CMStatistics 2015: Start Registration
View Submission - CMStatistics
B1138
Title: Some new insights into large commercial risks Authors:  Davide Benedetti - Imperial College Business School (United Kingdom) [presenting]
Enrico Biffis - Imperial College Business School (United Kingdom)
Abstract: Large commercial risks, such as commercial property and liability, present modelling challenges due to the paucity of data available for model estimation/validation, and the complex relation between hazard events and realized losses. There is therefore a tendency for insurance practitioners to apply a considerable degree of judgement in pricing, reserving, and capital modelling. Some new evidence on large commercial risks will be presented based on unique datasets on large commercial risks based on contributions from two leading Lloyd's of London syndicates, and a global reinsurer. The datasets contain granular information on more than 3500 losses and exposures, which are used to shed some light on the risk profile of medium to high layers of exposure as a function of different rating factor configurations, geographic region and period. In particular, differences between tail risk profiles in the North American, European, and Asia-Pacific region are explored, and structural breaks around major events are tested. Small sample issues and tail risk heterogeneity are addressed with weighted Hill estimators, tail index regressions, and Generalized Pareto Distribution with covariates. Finally, it is carried out a benchmarking exercise in which it will be quantified the risk premiums embedded in market rates as proxied by suitably constructed indices.