Switzerland: Accounting mortality
Benno Amrosini and Ruben Lombardi assess the application of generational mortality tables, and their impact on the balance sheet and income statement
With the publication of the BVG 2010 actuarial tables, Switzerland now has its first generational table based on observations of mortality in predominantly private pension funds. Generational tables take account of future trends in mortality - in accordance with the model used - and therefore comply fully with the ‘best estimate' requirements of IAS19.
IAS19 requires various assumptions such as the discount rate or expected salary increases to be taken into account when calculating a company's pension liabilities (present value of the defined benefit obligation or DBO) and pension expense. One of the key assumptions, especially when calculating these figures for Swiss pension funds, is the expected mortality rate of beneficiaries (active members and pension recipients).
Under IAS19.73 assumptions - including mortality rates - are required to represent what are termed ‘best estimates'. On 29 April 2010, the International Accounting Standards Board (IASB) published a series of proposed amendments to accounting standard IAS19 (Exposure Draft ED/2010/03). These confirmed the principle of best estimates for assumptions and also stipulated that they should include ‘current estimates of the expected mortality rates of plan members, both during and after employment'.
Static or generational?
Until now, IAS19 calculations in Switzerland have chiefly been made using static tables. These indicate the probability of a person of a specific age (eg 60) dying the following year. The expected one-year mortality rates depend solely on the person's age and gender. They are based on the mortality of a particular group of people during a particular observation period, generally covering a few years. The expected mortality rate does not change over time. The probability that someone currently aged 65 will die today is the same as the probability that someone who is now 50 will die in 15 years' time at the age of 65. That is why this type of mortality table is termed static.
Over the last century, however, mortality has not remained constant, but has steadily decreased. Life expectancy at birth has almost doubled. This trend has persisted in recent decades, a fact confirmed by comparing the latest actuarial tables (BVG 2010, published December 2010) with those of BVG 2005. For example, the life expectancy of a 65-year-old man has risen from 17.9 years under the 2005 actuarial tables (survey mid-point 1 January 2002) to 18.93 years under the 2010 tables (survey mid-point 1 July 2007). If we assume that life expectancy will continue to increase in future, in other words that mortality at a given age will decrease, the expected mortality rate will depend not only on age but year of birth, since the later a person's year of birth, the greater their longevity. Generational tables take account of this mortality trend and indicate expected mortality rates that are dependent on both age and year of birth.
As mentioned the BVG 2010 actuarial tables were published in December 2010. The static tables they contain are based on observations of the mortality of beneficiaries in fourteen large, predominantly private pension funds in the years from 2005 to 2009. The BVG 2010 tables were also compiled in generational format and reflect the current state of research regarding the extrapolation of changes in mortality (Menthonnex).
As indicated for the purposes of IAS19 calculations the assumptions and the expected mortality rate must represent ‘best estimates'. If we assume that life expectancy will continue to rise, the expected mortality rates in a generational table constitute the best estimate. If, on the other hand, we believe that life expectancy will remain unchanged, the current static tables are the best estimate. Mortality observations over the last century, the last few decades and the last few years indicate that the trend towards declining mortality rates is set to continue. This means that generational tables can be assumed to represent the best estimate of future mortality rates. In other countries, such as Germany, the UK and the US, they have already been regarded as best practice for some years.
Impact on pension liabilities and service cost
Pension liabilities under IAS19 depend on the actuarial tables used (and in particular the expected disability and mortality rates). We calculated the impact of the BVG 2010 generational and static tables on the pension liabilities and service cost of a sample population (active members and pension recipients) of a typical pension plan, compared with the previous 2005 static tables. This is merely an example, and the results for other populations and/or pension plans may be significantly different. The previous IAS19 calculations using static tables typically assumed a future increase in life expectancy, and a percentage loading was added to the pension liabilities and service cost to take account of this. We have factored this practice into our calculations. The results are set out below.
The changeover to the BVG 2010 static tables (ie, excluding a switch to the generational tables) results in an approximately 6.1% decrease in the pension liabilities towards active plan members. This reduction is primarily attributable to the significant decline in expected disability rates in BVG 2010 compared with BVG 2005, which is independent of the type of mortality table and is therefore equally evident in the static and generational tables. As might be expected, the liabilities are higher if they are calculated using the BVG 2010 generational tables rather than the static tables. Nevertheless, for the population of active members investigated, the liabilities are around 1.0% lower overall than they were with the old BVG tables. The effect on the pension liabilities for active members resulting from the marked fall in the expected disability rates is thus larger than the impact of the changeover to generational tables.
In our example, these two contrasting effects enable a changeover to the BVG 2010 generational tables to be carried out without any major change in pension liabilities - at least for active members. For pension recipients, however, the liabilities increase: a switch to the BVG 2010 static tables causes a rise of just 0.7%, but if the new generational tables are used, the figure is more substantial, at around 5.2%, or CHF12.8m (€10.1m). The modest increase when changing over to the BVG 2010 static tables is due to the fact that the effective increase in life expectancy between the BVG 2005 static tables and their BVG 2010 equivalents is higher than is modelled by the percentage loading in the calculations using BVG 2005.
For the plan population we examined, a changeover to the current BVG 2010 static tables leads to a reduction in liabilities of about 2.6%, compared with a rise of 2.2% when using the new generational tables. The change in the level of liabilities depends not only on the ratio of active members to pension recipients, but also on their respective ages. The older the population of active members, the lower the influence of expected disability rates, and the higher the impact of life expectancy upon retirement. The changeover to a BVG 2010 actuarial basis therefore causes the DBO to increase more sharply for a plan participant of retirement age than it does for a younger one.
Another key factor determining whether the liabilities are higher or lower using the BVG 2010 generational tables is therefore the structure of the plan population. The use of different actuarial tables affects not only the pension liabilities but also the service cost, which is relevant for the calculation of pension expense in the IFRS income statement. A changeover to the BVG 2010 static tables leads to a fall of around 11.7% in the employer's share of the service cost; using the BVG 2010 generational tables the figure is around 19.2%. This is attributable to two main effects:
• The reduction in expected disability rates under BVG 2010 that was mentioned earlier; and
• The different methods used to take account of decreased mortality (percentage loading for static tables, dynamic adjustment for generational tables). The economic benefit under IFRIC14, which may in some cases have an impact on the asset recognised in the balance sheet under IAS19, is also reduced by a corresponding amount.
Balance sheet and income statement impact
In the previous section, we demonstrated that the application of the BVG 2010 generational tables causes a change in pension liabilities and service cost. This change has an impact on the pension provision in the balance sheet and the pension expense in the income statement. If the actuarial gains/losses are reported directly in ‘other comprehensive income' (the ‘OCI method'), the changeover to the generational tables has an impact on equity.
As demonstrated, the changeover to the BVG 2010 generational tables results in a 5.2% increase in pension liabilities (DBO). These are actuarial losses that must be reported immediately under OCI. The pension provision in the balance sheet shows a corresponding rise of CHF12.8m (€10.1m) as of 31 December 2010 when compared with that calculated using the earlier BVG 2005 static tables. The changeover to the BVG 2010 generational tables as of 31 December 2010 has no impact on the 2010 income statement because the components of pension expense for 2010 were still calculated using the BVG 2005 static tables. The pension expense for 2010 stands at CHF14.5m, both with and without the change of actuarial tables.
However, the changeover to the BVG 2010 generational tables results in a reduction of around CHF3.8m in pension expense for 2011 in the income statement, from CHF14.2m to CHF10.4m. This is mainly due to the reduction in the service cost. The changeover has no further impact on OCI in 2011. The effects shown here relate to this specific example and cannot necessarily be used to draw general conclusions. The changes in the balance sheet and income statement derive from a number of factors including the population of plan members (age structure, ratio of active members to pensioners, etc), the current method of calculation (eg, loading to reflect increased life expectancy), the method of reporting actuarial gains and losses (OCI method, corridor method) and the current IAS19 data (overfunding/underfunding, limit on the recognised asset, etc).
It is advisable to examine the impact of the changeover carefully in advance, in order to avoid unwelcome surprises. Under some circumstances, the changeover to BVG 2010 generational tables can be linked to a switch from the corridor method to the OCI method.
Benno Ambrosini and Ruben Lombardi are both members of the management committee of LCP Libera, Zurich, and Swiss certified pension fund experts