Written evidence from Professor Iain Clacher (Professor of Pensions and Finance at University of Leeds); Dr Con Keating (Chair at Bond Commission of the European Federation of Financial Analysts Societies) (LDI0056)

We thank you for your letter. Below is a response to the questions raised in that letter.

We have taken each question in turn and tried to answer these as fully as possible. The specific questions are highlighted in bold, with our responses underneath. If additional detail on any answer would be useful, we are more than happy to clarify or provide more information.

What is meant by ‘88% of non-pension liabilities of UK DB schemes was repo’?  (Q10)

At the end of Q3 2021, the ONS reported scheme liabilities of some £221 billion which were not pension-related liabilities, and of those liabilities 88% were attributed to outstanding repurchase agreements (repo).  Economically, a repo agreement is borrowing, a liability. The 88% number was published by ONS in the summary accompanying the release of that data and confirmed in email correspondence between Con Keating and Harry Butterworth of the ONS.

The note you promised on lessons to be learned from the relative behaviours of index-linked gilts and conventional gilts immediately before and immediately after the crisis (Q16).

Indeed, Con did promise to send over a note on the relative behaviour of index linked gilts and conventional gilts – we simply have not had time to write this up, and for that delay we apologise.  ILGs are far more volatile that would be justified by their fixed income characteristics alone and retail price inflation is a relatively slow-moving variable.

Our working hypothesis is that this is a result of the concentration of ILGs in the hands of DB pension funds – they own more than 80% of the outstanding ILGs in issuance and that creates an effective ‘free-float’ problem for the Gilt-Edged Market Makers (GEMMs) trading ILGs.[1]

Your estimate that roughly £500 bn is missing from the asset side of pension schemes and that this was not a paper loss but a real loss, because pension funds were selling assets to meet their collateral calls (Q22)

In our written evidence, we refer to the £500 billion shortfall specifically, as being from the beginning of this  year. See: Page 17 (towards the bottom) of our written submission in evidence.

We believe this was available to John Ralfe.

We have received several enquiries about the £500bn number, which is a large part of the reason we have yet to undertake the modelling of the relative behaviours of index-linked gilts and conventional gilts immediately before and immediately after the crisis.

We have also shared our figures and analysis below with David Fairs of the Pensions Regulator as he (David Fairs) had raised a question as to the £500 billion figure. We have yet to receive a response to our numbers, but we have received an acknowledgement email of receipt of our email to him.

Con has developed several models of our DB pension market over the past fifteen years, so we have been following these very closely.

The first the question to ask in all of this is how big the losses should have been had LDI worked perfectly. The PPF has year-end liabilities of £1,761.3 billion and at end September this was £1,076.0 billion. We do love the spurious accuracy of these numbers.

Daily movements in gilt yields in late September had the liability estimate gyrating wildly – there was one day in late September when that liability value varied by 17.5% (circa £181 billion).

Our own estimates of the year-end liabilities of £1.81 trillion and £1.04 trillion at end September 2022.

So, had everyone been perfectly hedged against interest rates, by PPF reckoning the losses would have been £685 billion (down 39%) and by ours £770 billion (down 43%). We would be happy to explain the sources of those differences, but those technicalities would distract from the thrust of things here.

Of course, not all schemes have engaged in LDI – TPR estimates that as being just 60% – we think that 75%-80% is a better practical estimate. We also cannot distinguish between leveraged and unleveraged LDI – no-one can.

We consider exposure to gilts as for example in a traditional 60/40 equity/gilt fund to be exposure to 40% coverage. It is also clear that many schemes have been very much over-hedged this year. They had considered linkers (ILGs) to have their arithmetic durations but that has never held true of ILGs. This is a product of the concentrated ownership of ILGs by pension funds. Over the past quarter, the empirical volatility of ILGs has been close to twice the theoretical. The effect of that on the cash ILG holdings of schemes alone has been to add £146 billion to scheme losses.

The PPF has schemes having assets of £1,842.7 billion at year-end (our figure £1.85 trillion is not materially different). But the assets at risk are larger than this as we have extensive use of repo. At year-end we think this totalled £210 billion (ONS Q3 figure £194.5 billion), so total assets at risk at year end were £2.05 trillion. We think that repo peaked at around £225 billion in March 2022.

The PPF reports an asset estimation of £1,450.5 billion in the September 7800 Index. But recognises that their roll-forward methodology can seriously mislead – they added the note below:

“Note for 30 September 2022 update The September 7800 update has been produced using our standard methodology, which is summarised in Note 4 on page 7 of this document. This means the approach will capture the impact of government bond yield increases on the liabilities but that the impact on assets will be less accurate. This is because we do not hold sufficient data to capture the impact of any structural changes to asset allocations made to meet collateral calls over the month nor do we have sufficient information to accurately capture changes in any leveraged LDI portfolios. This will also impact the accuracy of the reported proportion of schemes in deficit.”

This methodological note had not appeared prior to the gilt crisis but has been in every 7800 Index publication since.

Historically, our model has tracked PPF asset estimates quite closely, but this year since March we have seen a growing divergence between their figures and ours. By the end June of this year, the divergence was £110 billion. Their figure suggests assets have only declined by 21.3% but twenty-year duration gilts were down by 43% over that period, and gilts and fixed income were around 76% of scheme assets. This PPF figure contrasts sharply with reported results such as BT’s.

In terms of scale, if we consider the only losses to be on the gilt and fixed income portfolio, and all else were to remain unchanged, scheme assets would be £1,240 billion on PPF figures

To add some further elements, we think that the equity in pooled funds peaked at around £200 billion and is now around £100 billion, so a loss of asset value of £100 billion. The repo-based leverage we estimate to have lost about £67 billion, and their use to have declined to around £90 billion recently.

There will also be some confounding effects from pension payments and inflation effects, or deficit repair contributions (some of which have been accelerated in response to cash calls on sponsors) or indeed estimates of loans from parents but it will be possible to strip these effects out from scheme accounts in the future.

Net, net, our modelling suggests that at end September, scheme assets lay in the range £1,150 billion to £1,200 billion, not the PPF’s £1,450 figure. That is a loss of £650 billion to £700 billion in asset values since the beginning of the year, of which £500 billion is attributable to LDI activities.

As for sales of assets, some £250 - £300 billion to meet collateral calls before September, £125-£150 billion in September, and then we have sales to increase collateral buffers of some £200 billion.

It is also worth noting that the market value of all gilts outstanding fell from £2.6 trillion at year-end to £1.9 trillion at September-end – a fall of 27.3% or £731 billion.[2] Between their cash holdings and derivatives referenced to gilts, UK DB schemes were in fact exposed at the beginning of the year to something of the order of 1.6 times the outstanding gilt market, and their total losses to that date in the absence of any sales would have been of the order of £1 trillion.

On the Mini-Budget

The response to the Mini-budget appears to have been rather modest. However, the problem is that it cannot be separated precisely from the previous day’s Bank rate rise and QE disposal announcements.

On the Wednesday (BoE day) we saw a fall in conventional prices of just 2.5% in 15-year gilt prices and no meaningful selling of index linked securities (ILGs). Based on previous experience, such a decline would not have triggered a large-scale sell-off. Then on mini-budget day we saw a further fall in gilt prices of similar order (2.5%), but we also saw price declines in ILGs – slightly larger than those in conventionals.

We have heard of dealer reports that there was selling by pooled funds. As we understand it, this was not driven by liquidity or buffer concerns but rather by the need of the highly leveraged Luxembourg pooled funds to remain within their leverage limits. We understand that the Luxembourg authorities operate a maximum four-fold leverage rule. Based on fourfold leverage, which TPR reports is the average, the previous day’s 2.5% price decline would require the sale 12.5% (5 x 2.5%) of the original gilt holdings. Of course, this would be reduced by the amount of liquidity buffers held – say 5% (£50 billion) or possibly a bit higher. It would also be reduced by the amount that pooled funded managers might expect to receive in new subscription calls – past experience suggests this would have been high – say 90%. So pooled funds would have had about £7.5 billion to sell, and indeed the gilt market was soft ahead of the Mini-Budget. Both conventional and linkers were down, of the order of 0.5% to 1.5%.

These sales hit the market quite heavily coming as they did on the top of the prior day’s actions.

But on Budget Day, conventional 15-year gilts were only down by 2.5% and ILGs by a little more at day’s end. However, this decline saw a repetition of the call process for action on Friday morning. And of course, these funds had largely depleted their liquidity buffers, the sales pressure from this source was larger. We also heard of some self-managed funds which were considering how to replenish their buffers, but held off as they did not like the price quotes they had received.

The price action was larger on the Friday – a little over 3% in both conventionals and ILGs.

There were further statements over the weekend from the Prime Minister, Liz Truss, and the Chancellor of the Exchequer, Kwasi Kwarteng, that reinforced the impression that they would not be changing their policy position. So, on Monday morning, there were sellers lined up on all sides and the market simply gapped down. The spiral now saw distressed selling, no longer the relative politeness of the rebalancing. This distressed selling was from pooled funds, segregated managed accounts, and self-managed funds. The ILG market simply dissolved. The 1/8th % ILG 2068 on the Friday was 94.75% clean, late on the Monday this was 63.25% clean, and on the Tuesday, this was at 48% clean, and we suspect they may have traded at £43-ish. It is worth noting that in November/December of last year, the price of this bond was around £334.

The weakness of sterling on the Monday morning also meant that overseas buyers were completely absent – and they hold slightly over 30% of gilts in issuance, just a little less than the Bank of England.

On the Monday, we saw the banks beginning to have concerns – less with their credit exposures to pension funds under swaps,  and more with the effect that three days of large movements would have on their authorised regulatory capital models. Three exceptions in 3 days is very significant, and technically for most, which operate to a hundred day window, another two days would have invalidated those models.

On the Monday and then Tuesday there was distressed selling all around the gilt and corporate bond markets. By Tuesday there were stories (reported in the FT) of schemes even trying to unload some private placements and other illiquid assets. On Tuesday it was a question simply of can I raise some cash from this. The value of the asset being sold really did not enter consideration.

The banks had also responded and haircuts on repo had risen from 0%-0.5% before the crisis to almost arbitrary values – the FCA quoted them as having risen to the range 2.5% to 8% - and for many schemes there was simply no quote at all.

Schemes also started selling overseas assets on the Monday and Tuesday – bonds and equities were followed by private instruments.

The Bank of England was well-advised to intervene – it is close to impossible to estimate how bad things were before that, but it is clear that losses for pension funds were well above £300 billion from the episode beginning on the 21st. And they were down by around £500 billion since the beginning of the year. After the Bank intervention the crisis left schemes with losses of £125-£150 billion and some who had sold in distress in even worse positions.

Additional points and developments

It is also worth noting again that there have been real losses in defined benefit pension assets, but that the improvement in liabilities is only an improvement on paper. A point we should have made more clearly at the hearing.

Since giving evidence, it is also worth noting the new requirements for increased collateral buffers that is now being asked of pooled LDI funds in Ireland and Luxembourg, which are of the order of 300-400 basis points. Though there can be little doubt about the costs of this, we have reservations about its efficacy. Prior to and during the crisis LDI funds were being routinely stress-tested to 100 basis points, but a two-day move of just 37 basis points was sufficient to trigger the collateral call spiral. We hope to address this issue more fully along with our analysis of conventional and ILG volatility.

Finally, the first line of Dr Keating’s answer to Q20 refers to the Pensions Regulator. Should this be the Prudential Regulation Authority?

Con’s reference to the Pensions Regulator should be to the Prudential Regulation Authority. To use that awful American expression: Con misspoke.

We will endeavour to write-up our work on the relative performance of ILGs and conventionals, and we will share this with the Committee as soon as possible.

We do have two articles in preparation, which cover some of this and some other aspects of the crisis and LDI more generally. We will share those with you as soon as they are complete.

December 2022