(The report is dated 16 October 2001. You might be surprised that a report purporting to reflect the financial prospects of the fund at December 2000 can have its parameters adjusted over a period that includes Sept 11 2001 but the regulations do allow that.)

The starting point for us is the calculation of the reserves, also known as surplus.

• surplus at December 1997 was 548 million pounds.
• add "Expected increase in surplus" of £386M.
• add "Higher than assumed investment return" of £265M
• subtract "Effect of contribution holiday" of £243M
• subtract "Changes in valuation basis over the intervention period" of £340M
• subtract "Miscellaneous" of £31M

The result is a surplus at December 2000 of 585 million pounds.

Taking the components separately, "Expected increase in surplus" is not quite what the title suggests. It means "Expected increase in surplus if the employer made contributions". The employer did not make significant contributions, and was never expected to. That is shown as "Effect of contribution holiday". So the really expected increase in surplus was £386M - £243M which is £143M.

(You might ask why a scheme which in 1997 had 23% more funds than it needed (represented by the £548M) was setup to increase its reserve by another £143M in the next three years. The answer is that the company's benefit, the £243M, was as big as it could be without the company being given a direct payment from the reserves. The only way to avoid planning for the extra £143M of surplus would have been to give the members better benefits.)

The £243M comprises both a holiday on the M-plan and a holiday on the C-plan. We know of the cost of the M-plan for 1998 to 2000 inclusive was £40M. The C-plan part was actuarially assessed as 17% of the total pensionable salary bill.

Because two of the three years involved were good years for investments, the £143M proved to be an under-estimate by £265M, the "Higher than assumed investment return". The assumed return was 8.25% pa. You can get a rough check on what actually happened from your Members' Reports which show the gains (without accounting for contributions and for benefits paid out) to be £1108M over the three years.

So in terms of the December 1997 Actuarial Report and what actually happened in 1998-2000, the new actuarial report says the increase in surplus would have come to an extra £143M (planned) plus an extra £265M (unplanned) giving a new total surplus of £956M. However, that number does not appear in the new report because the new report says the C-plan and M-plan together are going to cost £371M more than it was thought they would cost in 1997. After taking that £371M extra cost into account, the £956M becomes the latest actuarial surplus of £585M.

So in terms of the past, the new report reasonably accounts for how events differed from the 1997 assumed plan. The £371M of extra costs is a more difficult story to follow. First we consider the split between "Changes in valuation basis over the intervaluation period" (£340M) and "miscellaneous" (£31M). One dictionary of pensions terminology (ISBN 1 898785 31 7) defines "valuation basis" as "Commonly used by actuaries to mean valuation method and/or actuarial assumptions". In our case the meaning must be "actuarial assumptions" because the valuation methods were the same in both the 1997 and 2000 reports. (They are called the "Attained Age Method" and "Market Related Value", if you care to know). Both of the reports have an appendix listing the actuarial assumptions, and the assumptions differ. This is the cause of the £340M extra cost.

The miscellaneous £31M is a lot of money. It would comfortably pay what was needed to recompense the IBM retirees for the erosion in the value of their pensions in the nineties compared with the non-erosion of pension seen by the average UK retiree (although it is not enough to extend that benefit into the future). We cannot say how the £31M comes about, as we are not told, but we can say what it is not. It is not a cost that was forseen in 1997, since it is an extra cost. It is not related to poorer conditions in the financial markets, since these are covered in the actuarial assumptions. Extra costs for defending trustee actions in the High Courts are a possibility, but could hardly come to that much.

The next question is how did the changes in actuarial assumptions lead to a change of £340M in the predicted costs? The report does not tell us this but we can make some calculations of our own. One factor is that the "real" rate of return, which is the rate at which the actual return on investment exceeds inflation, has been changed. For the calculations about 1997 status it was assumed to be 4.25%, and for the calculations about 2000 status it was assumed to be 4.00% except in one case. For that case, calculating the future cost of actives and deferreds when they have become retirees, the return was assumed to be 3.50%. (This is an unexplained accounting artifice - if funds are earmarked for specific purposes, why would the fund managers get 3.5% on the funds for one purpose when they were getting 4% on funds earmarked for another purpose?)

It might seem strange that after a period when the actual real return was much higher than expected (7.8% versus 4.25%), the expected return for the future should be decreased rather than increased. However, you will probably agree reduction is wise in the light of events in 2001.

A second factor is the change in assumptions about how long we will live. The assumptions in the new report are, for men in service, "A67/70 ultimate rated down 3 years", for women in service "FA75/78 ultimate rated down 1 year", and for retirees "Standard tables PMA80/PFA80, calendar year 2010 rated down 1.5 years". Deciding amongst various tables to be used for assumed longevity is part of the actuary's "know-how". The numbers in the names of tables refer to the time when the data was collected, so A67/70 uses data collected in 1967 to 1970. It may seem strange to use such old data but the "rated down three years" is intended to correct for changes in longevity since then. We cannot work out the effect of using these tables because we do not know what tables the 1997 Actuarial Report used. The 1997 report just gives a few example numbers and says that "The allowances for death are based on standard published tables". It is just possible that the 1994 Actuarial Report would help since the 1997 report refers to it. (The pension regulations require the trust to give us copies of the current actuarial report but not previous ones.)

In the new report, various numbers are "out of a hat" in the sense that they are not explained and are not derivable from other numbers. In particular, the actuary recommends that the employer should contribute £19M to the pension funds in year 2002. While that is just a few percent of what the IBM balance sheet will gain from contribution holidays in 2002, it is welcome news. The actuary's recommendations are usually acted on; the trust and actuary agree the actuary's report before it is published. This change may herald a softening of the existing policy of moving sums from the reserves to IBM as fast as possible.

In summary, the good economic conditions in 1997 to 2000 would have made the reserves about a billion pounds but for the fact that the actuaries have assumed that future costs will be higher, reducing the reserves to about half a billion. There is some expectation that the reserves will not be used for IBM's benefit at the fastest rate possible.