Michael Anderson (of RGA), a member of the CMI Annuities Committee, reflects on the key findings of their analysis of mortality experience by IMD to 2022.
The CMI Annuities Committee has published our second analysis of mortality experience by socio-economic status (Working Paper 204). It is based on the subset of the 2022 pension annuities in payment dataset that includes indicators based on the Index of Multiple Deprivation (IMD), and region. We first analysed experience by IMD for 2015-2018 in Working Paper 138. Coverage of data including IMD fields has since significantly increased from 48% of the full 2015-2018 dataset in Working Paper 138 to 77% of the 2022 dataset, now comprising over 3.4 million life-years of exposure and nearly 134,000 deaths.
The Committee is very grateful to data contributors for their efforts in providing the data to enable high resolution analysis of the annuities experience data.
We used two-way analysis of Age Standardised Mortality Rates (ASMRs) and Age and Deprivation Standardised Mortality Rates (ADSMRs), as well as a simplified Generalised Linear Model (GLM) to explore the drivers of mortality differentials by socio-economic status and region. Many of the key findings from Working Paper 138 have been confirmed by the latest analysis, which is reassuring as they are in line with common prior expectations regarding the relationship between socio-economic status and mortality outcomes. There is a clear gradient in mortality experience, with the most deprived decile (IMD 1) consistently exhibiting the highest mortality rates and the least deprived decile (IMD 10) the lowest. This remains the case when comparing experience across several variables, including geographic region, annuity amount band, and product type.
The GLM analysis shows that most of the variation in mortality experience in the IMD dataset is associated with socio-economic status. It was also interesting to see that the degree of inequality has been relatively stable over time, with the slope of ASMRs by IMD decile being relatively similar across all calendar years since 2015, even during the pandemic years. Shifts in the ASMRs were of a relatively parallel nature as mortality increased in 2020 and 2021.
As was the case last time, the latest analysis also reveals significant regional variations in mortality experience. For instance, Scotland exhibits higher ADSMRs compared to England and Wales, with notable regional disparities within England itself. The "North/South divide" is evident, with higher mortality rates in the North of England compared to the South, even after controlling for IMD. This regional variation underscores the importance of including geographic factors in mortality analysis and highlights the persistent mortality differences across different parts of the UK. It may be interesting to consider whether the existence of regional variation after controlling for socio-economic differences suggests that there are additional socio-economic factors that the current metrics are not capturing, or whether there might be some other reason for regional variation, for example lifestyle factors not directly linked to socio-economic status.
At the same time as many of the key results have remained the same as in the earlier analysis, there have been some changes as well. When comparing the IMD dataset to the general population, the ratios of ADSMRs for different regions have become more similar to each other (averaging around 85% for males and females). Previously, we saw greater disparity between English regions. This suggests that after accounting for socio-economic status, regional variation in the annuitant population is in line with that of the general population, which makes more intuitive sense. It was also interesting to see that the ASMRs for Wales in the 2022 data are much more aligned with those of England for lower IMD deciles, but higher for higher IMD deciles. This could suggest less inequality in Wales than in England, but care needs to be taken in making inferences given lower data volumes and different definitions of IMD in the two nations.
We have also been able to consider whether the Pension Freedom reforms have had an impact on mortality outcomes. Interestingly, despite expectations of greater select effects as individuals have been able to take their pension savings as a lump sum instead of an annuity, there is no significant difference in experience by commencement period. ASMRs and ADSMRs for both the pre-freedoms and post-freedoms groups are similar for both sexes, suggesting little difference in mortality levels for individuals purchasing annuities after the introduction of pension freedoms relative to before. This is potentially a surprising result. Data for the post-freedoms period, however, remain relatively limited and so any conclusions at this stage can only be tentative.
Tracking progression of the IMD annuitant dataset by considering ADSMRs by year since the 2015-2018 analysis allows us to consider the impact of COVID-19 on our dataset. ADSMRs for 2020 are significantly elevated compared to previous years, reflecting the heightened mortality during the peak of the pandemic. It was interesting that by 2022, ASMRs were aligned with the pre-pandemic level for both males and females suggesting a relatively rapid return to normal (in a mortality sense) for annuitants.
The data confirms that the mortality impact of COVID-19 was more pronounced for males than females, and that regional disparities in mortality rates were clearly apparent during the pandemic as in normal times. For example, the ADSMRs for males in Scotland and Wales during the COVID-19 period were notably higher than those for England, with Scotland experiencing similar elevated mortality levels in both 2020 and 2021. This suggests that the pandemic's impact was not uniform across regions and highlights the need for care when analysing mortality using data from this period.
The analysis also shows that those in the highest annuity amount band experienced a much more muted impact from the pandemic, especially for males who also saw an apparently delayed impact, with a minor peak in mortality in 2021, rather than 2020. This was not the case for females, which could be because of a potentially weaker correlation between a higher annuity amount and a higher socio-economic group than for males.
Finally, there was a clear difference in experience by product type, with Individual External annuities not only experiencing much lower mortality rates across the whole period since 2018, but, especially for males, this group experienced a much more muted impact from the pandemic.
A notable aspect of the analysis is the variation in mortality experience by product type. The data shows that Individual External annuities have the lowest mortality rates by some margin, followed by Individual Internal and Bulk annuities with the highest mortality (for all IMD deciles). This pattern is consistent across both ASMRs and ADSMRs, suggesting that the differences in mortality experience are not solely due to socio-economic factors.
The lower mortality rates for Individual External annuities may be attributed to selection effects. This is likely to be linked to the impact of the enhanced annuity market, with those in poorer health who seek external quotes being diverted to that product due to the better income it will offer them. The data we are analysing here consists of those who did not purchase an enhanced annuity, which is likely to therefore comprise a healthier population in the “External” group than those in the “Internal” category who did not consider other options. It will be interesting to see whether this differential persists, with market practice tending towards more emphasis on ensuring that customers receive the best possible income from an annuity purchase.
It is well known that socio-economic status is a key driver of mortality, and that has been reinforced by our latest analysis. However, it is not the full story, and allowing for additional factors (for example product type and geographic variation) that give a more rounded insight into what is driving mortality among our annuitant population may be well worth considering.
