Predictive analytics solutions — Life and annuities

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

Milliman has experience in applying predictive analytics to life insurance and annuities. We set assumptions for financial projections- mortality, policy continuity, and policyholder use of options such as riders and transfers. Assumption setting was previously done using a tabular experience analysis approach. Predictive models can tease out more detailed relationships that can have a large impact on expected future cash flows. Our work in predictive analytics can use third-party data to improve the profitability profile of business sold in the future (through product design and targeted marketing) and in the past (through inforce management).

Milliman focuses on helping companies set policyholder behavior assumptions for variable annuities (VA). Specifically, we have studied two policyholder behaviors extensively: lapse and guaranteed lifetime withdrawal benefit (GLWB) rider usage. We have also studied lapses among fixed annuities and market value-adjusted annuities, lapses on VAs with other riders, and lapses for post-level term life insurance policies.

Industry studies we have performed have given companies insight into policyholder behavior they haven’t gotten elsewhere. We have shown how product features and past behavior—among other things—can be applied to better predict how policyholders are likely to use their products in the future, in particular in relation to changes in interest rates or equity markets. This information has been useful, especially in helping companies determine whether a more sophisticated assumption was needed. We also showed that policyholders act differently at differently companies, which has assisted in discussions with regulators currently working to update key VA reserve requirements.

We have recently been delving deeper into our industry studies. For example, we built a web-based software platform to deliver the same insight, but also to allow companies to dive into the data and generate predictions themselves. Additionally, we are assisting in delivering a mortality study in the same way and are working to develop the capability to allow clients to build their own predictive models within the platform.

Technology is connected to every aspect of our predictive analytics work. We use machine learning and generalized linear models to analyze data and the R platform to perform analyses and build our website.