Methodology & sources

Data last reviewed July 2026.

This page explains exactly how a percentile is produced, so you can judge how much weight the number deserves. The short version: statistical agencies do not publish percentiles, so the tool fits a standard distribution to the summary figures they do publish and reads your position off that curve.

The core problem

To say "you are at the 70th percentile" you need to know the whole distribution of a population. Almost no agency publishes that. What they publish are summaries — an average, a median, a Gini coefficient measuring inequality. The tool therefore assumes a distribution shape, fits it to those published summaries, and evaluates its cumulative distribution function at your value.

This assumption is the single largest source of error. It is a good assumption for height, a reasonable one for income and wealth, and a rougher one for BMI.

Income and net worth — the lognormal fit

Income and wealth are strongly right-skewed: most people cluster low, a few sit very high, and the average lands well above the median. A lognormal distribution captures that shape and is the standard choice for modelling income.

For a lognormal, the Gini coefficient depends only on the shape parameter σ, so it can be inverted directly. With the arithmetic mean it fixes the other parameter:

σ = √2 · Φ⁻¹((G + 1) / 2)
μ = ln(mean) − σ² / 2
percentile = Φ((ln(x) − μ) / σ)

where Φ is the standard normal cumulative distribution function. For net worth, UBS publishes both a mean and a median, which pins the same two parameters more directly and without relying on a Gini figure at all: μ = ln(median) and σ = √(2 · ln(mean / median)).

A worked check: at a Gini of 0.40 and a mean of $70,000, the implied median is about $53,200 — and the tool places that median at the 49.8th percentile and the $70,000 mean at the 64.5th. That the mean sits well above the middle is not a bug; it is the skew, and it is why the average income in most countries describes a person who is already better off than roughly two thirds of people.

Height and BMI — the normal fit

Adult height follows a normal distribution unusually closely, which makes it the most trustworthy of the four categories. The calculation is simply percentile = Φ((x − mean) / SD).

NCD-RisC publishes mean height by country and sex but not a per-country standard deviation, so the tool applies a constant: 7.1 cm for men and 6.5 cm for women. These are well-established figures for adult populations, but they are constants, and a country whose true spread differs will produce percentiles that drift toward the extremes.

BMI uses the same normal model with a constant SD of 4.8 for men and 5.6 for women. BMI is genuinely somewhat right-skewed, so the normal model is a weaker fit here than for height — results near the middle are sound, results in the far tails less so.

Why weight is compared through BMI

The WHO publishes mean BMI, not mean body mass. There is no published distribution of raw kilograms by country, so comparing your weight directly against one is not possible. Instead the tool asks for your height, computes BMI = kg / m², and compares that. The result is labelled as a BMI percentile throughout, because that is what it is.

BMI is a population-level statistic. It does not distinguish muscle from fat and was never designed to assess an individual. A percentile here tells you how your BMI compares with others', and nothing about your health.

Currency conversion

Every monetary figure in the underlying datasets is US dollars. For income and net worth you can switch the input between USD and your country's own currency; entering a local amount converts it to USD before any comparison is made, and the result is then converted back for display. The percentile itself is identical either way — conversion changes the units on screen, not the answer.

Rates come from a public exchange-rate feed, cached for twelve hours on our side, and the rate used is shown beneath every converted result. If that cache is unavailable the browser fetches the same feed directly, so the toggle keeps working. Two limits are worth knowing:

• It is a market rate, not purchasing power. Converting a salary at the market rate does not tell you what that money buys locally. A PPP-adjusted comparison would be a different, and for some questions more useful, calculation.
• Volatile currencies age badly. For a currency moving quickly against the dollar, a rate cached for hours can drift noticeably.

Where a rate is unavailable, the local option is disabled and amounts stay in USD rather than being converted at a stale or guessed rate. Countries that already use the US dollar get no toggle at all.

What age does and does not do

The tool asks for your age and uses it to confirm you are being compared against an adult population, but it does not narrow the comparison to your age group. The underlying figures are age-standardised across all adults, so a 25-year-old and a 65-year-old in the same country are compared against the same curve.

This matters most for income and wealth, which typically rise across a working life and fall in retirement. A 25-year-old's income percentile against all adults will understate their standing among their peers; an older person's will often overstate it. Age-stratified data exists for some countries and categories, and incorporating it is the most valuable improvement this tool could make.

Why gender is asked for some comparisons and not others

Height and BMI differ enough between sexes that a single combined curve is bimodal, and a percentile read off it is close to meaningless. NCD-RisC and the WHO both publish these figures separately for men and women, so the tool asks and uses it. Selecting "prefer not to say" pools both sexes — the tool averages the means and widens the standard deviation to absorb the between-sex variance. That is a deliberately conservative approximation, and results are correspondingly less precise.

Income asks, and uses it where a pay gap has been published. The World Bank publishes only a single Gini coefficient and a single GNI per capita per country, so the national curve itself is not split by sex. The ILO separately publishes a gender wage gap — the difference between average male and average female earnings — for most countries. Where one exists, the tool shifts the national curve into a male and a female version in that ratio.

The shift is calibrated so the two halves, weighted equally, still average to the published national figure. That has a useful consequence: choosing "prefer not to say" returns exactly the national result, and a country with no measured gap produces an identical curve for everyone. The adjustment can only move people relative to each other, never move the country as a whole.

The assumption to be aware of: both curves keep the national spread. A pay gap is a difference between two averages and says nothing about the shape of either distribution, so the tool assumes inequality within each sex matches the country's overall inequality. That is almost certainly not exactly true. It is the weakest step in this calculation, and it is why the adjustment is applied only where a gap is actually published rather than estimated from a neighbour.

Currently 35 of the 49 bundled countries have a usable figure. Readings older than 2015 are discarded — the raw series contains values from the 1980s that say nothing about today — as are a handful of implausible outliers, including one country-year reported at −125%. Countries without a figure, among them Germany, Japan, Canada and Australia, fall back to the national distribution and the result says so explicitly.

Net worth does not ask. UBS publishes one mean and one median wealth per adult, and there is no comparable per-country series on wealth by sex to layer on top. Gender wealth gaps are real and well documented in research, but nothing exists at this coverage to measure them country by country, so no adjustment is made and none is implied.

Sources

Every lookup is served from a bundled dataset first, so the tool works instantly and never breaks when an upstream API is slow or down. Where a live source exists, fresher figures are fetched and overlaid. Each result is labelled live or bundled so you always know which you are looking at.

The biggest caveat: income means national income

GNI per capita is a country's total national income divided by its population. It is not the average wage. It includes corporate and government income and is pulled upward by capital income concentrated at the top, so it runs considerably higher than what a typical worker earns.

The practical consequence: your income percentile here is conservative. A salary compared against national income per head will rank lower than the same salary compared against an earnings-only distribution. Read it as "where I sit in the national income distribution", not "where I sit among wage earners".

Countries with partial coverage

Income and BMI are available for most countries through the live APIs. Height and net worth are only available for the countries in the bundled dataset — roughly 47 of the largest economies and populations. For any other country those two categories report no data rather than substituting a regional average, because a fabricated comparison is worse than none.

Known limitations, in order of severity

1. Income is modelled from national income per head, not personal earnings.
2. Age is collected but does not narrow the comparison group.
3. Height and BMI use constant standard deviations, not per-country ones.
4. The lognormal fit understates concentration at the very top of the wealth distribution.
5. Net worth figures are transcribed from an annual PDF and carry the review date above.
6. BMI is mildly right-skewed but modelled as normal, weakening the tails.
7. Local-currency amounts convert at a market rate, which is not purchasing power.
8. The gender pay-gap split assumes each sex shares the national spread, and covers only 35 of 49 countries.

← Back to the tool