On 9 March 2022, the Continuous Mortality Investigation (CMI) published the latest version of its annual ‘mortality projections’ model, CMI_2021.
The CMI mortality projections model is widely used by Defined Benefit (DB) pension schemes (as well as insurers and reinsurers) to predict how current levels of mortality might change in the future. The new model requires trustees of DB pension schemes to make a conscious decision: do they feel the pandemic has influenced future mortality rates and, if so, by how much?
A period of extreme mortality
The Covid-19 pandemic resulted in another year of abnormally high deaths in 2021. The following chart illustrates this, and shows mortality rates (relative to mortality in 2000) for the England & Wales general population since 2000.
Standardised mortality rates over ages 65 to 100
Source: BW calculations based on CMI_2021
Mortality in 2021 was around 8% higher than in 2019 (the last pre-pandemic year). While this is not as high as 2020 (for which mortality was around 14% higher than in 2019), this is still extremely abnormal compared to recent years and the CMI has commented that you need to look back as far as 1940-41 to find two consecutive years as unusual as 2020 and 2021.
The CMI model works on the assumption that recent levels of deaths (or mortality) can be used to predict short-term trends in future mortality. This process works well in ‘normal’ times, when annual changes in mortality are within a relatively small range. Unfortunately, in abnormal times the model does not react well to short term ‘spikes’ in mortality and leads to significant and unreasonable falls in life expectancy (and therefore liabilities).
To address this issue, the CMI has taken a consistent approach in CMI_2021 as for CMI_2020 and introduced the new ‘2021 weight parameter’ to exclude data for 2021 from the default version of CMI_2021. This means that the default version of CMI_2021 excludes data for both 2020 and 2021.
What is the ‘weight parameter’?
Example 1: A weight parameter of 100% for 2021 means that the 2021 death data would be used as normal in the model.
Example 2: A weight parameter of 50% for 2021 means that the 2021 death data is used, but the impact of this data on the results is reduced by broadly half.
Example 3: A weight parameter of 0% for 2021 means that the 2021 death data is ignored by the model. This is the default approach used by the new mortality model.
How does the output of CMI_2021 compare to previous versions of the model?
Given the default versions of CMI_2021 and CMI_2020 have such similar datasets, because they both ignore data from 2020 onwards, as you might expect the results from both models are very similar.
In the default version of CMI_2021, life expectancy is around 0.2% lower than in CMI_2020 for both males and females.
If you are considering adopting CMI_2021 then this may be because you are in the process of your triennial valuation, and at your previous valuation three years ago you may have used CMI_2018.
In the default version of CMI_2021, life expectancy is around 0.3% lower for males and 0.1% higher for females than in CMI_2018.
Allowing for the impact of the pandemic on future life expectancy
As pension schemes are concerned about paying benefits to members in the future, currently the key discussion point for mortality assumptions is the extent to which the Covid-19 pandemic will affect mortality in the short and medium-term future (around the next 10 years).
How might the pandemic affect future mortality?
While direct deaths from Covid-19 is an obvious consequence of the pandemic that will likely result in an increase in future deaths, there are also a number of indirect ways in which the pandemic could affect future mortality:
| Effects that will increase future mortality (i.e. reduce longevity) | Effects that will decrease future mortality (i.e. increase longevity) |
| Missed diagnoses and treatments due to medical services operating at a reduced level during the pandemic. | If frailer individuals were more likely to die during the pandemic than stronger individuals, the post-pandemic surviving population might on average be stronger than the pre-pandemic population and therefore have longer life expectancy than before the pandemic. |
| Long-term health complications from catching Covid-19. | Better public hygiene (e.g. wearing masks indoors and washing hands more regularly). |
| A reduction in healthcare and other public spending due to the impact on the economy of the pandemic (e.g. increased public spending on furlough scheme, economic contraction during the pandemic). | Using tools developed during the pandemic (e.g. shielding the vulnerable, mRNA vaccines) to assist in combating other diseases such as influenza. |
| Possible emergence of vaccine resistant strains. | Increase in public healthcare spending due to the increased focus on the NHS. |
It still remains highly uncertain whether the pandemic will lead to more or less deaths in the future than we might have expected prior to the pandemic, and it is likely that a combination of many of the above factors (as well as other currently unknown factors) will occur.
What further factors do pension scheme stakeholders need to consider when setting mortality assumptions?
As well as the underlying mortality reasons listed above that will drive views on future mortality, pension scheme stakeholders (e.g. trustees, sponsoring employers, auditors, insurers, regulators etc.) have other issues to consider that will factor into their decision making on these assumptions.
| The Pensions Regulator guidance | Sponsoring employers – accounting |
|
The Pensions Regulator has published guidance to pension scheme trustees that any material changes to the scheme funding mortality assumptions due to the impact of the pandemic should be evidence-based and monitored. This may steer trustees to not make a pandemic deterioration allowance in the mortality assumptions. |
Auditors appear to be amenable to a reduction in liabilities of up to c. 1% as a result of the pandemic. As the directors of the sponsoring employer have discretion over the accounting assumptions adopted, the pandemic deterioration could be used to improve their accounting disclosures. |
| CETV basis | Bulk annuity pricing |
|
As a best estimate basis, trustees and sponsoring employers can be less cautious when setting the CETV basis than in the prudent scheme funding basis. The CETV basis could therefore fully reflect the views of the stakeholders in the pension scheme. |
Although it is too early to say for sure, given how cautiously insurers price their liabilities it is unlikely that bulk annuity pricing will fall in the short term due to insurers allowing for any pandemic deteriorations in their mortality assumptions. If you’re interested in reading more about the impact of the pandemic on bulk annuity pricing, read our report Bulk annuities: focusing on longevity. |
How can the impact of the pandemic be reflected in my assumptions?
When CMI_2020 was published, the 2020 weight parameter was a natural parameter to use to reflect views on any negative impacts on life expectancy arising from the pandemic.
With the release of CMI_2021 and the 2021 weight parameter, there are now four CMI model and weight parameter combinations that could be used to reflect the impact of the pandemic:
- CMI_2021; adjust the 2020 weight parameter; make no adjustment to the 2021 weight parameter (yellow line in below chart)
- CMI_2021; adjust the 2021 weight parameter; make no adjustment to the 2020 weight parameter (green line in below chart)
- CMI_2021; adjust both the 2020 weight parameter and the 2021 weight parameter (light blue line in below chart)
- Alternatively, trustees could also decide to adopt (or retain if they’ve previously adopted) CMI_2020 and adjust the 2020 weight (dark blue line in chart below).
With no guidance from the CMI as to which to opt for, trustees will need to decide how to make these adjustments with the weight parameters.
To aid with this discussion, the following chart shows life expectancies (for males at age 65 calculated using the S3PXA base tables) calculated using each of the above four methods. The chart for females is similar.
Indicative life expectancies from different
combinations of parameters in the CMI model
Source: BW calculations based on CMI_2021
This kind of chart will be helpful in discussions between different stakeholders in understanding how the mortality projection assumptions they put forward compare to assumptions put forward by other stakeholders. If you look across the dashed grey lines on the chart, pension scheme liabilities calculated using the four different approaches are broadly the same.
The following table shows how liabilities calculated using two 2020 weight parameters in CMI_2020 broadly compare to different weight parameter combinations in CMI_2021.
| CMI_2020 2020 weight |
CMI_2021 2021 weight only (nil 2020 weight) |
CMI_2021 2020 weight only (nil 2021 weight) |
CMI_2021 2020 and 2021 weight |
| 15% | 10% | 10% | 5% |
| 25% | 20% | 20% | 10% |
Interested to find out more information?
On 22 March 2022, Simon Bramwell (Principal and Head of Longevity Risk Transactions at Barnett Waddingham) and Will Rice (Partner and Scheme Actuary at Barnett Waddingham) will be hosting a webinar with industry expert Rhian Littlewood (Senior Business Development Manager at Standard Life) will review the emerging mortality data (including CMI_2021) and consider the implications for DB pension schemes and their journey plans. Register for our webinar here.
How does life expectancy affect your DB journey plan?
Longevity remains a hot topic for trustees and sponsors. Join our webinar to understand how the impact of life expectancy outcomes could affect DB pension schemes.
Register today
Stay up to date
Get the latest independent commentary and exclusive insights from a range of experts at the forefront of pensions, investment, insurance and risk – tailored to your preference.
Subscribe today