The Denominator Is the Argument
Picture the meeting where someone reports the return rate has fallen to three percent this quarter. Heads nod. Three percent sounds low, so the number does its job and the meeting moves on. Nobody asks three percent of what.
That question matters more than anything happening at the top of the fraction. A rate is two numbers doing one job: a count everyone stares at, and a population everyone assumes. The count gets treated as the argument. The population gets treated as scaffolding, decided before the real discussion started. Somebody decided who counts as the denominator, and that decision is doing at least half the work of the number you're being asked to believe.
Call it denominator amnesty: the habit of accepting the bottom number as given, because contesting it feels like changing the subject. Numerators get audited on reflex. Is that count real, or is it padded. Denominators get a pass, because disputing one means proposing your own population and defending why your line is the more honest one, which is actual work compared to pointing at a suspicious number and demanding receipts. So the original denominator usually wins, not because it held up to scrutiny, just because it was the only one anybody bothered to state out loud.
Unemployment is the example everyone reaches for, worn smooth from overuse, still correct. The rate is a share of the labor force, and the labor force is defined as people currently working or actively looking for work. Stop looking for six months and you exit the denominator entirely. You haven't found a job, the economy hasn't improved, but the rate can fall anyway, because the population it's measured against just got smaller. The count at the top told the truth the whole time. The floor under it moved.
Hospitals run the same trick with more paperwork. A mortality or readmission rate only means something once you've settled who counts as "at risk," and that population comes from a risk-adjustment model built by the hospital reporting the number. Widen the exclusions and a worse outcome can produce a better rate, nothing altered except the boundary around who was ever eligible to enter the fraction in the first place. School proficiency rates run on the same mechanism the moment a district gets to decide which students are exempted from the test that feeds the denominator.
None of this requires anyone to lie, which is what makes it durable. The numerator can be accurate to the decimal and the rate can still misdescribe the world, because accuracy at the top was never what was doing the persuading. The persuading happened at a boundary drawn before the counting started, and boundaries don't announce themselves as claims. They read as setup, so nobody argues with them, and the argument gets won before anyone realized one was underway.
Skip the top number next time one of these gets quoted at you. Ask who drew the line underneath it, and whether they'd have drawn it in the same place if the answer had come out the other way.