What the Calculation of Fat Percentage Is Actually Doing
The definition is simple arithmetic: body fat percentage equals fat mass divided by total body mass, times 100. A 180-pound person carrying 32 pounds of adipose tissue is at 17.8%. The difficulty is that nobody can weigh their fat separately at home, so every practical calculation of fat percentage substitutes something measurable — a circumference, a skinfold, a BMI value — for the quantity you actually want.
That substitution happens in one of two ways, and knowing which family a formula belongs to explains most of its behavior.
Direct regression equations skip biology and go straight to the answer. Researchers measured a few hundred people with a reference method, then fit coefficients that map tape measurements onto the reference number. The Navy and Deurenberg equations work this way. The coefficients have no physical meaning on their own; they are the fingerprints of one specific sample of humans.
Two-step density models estimate body density first, then convert density into fat percentage using a separate physics-based equation. Skinfold formulas such as Jackson-Pollock work this way, paired with the Siri or Brozek conversion. This route has one more step but a clearer logic: fat is less dense than muscle and bone, so an average density tells you the mixture.
Both families produce a number that looks precise to one decimal place and carries an error band of three to four points. The photo AI estimator and the Navy body fat calculator both live inside that same reality — the math below is what is running under the interface.
The US Navy Formula, Term by Term
The Navy equation is the most widely used field calculation of fat percentage because it needs only a tape measure. In imperial units, with all circumferences and height in inches:
Men:
BF% = 86.010 * log10(waist - neck) - 70.041 * log10(height) + 36.76
Women:
BF% = 163.205 * log10(waist + hip - neck) - 97.684 * log10(height) - 78.387
Why a difference, not a raw waist. The term waist - neck is the load-bearing idea. Neck circumference grows slowly with fat gain and tracks frame size reasonably well, so subtracting it partially removes skeletal size from the waist measurement. Two men with 36-inch waists but 14-inch and 17-inch necks are not carrying the same amount of fat, and the subtraction encodes that.
Why height appears with a negative sign. A 36-inch waist on a 5-foot-4 frame represents far more relative adiposity than the same waist at 6-foot-3. Height enters as a negative term so taller people get a lower estimate for identical circumferences.
Why logarithms. Circumference does not scale linearly with fat mass. The first inch added to a 30-inch waist represents a bigger proportional change than the tenth inch added to a 45-inch waist. Taking log10 compresses the upper end and makes the relationship close enough to linear for a straight-line fit.
Why women get a hip term. The female equation adds hip circumference because gynoid fat storage puts a meaningful share of total fat below the waist. Ignoring hips would systematically underestimate women with lower-body storage patterns.
There is also a metric form, which returns density-style output before converting:
Men (cm):
BF% = 495 / (1.0324 - 0.19077 * log10(waist - neck) + 0.15456 * log10(height)) - 450
A worked example. A man 70 inches tall with a 15.5-inch neck and a 36-inch waist: waist - neck is 20.5, and log10(20.5) is 1.3118. Multiply by 86.010 to get 112.82. Then log10(70) is 1.8451; multiply by 70.041 to get 129.23. Finally: 112.82 − 129.23 + 36.76 = 20.4%.
A worked example for women. Height 65 inches, neck 13, waist 30, hip 40. The sum waist + hip - neck is 57, and log10(57) is 1.7559, which times 163.205 gives 286.57. Then log10(65) is 1.8129, times 97.684 gives 177.09. So 286.57 − 177.09 − 78.39 = 31.1%.
What a measurement error costs you. Run the male example again with the waist half an inch larger and the result moves to 21.3% — a 0.9 point swing from a tape placement most people would not notice. A full inch of waist error moves the answer about 1.8 points. Neck error is equally powerful in the opposite direction. Height error is milder: one inch changes the result by roughly 0.4 points. This is why measurement technique matters more than which calculator you open.
Body Density and the Siri Conversion
The density route rests on a two-compartment model: your body is treated as exactly two substances, fat and fat-free mass, each with an assumed fixed density. Fat is taken as 0.9007 g/cc and fat-free mass as 1.100 g/cc. Given those two constants and your whole-body density, algebra gives the mixture:
Siri (1961): BF% = (495 / body_density) - 450
Brozek (1963): BF% = (457 / body_density) - 414.2
Where 495 and 450 come from. They are not arbitrary. They fall directly out of solving the two-compartment mass balance with those two assumed densities. Brozek's version uses a slightly different reference body composition, which is why the constants differ.
How sensitive the conversion is. A density of 1.070 g/cc yields 12.6%. A density of 1.050 yields 21.4%. That is nearly nine percentage points of body fat inside a 2% change in density — the whole physiological range of human body density is remarkably narrow. In practical terms, every 0.001 g/cc error in density estimation shifts the fat percentage by about 0.44 points.
Where the assumptions break. Fat-free mass is only assumed to be 73.8% water, with fixed bone mineral content. People who are dehydrated, very lean, elderly with low bone density, or of an ancestry whose average bone mineral density differs from the original reference sample all violate the constants slightly. Siri and Brozek agree within about one point through the 10–25% range and diverge at the extremes, with Brozek generally preferred for obese populations.
Jackson-Pollock Skinfold Equations
Skinfold formulas estimate density from the thickness of pinched subcutaneous fat, then feed that density into Siri. The three-site versions, where S is the sum of all sites in millimeters:
Men (chest, abdomen, thigh):
D = 1.10938 - 0.0008267*S + 0.0000016*S^2 - 0.0002574*age
Women (triceps, suprailiac, thigh):
D = 1.0994921 - 0.0009929*S + 0.0000023*S^2 - 0.0001392*age
Why there is a squared term. The relationship between skinfold thickness and density is curvilinear. Each additional millimeter of pinch means less at high thicknesses than at low ones, because subcutaneous fat eventually spreads rather than deepens. The small positive S^2 coefficient bends the line back up so the equation does not over-predict fat in the heaviest skinfold ranges.
Why age is in there. Two people with identical pinches at 25 and 55 do not have identical composition. With age, a larger share of total fat sits internally and viscerally rather than under the skin, while bone mineral declines. The age coefficient is a blunt correction for that drift.
A worked example. A 30-year-old man with chest 10 mm, abdomen 20 mm, and thigh 15 mm has S = 45. Then D = 1.10938 − 0.037202 + 0.003240 − 0.007722 = 1.06770. Running that through Siri: 495 / 1.06770 − 450 = 13.6%.
The age term in action. Keep those skinfolds identical and change the age to 50. Density falls to 1.06255 and the answer rises to 15.9% — 2.2 points added purely by the calendar. Whether that reflects your individual biology is unknowable from a caliper.
Seven-site versions exist with the same shape and gentler coefficients, and they generally perform better because they sample more of the body. The main practical limit is the operator: caliper technique takes real practice, and two people pinching the same abdomen can differ by 3–5 mm.
The Deurenberg BMI Equation
The cheapest calculation of fat percentage needs no tape at all — only height, weight, age, and sex:
BF% = 1.20 * BMI + 0.23 * age - 10.8 * sex - 5.4
(sex = 1 for male, 0 for female)
Reading the coefficients. Each BMI point adds 1.2 points of estimated fat. Each decade of age adds 2.3 points. Being male subtracts a flat 10.8 points. That last term is the most revealing part of the equation: it is a dummy variable, a single constant applied to every man regardless of build, encoding the population average difference in essential and storage fat between sexes.
A worked example. BMI 25.8, age 40, male: 30.98 + 9.2 − 10.8 − 5.4 = 24.0%. The same BMI and age for a woman returns 34.8% — exactly 10.8 points higher, by construction.
Why it fails predictably. BMI cannot distinguish muscle from fat, so the equation inherits that blindness completely. A lean 200-pound lifter at 5-foot-10 has a BMI near 28.7 and gets an estimate around 27.5%, when a scan might read 12%. Deurenberg's own reported standard error is roughly 4 percentage points, and that is for ordinary body types.
The equation is still worth knowing, because it is quietly embedded in many "instant" web calculators that ask for nothing but height and weight. If a tool gives you an answer without a single circumference, this is probably what it ran.
Putting the Formulas Side by Side
Take one hypothetical man — 40 years old, 70 inches, 180 pounds, BMI 25.8, neck 15.5 in, waist 36 in, and skinfolds of chest 12, abdomen 24, thigh 16 mm — and run every method above.
| Method | Inputs | Result |
|---|---|---|
| US Navy | Height, neck, waist | 20.4% |
| Deurenberg | BMI, age, sex | 24.0% |
| Jackson-Pollock 3-site + Siri | Skinfolds, age | 16.8% |
| Jackson-Pollock 3-site + Brozek | Skinfolds, age | 16.8% |
Four defensible calculations, one body, a 7-point spread. None of them is lying. They were fit on different samples, use different proxies, and carry standard errors of roughly 3.5 to 4 percentage points each. The Siri and Brozek rows land in the same place because the two conversions barely differ in this range.
The practical conclusion is not that the math is useless — it is that the absolute value is the least reliable output of any equation, while the change in that value over time is by far the most reliable. Every source of bias in these formulas is roughly constant for the same person using the same method: your frame size, your ancestry, your fat distribution pattern. Subtract two readings and most of that bias cancels out.
That is why it is worth running an independent second method rather than a second calculator from the same family. Pair the Navy calculator with the photo AI estimator, which infers from visual proportion rather than tape geometry, and add the visceral fat calculator if abdominal distribution is your actual concern. Consistent inputs, one primary method, and a long enough window will tell you more than chasing a decimal place. These are directional estimates for tracking, not medical diagnosis.
Frequently Asked Questions
Which body fat formula is the most accurate?
For tape-only measurement, the Navy equation performs best in most validation studies, at roughly ±3–4 percentage points. Skinfold equations can be more accurate with skilled caliper technique but worse with poor technique. BMI-based equations are the weakest for anyone with atypical muscle mass.
Why does the Navy formula use log10 instead of plain circumference?
Because fat accumulation and circumference growth are not proportional. Logarithms compress large values so a single straight-line fit works across a wide range of body sizes instead of over-predicting at the top end.
What is the Siri equation for?
Siri converts body density into fat percentage using assumed densities of 0.9007 g/cc for fat and 1.100 g/cc for fat-free tissue. It is the second half of any skinfold, BodPod, or hydrostatic weighing calculation.
Can I calculate body fat percentage from BMI alone?
You can, using the Deurenberg equation, but expect an error band around 4 points and much worse results if you carry above- or below-average muscle. Add at least a waist and neck measurement and use the Navy calculator instead.
How much does a measurement mistake change the result?
In the Navy equation, half an inch of waist error moves the answer by about 0.9 percentage points, and a full inch by roughly 1.8. That is often larger than a month of genuine fat loss, which is why identical tape placement matters more than the formula you pick.
Run the Numbers on Your Own Measurements
Understanding the equations is what turns an output into information. You now know why the Navy formula subtracts your neck, why Siri divides 495 by density, and why a BMI-only estimate can be 12 points off for a lifter.
Body Fat Estimator runs this math for you: the Navy body fat calculator for circumference-based calculation, the photo AI estimator for an independent visual read, and the visceral fat calculator for waist-based screening.
See pricing for unlimited runs and saved history. Take your measurements once this week, record every input alongside the output, and repeat in fourteen days — the delta is the number the formulas are actually good at.
