Pensioners under the microscope: zooming in on the mortality of bulk annuitants and private sector pensioners

The CMI has a target for closer collaboration between the Annuities Committee (which collects data for bulk annuitants from UK insurance companies) and the SAPS Committee (which collects data for self-administered pension scheme members), reflecting the increasing transfer of defined benefit pensions from pension schemes to bulk annuity insurers. The publication of CMI Working Paper 210, which compares data, published mortality tables and mortality experience for bulk annuities and private sector pension schemes in 2016-2023, marks the first step in our closer relationship. 

Comparing apples with oranges – breaking down the datasets

Our first task was to compare the data CMI collects for bulk annuities and SAPS pensioners. If we simply look at the entire Bulk annuity and SAPS datasets, it seems we are comparing apples with oranges. Some of the key differences are:

• Pensioner type: The majority of SAPS data includes an indicator of pensioner type and all analysis and mortality tables are differentiated into pensioners and dependants (and sometimes further into normal-health and ill-health pensioners). This is not the case for Bulk annuity data, which lacks a reliable indicator of pensioner type in sufficient volumes for analysis. In Working Paper 210, we created an aggregate SAPS dataset which is not differentiated by pensioner type (unusual for SAPS data) to enable more consistent comparison with the Bulk annuity dataset.
• Private/Public sector data: The SAPS dataset includes both public and private sector pension scheme members, and public sector members make up a particularly large proportion of the female pensioner and male dependant datasets. Our understanding is that the Bulk annuity dataset consists only of individuals who were previously members of private sector pension schemes. For the analysis in Working Paper 210, we have restricted the SAPS dataset to private sector data only. 
• Data collection timing and analysis periods: SAPS analyses typically cover an 8-year period with data weighted towards the start of any given period as data submissions tend to occur in triennial cycles and it takes time for data volumes to build up in the later years of a dataset. For bulk annuities, data is received annually for the period to the end of the previous calendar year and analysis tends to be for a quadrennium (or a single year since the COVID-19 pandemic). We have mainly focused on analysis of 2016-2023 in Working Paper 210, for consistency with SAPS analysis. A shorter period would see a dramatic drop in data volumes for SAPS and those schemes included in later years may not be representative of the dataset as a whole.

Those oranges are now looking more like apples on the outside but what about on the inside?

After these dataset adjustments, at overall level we observe little difference between the mortality experience of the Bulk annuity and SAPS datasets, particularly on a lives-weighted basis. We dig further into the factors we have available for analysis to see what is hidden under the surface.

How much of a difference do amount band and socio-economic group make?

The SAPS Committee produce a wide range of tables based on the SAPS dataset, including tables differentiated by pension amount band. A new addition for the “S4” Series of tables was the use of Index of Multiple Deprivation (IMD) deciles alongside pension amount bands to graduate new “IMD tables”. Such granular tables are not available for bulk annuities, so we wanted to see how well these differentiated SAPS IMD tables reflect experience of the Bulk annuity dataset.

We compared bulk annuity amounts-weighted experience relative to the projected “S4” Series IMD tables with experience relative to undifferentiated “S4” Series tables. As seen in the charts, experience relative to undifferentiated tables (the gold lines) shows a steep gradient, i.e. mortality gets lighter as IMD group increases. Whereas experience relative to the IMD tables (the blue lines):

• For males is fairly flat suggesting that these tables do a good job of explaining differences in experience by IMD group for the Bulk annuity dataset, even though the tables are based on data for SAPS members.
• For females, a gradient remains suggesting that not all of the difference is explained by these tables, although this might be due to the mix of pensioner and dependant data in Bulk annuity IMD dataset.

Chart: 100A/Es by IMD Group for the Bulk annuity dataset – Males (Left) and Females (Right)

We also considered experience of the SAPS dataset relative to the “S4” Series IMD tables and observed similar conclusions. This suggests that when we allow for differences in pension amount and socio-economic status, the underlying experience of the Bulk annuity and SAPS datasets are similar, which should be comforting for anyone using assumptions based on SAPS tables as part of bulk annuity transactions.

What’s in an age?

At overall level, it seems experience of Bulk annuities and SAPS are similar, but the age profile and experience by age do show some differences. These were interesting to observe, and readers may want to consider further.

We note that the Bulk annuity dataset is more mature than the SAPS dataset, with a higher average age measured both by exposure and (to a lesser extent) deaths. This is seen in very low data volumes at younger pensioner ages for the Bulk annuity dataset.

For age bands with reasonable data volumes (broadly 75+), we see similar experience for the Bulk annuity and SAPS datasets on a lives-weighted basis, both for males and females. At younger ages, experience is typically heavier for the Bulk annuity dataset than the SAPS dataset. The data doesn’t allow us to say for sure why this is the case, but we speculate that it may be due to the mix of pensioners and dependants; historical bulk annuity transactions were weighted more towards pensioner only deals potentially leading to a higher proportion of dependants at younger ages in the bulk annuity dataset than the SAPS dataset as these deals have aged. 

What next?

This analysis has been a really illuminating starting point for us and will aid our understanding of experience going forward. We intend to carry out further analysis of bulk annuities and SAPS data in future. If you have any feedback on the paper or suggestions for areas we could consider in future, please contact annuities@cmilimited.co.uk.