Rational choice theory, including models of social preferences, is challenged by decades of robust data from public good games. Provision of public goods, funded by lump-sum taxation, does not crowd out private provision on a one-for-one basis. Provision games elicit more of a public good than payoff-equivalent appropriation games. This paper offers a morally monotonic choice theory that incorporates observable moral reference points and is consistent with the two empirical findings. The model has idiosyncratic features that motivate a new experimental design. Data from our new experiment and three previous experiments favor moral monotonicity over alternative models including rational choice theory, prominent belief-based models of kindness, and popular reference-dependent models with loss aversion.
