How many years of retirement will a new scheme member enjoy after they reach their 65th birthday? What will happen to their life expectancy if they swap mid-career the construction site for a comfortable office chair?

The answers to such questions will always be unpredictable but as defined benefit (DB) schemes become increasingly unaffordable, accurately predicting mortality is becoming ever more important. Mortality, or longevity as it is often termed, attempts to predict how long us humans will live for – and therefore how long pension schemes will have to pay them after retirement.

Life expectancy previously improved slowly and predictably. However, it is now increasing by as much as two years every decade. How long can we assume that such progress will be maintained? “It is difficult to quantify,” says Mike Butterfield senior consultant, Watson Wyatt. “But trustees need to understand the risks about how sensitive valuations are to changes in life expectancy.”

Live and let live

Mortality models which support longevity predictions are necessary for trustees to complete valuations and determine whether risk-reduction exercises such as the newly-created longevity swaps (see box) are worthwhile.

The most widely used data on which these predictions are based comes from The Actuorial Profession’s Continuous Mortality Investigation (CMI), a research body that produces freely available mortality tables. These tables, originally based on life assurance data in UK and Ireland, were only intended for use on an interim basis but have become the pensions industry standard.

The CMI tables that trustees will be most familiar with were accompanied by short, medium and long ‘cohort’ predictions with this data, setting out three views about how long mortality improvements will increase. These tables have become out of date, having been released in 2002 and based on data ending in 1999.

In October 2008 the CMI created its first ever analysis based on Self Administered Pension Scheme (SAPS) data. Schemes approaching valuation now will use this suite of tables instead.

Further change is due this year. CMI SAPS tables will soon be enhanced by the CMI Mortality Projections Model. The prototype model was published in a working paper on 19 June and is the subject of a consultation process that will end on 31 August. The final model is expected to be published during October 2009.

Hewitt Longevity Solutions head Martin Bird said: “We are impressed with the way the new model can be adapted to cater for different views of the future. It has the advantage that its inputs are intuitive, unlike the opaque adjustments that are needed for the Interim Cohort Projections currently in use.”

One size does not fit all

Large consultancies can supplement the CMI tables with data collected from their clients. For example, Watson Wyatt breaks data down across industry sectors but this analysis is not published.

“SAPS tables are a good starting point, but it might not pick up the specifics of your individual scheme. There are a variety of reasons why mortality can vary,” says Butterfield. He cites the following as the three main measurable factors:

• Pension income
• Industry sector
• Postcode

Pension income is used to approximate lifetime earnings, the real driver of life expectancy. The flaw is that it embodies both earnings level and length of service. “It is a reasonable proxy using the data that you have got,” says Butterfield.

Educating Vita

Consultants and actuaries Hymans Robertson last year launched its own longevity modelling project, Club Vita, offering an alternative to the CMI data. It is based on up-to-date pension scheme data that is diversified across industries and geographic regions. The pilot phase is now over and enough data has been collected to be used for valuations.

“The Regulator is very keen that an objective approach is used for mortality assumptions and schemes allow for the specific nature of their workforce,” says Andrew Gaches of Club Vita. “Club Vita does exactly that.”

Club Vita research found that actual improvements in lifespan do not follow any of the CMI’s three interim cohort projections. Gaches says that this is “perhaps no great surprise” as the data is not from pension schemes and is out of date.

The data has also uncovered the extent to which salary affects life expectancy: an individual between 60 and 65 years old earning less than £15,000 is more than three times as likely to die in the following year than someone earning more than £35,000.

“We have seen a rapid rate of improvement of two years per decade,” says Gaches. “It is significant but not as rapid as predicted by CMI cohort projections.”

He says governance processes should include continuous monitoring of the scheme’s mortality improvement allowance, so trustees are in a position to consider risk-reducing opportunities. “Before taking action trustees need to understand their own position,” he adds.

External analytics can help where there has been a significant workforce transition – for example from manual to office-based – that makes pensioners’ mortality misleading.

Scheme maturity also plays a role. Longevity is very real and current for mature schemes; young schemes have insufficient data and can change the benefit structure.

Even the best laid plans…

It is important to realise that the best models can still be blown apart by unexpected events, medical break-throughs or simply changes in lifestyle.

Says Butterfield: “On the one hand a cure for cancer may be around the corner and on the other swine flu may be devastating. How can anyone model for that?”

Even on the potential porcine problem there is a matter of opinion. Gaches says: “Even if swine flu is much worse than anticipated it would not have a ‘beneficial’ impact on funding – the economic turmoil would outweigh any effect.”

Trustees must adopt some measure of future liabilities and this should be based in reality as much as possible. However, as the answer is essentially unknowable, trustees need to think carefully about how much of their resources to dedicate to the task. Modelling longevity is still as much of an art as it is a science.