[Skip to content]

Sign up for our daily newsletter
The Actuary The magazine of the Institute & Faculty of Actuaries

Longevity: Goldilocks and the three cohorts

Pensions actuaries have commonly relied on one of the Continuous Mortality Investigation’s (CMI’s) three cohort projections – short, medium or long – since they were first published as ‘interim adjustments’ in 2002. Many actuaries have tried out all three in successive actuarial valuations, in the search for that ‘just right’ feeling.

The three interim cohort projections were derived from an analysis of insured pension scheme data. Until recently, it had not been known whether the same patterns are evident for pensioners of occupational pension schemes. Club Vita has now collected and analysed long-term data from a diverse collection of workplace schemes from the private and public sector. This has enabled a comparison between the rates of mortality improvement shown in the insured schemes, with those of occupational schemes.

Just right?
The CMI long cohort projections were derived from insured scheme data for men up to 1999. Hot colours show ages and years where the most rapid improvements in longevity are assumed to occur. Club Vita’s data shows the corresponding smoothed experience for a broad cross section of occupational schemes.

There are clear differences between the two pictures: longevity improvements in occupational schemes have been different to those assumed under the long cohort assumption. More specifically, the colours on the right hand chart are typically cooler. This shows that longevity improvements observed in occupational schemes have generally been less rapid than those projected under the long cohort assumptions. While improvements have been significant at around two years’ increase in life expectancy in the last decade, they have been less rapid than the long cohort assumes.

The long cohort projections assume that the rapid improvements observed for the cohort of lives born around 1926 would continue through the early part of this century. This can be seen by the dashed diagonal trend line, stretching up from age 68 in 1994 through to age 81 in 2007. In comparison, in the occupational scheme data for men, there appears to be a clear diagonal trend (highlighted on the right hand chart), but it is centred on younger ages. The most rapid improvements are being experienced by those now in their early 70s.This difference affects schemes in different ways, with the financial impact depending on the maturity profile of each scheme.

It is also worth noting the green area at younger ages in the occupational schemes chart. This suggests that pensioners now aged 50 to 60 are seeing little, if any, improvement in life expectancy. Is this the start of a trend? Does this provide evidence that supports reduced allowance for improvements in the future? We would caution against drawing any firm conclusion at this stage. After all, there are fewer deaths at younger ages, so there is a smaller weight of evidence underlying this element of the chart. However, it does remain a feature of interest, worthy of further research, and requiring regular monitoring.

What might explain the different patterns? Is the mortality of different socio-economic groups in society changing at different rates? This is a subject that the Club Vita team is currently researching with the Oxford Institute of Ageing. The results will be presented at the multi-disciplinary longevity conference being organised by the Actuarial Profession in Edinburgh in October.

Better handling of longevity uncertainty
Of course, the actual increases in lifespan are likely to vary greatly between projections. We appear to be going through a period of great change. Much is understood about the reasons for current differences in death rates across society, with differences in lifestyle and other factors having a profound effect. However, there is no consensus in the medical or biological worlds on the impact of advances in medical technology, lifestyle changes and healthcare or, in these difficult economic times, our ability to pay for them all. A quick look back in history also reminds us of our fallibility in projecting the uncertain. For example, the research of Oeppen and Vaupel (Broken Limits to Life Expectancy, Science Vol 296, 2002) shows that demographers have repeatedly under-estimated limits on lifespan.

The long-term effects of longevity improvement should be regarded as a risk management issue for trustees. It should be handled as part of the scheme’s governance process, rather than just left to the scheme actuary to deal with every third year.

Get yourself a Plan B
Given all the uncertainty, it is dangerous to place too much faith in any single projection of the future, without some contingency planning. Trustees should appreciate the range of possible outcomes, and then stress-test the funding of their own scheme. There is a case for swift action too, particularly in those schemes where future accrual is significant. Today’s pension laws prevent benefits that have already been earned being reduced. However, schemes can be made more durable by building in a power to alter pension amounts earned in future (or to delay their payment) if longevity changes.

Each scheme’s approach to risk management must be tailored to its own circumstances; the design of the benefits, its maturity, its funding and the control levers that the employer can pull, are all reasons for differences. However, the common ingredient in any form of longevity risk management is regular monitoring of the emerging evidence, using relevant and up to date longevity data.

And the moral of the tale?
Had Goldilocks woken up to the true risk she was taking, she would have realised that the ‘just right’ feeling would lead her to a false sense of security.


Douglas Anderson is the founder of Club Vita. Andrew Gaches is the longevity expert for Club Vita. Club Vita is a recently established longevity comparison club which offers statistical analysis allowing the pensions community to benefit from pooled pension fund data and scheme specific analysis.


Figure 1: CMI Long Cohort projections vs. Club Vita’s analysis