Week 1: The digits are a claim you did not make

Why written precision overstates what is known, and why the first question in any estimate is what the decision needs

22 February 2027

Open the slides

Every digit you write is an assertion about what you know. Write eight of them and you have asserted eight, whether or not you meant to.

This is why most weak estimates are not wrong. They are overstated. They claim a resolution nobody established, and they do it silently, in the notation, where no reader thinks to look for a claim. The fix is not more care in the arithmetic. It is asking, before the arithmetic, what resolution the decision actually needs.

Four questions people run together

Precision is how finely a number is written. Accuracy is how close it sits to the thing it describes. Uncertainty is the spread it is really known within, shown or not. Sufficiency is whether it is good enough for the decision in front of you.

Only sufficiency is yours to settle, and it does not depend on the number at all. It depends on what the answer is for. A figure can be imprecise, uncertain and entirely sufficient. A figure can be precise to eight digits and useless.

The ABS is not the villain of this lecture

It would be easy, and wrong, to run this week as official numbers deceive you. The ABS publishes a population clock resolved to the individual person and states on the same page that the projections behind it are not predictions or forecasts. That is disclosure working exactly as it should.

What the disclosure cannot do is change how a number looks. Eight digits read as eight digits’ worth of knowledge, and the reader supplies the overconfidence the publisher explicitly declined to. The lesson is about the reader. It is about you, next time you paste a figure into something.

Two coverage uncertainties, one census

Separately from the clock — a different measurement, of a different thing — the ABS assesses how well a Census covers the population it counts. After 2021, net undercount was 0.7% nationally and 17.4% for Aboriginal and Torres Strait Islander people.

Hold those side by side. The methodology did not change between them. What changed is which population is being asked about, and therefore which decisions the resulting figure can carry. A number that is comfortably sufficient for a national question can be the wrong tool entirely for a question about a smaller population measured with more difficulty — and the digits look identical either way.

What this course will not teach you

No hypothesis tests, no distributions, no interval arithmetic. Not because that machinery is worthless, but because the judgment this course trains sits upstream of it: deciding what needs to be known, how well, and saying so in language a decision-maker can act on. You will express uncertainty in words and ranges all semester, and you will never need a formula to do it.

Outline