Quick links The costs   The actuarial approach   The impact of projecting mortality   The risks for companies   Typical practice   What companies can do   Conclusion     Main Global Retirement page   About the author    Duncan Smithson    +1 312 902 7003     Duncan Smithson is an actuary based in Chicago and is head of Mercer’s International Retirement consulting group in the Midwest.

“I don’t want to achieve immortality through my work; I want to achieve immortality through not dying.” – Woody Allen

Regardless of your views on immortality, one of the great achievements of the modern age, thanks to advances in health care and improved living standards, is that people are living longer. And in some countries, and for some groups at least, the rate of improvement itself seems to be increasing. For example, the average life expectancy at birth for a US male has risen from 45 in 1900 to over 75 today. Similar dramatic improvements have been seen in other countries. Improvements in longevity also vary among different “cohorts” or generations of people. For example, the greatest increases in the UK have been for those born in the 1920s and 1930s, and in Japan, for those born in the 1910s.

Given these variations, there is great uncertainty about future rates of improvement and about whether there is an ultimate age beyond which human survival is not possible. Although the longest documented life span is slightly over 120 years, experts are still debating whether there are indeed any finite limits to human life. So Woody Allen may yet achieve his goal of “not dying.”

The costs

These welcome improvements in longevity have come at a cost. Although there are wide-ranging, socioeconomic costs associated with improved longevity, it is the pressure on retirement income that has frequently caught the attention of governments, companies, the press and the general public. State social security retirement systems have come under increasing strain as finite amounts of capital are used to provide retirement income for longer periods than were ever imagined when the plans were designed. Pay-as-you-go systems have also had to cope with a declining proportion of active workers to pay for the current benefits of an increasing proportion of pensioners, and governments everywhere have reacted by reducing or changing benefits in some way or by passing the additional costs on to taxpayers. For both state and private pension plans, this crisis has been exacerbated by a trend of people retiring earlier, when arguably, retirement ages should have been increasing if costs were to be kept stable. Governments in many countries have started to address this issue, but to fully understand the impact on companies that sponsor defined benefit (DB) plans, we must first review how actuaries model mortality in their calculations.

Back to top

The actuarial approach

In the first half of the 20th century, to value pension liabilities and costs, actuaries used mortality tables that simply showed the probability of a person aged x dying before reaching age x+1. These calculations were based on the experience of a particular group of lives (for example, the general population, life office annuitants or pension fund members) at a specific point in time. As it became increasingly clear that this approach did not allow for improvements in mortality over time, actuaries began using projected mortality tables. These tables still show the probability that a person aged x will die before reaching age x+1, but they might be set at 80 percent of the corresponding probability in the base (unprojected) table to allow for future improvements. However, in recent years, actuaries have realized that despite their best efforts to accurately model future mortality, they have often understated mortality improvements, as experience has consistently turned out to be better than even the most optimistic projections.

To better capture mortality assumptions, actuaries in countries such as France, the US, Germany, Spain and the UK have developed so-called generational tables that show the probability that a person aged x who was born in year y will die before reaching age x+1. In addition, actuaries in countries where there is insufficient data to produce generational mortality rates often use projected tables adjusted in ways that attempt to mirror the generational effect.

A “generational” approach

The rates of mortality table (shown below) illustrates different ways in which actuaries construct sets of mortality rates. The rates of mortality circled in red are taken from a static or base table. The probability that a 55-year-old will die in the next year is 36 in 10,000, regardless of when he turns 55. The rates circled in green show the same base rates projected forwards five years, allowing for a small improvement each year. The probability, though now reduced to 28 in 10,000, is still assumed to apply regardless of when the person reaches age 55. The blue arrow shows a possible progression of generational rates. The probability is 36 in 10,000 for someone turning 55 in 2006, but only 30 in 10,000 for someone turning 55 in 2007, 26 in 10,000 for someone turning 55 in 2008.

While not wanting to increase company costs needlessly, Mercer now requires its actuaries to use up-to-date projections of mortality when calculating pension plan solvency or contribution requirements. This is, however, merely part of an overriding approach to meet regulatory requirements and to use assumptions that reflect the specific circumstances of each pension plan. Thus, there may be specific evidence (for example, a study of actual plan mortality experience) to suggest that plan members will experience heavier-than-average mortality.

It is interesting to note that there is so much uncertainty surrounding future improvements in mortality that the very latest UK actuarial tables do not even allow for projected improvements or generational effects; instead, individual actuaries decide what allowance to make. In contrast, the US Social Security trustees project a rate of mortality improvement for the next 75 years that is only slightly less than the actual rate of improvement experienced during the last few decades. The point being that estimating future mortality experience is fraught with uncertainty.

Back to top

The impact of projecting mortality

Planning ahead The outer bound of how much mortality changes can affect a capital cost is set by the “immortality annuity” level. For example, if you could earn 8 percent per annum interest forever, you would only need $12.50 to provide a $1 of income every year. Changing mortality will not have much effect. But if the income were indexed to prices and the yield available dropped to 1 percent (interestingly, the real yield on UK government index-linked stock in January 2006), the equivalent capital cost is $100 – plenty of room for life expectancy to have a major effect!

So how significant are the differences between the various types of mortality tables? Or more important, what is the impact on pension liabilities and costs of changing the assumed level of future mortality? The answer, of course, depends on the current mortality assumption and on the prevailing view of future yields.

In the UK, where pensions are index-linked (and for which the impact of changes in mortality are therefore more significant*), a pension liability calculated using a base (that is, unprojected) mortality table might typically increase by 30 percent when switching to the same table projected forwards to allow for future improvements, and by 35 percent when switching to a generational table based on the same base mortality table. (Figures are based on the approximate increases at age 65 for a UK male born in 1950. For older plan members and for current pensioners, the impact will be less.)

In the US, where (private sector) pensions are not generally index-linked, the impact of changes to assumed mortality is less dramatic. Moving from a base table to a projected table would increase the liability for a US male born around 1950 by approximately 8 percent. Moving to a generational table would increase the liability by 10 percent to 12 percent. For various reasons, increases for female members would be much lower, and as in the UK example, the impact would also be less for older plan members and for current pensioners.

For many (if not most) pension plans, these potential increases in liabilities will already have materialized because very few actuaries today would be recommending the use of the unprojected base tables for the examples above (even though such mortality tables or broadly equivalent tables were certainly in use until relatively recently). Rather, the issue at stake is the extent to which such significant increases in liabilities and costs will be repeated in the future.

Back to top

The risks for companies

Obviously, a risk for companies that sponsor DB plans is that they and their actuaries may use mortality tables that significantly understate the ultimate length of time for which future pensions will be payable – and therefore significantly understate the cost of providing those pensions. For funding valuations where tables have not already been moved to a projected basis, companies will likely face pressure from plan trustees, management boards, employee representatives and, possibly, regulatory authorities to move to more realistic mortality assumptions, leading to increased cash contributions to the plan.

Companies that address the issue proactively should be able to manage the timing of any increase in cash costs and ensure that necessary investigations are done at a time suitable to them and which gives them ample time to benchmark their own specific experience. A less-favorable outcome for companies that resist moving to more up-to-date mortality assumptions would be unexpected future increases in the cash cost of the plan to keep the fund solvent.

For accounting valuations, the pressure to use more realistic mortality assumptions is likely to come from regulators and auditors for whom the mortality assumption has currently become a hot topic. If companies assume heavier mortality than their plan members actually experience over time, actuarial losses will emerge as pensioners live longer than expected. Under some accounting standards, these losses will need to be recognized immediately as they arise. Companies may prefer this (especially if recognition is through a secondary revenue account instead of through earnings) to the alternative of allowing for lighter mortality, and therefore booking a bigger expense through earnings each year. But bear in mind that companies typically use the same mortality assumption for their accounting valuations as for their funding valuations. While a typical approach, there is an argument that the prudence built into a funding valuation to calculate plan solvency and contribution requirements is not appropriate for a more realistic accounting valuation.

Back to top

Typical practice

So the question arises – which mortality tables are companies typically using?

The answer is often difficult to obtain from publicly available information. Very few companies disclose what they are using for accounting purposes (none of the major accounting standards require this to be disclosed), and in only a few countries are funding valuation results publicly available. Furthermore, where information is available, it is not always clear whether other assumptions have been adjusted to make approximate allowance for improving mortality (for example, using a static mortality table, but reducing the discount rate by 0.5 percent to compensate).

A recent Mercer study of the little information that is publicly available showed wide variations in the mortality tables used. Even in countries for which generational tables are readily available (such as the US and the UK), only a relatively small proportion of companies use them and of these, most use the least-optimistic projection basis. Most companies around the world seem to use static tables, projected by 10 to 15 years into the future. (In some countries, this may be approximately equivalent to using a generational table, but this is not the case universally.)

Why? Companies may be reluctant to make changes that cannot easily be reversed. If discount rates, for example, reduce, then presumably at some point they may increase, so increases in estimated liabilities now may be offset by reductions in estimated liabilities in the future. For mortality, however, companies may feel that once they have moved to a lighter mortality table, it will be difficult to justify (to auditors) moving back to a heavier table in the future. It is seen as a one-way street.

For many companies, the uncertainty surrounding improving mortality has been just one more reason (along with cost volatility and an everincreasing regulatory burden) to consider changing the design of their pension arrangements, often to a defined contribution (DC) design or to some other plan form that delivers a lump sum benefit to retirees (either directly or to purchase an annuity). In some countries, such as Japan and Germany, DC plans, while legally possible, have limited tax advantages and can also present cultural challenges, so moving to a DC plan may not be as straightforward as it seems. In other countries, DB retirement plans are mandatory (for example, collectively bargained retirement indemnities in France), so moving to a DC plan is not an option. But even for DC plans, the problem of increased longevity remains with associated risks shifted to the employee (unless the employer provides annuitization) and with the employer’s role focusing on the adequacy of benefits and on communicating the risks to individuals who might be ill-equipped to understand or manage those risks.

Companies that have retained a DB plan design have considered increasing employee contributions or reducing benefit accrual formulae, among other things. Interestingly, although many governments worldwide have increased state retirement ages, few private companies seem to have raised the age at which plan members are entitled to receive retirement benefits (unless compelled to do so by legislation, as in Japan). In contrast, some jurisdictions will not allow increasing the age at which unreduced benefits may commence (as under ERISA in the US). Many employees are facing potentially lower employer-sponsored retirement benefits at the same time as potentially lower state retirement benefits.

Back to top

What companies can do

So what else can companies do to manage the risk of surprises apart from changing the design of their retirement plans? Here are some suggestions, in roughly ascending order of difficulty.

1. Review current processes: Many companies already manage the process supporting their selection of funding and accounting assumptions well. But even so, companies should make an effort to understand their plans’ assumptions (not just in the headquarters country, but also for their major plans around the world) and how they relate to their specific experience. And they should thoroughly review their plans, possibly analysing industrywide experience or segmentation of their plan membership in more detail. 
   
   
2. Consider disclosure: “Forewarned is forearmed.” Pressure for disclosure will not go away, so while disclosing financial implications clearly and concisely can be a challenge, auditors and the investment community are likely to react favorably to greater transparency.
   
   
3. Analyse the risks: Carrying out an analysis to identify and quantify the sponsor’s main risks, including mortality, in providing retirement benefits to employees and ex-employees will let companies explore how these risks can be mitigated or managed. 
   
   
4. Buy out the risk: In countries with developed insurance markets, buying out the pension liabilities using group annuity insurance policies is a possibility. In theory, this is one way to settle all aspects of the pension obligation (not only mortality risk, but also investment and expense risk) and may be the only way of terminating a plan. In practice, however, there has been limited insurance capacity, and the margins charged to cover uncertainty around future mortality improvements have made such policies appear to be very poor value for the money. (Insurance buy-out costs can easily be 20 percent to 30 percent or more above the liabilities estimated under US GAAP or IAS 19.) But this might be changing. In the UK, there were only two main insurers in the market a year ago, but several others have now entered the market. 
   
   
5. Insure against volatility: Innovation is leading to other ways of carving up the mortality risks of pension funding more cost-effectively. Reinsurance of longevity risk is one option, although this is very challenging to set up and has not yet been used extensively, as well as various solutions provided by investment banks. BNP Paribas recently offered a “longevity bond” to UK pension funds. Although this product was subsequently withdrawn, it is only a matter of time before other novel solutions are brought to the market.

Back to top

Conclusion

Mortality rates have decreased significantly over the last few decades, and the improvement continues. But there is considerable uncertainty over future trends in mortality. Companies that sponsor DB plans and their actuaries are left with difficult decisions about what allowance to make for future age improvements when determining cash and accounting costs. While many companies continue to offer DB plans to employees, uncertainty around mortality is just one more factor driving the global trend toward DC retirement designs. So while “not dying” seems to be something everyone can aspire to, financial security in old age is becoming much less certain and much more urgent.

“I don’t want to live on in the hearts of my countrymen; I want to live on in my apartment.” – Woody Allen

Back to top

* The impact of mortality improvement on index-linked pensions is greater than on flat pensions because the amounts paid to those who live long are much greater.