It is often said that “What gets measured gets managed” and it is simple to find many examples, inside and outside of the pensions arena, where this adage applies. However, in some contexts, especially when measuring uncertain financial liabilities and risks, it is not merely “what” is measured that matters but also “how” it is measured that is important. Applying different principles to measurement can significantly influence the actions taken with the potential for material financial ramifications.
One common measurement of risk for DB schemes is the one-year ‘Value at Risk’, or ‘VaR’. When first introduced, it was calculated in many different ways and supplemented by many different metrics, just like climate risk is today. The pensions industry soon settled on a simple approach that, at the time, was fit for purpose. Fundamentally, an approach that worked well when risk levels in DB schemes were high breaks down when risk levels are low; for small measurements the quality of our ruler is much more important.
As VaR has traditionally been employed in a pension scheme context to help inform choices between investment strategies, modelling the assets has been a primary consideration in its calculation – i.e. ensuring that the risk/reward characteristics of different asset classes are sufficiently taken into account. For a host of reasons, including convenience, capabilities and perceived objectivity, the liability modelling involved in calculating VaR has not necessarily had the attention it deserves for a while. Almost invariably today, the modelling of liabilities for calculating VaR implicitly assumes that liability discount rates always move commensurately with gilt yields. This is not a given; discount rates for the vast majority of schemes should usually be reviewed at each future valuation, if not more frequently, and may well respond differently in the light of significant market movements.
For many years, most schemes have been far from their ‘endgame’ state in terms of funding levels, investment strategies and time horizon, and as such refining risk measures hasn’t been a first order consideration. Closing significant deficits has been the key consideration for many schemes.
Many will have observed an improvement in their schemes’ financial positions over the last few years given significant tailwinds from rising yields and falling life expectancies. This has presented opportunities and appetite for much greater and/or earlier risk reduction than may have been expected, bringing endgames much more closely into sight. As such, there is now more to lose. Recent years have also given trustees and sponsors a series of vivid experiences of extreme risk scenarios playing out in practice and so risk management is very high on the agenda.
As a result, a lot of schemes that had historically been running funding and investment strategies with relatively low levels of hedging have now advanced their hedge ratios towards their maximum preferred position. However, in this context, familiarity with VaR measures could breed contempt. Those managing schemes may be continuing to lean on crude measures of risk without being aware of the limitations or potential pitfalls; with risk management decisions becoming more acute, broad-brush approaches to measurement could stray over the lines and lead to suboptimal decision making.
Regardless of whether a scheme is sure of its ultimate ‘endgame’, its destination is very unlikely to be to invest entirely in gilts. In contrast, the nature of the cashflow matching portfolios that the scheme may pursue, or that of the insurer with whom the scheme may transact, is likely to involve a material allocation to assets with credit risk exposure – e.g. corporate bonds.
If the VaR calculation is predicated on changes to liability discount rates being driven solely by movements in gilt yields, then a scheme with significant allocations to credit will perceive the scenario of credit spreads widening (i.e. the yield on corporate bonds increasing relative to the yield on gilts) as one which poses a material amount of the scheme’s total risk.
In effect, the risk measure would be indicating that losses in the scheme’s balance sheet would be created (i.e. a deficit worsen or a surplus reduce) if the market value of credit-based assets falls relative to gilts. However, in such situations:
Therefore, such market movements would not necessarily be a cause for concern unless they were perceived to be indicative of higher future losses through increased risk of issuer defaults or other needs to sell at depressed valuations. An antidote would be to recognise the contribution that higher credit spreads could have on the discount rate used to measure liabilities.
Well, back to what gets measured gets managed. Overstating VaR or omitting drivers of risk could lead to misguided risk management choices – in short, poor decision making causing suboptimal actions. Here are two examples.
For this example, let’s consider a scheme with a VaR measure that incorporates an allowance for longevity risk but that is calculated in two different ways: one with a gilts-based discount rate and another with a ‘dynamic’ discount rate that is also responsive to changes in credit spreads. In this situation, let’s suppose the potential impact on the scheme’s total risk of hedging 75% of its longevity risk is being examined.
| Measurement approach | VaR: No longevity swap | VaR: With longevity swap | Risk reduction |
|---|---|---|---|
| Gilts-based | 141 | 103 | 27% |
| Dynamic | 105 | 41 | 61% |
Under the first analysis, hedging 75% of the longevity risk appears to reduce total risk by around 27% whereas under the second analysis the total risk is reduced by 61%.
Because the gilts-based approach overstates the real impact of credit spreads widening, it masks how effective the longevity hedging action is; diluting its apparent effectiveness by more than a half.
This level of difference could weigh very heavily in discussions and on the conclusions reached on the best course of action for a scheme, influencing a trustee’s or sponsor’s risk management decision making with real long-term financial consequences. It presents a significant risk of schemes prioritising the reduction of risks that are artefacts of how risks are measured (e.g. reporting or measurement risks) and de-prioritising the reduction of the real risks (e.g. financial or demographic risks) that pose proper threats to the financial security of member benefits. This is not confined to the topic of discount rates; analogous issues can arise elsewhere, such as in the calculation of liability proxies for inflation hedging purposes[1].
In this second example, let’s assume that the scheme is considering reducing its target return in order to reduce the level of risk being run. We’ll compare the situation where the scheme measures VaR focusing only on investment risks, with an alternative where the VaR measure also makes allowance for longevity risk.
| Measurement approach | VaR: Gilts+1.0% pa target | VaR: Gilts+0.25% pa target | Risk reduction |
|---|---|---|---|
| Excluding longevity risk | 150 | 50 | 67% |
| Including longevity risk | 180 | 112 | 38% |
Under the first analysis, the trustees or sponsor might believe that two-thirds of the risk in running the scheme would be addressed through the investment de-risking action. However, taking longevity risk into account in the risk measure would have shown the de-risking manoeuvre to be about only a half as effective in reducing total risk. Whilst closer to reality, this example is still simplified; VaR does not measure all investment risks (e.g. the LDI crisis) and longevity is not the only non-investment risk.
The blind spots that can arise when the view of risks is not holistic could lead to a narrower set of risk management actions being considered, or the impact and efficiency of certain actions being overstated. This presents a danger of making inefficient and suboptimal risk management and capital allocation decisions.
Continuing to do things as they may have been done for a long time does not necessarily mean they remain fit for purpose, particularly when the surrounding context has changed so much. This is a reason why the Pensions Regulator’s General Code[2] requires governing bodies to “be prepared to monitor, challenge and review their risk evaluation process and outputs”.
Thoughtful integration of the funding and investment approach, with the scheme actuary and investment consultant working collaboratively, has the potential to lead to significant improvements in reporting, advice, understanding, decision making and risk management – and reduces the scope for trustees or sponsors to waste time and capital managing artificial ‘reporting risks’ at the expense of making great risk management decisions that attack the scheme’s true risks with the right force.
01
02
03
04
All schemes are about to contend with a new funding framework and will need to review and articulate their ultimate funding and investment strategy. Use this regime change as an opportunity to lift the lid on how your scheme measures its liabilities and risks in order to identify where there is room for improvement. If you do so, we think that you’ll see things more clearly and have much greater confidence that what gets measured for your scheme is indeed the right thing to manage.
