The complaint that technical decisions have been removed from democratic control is usually met with the observation that voters cannot evaluate epidemiological models or monetary transmission mechanisms, which is true and beside the point. The delegation at issue is not of the analysis but of the judgement that follows it, and the two are separated with less care than the defence assumes. A recommendation about how much unemployment is worth accepting in order to reduce inflation contains an economic estimate and a valuation of two kinds of harm, and only the first belongs to the economist. Institutions have generally been designed as though the composite were technical throughout, which is convenient for everyone: politicians avoid an unpopular trade-off, experts acquire authority beyond their competence, and the arrangement is stable until the moment somebody notices.