What are the equilibria in public-good experiments?

Most social-preference models have been tailored to yield only a full-defection equilibrium in one-shot linear public-good situations. This paper determines the Nash-equilibrium sets that result from experiment participants' elicited preferences. The data show that multiple equilibria are relatively frequent even in a standard three-player setting. In this perspective, the common finding of close-to-omnilateral defection at the end of repeated public-good games is surprising and raises the question of why the dynamics of play seem to select this equilibrium out of the existing equilibria.