Uncertainty in Hurricane Risk Modeling and Implications for Securitization

This paper presents a simple procedure for quantifying the effect of the historically limited meteorological record on insurance losses calculated by hurricane catastrophe simulation models. Using a standard actuarial approach, the uncertainty in the historical record can be decomposed into uncertainty in the mean annual hurricane frequency, and uncertainty in the severity, where the severity corresponds to the distribution of insurance losses resulting from a single random hurricane event. Uncertainty in the mean annual frequency can be estimated directly from the historical data. Uncertainty in the severity is more difficult to evaluate, but can be approximated using a parametric bootstrap method which captures the effects of the finite historical record. Confidence intervals on the insurance loss as a function of return period are calculated for several different portfolios. These results are placed in the context of a recent insurance-linked securitization transaction, and are found to be consistent with other independent measurements of pricing uncertainties in similar instruments. The historical volatility of one-year corporate bond default rates is also compared to the magnitude of the uncertainty in the hurricane simulation models. This comparison suggests that volatility of one-year corporate bond default rates is greater than uncertainty in the hurricane models.