Milliman Claim Variability Benchmarks™ (CVB) are new industry benchmarks to help assess the quality of stochastic unpaid claim distributions used for enterprise risk management (ERM) and dynamic financial analysis (DFA), including correlation for aggregate distributions. The CVB also stochastically support deterministic ranges used for reserving.
The CVB product is the only tool available to help dynamically assess estimates of ranges, distributions, and correlation assumptions. By using the CVB product, companies can help avoid the costs of mis-estimating risk, which can lead to undercapitalization, underpricing, inadequate risk margins, and other risk management issues, and improve the bottom line.
Derived from extensive testing of common models, using more than 30,000 data triangle sets involving all long-tail Schedule P lines of business, the CVB are aimed at understanding an insurance entity’s risk. The extensive data also allows for customizable benchmarks that are suitable for companies of all sizes, including companies outside the United States.
For any model(s) or approach to quantify risk, the CVB guidelines can be used to assess model assumptions and results. In addition, the multiple model approach used in Arius® provides a more robust estimate than the single model approach by taking model risk into account.
Insurers typically lack industry benchmarks to gauge the quality of unpaid claim variability estimates. The CVB can help prevent underestimation of the tail of the distribution even after aggregation. They can be used with deterministic estimates to provide distribution information from industry data and can provide reasonableness tests for stochastic estimates.
The CVB includes loss development factor (LDF) patterns, unpaid distributions, and correlation values that are adaptable based on company size, development patterns, and more. The CVB are available as an easy-to-use Excel add-in and templates are included so you can immediately benefit from using the product.
Benchmarking a reserve range
Benchmarking an unpaid claim distribution