Life insurance companies in the UK are responding to a range of factors that are affecting mortality, from the effects of the pandemic on healthcare services to the adjustment of historical population estimates following the 2021 census.
A key component in the estimation of EoL is the rate of improvement in mortality at each age and in each future calendar year. UK life insurers typically estimate these mortality improvements using one of a series of models produced by the Continuous Mortality Investigation (CMI). These CMI models are highly flexible, allowing insurers to reflect their own views about likely changes in mortality in both the short-term and the long-term future.
In our annual survey of life insurers’ intended end-year demographic assumptions, WTW collected information on the parameterisation of these models, representing the best-estimate views of the participating insurers for the prior year-end (end 2021) and for the subsequent year-end (end 2022). Calculating the implied EoL for each insurer at a consistent date and using a common base mortality assumption allows us to see how improvement assumptions have changed over 2022.
This year’s results indicate that insurers are increasingly pessimistic about future improvements; male EoLs fell by an average of 0.5% and female EoLs by 0.33% compared with end-2021, though some of the changes in improvement assumptions implied reductions as large as 2% for males or 1.2% for females.
Most of the changes seen involved a reduction in insurers’ assumed short-term (or ‘initial’) improvements. The reasons for such changes included the effects of COVID-19. We expect that firms are concerned about delays in the provision of care by the NHS, exacerbated due to the pandemic, as well as the potential burden of cancer cases which may not have been diagnosed during the pandemic. Staffing crises, strikes and concerns around the funding of healthcare are also likely to be concerns affecting short-to-medium-term views on changes in mortality.
The changes are not entirely event-driven, however. Insurers are also carefully considering the implications of the 2021 census, which indicated that the Office for National Statistics’ (ONS) estimates of the population at high ages have been overstated in recent years. Updating the data underlying the historical improvements in the CMI models results in an increase in estimates of old-age mortality and, correspondingly, a reduction in the rate at which mortality has improved. These revisions might lead to reductions in male and female EoLs of 0.3% and 0.2% respectively. The full impact of the census on the CMI models will not be known until later in 2023 when the ONS releases further data.
Aside from changes in mortality improvements, insurers are tentatively facing up to the challenge of carrying out assumption-setting in what we might call the ‘post-COVID-19’ era.
Mortality assumptions are typically informed by insurers’ analyses of the mortality experience within their own portfolios, reflecting the underwriting (if any), socioeconomic profile and maturity of the portfolio. These factors are likely to mean that mortality within a portfolio differs significantly from the population average. There is a fundamental assumption that past mortality experience is representative of likely future mortality (once plausible improvements have been allowed for).
COVID-19 has thrown a spanner into the works; data from 2020 onwards may include additional deaths which insurers generally do not believe to be representative of future mortality, so they have (until now) largely ignored such data – excluding it from their investigations or adjusting to remove the effects of COVID-19 as far as possible. But disregarding these years introduces another problem: the historical period analysed to inform assumptions becomes increasingly dated, which is likely to make it less representative of future mortality.
The WTW Demographic Assumptions Survey indicates that the approach here is now changing, with an increasing number of firms including some data from 2020 or later in their analyses. This is true for analyses underlying protection business as well as annuity business.
Whilst the exact impact of including that data will vary across insurers, in general we would expect that including such data would (without adjustment) lead to a gradual increase in the base mortality assumptions. In combination with lower rates of improvement, this suggests a gradual release of reserves by annuity writers and, in principle, the opposite for protection writers (though the impacts at younger ages are more difficult to determine).
As we progress through 2023, we expect insurers to devote more time to analysing the extent to which the unusual mortality experience of late 2022 is representative. This will require a difficult ‘disentangling’ of factors as well as an appropriate allowance for endemic COVID.
WTW has extensive experience in modelling the ‘drivers’ or ‘catalysts’ of change in mortality for UK insurers, based on a detailed understanding of the range of such drivers and familiarity with the epidemiological evidence base for links between such drivers and different causes of mortality.
As well as the experience of the team and our customisable driver-based model, we have a range of other tools that we use to support our analysis, from a postcode mortality model to our medically-informed model of mortality (WTW PulseModel). We are able to help insurers to:
We have also worked with a range of insurers and reinsurers on the likely impacts of the pandemic, from forward displacement and long COVID to the impact of delayed diagnoses, healthcare pressures and a plausible endemic COVID outlook.
Our approach considers differences between short-term, medium-term and long-term changes in mortality based on the drivers which most likely affect mortality over those periods, making use of extrapolative methods in the short term, driver-based projections in the medium term and more judgment-driven approaches in the long term.
We believe that best-estimates should not stem from your choice of CMI model or its parameters, but that your choice of CMI model and parameters should reflect your best-estimate view.
For more information, please contact a member of our team or contact us at ict@wtwco.com.
Richard Marshall is a Director in WTW’s Insurance Consulting and Technology business and leads the development of mortality and demographic risk models for our UK business.
