Public goods are dealt with in two literatures that neglect each other. Mechanism design advises a social planner that expects individuals to misrepresent their valuations. Experiments study the provision of the good when preferences might be non-standard. We introduce the problem of the mechanism design literature into a public good experiment. Valuations for the good are heterogeneous. To each group we add a participant with power to impose a contribution scheme. We study four settings: the authority has no personal interest and (1) valuations are common knowledge or (2) active participants may misrepresent their types; the authority has a personal interest (3) and must decide before learning her own valuation or (4) knows her own valuation. Disinterested social planners predominantly choose a payment rule that gives every group member the same final payoff, even if misrepresentation is possible. Authorities are overly optimistic about truth telling. Interested social planners abuse their power, except if the opportunity cost of a more balanced rule is small.