Do Self-Reported Vegetarians Eat Meat?

Published: 12 October 2026

Many studies on vegetarian diets start with a simple question: “Do you consider yourself a vegetarian?” A study published in October 2026 in the International Journal of Epidemiology checked how well that answer matches what people actually eat. In a large US survey, most adults who called themselves vegetarian reported eating meat, poultry or fish on the days their diet was recorded. The finding does not say whether vegetarian diets are healthy. It says something about how much weight to put on studies that rely on that one question.

The study

The research team came from Charité – University Medical Center Berlin, the German Federal Institute for Risk Assessment (BfR) and the Boston University School of Public Health. They used data on 11,180 US adults from NHANES, the US national health and nutrition survey, collected in 2007–2010.

Each participant was asked whether they considered themselves vegetarian. Each also completed up to two 24-hour dietary recalls, in which they listed everything they had eaten and drunk the previous day. The researchers counted a diet as vegetarian if it had no meat, poultry or seafood, and compared people’s answers to the question with their recalls.

What they found

Few vegetarians: only 2.2% of participants said they were vegetarian.
Most of them reported meat or fish: about two in three self-described vegetarians (68%) reported eating meat, poultry or fish in their recalls, most of them on both recall days.
The reverse was rare: very few people who said they were not vegetarian (1.6%) had no meat, poultry or fish in their recalls.
Meat-free days do not make a vegetarian: among people whose recalls showed no meat, poultry or seafood, only about one in three (32%) called themselves vegetarian. This does not mean many people were hidden vegetarians. Most of them were meat-eaters who simply had no meat on the recorded days.

The error was also uneven. The authors call this “differential” misclassification: how well people’s answers matched their recalls differed between people with and without each health condition studied.

This is not new. An earlier US survey of 13,313 people in the 1990s also found that self-defined vegetarians “may consume red meat, poultry, or fish”. In that survey, people who called themselves vegetarian and had almost no meat in their recalls made up only 0.9% of the sample.

Same data, different answers

The study’s main aim was to show how much this error can change results. The researchers looked at four health conditions: low bone density (osteopenia), high blood cholesterol, depression and bowel health. They grouped people in three ways: by their answer to the question, by their dietary recalls, and by using both together (vegetarian only if both agreed).

Depression shows the problem most clearly. Compared with non-vegetarians, the odds of depression in vegetarians looked lower when based on their own answer, higher when based on their recalls, and much lower when both were used (odds ratios of 0.76, 1.14 and 0.36; below 1 means lower odds). The same data pointed in opposite directions. None of these results was statistically significant, and the last one rested on only four vegetarians with depression, so it is the least certain of the three.

The researchers then estimated how far the misclassification could shift the results under different assumptions, using a set of methods called quantitative bias analysis. These methods do not reveal the true result. For bone density, the estimate changed noticeably with the assumption chosen, and the highest estimate came from an assumption the data did not support.

The authors stress that the aim of the study “was not to make assumptions about the effect of a vegetarian diet”. The study is about method: it does not show that vegetarian diets harm or protect bones, cholesterol, mood or bowel health.

What this means when you read a study

The authors conclude that defining vegetarians by self-report, by recalls or by both “led to divergent results”. For readers, three questions help:

How was the diet defined? Did the study rely on a single “Are you vegetarian?” question, or did it also record what people ate? In this survey the two methods often disagreed, and neither is a gold standard.
How big is the vegetarian group? In this survey vegetarians were a small group. Results from small groups are less certain, and a small change in who is counted can change the answer.
Do the authors discuss misclassification? The paper cites an earlier review: most systematic reviews in nutrition research (60%) consider this kind of error, but only 13% discuss which way it might push the results or by how much.

The data come from US adults surveyed in 2007–2010, so the exact figures may differ elsewhere. The study also did not look at vegans separately. Our guide to how to read nutrition research covers other questions to ask. Large studies that we have covered, such as the heart health meta-analysis and the study of vegetarians over 65, are worth reading with this in mind.

Limits of the study

The authors list their own limits. Two days of dietary recalls are not a gold standard for identifying vegetarians. The study looked only at errors in diet classification, and assumed the outcomes and other factors were measured without error. Two of the outcomes, depression and bowel health, were also self-reported. The authors also note that people choose a vegetarian diet for different reasons, including cultural and religious ones, which may affect both how they describe their diet and their health.

The authors suggest that future studies could treat diet as a continuous measure, such as how much animal food a person eats, rather than as a strict yes-or-no category.

Sources

Ghoreishi N, Kurth T, Knüppel S, Fox MP. Exposure misclassification of self-reported vegetarian diet and its impact on health outcomes. Int J Epidemiol 2026;55(5):dyag168.
Haddad EH, Tanzman JS. What do vegetarians in the United States eat? Am J Clin Nutr 2003;78(3 Suppl):626S–632S.

Editor: Anıl Güler, Pharmacist
Last reviewed: 12 October 2026