What does “use it or lose it” budgeting actually cost?

The finding

Liebman and Mahoney, publishing in the American Economic Review, examined US federal spending data and found the expected spike: in the final week of the fiscal year, when unspent budget simply evaporates, spending ran at roughly 4.9 times the weekly average.

The valuable part is what they did next. They measured the quality of what was bought. Across around $130 billion of IT projects, work procured in the year-end rush was between 2.2 and 5.6 times more likely to score below median. The spike is not merely a timing artefact — it is a measurable degradation in what the money buys.

The distortion chain, which is worse than the spike

Expiring money produces a fourth-quarter sprint. The sprint inflates the year’s spending baseline. Next year’s budget is then built on that baseline — so the padding becomes permanent, and it is defended with the unanswerable observation that the money was, after all, spent.

Capital discipline weakens from there, because a budget that has been padded for two consecutive years is no longer telling anyone anything.

The fix is structural, not exhortative

The same research shows what works: where rollover was permitted — where unspent money carried forward instead of vanishing — the spike disappeared.

That is the important word, and it is worth being clear about why. The year-end sprint is not a failure of discipline or a moral lapse. It is a correct, rational response to a rule that punishes underspending. Telling people to stop will not work, and has not. Changing the rule does.

Private owners can copy the mechanism directly: allow defined carry-forward for capital and project budgets, with governance around it, rather than a hard December expiry that guarantees the behaviour you are trying to prevent.