How should we estimate the benefits of policies that lead to large (non-marginal) changes in environmental outcomes? We show how a locational equilibrium (‘horizontal sorting’) model can be used on data from Indonesia to address this question, making this a rare application of a structural approach to non-market valuation in developing countries. Specifically, the model exploits migration patterns induced by spatial variation in air quality, wages, and cost-of-living to value air quality. Additionally, we are also able to control for (a) migration distance and religious preferences, and (b) endogeneity of air pollution by using an instrumental variable strategy (source-, wind- and distance-based). Using a combination of household and satellite-derived data, we estimate MWTP of PM2.5, and find that our estimates are higher than results from hedonic property models. Our simulations of a tropical conservation policy to control air pollution, based on the structural model, consider two measures of partial equilibrium – instantaneous change and adjustments for wages and housing prices. We find such a policy benefits residents across
Indonesia, even for those living in districts outside of where the policy is implemented. In sum, our paper demonstrates a potentially novel and rigorous tool to evaluate environmental policies and value non-marginal changes in environmental quality.