Grantsmanship and the design of funding contests

A funder divides a fixed budget between an equal baseline and a merit pool awarded through a contest on a measured signal. Applicants produce the signal with two efforts: research, which the funder values, and polishing, which only raises the measure. The two-effort contest reduces exactly to a one-effort contest governed by two numbers, the signal's cost and its research content, and the design problem runs on the gap between them. Better polishing technology strictly shrinks the optimal pool; costly scrutiny of proposals re-expands it, and full scrutiny is never optimal. Voluntary participation removes researchers first and retains polishers, so the equal baseline is what shelters the research-rich; a shortlist ranked on contest strength admits exactly the entrants that shrink the pool. The same structure appears wherever fixed money is split between a flat share and a contest on a measured signal that mixes valued production with measure-improving effort, from exam-based awards to merit pay and promotion tournaments. The rule is deterministic throughout: merit still decides who wins, and the design decides how much merit pays.