MG-ALFA — Advanced analytics


  • Print
  • Connect
  • Email
  • Facebook
  • Twitter
  • LinkedIn
  • Google+

As the insurance industry changes, companies will need to perform much more sophisticated calculations and the number of compute cycles required will increase by many orders of magnitude. Swiss Solvency Test, Solvency II, economic capital, and MCEV all require stochastic calculations for a single valuation and nested stochastic calculations to project values forward.

Stochastic modelling uses a combination of probability and random variables to forecast financial performance or, in the case of reserve setting, to forecast financial requirements. Nested stochastic models, as the name implies, are stochastic models inside other stochastic models. Because setting reserves and capital using a principles-based approach is based on stochastic valuation, earnings projections will require stochastic projections at each future projection date, across all scenarios. Nested stochastic models are needed to appropriately manage the business, price new products, project earnings, or measure risk. MG-ALFA® delivers this computational ability today.

Recently, a large life company migrated to MG-ALFA and developed economic capital models across the organization. Using our nested stochastic capabilities, it was able to project economic capital requirements by line of business and then use our aggregation functionality to demonstrate to rating agencies that it had inherent hedging between lines within the company. Using MG-ALFA, the company was able to reduce its overall capital requirements well below the sum of the capital requirements for the separate business lines. The results, of course, were unique to this client and its situation.

Advanced cluster modelling

With actuarial models growing in complexity and with increased use of stochastic and nested stochastic approaches, running large models is time-consuming and unwieldy. Milliman has developed a new type of automated model compression process called "cluster modelling". Cluster modelling enables users to efficiently model millions of policies into significantly fewer—hundreds or thousands—model points. The process can accurately reproduce the results of the original seriatim model across a range of economic or experience scenarios and only requires a small number of calibration runs.