Insurance and Natural Hazard Losses: A Case Study of Florida Counties, 1997-2006
Advisor: Dr. Susan L. Cutter
Across the United States, natural disaster losses have been rapidly increasing over the past few decades (Cutter and Emrich 2005). This has been a major concern for both researchers and practitioners. Reducing vulnerability and increasing the population's resiliency to natural hazards is a clear priority. Researchers believe that insurance is one possible method that can be utilized to help populations recover more quickly and decrease losses (Kunreuther 2006); however, the relationship between insurance and losses has not been fully evaluated in the literature (Klein et al. 1998). The insurance situation in coastal states, particularly Florida, is currently at a precarious turning point. This research evaluates the relationship between the spatial variability of insurance, social vulnerability, and natural hazard losses. It also examines the ability of social vulnerability and insurance variability to predict natural hazard losses.
The research is based on a ten year case study of Florida counties from 1997 to 2006. The study examined the spatial variability between Florida's insurance data, natural hazard losses, and social vulnerability through a cluster analysis with local Morans I spatial statistics. Spatially, regional patterns are found throughout most of the variables. In the north central and northeastern counties, there was a low presence of insurance, whereas the coastal areas, especially the southern tip of Florida, had a high incidence of insurance. A Pearson correlation analysis and Kruskal-Wallis difference of means test gave insight into the ability of insurance to act as an indicator of resilience.
The correlation and Kruskal-Wallis test illustrated that insurance alone may not be effective at lowering natural hazard losses due to purchasing behavior and the variability of premiums costs across the state. A linear multivariate regression was performed to evaluate the ability of insurance and social vulnerability data to predict hazard losses. The linear regression analysis demonstrated that social vulnerability and insurance could not produce a reliable regression with predictive power. Future research should examine a greater time period and a larger geographic area. A geographically weighted regression should also be explored in order to account for regionality found within the data.Thesis