What Is a Good Body Fat Percentage? The Chart Everyone Quotes Was Built From BMI, Called Provisional, and Never Replaced
Here is the first sentence of the paper that every body fat chart on the internet ultimately traces back to: “no accepted published ranges of percentage body fat currently exist.” [1]
That was published in 2000. The ranges it went on to propose are the ones you have been reading ever since, and the authors called them provisional — groundwork, in their words, for someone else to finish. [1]
Nobody finished.
Where the numbers actually came from
The researchers measured body fat in 1,626 adults across three ethnic groups at three universities, using proper reference methods — a four-part body composition model and DXA scanning, not a bathroom scale. [1] That part is solid work.
Then look at how they turned it into ranges. They took the existing BMI cut-offs — underweight below 18.5, overweight at 25, obesity at 30 — and worked out what body fat percentage those corresponded to. [1]
So the body fat chart is a translation of the BMI chart. It is not built from anybody’s heart attacks, diabetes or death. It is built from the numbers somebody already drew a line at, converted into a different unit.
Which is worth knowing before you treat crossing one of those lines as a medical event.
And the equation needs things your chart threw away
The prediction in that paper depends on four inputs: your BMI, your sex, your age, and your ethnic group. [1]
Every chart you have seen gives one column for men and one for women. Two of the four inputs are simply gone. And the prediction, even with all four, carries a standard error of 2.8 to 5.4 percentage points of fat — which is wider than the gap between most of the categories those charts print. [1]
Then there is whatever you measured yourself with
| The ranges | Derived from BMI cut-offs rather than from health outcomes, and called provisional by the authors who built them. [1] |
| The prediction | Standard error of 2.8 to 5.4 percentage points of fat, and the equations need your age and ethnicity — which the chart you found has dropped. [1] |
| Your measurement | A good body-composition analyser sat between 5.0 points below and 3.7 above DXA. Its authors recommend it for groups, not individuals. [2] |
That third row is the one people underestimate. A multi-frequency body-composition analyser — a good one, not a cheap bathroom scale — was compared against DXA in 48 adults. The correlation was strong, 0.921. [2]
But the limits of agreement ran from 5.0 points below to 3.7 above. [2] A reading of 22% is compatible with a real value anywhere from about 17 to about 26. The authors’ own conclusion is that the method can be recommended “for groups of populations” — not for one person. [2]
That is a study of 48 people and we are not going to oversell it. But it points the same way as the arithmetic above: the number is softer than the display on the device implies.
Does the number predict anything at all?
Yes — and here the honest answer gets more complicated than either side of this argument usually admits.
A study of 500,107 UK adults followed for over a decade found every measure of body fat positively associated with getting a chronic disease — cardiovascular disease, diabetes, several cancers, osteoarthritis. [3] Carrying more fat tracked with getting more ill. That part is not in dispute here.
Mortality was a different story. After full statistical adjustment, the only measure still associated with dying was waist-to-hip ratio. Body fat percentage and BMI ran the other way. [3]
We are going to be careful with that, and you should be too. The authors offer it as support for a hypothesis, not as a finding to act on. Inverse mortality results after heavy adjustment are the signature of a long-running statistical argument, because adjusting for things that sit on the causal path between fat and death can manufacture exactly this pattern. It is also one cohort, 94.7% white, aged 40 to 69. This is the kind of result we flag rather than repeat.
What survives all of that is narrower and more useful: where the fat sits appears to matter more than what the total says, and a tape measure round your waist and hips captured that better than a body fat percentage did. [3]
So what is a good body fat percentage?
Nobody can hand you a number, and the people who built the ranges said so first. That is not us dodging the question — it is the answer, and every chart that gives you a confident band is quietly discarding the uncertainty its own source insisted on.
What to do instead, given all three layers of slop:
— Use your own trend, not the category. The same device measuring the same person over months is far more informative than one reading against a chart, because a consistent error cancels out of a comparison with yourself.
— Measure the same way every time. Same device, same time of day, same hydration. Switching methods produces a change that is not a change.
— A tape measure is not a downgrade. In the largest study here, waist-to-hip ratio outperformed body fat percentage for the outcome people care most about. [3]
And the thing charts are actually used for is not health. Most people looking up a target are looking up how they want to look, which is a legitimate thing to want and a different question from whether a number is healthy. It is worth being honest with yourself about which one you are asking, because only one of them has research attached.
Related: a pound of muscle and a pound of fat weigh the same, and the difference is volume, you cannot choose where it comes off, telling remaining fat from loose skin, and why the calorie arithmetic stops working after a few months. For a measurement that does have hard outcome data behind it, see fitness and grip strength.
