Abstract:
|
Due to non-catastrophic weather events, such as, for example, higher than normal precipitation or wind speed, insurance companies increasingly frequently incur high cumulative losses. Climate change further amplifies weather hazards. Weather-induced insurance risk generally represents a non-stationary non-separable space-time process, which often implies that more conventional methods of statistical inference for spatio-temporal data are inapplicable. We discuss how we use approaches for complex multi-layer networks and topological data analysis to shed more light on weather-induced hazards and associated peril maps, while accounting not only for space-time dependencies but for such key risk information as socio-demographics, building codes, and real estate prices.
|