Modelling emergency disaster mortality around the world: a network-based distributional inference approach

Abstract In this paper, we leverage Optimal Transport theory and network analysis to propose a novel distributional framework for the inference of mortality risks arising from worldwide emergency disasters. We propose a propagation method grounded on Wasserstein barycentre and similarity networks, based on climate, social, economic, and demographic variables, able to model mortality risk distributions, when country-level information is incomplete or unavailable. We apply our proposal to the International Emergency Events Database showing that the method is able to coherently reconstruct the mortality risk distributions related to natural and technological disasters via the information embedded in country similarity networks. We provide a comprehensive indicator of countries potential extreme losses from various types of disasters through the Gini-based Value-at-Risk (Gini-VaR). This measure provides country-specific and disaster-specific probabilistic impacts of extremely severe hazards. We also shed light on the relationship between disaster risk preventive strategies put in place by nations and their mortality risk profiles based on the Gini-VaR, highlighting major inequalities across countries.