How accurate is the Navy body fat formula?

Where the formula comes from

In 1984, Hodgdon and Beckett at the Naval Health Research Center measured body density by hydrostatic weighing in several hundred Navy men and women, then regressed it against simple circumferences. The result: for men, waist minus neck plus height predicts body density remarkably well; for women, adding the hip measurement does the same. The Navy adopted it because it needed something administrable to an entire fleet with a tape measure: accuracy per unit of logistics, which is exactly why it works for you at home.

One technical note our test suite enforces: the widely-quoted density form of the equation takes measurements in centimetres, while the equivalent direct percentage forms are defined on inches. Sites that mix these up run 6–7 points hot. Ours cross-checks both forms against each other on every build.

The design choice worth appreciating is what the equations leave out. There is no weight term. A body fat percentage estimate that never asks what you weigh sounds like it should be worse than one that does, and for a single snapshot it probably is, but it means the estimate cannot be dragged around by water weight, a heavy meal or a full bladder, which is precisely the failure mode that makes impedance scales frustrating.

The equations, written out

These are the direct body fat percentage forms as used in the validation literature. All circumferences and heights are in centimetres in the versions below; the division by 2.54 converts each to inches, which is the unit the published coefficients were fitted on. Getting that conversion wrong is the single most common implementation bug on calculator sites, and it produces results several points too high.

The male equation depends on the abdomen minus the neck. Because that difference sits inside a logarithm, the estimate is most sensitive to error when the difference is small, which is to say, in lean, thick-necked people, exactly the group the method is already known to misread. A centimetre of neck error in a man with a 52 cm waist–neck gap moves the result by roughly half a percentage point; in a man with a 30 cm gap it moves it by nearly a full point.

• Men: %BF = 86.010 × log₁₀((abdomen − neck) ÷ 2.54) − 70.041 × log₁₀(height ÷ 2.54) + 36.76
• Women: %BF = 163.205 × log₁₀((waist + hip − neck) ÷ 2.54) − 97.684 × log₁₀(height ÷ 2.54) − 78.387
• All measurements in centimetres; the ÷ 2.54 terms convert to inches, the unit the coefficients were fitted on.
• Men's abdomen is measured at the navel; women's waist is measured at the narrowest point of the torso and the hip at the widest point of the buttocks.
• The output is a percentage of total body mass, which you can turn into fat mass and lean mass by multiplying by body weight.

What validation studies actually find

Across independent comparisons against DEXA and four-compartment models, the standard error of estimate lands around 3–4 percentage points, with mean bias near zero for average builds. Translation: if the tape says 22%, the honest read is 'very likely 18–26, most probably close to 22'. That sounds loose until you compare alternatives. Consumer BIA scales can swing that much before lunch, and the tape's error, crucially, is stable: it points the same direction for the same body every time, which is what makes month-over-month changes meaningful.

The predictable failure cases: very lean, muscular people (thick necks and tight waists push the male formula to overestimate; our calculator flags results under 10% male / 16% female for this reason); people with proportionally narrow waists but high visceral fat, whom it underestimates; and anyone whose measurements were taken sucked-in, over clothing, or at different landmarks each time, which isn't the formula's failure, but produces most 'this calculator is broken' complaints.

The two most useful recent validations were run in military populations, where the equations are used in earnest and sample sizes are large enough to say something. Both compared the same 1984 equations against DEXA. Both found the scatter the literature has always reported. Neither found the same average offset, which is the finding that actually changes how you should use the method.

Why the direction of the error is not random

A regression equation reproduces the population it was fitted on. The 1984 sample was US Navy personnel of the era, a group that was, on average, fitter and younger than the general public and carried a particular relationship between waist size and total fat. Feed the equations a group whose shape differs systematically from that sample and the predictions will be off systematically, in the direction of the difference.

That explains the pattern in the table above. Marines measured at the fit end of the distribution came out close to DEXA, a couple of points low for men. Army recruits measured before any training (a broader, less selected group) came out six points low. The equations were not more or less accurate in either study; they were being asked to describe two different populations with one set of coefficients.

The practical consequence is the opposite of discouraging. Because the error is systematic, it is largely constant for you. Take one DEXA, air displacement or skinfold measurement from a competent operator, subtract your tape estimate from it, and you have a personal correction that stays roughly valid for years. The scatter in the table is the spread across many people; your own reading sits at some fixed point inside it, and a single calibration finds that point.

There is one place this reasoning breaks down, and the Foulis results show it clearly. Over eight weeks of hard training the recruits' body composition changed in ways the circumference equations could not see: the tape registered two-thirds of the men's fat loss and none of the women's. A tape reads shape, and eight weeks of simultaneous fat loss and muscle gain can leave shape almost unchanged. Over four to six months of ordinary fat loss this is much less of a problem, but it is a real argument against reading too much into a six-week block.

Who it misreads

Four groups come out consistently wrong, and knowing which one you are in is more useful than knowing the average error. Very lean, muscular men read high, because the formula interprets a thick neck as a proxy for frame size rather than muscle and a tight waist cannot compensate enough. Women with narrow waists but a high proportion of body fat stored on the hips and thighs read high too, because the female equation adds the hip circumference into the fat term.

In the other direction, people with a narrow waist and substantial visceral fat, a pattern that becomes more common with age, read low, since the tape sees a modest circumference and the equation has no way to know what is behind it. And anyone whose neck is large for reasons unrelated to body composition, including some patterns of muscular development and some clinical conditions, will read lower than they are.

Pregnancy invalidates the estimate outright, as does significant abdominal swelling from any cause. Under-18s should not use it at all: the equations were fitted on adults, and body composition in a growing body is assessed on age- and sex-specific charts instead. Our calculator refuses adult formulas below 18 for this reason.

Getting the most out of it

Measure in the morning, same landmarks, tape snug not tight, normal exhalation, two or three passes averaged. The full technique is in our waist measurement guide. Then treat the number as a calibrated trend instrument: the absolute value carries the ±3–4 band, but the direction of change over a month carries far less, because your personal bias cancels out of the subtraction. If you ever get a DEXA scan, note the difference from your tape reading and apply it as a personal offset from then on.

Three habits do most of the work. First, always record the raw circumferences, not just the computed percentage: if you later discover you were measuring the neck too low, you can recompute the whole history rather than lose it. Second, measure the neck with more care than feels proportionate: it is the smallest number in the equation and therefore the one where a centimetre does the most proportional damage. Third, give the method four to six months before judging it. Eight weeks is inside its blind spot.

And be honest about what a change means. If your tape estimate moves from 22% to 20% over a quarter, the defensible statement is that your waist-to-neck geometry changed in the direction that usually means fat loss, by an amount consistent with about two points. It is not a measurement of two points of fat. That distinction costs nothing to observe and saves a great deal of pointless argument with a scan that disagrees.

What the Navy itself does now

The tape method is no longer the first thing the US Navy applies to its own people. Under OPNAVINST 6110.1L, issued in December 2025 and effective from January 2026, the Body Composition Assessment became a two-step process: a sex-neutral waist-to-height ratio screen first, with the circumference-based body fat calculation applied only to those who fall outside the screen. The maximum allowable body fat limits remain 26% for men and 36% for women.

That is a striking endorsement of the waist measurement rather than a repudiation of the formula. The service that produced the equations concluded that for a first-pass screen, the simplest possible ratio, one circumference divided by height, does the job, and the full estimate is worth computing only when the screen is not decisive. Civilian guidance moved the same way in 2025, when the Lancet Commission on clinical obesity required a fat-distribution measure alongside BMI.

For you, the practical reading is the same order of operations. Take the waist measurement first, because it is a direct measurement with a single threshold and no equation attached. Compute the body fat estimate second, because it converts your measurements into the number most people find intuitive, and because it responds to neck and hip changes the waist alone cannot see. Neither replaces the other, and the Navy's own instruction now says so in policy.

Frequently asked questions

Put it into practice

Run your own numbers through the body fat calculator, the army body fat calculator and the lean body mass calculator. Related reading: Body fat methods, Measure your waist and Track without DEXA.

Sources

1. Hodgdon JA, Beckett MB. Prediction of percent body fat for U.S. Navy men/women from body circumferences and height. NHRC Reports 84-11 and 84-29, 1984. apps.dtic.mil/sti/citations/ADA143890
2. Potter AW, Tharion WJ, Holden LD, Pazmino A, Looney DP, Friedl KE. Circumference-based predictions of body fat revisited: preliminary results from a US Marine Corps body composition survey. Front Physiol 2022;13:868627. doi.org/10.3389/fphys.2022.868627
3. Foulis SA, Friedl KE, Spiering BA, et al. Body composition changes during 8 weeks of military training are not accurately captured by circumference-based assessments. Front Physiol 2023;14:1183836. doi.org/10.3389/fphys.2023.1183836
4. Hodgdon JA, Friedl KE. Development of the DoD body composition estimation equations. NHRC Technical Document 99-2B. apps.dtic.mil/sti/citations/ADA370158
5. Rubino F, et al. Definition and diagnostic criteria of clinical obesity. Lancet Diabetes Endocrinol 2025;13:221–262. doi.org/10.1016/S2213-8587(24)00316-4