Spatial dependencies in wind-related housing damage

This article examines the spatial dependence among housing losses due to tornadoes using data from the May 1999 Oklahoma City tornado. In order to examine the existence of spatial dependence and its impacts on the damage analysis, we compare an estimation based on a traditional ordinary least square model with the general spatial model. The results show that housing damage in this disaster area is highly correlated. Monetary losses not only depend on the tornado that struck residences, but are related to the damage magnitudes of neighboring houses. Average losses as well as the loss ratio increase with the Fujita Scale damage rating. We conclude that the general spatial model provides unbiased estimates compared to the ordinary least square model. In order to construct appropriate home insurance policies for tornado disasters or to improve the damage resistance capabilities of houses, it is necessary for insurance underwriters and builders to consider spatial correlation of tornado damage.