The ERS Food Price Outlook (FPO) periodically updates and improves the FPO forecasting methods. This page compiles updates and revisions to the data since 2023.
For information about methods used to produce the FPO, see the documentation.
January 23, 2026
In January 2026, ERS resumed FPO updates after the October 1, 2025, to November 12, 2025, Federal government shutdown. The Bureau of Labor Statistics (BLS) suspended Consumer Price Index (CPI) and Producer Price Index (PPI) data collection during this time. BLS did not publish CPI data for October 2025. FPO forecasting methods rely on 13 to 14 years of complete data. Values are needed for each month of each of those years. To generate FPO forecasts, ERS researchers estimated values for the missing BLS data using the autoregressive integrated moving average (ARIMA) model detailed in the ERS report, Time-Series Methods for Forecasting and Modeling Uncertainty in the Food Price Outlook (TB-1957). ERS researchers continue to use these estimated values when forecasting food price changes as CPI data from the BLS remain unavailable for October 2025.
September 25, 2023
In September 2023, ERS revised the historical data series to enable reproducible estimates from the pseudo-random number generation process used to produce the prediction intervals. Monthly forecasts as originally published between January and August 2023 are available in the archived data files.
July 25, 2023
In July 2023, ERS discontinued producing forecasts based on the forecasting methods used since 2011, described in the ERS Technical Bulletin How USDA Forecasts Retail Food Price Inflation (TB-1940). Records of monthly forecasts as originally published are available as archived data files since 2003 for the CPI and 2014 for the PPI.
January 25, 2023
In January 2023, ERS updated the primary forecasting method used in the USDA Food Price Outlook, as described in the ERS Technical Bulletin Time-Series Methods for Forecasting and Modeling Uncertainty in the Food Price Outlook (TB-1957). The food price forecasts produced using these methods are based entirely on statistical models that are fitted to recent trends in the data. The forecasts are presented as a midpoint and a 95-percent prediction interval, a change from the previous approach that presented the forecasts using fixed-width forecast ranges. The prediction intervals vary in size based on the level of uncertainty about the forecast and narrow over the forecast period as more data become available.
The updated methods are described in TB-1957, with the exception that natural log transformations are applied to all series during forecasting. Using natural log transformations ensures all price forecasts are above zero. Following forecasting, exponential transformations allow the forecasts to be included in Equation 1a from TB-1957.
For a summary of the changes to the forecasting methods implemented in January 2023, see ERS Refines Forecasting Methods in the Food Price Outlook.