Absolute and relative risk describe the same event and feel nothing alike

Two ways of stating one number

Suppose some outcome occurs in a small number of people out of a large group, and some exposure raises that number. There are two honest ways to report the change. The relative version says how much the risk went up as a proportion of what it was: doubled, up by half, increased by a third. The absolute version says how much it went up in cases per number of people.

Both describe the same underlying fact and neither is a lie. But they behave completely differently in the mind, because relative change discards the starting point and absolute change keeps it. A doubling of something that almost never happens is still something that almost never happens, and a doubling of something common is a serious public health matter. The relative figure cannot distinguish these two situations, and that is not a flaw in the reader.

Why the relative figure dominates reporting

Relative change produces bigger, rounder, more quotable numbers, and it does so without requiring any explanation of the underlying rate. It also survives translation between populations with different baselines, which is a genuine technical advantage in research and one reason the literature uses it so heavily.

The result is a pipeline problem rather than a conspiracy. A study reports a relative measure because that is the convention, a press release repeats it because that is what the study said, and coverage repeats the press release because that is what arrived. By the time it reaches a reader, the baseline that would make the number interpretable has been dropped three times, and nobody involved has to have intended anything.

The question that restores the meaning

Out of what? A relative change becomes interpretable the moment you know the baseline rate, and it stays uninterpretable until you do. If a report gives you a percentage increase and no underlying frequency, you have been handed half a fact, and the missing half is the half that determines whether it matters.

A useful reflex is to convert to counts. Instead of a percentage change, ask how many additional cases per ten thousand people, or per hundred thousand, over what period. Counts carry their own baseline and are much harder to misread. This is the same reason frequency formats improve reasoning about test results: keeping the denominator in the sentence stops it from being forgotten.

The mirror image, and where it is abused

The same asymmetry works in reverse. A benefit can be made to look small by reporting it in absolute terms across a whole population when it is concentrated in a subgroup, or made to look large by reporting it relatively when the baseline is negligible. Both moves are available to anyone with an argument to make.

A related move is the change in relative terms of a number that was itself relative, which compounds the confusion so thoroughly that even careful readers lose track. Percentage points and percentages are not the same quantity, and a shift from two per cent to three per cent is one percentage point and a fifty per cent increase simultaneously. Both statements are true. Only one of them tells you what happened.

What to do with a number you cannot complete

Frequently the baseline is simply not given and cannot be found in the time available. The reasonable response is to treat the claim as unresolved rather than to guess, and specifically to resist the impulse to fill the gap with an intuition about how common the outcome feels. Intuitions about the frequency of rare events are poor, and they are systematically skewed by how often the event is mentioned rather than how often it occurs.

Where the decision is consequential and the baseline is missing, that absence is itself information about the source. Reports written for people who will act on them tend to include the underlying rate, because the writer knows it is needed. Reports written to be striking tend not to.

Why this is worth the effort

Most everyday numerical claims about risk arrive in relative form, and most of the disagreement about them is really a disagreement about baselines that neither party has stated. Two people can hold identical views about the underlying facts and argue bitterly because one is thinking in relative terms and the other in absolute ones.

Getting into the habit of asking for the denominator does not require any statistical training and it resolves a surprising proportion of these arguments before they start. It is also the single most reliable filter for telling careful reporting from careless reporting, which is worth having for reasons well beyond any individual number.