Validation of a probabilistic model for hurricane insurance loss projections in Florida

• Published in: Reliability engineering & system safety 2008-12, Vol.93 (12), p.1896-1905
• Main Authors: Pinelli, J.-P., Gurley, K.R., Subramanian, C.S., Hamid, S.S., Pita, G.L.
• Format: Article
• Language: English
• Subjects: Applied sciences; Calibration; Damage; Exact sciences and technology; Hurricane; Operational research and scientific management; Operational research. Management science; Risk model; Risk theory. Actuarial science; Validation
• ISSN: 0951-8320; 1879-0836
• DOI: 10.1016/j.ress.2008.03.017

The Florida Public Hurricane Loss Model is one of the first public models accessible for scrutiny to the scientific community, incorporating state of the art techniques in hurricane and vulnerability modeling. The model was developed for Florida, and is applicable to other hurricane-prone regions where construction practice is similar. The 2004 hurricane season produced substantial losses in Florida, and provided the means to validate and calibrate this model against actual claim data. This paper presents the predicted losses for several insurance portfolios corresponding to hurricanes Andrew, Charley, and Frances. The predictions are validated against the actual claim data. Physical damage predictions for external building components are also compared to observed damage. The analyses show that the predictive capabilities of the model were substantially improved after the calibration against the 2004 data. The methodology also shows that the predictive capabilities of the model could be enhanced if insurance companies report more detailed information about the structures they insure and the types of damage they suffer. This model can be a powerful tool for the study of risk reduction strategies.