Survey shows widespread use of the latest CMI Model

Looking through data

This blog provides a high-level summary of the CMI’s latest survey on the use of the CMI Mortality Projections Model by insurers and reinsurers, focussing on how users are reacting to the pandemic. There is more detail in Working Paper 163 (restricted to Authorised Users).

The CMI publishes the CMI Mortality Projections Model to enable users to make projections of future mortality. We provide “Core” assumptions for all bar one of the Model’s parameters, but encourage users to consider modifying parameters. This can be done to tailor the Model’s output to a specific population, or to reflect a user’s own view of the outlook for mortality improvements.

It is helpful to know how the Model is being used – users can see how their peers are using the Model, and the CMI itself can focus research and future developments on the parts of the Model that users are most interested in. To achieve this, we have recently completed our second annual survey of how insurers and reinsurers use the Model.

We received 20 responses to the latest survey, covering 40 books of business – a similar level of response to the 2021 survey. We would like to thank all those who responded to the survey. Given the good responses to date, we intend to repeat the survey in 2023.

Impact of the pandemic

In response to the COVID-19 pandemic, and following consultation with users, we adapted the Model to allow users to reduce the weight placed on data for certain years. In the standard version of the Model, we place no weight on data for 2020 or 2021, but users can choose to put some weight on these years if they wish.

We are pleased to see from the survey that the latest versions of the Model are being used. At 31 December 2021, the most widely-used Model version was CMI_2020, the latest version available at that time; and many respondents expect to use the newer version, CMI_2021, at 31 December 2022.

The survey shows that users are comfortable with the CMI’s approach of placing no weight on data for 2020 or 2021 – no respondents place any weight on data for 2020, and only one expects to place any weight on data for 2021.

To better understand how the pandemic has affected views of future mortality improvements, we asked respondents for their best-estimate views of the impact of the pandemic on cohort life expectancy at age 65 at the end of 2022. Over half of respondents indicated that their best-estimate view of the pandemic was no (or close to no) change in life expectancy, with the remaining respondents indicating falls of either about 0.5% or about 1% in life expectancy.

Other parameters

Although one respondent modifies six types of Model parameter, most users of the Model make relatively few changes to the parameters and over 40% of the books of business use the Core version of the Model.

The most commonly varied parameter is the “initial addition to mortality improvements”, which is modified in over 40% of books. The Model is calibrated to the general population for England & Wales, but the initial addition parameter allows it to be tailored to reflect socio-economic characteristics of the specific population that the Model is being used for. 

The next most common adjustments are to the “period smoothing parameter”, which controls the responsiveness of the Model to new data, and the age profile of the long-term rate of mortality improvements.

Further information

We encourage Authorised Users to look at Working Paper 163. This has detailed analysis of the survey responses including: liability-weighted results; comparisons of results for insurers and reinsurers; comparisons of mortality protection and longevity books; and analysis of life expectancies, including 1-in-200 stressed mortality improvements for Solvency.


Read more blogs by the CMI here.

Topics