We study the relation between mechanism design and voting in public-good provision. If incentive mechanisms must satisfy conditions of robust coalition-proofness as well as robust incentive compatibility, the participants’ contributions to public-good provision can only depend on the level of the public good that is provided and that level can only depend on the population shares of people favouring one level over another. For a public good that comes as a single indivisible unit the outcome depends on whether or not the share of votes in favour of provision exceeds a specified threshold. With more provision levels for the public good more complicated mechanisms can be used but they still involve the counting of votes rather than any measurement of the participants’ willingness to pay. The paper thus provides a foundation for the use of voting mechanisms.