General insurance pricing often involves the use of statistical techniques to estimate the expected claims cost associated with a relevant risk profile. These techniques include GLMs and policyholder specific rating factors. In recent years, the insurance industry has been increasingly impacted by weather-related events, which are harder to price for. These events are expected to be exacerbated by both climate change and meteorological phenomena such as El Niño. Consideration of these risks has traditionally been captured in the pricing process indirectly using geographical features such as area codes. hi these may not directly capture the risks that need to be modelled. This talk: provides a framework for thinking about modelling climate-related risks, then deep-dives into how the pricing process can be improved directly links an exposure dataset of buildings risk and associated claims experience to a high-resolution gridded precipitation dataset to investigate the predictive power of this feature on both an actual and forecasted basis establishes a modelling framework that allows for the estimation of both the frequency and severity of buildings claims given the combined dataset considers the relative importance of the added precipitation feature against traditionally used rating factors and the sensitivity of the underlying claims frequency and severity to changes in precipitation Finally, by considering different precipitation scenarios, we show how the risks of excessive precipitation can be quantified, allowing for more accurate forecasts of financial performance to be made, and risk mitigation strategies to be investigated. 

Speakers: Ronald Richman, Kovlin Perumal

Chair: Sheena Suchak