Initial census results suggest lower life expectancies in the general population

The globe with data statstics zooming in

This blog considers how initial results from the 2021 census in England & Wales might lead to lower estimates of recent and future mortality improvements and cohort life expectancies from the CMI Model than previously thought.

Population estimates

The CMI Mortality Projections Model is calibrated to data provided by the Office for National Statistics (ONS) for deaths and populations in England & Wales. The mid-year population estimates are based on censuses held every decade and are rolled forward between census dates using data on births and deaths and estimates of migration. When new census results become available, population estimates are typically restated for the period since the previous census.

Initial census results

The ONS has recently published initial results from the England & Wales census held on 21 March 2021. At this stage it has only published age-grouped population figures at the census date. While this enables an initial estimate of the impact on results from the CMI Model, we stress that our calculations are approximate at this stage and will be refined as the ONS publishes further results. We expect the ONS to publish mid-2021 population estimates in the autumn, and revised estimates for mid-2012 to mid-2020 in early 2023.

As a first step, we have estimated the mid-2021 population based on the census data and compared this to the estimated mid-2021 population underlying the latest version of the CMI Model, CMI_2021. Since the CMI Model covers ages 20 and 100 we restricted our analysis to this age range.

The census and CMI_2021 mid-2021 estimated populations are similar in size, but there are material variations in the composition by age and gender:

  • The largest differences are at young ages. The census population is 6% lower than the CMI_2021 for males aged 25-29 and 6% higher than CMI_2021 for females aged 30-34.
  • There are also significant differences for ages 85-89, where the census figures are 4-5% lower than CMI_2021 for males and females.
  • Compared to the CMI_2021 population, the census population is around 0.5% higher for ages 20-64, 1% lower for ages 65-84, and 2.5% lower for ages 85-100.

Impact on the CMI Model

We have considered what impact the initial census results would have on life expectancies from CMI_2021.

To do this, we have estimated revised population estimates for mid-2012 to mid-2020 as well as mid-2021. We have taken a simple approach, and the eventual ONS figures may well differ, particularly given the age-banding in the initial census results and complications of mortality during the pandemic. Our calculations use the Core version of CMI Model, apart from modifying the population data, and we have considered cohort life expectancies using the S3PxA tables with an illustrative long-term rate of 1.5% p.a.

For males, initial mortality improvements from 2021 are lower at all ages when using the revised population estimates, with a difference of up to 0.25% p.a. For females, revisions to the initial mortality improvements are smaller, and negative at some ages.

The impact of the census results on life expectancy depends on the approach taken by a user of the CMI Model:

  • If mortality improvements are based on the general population:
    • Considering only future mortality improvements, from 1 January 2022, the revisions might lead to reductions in cohort life expectancies at age 65 of around 0.3% for males and 0.2% for females
    • There will be a greater impact if the census data also affects views of historical improvements. Considering mortality improvements from 1 January 2013, the effective date of the “S3” Series mortality tables, the revisions might reduce cohort life expectancies at age 65 by around 0.7% for males and 0.3% for females.  
  • We encourage users to tailor the CMI Model to reflect the characteristics of the specific population that they are using it for. Where views of mortality improvements are based on other data, the impact of the census figures on results is likely to differ from that for the general population.

The impact of the initial census results on CMI_2020 is similar, although slightly smaller.

What next

We intend to revise this initial analysis as the ONS publishes further census results and expect to include further analysis in our “interim update” working paper towards the end of the year once we have ONS’s official mid-2021 population estimates. Even then, it will only be after the revised estimates for mid-2012 to mid-2020 are made available in 2023 that we’ll get a clear picture of the overall impact on the results of the Model.

We are also considering the timing of the release of CMI_2022. In recent years, new versions of the CMI Model have typically been published in March. However, as it appears that revised mid-2012 to mid-2020 population estimates are likely to affect our projections we are minded to wait for ONS to publish these figures even if it would delay CMI_2022. We aim to communicate a decision on timing by the end of this year when we’ll have greater clarity about when the revised population estimates will be released.

Read more blogs from the CMI here.