How to project trends in mortality after Covid-19: the CMI consultation

Published by Will Rice on

Estimated reading time: 6 minutes


On 21 September the Continuous Mortality Investigation (CMI) published a consultation on a modification to its mortality projections model to allow for the effect of the Covid-19 pandemic. All sounds very technical, so what does it mean for Defined Benefit (DB) pension schemes?

DB pension schemes in the UK use the CMI model to make assumptions about how life expectancy will change in the future. This affects all aspects of pension scheme funding, including funding valuations, company accounting disclosures and transfer values for members. A change to the model could mean deficits growing or shrinking, and good or bad news for members transferring their pension to another scheme. Understanding how the model may change is key for trustees and sponsors to anticipate upcoming shocks to their scheme’s financial position.

Important choices for DB schemes

The CMI model is deliberately flexible. Schemes can choose how to pull the levers in the model to reflect their view on future life expectancy. This means the impact of the release of a new model isn’t “one-size-fits-all”. Instead trustees and sponsors (and their actuaries) need to look at emerging mortality data and any information they have on the likely longevity of their members, to work out how to pull the levers.

In their consultation, the CMI are proposing the introduction of a new lever to pull, which is all about Covid-19.

Why Covid-19 requires a new lever

The CMI model uses data on trends in historical deaths to project how deaths in the future may pan out, under the assumption that the recent past is a good guide to the near future. This is illustrated in the following chart. The dots in the chart show actual mortality improvements, averaged over five years and the lines show fitted improvements from the model, which smooth out the dots and blend into the future improvements.

However, the Covid-19 pandemic has led to deaths in the UK that are so extremely abnormal that this has called into question whether using the recent past as a guide still makes sense. At the time of writing (October 2020), there have been about 60,000 more deaths in 2020 in the UK than what would have been expected based on deaths in 2019. This is equivalent to a worsening of mortality rates by 10.4%. The CMI model, in normal years, would say that an increasing trend in deaths means that this trend will probably continue for a few years. However, this isn’t a normal year and the recent dramatic spike in deaths is (we all hope) a one-off event, and not a step change in trend that will continue to play out in similar fashion for many years to come. 

So a change to the model is needed. The CMI are proposing to add a new lever to the model so users will be able to adjust how much weight is placed on the 2020 data. 

The impact of varying the weight parameter is shown in the following charts, taken from CMI Working Paper 137. The vertical axis shows the impact on life expectancy, for the different values of the weight parameter on the horizontal axis. The different colour lines show what the impact might look like depending on how the mortality data emerges over the remainder of 2020. At the time of writing the 2020 data is in line with the yellow “-10% improvement” line. If that is the case for the whole of 2020, then giving full 100% weight to data for 2020 would lead to life expectancy falling by over 4% for males.
 

Of course, the future of Covid-19 remains highly uncertain. Even in normal years, making assumptions about future mortality is challenging, but Covid-19 has made even the near future very unpredictable. But DB schemes must set these assumptions, and so a key question will be how far to pull the new lever. 

What is the impact of the “2020 weight” lever?

Weight of 0%: Using a weight of 0% is broadly equivalent to ignoring the data for 2020 altogether, as if 2020 had never happened. Schemes taking this option will be taking the view that future mortality will be largely unaffected by the experience observed in 2020, and mortality data will revert to pre-Covid-19 levels in the future. Life expectancies under this approach would show only a small fall compared to the previous 2019 version of the CMI model.

Weight of 100%: Using a weight of 100% fully incorporates the data for 2020 into the model, and projects this increase in deaths forward into future years. It is important to understand that this doesn’t mean the model assumes 2021 is the same as 2020, in fact the model averages and blends the 2020 data with previous years, and then assumes that the trends in death rates fade away over time. Nevertheless, schemes taking this option will be taking the view that there has been a significant step change in mortality that isn’t just affecting mortality rates now, but for many years to come. Using a weight of 100% may fit the view that there will be future waves of Covid-19, or exposure to the virus leads to a frailer population, or both. As the charts above show, life expectancies (and by extension, liabilities) under this approach would fall dramatically (for example, by c. 4% compared to the 2019 version of the CMI model assuming the “–10% improvement” scenario continues to hold out for the rest of 2020).

Weight of between 0% and 100%: Setting the weight parameter between the two extremes would reflect a balance between the views above, and would result in a liability outcome between the two extremes. Perhaps this feels like a balanced approach, although we should all be careful of the effect of framing; when presented with any three options the “one in the middle” always seems a good compromise, but we need to consider whether each option is reasonable objectively and on its own merits. 

The big question is therefore how far to pull the lever – what weight value is appropriate? This is a question that all schemes will need to grapple with when they adopt the new version of the model. 

It is important to approach this question with care, and understand the risks and uncertainty involved with setting an assumption. We should question whether this assumption is a significant factor for the scheme. If so, we should be wary of reaching a conclusion before having considered the alternatives, as the choice of approach may have a significant impact. 

Want to know more?

Our specialist Longevity Consulting team are considering how to amend our views of future life expectancy in light of Covid-19, including how we would propose to pull the lever in the CMI_2020 model. We will be discussing our thoughts with Alan Baker, Trustee Director of LawDeb, in our webinar “What is the impact of Covid-19 on life expectancy” on 3 November. 

How has Covid-19 changed life expectancy?

Join experts Simon Bramwell, Principal, Head of Longevity Risk Transactions, Will Rice, Partner and Alan Baker, Trustee Director from LawDeb for a joint webinar in which they discuss how trustees should be reflecting on the significant mortality trends in recent years in light of the impact of Covid-19. 

Register now

Stay up to date

Get the latest independent commentary and exclusive insights from a range of experts at the forefront of pensions, investment, risk and insurance – tailored to your preference.

Subscribe today