What a percentile means
A percentile ranks one value against a whole population's distribution of values. Standing at the 70th percentile for income means seventy people out of a hundred in the comparison group earn less, and thirty earn more; it says nothing about the dollar amount itself, only about where that amount sits relative to everyone else being compared. The 50th percentile is, by definition, the median — the same figure covered in the income and net worth measure guides on this site, expressed here as a rank rather than as a dollar value.
A percentile is only meaningful relative to the specific group it is computed against. "The 80th percentile for income" answers a different question depending on whether the comparison group is the whole Canadian population, a single province, a single age band, or some combination of the three — a figure that ranks in the 80th percentile nationally might rank lower within a narrower, higher-earning peer group, and higher within a broader one that includes retirees and part-time workers.
A percentile also translates directly into the more familiar "top X%" phrasing: the 90th percentile is the boundary of the top 10%, the 99th percentile is the boundary of the top 1%, and so on, with the two phrasings describing exactly the same rank from opposite ends of the distribution. Peer group choice matters just as much under this phrasing — "the top 10% nationally" and "the top 10% for a specific province and age band" are two different lines drawn in two different places.
This is why this app asks for province, age, gender and education before reporting a percentile at all, rather than simply asking for an income figure. Each of those inputs narrows the comparison group the percentile is computed against, and a narrower, more specific peer group generally produces a different — and, for the purpose of self-comparison, more relevant — rank than the broadest possible national comparison would.
Why the top of the range stretches out
Because income is right-skewed — a small number of very high earners pulling the distribution's tail out far beyond where most people sit, covered in the income measure guide — percentiles are not evenly spaced in dollar terms. The dollar gap separating the 10th percentile from the 20th percentile is typically much smaller than the dollar gap separating the 80th percentile from the 90th, even though both gaps span the same ten percentile points. Near the very top of the distribution that spacing widens further still, so the dollar distance between a percentile rank of 90 and one of 99 is usually far larger than the distance between 50 and 59, despite one span covering nine points and the other ten.
This unevenness is exactly why a percentile is a more informative way to describe standing than a raw dollar comparison on its own: it absorbs the stretching-out at the top automatically, reporting a consistent rank even where the underlying dollar gaps are wildly inconsistent in size from one part of the range to another.
How this app estimates a percentile
This app does not have access to a real, published table of income percentile thresholds for every combination of province, age, gender and education it lets a reader select. Instead, it estimates a percentile by modelling the peer group's income as a log-normal distribution — a standard way to approximate a right-skewed distribution mathematically — fitted using only two numbers: the peer group's median income and its average income, both of which are real Census figures rather than assumed ones.
From those two numbers, the model derives the log-normal distribution's two governing parameters directly: the median sets the distribution's centre in log space, and the ratio between the average and the median sets its spread, since a log-normal distribution's average exceeds its median by an amount that grows with how spread out, or skewed, the underlying values are assumed to be. A user's income is converted into a percentile by standardizing it against that fitted curve and reading off the corresponding position on the distribution — the same underlying logic as a z-score, applied to income measured on a log scale rather than a raw dollar scale.
Why log-normal, not real thresholds
Statistics Canada does publish detailed income percentile thresholds for the Canadian population, broken out by many of the same categories this app uses, but only through an interactive online tool, the Census Income Explorer, rather than as a bulk-downloadable data table. Building this app's comparisons directly on that tool's output would mean manually querying and recording a very large number of individual category combinations one at a time, rather than working from a dataset that could be verified, updated and reproduced as a whole in the way the rest of this site's figures are.
The log-normal estimate is this app's way of approximating that same underlying picture from data that is available in bulk: the real median and average income figures by province, age, gender and education, rather than the real percentile thresholds themselves. This is a different situation from the net worth side of this app, where a separate guide describes how a real, published five-point distribution curve is used instead of a modelled estimate — net worth percentiles on this site rest on real breakpoints; income percentiles rest on a mathematical approximation of a distribution's shape built from summary statistics.
What the estimate costs in accuracy
A log-normal curve fitted to just a median and an average is a compact summary of a distribution, not a description of its actual shape. Real income distributions do not always follow a log-normal curve precisely — they can be more concentrated in the middle than a log-normal shape predicts, or carry a heavier top tail than a two-parameter curve can capture, especially within narrow peer groups where a handful of unusually high earners can distort the average without the log-normal fit having any way to detect that distortion beyond its effect on the ratio between average and median.
The estimate is most reliable near the middle of the distribution, where a log-normal shape tends to sit closest to how income actually spreads across most people, and least reliable at the extremes — very low and very high percentiles — where a true distribution's shape can diverge furthest from the smooth curve the model assumes. A percentile reported near 1 or near 99 by this app carries more uncertainty than one reported near 50, even though every percentile figure on this site is generated by the same underlying method and displayed with the same apparent precision.
This limitation is disclosed here rather than left implicit because a percentile figure invites a false sense of exactness — a single number, to the nearest integer, can look like a precise measurement rather than the output of a model fitted to two summary statistics. An income percentile from this app is an estimate of standing built from two real numbers and a mathematical assumption about shape, not a precise measurement read off a published table the way the net worth percentile guide describes for wealth data.
None of this means the estimate is arbitrary. The median and average feeding it are real, verified Census figures for the selected province, age, gender and education, and the log-normal assumption is a widely used approximation for exactly this kind of right-skewed economic data, not an invented shortcut specific to this app. The caveat is about precision at the edges of the range, not about whether the method is a reasonable one to use in the first place.