Modelling economic losses from earthquakes using regression forests: Application to parametric insurance

Parametric insurance has developed in response to the increasing economic losses from natural disasters over the past two decades, but there is still a large gap between the total economic loss and the insured loss, especially in China. This paper explores regression forests to model economic losses from earthquake disasters. Using historical economic loss data from China between 1974 and 2020, we apply Generalized Pareto regression model, mean and quantile regression forests to investigate the effect of different risk factors on the severity of earthquake disasters. The results show that regression forests can effectively improve prediction accuracy, and quantile regression forests perform best. Earthquake magnitude is the main factor influencing economic losses. A new coverage of parametric insurance for earthquake catastrophes in Dali is then obtained by quantile regression forests. It is shown that the gap between the total economic loss and the insured loss is significantly reduced.

Automated summary

This paper uses regression forests to model economic losses from earthquake disasters and finds that they can effectively improve prediction accuracy, with quantile regression forests performing the best. Earthquake magnitude is the main factor influencing economic losses. A new coverage of parametric insurance for earthquake catastrophes in Dali is obtained through quantile regression forests, significantly reducing the gap between the total economic loss and the insured loss.