New mortality research: Embracing uncertainty

By Tulio Walles, FCIA, FSA | Murray Wright, FCIA, FFA | Ian Zinck, FCIA, FSA

Insights – January 2025

In our previous article, we provided insights on the potential impact of new Canadian mortality improvement (“MI”) research published in April 2024 by the Canadian Institute of Actuaries (“CIA”).

The research provides a recommended MI scale and highlights that “our analysis shows that best-estimate results for long-term improvement rates are subject to great uncertainty…”. This finding is consistent with existing actuarial guidance, including the CIA’s Educational Note on the Selection of Mortality Assumptions for Pension Plan Actuarial Valuations, which observes that “assumptions in respect of future mortality improvement rates are subject to a high level of uncertainty and debate”.

In this follow up article in our three-part series on the CIA’s new mortality research, we will discuss the potential impacts of that uncertainty for pension and benefit plans. While this article does not intend to provide guidance on the selection of an appropriate MI assumption, our goal is to provide research-based insights to support the decision-making process. A future article will review the considerations of appropriate MI assumption setting.

A reasonable range

The research paper recommends a long-term MI rate of 1.3%. This means that in the long run mortality rates will fall by 1.3% a year, and consequently life expectancies will increase. To derive its long-term MI rate of 1.3%, the research paper applied models to historic Canadian population data and projected these models into the future. It is worth noting that this is a change from the approach used for existing MI scales in Canada which rely heavily on expert judgement.

The paper also states that “a reasonable range for the long-term MI assumption…is between 1.0% and 1.9%.” We commend the researchers for explicitly including ranges in the paper because it highlights that while 1.3% is the recommended long-term rate based on historic Canadian population data, a range of alternative assumptions is plausible. It also highlights how difficult it is to develop a mortality improvement assumption. Even with data from tens of millions of Canadians, using the past to predict the future is a difficult and uncertain exercise.

It is important to note that a mortality improvement scale is, after all, just an assumption. In practice, long-term mortality improvements could be significantly higher or lower than the expectation. This article discusses the proposed range for an appropriate best-estimate assumption, and how the choice of a mortality improvement scale that predicts higher life expectancy, will make it more likely that liabilities aren’t understated, and vice versa.

In our opinion, the proposed range of 1.0% to 1.9% appears reasonable and is broadly consistent with the range of assumptions used by UK pensions plans which are required to choose their own long-term MI rate when setting assumptions (see chart below):

The image is a donut chart titled "Long-term Mortality Improvement Rates of FTSE100 Companies." It illustrates four segments, each representing a percentage value: 1.00% in teal 1.25% in pink 1.50% in dark pink 1.75% in gray Each color corresponds to the respective improvement rate, providing a visual comparison of how frequently these rates are selected by FTSE100 companies.

Source: FTSE 100 annual reports.

For the purpose of this article, we propose to extend the bottom end of the range from 1.0% to 0.8% to include the long-term MI rate of the CPM-B scale, which is currently the most commonly used MI assumption by Canadian pension plans.

How would adopting a long-term MI rate of between 0.8% and 1.9% impact pension plans?

Approximate increase in life expectancy at age 651 for a 65-year-old
CPM-B MI-2017 MI-CAN-2024 [0.8%] MI-CAN-2024 [1.3%] MI-CAN-2024 [1.9%]
Long-term MI rate 0.80% 1.00% 0.80% 1.30% 1.90%
Males +0.3 years +0.8 years +1.1 years +1.6 years
Females +0.5 years +1.0 years +1.4 years +1.9 years

¹Based on 2014 baseline Mortality Table (CPM 2014).

In our prior article, we referred to the new MI scale as the APCI scale, named after the method used to examine the impact of age, period and birth cohort in the new scale. In this article, and in future articles, we will use a more intuitive name: MI-CAN-2024.

Adopting the new MI-CAN-2024 scale with a long-term rate of 0.8% (similar to the long-term rate of the CPM-B scale) still results in an increase in life expectancy at age 65, and therefore pension liabilities. This illustrates that while the long-term MI rate is a critical driver of the results, other differences in the scales (including specific model choices and more up to date data) also have a material impact.

Adopting the new MI scale with a long-term MI rate of 1.9% leads to significantly higher increases in life expectancies, even for a member currently aged 65.

Approximate increase in life expectancy at age 651 for a 45-year-old
CPM-B MI-2017 MI-CAN-2024 [0.8%] MI-CAN-2024 [1.3%] MI-CAN-2024 [1.9%]
Long-term MI rate 0.80% 1.00% 0.80% 1.30% 1.90%
Males +0.7 years +1.2 years +2.3 years +3.6 years
Females +0.9 years +1.4 years +2.4 years +3.6 years

¹Based on 2014 baseline Mortality Table (CPM 2014).

The main observation from the results above is that the impact of different long-term rates is much more significant for younger members. This is expected as the younger the person, the more exposed they will be to higher mortality improvements in their lifespan and the bigger the difference in their life expectancy.

If we translate this at the plan level, the choice of long-term MI rate will be more impactful for less mature plans, and on “normal cost” which considers only active members, compared to overall plan liabilities which includes retirees.

Liability impact

The chart below shows the potential impact of replacing a current MI scale (either CPM-B or MI-2017) with the new MI-CAN-2024 scale under different long-term improvement rates for a hypothetical pension plan¹. Impacts are shown with and without post-retirement indexation, and with different discount rates of 4.50% and 5.50%.

Image

It is interesting to note that the impact of adopting the new scale with a 0.8% long-term improvement rate (equivalent to that used by CPM-B) makes up more than half of the liability change. Again, this highlights that the choice of long-term improvement rate is only part of the story.

As we noted earlier, the actual impact of adopting the new scale will depend on the characteristics of the particular plan. For example, the impact for a plan with a higher proportion of active members will be greater than for a plan with a higher proportion of retired members. To get the actual impact of adopting the new scale for your plan(s), your actuary would need to “run the numbers.”

The ‘shape’ of mortality

It is also interesting to look behind the headline numbers and examine the ‘shape’ of mortality. Mortality improvements take time to have an impact because they compound as members age. The chart below shows the impact of different mortality assumptions on the expected benefit cashflows for a hypothetical pension plan¹.

Image

This highlights that even when the impact on life expectancy and pension liabilities is relatively small, the impact on the underlying plan cashflows could be more material. For this hypothetical plan, the expected cashflows over the next 100 years are $6.3bn under a CPM-B assumption, and $7.4bn under the MI-CAN-2024 assumption with a long-term improvement rate of 1.9%. If the 1.9% improvement rate was borne out in practice, the ‘extra’ billion dollars would need to be met by either future investment returns or additional cash contributions. This shows us that the assumption used is consequential and must be carefully considered for strategic decision-making and asset-liability modelling, including stress and scenario testing to help plans understand the potential impact.

We believe it is important to recognize that a level of uncertainty exists and in doing so we can better understand, monitor and manage longevity risk no matter which assumption is adopted in practice. In fact, the CIA has updated its Educational Note on the selection of mortality assumptions to state that any available scale (CPM-B, MI-2017 or the new MI-CAN-2024 scale) is appropriate to be used in the absence of credible information to the contrary. However, pension and benefit plans should consider the range of plausible assumptions and the corresponding impact on the plan if those scenarios were to be borne out. Some plans may wish to consider adopting different long-term MI rate assumptions for a range of specific uses, including for their financial statements. We will continue to explore these topics in future articles.

¹Salary-related pension plan, with approximately 45% active plan membership. Where applicable, post-retirement pension increases are 2.0% per year. Baseline mortality based on the CPM 2014 Combined mortality table.