Earthquake parametric insurance with Bayesian spatial quantile regression

With its transparent and fast claims payment, parametric insurance has been widely used to insure nature-related risks such as earthquakes, floods and hurricanes. In 2014, earthquake parametric insurance was introduced to provide coverage for earthquake losses occurred in Yunnan Province of China. However, as a main limitation of parametric insurance, basis risk is inevitable. In this paper, a Bayesian spatial quantile regression model is proposed to reduce the basis risk of earthquake parametric insurance. The effect of earthquake hazard, risk exposure, and vulnerability on economic loss are analyzed and considered in the quantile regression model. Since risk exposure and vulnerability at the epicenter cannot be observed, they will be treated as latent variables in the quantile regression model. Bayesian approaches are applied, and spatial correlation is considered to construct the prior distributions for the latent variables. Earthquake losses in Yunnan Province from 1992 to 2019 are collected and analyzed by the proposed model and methods. The payment mechanism and the corresponding premiums of 16 regions in Yunnan Province are then calculated. The results show that the loss ratio is more reasonable than the current earthquake insurance, and the basis risk is then reduced. (c) 2022 Elsevier B.V. All rights reserved.

Automated summary

Parametric insurance is widely used to insure nature-related risks. This paper proposes a Bayesian spatial quantile regression model to reduce the basis risk of earthquake parametric insurance. The model analyzes the impact of earthquake hazard, risk exposure, and vulnerability on economic loss, taking into account spatial correlation.