Bayesian space–time varying coefficient modeling for climate econometrics: A spatial–temporal Gaussian process approach
Abstract
Quantifying the economic impacts of weather variables has been a key empirical topic in understanding the impacts of climate change. While panel methods, which utilize cross-sectional differences in within-unit over-time changes in weather variables, have been widely used to credibly identify the average effect of interest, there has been a growing interest in understanding the distribution of potentially heterogeneous weather effects. The heterogeneity of the effects across space and time, which can shed light on the differences in the local capacity of adaptation across regions or its evolution over time, provides important insights into developing climate policies. In this paper, we propose a novel Bayesian hierarchical modeling approach to estimate heterogeneous effects of weather variables using panel data. Our proposed approach models space–time varying coefficients via Gaussian process priors. Using the data of the corn yield from the US corn belt, we show that our proposed model outperforms several alternatives. Our findings are consistent with the literature and also provide additional insights into the localized nuances of climate change adaptation.