The data behind every number
Every figure in this article comes from one free, public source: the August 2021–August 2023 cycle of the National Health and Nutrition Examination Survey (NHANES), the federal survey in which trained interviewers walk each participant through everything they ate and drank in the previous 24 hours. We used the CDC's individual-foods file for the first day of recall, which records each food a person reported, how much they ate, and its nutrient content — including grams of saturated fat.
We kept only recalls the survey itself classifies as reliable and complete, applied the survey's day-one weights so the results represent the US population, and limited the analysis to adults. That left 3,684 adults aged 40 and older (1,331 aged 40–59 and 2,353 aged 60 and older), plus 4,798 adults 20 and older for the all-adult column.
Sorting 100,000 food records into categories
Each food someone reports carries an eight-digit food code. The US Department of Agriculture groups those codes into about 170 "What We Eat in America" food categories — cheese, pizza, burgers, cookies and brownies, and so on. We matched every one of the 100,116 food records in the file to its category using the USDA's published food list for this survey cycle. Not a single record went unmatched.
A category's share is simple arithmetic: the saturated fat everyone got from that category, divided by all the saturated fat everyone ate. We also calculated each category's share of total calories, so you can see whether a food ranks high because it is rich or simply because it is popular.
Checking our math against the government's
Before trusting any new cut of the data, we checked that our pipeline reproduced the numbers the federal government publishes. For adults 20 and older, our build returns the same sample size (4,798), the same average calorie intake (2,036 kcal) and the same average saturated fat intake (27.6 grams a day) as the USDA's official What We Eat in America nutrient table. It also matches the published 12% of calories from saturated fat.
We then checked every cell of the final table for statistical reliability. Each one clears the flagging rule USDA uses in its own tables — no estimate rests on too few people or carries too much uncertainty — so nothing below needs an asterisk.
Categories that sit a few tenths of a point apart are statistically neck and neck. Cheese (5.6%) and ice cream (5.2%) at the top, for example, are effectively tied. Treat small gaps in rank as "about the same," and larger gaps as real differences.