Continuous Mortality Investigation (“CMI”) Working Paper 197 (“WP197”) consults upon a set of proposed methodological changes to the CMI_2024 mortality projections model. Club Vita have responded to the consultation in our capacity as a provider of mortality and longevity analytics services to both pension schemes and insurers. Our full response can be downloaded using this link.
Key themes of our response
1 Does the new “Fitted Overlay” approach meet the desirable characteristics of a projection model?
We believe the desirable characteristics for a mortality projection model in the aftermath of the COVID pandemic are:
- A sensible fit to observed mortality rates pre COVID
- A sensible fit to observed mortality rates post COVID
- A plausible launching off point for best estimate improvements
- Robustness over time as new data emerges
We have used this framework in forming our response. Our focus here has been the new proposed Fitted Overlay since that methodological change relates most directly to the ongoing implications of the COVID pandemic.
We conclude that the proposed Option (“D”) broadly meets these desirable characteristics but have a concern regarding how the model will respond if 2025 is “as expected”.
2 What are the practical challenges when using the proposed model to generate mortality rates?
The two key uses of a mortality projections model are to generate prospective mortality rates (for use in cashflow projections and valuations) and historical mortality rates (for use in experience studies). Both require the combination of a selected projection with an appropriate base table.
CMI_2024 as proposed implies “spiky” year on year mortality rates over the course of the pandemic, in contrast to previous versions which excluded 2020 and 2021 data from the fitted trend. Particular care is needed when combining the projection with a base table calibrated against data including 2020 and beyond. Further, the shape of the Fitted Overlay is based on an estimate of excess mortality in the overall England & Wales population. Other populations (both in the UK and beyond) may have exhibited different levels or year-on-year pattern of excess mortality to the England & Wales data. The overlay fitted to England & Wales data is unlikely to match the excess mortality rates seen in a particular book or required for a particular use case.
We conclude that many users will need to adapt the proposed core projection to meet these challenges and that the model would benefit from some simple changes to make these adaptations easy for users – specifically easier access to the projection excluding the overlay and more flexibility in defining the overlay to be used for populations other than the England & Wales population built into the model.
3 Wider changes to CMI Model: Is the introduction of three period terms necessary?
The most material of these wider changes to the model is the extension of the existing age-period-cohort improvement (“APCI”) model to have three period terms. This materially increases the complexity of the model for users and stakeholders – which may be to the detriment of understanding and of stability.
In our response we look specifically at:
- The rationale for introducing multiple period terms in the context of the emerging data.
- Whether the resulting updated projections at younger ages are plausible, explicable, and useful.
- Whether alternative interpretations of the emerging data could lead to a different, simpler, model
As part of this discussion, we suggest two alternative approaches which could address the challenge of capturing the current age pattern of improvements, but which are materially less invasive.