Forecasting Mass Action Losses Using a Hybrid Development Model

Mass action losses often emerge differently than other losses for a line of business. Using asbestos as an example, general liability development began to show some unexpected late development in the late 1970’s and early 1980’s. After some investigation it was concluded that much of this development could be attributed to asbestos related claims. In addition these claims did not seem to exhibit the dependence on accident year age that other general liability losses usually experience. Thus, it could be concluded that normal development methods may not be appropriate for forecasting such losses. One alternative that has been considered is to assume that future emergence of asbestos losses will depend not on the age of the particular accident year, but on the valuation year of the particular losses. This assumes future development of all losses would be the same, independent of the accident year. In this paper we will propose an alternative, hybrid, of these two models (pure accident year and pure calendar year). In the hybrid model we will allow the data to dictate what mix of the two models best fits the experience emerged to date. The method itself is not very difficult to implement in practice. Given numerical solution methods available in current personal computer spreadsheet software, PC solutions can be generated in a very short amount of time. We discuss the concept and support the discussion with examples applied to some real-but-disguised data. We then explore an approach that to apply which is learned from this asbestos example to other situations with example hazardous waste data as an example. The concepts could apply to other mass action types of exposure and provide a separate, independent, test of results implied by other forecast methods. Keywords: Pollution, Environmental, Toxic Torts