Adaptation to Rare Natural Disasters and Global Sensitivity Analysis in a Dynamic Stochastic Economy

This paper aims to investigate rare natural disasters and studies adaptation decisions in a dynamic stochastic economy. We examine the optimal balance between investment in productive capital and adaptive capital stock, which is inherently non-productive but alleviates the damage caused by a rare natural disaster. We present a modeling way to include uncertain rare natural disasters in discrete time and solve the model by using the time iteration collocation with the adaptive sparse grid. We perform a global sensitivity analysis to screen which uncertain parameters should be primary calibrated based on the Sobol' indices and compute the univariate effects to identify in which parametric region the model outcomes are most sensitive. To speed up the solving processes, our implementations are massively parallelized on high-performance computing architecture in a distributed memory fashion. We claim that the optimal adaptation to rare natural disasters is to advance our economic development; however, when the economy is developed enough, the growth rate of the adaptive capital stock exceeds that of productive capital stock to precautionary prepare for the future uncertainty.