Pension Risk Transfer – Improving Mortality Rate Prediction by 30%

Business Objective

Our client is a recognized leader in insurance, annuities, and employee benefit programs, with 100 million customers in over 50 countries.

Pension fund liability fluctuates due to many factors with longevity being one of the prime drivers. As the mortality tables are updated only once in a decade and it is for the overall population, the client’s pension risk transfer team wanted to develop a better solution for longevity prediction for improved and effective pricing. The client wanted to:

— Improve mortality rate predictions using a predictive model
— Leverage external data for mortality rate prediction
— Explore advanced modeling methods and run multiple experimentations to boost performance

  • Identifying the relevant external data sources to assess and evaluate their impact each external data source and its features had to be well researched and validated from the various business teams
  • Capturing and outlining the incremental impact of each data source/variable and modeling technique to obtain technical and business validation for each experimentation

Solution Methodology
  • Gained a thorough understanding of existing internal core model and data sources
  • Identified several external sources and tested their relevance
  • Evaluated multiple ML modeling methodologies/techniques to estimate the mortality probability for each customer
  • Designed multistage models to improve prediction and isolate effect
  • Ensured in-depth technical and business validation at individual factor analysis level to verify model stability
Business Impact
  • 30%+ improvement in mortality rate prediction

  • Improved and more effective pricing as mortality rate was one of the key inputs in pension risk transfer pricing

Copyright © 2023 Tiger Analytics | All Rights Reserved