Applying the concept of “Value of Information” within earthquake loss modelling for insurance purposes

Successful earthquake insurance, both for the property owner and the (re)insurance company providing the cover, relies on an accurate assessment of the property's expected future financial loss, which itself depends on the earthquake hazard and the property's vulnerability to that hazard. Both these aspects (hazard and vulnerability) are associated with large epistemic uncertainties due to a lack of knowledge and data concerning future earthquake ground motions and how the property will respond to these ground motions. Differences of 50% amongst assessed earthquake hazard (in terms of, for example, peak ground acceleration for a given return period) by different projects for the same location are not uncommon, even for well-studied regions such as Italy. The uncertainties in assessed vulnerability are often considered to be smaller but this is dependent on accurate information on the building type, material properties, geometry, and the building's condition, which are not always available to the (re)insurer.

Decisions on what level of earthquake insurance is appropriate for a given building could be improved by collecting additional information on, for example, the geotechnical ground conditions (e.g., average shear-wave velocity in the top 30m, Vs30) under the building, the building’s typology (e.g., construction materials, numbers of storeys or age of construction) or whether the building’s location is susceptible to secondary hazards (e.g., landslides or liquefaction). Additional information has the potential to reduce the associated uncertainties, which in turn could potentially mean that the building is not over- or under-insured. Collecting this information, however, costs resources in terms of time spent in identifying and interpreting the data, employing an external consultant to provide information or, potentially, licensing costs in accessing data. Therefore, there is a balance to be struck between collecting as much data as possible (thereby reducing uncertainties to the lowest possible level) and collecting no additional data (accepting the default uncertainties). The concept of “Value of Information” (VoI) can help guide the decision as to whether it is worth collecting new data and which data has the highest potential benefit for cost-effective uncertainty reduction.

The aim of this PhD project is to extend the VoI approach developed in this recent article in Soil Dynamics and Earthquake Engineering

(https://www.sciencedirect.com/science/article/pii/S0267726122004997) and apply it to earthquake catastrophe modelling. In a first step, the existing developments could be applied within a loss model developed by Aon Impact Forecasting, the industrial partner of this project, to understand when refinements to the soil layers of this model are warranted. Further developments could include applying the approach to investigate when improvements to the exposure data (e.g., which building types are present in the region) are beneficial, the vulnerability component (e.g., seismic behaviour and costs of the various building types) or the assessment of secondary perils (e.g., how susceptible is the region to liquefaction or landslides).

This PhD project will result in a better understanding of the uncertainty in key insurance metrics and consider how these can be reduced by collection of more data on the hazard (e.g., on the local site conditions under a property) or the property itself (e.g., its structural system). The project will have a considerable industrial impact given the vast portfolios of national and international (re)insurance companies and governments. The proposed method could also be easily extended to loss assessments under other hazards (e.g., floods or extreme wind).