Robust simulation for sensitivity analysis of catastrophe risk losses

Enormous advances in information technology have significant implications for catastrophe risk analysis. It is now possible to perform numerous simulations and avoid artificial smoothing techniques that can mask actual variability in outcomes. In seismic risk studies the presence of competitive models has led to various weighting systems that sometimes have unexplored as well as undesirable outcomes. An alternative approach uses competitive models to characterize uncertainties. An example of a robust simulation using a hurricane catastrophe risk model for the mainland United States is used to show the range of possible outcomes based on alternative views of hazard and exposure. Reasonable alternative views of hurricane activity produce a 20% variation in AAL whereas extreme differences in hurricane activity produce AAL that varies by a factor of three. The impacts of different assumptions for exposure are also explored.