research team at Cass Business School, City University has carried out an analysis of the mortality assumptions used around Europe in the valuation of company DB pensions liabilities. The outcome is a detailed report (available from www.cass.city.ac.uk/media/stories/story

_12_18491_57598.html), the main points of which were presented at a seminar at Cass in November 2005 and which will also be discussed at the upcoming International Congress of Actuaries in Paris.

Background

There has recently been an upsurge of interest in the press and from analysts in the mortality assumptions used in pension liability calculations, which have a key effect on the balance sheets of many companies. Changes in financial regulation have meant that key assumptions must be disclosed, but, to date, little attention has been given to the mortality assumptions.

The report presents an analysis and comparison of the mortality tables commonly in use, suggests some possible improvements to current practice, and discusses possible methods for disclosure. There was a lively discussion from the audience at the presentation at Cass, mostly concerning the disclosure issues and the appropriate approach to setting mortality assumptions. It is clear that the view that mortality assumptions are too complicated to be disclosed meaningfully and that, therefore, nothing should be disclosed, will not be tenable in the future.

The project compared the mortality assumptions used in pension liability calculations across EU countries, and compared these with the population mortality tables within each country. We have added the US and Canada for comparison. It was not the aim of the project to produce new tables, but simply to carry out a comparison of the general practice in each country. It is recognised that often adjustments are made to these tables to take into account the particular circumstances of an individual company. However, in many cases, the standard tables are used unadjusted, and can be used as a proxy for the strength of the pension mortality assumptions used in each country. The research also considered a single statistic which could be disclosed as a measure of the strength of the mortality assumptions used in calculating the pension liabilities.

The results

The full report contains a detailed analysis for each country. The overall outcomes are presented in figure 1, which shows the difference between observed national population life expectancy and assumed life expectancy for company pension schemes, for a male aged 65. We see that the range of differences is very wide: in Denmark it is about zero while in France it is about 71/2 years. This reflects the fact that the between-country variation in observed population life expectancy for a man aged 65 is slightly more than two years, but it is much wider for a male 65-year-old member of a company pension scheme.

Pension liabilities are driven by the discounted value of annuity payments, so quoting a difference in future life expectancy is perhaps not the most useful way to indicate the effect of the mortality assumptions. An alternative is to calculate an equivalent discount rate. This is illustrated in figure 2, which shows the discount rate needed to keep the pensions liability constant when changing mortality assumptions. So, for example, the mortality assumptions used in Germany are less conservative than in the UK, and the difference is equivalent to an increase of over 1% in the discount rate. This is useful both because the discount rate is a familiar quantity for financial analysts, and also because it illustrates the relative importance of the mortality assumptions.

Discussion

It is to be expected that mortality assumptions in company pension schemes will vary between countries, because of variations in underlying population mortality as well as in variations in the typical membership of a company scheme. However, the results indicate that current variations in mortality assumptions are much greater than would be justified by these factors alone. A part of the variation arises from the fact that some countries incorporate an allowance for expected future improvements in mortality, while others do not.

The financial press and company analysts have realised that the mortality assumption can have a significant effect on the liabilities in company balance sheets. In order to illustrate this, figure 3 shows what the actuarial deficit would be using the mortality assumptions currently in use in other countries, for a pension scheme with assets of £800m, liabilities of £1,000m, and therefore an actuarial deficit of £200m when calculated using assumptions generally used in the UK.

Clearly, longevity is a complex subject, influenced by a range of economic, social, and environmental factors and it could have a significant impact on disclosed company pension costs and insurance company solvency. We certainly cannot predict with accuracy what mortality rates will be in 20 years’ time. However, it is clear from this project that an indication of the mortality assumptions used in valuing pensions liabilities needs to be included in company accounts so that a fair assessment of the financial statements may be made. The best method for disclosure, however, is not so clear. It is often suggested that assumed life expectancy at certain ages would be the most easily understood measure. So, for example, we could envisage a situation in which the disclosures say something like:

The standard current population tables give a life expectancy at age 65 of 17.2 years. The mortality assumptions in these accounts assume 24.4 years, which includes 5.4 years for expected future improvements in longevity and 1.8 years because of the make-up of the company scheme.

This would be a readily understandable way to express the mortality assumptions, but it may not be the most useful for a financial analyst. What they need is a figure which allows a quick, rough-and-ready, recalculation of the liability under alternative assumptions. For this, we would recommend that more discussion is needed, and that measures that could be considered are equivalent discount rates (as in figure 2) and relative annuity values. For the latter, we could use the ratio of the value of a standard annuity under the mortality assumptions used in the accounts to the value under a standard set of assumptions for a particular country.

06_05_06.pdf