Restaurant vs. Grocery Price Inflation Since 2020

This comparison uses the BLS Consumer Price Index for All Urban Consumers (CPI-U), read after the October 14, 2026 release. The BLS September release reports the latest annual changes; the longer comparisons below are our calculations from its published index levels. “Eating out” means the broader food-away-from-home category, not a survey of independent restaurants alone.

How much have restaurant prices risen since 2020?

A lower annual inflation rate does not mean prices returned to their 2020 level. The 3.5% annual eating-out increase is a change from September 2025; 37.6% is the accumulated change from February 2020. The cumulative eating-out increase exceeds the grocery increase by 5.8 percentage points. That difference does not say that a restaurant meal costs only 5.8% more than cooking it at home: these are separate baskets and indexes, not the dollar prices of the same dinner.

Compare 23 local areas without mixing their dates

BLS publishes the restaurant-price category monthly for New York, Chicago and Los Angeles, and every other month for the other areas below. Food-at-home observations may be available more frequently; we use the same dates as the eating-out series in each comparison. January is the last available pre-March-2020 restaurant observation for the odd-month areas; February is used for the others. September is the latest common month for the monthly and odd-month groups, and August for the even-month group.

The three tables deliberately have different date labels. Read the two categories within an area together; do not turn these unequal windows into a national ranking. Local names are abbreviated and may cover several counties or states. The BLS regional directory identifies the exact areas; each row links to its eating-out series. Urban Alaska and Urban Hawaii retain their BLS area names.

Monthly areas: February 2020 to September 2026

Odd-month areas: January 2020 to September 2026

Even-month areas: February 2020 to August 2026

New York illustrates the distinction: its eating-out index increased 36.8% and groceries 29.7% over February 2020–September 2026. San Francisco shows the reverse ordering over its own February 2020–August 2026 window: 34.3% for eating out and 37.8% for groceries. Neither comparison identifies the causes, measures an individual restaurant’s margin, or establishes which area has the more expensive meal.

What these numbers can and cannot tell an operator

• —Consumer prices are not your ingredient costs. Grocery CPI describes consumer food purchases. It does not measure a restaurant’s distributor quotes, specifications, pack sizes or usable yields. An ingredient may move differently from the aggregate basket.
• —A regional index measures change within its area. BLS warns that local index levels cannot compare the cost of living or price levels across areas. A larger cumulative increase does not make an area the most expensive. Local estimates also have smaller samples than the national index.
• —Eating out is broader than a printed menu. The category includes full-service and limited-service meals, as well as other away-from-home food. BLS includes applicable sales taxes and mandatory service charges; it is not a clean measure of your pretax menu-price edits.
• —A price increase is not a profit increase. These series contain no evidence about your sales volume, food-cost contribution, rent, payroll or customers’ willingness to pay.

These interpretation limits follow the BLS CPI questions and answers, its item definitions, and the CPI technical documentation. Keep the market benchmark beside your own records instead of substituting it for them.

A dish can earn more dollars and still have a higher food cost percentage

Consider this hypothetical dish. Its recipe and portion are held constant for the example; actual supplier inputs, quantities and yields would need review. Selling prices below exclude sales tax, tips and delivery charges. The example numbers are assumptions, not estimates generated by applying CPI to a restaurant.

The menu price rose 20.0%, while the assumed ingredient cost rose 35.7%. Ingredient contribution increased by $1.50 per plate, yet food cost took a larger share of the selling price. That contribution is still before labor, occupancy and other expenses; it is not net profit. A chosen 30% food-cost target would imply $5.70 ÷ 0.30 = $19.00 before rounding. It does not establish that customers will buy the dish at $19 or that 30% is the right target for this business.

Start with a complete recipe cost, including garnishes and usable yield. Compare the resulting price with your menu and actual sales evidence using the menu pricing method. If you decide a change is warranted, the price-increase notice guide helps explain it without turning a national inflation figure into a claim about your own costs.

Sources, dates and calculation method

1. 1.We read the October 14, 2026 BLS bulk release: observations, series metadata, and area definitions. These files update in place; the tables here describe the dated snapshot and periods shown, not a live feed.
2. 2.Select CPI-U, unadjusted, regular monthly-period records: series prefix CUUR, item SEFV for food away from home and SAF11 for food at home. The national full- and limited-service items are SEFV01 and SEFV02. Local area codes appear in each linked series ID.
3. 3.Calculate 100 × (end ÷ start − 1) from the published index levels and round once to one decimal. For a 12-month comparison, use the same month one year earlier. Exclude M13 annual-average records. No monthly percentages are added together.
4. 4.Unavailable observations stay unavailable. The October 2025 rows in this selection are absent or marked X with no numeric observation. We neither insert zero nor interpolate that month; it is not an endpoint in these calculations. A missing month prevents treating an incomplete year as an ordinary 12-month average.
5. 5.As a second extraction check, we compared 845 monthly records from 12 selected series with the BLS public API, including national, New York, Boston, Detroit and San Francisco series. Values matched. This checks extraction consistency, not sampling precision or any restaurant’s operating results.