Covid-19 – how unusual is recent mortality?

Estimated reading time: 5 minutes

Our previous blogs discussed how the Covid-19 pandemic might affect longevity for pension schemes and annuity portfolios. Here we have updated our analyses to reflect recent coronavirus data and will compare mortality in 2020 to earlier years to show just how unusual the recent experience is. 

Recent Covid-19 data and excess deaths

We start, as in previous blogs, with the widely-quoted Public Health England (PHE) data for reported daily deaths of those hospitalised and with a positive test for Covid-19. The figures are affected by weekends, typically being lower on Sunday and Monday, so we use an average to smooth out this effect. 

Chart 1 shows seven-day averages, plotted against the middle of the seven-day period. We also show the hypothetical position if the fall in deaths mirrors their rise. The chart shows that, although we appear to be well beyond the peak for hospitalised Covid-19 deaths, the current decline in reported deaths is much slower than the earlier rise.

Chart 1 – Weekly average of reported UK Covid-19 deaths in hospitals

While the PHE data gives some indication of the current trend, we are most interested in the impact of Covid-19 on total mortality, including deaths from all causes and those outside hospitals.

Chart 2 compares weekly registered deaths in England & Wales in 2020 (highlighted) with those in the corresponding weeks of 2005 to 2020. The latest week (11 to 17 April 2020, or “week 16” of 2020) has a higher number of deaths than any other week shown.

Chart 2 – Weekly registered deaths in England & Wales, 2005-2020

The Continuous Mortality Investigation (CMI) used the ONS data to update its analysis of “excess” deaths, comparing actual deaths to those that would have been expected in the absence of a pandemic. The CMI estimates around 41,000 excess registered deaths in England & Wales by 27 April 2020. This estimate excludes those that occurred by then, but had not been registered.

A striking feature of the ONS data is the difference between excess deaths in care homes and hospitals. Chart 3 compares deaths in each week to those in week 12, before the pandemic had a material impact. 

Chart 3 - Excess registered deaths in England & Wales by place of occurrence

There has been a much more rapid increase in excess deaths in care homes than in hospitals or homes. This suggests that the peak for total excess deaths is likely to be later than the peak for hospital deaths, and the PHE figures for deaths in hospitals may not be a reliable guide to the developing impact of the pandemic on total excess mortality.

2020 in historical context

To put the mortality experience of 2020 into context, Chart 4 shows historical mortality improvements in England and Wales. These improvements are standardised to show the impact on a consistent population with a modern-day age structure. 

Chart 4 – Annual mortality improvements in England and Wales

We have added a grey “bottle” shape that covers all of the points, to emphasise that mortality improvements were more volatile in the distant past. There are various plausible reasons for this, including the reduced impact of communicable diseases in recent times (apart from this year), although the improvements may be affected by lower data quality in the earlier years. 

The last 40 years, within the “neck” of the bottle, all have improvements of between -4% and +6%. However, we are already well outside that range for 2020. The CMI’s estimate of excess registered deaths to 27 April would contribute around -8% to the mortality improvement for 2020, and further excess deaths could lead to a much lower (i.e. more negative) improvement for the year as a whole. 

Most mortality models and datasets use data for calendar years, as we have done in Chart 4, and this choice has an impact on our analysis and on longevity models. A period of excess mortality has a larger impact on annual mortality improvements if it falls entirely within one year, rather than spanning two years. 

Challenges for longevity models

When modelling mortality, it is common to treat the recent past as a good guide to the short-term future, and to think of observed mortality as a combination of a smooth underlying trend and short-term volatility. We would like to remove the volatility in order to understand the underlying trend. This can be done by simply averaging over a number of years, or by fitting a more sophisticated model. This approach works well during a period of modest volatility, as seen in the last 40 years, but the extreme mortality experience in 2020 will be a challenge for existing methods.

Chart 5 compares historical standardised mortality rates in England & Wales to the underlying trend, using the Core calibration of the CMI_2019 Mortality Projections Model. 

We have included an illustrative scenario for 2020, based on a mortality improvement of -15%. We stress that this is just a scenario, using a round number near the bottom of the historical range in Chart 4, rather than an estimate. There are too many uncertainties to make a reliable prediction of the eventual mortality improvement for 2020 at this stage.

Chart 5 –Standardised mortality rates, and underlying smooth trend from CMI_2019

The trend in mortality rates before 2020 is fairly clear, at least with hindsight. While views may differ on exactly how to fit the pre-2020 trend, all of the actual mortality rates in Chart 5 lie within 3% of the trend shown (in relative terms) and it would be difficult to justify a trend outside this range. But we expect that mortality in 2020 will be well above the previous trend.

The unusual mortality experience in 2020 will make it more important than usual to form an opinion on how future improvements may differ from the previous trend, rather than relying on historical data and smoothing algorithms. Will 2020 be a one-off, with mortality in 2021 and beyond returning to the previous trend? Will the coronavirus continue to circulate in later years? Will an effective vaccine be developed? What will be the impact of diagnosis or treatment of other conditions, such as cancer, being delayed due to the pandemic? And how will a likely economic recession impact mortality?

If you would like to talk about this topic or have any questions in general, please get in touch with your usual Barnett Waddingham contact to find out how we can support you. Alternatively, please contact me below.

Visit our Covid-19 hub for the latest expert analysis on the coronavirus impact on business, investment, insurance and pensions.