USDA ERS Food Price Outlook
Datadory delivers USDA ERS Food Price Outlook data: monthly forecast vintages of annual US food CPI and PPI percent changes across roughly 22 categories - beef and veal, pork, poultry, dairy, fats and oils - reaching up to 18 months ahead, with annual history back to 1974 and prediction intervals on every figure. Delivered daily, weekly, or hourly.
What is the USDA ERS Food Price Outlook?
The federal government's food-inflation crystal ball, and the only one with five decades of receipts. USDA ERS Food Price Outlook is the Economic Research Service's forecasting program for American food prices: historical Consumer Price Index and Producer Price Index series for food, plus model-based forecasts of annual percent changes reaching up to 18 months into the future. Roughly 22 food categories carry their own line - All food and Food at home down through Meats, Beef and veal, Pork, Poultry, Eggs, Dairy products, Fats and oils, Fresh fruits, Processed fruits and vegetables, Nonalcoholic beverages and the rest.
Every figure ships with its uncertainty attached: each forecast row reports the lower bound, mid point or upper bound of the prediction interval, so a planning range arrives built in rather than guessed at afterward. It anchors the forecast side of the Packaged Foods & Meats catalog. Get a sample of this dataset and lay the category roster against your own SKUs.
What do sample rows from the dataset look like?
One forecast, six identifying facts. Two consecutive-vintage rows exactly as they ship, from the historical CPI forecast series:
Consumer Price Index item : All food
Month of forecast : 7
Year of forecast : 2002
Year being forecast : 2003
Attribute : Mid point of prediction interval
Forecast percent change : 1.9
Consumer Price Index item : All food
Month of forecast : 8
Year of forecast : 2002
Year being forecast : 2003
Attribute : Mid point of prediction interval
Forecast percent change : 1.8Read the pair together and the design shows itself: one vintage later, same forecast year, the mid point moved from 1.9 to 1.8 percent. Swap All food for Beef and veal and the identical six fields hand you the protein line. Vintages stack year upon year - which is precisely what makes scoring forecasts against outcomes possible at scale.
What fields does the dataset include?
Six documented fields, verified against the published forecast-series layout: four identify the observation, one selects the statistic, one carries the number. Nothing inferred, nothing undocumented - the dictionary below is the complete verified schema for the CPI forecast series.
Where does coverage reach?
- Geo: United States, national level - no state, regional or metro cut exists anywhere in the series
- Temporal: Annual percent-change tables running 1974-2025; recent-year change tables covering 2024-2027; historical CPI forecast vintages back to the early 2000s; archived vintages to 2003 for CPI and 2014 for PPI
- Granularity: Monthly forecast vintages of annual, category-level percent changes - one row per item, per vintage, per interval bound
That footprint is the niche: most price series tell you where food inflation has been, while this one is structured around where the forecasting model says it goes - and keeps every past call on record to be graded.
How is the data delivered?
API, files, or your warehouse. Daily, weekly, or hourly.
Pick the categories, pick the cadence, pick the landing zone - the same six-field rows arrive whichever way you take them. The sample comes first, so the category roster, the vintage depth and the archive handling are settled facts before any commitment.
Who builds on it?
- Investors and quant researchers grade the model itself: each monthly vintage can be scored against realized CPI and PPI for meats, dairy and fats, turning forecast accuracy itself into a testable factor for food-margin calls in backtests. The investors quants use cases page covers vintage-scoring setups.
- Demand planners fold forecast food-at-home CPI into volume and revenue models; the demand forecasting use case collects the datasets built for it.
- Grocery pricing teams set repricing guardrails off forecast beef, pork and poultry inflation instead of last quarter's headlines; see price monitoring.
- Market researchers and consultants restate category revenue histories in constant dollars using the CPI and PPI detail before quoting any growth rate.
- Competitive intelligence teams read a rival's price move against forecast category inflation - tracking the pack or taking margin becomes visible. The competitive intel product teams use cases page walks the comparison.
- Journalists, academics and students quote USDA ERS forecasts with clean attribution in inflation explainers and coursework.
Which personas get the most value?
Data scientists and ML engineers engineer food-inflation features from monthly vintages of annual percent-change tables stretching back to 1974 - point-in-time correct by construction. Investors and quant researchers treat the vintage archive as a lab for testing how quickly official forecasts absorbed price shocks. Market researchers and consultants get deflation-safe growth numbers. E-commerce operators calibrate grocery pricing corridors from the meat and poultry lines. Developers and data-product builders drop the flat percent-change tables straight into customer-facing inflation widgets. Journalists, academics and students cite the canonical federal food-price outlook by name.
What should I know before requesting a sample?
Three things worth knowing upfront. First, the series carry percent changes, not price levels - pair them with a level series such as Meat Price Spreads when dollar magnitudes matter. Second, a 2023 methodology replacement retired the 2011-era forecasting approach and the historical series was recomputed; archived pre-2023 vintages were produced under the older method and must not be mixed with the recomputed series. Third, coverage is strictly national - there is no regional cut, so location-specific questions need a companion dataset. Name the categories and years you need and the sample confirms exactly which tables carry them.
Which datasets and notes pair with it?
- USDA ERS Meat Price Spreads - the level detail behind these percent changes: farm, wholesale and retail values for Choice beef, pork and broilers since 1970.
- USDA ERS Food Expenditure Series - where the food dollar actually goes, the spending counterpart to price change; the head-to-head sets out the trade-off.
- USDA ERS Food Availability (Per Capita) Data System - per-capita quantities, the supply side of the same food categories.
- USDA AMS Market News - Livestock, Poultry & Grain - daily wholesale protein reporting; the intramonth texture behind smooth monthly CPI categories.
- Food price forecast, the glossary note - what the category covers and why archived vintages cannot be mixed with recomputed series: food-price-forecast.
For placement in the wider slice, see the ranking of the best Packaged Foods & Meats datasets.
Field dictionary
Every field below is documented against real records. The full dictionary ships with the sample.
| field | type | definition | example |
|---|---|---|---|
Consumer Price Index item | string | CPI or PPI series being forecast; roughly 22 food categories per vintage, from All food down to individual proteins. | Beef and veal |
Month of forecast | integer | Month in which the forecast was issued; combined with the year it identifies the vintage. | 7 |
Year of forecast | integer | Year the forecast was issued - the vintage stamp that lets successive calls on the same target year be compared. | 2002 |
Year being forecast | integer | Calendar year the percent change applies to; the current year, joined by the following year once July passes. | 2003 |
Attribute | enum | Which statistic of the predictive distribution the row reports: Lower bound of prediction interval, Mid point of prediction interval, or Upper bound of prediction interval. | Mid point of prediction interval |
Forecast percent change | number | Forecast annual percent change in the index for that item, vintage and interval bound. | 1.9 |
Questions buyers ask
How far forward do the food price forecasts reach?
Up to 18 months ahead. Forecast tables cover the current calendar year, and from July onward the following year joins them, so a late-year planning cycle reads next year's outlook alongside the current one. Every figure comes as a lower bound, mid point and upper bound rather than a single guess.
How far back does the history run?
Annual percent-change tables for selected CPIs and PPIs run 1974 through 2025; the historical CPI forecast series reaches back to the early 2000s; archived vintages extend to 2003 for CPI and 2014 for PPI. Forecast-versus-outcome studies get half a century of context.
What do the three attribute values mean?
Each forecast is a predictive distribution, and the Attribute field picks what a row reports: Lower bound of prediction interval, Mid point of prediction interval, or Upper bound of prediction interval. Uncertainty arrives native, so planning corridors need no invented error bars.
Which food categories get their own line?
Roughly 22 per vintage: All food, Food at home, Food away from home, Meats, Meats poultry and fish, Beef and veal, Pork, Poultry, Other meats, Fish and seafood, Eggs, Dairy products, Fats and oils, Sugar and sweets, Cereals and bakery products, Fresh fruits, Fresh vegetables, Processed fruits and vegetables, Nonalcoholic beverages and Other foods.
Can archived forecasts be mixed with current-series figures?
No. A 2023 time-series methodology replaced the 2011-era approach, and the historical forecast series was recomputed under it. Pre-2023 archives were produced with the older method and are not comparable, so any backtest should segment them by vintage rather than splice them together.
How do the CPI and PPI sides differ?
The CPI side tracks retail prices consumers face, split into food at home and food away from home. The PPI side captures price changes at earlier stages as processed foods leave the manufacturer. Reading both shows whether upstream cost pressure actually reached the shelf.
See the rows before you pay anything.
Name this dataset and we send real records from it — scoped to the fields you asked for.