Making sense of an uncertain future
New mortality research conducted for the UK Actuarial Profession should help guide future assumptions, argues Dave Grimshaw
Mortality is at the heart of many actuarial calculations relating to defined benefit pension schemes. The expected cost of providing a pension depends upon how long an individual is expected to live (and other factors such as expected investment returns). Of course, the actual cost depends on how long he or she does actually live, as well as the actual investment returns, and is not known with certainty until the pension ceases.
In recent years there has been an unprecedented debate on actuaries’ assumptions about longevity - in part arising from a period of unprecedented improvements in mortality and also from changed economic scenarios, meaning that mortality has become a relatively more important assumption.
It is helpful to separate mortality assumptions - the assumptions about the current mortality experience and the assumptions about future changes in mortality. Making this separation clear can help trustees and employers better understand the nature of the assumptions that their actuaries are using.
It is impossible to tell exactly what the future will hold, and in this regard mortality is no different from other assumptions. Mortality rates have undergone significant changes in the past, but this does not necessarily mean that similar changes will occur in the future. Much of the improvement in longevity for males in the UK has been the result of a fall in death due to coronary heart disease, which have arisen at least in part from the reduction in smoking. Such gains may not be repeated. In the male population as a whole, cancer is likely to be the main cause of death in the coming years and - even if deaths from heart disease are eliminated entirely - it is trends in mortality from cancer and other causes that will drive future changes in mortality.
Given the inherent uncertainty surrounding future mortality, it is inevitable that assumptions will engender debate. For many years the mortality tables produced by the CMI on behalf of the Actuarial Profession contained built-in assumptions future mortality improvements. In 2002 the CMI published the Interim Cohort Projections which recognised the particularly fast rates of improvement in mortality among the group of people born in the late 1920s and early 1930s, and offered three variations on how long these faster rates of improvement might persist.
More recently, the CMI has conducted research into extrapolative methods for the generation of future assumptions in mortality. These methods have not produced assumptions that have achieved any degree of consensus among actuaries. Instead a variety of assumptions has emerged, including the use of a minimum rate of improvement on a cohort projection, in order to strengthen the assumption. This particular approach has been used by many actuaries, albeit with considerable variation over the minimum being applied.
In an attempt to standardise the method by which such adjustments are being applied by actuaries, the CMI last year published its ‘library of mortality projections’. This allows actuaries (especially those in smaller firms) easier access to a range of projections produced by the CMI and others. It was not intended to provide guidance to actuaries on what assumptions they should be making about future mortality, and the debate in this area looks set to continue.
Assumptions on current mortality have attracted less attention than assumptions on future changes, but can also be subject to considerable uncertainty.
Actuaries have for many years differentiated mortality rates by age and gender. In the UK, mortality generally increases as people get older and female mortality is generally accepted to be significantly lower than for males. Even allowing for differences by age and gender, mortality rates are not the same for everybody but are affected by a wide range of factors including health, wealth and lifestyle.
The mortality of any specific group of lives is unlikely to equal that of the general population. For example, population data will include those in ill-health who have never entered the work force but pensioners of final salary pension schemes will have been in sufficiently good health to be employed, at least for part of their working lives. Pension schemes and insurance portfolios can therefore be expected to exhibit lighter mortality on average than the general population.
For larger schemes, there should be sufficient data to measure the past mortality experience over recent years. This may provide a useful guide to the assumption on current mortality but care is required in a number of areas. Most deaths will occur at the older ages but this group may not be representative of the pension scheme membership as a whole, if the sponsoring employer has undertaken a substantive change in the nature of its operations. Often, past mortality experience will be compared to a published table of mortality rates. This allows actuaries to use the experience of the much wider group of lives underlying the table, and to adjust this to better fit the particular group they are considering.
For smaller schemes, recent mortality experience will be subject to considerable random fluctuation, so is of little value in assessing the likely current experience. For such schemes, reliance on published tables is perhaps greater still, although actuaries will exercise judgement to select the most appropriate table and to adjust it appropriately.
Until recently the choice of table was limited to one derived from population data or one published by the CMI based on its analyses of mortality data collected from insurance companies.
In February, the CMI published in draft form its first tables based on the mortality experience of UK self-administered pension schemes (SAPS). The SAPS investigation collects data from scheme actuaries of self-administered pension schemes with at least 500 current pensioners. The data underlying the new tables contains nearly 400,000 deaths aggregated from around 350 schemes. The data spans the years 2000-2006 and is centred on 2003, hence the tables have provisionally been entitled the ‘SAPS 03’ mortality tables.
Separate tables have been produced for male and female pensioners, each based on numbers of pensioners and amounts of pension. Tables have also been produced for those retiring in normal health, those retiring in ill-health, and for widows who continue to receive a pension after the death of their husbands. These groups exhibit quite different mortality rates.
The tables should help actuaries in setting assumptions on current mortality, but are by no means the end of the story since they necessarily represent the mortality experience of an average pension scheme, based on the data available.
Previous CMI research has demonstrated a marked correlation between the amount of pension and the mortality experience. This has been recognised in the inclusion of separate ‘light’ and ‘heavy’ tables, where the draft light tables are based on male pensioners receiving over £8,500 (€11,000) annually and female pensioners receiving over £3,000 annually. These pensioners average lower mortality. While this provides a valuable indicator of the range of experience, the amount of pension is not necessarily a good predictor of mortality experience. Further research will look at the SAPS data segregated for different industries and one recent initiative has narrowed this categorisation still further in looking to produce a study of mortality experience specific to retailers.
Work is also planned to look at the improvements in mortality rates in pension schemes, but this may not yield any clear messages for assumptions on future changes in mortality until the data covers a longer period of years.
Whatever assumptions are used - for current mortality or for future changes - it is inevitable that the actual mortality, when it becomes known at a point in the future, will be different. It is important that actuaries help trustees and employers to recognise the extent of the uncertainties in mortality rates.
The draft SAPS 03 mortality tables can be found on the Actuarial Profession’s website: www.actuaries.org.uk