Influence of Community Characterization on the Output Uncertainty of the Florida Public Hurricane Loss Model

Abstract

The Florida Public Hurricane Loss Model projects insured hurricane losses for residential buildings. It is a comprehensive multi-disciplinary probabilistic model that computes building and contents losses, as well as time-related expenses, for any given portfolio of residential exposure in Florida. The model can provide probable maximum losses and annual average losses for the entire state or any region of the state, as well as scenario analysis for real or hypothetical events. The library of vulnerability models covers the most common building classes of personal (manufactured homes and single-family homes) and commercial (1 to 3 story low-rise and more then 3-story mid/high-rise buildings) residential buildings. The models include different mitigation measures that can be activated to explore possible ﾓwhat-ifﾔ scenarios.

Significant epistemic uncertainty exists in the model, including uncertainty related to: the hazard; he strength capacities; the water absorption capacities; the costs; the replacement values; the building orientation; and the characterization of the exposure. The later implies the matching of every building in the exposure set with the proper building class in the library of vulnerability models of the FPHLM. A set of building parameters defines each building class. Typically, much of this information is missing from insurance portfolios, and the modelers rely on statistical studies of the regional exposure to make-up for missing information. The missing information for any given property is assigned based on statistics, and the appropriate building class is subsequently assigned to the property. Any incorrect assignment shall affect the accuracy of the insured loss.

The purpose of this paper is to quantify the effect of the uncertainty regarding the building class assignment on the overall uncertainty of the projected loss. The FPHLM modelers have different sets of exposure statistics from less to more exhaustive or accurate. Their latest set of statistics covers the entire State of Florida, as opposed to the previous ones, which included a limited number of counties, and where the missing information was extrapolated from neighboring counties. The authors propose to run portfolio analyses using previous and most recent statistics to demonstrate and quantify the range of uncertainty associated with community characterization. They propose to do this for multiple regions and scenario events, for which they have the associated claim data. The comparisons of the aggregated losses and their contrast with the spread among like-structures should help quantified the uncertainty due to building characterization and should inform strategies to reduce that uncertainty.