Analysts are expending a great deal of effort to devise new ways to perform stress tests for capital allocation under Solvency II. According to Martin Neil and Neil Cantle, many of the answers can be found in the work of 18th-century mathematician Thomas Bayes—his famous theorem provides a way of reasoning under uncertainty, where data and judgements are combined to provide the best possible risk estimate given the model at hand.
Bayes’ work can now be implemented to help with stress testing. Proponents of this work realize that data alone is insufficient for stress testing because there will never be enough data to fit an analytical model to the data—even for the rare scenarios that are of interest. In fact, there may be no data at all. This article, published by InsuranceERM (subscription required), explores what can be done in this situation to allow stress testing.