Open Mortality: Life Tables for Swiss Pension Funds from Public Data

Swiss pension funds value their liabilities with proprietary life tables based on data from 14 large funds. This project tests whether comparable tables can be derived from public data (FSO, HMD) using Lee-Carter and CBD models, and provides an open, reproducible data pipeline.

Key data

Description

Swiss pension funds are legally required to value their liabilities using actuarial life tables (technical bases BVG). These are published by Aon Switzerland and Libera AG and are based on observations from 14 large autonomous funds combined with trend models from the Federal Statistical Office (FSO). They are licensed products and costly for smaller funds. From a data science perspective, this project investigates whether high-quality life tables for the Swiss occupational pension context can be derived from publicly available data (FSO, Human Mortality Database).

Two questions are central: How closely can classical models (Lee-Carter, Cairns-Blake-Dowd) replicate the published BVG benchmarks? And how large is the selection bias between population and pension fund mortality, and can it be quantified without proprietary data? In addition, earlier BVG generations are validated retrospectively against subsequently observed mortality. In a second phase, the results are translated into conversion rates, pension capital and longevity risk margins.