One of these three numbers is a finding. The other two are also the finding.
A study estimates "between 8,000 and 34,000" affected. A week later the news says "around 30,000". What most likely happened?
Yes.A range is hard to headline. Each retelling has to pick one number, and the one that gets picked is the one that makes the sentence worth reading.
Not quite.That happens, and when it does the new work is usually named. A number that hardens without a new source has just lost its error bars.
Not quite.The study said something careful. What is being quoted is no longer what the study said.
Numbers harden as they travel. Each retelling drops a qualifier, and none of the drops is a lie on its own.
Put in order how a careful estimate becomes a hard fact.
Tap them in order — first to last.
A range→The striking end→One number→A known fact
Four steps, no invention, and the uncertainty is gone.Yes.No step invents anything. Each one drops something — the range, then the assumptions, then the source — and at the end the number is both famous and unfalsifiable.
Which of these phrases keep a number honest, and which quietly harden it?
"…estimated at between 8,000 and 34,000"
"…as many as 34,000"
"…on assumptions the authors describe as conservative"
"…the widely cited figure of 30,000"
"…according to a 2024 survey of 900 households"
Yes.Honest phrasing tells you where the number came from and how much room it has. Hardening phrases attach it to nothing — 'widely cited' names no author, no method and no date.
The same finding, retold. Slide through the retellings and watch what is left.
The studyThe press releaseThe news reportThe column, a year later
8 of 8 · qualifiers still attached
The studyA range, a method, a sample size, and the authors' own list of what could be wrong.
4 of 8 · qualifiers still attached
The press releaseThe high end, the institution's name, one caveat near the bottom.
2 of 8 · qualifiers still attached
The news reportOne number, one attribution: "researchers say".
0 of 8 · qualifiers still attached
The column, a year laterThe number alone, in a sentence about something else.
At the last frame the number cannot be checked. Why not?
Yes.A number with no source cannot be wrong, because there is nothing to compare it against. That is what makes a hardened figure so durable.
Not quite.An old study can still be checked. It is the missing pointer back to it that does the damage.
Not quite.It often has not changed at all. What changed is everything around it.
Move the control to see what changes.
A survey of 900 households reports 62%, with a margin of error of 3 points. Two months later a similar survey reports 64%. How many points of real change does that prove?
points
Yes.Both figures sit inside each other's margin. A two-point move between two surveys with three-point errors is a result that is fully consistent with nothing having happened.
What is the most useful question to ask of any striking figure in a news story?
Yes.It reaches both halves at once: the source, and the assumptions. Almost every hardened number fails the second half instantly, because nobody can say what its assumptions were any more.
Not quite.Plausibility is what a hardened number is optimised for. It has already survived several retellings by sounding right.
Not quite.They almost always do, and it means very little — you already know that agreement between outlets is usually one source repeated.
Lesson complete
Numbers harden as they travel, and each retelling drops a qualifier.