Projecting Life Expectancy
Barnett Waddingham LLP discuss the most recent developments in projecting life expectancy.
The longer people live the longer pensions need to be paid, and so mortality risk has always been a key concern for all pension schemes. In the current period of sustained low interest rates, future payments have a higher value, making the cost impact of living longer even more pronounced.
The second report of the Pensions Commission indicated that increasing longevity will require increases in State Pension Age in due course (and, in response, the Actuarial Profession suggested that Lord Turner's proposals perhaps did not go far enough in this respect).
This note looks at recent developments in analysing and projecting mortality trends; in particular recognising the high level of uncertainty about projecting future increases in life expectancy. To understand the risks involved and to develop appropriate mitigation strategies, trustees may want to consider the merits of commissioning a scheme-specific mortality analysis.
The mortality tables in most common use currently are the "92" Series. These relate to a study of insured pensioner mortality during the period 1991 to 1994 by the Continuous Mortality Investigation (CMI), a research project organised by the Actuarial Profession. During the course of 2005 the CMI published its proposals for a new set of standard mortality tables - the "00" Series. These tables are based on the experience of 1999 to 2002. We expect that final tables will be officially published around Spring 2006.
The CMI are also supporting a separate series of studies looking at the experience of non-insured occupational pension schemes. These studies already relate to a larger number of pensioners than the insured pensioner surveys, but only have a short history to date. So far, these studies appear to show a very wide variation in experience. For example, those with larger pensions appear to be experiencing lower mortality than the insured population, whereas those with smaller pensions appear to be experiencing higher mortality.
Base mortality tables are, however, just a snapshot of the experience at a particular point in time. Of equal importance is the question of what will happen to life expectancy in the future. For some time the mortality studies completed by the CMI have therefore included projected improvements in mortality for use with the base tables.
The original projection accompanying the "92" Series tables took account of the then apparent long-term trends. However, rapid improvements in longevity for the generation born between 1920 and 1945 subsequently became apparent (known as the "cohort" effect) and so interim "cohort projections" were also published. The CMI recognised the uncertainty around these projections by publishing three projections, which varied according to how long the cohort effect was projected to continue in the future.
The emergence of the cohort effect is an example of how projections made in the past have often turned out, with hindsight, to be inadequate. In the light of this the CMI wishes to ensure that, in future, users are fully aware of and take account of the inherent uncertainty involved in longevity projections.
The graph above shows a projection of life expectancy improvements for 60 year old men using one of the latest methods. The mid point of the new projection is shown by the dark blue line, whilst a reasonable range of future life expectancies is illustrated by the light blue area. The original projection accompanying the "92" Series tables (the red line) is also shown as a comparator.
For example, the new projection suggests that the projected life expectancy for a man reaching age 60 in 2006 might be between 85.75 and 92.75 years with a central estimate of 88.75. This is between 1.25 and 8.25 years more than the 84.5 years previously projected.
The graph brings out two important features of projecting life expectancy:
- The uncertainty at a point in time (the range of possible results).
- The increase in that uncertainty over time (the widening of the range of results).
It is important to note that the life expectancy using the original "92" Series projection is lower than the mid point of the new projection. This is true even for life expectancy in the years immediately following 1992 because of lower projected mortality during the later stages of retirement.
The following graph shows the range of annuity values, using the above underlying longevity projections, that might have applied in the calendar years following 1992. Annuity values allow for the effect of interest earned during retirement (at the rate of 4.5% per annum in the example shown) and therefore the values are lower than the simple life expectancy. The range of results is also narrower (and therefore the scale of the graph covers a much shorter interval) as the effect of interest means that less weight is placed on the later stages of retirement where there is greater uncertainty.
This graph shows that if an actuary updated the assumptions used for valuing pension schemes or pricing annuities based on the new projections, he might recommend increasing the 2006 actuary value at age 60 by over 6% based on the central projection. The impact of using the new projections is even greater for deferred pensioners who will reach retirement age in later years.
The potential impact on the value of liabilities is illustrated in the table below. Figures calculated using the new projections are compared to both the original "92" Series projections and the CMI's interim medium cohort projections.
Scheme-Specific Mortality Studies
A mortality projection will be affected by a number of decisions, including the choice of model, the dataset to feed into the model (e.g. the scheme's own data or a published dataset) and other inputs required by the model.
For any pensioner population, an understanding of the recent mortality experience is necessary in deciding the underlying base mortality table to be used. For larger pension schemes (say more than 1,000 pensioners) there should be sufficient data available to inform this decision directly but, for smaller schemes, a greater degree of judgement and understanding of those factors distinguishing the scheme from the general pensioner population will be necessary (e.g. the particular mix of socio-economic groups).
Modelling projected improvements is only possible for the largest pension schemes with data covering periods longer than 20 years. For others, mortality projections can usually be tailored by making suitable adjustments to projections based on published datasets - either population data or the insured pension scheme studies.
For smaller schemes, or for schemes where the pensioner profile is dominated by a few large pension amounts, there may also need to be an allowance for random variation in mortality experience. For example, the mortality of a population of 10,000 pensioners could be expected to follow the projections closely, whereas the experience of just a few pensioners is more unpredictable (as any one pensioner could either live for only a short while longer or keep going to the age of 120!). Therefore, if a scheme's pension profile is dominated by a few individuals with large pensions, then the costs for the scheme will depend on the actual life expectancy of those few individuals, irrespective of whether the average experience for all its pensioners follows the projections.
For many years insurers have developed an understanding of their pensioner portfolios and carried out the above analyses so that they could calibrate the projections, as well as the base tables, to their portfolios.
However, pension schemes have generally not been as sophisticated. While many pension schemes have conducted occasional analyses of their own past mortality experience to calibrate base tables, few have carried out analyses to allow them to calibrate the projections. Instead, they have generally used the standard CMI projections for future improvements in longevity. In the future we expect that many schemes will choose to go further and conduct scheme-specific projections of likely future mortality trends. This will enable trustees to build up a better picture of the mortality risks facing their own scheme.
Why is Understanding Scheme-Specific Mortality Risk Important for a Pension Scheme?
Scheme-specific analysis will provide greater understanding of the uncertainty about future improvements which could suggest that the scheme's experience is likely to fall in a narrower (or wider) range than that from the CMI studies. It may also indicate whether longevity is likely to be higher or lower overall than the average (for example, a scheme with mostly wealthy pensioners might expect to experience higher than average life expectancy).
Modelling scheme-specific mortality experience helps identify the pensioners contributing most to a scheme's mortality risks. This will allow trustees to seek appropriate mitigation strategies such as selectively buying annuities for the few pensioners who have much larger pensions than the rest of the scheme's members.
Trustees could allow for the scheme-specific analysis in setting assumptions for Scheme-Specific Funding valuations, or accounting for pension costs under FRS17 or IAS19 (for which it is likely that the accounting profession will soon require explicit disclosure of the longevity assumptions made). Representations could even be made to the Board of the Pension Protection Fund for PPF levy valuation purposes.
Scheme-specific mortality analyses could also prove of benefit in negotiating bulk annuity purchase prices with insurers, and may be useful in the setting of terms for bulk transfers into or out of schemes (for example on corporate transactions or scheme mergers).
It may also help trustees in their negotiations with employers over Statements of Funding Principles and future funding rates. Furthermore, an understanding of mortality risk (as well as investment risk and other areas of uncertainty) would help the trustees to appreciate the reliance of a scheme on the sponsoring employer's covenant, which may be relevant to issues such as deficit spreading periods.
How can Barnett Waddingham help?
Barnett Waddingham LLP has significant experience in undertaking mortality analyses and projections (for the CMI and others). We have a team of twelve consultants (including four of our 38 qualified actuaries) with particular expertise in this area.
This means we are well placed to provide the technical services required for a scheme-specific analysis, and offer a clear and understandable presentation of the results.
Please speak to your usual Barnett Waddingham contact for further information about our mortality analysis services.
Barnett Waddingham LLP, January 2006.