Week 10: The one thing forecasting adds

Why a claim about the future needs no new machinery beyond what you already have, and the one structural difference that changes everything

10 May 2027

Open the slides

Nine weeks have built one skill, applied to numbers that were always already sitting somewhere, waiting to be found. Today the skill does not change. The thing it is applied to does.

No new machinery

There is a temptation to treat a claim about the future as requiring a different kind of thinking altogether — models, trends, some apparatus this course has not given you. Resist it. A chain of assumptions is still a chain of assumptions whether the quantity at the end of it has already happened or has not. A range still needs a low bound, a high bound and an honest width whether the truth is sitting in an archive or has not yet occurred. Nothing this course has taught you stops applying the moment the target moves into next month instead of last year.

The one variable that does change

What is genuinely different is not the reasoning. It is the check.

For nine weeks, every estimate this course has asked you to make could, in principle, be checked the moment you finished making it — the number was sitting somewhere, and looking is all that stood between your range and the answer. A forecast removes that option entirely. There is no archive to consult, because the fact your range is about has not happened yet. The only way to know whether a forecast was any good is to wait for the date it named and then look.

That single change has a large consequence: a forecast cannot be evaluated by argument, however careful. Two people can each present a reasoned case for opposite ranges, and no amount of further reasoning settles which was closer without the world actually arriving at the date in question. Quality, for a forecast, is not a property you can establish at the time you make one. It is a property a record establishes later, once the date has passed and there is something to compare the range against.

Why the record has to exist before you know you need it

This creates an obligation that has not applied to anything else in this course. A retrospective estimate can be judged the moment it is made, because looking up the answer is always available as a check. A forecast can only be judged by whoever kept a record of what was claimed, precisely enough, before the outcome existed to bias the memory of it.

Write a forecast down honestly today and forget to keep it, and by the date it resolves you will have no way to know whether you would have been impressed with yourself. Memory reconstructs confidence to match outcomes without being asked to. A sealed, dated record is the only defence against your own hindsight, and it has to be built at the moment of commitment or it does not exist when you need it.

A public institution, scored by its own record

The Reserve Bank case this week is not a lesson in economics. It is a lesson in what becomes possible once an institution puts a specific, dated, conditional claim about the future on the public record: the record itself later does the scoring, without requiring anyone to adjudicate whether the institution had been careful. The claim either matched what happened or it did not, and the calendar was the only judge required.

That is what today’s forecast borrows from the case — not the economics, the discipline of writing something falsifiable down before the world decides.

Outline