BMI and Body Fat Percentage Cannot Be Reconciled

This analysis does not need a population. Both classifiers are closed formulas, so their disagreement is a property of the arithmetic that can be derived exactly rather than estimated. The method is to hold one method's inputs fixed and move the other's.

We used bmi() and bmiCategory() from the BMI calculator, and navyBodyFat() and bodyFatCategory() from the body fat calculator. BMI bands are the WHO adult thresholds; body-fat bands are the American Council on Exercise categories, which are sex-specific.

The reference body is 175 cm, with a 38 cm neck and 88 cm waist for men, and a 32 cm neck, 78 cm waist, and 98 cm hips for women. The sensitivity figures come from moving one measurement a millimetre at a time and recording where the estimate crosses a threshold, rather than from extrapolating a slope — the formula is logarithmic, so its slope is not constant.

This is the single most surprising thing about the method and it is easy to miss, because most body-fat calculators ask for weight somewhere on the page anyway. Look at the equation: it takes the base-10 logarithm of waist minus neck, and the logarithm of height. Weight is not in it. Calculators that ask for it are using it only to convert the percentage into kilograms of fat mass and lean mass afterwards; the percentage itself is decided entirely by your height and your tape. Our body fat calculator does not ask for your weight at all, which is the honest reflection of what the formula reads.

Read either table down the last two columns and nothing happens. Read the first three and BMI travels 10 points across 3 categories. Be clear about what these tables are: they are the degrees of freedom the arithmetic permits, not a claim that a real person can gain 30.6 kg with an unchanged waist. The middle rows are ordinary bodies; the ends are the extreme the equation tolerates. The finding is the direction of travel — added weight is completely invisible to this method, whether it is muscle, fat on the limbs, or water.

That invisibility cuts both ways, and it is why the method is popular with lifters. Add muscle without changing your waist and the tape estimate holds steady while BMI marches you into a worse category. But it also means the method cannot see fat that is not on your waist, neck, or hips, and it cannot see the difference between a 61.3 kg frame and a 91.9 kg one.

Now the mirror image. Fix height and weight — so BMI is frozen — and move the waist. Every row below is the same 76.3 kg at 175 cm, a BMI of 24.9, which sits at the very top of the normal weight band.

That is a spread of 18.9 percentage points in estimated body fat for men, and 12.2for women, at a single unchanging BMI. The men's column passes through 4 of the five ACE categories. Every one of those people would be told the same thing by a BMI chart.

This is the substance behind the familiar complaint that "BMI does not work for athletes". The complaint is usually made in one direction — muscular people classed overweight — but the table shows the other direction is just as available, and clinically it matters more. The person at the bottom of the men's table has a normal BMI and an estimated body fat above the ACE average band. That pattern has a name in the literature, normal weight obesity, and it is associated with worse metabolic markers than the BMI alone suggests.

The tables above isolate one variable at a time. Here are four plausible bodies, each with a full set of measurements, where the two methods return opposite verdicts. These are illustrative bodies, not sampled people — but every measurement is inside an ordinary adult range, and every figure to the right of them is computed.

Take the two men. The lighter one, at 72 kg, has a normal weight BMI of 23.5 and would pass any weight-based screen. The tape puts him at 26% — above average. The heavier one, 23 kg heavier with a thicker neck and a narrower waist, is BMI-obese at 31 and estimated at 14.5% — fitness. Two men, same height, and the methods rank them in opposite orders.

The women's pair is the sharper case, because the inversion survives even within the body-fat method alone. The 62 kg woman is estimated at 32.9% fat and the 82 kg woman at 30.9%. Their waists and hips are identical; the only difference the formula can see is the 4 cm of neck. So the tape says the lighter woman carries proportionally more fat, which is plausible — and a BMI chart says the exact opposite, with 22.8 against 30.1.

Neither result is evidence that one method is right. What the pair establishes is that the two cannot be reconciled by adjusting thresholds or adding a correction factor. They rank people differently because they are looking at different things.

If the tape estimate is the one you trust, it is worth knowing how precisely it has to be taken. The Navy method's published accuracy is roughly plus or minus 3 to 4 percentage points against a laboratory reference. Measurement error sits on top of that.

Two things follow. First, the neck row is the exact mirror of the waist row, because the equation never sees the two separately — only waist minus neck for men, and waist plus hips minus neck for women. Measuring your neck one centimetre too large has precisely the same effect as measuring your waist one centimetre too small. Second, and more usefully: a 4.2cm waist error is enough on its own to swamp the method's entire accuracy budget. A centimetre or two of tape tension, or measuring at the navel one week and at the narrowest point the next, is a real 0.7 to 1.4 point swing that has nothing to do with your body.

The practical rule is that changes smaller than about two percentage points are not distinguishable from measurement noise, and that the same person should take every measurement the same way at the same time of day. A consistent method with a small bias is far more useful for tracking than an inconsistent one with none.

Height is the only thing both methods read, and it is worth checking whether they at least agree there. They do, in sign: at a fixed weight, being taller lowers BMI, and at a fixed waist, being taller lowers the estimated body fat. But the size of the effect on the tape estimate is larger than most people expect.

The same 88 cm waist and 38 cm neck yields 21.9% at 160 cm and 16.8% at 190 cm — a 5.1-point range for men and 7.6 points for women, crossing a category boundary in both cases. That is defensible modelling, since a given circumference represents a smaller share of a longer torso, but it also means an error in your stated height propagates into the estimate.

And here is the scale of what the tape cannot see, expressed in BMI's own terms. At 190 cm, the weight range from the bottom of normal BMI to the bottom of obese runs from 66.8 to 108.3 kg — a 41.5 kg span, none of which enters the body-fat formula.

The honest answer is that the question is malformed, and the tables above are why. These are not two measurements of one quantity, one of them more accurate. BMI is a ratio of weight to height squared, designed in the nineteenth century for describing populations and adopted for screening because it needs only a scale and a wall. The Navy method is a regression fitted to circumference measurements that predicts a body-fat percentage. They correlate across a population and diverge for individuals, and no threshold adjustment can fix that, because they take different inputs.

What the analysis does support is a division of labour:

• BMI is a fast screen, not a verdict. It is reproducible to within a rounding error, needs no technique, and is the number nearly every risk study is indexed against. Its weakness is that it is blind to shape, so it should never be the only number.
• The tape estimate is for tracking, not for a label. Its absolute accuracy is limited, and the sensitivity table shows how easily technique moves it. Measured consistently, the direction it moves over months is more informative than its value on any single day.
• Waist circumference on its own is the underrated number. It needs no formula, requires one measurement, and is what carries most of the signal in the Navy equation anyway. Waist relative to height is a well-studied risk marker in its own right.
• When they conflict, that is the finding. A normal BMI with a high tape estimate points at low muscle mass and central fat, and is worth acting on. A high BMI with a low tape estimate points at muscle. Both are more informative than either number alone.

We have added a note to the results panel of the body fat calculator stating explicitly that body weight does not affect the percentage, what a centimetre of tape error is worth, and that changes under about two points should be treated as noise — because a reader who does not know the first of those cannot interpret the result, and a reader who does not know the second will read technique drift as progress.

We deliberately do not report how often the two methods disagree. That would be the most quotable number on this page and we do not have the data to produce it honestly. A disagreement rate requires knowing how body weight and waist circumference are jointly distributed among adults, and we have no such dataset. Sweeping the two independently, as this analysis does, generates combinations nobody has — which is fine for establishing what the equations permit and useless for estimating what happens to real people. Every figure here is a property of the formulas. None of it is a population statistic.

The illustrative bodies are chosen, not sampled. The four crossover cases were selected to show the contradiction clearly. Their measurements are inside ordinary adult ranges, but choosing them proves only that such bodies are arithmetically possible and physically plausible, not that they are common.

Neither method is a body-composition measurement. The Navy formula is a regression, originally fitted to a military population, and its output is validated against underwater weighing to roughly plus or minus 3 to 4 percentage points. A DEXA scan or hydrostatic weighing would disagree with both methods, and where they all three disagree this analysis cannot say which is closest to the truth. That would need measured reference data, not formulas.

The category boundaries are conventions.The BMI cut-points are WHO thresholds for adults and are known to fit some populations poorly. The body-fat bands are ACE's, are sex-specific, and are not age-adjusted even though body fat rises with age at constant fitness. Several findings above are stated as "crosses a category", and those statements are only as firm as the conventions. The underlying percentage spreads do not depend on them.

The formula has an undefined region. The male equation takes the logarithm of waist minus neck, so it returns nothing usable when the waist is not larger than the neck — which is attainable for a very lean, heavily muscled person. Our calculator rejects those inputs rather than reporting a result, which is the right behaviour but is also a real limit on who the method can serve.

Sex is binary in both classifiers. The Navy formula has separate male and female equations, and the ACE bands are sex-specific, so the analysis inherits that. There is no defined behaviour outside those two sets of coefficients.

Every table can be rebuilt with the two calculators on this site and nothing else. To check the first finding, open the body fat calculator, enter 175 cm with a 38 cm neck and 88 cm waist, and then change the body weight to anything you like. The percentage will not move. To check the second, hold the same height and weight in the BMI calculator and step the waist through the values in the table.

The sensitivity figures need only a calculator: enter your measurements, note the result, add a centimetre to the waist, and note it again. The body fat percentage chart and the BMI chart list both scales across their full ranges if you would rather read the numbers off a table.

The underlying equations are in the sources below in their published form, so a spreadsheet reproduces all of this independently of our code. If your rebuild disagrees with anything above, the contact page reaches a person who will check it.