Economic Damages from Climate Change and Model Uncertainty

Climate change damages are increasingly characterized by the empirical relationship between temperature and economic outcomes, like gross domestic product. Theory, however, does not posit an estimable form for these relationships. This paper, therefore, considers the implications of model uncertainty for estimates of global climate change damages using cross validation methods to evaluate out-of-sample predictive accuracy of 400 variants of prominent models. Model performance is shown to be insensitive to specification of the temperature relationship. However, consistent with the recent literature, we conclude that a concave function best describes the relationship between economic activity and temperature, implying a temperature for optimal production that is extremely sensitive to model selection. Given the concentration of production in regions that typically experience temperatures in the range of potential optima, estimated long-run GDP impacts of warming are also very sensitive to model choice, varying in magnitude and sign across models....