Budget scoring windows

Why does a tax cut everyone expects to be permanent get priced as though it will expire?

A legislature passes a tax cut. Its sponsors say publicly that it is permanent and will fight anyone who lets it lapse. The bill itself says it expires. The official cost estimate, dutifully, counts the revenue lost only up to that date.

Everyone in the room knows the expiry is unlikely to happen. So who is being deceived? Nobody, apparently — and that is the puzzling part. What must a rule look like for a manoeuvre to be simultaneously obvious, legal, and effective?

Start with why any window exists. A legislature that wants fiscal discipline needs a number attached to each bill: what does this cost? But cost over what period? A tax provision has no natural end, so the true answer runs to infinity, and an infinite-horizon figure is dominated by assumptions about growth, interest rates and demographics decades out — assumptions nobody can defend and everybody can dispute. The number would stop constraining anything, because any advocate could produce a favourable one.

The fix is to fix the horizon. Score everything over the same span, so estimates are comparable and the arguing happens about the policy rather than the discount rate. The United States uses a ten-year window as its standard span, if I recall the current convention correctly. The particular length matters less than that it is fixed and shared.

Now the second ingredient, which turns a convention into a lever. In some procedures the score is not merely informative — it is binding. The fast-track budget process in the US Senate carries a rule barring a reconciliation bill from increasing deficits beyond the window. A bill failing that test loses its procedural protection and needs a supermajority it does not have.

Put the two together and the incentive is immediate. The bill must show no cost beyond year ten. A permanent provision fails; one that expires inside the window passes. So the drafters write an expiry date — and everything they expect afterwards, the extension votes, the renewals, the eventual permanence, falls outside the measured span and therefore does not exist as far as the rule is concerned.

Notice the shape of this. The score began as a measurement of a policy. The moment it became a binding constraint, the policy started being shaped to fit the measurement. The expiry date is not a feature anyone wants; it is a feature the scoring rule pays for. The measure has stopped being a good measure precisely because it became a target.

What would falsify the account? If sunsets are artefacts of the window rather than genuine policy choices, they should cluster at the window's edge and appear disproportionately in bills moving through the procedure where the score binds — and provisions in the same bill that fit the constraint anyway should be written permanent. The 2017 US tax legislation is the clean case, as I recall it: the corporate rate reduction was permanent while the individual-side provisions carried an expiry date near the end of the window, in one bill, drafted by one set of hands. The refuting observation would be sunsets scattered evenly through the window and appearing at the same rate in ordinary legislation where no procedural rule requires them. That would mean legislators sunset things because they genuinely want to revisit them, and the window is incidental.

Three things the simple story leaves out.

First, the manoeuvre is not free. A real expiry date creates real uncertainty, because the extension is likely but not certain. Households and firms plan against a rate that may change, investment decisions get deferred toward the edge, and the legislature must schedule a renewal fight it would rather not have. So the scoring rule does not merely mislabel a permanent policy — it converts a permanent policy into a temporary one that is usually renewed, which is a genuinely different thing with its own costs.

Second, those renewal votes become valuable in themselves. A must-pass extension is a vehicle: it attracts amendments, and the recurring need for it gives leverage to anyone whose vote is required. So the sunset generates a stream of legislative occasions that would not otherwise exist, and the participants come to depend on them.

Third — and this is where the honest account has to slow down — the window is not simply a mistake. Every alternative has a defect of its own. Score to infinity and the number becomes unfalsifiable. Score over a shorter span and the gaming gets easier, not harder. Score on the assumption that expiring provisions will be extended, and you have handed the scorekeeper the job of predicting future legislatures, which is both beyond its competence and a large transfer of power to an unelected office. Some systems address the problem from the other end, requiring a medium-term fiscal framework or an independent forecaster to publish the long-run path alongside the official score, so the manoeuvre is visible even where it is permitted. That helps with visibility and does nothing about the incentive.

And there is the plain fact about why this persists. Whoever is in the majority finds the technique useful, and the minority expects to be in the majority eventually. Reforming it would require the current majority to give up an instrument it is presently using, in exchange for a constraint on an opponent who does not yet exist. A rule that everyone can criticise and everyone can use is unusually stable — not because it is defensible, but because the change would have to be made by its beneficiaries.

• Whether independent fiscal institutions that publish long-run paths alongside official scores change legislative behaviour, or only the commentary.
• How the same edge effect appears in quarterly corporate reporting and in agency spending before a fiscal year closes.

Every term the collection defines is gathered in the glossary.