key Information

• Forecast Accuracy:
  • The average difference between initial forecasts and actual CPI values was reduced from 2.6 to 2.0 percentage points.
  • The average size of forecast revisions dropped from 2.6 to 2.1 percentage points.
• Forecasting Models:
  • Four vertical price transmission models are used, with variations in methodology (e.g., Error Correction Model (ECM) and Autoregressive Distributed Lag (ARDL)).
  • For categories with reliable data, the VPT method is preferred; otherwise, ARMA is used.
• Structural Breaks:
  • Structural breaks are identified using statistical tests, such as the Sup-Wald test.
  • Dummy variables are used in regression models to incorporate structural breaks.
• Forecasting Process:
  • The VPT method involves passing through price forecasts from farm to wholesale to retail levels.
  • Farm PPI is forecasted using farm price projections, which are then used to predict wholesale and retail prices.
• Data Limitations:
  • Some categories only have quarterly farm price data, which is adjusted to monthly for forecasting purposes.
  • The BLS provides relative importance shares to weight CPI subcategories in aggregate forecasts.

Forecasting Methodology Overview

ERS employs two main forecasting approaches:

• Vertical Price Transmission (VPT):
  • Uses input prices at each stage of the food supply chain.
  • Incorporates farm, wholesale, and retail price data.
  • Includes Error Correction Model (ECM) and Autoregressive Distributed Lag (ARDL) techniques.
• Autoregressive Moving Average (ARMA):
  • Relies on historical CPI data and a time trend.
  • Applied when data for VPT is insufficient.

Forecast Performance

• Accuracy Improvement:
  • The revised methodology improved forecast accuracy compared to previous univariate approaches.
  • Performance was evaluated over 2011-2013 for VPT and 2003-2010 for ARMA.
• Revisions:
  • The number of revisions per food category decreased from 3.7 to 3.2.
  • The average size of revisions also decreased, indicating more stable forecasts.
• Limitations:
  • More data is needed to fully assess the long-term performance of the new methodology.
  • Structural breaks remain a challenge, as they can significantly affect price trends.

Conclusion

ERS continues to refine its forecasting methods to enhance accuracy and reliability in predicting retail food price inflation. The updated approach, which incorporates vertical price transmission and structural break analysis, has shown improvements in forecasting performance. Future work will focus on further refining these methods and integrating new data sources to better capture the complexities of the U.S. food supply system.