How accurate are the calories listed on your food?

Four ways people check a food's calories

Most calorie numbers come from one of four places: a label's own math, a database such as USDA FoodData Central, a restaurant menu, or memory ("that was about a sandwich and fries"). Each one is an estimate of a real, hard-to-measure number, and each has a different error built in.

US food-label rules (21 CFR 101.9, via govinfo) let companies calculate calories from the grams of protein, carbohydrate and fat, using about 4, 4 and 9 calories per gram. The total is then rounded, by law, to about the nearest 5 calories up to 50, and the nearest 10 above that. The same rule lets the true, lab-measured value run up to 20% above the printed number before it counts as a violation. USDA FoodData Central, the government's free nutrient database, documents the same approach for most entries. It calculates most energy values using the Atwater general factors of about 4, 9 and 4 for protein, fat and carbohydrate.

What happens outside the lab

Menus and memory are a different story, because nobody rounds consistently for you there.

A 2011 study in JAMA measured the real energy content of 269 items from 42 US restaurants with a bomb calorimeter. It then compared that to the calories stated on the menu. The overall average gap was small, only about 10 calories per portion higher than stated. But that average hid real spread. About 40% of items measured at least 10 calories over, 52% measured at least 10 calories under, and close to one in five items (19%) came in over 100 calories above what the menu said. Items listed at or below about 625 calories tended to measure higher than stated; pricier, higher-calorie items tended to measure lower.

Estimating from memory is harder still. A 1992 study in the New England Journal of Medicine used indirect calorimetry to measure actual intake in obese adults who were not losing weight despite reporting diets of under about 1,200 calories a day. Their measured intake ran an average of about 47% higher than what they had reported. This was not dishonesty; the researchers found people also overestimated how much they exercised. Careful food recall is genuinely hard to do in your head.

A worked example

Picture a day built around the FDA's general reference point of about 2,000 calories: a home-cooked, weighed breakfast logged against a database; a sit-down restaurant lunch read straight off the menu; and a dinner guessed from memory.

The breakfast carries the smallest error. Weighing food and matching it to a database entry mostly just inherits the database's own Atwater-based rounding, a matter of a few calories either way. The restaurant lunch is the wildcard. Per the JAMA study above, if the menu lists it under about 625 calories, the real plate is more likely to run higher than stated, not lower. The dinner is the least reliable. If memory says it was about 700 calories, the Lichtman study's average 47% underreporting would put the real total closer to ≈1,030 calories, if that same gap applied here. None of these numbers is exact. But they are not wrong by the same amount. The weighed, logged meal is probably close, and the recalled one is the one most worth double-checking.

Common mistakes that widen the gap

• Treating a menu calorie count as a lab result. On average menus are close, but any single item can be off by about 100 calories or more.
• Trusting memory for a full day's eating. The 1992 study found even careful, motivated dieters underreported by close to half.
• Assuming every app uses the same database. Different entries for "chicken breast" or "pasta" are not the same food, even with identical names.
• Skipping the scale for home cooking. A database figure is only as accurate as the weight you match it to.

What to do instead

• Weigh what you can. Pairing a kitchen scale with a database such as USDA FoodData Central removes the biggest source of home error.
• Treat restaurant numbers as a rough guide, especially for items on the cheaper, lower-calorie end of a menu, where the JAMA data suggests the real number tends to run higher.
• Be most skeptical of memory-based totals, and write meals down close to when you eat them rather than reconstructing a day later.
• Use a tool for the arithmetic. The macro calculator applies the same 4-4-9 math labels and databases use, so your own totals are at least consistent.
• Read how calories in food are calculated to see why even an accurate label is a rounded estimate, not an exact figure.
• Judge your counting over weeks, not one meal. A single plate being off by about 100 calories matters far less than a consistent daily bias in one direction.