Aggressive Backtesting of Stochastic Loss Reserve Models – Where It Leads Us
ISO Innovative Analytics,
New York USA.
In 2012 the was posted on the Casualty Actuarial Society website. This database includes several hundred loss triangles compiled from the American Schedule P exhibits that are reported to the American National Association of Insurance Commissioners. The database includes subsequent outcomes that were reported after the original loss triangle was reported.
Since the database was compiled, the speaker has used this database test the predictions of two currently popular stochastic loss reserve models and found some shortcomings of these models. The talk will discuss new models that are fit with Bayesian MCMC algorithms that address these shortcomings. The talk will then go on to show how these models can be used to address the current problems of dependencies between lines of insurance, and cost of capital risk margins.
Keywords: Bayesian MCMC, Stochastic Loss Reserving, Dependencies, Risk Margins, Schedule P