The first results from the 2021 census show the population of England and Wales getting bigger (up 6.3% since 2011) and older (with the proportion of people aged 65+ rising from 16.4% to 18.6% and now outnumbering the under 16s).
This direction of travel should come as no surprise, but the precise numbers are not quite what earlier estimates had suggested the census should show. The census recorded fewer people than expected across many age groups, particularly at older ages. For example, there seems to be almost 35,000 fewer people aged 901 and over than expected in England and Wales when compared with the ONS’s population projections.
The reason for the discrepancy will likely be the subject of some scrutiny by ONS, but small errors can quickly magnify over time when trying to track a small cohort of elderly people that naturally but rapidly shrinks – as the ONS population estimates attempt to do. For those managing defined benefit pension schemes, the more important question is how this affects the value of their liabilities.
If there are fewer people aged over 90 now than thought, it means those over 80 in 2011 who have since died represent a greater proportion of the elderly population – ie, mortality rates over the past decade were higher than previously believed. As a crude example using ONS data for England and Wales, 43,654 men aged 90+ in died in 20212. This is 23.6% of the 185,339 male population aged 90+ previously estimated by the ONS for mid-2021, but 25.8% of the 169,200 male population found in the latest census for 21 March 2021. This in turn also means improvements in mortality rates over this period will have been slower than thought (and they were already quite slow in historical terms).
All in all, this could mean that pension scheme member life expectancies have been overstated: actuarial models use Office for National Statistics data to calculate mortality rates and mortality improvements, which in turn feed into estimates of what mortality rates will be in future.
A similar issue occurred 10 years ago on the release of the 2011 census. The UK actuarial profession have already estimated that a comparable data correction in 2021 would reduce male life expectancies by around 1%, with female life expectancies reduced by around 0.5%3. That could translate to a reduction in liability values of up to 1% for a typical defined benefit scheme – more than £10bn across all DB schemes outside the public sector. The revisions at ages 90+ are smaller this time, but we’re also seeing larger deviations at ages below 90 in the 2021 census which could translate into even higher reductions.
It’s too early to know whether this year’s census correction will lead to that big a reduction in life expectancies and liability values – it takes time to work through the revised population figures for past years, and the updated mortality models won’t be available until next Spring in line with the usual annual release calendar. Further, the full impact feeds through on the basis of accepting the new census data is correct and that the error lies in past figures/tracking – highlighting the uncertainty in life expectancy estimates. Nevertheless, it is looking likely that, all else equal, reflecting the census data in mortality assumptions will improve funding levels.
1 Comparing census figures (for 21 March 2021) to the mid-2021 estimate from the ONS’s 2020-based population projections.
2 Source: ONS, “Deaths registered weekly in England and Wales, provisional”.
3 The 2011 census population correction was much smaller for females compared with males, primarily due to the larger population which in turn reduces the likelihood of any population errors.
