At first sight, it is natural for an individual to consider the ages that other people are dying at, and use that to gauge the age they might achieve.  But that oversimplifies - lifespans not only have increased but will increase (almost certainly).

In the pensions arena, trustees and actuaries have been slow to acknowledge the extent of this effect, and valuation after valuation has needed to correct an underestimate from a previous valuation.  AMIPP has explained the updating between the 2003 and 2006 valuations.  Here we consider whether the advance between the 2006 and 2009 valuations is sufficient to catch up with the current statistical evidence.  That requires explaining the 2009 gobbledegook:

“Mortality: males – 115% of SAPS standard table “S1” Light for males with improvements up to the valuation date in line with the CMI medium cohort projections. Females – 100% of SAPS standard table “S1” Light for females with improvements up to the valuation date in line with CMI medium cohort projections. Allowance is then made for improvements in longevity from the valuation date onwards in line with CMI medium cohort projections with a minimum rate of improvement of 1.25% pa.”

There are no regulations about how these longevity calculations should be done, other than the requirement for the scheme actuary to be qualified.  Common practice is to adopt the approaches of the Continuous Mortality Investigation (CMI) which is sponsored by the Actuarial Profession.  The methods for a valuation compute in two steps - what death rates would be if they stabilized at today's figures, and then what improvements are expected.

The first step is easily parameterised - for a category of people we need about 50 numbers:  the proportion of them that survive to age X but not to age X+1, for a range of X up to age 120.  (It is a reasonable approximation for establishing costs that they all die by age 120.)  Producing a table for the first step just needs counting deaths by age that have happened (and perhaps applying some smoothing and weighting to the results).

The CMI provides such tables for males and females, rich and poor (as measured by the size of their pension).  The latest published tables are the "S1" series.  On average the more affluent live longer than the less affluent and these are referred to as "light" and "heavy" death rates.  Thus the short name for the table for more affluent male pensioners is S1PMA_L.  The raw data for such tables was gathered from the records of Self-Administered Pension Schemes, and the "S1" series is also known as the SAPS data.   

As you would expect, newer series make the older ones obsolete.  In our case the "S1" series, where the midpoint of the data gathering was around 2003, replaces the "00" series (used for our 2006 valuation) where the data was gathered around year 2000.   "Light": shows mortality rates for those pensioners with the largest pensions (in excess of £13,000 pa for males and £4,750 pa for females). These members tend to experience "lighter" mortality i.e. they live for longer.  "Heavy": shows mortality experience for those pensioners with the lowest pensions (below £1,500 pa for males and dependants and below £750 pa for females).  These pensioners tend to experience "heavier" mortality.

The average IBM UK pension is £15,900 p.a.  So for male scheme members table S1PMA_L, or something lighter,  is the obvious choice.

Remarkably, without giving any reason, the 2009 valuation assumes our scheme members will have death rates 15% heavier than S1PMA_L.  It is not statistically sound to suggest IBM UK experience as justification for overriding the SAPS experience since IBM experience will involve some 1000 instances and SAPS experience is of order 100 times greater than that.  If it were true that IBMers had 15% greater death rates than nationally evidenced rates, the appropriate response would be to establish the cause.  (There are those who believe chemical exposures in the IBM US East Fishkill plant damaged the health of those who worked there.)    The 15% loading corresponds typically to a year off the lifespan.  A year off every member's pension, if realised, would gift some £300M to IBM shareholders.

After figures for the death rates at some particular time have been adopted, the next stage is an assumption about how those rates will go lower with time.  The CMI provides tables which extrapolate historical experience, with a number for each age and future calendar year combination.  The number is an adjustment downward of the death rate for a person having that age in that calendar year.   The CMI has developed an updated and flexible way of using such tables, "... in response to the continuation of significant year-on-year increases in life expectancy, and to concerns over the continued widespread use, albeit with modifications, of the Interim Cohort Projections which have inevitably become increasingly out-of-date."

Nevertheless, our trust has used one of the cohort projections.  The name "cohort" projections arose in recognition that children who grew up during World War II tended to have noticeably longer lives.  (By reason of diet, attitude to life, or something).  The method enabled this cohort to be allowed for in calculations.  The tables came in "short", "medium" and "long" varieties, allowing assumptions about how far the effect would persist into the future.  The tables could be adjusted to add a little prudence by ensuring a minimum improvement year-on-year of (typically) one or two percent.  This explains the phrase "CMI medium cohort projections with a minimum rate of improvement of 1.25% pa.”

The computer program doing the valuation needs to aggregate the anticipated futures of the many members of our scheme.  Here is an illustration of what the component of the calculation is for one typical member, a recently retired 63 year old male. The three colours in the chart correspond to calculation on the IBM 2006 valuation basis, the IBM 2009 valuation basis, and the S1PMA_L with medium cohort improvement.  The red bars show how the 2006 basis failed to recognise the improvements in lifespans that are now acknowledged.  The green profile is flatter than the blue, showing that the 2009 basis mismatches the closer-to-raw data.  The green is higher at the lower ages because of the 15% loading and also higher at the highest ages because of the minimum 1.25% p.a. improvement assumption.

All extrapolations lose credibility if extrapolated too far.  The issue here is whether the experience of the several decades up to now should be used to predict for the next few decades, which cover the majority of the fund's outgoings.  The most recent published data is hard to follow as it consists of a presentation without the accompanying words, but it does not suggest a slowing down of the pace of lifespan improvement.

DOES ALL THIS MATTER?

- As far as having your anticipated pension delivered, it probably does not matter.  The valuations get adjusted every three years and underestimates of cost will eventually get corrected, unless the company fails completely.   Underestimates of this cost do not reduce the company's costs, they merely move them later.

- It does matter to a person starting on a pension that is calculated on a "cost-neutral" basis.  The idea of "cost-neutral" is sound - that someone retiring early should have a lesser pension per month because the pension is expected to be paid over a longer period.  However the extent of the earliness will be known and fixed whereas the lifespan is a prediction;  if the lifespan is underestimated the reduction needed to balance overall will be overestimated.  This over-reduction is gain for the shareholders - no other members' pensions are increased because of it.

- It matters to everyone in their financial planning, in assessing the risk that they will live "too long" for their inflation-ravaged pensions to support them.   Failure by the trustees to adopt the best predictions, and inform the members in understandable terms, will lower the quality of this financial planning.  In this planning, retirees will need to recognise that inflation experienced by pensioners exceeds the experience of the broader community.