CMI 2022: what weight works?

Despite daily life returning to some form of normality in 2022, the number of deaths in the UK continue to be high. This article outlines the challenges of using recent mortality data to set mortality assumptions and explores ways that companies could implement the necessary changes in approach. For an overview of why deaths in 2022 continued to be high read our previous blog.

Setting post-pandemic improvement rates

Before the pandemic, the common approach to setting mortality improvement rates was to use extrapolative models, such as the Continuous Mortality Investigation (CMI) Mortality Projection model. These types of models are typically calibrated to historical population data and project future mortality rates. They work well in times of relatively stable deaths but are very sensitive to extreme events such as the pandemic.

The extreme mortality observed in 2020 and 2021 is unlikely to be indicative of future mortality, but using recent data could prove valuable. 2022 potentially represents some form of “new normal” so it makes sense to consider taking account of this. There may be reasonable grounds to justify a reduction in future life expectancy since 2020 but the big question that remains is how much importance to place on this recent data.

The new CMI 2022 model

In response to this challenge, in their latest model (CMI_2022), the CMI have retained the same structure and core parameterisation as CMI_2020 and CMI_2021, with the flexibility to specify the weight placed on recent years of data[1]. Under the core parametrisation, but placing the full weight on 2022 data, there is a noticeable drop in projected life expectancy from age 65[2] of the order of 3-4% for both males and females. If instead we use the core parameterisation of 25% weight on data from 2022, this drop is closer to 2-3%.

The value of this weighting parameter has a material impact on the final result and is also highly subjective. Given the value of 25% has been stated as the default value, there is a risk of herding in the market around this number. It is important that companies give thought to what weight is appropriate and can justify their rationale for the weighting parameter they choose to use.

Our view is that a fundamentally different approach to setting mortality projections is now required, which focuses on the real-world drivers of future mortality. This requires greater use of expert judgement and consideration of other sources of information beyond the population statistics that have traditionally been used.

Drivers of post-pandemic mortality experience

To establish a view of “post pandemic trend”, a view needs to be formed on how the main drivers of excess mortality will evolve. This involves identifying what has been driving excess mortality to date and considering what could influence mortality rates in the future.  Some of these drivers are set out below.

• COVID-19: Around 22,000 deaths in 2022 were directly attributable to COVID-19, and at the end of 2022 the ONS estimated that 2 million people were experiencing self-reported “long-COVID”, experiencing signs and symptoms of COVID-19 weeks after having the infection. As the disease is still relatively new, the potential long-term impact of COVID-19 on health is not yet fully understood.

• Delays to healthcare: The pandemic created significant disruption to normal healthcare services in the UK. Data has subsequently emerged showing huge backlogs and increased waiting times for patients. Based on our analysis, long A&E wait times are thought to be one of the key drivers behind excess mortality in 2022.

• Strains on NHS funding: The COVID-19 pandemic created massive economic disruption worldwide, as large swathes of the economy were shut down for months to control the virus. The recent high inflation environment has put further downward pressure on the funding of an already stretched health and social care sector.

• Cost of living: With prices rising faster than household incomes, recent high levels of inflation are putting pressure on what individuals can afford. The number of people living in fuel poverty and the associated deaths from living in cold homes has been reported to have increased. Living in cold, damp homes has been attributed to increased risk of cardiovascular and respiratory conditions.

Most of the research and insights available on these drivers are based on population level statistics. It will be important for insurers and reinsurers to consider how the experience of the generally higher affluence insured population compares to that of the general population. 

How to implement the change

Once a view of future mortality has been formed, there are several ways of practically implementing changes. Implementation may vary depending on the purpose (e.g., experience analysis, valuation or pricing) and it is important to carefully consider the interaction between the baseline and trend assumptions.

Expert judgement overlays can be constructed, which reflect the real-world change that may be expected relative to a “pre-pandemic” view i.e., a view of assumptions before the pandemic. The pre-pandemic assumptions with the overlay applied on top can be re-expressed using the CMI model, back-solving to find an alternative parameterisation which has the same impact. Parameters which could be changed include the long-term rate, the weight placed on 2022 data, and the period smoothing parameter. In doing this, you should be careful that the shape of improvements is in line with expectations.

If it’s believed that the pandemic and subsequent fall-out has caused a “step-change” in mortality, then it might be more appropriate to make a change to the base table rather than changing the future mortality improvements. Care will be needed when considering the implementation of any changes. 

We have significant experience in supporting clients with their assumption setting and internal model developments. We would be delighted to discuss any of the above or any other related topics with you in more detail. For more information, please get in touch.