Norman Dreger and Andrew Arbour outline why companies with pension obligations in multiple countries should consider carefully which mortality tables to use for accounting valuation
Mortality assumptions used for the valuation of pension benefits can have a large impact on the calculation of pension liabilities, so the selection of appropriate mortality tables for companies operating in a single country is crucial. For multinational companies with pension obligations across several states, the selection of mortality assumptions is even more complicated. There is pressure to balance the desire for a globally consistent approach to assumption-setting with the need to consider the mortality tables used as standard in different countries. This article examines the impact of country-specific mortality assumptions on the pension liabilities of multinational companies and describes the options these companies should consider in selecting appropriate mortality tables.
Life expectancy and mortality tables
There are two basic types of mortality tables - static mortality tables and generational mortality tables. Static or ‘period' tables contain probabilities of death for each age, typically separately for men and women. These tables are relatively easy to construct and for actuaries to work with. The downside with this approach is that, as life expectancy continues to improve, these tables can quickly become out of date and may need to be updated on a regular basis.
Generational tables are more sophisticated and include assumptions with regard to current and expected future rates of improvements in mortality.
Such tables are more difficult to construct, but technology can assist in the production of these more complex models. These tables are likely to more accurately reflect the actual life expectancies of plan members, provided that the underlying assumptions are reasonable. An added bonus is that, since future improvements are built in to the tables, they may not need to be updated in the future as frequently as would be necessary with a static mortality table.
During the past 10 years, the use of generational mortality tables has become the most common approach in all major developed countries with material pension liabilities. New generational tables have been introduced in Austria, France, Germany, the Netherlands, Switzerland and the UK. The exceptions have been Japan and the USA.
In Japan, the choice of mortality table typically does not have a major impact on the liability amount calculated, as most benefits are paid either as lump sums or fixed-term instalments at retirement. As such, the use of a computationally more complicated generational mortality table has not yet become standard practice in Japan.
In the US, it is also quite common to pay out lump-sum pension benefits at retirement, and when a lifetime pension is granted, it is uncommon for the pension benefits to receive indexation adjustments. As a result, pension liabilities in the US are typically weighted more heavily to the shorter term and are therefore less
sensitive to the choice of mortality table as compared with many other developed countries. However, although it is still common to use a static mortality table in the US for accounting purposes, the use of generational tables is increasing, and even when a static table is used, it is common practice for some type of mortality improvements to be taken into account. For instance, when using the RP 2000 table, a projection reflecting mortality improvements from 2000 to the valuation date plus seven years is typically used, in order to reflect average mortality improvements for the population to be valued as of the valuation date.
Comparison of statistical data with common mortality tables
The issue that the mortality tables aim to model, life expectancy, is of course, complicated. Life expectancy varies from country to country depending on, for example, levels of access to health care for the population, relative affluence, diet and the genetic disposition of the population. This is part of the reason why, when conducting actuarial valuations, different mortality assumptions are used in different countries. Another reason is simply that people just use the tables that they always have used.
However, actual differences in life expectancy may not be the only factors which result in differences between the ‘observed' mortality data between countries. For instance, the sampling and statistical approaches used to measure life expectancy will vary, or information might only be available for part of the population.
The following example shows how data can differ. In the analysis, mortality statistics were gathered from the UN Demographic Yearbooks from 2000, 2002, 2005 and 2008 (observed mortality data for the years 2009-11 had not been published as of the time of writing). In order to make the most up-to-date comparison and to bring the mortality data onto a common basis for comparison purposes, the life expectancies from the various UN reports were projected to 2011 using extrapolation methods. Certain outliers in the UN data were also removed as part of this exercise.
Figure 1 shows the estimated period life expectancy of a man currently aged 65 in each of the countries considered. The number represents the average number of years that a person aged 65 can be expected to live if mortality rates did not change over their remaining lifetime. It's worth noting that these figures make no allowance for future improvements in mortality.
Based on this UN data and the analysis performed, the highest current life expectancies for 65-year-old males are for Switzerland, Spain and Japan at just over 19 years. The country with the lowest projected life expectancy examined is Portugal with a projected life expectancy of just over 17 years.
However, contrast this with figure 2, which shows life expectancies derived from mortality tables typically used in 2011 for accounting valuation purposes.
According to this data, the countries with the highest assumed life expectancy for a 65-year-old male are France with over 23 years, with Spain and the UK at around 22 years. The countries with the lowest life expectancy derived from accounting mortality tables are Portugal and Japan at around 18 years. Thus the range in these figures is more then five years, considerably wider than the period estimates would suggest.
Figure 3 shows the difference between the estimated period life expectancies, and the life expectancies assumed for accounting purposes under the commonly used mortality tables, for 65-year-old males in each country.
In all countries examined, with the exception of Japan, the mortality tables commonly used for accounting valuation purposes assumed a longer life expectancy than one might expect based on the historic statistical data. The country with the largest difference is France where 65-year-old males are assumed to live more than 4.5 years longer than current rates of mortality would imply. The UK follows with an allowance of just under four years.
It may seem strange that in some countries the life expectancy assumed in current mortality tables appears to substantially deviate from actual observed life expectancy. However, there are two key differences in the life expectancy figures being compared above:
• The main element appears to be the allowance for future mortality improvements, ie, assuming that plan members will live longer than current rates of mortality would suggest; and
• A second adjustment is to take account of the differences in mortality experience between plan members and the general population. As pension scheme members tend to live longer than the general population, we would expect this effect also to make a positive contribution to the above figures.
The figures for each country also need to be considered in their own context:
• The mortality assumptions used in Japan appear, at first, to be relatively weak, which one might suspect may lead the liability amounts in Japan to be understated. However, as the majority of pension benefits in Japan are payable in the form of a lump sum or instalments, there is little longevity risk, so the mortality assumption chosen does not typically have a large impact on liability levels.
• In France, it appears that the life expectancy assumed under the typical accounting mortality tables is significantly longer than the life expectancy observed in the general population. When one considers, however, that in France usually only highly paid employees would have entitlements under a defined benefit pension plan, the assumed life expectancy for valuation purposes may, in fact, be broadly reasonable for most plans.
However, there are also some cases where it appears that there may be some discrepancies which are not so easy to explain. For instance, although German and Austrian period life expectancies are within half a year of each other, the Austrian mortality tables assume that a 65-year-old male will live almost two years longer than his German counterpart. If Austrian mortality tables were to be applied for the valuation of German pension obligations, this could result in a material increase in the cost of providing German pensions.
As a rule of thumb, one might expect that an increase in life expectancy of one year might lead to an increase in pension obligations by 2-4%. Assuming that a pension plan has pension assets equal to 90% of the pension obligations, a 2-4% increase in pension obligations could lead to a 20-40% increase in the liability shown in a company's accounts. Thus even minor adjustments to actuarial assumptions such as mortality can have a very material impact on the pension liability shown in a company's balance sheet.
Discussion and conclusions
When deciding on which mortality assumptions to use for accounting valuation purposes, companies with pension obligations in multiple countries are likely to need to balance the desire for global consistency with the need to comply with local market practice. Often these two objectives will be at odds with each other; companies that insist on using consistent mortality tables (or even that the same type of tables, eg, generation tables) in every location where they have pension liabilities, will find that they are acting out of line with local market practice in some cases. When this happens, a company should make a conscious decision whether to adhere more to local practice, to strive for global consistency, or to make some type of a compromise, in order to be able to ultimately defend the approach used to auditors, plan participants and other interested parties.
Norman Dreger is a principal in Mercer's international consulting group, based in Frankfurt am Main, Germany. He also chairs a global working group at Mercer which deals with the selection of actuarial assumptions for accounting valuation purposes.
Andrew Arbour is a mathematics student in his final year at the University of Freiburg, Germany