Why population statistics need denominators, not just headlines

A number on its own is not a statistic. "340 cases" means something entirely different depending on whether it's drawn from a population of four thousand or four million, over a month or over a decade, counted once or counted continuously. Strip the denominator away and what's left is a headline — persuasive, memorable, and unverifiable.

This matters more in public health than almost anywhere else, because the headline number is usually the only part that survives the trip from a data release to a news article to a conversation. The denominator, the time window, and the counting method get left behind, and what people remember is the number that sounded biggest.

The shape of the problem

Take a claim like "cases doubled this year." It's true, and it's also compatible with completely different underlying realities:

• A genuine, alarming increase in incidence.
• A small base rate where going from four cases to eight is a mathematical doubling and a rounding error in practice.
• A change in surveillance — more testing, a new reporting requirement, a case definition that widened — that increased detection without any change in the underlying rate at all.

All three produce the same headline. Only one of them is the story most readers assume they're reading. Distinguishing between them requires the denominator (how many people were at risk), the rate (cases per capita, not raw counts), and the method (has the way we count changed?) — three things a headline structurally has no room for.

Why this isn't pedantry

It would be easy to treat this as a statistician's quibble — technically correct, practically beside the point. It isn't, for two reasons.

First, decisions get made on headline numbers. Funding allocations, service prioritisation, and public concern all respond to the number that gets repeated, not the number that was actually most defensible. A rate presented as a raw count can misdirect real resources.

Second, the credibility cost compounds. The first time a "doubling" turns out to be four cases becoming eight in a population of two hundred thousand, readers don't just discount that claim — they discount the next one, even when it's solid. Precision that can't be traced back to a denominator isn't precision. It's a withdrawal against the same trust that made the headline worth writing in the first place.

What we do about it

In practice, this means a small, boring discipline applied consistently: every figure we publish carries its denominator, its time window, and a note on whether the underlying counting method changed. Where we only have a raw count and no reliable denominator, we say so and hold the claim as provisional rather than dressing it up as a rate.

None of this makes for a punchier headline. It makes for a number you can still stand behind a year later, when someone checks.