There is an aspect of the letter from TPR to Work and Pensions Select Committee which we believe warrants unpacking:
“Schemes that used LDI strategies to lessen the impact of falling or volatile interest rates over the past decade have been impacted less by previous market shocks such as the 2008 financial crisis and the recent pandemic”.
The references here to the pandemic and great financial crisis are rather odd. The 2018 – 2020 pandemic period was one of remarkable stability. As Chart 1 shows, the yield curve remained positively shaped, and the only period of rising rates was from the late summer of 2019 until the end of December (The dates on all charts should read as month ends not the 1st of the month – a quirk of Excel).
Chart 1: Gilt yields over the pandemic
Over this period, with rates falling, schemes would have been receivers of collateral, not payers. The 2019 increase in rates was modest, only 35-37 basis points, a price decline of 7% over four months.
The reference to the Great Financial Crisis is equally odd. Chart 2 shows the evolution of the fifteen and twenty-year gilts together with the six-month rate, the curve is negatively sloped until the crisis hit in earnest in September 2008. At that point, aggressive Bank of England intervention in the money market saw short rates plummet and the curve then reverts to being positively sloped.
Chart 2: Gilt yields during the Great Financial Crisis
Gilt yields also fell marginally as a result of the intervention as is most easily seen in Chart 3.
Chart 3: Gilt yields 2007-2009
At the time of the Great Financial Crisis there was rather little take-up of LDI by pension schemes – perhaps a total of £200 billion, and those who did take it up tended not to engage in aggressive leverage. Leverage and borrowing came later with sustained low short-term rates and a sustained belief that all rates would be lower for longer tending to lower forever.
Throughout this period, schemes were using the collateral received to buy more gilts and other securities and the response from TPR recognises that obliquely:
“Over the past 20 years, as long-term interest rates fell to historically low levels and through market events seen during the COVID-19 pandemic, LDI has meant that the assets in DB schemes increased.”
During this period of declining rates, the same self-reinforcing spiral of LDI was at work; yields went down, collateral was received and invested in more bonds, so bond prices went up again. The spiral was in a sense positive, but it was largely unrecognised, and what was missed by TPR was the self-reinforcing nature of these interactions – there were significant concentrations of risk being built up.
How influential were pension schemes on the index-linked gilt market?
It is possible to offer empirical evidence in support of the proposition that the influence of pension schemes was rather substantial.
If we look to the Index Linked Gilt (ILG) market, we may observe that the high to low yields on the 1/8th % 2068 has ranged from RPI minus 213 basis points last December to RPI plus 212 basis points in the recent crisis – a decline in value of 85%.
What is important to note here is that The Bank of England’s Quantitative Easing programme in relation to the GFC or the Covid response did not operate in the ILG market, so while the Bank waded into the market and offered to buy long-dated bonds , they did not initially offer to buy linkers. Pension funds are the dominant players in the linker market.
So, if we now compare the forty-year conventional gilt yield reported by the Bank of England, we see that has experienced a range of just 214 basis points over the same period, half that of the ILGs.
The effect of LDI strategies was to drive prices up beyond the levels which might otherwise prevail when gilt rates are declining, and as we have seen recently to drive them down as much when gilt rates are rising.
The volatility induced by LDI can be expected to raise the cost of new government finance and making gilts appear riskier can only serve to lower foreign participation in auctions.
Hedging discount rates
The discount rate used in determining the present value of pension fund liabilities is assumed to be anchored to the gilt yield. That is the rate we would expect to earn by buying the gilts with cash. There is no legislative reason why, for UK pension scheme valuation and funding purposes it need be.[i] Crucially, in relation to LDI strategies, neither repo nor interest rate swaps offer the gilt yield as a pay-off. In other words, neither instrument is a sound hedge of the interest rate sensitivity of liabilities. In the jargon, this means there is a ‘basis risk’ and it can be very substantial.
This is most easily seen with repo. The gilt is owned, and effectively continues to be during the life of the repo, as proceeds from the security repo-ed continue to accrue to the seller, the pension fund. The scheme earns the return of the gilt and must pay the cost of borrowing, so its pay-off is the gilt yield minus the cost of borrowing for the term of the repo. Borrowing very rarely comes at zero cost.
Interest rate swaps have a similar problem. The pension fund receives a fixed rate (the gilt yield) but must pay out a short rate (usually the six-month SONIA or previously LIBOR) at intervals over the life of the swap. A swap is priced by equating the present value of the fixed payments leg with the present value of the expected future short period rates. Fair value pricing, that is no initial cost, is achieved by adding or subtracting a spread to the six-monthly payments.
The problem is that estimating the future path of these short-term rates is no simple matter. Using the present implied six-month forward rate curve over the term of the swap is notoriously unreliable. More sophisticated models are needed, and these involve consideration of the riskiness and the volatility of the future short rate curve. There is room for considerable disagreement over this detail – particularly when markets are unsettled. The Great Financial Crisis saw many disputes over derivatives valuations and collateral calls resulting from them.
It is worth noting that there has been a significant improvement in that many derivatives are now centrally cleared, and it is the Clearing House which sets the levels of maintenance margin collateral calls. While this is usually done on close of business prices, the Clearing House may make calls intra-day when times are turbulent.
It would be informative to hear about the Clearing House’s experience in the recent crisis.
Now the central point of this elementary exposition is that the holder of the fixed interest leg (the fixed receiver[ii]) of the interest rate swap) is not provided with a perfect hedge of a gilt. Its pay-off is that of a gilt less the short rate paid away every six months. The interest swap is not a bet on the level of gilt yields, matching changes in which are the ambition of the LDI hedge, but rather they are bets on the shape of the yield curve.
In Chart 2 above we see gilt yields fall a little after Central Bank intervention and short rates fall off a cliff, the nirvana of the holders of the fixed legs of interest rate swaps. It is only when the entire curve moves up or down uniformly, a level shift, that gilt yield changes are closely approximated, and that is rather more the exception than the rule.
The presence of the short-term cost and the lower net return from gilts is that higher nominal or notional amounts are needed to cover the so-called interest rate exposure of the liabilities.
We illustrate this most simply with a repo transaction.
If the interest rate differential is 2%, say between 3% short (the money market interest rate cost) and 5% long (the yield on the gilt), then 250% of gilt exposure is needed to fully cover the liability exposure. To replicate the 5% yield on the gilt, we need to leverage the 2% net margin of the repo 2.5 times. If the differential is 1%, (say 4% and 5%), 500% nominal or notional and so we need to leverage the 1% net margin 5 times.
If we think of this in terms of a buy to let property investor, the investor puts up 25% of the purchase price, borrows 75% of the purchase price from the bank (3 x leverage) at a floating rate of interest and rents the property at a fixed rent which should cover the costs and repay some of the borrowing. If the floating interest rate goes up… The comparison is not perfect, but it may help to draw out that “leverage” is economically the same as borrowing. What is critical to this point is that pension schemes are only allowed to borrow for temporary liquidity purposes.[iii]
With fair pricing of an interest rate swap, the situation is even more strained. The swap by construction has no net margin in favour of the long gilt. We may be receiving the fixed rate, but we are paying away its present value equivalent in short rates. The LDI manager has managed to reduce the long-term nature of a pension fund to a series of short-term payment obligations.
When this so-called ‘hedging’ is concentrated in the derivatives rather the cash market, we have a derivatives market far larger than the cash; the harm that can arise from that situation was painfully demonstrated by the role of mortgage derivatives in the Great Financial Crisis of 2007-9.
There are also widespread reports that repo was being used to “juice up” returns, to purchase or allow the purchase of riskier assets with higher expected returns.
Well, that strategy is even more plainly ‘ultra vires’, that is beyond a trustee’s legal power or authority, on any analysis as we have set out elsewhere. We understand that in many cases corporate bonds are being bought in search of the additional returns from the credit risk spread they carry. In all too many instances this is regarded as a simple fixed risk premium, when in fact it is in part compensation for the liquidity risk of the security rather than solely credit risk. These spreads are both highly variable and cyclical.
In the absence of any further leverage of gilts by the fund, the returns which will in fact pay pensions come only in small part from gilts in the case of repo, the net margin. Where corporate bonds have been bought using the proceeds of a gilt repo, those required returns, to cover the net margin shortfall, are coming (or may not be coming) from the yields on those corporate bonds.
In the case where interest rate swaps have been used, all of the returns must come from other assets held (as the starting position in an interest rate swap is that both legs have the same value the swap is “at the money”).
This brings into question the use of a gilt-based discount rate. There is the quite separate question of investing in a fixed interest gilt market where rates have been artificially suppressed by the Bank of England in pursuit of financial stability through the use of quantitative easing.
It is not evident from the utterances and blogs that the Pensions Regulator understands any of this or would like scheme funds to be aware of it. Supervision by blog is rather an interesting concept.
Accounting for pension liabilities
The central problem with current accounting practice is that it introduces the volatility of market movements in yield into the valuations of pension scheme liabilities in the accounts of the sponsoring employer. This is what motivated the use of Liability Driven Investments strategies – and though some are using these now as purely speculative arrangements, this still motivates most market participants. Some defenders of LDI strategies have asserted that they have “helped to stabilise pension funding over the past two decades”[iv] . But then the turkey being fattened for Christmas also thought that interest rates would stay low forever.[v]
It is necessary to emphasise that, from a funding scheme perspective, the legislative requirement in Regulation 5 of the Occupational Pension Scheme (Scheme Funding) Regulations does not require a discount rate to be derived from gilt yields and that the Pensions Regulator is required, in relation to the exercise of its powers in relation to scheme funding, to minimise any adverse impact on the sustainable growth of an employer.[vi]
This takes us back to the fundamental question of how should the discount rates be determined?
The desirable properties of a discount rate are that the rate should be fundamentally invariant and specific to the scheme, and this discount rate should change only when the experience or expectations of the determinants of the projected ultimate benefits change.
There is such a rate, which we have called the Contractual Accrual Rate (CAR). It is that rate, the internal rate of return, which equates the contribution received with the projected benefits ultimately payable under that award. It is in other words, the rate of return promised by the sponsor employer on their contribution and that of a member.
In setting the contribution at inception of an award for a year of service, the trustees and sponsor may take account of the expected returns available from markets but would not be required to. The principal attraction of this rate is that it will only change with revisions to the projected benefits; though these may come from differences between the assumptions made and subsequent experience or changes to those assumptions, they are usually small in magnitude. Unlike market-based discount rates, changes to the CAR are reflections of real changes to the benefits to be paid after retirement. The volatility of the CAR would be an order of magnitude lower than that of market yields.
For a scheme the CAR is simply the weighted average across members and time.
It would be feasible to adopt this approach under existing regulations. The Occupational Pension Schemes (Scheme Funding) Regulations 2005 require for the calculation of technical provisions:
5 4) The principles to be followed under paragraph (3) are
(b) the rates of interest used to discount future payments of benefits must be chosen prudently, taking into account either or both—
(i) the yield on assets held by the scheme to fund future benefits and the anticipated future investment returns, and
(ii) the market redemption yields on government or other high-quality bonds;[Emphasis Added]
If the assumptions driving the values of projected benefits are prudently chosen, then the discount rate, the CAR, derived must logically also be prudent. Moreover, prudently chosen assumptions will lower the need to revise projections in the light of minor variations of those experienced projections, further reducing the variability of the CAR.
The key to acceptance of this rate under these regulations would be recognition of the sponsor covenant as an asset of the scheme, which clearly it is.
This may be explicitly documented, for example as a call option (or strip of them) for any shortfall of return from the investment portfolio relative to the CAR value. In fact, such an option need not be called upon by trustees. Its value in-the-money and uncalled is not less than its value exercised.[vii] This would leave flexibility for sponsor and trustees to collaborate for the benefit of all.
The Pensions Regulator refers often and in many contexts to the sponsor covenant but fails anywhere to recognise it explicitly as an asset of the scheme – an elementary failing.
This approach changes the role of the pension fund from being to provide the wherewithal to pay pensions when due, to defraying the costs of the sponsor employer in providing the pensions. This is a subtle but fundamental change in how the scheme is viewed.
This approach will clearly remove any and all incentives for LDI with respect to discount rate hedging. The real risks of a scheme which remain may be hedged using assets, such as index-linked securities for inflation, and insurance policies for longevity, and indeed the trustees may even, from time to time, find it reasonable to buy default risk insurance cover on the sponsor.
Perhaps the most pleasing aspect of using this approach is that the funding ratio of a scheme as at a valuation date would go up or down in line with the change in the market value of the assets of the scheme, and with rising gilt yields and falling asset prices, have fallen and in that concurring with common-sense and reality. It would also remind us that what matters to a pension scheme is the ability to pay benefits as and when they fall due and that with long term investment time horizons, the funding course can be gradually corrected over time.[viii] All too many are taking false comfort in the fact that under the current practices the present value of liabilities has fallen, and scheme funding ratios improved, even though the scheme has suffered material losses on assets held.
As for change in the accounting standards, a quotation from Baroness Bowles of Berkhamsted on accounting standards is appropriate.
“It is negligent if government, BEIS, the Treasury, regulators and the Bank will not get their heads around issues in accounting standards. It is no defence to say the accounting standards are independent; they are a closed shop defended by their acolytes. We are not all bamboozled, but those with power must take the blinkers off.”
What about accountability?
On the question of accountability, we note that the Pensions Regulator has stated:
“It is right that our expectations of those accountable for delivering the retirement outcomes that savers expect are clear and properly enforced.”
However, we find no conformity with clarity and proper enforcement in their actions and accountability with respect to Liability Driven Investment, with its use of borrowing and derivatives. Indeed, the Pensions Regulator’s guidance was positively supporting the use of leverage by DB schemes. It begs the question whether Regulator understood what it was regulating or whether this was regulatory capture by those on secondment to the Regulator.[ix]
By way of ending, we would like to share two highly relevant insights offered to us by a former central banker, and current pension scheme trustee, John Nugee[x].
“Two lessons emerge from all this:
1) What you measure is very important as people will manage to the measure, not measure the management. If you measure the quarterly coverage ratio (assets vs liabilities) people will try to manage the quarterly coverage ratio and in particular manage its volatility. In short, the measure ceases to be an indicator and starts being a target – and in the process loses its usefulness as an indicator. Or, as a very similar case was once summarised to me, “the measure ceases to be a window through which management see what is really going on and becomes merely a mirror in which they see what they themselves have done”.
2) What a financial system needs above all for systemic resilience is diversity of business models. A system with many diverse business models will be more resilient than a system with uniformity of business models, even if the one standard business model adopted in the latter is superior to many (perhaps even all) the models in the former. The problem with LDI is not that it makes any one pension fund safer or less safe, but that if everyone employs it, it makes the market overall less safe, because when it moves, it all moves in the same direction.
None of our regulators have paid enough attention to the first of these, few would even understand the second.”
[i] The legislative requirement is that the rate is prudent and can be based on the yield on the scheme’s assets and anticipated future investment returns or the market redemption yields on government or other high-quality bonds (see the Occupational Pension Schemes (Investment) Regulations 2005, Regulation 5. We know that the Bank of England’s quantitative easing programme has been rigging the fixed interest gilt market to reduce yields: https://www.bankofengland.co.uk/monetary-policy/quantitative-easing and the expected investor behaviour by the Bank of England is “Here’s an example. Say we buy £1 million of government bonds from a pension fund. In place of those bonds, the pension fund now has £1 million in cash.
Rather than hold on to that cash, it will normally invest it in other financial assets, such as shares, that give it a higher return.
In turn, that tends to push up on the value of shares, making households and businesses holding those shares wealthier. That makes them likely to spend more, boosting economic activity.”
The conventional wisdom of using the “risk free rate” derived from fixed interest gilts to determine the expected return (or risk premium) on “riskier” assets does not work when the fixed interest gilt market is being rigged by the Bank of England using quantitative easing for financial stability purposes.
[ii] A fixed receiver interest rate swap is one is which the pension fund receives a fixed, say a gilt yield, in each period from the counterparty, and pays a short-term rate, say six-month SONIA plus or minus some fixed spread, to the counterparty.
[iii] See Regulation 5 of the Occupational Pension Schemes (Investment) Regulations 2005, transposing Article 18(2) of the IORP I Directive (now consolidated into the IORP II Directive Article 19(3) ). These directives form part of retained EU legislation not withstanding Brexit and are required to be interpreted purposively (see eg Pfeiffer (C-397/01) at para 113: https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:62001CJ0397&from=EN).
[vi] Under the Pensions Act 2004, Section 5(1)(cza).
[vii] The value of an unexpired option, even when in the money, is strictly greater than its immediate value if exercised. This arises from the remaining optionality.
[viii] The incidence of employer insolvency over time and the protection of members’ benefits can be addressed much more cost effectively for the taxpayer with a higher level of protection from the Pension Protection Fund as well as allowing an investment strategy which allows for the long-term investment time horizons of pension schemes to invest for growth rather than to bury their “talents” in the ground: https://en.wikipedia.org/wiki/Parable_of_the_talents_or_minas
[ix] https://www.thepensionsregulator.gov.uk/en/about-us/work-for-us/secondments. We have also been informed that “The TPR has a lot of staff. A bunch of civil servants from the DWP and then a load of secondees from the DB industry and with the majority of staff recruited over time from the DB industry and going back to the DB industry after a number of years (a period working at TPR is considered by the DB industry to be very good for your CV generally). You will see that loads of consultants, lawyers etc highlight their time at TPR on their DB industry CVs.”