Iain Clacher & Con Keating – authors of this blog

Many of the advocates of leveraged LDI have claimed that they were able to weather the March 2020 pandemic panic well.

We examine the price movements (using BoE data) over the month of March 2020 below as Figure 1. This shows daily returns on the 20-year gilt and the cumulative performance, a measure of net collateral movement, over the month.

It is clear that the net collateral position ahead of the market weakness was positive.

Figure 1: March 2020 Gilt Price Movements

The 6% decline on 17^th March, which might have triggered sales of gilts in the market the following day but did not, is in fact consistent with schemes holding buffers of 100 basis points, these would cover an 18% decline in price. It is notable, and fortunate, that the following day, the Bank of England announced that bank rate would fall to 0.1% and triggered the market recovery as shown by the red column.

However, the previous announcement on 11^th March, that the bank rate would be cut from 0.75% to 0.25%, had produced no significant effect on gilt prices.

By contrast, in September 2022, the Bank of England was increasing bank rate once again after having undertaken a series of previous increases that had commenced late 2021.

Perhaps the most important difference, however, was that the Bank of England was actively buying gilts as part of the QE programme in 2020. In March 2020, the Bank held £645 billion, which by June had increased to £745 billion and reached £895 billion in November. In this period, there was issuance of £15 billion and redemptions of £28 billion, for a net redemption of £13 billion.

By contrast with the earlier episode, for the September 2022 period, the prior month’s actions had tended to deplete liquidity reserves. The daily returns and cumulative collateral position for this period is shown as Figure 2. The first and most obvious point is that the cumulative collateral position was everywhere negative. Indeed, a buffer of 100 basis points held at the beginning of the month would have been exhausted by Friday 23^rd, had there been no replenishment of the buffer. The spiral was well under way at this point, and it took Bank of England intervention to break that spiral.

Figure 2: September 2020 Gilt Price Movements

Unfortunately, the published data on transactions volumes is weekly. Table 1 shows the reported volumes of client turnover in the periods leading to the disruptions in March 2020 and September 2022.

Table 1: Market Turnover (Client)

The first point to note is that trading turnover was generally lower in 2022 than 2020. It is also clear that in both periods the stress did lead to increases in traded  volumes. The average daily volumes in both cases are rather modest, £16 billion in 2020 and £13 billion in 2022. Turnover rises to around £25 billion per day in times of stress.

It is worth nothing that the difference in trading turnover cannot be explained by the size of the gilt market. In March 2020, the market value of all gilts was £2.16 trillion amd in September 2022 it was £1.9 trillion. This latter figure was down from £2.65 trillion at year-end 2021, and £2.2 trillion at end-June 2022. The difference in market value between year end and September is a good first order proxy for overall gilt losses in 2022 – £731 billion. There were, net, £9.8 billion gilts redeemed in this period. In the quarter ended in September, the net redemption of gilts was £16.2 billion. Moreover, over this 2022 period, the average maturity of the outstanding gilt stock has fallen from 17.5 years to 15 years.

Perhaps the greatest surprise is the low volumes of long-dated gilts offered to and purchased by the Bank of England under the intervention programme.

Table 2 Gilts Offered/Purchases 28^th – 30^th September 2022

It is notable that even after the size limit on purchases was increased and ILGs became eligible, the volumes of securities offered and purchased did not exceed the previous limit of £5 billion.

Table 3 Index Linked Gilts and Conventionals Offered/Purchases 11^th October -14^th October 2022

It is difficult not to conclude that these interventions were actually small relative to daily turnover. It appears that confidence had a rather larger effect than the amounts of liquidity supplied.

These daily average turnover volumes are actually rather small amounts, particularly so when the volume of derivatives written referencing gilts is considered.

We wonder if the GEMMs are constrained by their relatively light capitalisation from holding larger inventories and managing larger trading books, rather than by their regulatory capital requirements and the normal levels of demand for their services.

Since writing this article, the Bank of England has published its latest Financial Stability Review in which a diagram appears which indicates that the net sales of gilts by pension funds with maturities of 20 years or more, to 30^th September was approximately only £7.5 billion.  This was accompanied by the following text: “In total, DB pension schemes were responsible for around £14 billion of gilt sales between 23 September and 14 October, compared to around £23 billion of sales from LDI funds over the same period[1]. These gilt sales are smaller than the total margin and collateral calls faced by LDI funds and pension schemes in this period, which Bank staff estimate to be in excess of £70 billion[2].

This reflects the fact that LDI funds and pension schemes were also able to sell assets other than gilts and use existing cash buffers in order to meet these obligations.”

We believe that these estimates will be low biased and also paint a misleading picture. The analysis of GEMM turnover by maturity band throws some light on this.

Table 4; Client Turnover

With 20+ year gilts accounting for 76% of gilts outstanding of that 15+ year tenor, we might attribute £36 billion of the turnover increase to 20+ year gilts. In which case, an imbalance of £7.5 billion becomes plausible. However, perhaps the most important point to note is that far greater increases in turnover occurred in the 0 – 3-year range, which is almost surely attributable to LDI buffer sales. Indeed, the greatest increase in turnover occurred in the ten-year range, and that is generally considered the most liquid of gilt issues.

This raises a further issue. In Figures 1 & 2 above the experienced returns and their accumulations assume that collateral calls are met from other scheme assets. If this were the case there would, of course, have been no LDI gilt sales-driven price spiral, though the markets for the assets sold would very likely have suffered.

If collateral calls are met by sales of gilts, then the exposure of the scheme to gilts will have declined much more sharply than in the case where these calls are met from other resources. This is illustrated as figure 3, which shows the evolution of a 20-year gilt using the returns experienced during the crisis in the case where collateral calls are met from other assets and in the case in which all calls are met from  the sale of gilts. Obviously, the reality recently experienced lies between these boundaries. The decline, at 14 October, is 21.04% in the no sale case and in the sale scenario, the decline is 40.7% of which 20.35% are collateral calls. The most important aspect of this is the fall in the coverage of the discount rate effects in the gilt sales case, from the starting 100% to just 75% of the correct hedge amount on 14^th October.

Figure 3: Exposure to gilts – under sale and no sale conditions – 20-year duration

If we assume £1 trillion of exposure at the beginning of this period, as the Bank did, this would imply collateral calls over the period closer to £210 billion rather than the £70 billion they suggest. Even If we were instead to consider the scheme and its gilt portfolio to have a 15-year duration that would only fall to around £147 billion.

Unfortunately, we are most unlikely ever to be able to resolve that conundrum.

Lessons for the gilt market and pension funds

With the estimated notional volume of derivatives outstanding referencing gilts of £1.5 trillion notional, a 1% decline in price (5 basis points increase in yield) is sufficient to invoke collateral calls of £15 billion, and result in sales which swamp the gilt market, producing significant further downward pressure on prices.

This suggests that, as part of any long-term stabilisation of the gilt market, there needs to be restrictions on the aggregate notional amounts of derivatives referencing gilts.

Market participants have repeatedly stated that the size and speed of the increase in yields in late September surprised them, and some have claimed this was unpredictable. Far from being unpredictable, it is evident that relatively small increases in yield, or declines in prices may generate collateral calls which overwhelm normal gilt market functioning. The speed with which the process progresses once initiated is a consequence of the daily collateral margining  and the speed is assured by the structure of these contracts.

It is not at all obvious that increased buffers will suffice. Their adequacy will depend significantly on their rate of replenishment, that is the period over which sales are made and further cash raised, after payment of initial margin calls. The most important lesson to be taken from this episode is that liquidity has a cost, and the more liquidity that is required, the lower the assets that are at the disposal of the pension fund to invest for the long-term, and thus the lower the returns available to pension funds.

[1] DB pension schemes may include LDI segregated mandates. The estimates subject to the same caveats as set out in footnote (a) of Chart 5.2.

2022. a) The chart captures total net sales of gilts by LDI and pension funds that reported an outstanding open gilt repo or interest rate derivatives position between 22 September and 21 October 2022. LDI funds have been identified within the broader fund category by combining existing sectoral classifications and entity-level name screening. Gilt sales are calculated on a best-endeavours basis and expressed in market-value terms. Market values are estimated using the nominal amount transacted (available in MiFID data), intraday Bloomberg prices and Index Ratio data from the Debt Management Office.

[2] This figure is likely to be a lower bound for total margin and collateral calls as the data sources used to estimate such calls (Sterling Money Market Dataset and Trade Repository data) for outstanding repo and interest rate derivative positions may not capture the entire LDI and pension fund universe (eg for some trades the counterparty identifier is not available).