How can insurers get prepared to catastrophes? Assessing earthquake expected losses from historical catalogue

Countries around the world have had to face huge economic losses due to natural disasters over the past decade. This, of course represents a source of great concern for Na-tional Governments and even more so for the insurance industry. In the aftermath of a natural disaster, insurance and reinsurance markets are prone to severe insolvencies and destabiliza-tion. Therefore, the finance industry is looking for more reliable loss estimation procedures and insurance models, as effective means for resilience improvement. The present paper proposes an engineering-based methodology as a support for innovative insurance models. The study aims at defining a scientific instrument supporting insurers and reinsurers in forecasting expected losses and in mitigating the potential lack of financial ca-pacity. This allows for catastrophe-linked modeling to be performed according to a risk-based framework. The proposed methodology is applied to the Italian residential building stock subjected to seismic risk. Expected losses are evaluated following the procedure out-lined in Asprone et al. (2013)[1] for earthquake scenarios from the catalogue of historical earthquakes, of the National Institute of Volcanology and Geology (INGV) [2] and assuming present-day exposure characteristics. Hence the procedure can be implemented anywhere else a detailed catalogue collecting information about earthquakes from the past is available, as for Italy. Statistical simulations of ground motion intensity (peak ground acceleration, PGA) using multivariate normal distributions are performed for each earthquake. The simulated PGA values are calculated based on the ground motion prediction equation of Sabetta and Pugli-ese (1996)[3], whose coefficient are re-estimated by Bindi et al. (2009)[4], for each Italian Municipality. A set of different fragility curves from the literature has been selected and averaged for each building type, also accounting for seismic and non-seismic design. In the next step, the annual expected losses for insurers are evaluated and the results are aggregated in order to calculate total losses for the entire National building stock. Linear regression analysis is performed for predicting the expected loss as a function of earthquake magnitude. The result-ing loss model can be used for efficient and rapid loss estimation for a given earthquake scenario.