The Revenue Estimate Decides the Bill

A legislature cannot vote on a proposal until it knows what the proposal costs. That requirement is sensible and it creates a role of extraordinary quiet influence.

Somebody has to produce the number. The office established to do it was created precisely so the number would not come from the people proposing the spending.

Independence was the whole point and it has largely held, which is a real achievement and not the thing worth examining.

What is worth examining is that the output is an estimate presented as a figure, and a figure behaves differently in an argument than an estimate does.

An estimate carries assumptions. It rests on a projected growth path, a behavioural response, a take-up rate, and each of those is a judgement that could reasonably differ.

Behavioural assumptions are the sharpest disagreement inside the exercise and the least visible outside it.

By the time it reaches a chamber it is a single number, quoted to the nearest billion, and disagreeing with it sounds like disagreeing with arithmetic rather than with a forecast.

Whether a change alters what people do, and by how much, is a genuinely open empirical question on which serious economists differ by large margins.

The budget documents carry the aggregates, and the aggregates are what get reported, which strips the assumptions out entirely.

The estimate must nonetheless pick one answer, because a legislature cannot vote on a range, and the picking happens inside a process with no external participation.

The window is the mechanism that does most of the work. A proposal is scored over a fixed number of years, and the choice of window is not neutral.

Publishing the range alongside the point estimate would change very little procedurally and would make the uncertainty part of the argument rather than a technicality behind it.

A measure that costs money early and saves later scores badly. One that saves early and costs later scores well. The underlying merits may run the other way entirely.

It would also protect the estimators, who currently absorb political attacks for producing a single figure that the method never claimed was certain.

Legislators respond to this rationally, by designing around it. Provisions expire just inside the window; costs are pushed just outside it; timing is chosen to score rather than to work.

The introduced text of any large fiscal measure shows the fingerprints: effective dates and sunsets that make no policy sense and perfect scoring sense.

Nobody involved regards this as dishonest. The window is a published rule and designing to a published rule is what legislators do.

The consequence is a body of fiscal law full of expiry dates that everybody expects to be extended, which means the scored cost and the real cost diverge permanently.

The extensions then arrive as separate measures, each scored on its own, so the aggregate is never presented to anyone as a single number.

The published statements eventually record what actually happened, years later, when nobody is arguing about it any more.

That lag is the structural problem. The estimate arrives when the decision is live and the outturn arrives when it is not, so the two are never compared in public.

A comparison would be genuinely useful and is not produced. There is no routine document that sets the estimate for a major measure against what it actually cost.

Producing one would be unwelcome to everybody: to the estimators, who would be marked; to the sponsors, whose projections would be checked; to the opponents, whose warnings would be too.

Which is a reasonable description of why it does not exist, and not a reason it should not.

The narrower fix is easier. Publish the central assumptions alongside the number, in the same document, at the same prominence, so that a disagreement about a forecast can be conducted as one.

At present the assumption is available to anyone who goes looking and the figure is what reaches the floor, and those two facts together decide a great deal of what a country does.