[Skip to content]

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

The ‘00’ series

he ‘00’ series mortality tables, based on CMI experience in 19992002, are the latest in a line of mortality tables produced by the CMI for UK actuaries, stretching back to the A192429 tables. Considerable work is involved in producing new tables, so they are only produced when the underlying experience is judged to have altered significantly from that underlying previous tables.
After the ‘92’ series tables were published, based on experience in 19911994, it soon became apparent that mortality was improving at a rapid rate, by historical standards. That, in itself, might not have necessitated new tables after such a short interval, but the improvements were not consistent by age. For example, evidence was emerging of a ‘cohort effect’ in both population and CMI data. This effect appeared to show that a group of lives (born somewhere around the late 1920s/early 1930s) was experiencing even more rapid improvements in mortality than the rest of the age groups (see, for example, ‘Longevity in the 21st Century’ by Willets et al). Given these people had recently reached (or would soon reach) retirement age, this had considerable significance for the profession, if actuaries continued to use the ‘92’ series tables.
Even before all of the 19992002 data had been submitted to the CMI, the Mortality Committee was preparing for the new tables. Figure 1 shows the improvements in assured lives mortality between successive quadrennia for selected age groups. This shows that mortality has improved faster for those aged 51 to 75 and most rapidly for those aged 61 to 65.

More tables
The ‘00’ series tables are not just an update of the previous tables, though. Before embarking on the graduations themselves, the Mortality Committee studied the data underlying each investigation to assess if any of these no longer warranted graduation, or if any new graduations should be produced, and consulted with the profession on its proposals.
The outcome is that a number of tables have been published for the first time, including:
– Separate smoker and non-smoker tables for assured lives and temporary assurances (as well as the traditional aggregate tables).
– Experience under personal pension contracts. Such contracts have only existed since 1988, so there has not been sufficient experience to produce tables previously.
As a result the ‘00’ series contains 40 tables, each of which represents a different segment of the mortality data collected by the CMI.

What does ‘adoption’ mean?
The UK actuarial profession has ‘adopted’ the ‘00’ series mortality tables. There was a debate within the profession over the meaning of ‘adoption’, especially following the formation of the Board for Actuarial Standards (BAS). It was agreed that it was still appropriate in the new environment for the profession to adopt the tables as a part of its remit to provide its members with useful tools, and that this does not stray into the responsibility of the BAS for setting technical actuarial standards.
In the past, the tables were commonly referred to as ‘standard tables’. This is no longer appropriate. Indeed, the CMI has never seen itself as responsible for setting standards for the profession regarding base mortality or future projections and all tables have carried warnings to the effect that:
‘It is the responsibility of any actuary or other person using a published table to ensure that it is appropriate for the particular purpose to which it is put.’
By approving the publication of the ‘00’ series tables in the name of the UK actuarial profession, FIMC confirmed the CMI’s view that the tables are well-constructed and suitable for use by actuaries. It is certainly not the case that actuaries should assume the tables can be used blindly nor that they are mandated by the profession in any sense. Clearly, though, the CMI hopes that the tables will be a welcome addition to actuaries’ toolkits, whether they work in the life or pensions sector.

Mortality projections
Previous series of tables for pensioners and annuitants, from the a(90) tables up to and including the ‘92’ series, have incorporated a projection, giving a single view of future mortality. These projections were principally based on analyses of past trends in the various CMI investigations and in the wider population.
Perhaps the most notable feature of the ‘00’ series tables is the absence of projections of future mortality. It is worth explaining the background to this.
The projections incorporated in the ‘92’ series tables were quickly found to understate the level of mortality improvements that were actually occurring in CMI experience. Indeed, this had tended to happen with previous projections too. In addition, evidence was emerging of the ‘cohort effect’ (see above). The CMI responded by publishing the ‘interim cohort projections’ late in 2002. These projections were ad hoc adjustments to the original ‘92’ series projections for lives at the important ages that reflected actual improvements in mortality to 1999. A choice of three projection bases was offered short, medium, and long in which the cohort effect was assumed to persist for different periods. The three bases were not intended to carry any probabilistic interpretation and there was no indication of which one an actuary should use if any!
The use of the term ‘interim’ in the name reflected the ad hoc nature of these projections and that the CMI would undertake further work in this area.
Since publishing the interim projections, the CMI has been undertaking research into possible approaches to mortality projections, most recently evaluating two methodologies for mortality projections: P-spline and Lee-Carter. The CMI is hoping to assist actuaries by explaining the features and benefits of each. Actuaries will then need to consider the appropriateness of these and other methods for use in a particular situation.
The absence of projections within the new tables has been widely communicated to the profession over the past year. It was also publicised when the profession issued a press release (‘Actuarial profession highlights need to allow for uncertainty in future mortality as it unveils sharp improvement in pensioner mortality figures’). This generated considerable interest in the press and on the radio. An example is given below.
‘In a world of shifting sands, actuaries lived in an oasis of certainty. But no longer. Life expectancy has risen so fast in recent years that these masters of mathematical projection, who traditionally make auditors look like careless dilettantes, have acknowledged that mortality rates are too slippery to predict. A symbolic throwing up of hands may kill the traditional annuity, at least for anyone under 75. But the admission of ignorance is also refreshing since no one can really know what the future holds, and sound planning will provide for unexpected, as well as expected, events.’
The Times business section
30 September 2005

How does an actuary produce mortality projections?
The CMI has made software freely available that allows actuaries to experiment with the P-spline and Lee-Carter methods. However, this is perhaps mainly for those with a keen interest in the subject, and some actuaries will probably want to use a simpler approach.
Some results of using P-spline projections have been published in Working Paper 20, and presented at CILA and Current Issues in Pensions seminars. Unfortunately work on the Lee-Carter method has proved troublesome and is lagging behind, but we hope to provide comparable results soon.
Actuaries may choose to use these results or those from other methods to adjust existing approaches.
Unfortunately, it is not just the choice of method that actuaries have to consider. For example, there is also the question of what data source to use. There are many similarities in the patterns of mortality improvements in CMI data and population data, such as the main cohort effect. But there are differences too for example, population data has shown worsening mortality at younger ages during the 1990s while CMI data continued to show improvements, albeit at lower rates.

Where does this leave actuaries?
There is an understandable desire for guidance in a complex area, particularly perhaps from actuaries in small firms. In the past the CMI has met this need (of actuaries generally) through the ‘92’ series (and prior) projections. While these projections made life easy for actuaries, one could question with the benefit of hindsight whether the profession was best serving the public interest by conveying a single view of the future and giving an impression of certainty around future mortality, particularly as these projections consistently understated the improvements that subsequently occurred.
Indeed, recent trends in CMI experience suggest that the medium-cohort projection basis may now be understating likely future improvements. Reliance on a single view of the future is therefore misleading and potentially dangerous.
Within the life insurance sector there is an increased emphasis on the use of stochastic modelling and actuaries have to illustrate the impact of uncertainty in their individual capital assessments and realistic balance sheets.
In the pensions sector, my impression is that many actuaries have continued to use a single basis and have not communicated the inherent uncertainty to employers and trustees.
Actuaries now face a complex choice of models and datasets and need to consider what is appropriate for their particular portfolio. Such decisions may be made easier by an understanding of the factors that have contributed to past improvements in mortality and a view on how these may change in the future.

What happens next?
The CMI is committed to completing its current research into P-spline and Lee-Carter projection methods. We will continue to communicate the results of this via the website, publications, presentations at seminars, etc. The precise nature of these results is not yet known. For example, different methods, including the two currently under investigation, can give very different projections of future mortality and this is one way of conveying ‘model risk’. It is possible that any steps towards more realism will mean moving away from the simplicity of the cohort projections, which themselves were meant as a step away from the single projections actuaries have previously been used to.
The investigation into the mortality of self-administered pension schemes has now been brought into the CMI and will yield valuable insights into the experience of such schemes. These may prove more relevant to pensions actuaries than the existing investigations into insured experience and may, in time, yield another valuable database to enhance our knowledge of mortality trends.

06_09_06.pdf