Developments in Pension Fund Risk
Paul Sweeting
Fidelity Investments International
Introduction
In this chapter, I look at the interaction between pension schemes and their sponsoring employers, the risks that pension schemes create for their sponsors, and the ways in which these risks can be mitigated, if not removed. Much of the analysis here relates to funded pension schemes, where assets to meet pension payments are held apart from the sponsoring employer. Many of the issues regarding risks, however, can also be applied to unfunded book reserve schemes, where the pension scheme liabilities remain wholly on the corporate balance sheet.
Scheme and Sponsor—the Economic Relationship
Funded pension schemes have historically been regarded as completely distinct from the firms that sponsor them. While in many countries this view reflects the legal situation, it does not adequately describe the economic reality of the relationship between scheme and sponsor, at least not for defined benefit (DB) pension schemes. For a defined contribution (DC) pension scheme, the employer pays a contribution that is a fixed proportion of each employee’s salary. These contributions are then invested and the proceeds of these investments, whatever their size, are used to purchase pensions at retirement. The only liability the employer has, in economic terms, is to pay those fixed contributions. The assets of a DC pension scheme are always equal to the liabilities, assuming that contributions are paid.
The situation for a DB scheme is more complicated. Members accrue pension benefits and contributions are paid into a pension fund. These contributions are invested in a range of assets and the proceeds of these assets are used to pay the benefits. If, however, the assets are insufficient to meet the benefits, then the employer may well be expected to pay the difference; and if the assets exceed the value of the benefits, then the employer may well offset that surplus against contributions paid in respect of future pension accrual. In this way, the scheme and sponsor remain economically linked. Even if the sponsor decides to wind up a pension scheme, giving the assets and liabilities to an insurance company, the sponsor will often remain liable for the excess of the pension scheme liabilities over the assets.
This view of the relationship between pension schemes and their sponsors is not new—Graham and Dodd (1951) suggested that when analyzing securities, pension scheme liabilities should be treated as a debt of the firm and pension scheme investments should be treated as assets. This being the case, it is important that anyone responsible for managing the risk of a firm is aware of the contribution to that risk of the pension scheme. Furthermore, it is in the interests of the shareholders for the risk contributed by the pension scheme to be well managed.
The first risks considered relate to the amount of money going into a pension scheme. Then I consider the two categories of risk that affect the ongoing health of the pension scheme: financial and demographic.
Funding Risk
From a tax point of view, many pension schemes should be fully funded. Firms get tax relief on contributions to pension schemes and on the interest payment on debt that they issue. Borrowing to fund a pension scheme rather than running a deficit accelerates the tax relief received, potentially leading to a significant present value tax saving, as shown by Tepper and Affleck (1974).
This is, of course, a highly stylized interpretation of tax, capital structure, and pension schemes. In particular, it ignores the rate of interest that the employer must pay on issued debt. The tax saving will be insufficient to justify additional funding if the net-of-tax interest rate paid by the firm is greater than the valuation rate of interest used to value the pension scheme liabilities. In this situation, a pension scheme deficit becomes a cheap source of additional funding for a firm. Another way of looking at this is to consider the firm as having a put option on any pension scheme deficit, an insight highlighted by Sharpe (1976). The more likely the firm is to default, the more valuable this put option becomes, as noted by Francis and Reiter (1987).
One issue that can arise if a pension scheme is well funded is that a surplus might develop that is then trapped in the scheme and unavailable to the sponsor. This is a particular problem if the surplus is then used to fund discretionary benefits. In many cases, the surplus can be offset against ongoing contributions if there are active members still accruing benefits. It leads to the question, however, of whether an asset allocation that can lead to a volatile level of funding should even be in place. This question is covered in the next section.
A less obvious issue that can arise in relation to pension scheme funding relates to the resources under the control of company management. Jensen (1986) points out that the smaller these resources are, the more rigorous the monitoring by capital markets can be, as new projects then need to be funded externally. This suggests that management might be prone to use a pension scheme as a “slush fund” to over- and underfund in such a way as best meets its needs. It also suggests that shareholders should prefer pension schemes to have full and stable funding so that the funds available to managers are controlled. This again suggests a role for asset allocation.
Financial Risks
Investment Mismatch
Graham and Dodd’s analysis suggests that pension scheme assets and liabilities should be regarded as part of the capital structure of the firm. Bagehot, aka Treynor (1972), formalizes this with the idea of the pensions-augmented balance sheet. This is a concept whereby the values of pension assets and liabilities are added to firm assets and liabilities, with the value of corporate equity being the balancing item.
Knowing that the pension scheme can be regarded as part of the capital structure of the firm is all well and good, but what is there to say about the impact that this has on the management of the pension scheme? After all, Modigliani and Miller (1958) showed us that capital structure does not matter and, in the first order, there is no reason why it should matter more if the capital relates to the pension scheme rather than traditional debt and equity in issue.
Investment mismatch does, though, matter in the second order. I define investment mismatch as investment in assets with an expected risk premium over matching bonds—equities would be the most common example. There are, in fact, many reasons for removing as much investment risk from the pension scheme as possible by investing in matching, debt-like assets. Not only does such an approach reduce the risk of the emergence of a surplus or deficit, but it also means that less monitoring of the assets is needed. Just as importantly, it means that investors in the company’s shares are getting returns based on the company’s core business rather than a mixture of the core business and assets in which the pension scheme is invested. The arguments that Jensen (1986) makes about free cash flow also apply to pension scheme asset allocation. An even stronger argument relates to the tax implications of mismatching, however, as demonstrated by Black (1980) and Tepper (1981). It is more tax-efficient for a firm to raise capital through debt finance than through equity finance, as interest on debt is tax-deductible whereas dividend payments to shareholders are not. Given that there is no tax advantage to having a particular asset allocation in a pension scheme, it makes sense for a firm to obtain leverage directly through debt issuance rather than indirectly through pension scheme equity investment. Again, the proviso regarding the net-of-tax cost of borrowing relative to the interest rate used to value the liabilities applies, but for firms that can borrow cheaply, there are strong arguments in favor of reducing the risk in their pension schemes and increasing it in their traditional capital structure.
The degree to which the market recognizes such tax arbitrage must, of course, be considered—there is no point in trying to carry out such arbitrage if the market ignores the pension scheme and recognizes only an increase traditional leverage. Research by Jin et al. (2006), however, does suggest that firms’ equity returns do reflect the asset allocation of pension schemes, for the US at least.
Interest Rate Mismatch
There are many firms for whom it is sensible to take investment risks in their pension schemes. Even if it is not the best approach from a theoretical standpoint, investment mismatch risk can always be justified with reference to the expected risk premium of the risky investments. Whether this claim is valid or not, the same claim cannot be made in relation to interest rate mismatch risk. This risk occurs when the interest rate sensitivity of the debt assets held does not match the interest rate sensitivity of the liabilities. This is fine if a conscious decision has been made to take on this risk, either because of a strong view on interest rates or due to the integration of the pension scheme’s interest rate profile into the firm’s overall interest rate profile, but this is very rarely the case. In particular, the magnitude of the interest rate risk taken within the pension scheme is often far larger than would be allowed if it were a firm-level interest rate decision. Because there is rarely an expected, long-term, out-performance expectation from interest rate mismatch (unlike investment risk mismatch), the risk being taken can be defined as an unrewarded risk.
There are two broad types of interest rate mismatch. The first is a fixed/real mismatch. This occurs when fixed-interest investments are held in order to match inflation-linked pension scheme liabilities. Inflation linking is often applied to increases to pensions in payment and, in the run-up to retirement, including through salary inflation. If this type of mismatch occurs, the pension scheme is exposed to the risk that an expected increase in inflation will cause the value of fixed-interest investments to fall while leaving the value of the liabilities unaffected (assuming that the increased inflation expectation increases nominal yields but leaves real yields unchanged). This risk is best dealt with by treating inflation-linked and nominal liabilities separately when possible for the purposes of matching.
The second type of interest rate mismatch is more subtle. This occurs when a change in interest rates affects the assets and liabilities to different degrees. It is predominantly this type of risk that is addressed in the following analysis.
The framework within which interest rate risk can be addressed is known as liability-driven investment (LDI). Superficially, this can sound like asset liability modeling (ALM), but the key to LDI is that it focuses on the interest rate sensitivity of the liabilities. LDI does not necessarily require complete investment in risk-free matching assets; however, what it does do is distinguish between rewarded and unrewarded risk and attempts to manage unrewarded interest rate risk. There are a number of ways in which this can be done.
The most basic—and obvious—approach is to cash flow match the assets and liabilities. Thus, if Exhibit 11.1 shows the cash flows of a portfolio of liabilities worth £100m (using a discount rate of 5 percent per annum)—which are shown as negative as they are payments out of the pension scheme—then Exhibit 11.2 shows the profile of these liabilities together with the cash flows that would be required from the assets for exact cash flow matching.
Exhibit 11.1 Cash flows for a portfolio of liabilities
Source: Author’s own.
Exhibit 11.2 Cash flows for matched portfolios of assets and liabilities
Source: Author’s own.
One way to get cash flows is to hold bonds that produce coupons and redemption payments at the times that you need them to pay liabilities. If bonds are used, however, then it is unlikely that such a good match between assets and liabilities could be obtained, particularly as the term of the liabilities increases. Exhibit 11.3 gives a more likely outcome, together with the annual cash flow mismatch. Here, matching has been possible for the first 10 years, but reinvestment risk has had to be borne for longer periods—the “lumpiness” of bond payments means that it is difficult to get a good match.
Exhibit 11.3 Cash flows for a short-term matching strategy
Source: Author’s own.
The cash flow mismatches in Exhibit 11.3 grow to be significant. In present value terms, however, as shown in Exhibit 11.4, the economic impact of these mismatches is less severe.
Exhibit 11.4 Present values for a short-term matching strategy
Source: Author’s own.
A more fundamental issue with bond cash flow matching is that it removes the opportunity for returns above those provided by the bonds—risk reduction is tied to low returns. If a higher expected rate of return is achieved, then the compromise is that not all of the pension scheme cash flows are being matched. One solution to this is to use swaps instead.
For non-increasing pensions, or pensions that increase at a fixed rate, the approach would be to enter into an interest rate swap agreement where the fixed payments received would match the fixed cash flows payable. The floating payments due on the swap could be met by holding cash. Absolute return funds or a balanced portfolio could be held, however, if there was a desire to take more risk with the investments.
For pension payments with some linkage to inflation, the interest rate swaps would be combined with inflation swaps. Here, part of the fixed payment received with the interest rate swap would be used to pay the fixed leg of an inflation swap; in return, inflation would be received.
As attractive as the swap approach sounds, there are still several issues. First, cash flows change from year to year. This means that the swap portfolio would need to be rebalanced significantly on a regular basis, the rebalancing not being free due to the spread on swaps. In any case, exact cash flow matching is very difficult, if not impossible, due to the uncertainty surrounding a whole host of financial and demographic factors. Furthermore, providing a proportion of the assets held is sufficiently liquid, the assets can be sold to meet pension payments. This suggests that exact cash flow matching is unnecessary and a more approximate approach might produce just the same sort of reduction in risk.
The most approximate approach to managing risk is to give the assets the same monetary duration as the liabilities. Monetary duration (sometimes referred to as dollar duration) gives the change in value of the assets or liabilities for a 1 percent change in the interest rate.
This means that only assets that have a price that is defined by the rate of interest, such as bonds or swaps, are treated as having duration. Although equities might have a relationship with both long- and short-term interest rates, such a relationship is not instantaneous and does not define the price of an equity. Equalizing the monetary durations can most easily be achieved for fixed liabilities by adding a long or short position in a long-bond future to the existing asset allocation. The size of the position needs to be such that the change in the value of assets for a 1 percent change in the interest rate along the length of the yield curve is the same as the change in the liabilities.
A similar approach to monetary duration matching is immunization. To immunize a set of liabilities, three conditions need to hold:
- the present value of the assets must equal the present value of the liabilities;
- the duration of the assets must equal the duration of the liabilities; and
- the convexity of the assets must exceed the convexity of the liabilities.
In this context, duration refers to the percentage change in assets or liabilities for a 1 percent change in the valuation rate of interest rather than the monetary change. Again, this means that only assets that have a price that is defined by the rate of interest, such as bonds or swaps, are treated as having duration. Having the present value of assets and liabilities equal means that when these durations are equal, so are the monetary durations.
The convexity adds a new dimension, however. Whereas duration effectively looks at the slope of the price/interest rate relationship (the rate of change of the former with respect to the latter), convexity looks at the curve of the slope (the rate of change of the rate of change). This is important because the relationship between price and interest rate is not linear.
These criteria mean that, for a small change in interest rate along the yield curve, the assets will either increase in value by more than the liabilities or fall in value by less.
Bonds can be used for immunization, but the scarcity of very long-dated bonds makes this difficult for many pension schemes. In addition, if the value of assets is less than the value of liabilities, then the first condition of immunization cannot be met. The inability to separate risk reduction from return generation is also an issue when it comes to using bonds for immunization.
Bond futures can also be used, but these generally only give exposure to a single point in the yield curve. The tools that are really needed are, again, swaps. If swaps are used, then the present value and duration of the assets is based on the fixed and inflation legs of the interest rate and inflation swaps. As swaps with a greater notional amount than the underlying assets can be held, the equal present value and duration criteria can be met. In theory, only two swaps are needed, one with a very short duration and one with a very long duration. If the floating leg of the swaps and the assets held to fund the floating legs are ignored, then the profile of the present values of the cash flows is as shown in Exhibit 11.5.
Exhibit 11.5 Present values for a duration matching strategy
Source: Author’s own.
Such approaches are really only suitable if the interest rate risks in the portfolio are swamped by other investment risks—for example, if the majority of the pension scheme assets are invested in equities. This is because the implicit assumption that interest rates will change by the same amount along the entire yield curve is strong, to say the least. Equalizing monetary durations can still leave the assets and liabilities exposed to changes in very different parts of the yield curve and immunization gives only a small improvement.
One way of correcting for these shortcomings is to equalize the convexities of the assets and liabilities rather than to ensure that the convexity of the former is greater than that of the latter. To match convexities, an additional bond (or swap) is needed. The cash flow present values for a swap strategy are shown in Exhibit 11.6.
Exhibit 11.6 Present values for a convexity matching strategy
Source: Author’s own.
This principle can be extended to higher moments, but another approach is to look at key rate durations. Here, assets are held such that the change in value of the assets is equal to the change in value of the liabilities for a 1 percent change in the interest rate at each maturity while holding the interest rate at the other maturities constant. If the maturities investigated are at five-year intervals, a portfolio of six or so bonds or swaps can give very good interest rate protection. Such a strategy is shown in Exhibit 11.7.
Exhibit 11.7 Present values for a six-bond matching strategy
Source: Author’s own.
Various combinations of these approaches are also used in practice. A common variant is horizon cash flow matching. This involves matching the cash flow for, say, the first 10 years, often with bonds or combinations of bonds and swaps, and then matching only the duration of the longer-term liabilities, or even taking the view that over a long time horizon, the return on risky assets is likely to exceed the return on risk-free investments. If the latter approach is used, there are caveats. Stock markets can underperform other markets for long periods of time, sometimes decades. It is also not certain that the sponsoring employer will be around for the length of time needed to give a reasonable degree of certainty about the outperformance of the risky assets. Indeed, the circumstances that might result in poorly performing stock markets might also be the circumstances that result in the insolvency of the employer. Even if the sponsor survives, accounting rules generally also require the marking-to-market of pension scheme assets and liabilities, so short-term volatility will arise. This volatility will not be helped by the fact that the liabilities against which no matching assets are held are those which have the longest duration and hence the greatest volatility.
The solutions discussed so far imply that the assets will be managed directly against the liabilities, so the liabilities themselves will be the benchmark. In practice, this is not necessarily practical or desirable. Some changes in the liabilities cannot easily be hedged and instruments with the same maturity as the longer-dated liabilities do not exist. This means that no matter how good an LDI strategy is, there will always be some mismatch between the assets and the liabilities. Given that a good benchmark should be investible, it makes more sense to construct a benchmark that is a good proxy for the pension scheme liabilities rather than to treat the liabilities themselves as the benchmark. Any proxy benchmark used should also be verifiable by a party other than that carrying out the management of the LDI strategy—external auditing of performance against the benchmark is important. Interest rate and inflation swap indices are available in some countries and, if the processes described previously are used to find the combination of cash and swaps that gives the lowest predicted tracking error relative to the pension scheme liabilities, then this can be used as a proxy benchmark. This means that the manager of the LDI strategy can be measured effectively for each part of the process, with allowance being made for the inability to completely remove all risks.
A sample attribution analysis is given in Exhibit 11.8. Here, a multi-asset portfolio is being managed against a multi-asset benchmark which, together with a swap overlay, has been designed relative to a benchmark which is a proxy benchmark for the liabilities. The choice of multi-asset benchmark and the styles of management within this benchmark would have been chosen so as to best allocate the risk budget between market risk (the multi-asset benchmark decision) and active risk (the allocation away from the benchmark and within asset classes), allowing for the fact that part of the risk budget must be allocated to the difference between the liabilities and the proxy benchmark that cannot be hedged. Attribution analysis within the multi-asset portfolio, giving the return due to allocations away from the benchmark and that due to management within each asset class, can be carried out in the same way as would be done for any multi-asset portfolio.
Exhibit 11.8 LDI attribution analysis
|
Component |
Code |
Calculated as |
|
Return an assets |
A |
|
|
Return on multi-asset portfolio |
B |
|
|
Return on multi-asset benchmark |
C |
|
|
Return on proxy benchmark |
D |
|
|
Return on liablities |
E |
|
|
Performance of assets relative to liabilities |
F |
=A−E |
|
Unhedgable return difference |
G |
=D−E |
|
Benchmark decision value added* |
H |
=A−B+ C−B |
|
Mutli-asset value added |
I |
=B−C |
|
*includes the efficiency of the swap overlay |
||
Source: Author’s own.
The preceding analysis describes a segregated approach to LDI, where swap overlays and asset allocation strategies are carried out on a scheme-by-scheme basis. If swaps are involved, however, then expertise is required in carrying out an LDI strategy and smaller pension schemes might not feel that they have sufficient expertise either to design or to implement the swap overlay. Furthermore, unless the swaps traded have notional sizes in the tens of millions of sterling, then the cost of trading and of subsequently rebalancing the swaps may make such an approach uneconomical. The alternative is to use a pooled LDI fund. These are essentially collateralized swaps, available in both fixed and inflation-linked “flavors,” in which units can be bought as they can for any other fund. There are two ways in which a pooled LDI fund can work. The first is for each fund to have a fixed maturity date, for funds with initial maturities of, say, 10, 20, and 30 years to be available, and for the maturities of the funds to decrease over time, eventually leading to a redemption payment. Such funds could be designated as, say, “2016,” “2026,” and “2036” funds, and they are essentially the same as zero-coupon bonds. Another approach is to hold the maturities constant, so a fund initially launched as a 10-year fund would be kept as a 10-year fund by periodically resetting the swaps within each fund. Such funds would never mature. In both cases, the funds could be used to implement a key rate duration approach to LDI. The reducing maturity funds would be most suitable for a scheme where no new benefits were being accrued, so the maturity of the scheme would be reducing year-for-year with the assets; the fixed maturity funds would be most suitable for a scheme where benefits continued to be accrued, so the maturity of the liabilities would be more stable from year to year.
If the assets underlying the swaps in the pooled LDI funds are low-risk, cash-type assets, then the result of implementing an LDI strategy is potentially to reduce expected return as well as unrewarded risk. There are two potential solutions to this. The first is for the pooled LDI funds to have assets other than cash-type assets underlying the swaps. These might be absolute return funds with return targets in excess of cash, or a range of long-only assets such as equities. While this approach solves the problem, it means that a large number of funds needs to be launched in order to satisfy all potential clients. The alternative is to use levered LDI funds. The standard, pooled LDI fund is unlevered, so £10m notional of swaps has £10m of cash-type assets supporting it within the fund. However, a levered fund—for example, with 2x leverage—would have the same £10m notional of swaps, but would hold only £5m of collateral to support it. This would mean that a pension scheme with £10m of liabilities could hedge all of its interest rate risk by investing in only £5m of levered LDI funds, and could invest the remaining £5m of assets in a range of return-generating assets. It would also mean that a scheme in deficit, without sufficient assets to cover its liabilities, would be able to use a combination of levered and unlevered funds to fully hedge the interest rate risk. The various structures are shown in Exhibit 11.9. One complication with levered LDI funds is that the degree of leverage changes as the value of the swaps changes, something that does not happen with the unlevered funds. This means that, in practice, a 2x levered fund might operate with leverage between 1.5x and 2.5x.
Source: Author’s own.
Other Financial Risks
It is worth spending some time discussing increases to pensions in payment and in deferment in some more detail. While in many cases these increases are straightforward, in some cases there are caps and floors that make exact hedging using basic nominal and inflation swaps difficult, if not impossible. Investment banks will generally be able to provide tailored solutions for very large pension schemes, but what solutions are there for everyone else?
A first approximation can be obtained by looking at the expected value of the increase. For example, if the expected long-run rate of inflation is 3 percent per annum and pension increases are provided in line with inflation subject to a cap of 5 percent and a floor of 0 percent, then treating the pension increases as purely inflation-linked gives a sensible answer. Similarly, if the same increases are provided and the expected long-run rate of inflation is 7 percent, then treat the pension increases as being fixed at 5 percent and hedge accordingly.
This approach is less satisfactory when the expected rate of inflation is close to the cap or floor. In this case, a better approach is to use stochastic modeling and optimization to arrive at the mix of nominal and inflation-linked swaps or securities that best matches the level of increase under investigation.
Another key financial risk arising from the pension scheme that cannot be hedged is wage inflation. Whenever pay increases are provided, accrued pension is increased and this results in an increase in pension scheme liabilities. Although wage inflation is a risk, however, it is also something that is under the control of the scheme sponsor. The main form of risk mitigation here is for the sponsor to recognize not just the direct effect of wage inflation on corporate profits, but also the indirect effect resulting from the impact on pension scheme liabilities.
It is also worth noting that although nominal wage inflation has been volatile, real wage inflation—wage inflation net of price inflation—has been much more stable. The reason for this is clear: for the vast majority, regular pay increases are decided with reference to the underlying rate of price inflation. This means that treating pay inflation as price inflation plus a fixed amount can still result in a close match between assets and liabilities.
Demographic Risks
The main demographic risk for a pension scheme is mortality—or, more correctly, longevity—risk. This is the risk that pension scheme members will, on average, live longer than expected, meaning that current valuations will underestimate the true liability. However, mortality risk is actually two risks: one risk is that the projections are wrong; however, potentially just as serious a risk is that the projections are correct but that the pension scheme is unlucky. This can be thought of as binomial mortality risk.
Binomial Mortality Risk
Binomial mortality risk occurs when the underlying mortality projections for the population or the group as a whole are correct, but random fluctuations are an issue. This is particularly a problem for smaller schemes and the problem decreases as the scheme size increases, although it remains a major risk until the number of scheme members is well into the thousands. The key to managing binomial mortality risk is simply to increase the number of lives—this risk is a diversifiable risk. The number of lives can be increased by taking action such as merging separate pension schemes within a company, a group, or even an industry, or by giving all assets and liabilities to an insurance company. The final approach will clearly solve many more problems than just binomial mortality risk, but at a price.
Mortality Projection Risk
Improving longevity does not necessarily cause a problem for pension schemes—providing actuarial valuations make allowance for this improvement. Allowance is almost always made for some improvement in longevity, but unfortunately this allowance has often proved to be inadequate. Will this improvement continue? It is difficult to say; much of the improvement so far has concentrated on curing illnesses, but the key question in the long run is whether old age itself will remain incurable. If so, then there is a finite limit to the possible improvements in longevity; if not, then immortality will provide a challenge for pension schemes, to say the least. As Buettner (2002) points out, this distinction forms the two main schools of thought when it comes to long-term longevity projection. Whatever the answer is, the fact remains that mortality projection risk is not a diversifiable risk: no matter how large the pension scheme, the risk remains.
Pension schemes have been aware of longevity risk for many years, but the magnitudes of recent revisions to mortality projections have brought the issue into sharper focus. Perhaps this is why capital market solutions for these issues have been explored only relatively recently, first appearing in articles by Blake and Burrows (2001), Milevsky and Promislow (2001) and others.
Blake and Burrows (2001) are among the first to discuss capital market solutions with their idea of survivor bonds. The bonds in their paper are amortizing securities, the payments of which depend on the proportion of a reference population still surviving at the date of payment of each coupon. Survivor bonds differ from annuities in that the payments from the bond are not made to the reference population. The idea initially sounds appealing. When BNP Paribas looked at launching a Blake-and-Burrows-style longevity bond with the European Investment Bank, however, the reception could best be described as lukewarm and the bond was withdrawn without being launched. Blake et al. (2006) give a number of reasons for the bond’s lack of success, the three main reasons being:
- the duration of the bond was too short for many pension schemes;
- the bond tied the reduction in risk to a nominal bond return; and
- the basis risk between the proposed reference population (English and Welsh male lives) and age (65 only) and the pension scheme populations and ranges of ages would have made the risk premium charged seem unattractive.
It is possible, however, that the main reason was that it was a bond rather than some other instrument.
Dowd (2001) suggests a number of alternatives to survivor bonds, among them survivor swaps. Here, the population-dependent payments form the floating leg of the swap, with the fixed or preset leg being the expected amount of those payments assessed at the time that the swap is transacted. This solves a key problem with the longevity bond in that it separates risk reduction from return generation.
Given that survivor swaps would be over-the-counter (OTC) products, larger pension schemes (and life assurance companies wanting to hedge mortality rather than longevity) would be able to formulate swap contracts that best suited them. This could then lead to a reasonable trade in survivor swaps and, as trade grew, a degree of standardization of swap contracts. This might also lead to more success in the development of longevity bonds as, according to Blake et al. (2006), one of the constraints on the size of the BNP/EIB issue (€540m) was the lack of capacity for the swap contained within the bond. At present, however, survivor swaps are rare and survivor bonds are nonexistent.
Other Demographic Risks
Other demographic risks exist, such as underestimating the proportion of married members, or overestimating the number of members that will leave active membership before retirement, but these are difficult to hedge and are best dealt with through scheme design. For example, adopting a “career average revalued earnings” (CARE) structure rather than a final salary structure means that each year of benefit accrued is based on the salary in that year and then revalued to retirement in line with price inflation rather than the individual member’s wage inflation. This effectively treats each year’s benefit as being a slice of deferred pension, removing the risk of underestimating early withdrawal. It is also worth noting that such demographic risks are generally small in comparison with longevity risk.
Conclusion
Although there are many risks inherent in defined benefit pension schemes, the tools necessary to hedge these risks are improving all the time. Even now, the vast majority of these risks can be hedged, and there is no question that many should be.
It is also worth mentioning that the preceding analysis deals only with the risks relating to accrued benefits, but the future service cost of pension schemes should also be a major source of concern. Real and nominal interest rates are lower now than they have been for many years and this, coupled with increasing longevity, has led to the cost of ongoing pension accrual rising sharply. Fortunately, this is a risk that can be dealt with relatively easily, through scheme redesign or even the cessation of future accrual. Many schemes do try to deal with this risk by closure of the pension scheme to new entrants, but this action must be carried out with regard to the level and type of staff turnover. If staff turnover is low, then the number of active members will decrease very slowly and no significant change in cost will be seen for some time. Even if staff turnover is relatively high, there still might be little effect if the turnover is confined to a particular part of a firm. If most of the turnover is within a group of younger employees in a particular department, then there is still likely to be a core of older (and more expensive) employees in the rest of the firm accruing benefits for the foreseeable future.