On model selection criteria for climate change impact studies

Climate change impact studies inform policymakers on the estimated damages of future climate change on economic, health and other outcomes. In most studies, an annual outcome variable is observed, e.g. agricultural yield, annual mortality or gross domestic product, along with a higher-frequency regressor, e.g. daily temperature. While applied researchers tend to consider multiple models to characterize the relationship between the outcome and the high-frequency regressor, to inform policy a choice between the damage functions implied by the different models has to be made. This paper formalizes the model selection problem in this empirical setting and provides conditions for the consistency of Monte Carlo Cross-validation and generalized information criteria. A simulation study illustrates the theoretical results and points to the relevance of the signal-to-noise ratio for the finite-sample behavior of the model selection criteria. Two empirical applications with starkly different signal-to-noise ratios illustrate the practical implications of the formal analysis on model selection criteria provided in this paper.