The introduction of new regulations for Dutch pension funds brings forward an aspect of pension liabilities which so far has remained underexposed: the interest rate risk. In the headlines for the new regulatory framework (in Dutch ‘Nieuw Financieel Toetsingskader’, abbreviated as ‘NFTK’) and the new international financial reporting standards (IFRS) the pension agreement between employer and employee is pivotal. Pension liabilities are regarded as bonds issued by the pension fund (or sponsor) to the employee, whose value is determined analogously to the value of similar bonds traded in the capital markets. Most pension contracts are constructed around statements such as ‘nominal pension guaranteed, with the ambition to compensate for inflation'. Parliament has decided that the guaranteed component will be the starting point for the regulator. The result is that accrued pension benefits are valued at the nominal interest rate term structure rather than the flat and constant 4% actuarial interest rate used nowadays. If the current funding ratio is about 105% it is expected to increase to 115-120% under the new NFTK regulations, given a current long-term interest rate of about 5%. Another consequence of the transition to market value is that the interest rate sensitivity of the funding ratio will increase (if the investment policy remains unchanged). This is shown in the figure below.
Figure 1 yields three important insights. First, the interest rate risk in the current actuarial environment is relatively low compared to the market-value environment, where the interest rate and equity risks are of similar magnitude if the investment policy is not adjusted. This is because traditional bond investments have relatively limited interest rate sensitivity and the liabilities are currently totally insensitive to interest rate movements due to the constant actuarial interest rate of 4%.
In the new situation, where the value of the liabilities does fluctuate as a result of interest rate movements, the influence of interest rate movements is much bigger due to the difference in duration between assets and liabilities.
The second notable feature is that in the actuarial situation an increase in interest rates negatively affects the funding ratio, while under the new regulations a decrease in interest rates would lead to a substantial decline of the funding ratio. In the actuarial environment the duration of the investments is 2.5 against 0 for the liabilities, while in the market-value situation the same 2.5 is compared to 15. The small positive difference is thus changed into a large negative difference.
The third difference is that the total volatility increases from about 11% to 15% on an annual basis(1) , which means that without adjusting the asset allocation, the volatility of the funding ratio will substantially increase. While in the past the probability of underfunding with a funding ratio of 122% was equal to 2.5%, now a funding ratio of 130% is required for the same security. Note however that a funding ratio of 130% in the market-value situation is similar to 115% in the actuarial situation given current interest rates. This difference will be even greater if interest rates were to rise. The NFTK is therefore not by definition a more strict regime.
As mentioned before, the regulator focuses on the promise made to the employee, which in most cases will be the nominal pension. If the pension fund holds investments in its portfolio which generate exactly the expected pension payments (plus insurance against actuarial risks), the solvency test will reveal that no additional financial reserve is required. In that case there is virtually no risk for the pension fund anymore and the difference from an insured pension plan is relatively small in such a situation. This liability-matching solution, however, also does not allow the generation of extra return, which is required for indexing pensions. This is shown in Figure 2 with the dark-green dot. The yellow arrows indicate that indexation might be achieved by setting a higher contribution rate. Liability matching means that pension funds will have the opportunity to invest risk-free from now on. This is an interesting observation in itself, because in the past this economically risk-free solution was not considered as such due to the actuarial valuation of the liabilities.
The liabilities are thus the new starting point and constitute a natural benchmark for the investment policy. However, this does not mean that full liability matching is automatically the optimal strategy. In fact classic Markowitz theory will revisit pension funds: invest part of the portfolio risk-free and allocate a part to the optimal risky portfolio. The risk attitude of the stakeholders should determine the weights of investments which are risk-free, and those which do involve risk. Investments for which extra returns are expected (compared to the liabilities) can be used for indexing pensions.
However, the regulator requires that there are financial reserves for the extra risks involved with these investments. The size of these buffers is such that there is only a 2.5% probability that a situation of underfunding will occur in any one given year. If a pension fund’s reserve is below the required buffer, a recovery plan has to be submitted explaining how the solvency requirement can be met within 15 years. This can in fact be seen as a kind of ALM study, which indicates how certain measures can be used to improve the fund’s financial health.
The current allocation of an ‘average’ pension fund is indicated by the red dot in figure 2. The risk with respect to the liabilities requires a buffer of 30%. A logical first step is to extend the average duration of the bonds in the portfolio from 5 years to the duration of the liabilities, say 15 years. The consequence is that the expected return rises somewhat in normal market circumstances, because interest on long bonds is usually higher than on bonds with a shorter duration(2). This is indicated by the arrow from the red to the orange dot that points to the upper left-hand corner. The risk falls as well, because the interest rate sensitivity of the investments will be brought more into line with that of the liabilities. So if interest rates decrease by 1%, the loss will only be 7.5% instead of the 12.5% that is shown in figure 1. By using interest rate swaps the interest rate risk can be eliminated totally, without having to invest all assets in bonds. The swap market can also be used to hedge liabilities with a longer duration than the longest available bonds.
Increased pressure from IFRS and decreasing risk appetite may lead to a situation in which both employers and employees prefer a less risky investment policy, even if this is expected to require a higher average premium. This is indicated by the arrow from the orange to the yellow dot, which points to the lower left-hand corner, meaning that both risk and expected return will decrease. However, contrary to the duration extension this is a trade-off between risk and return, on which social partners will have to reach an agreement. Therefore, we indicate this step with a question mark in figure 2. If the current risk level remains acceptable, this step can be ignored.
The consequence of making the liabilities the starting point for composing an investment mix is that deviations against the full matching strategy are only useful when they generate extra return. Since the liabilities are denominated in euros and currency returns are expected to be zero, it would be logical to hedge the risk of foreign-currency investments with derivative instruments.
For most pension funds, the risky part of the investment portfolio currently predominantly consists of equities. However, interest in alternative investment categories is growing. Several large Dutch pension funds such as ABP and PGGM have already invested a substantial part (up to 20%) of their portfolios in alternatives. Examples of alternative investments include direct or indirect real estate and corporate bonds, but also hedge funds, private equity or commodities. The disadvantage of an investment policy which only includes equities is that when stock markets decline the pension fund's financial position will decrease accordingly. The diversifying nature of alternative investments can mitigate this effect significantly. The result is that the required financial buffer for the solvency test will turn out to be lower than when the fund only responds to one risk factor. This is depicted by the last dot in figure 2. The choice of portfolio composition does require estimates of the expected long-term return for all investment categories, while for many alternative investments only a limited data history is available. Nevertheless, even with conservative assumptions for the expected returns the added value of alternatives is still present. This diversification over various sources of return contributes to the ambition to index pensions to inflation.
A number of the recommendations mentioned above cannot be attributed directly to the new regulations, because these already applied to a certain extent under the current guidelines. The first example is the trade-off between risk and return to be made by social partners. A lower risk leads to a higher expected pension premium. Secondly, all risks for which one is not rewarded by means of a positive expected return should be avoided. Furthermore, pension funds could diversify using alternative investments in order to improve their risk/return profile.
However, a radical change in risk awareness will occur as a result of the introduction of market valuation of the liabilities. The fully matched solution will replace the strategic ALM portfolio as the natural starting point of the thinking process. For many pension funds the nominal liabilities will act as a hard benchmark (‘guaranteed value'). Besides the position compared to the guaranteed level, employees and retired persons will also have to keep an eye on the amount of resources that are available for indexation of pensions. In its 2003 annual report ABP indicates that the funding ratio based on a fully indexed pension declined from 86% to 85% while the actuarial funding ratio rose from 103% to 109%. These indexed liabilities thus function as a soft benchmark.
With this new risk approach it would also be logical to involve the risks run by the sponsor company in the pension fund's investment policy. In situations with high market returns, it will generally be easier for the company to pay a slightly higher premium, while during a bear market that same company will likely face disappointing profits which makes higher premiums, or even extra deposits in the pension fund, very undesirable. In general, the more sensitive a company’s profits are to the stock market, the less attractive a certain equity position in the pension fund becomes. The transition to IFRS will stimulate this manner of integrated risk management.
Summarising, we expect that pension fund allocations will change in three respects, as illustrated in figure 3. The added value of diversification over more sources of return already applied before the new headlines were introduced, but they are reinforced by the required solvency buffers that should be held. The increased interest rate risk as a result of market valuation can best be tackled by bringing the duration of the investments into line with that of the liabilities. Finally the increased attention to short-term risk is an incentive to invest more rather than less in bonds than currently is the case.
Roel Knol of Robeco Institutional Business Development.
Tel: +31 10 224 7171
(1) These calculations are based on assumptions, such as the volatility of equities and bonds. We assume, for instance, that the duration of equity markets is zero and thus equity returns have a zero correlation with bonds. From an empirical point of view, periods of positive correlations alternate periods with negative correlations. We do assume that interest rate movements on 5-year bonds are positively correlated with 15 year bonds.
(2) Here we assume that interest rate levels are not predictable. When interest rate levels are predictable and are expected to rise, the expected return on bonds with a longer life to maturity can be lower than that on bonds with a shorter life.