A0388
Title: Using precipitation forecasts to predict insurance claims
Authors: Haakon Otneim - Norwegian School of Economics (Norway)
Etienne Dunn-Sigouin - NORCE (Norway)
Sondre Hoelleland - Norwegian School of Economics (Norway) [presenting]
Mahsa Gorji - Norwegian School of Economics (Norway)
Geir Drage Berentsen - Norwegian School of Economics (Norway)
Abstract: Climate change is affecting insurers worldwide, as more extreme weather leads to increasingly severe and frequent damage to infrastructure. A framework for short-term property insurance claim forecasting is presented, which facilitates early customer warnings and efficient resource allocation. Using ensemble precipitation forecasts, weather-driven claims in Bergen and Oslo are modeled as a rare-event binary classification problem, employing probabilistic regression and machine learning methods. Models are evaluated on discrimination, reliability, and economic value. Results indicate that using weather forecasts enhances model discrimination, reliability, and operational cost efficiency, compared with baseline scenarios. Case studies of extreme rainfall events illustrate practical application, demonstrating how insurers can leverage publicly available forecasts to anticipate increased claim risks and improve their response strategies.
