I’ve seen a number of exercises recently which have looked to model potential scheme mortality costs in relation to the quality of health of the scheme membership. The rationale is that certain employers may have a workforce which is likely to be in poorer health and therefore have a lower life expectancy than might be assumed as ‘standard’. This can then be used as a basis to adjust the mortality assumptions and therefore reduce liabilities, deficit and ultimately costs.
Whilst the results of these exercises are often illuminating I would seek to add a note of caution to the process and those considering such a review need to consider the positives and negatives.
Clearly this is only one method of assessing scheme specific longevity. If the process is adopted using a sampling approach then clearly there could be something of a sampling risk with results being skewed purely by those actually selected for the sample. The greater the sample size the greater the accuracy but the higher the likely cost. Even with a comprehensive sample request it is unlikely that you would achieve a full response so there will be some sampling risk involved.
An alternative method would be to employ a post code assessment basis and for larger schemes this may provide a more cost effective and accurate method of estimation.
There also needs to be some weighting applied to the results. Scheme liabilities are not evenly spread across a membership so positive or negative movements are likely to be more relevant the greater the liability of any individual member. It is likely that, on average, larger liabilities are likely to be held by individuals with better than average life expectancy.
In previous blogs we have also provided details on our research in to the “small scheme effect” which shows that the smaller the scheme the greater the random impact of longevity. For small schemes a member dying a long time before their expected date or living a long time after it is likely to be much more significant than the impaired life impact across the scheme. The larger the scheme the closer it is likely to mirror standard longevity tables – it’s selecting the table and estimating the likely improvement which is likely to be relevant. In this case again post code assessments are likely to be more relevant.
There is also a concern that actually securing benefits using impaired life annuities may actually have negative rather than positive consequences particularly the greater the timescale which elapses from using the impaired life annuity and ultimately buying out the remaining benefits. A considerable timescale would mean that the scheme is losing out on a potential windfall from an early death which is likely to be much greater than that derived from the discounted purchase price of the impaired life annuity.
If impaired life annuities are used at a point close to ultimate buyout there must be a question if the ultimate buyout insurer will not make some adjustment to the buyout price to reflect the fact that only those lives who are likely to live longer remain in the scheme, thereby out-weighing any gain achieved via the use of impaired life annuities.
There may well be a place for the use of exercises such as these but potentially only as part of a wider ranging liability management review which considers other potential solutions such as enhanced transfer exercises, pension swaps or early retirement exercises to identify the relative value of an impaired life review. All this just reflects one simple fact – longevity assumptions are just assumptions and highly unlikely to be borne out in reality on average and certainly not at an individual level.