Menu calorie labeling changes some orders, but results vary

Menu calorie labels sometimes change orders, but findings vary. Learn why fewer calories purchased do not automatically mean less food eaten or weight lost.

日本語で読む（Japanese version）

You have almost chosen lunch from the menu photo when you notice the calorie figure beside it. The neighboring dish has a lower number. Should you switch, or choose the meal you wanted? With a queue behind you, comparing figures can suddenly feel like another task to finish before getting back to work.

It seems reasonable that more information would change the choice. Actual purchase records give a less uniform answer: some studies find a small reduction, others find little change, and reviews do not all reach the same conclusion. A menu can put information in front of you without deciding what you order.[1][2][5]

This article follows the information available just before a food choice. How often you eat out belongs to the separate feature on eating out and takeaway food; how much food is served belongs to the portion-size feature. Here the question is narrower: what do menu numbers change, under which conditions, and how far can those findings take us toward a claim about weight loss?

Menu numbers and scale numbers measure different things

The first question to ask of a study is what it measured. “Calories selected” might mean the energy content of dishes chosen from a hypothetical menu. “Calories purchased” refers to what was bought, calculated from purchase records. “Calories consumed” concerns what people actually ate. Systematic reviews, which gather and assess related studies using a defined method, include these different outcomes. They are connected, but they are not interchangeable.[2][8]

Think of ordering dinner on your phone. If the meal is shared with your family, the total on the order is different from your own intake. Leaving some food on the plate creates another difference. These are everyday examples rather than findings from a particular trial, but they explain why “a smaller order” cannot simply be rewritten as “less food eaten by that person.” Reviews report ordering and consumption as separate outcomes for this reason.[5][8]

There is another measurement gap between a meal and body weight. A change in what someone chooses at one eating occasion does not show what happens across later meals or establish a long-term weight change. The review by Bleich and colleagues, covering 53 studies of consumer and restaurant behavior, concluded that the extent to which labeling encourages lower-calorie purchases remained unclear, as did whether it translates into a healthier population.[2]

Pooled averages can shrink in actual restaurants

Long and colleagues pooled numerical results from menu-labeling studies in a meta-analysis. Across 19 studies, calorie labeling was associated with an average reduction of 18.13 kcal ordered or purchased per meal. The 95% confidence interval ran from −33.56 to −2.70 kcal. The entire interval lay on the side of fewer calories ordered or purchased, and the difference was statistically significant.[1]

The studies did not all produce similar results, however. When the analysis was restricted to 6 controlled studies conducted in actual restaurants, the average difference was −7.63 kcal, with a 95% confidence interval of −21.02 to 5.76 kcal. That was not a statistically significant reduction. “Controlled” means there was a comparison condition, helping researchers distinguish the change associated with labels from other changes occurring at the same time.[1]

The restaurant estimate therefore changes both the size of the difference and the statistical conclusion. Its interval includes a reduction and an increase. The comparison does not establish a reliable reduction under those conditions, even though the average estimate points downward. Reporting only the pooled result from all studies would hide an important qualification about the setting closest to an ordinary restaurant visit.[1]

A separate meta-analysis by Littlewood and colleagues found larger reductions. Their review covered 15 papers published in 2012–2014 and pooled results from 12 studies. Energy consumed was lower by an average of 100.2 kcal, while energy ordered in real-world settings was lower by an average of 77.8 kcal. Confidence intervals were not reported for these estimates.[5]

Among the 15 papers, 9 studies reported statistically significant reductions in energy consumed, ordered or selected. Three papers reported no effect. The review’s quality assessment classified 3 studies as good, 9 as fair and 3 as weak. A pooled average favoring a reduction therefore sat alongside variation in individual findings and study quality. Those qualifications belong beside the favorable result.[5]

The two reviews are not repeated measurements under identical conditions. Their publication windows and combinations of studies differ. Nor are the reductions in consumption and in ordering the same outcome. Reading the 100.2 kcal figure as though it directly contradicts the 18.13 kcal figure would erase both the outcome distinction and the differences in the evidence assembled.[1][5]

A useful reading habit is to keep each estimate inside its own sentence: which review, which set of studies, which setting, which outcome. That lets the smaller result in controlled restaurant studies remain visible alongside the larger real-world ordering estimate in the other review. The disagreement tells us to examine the conditions under which labels operate. A review showing reductions and a review finding smaller differences in actual restaurants both need to remain in the picture.[1][5]

Large purchase datasets also give different answers

Rummo and colleagues’ 2023 study used actual transaction data from a US national fast-food chain. It was a quasi-experimental cohort study: rather than randomly assigning labels to restaurants, the researchers compared differences arising through policy implementation. The sample included 2,329 restaurants, of which 474 had menu labeling, and tracked purchases against comparison restaurants without labels. This design addressed what customers actually bought.[4] Compared with asking people to select a dish in a laboratory, these records were closer to what passed through the checkout. That strength still leaves a difference between the unit recorded by the restaurant and the food eaten by an individual customer.

During the 3–24 months after labeling, purchases in labeled restaurants were associated with 24.7 fewer kcal per transaction relative to the comparison group. The 95% confidence interval, expressed as the size of the reduction, was 23.6–25.7 kcal. The records captured a small difference across many restaurants. However, a transaction is not necessarily one person’s meal, so these results cannot be relabeled as an equivalent reduction in individual intake.[4]

The findings also separated follow-up periods. Purchases were lower by 21.9 kcal per transaction during months 3–12 and by 25.0 kcal during months 13–24. The corresponding 95% confidence intervals for the reductions were 20.9–22.9 and 24.0–26.1 kcal. These results describe the measured periods after implementation, rather than an immediate response or a promise about how much any particular customer would change an order.[4]

The pattern varied within the purchases. The number of tacos purchased declined more than the number of other items, and the reduction in purchased calories was larger at breakfast than at other times of day. An overall transaction average can therefore cover different changes in item choice and different responses across eating occasions. It need not mean that every part of the menu changed by the same amount.[4]

Geography was another important qualification. Of the restaurants, 94.3% were in California. Results there resembled the overall estimates, but restaurants outside California showed no association during the 3–24-month period. The benefit of observing real purchases and the limitation of a geographically concentrated sample exist together. The overall reduction should travel with this regional finding whenever the study is discussed.[4]

A 2025 quasi-experimental study by the same researchers asked a different question. It compared 5,060 restaurants that introduced labeling under the national policy with restaurants that had introduced labels earlier under local policies. Relative to the value predicted from the comparison restaurants’ trend, transactions in the national implementation group contained 7.4 more kcal. The 95% confidence interval was 7.3–7.5 kcal. The authors judged the change in purchased calories to be negligible.[14]

This does not create a simple story in which labels worked in 2023 and became harmful in 2025. The comparison group changed. One study compared labeled restaurants with restaurants without labels; the other compared later implementation with restaurants already displaying labels. The latter addressed the additional change accompanying the nationwide rollout. Publication dates alone do not make the research questions or comparison conditions equivalent.[4][14]

Petimar and colleagues’ natural experiment adds a time dimension. A natural experiment uses a change occurring in the real world, such as a policy or environmental change, as an opportunity for study. The researchers analyzed approximately 67.11 million transactions from 104 restaurants in the southern United States. When the franchise began labeling, purchases dropped by 54 kcal per transaction, followed by a trend increasing by 3.3 kcal per transaction every 4 weeks.[6]

At nationwide implementation there was a further drop of 82 kcal per transaction, followed by a decreasing trend of 2.1 kcal per transaction every 4 weeks. The 95% confidence interval for the nationwide implementation change was −88 to −76 kcal; the interval for the subsequent trend was −2.9 to −1.3 kcal. The initial change and the later direction of movement are distinct parts of the finding.[6]

At the end of the study, the model estimated 73 fewer kcal per transaction, or 4.7% less, than would have been expected without labeling. The 95% confidence interval for the difference was −81 to −65 kcal. This is a model-based comparison with an estimated alternative course of events, not a measurement of what each customer would have bought in two different situations.[6]

The researchers identified possible remaining confounding from factors that changed over time and the absence of individual-level transaction data as limitations. They judged the findings to represent small improvements in the calorie and nutrient content of purchased meals after labels had been in place. Both the favorable endpoint and the uneven course leading to it matter: a single before-and-after number would leave out part of what happened.[6]

Numbers: menu calorie labeling

Across 19 studies: an average of −18.13 kcal ordered or purchased per meal, with variation between studies.[1]

In 6 controlled restaurant studies: an average of −7.63 kcal, without a statistically significant reduction.[1]

Real-world ordering in another meta-analysis: an average of −77.8 kcal, from a different combination of studies.[5]

Across 2,329 restaurants: 24.7 fewer kcal purchased per transaction, with geographic differences.[4]

In the national rollout study of 5,060 restaurants: 7.4 more kcal per transaction than predicted from the comparison trend, judged negligible.[14]