When I first read Toby Nangle’s July FT piece “the coming collateral call for UK pension funds” I kind of thought, well yeah, but … this is what is happening right now, or has already happened and in a pretty orderly way. It seemed a statement of what was happening at the time, as back then many LDI programs went through the orderly process of refreshing collateral as interest rates had risen. So far, so un-interesting. I didn’t have any idea how things would actually play out.

Author: Dan Mikulskis

Published: 19 October 2022

When I first read Toby Nangle’s July FT piece “the coming collateral call for UK pension funds” I kind of thought, well yeah, but … this is what is happening right now, or has already happened and in a pretty orderly way. It seemed a statement of what was happening at the time, as back then many LDI programs went through the orderly process of refreshing collateral as interest rates had risen. So far, so un-interesting. I didn’t have any idea how things would actually play out.

British Pound Sterling by Suman Bhaumik, Getty Images

British Pound Sterling by Suman Bhaumik, Getty Images

I want to tackle some of September’s events but you can’t do that without reviewing some of the bigger picture context and history, I’ve done that more thoroughly here, but in summary:

A quick history & purpose of LDI

LDI has its roots in accounting changes in late 90s that put pension deficits on balance sheets and made corporate management worry a lot about large and volatile deficits at the whims of bond markets.

The key driver of the deficit it turned out, wasn’t the value of assets going up and down, but the present value of the liabilities. The reason this was so variable was that it was discounted using the yields available on long-dated bonds, and these move around quite a lot.

The theoretical framework for defined benefit pension funds and liability driven investing was developed by actuaries in the late 90s[1], and as schemes closed and matured in the 00s the nature of the investment objective changed, becoming less about open ended growth investing and more defined relative to a set of now fairly fixed liabilties. This pointed toward the use of LDI.

LDI implemented with leverage was important to allow schemes to both hedge, and invest for returns at the same time, schemes couldn’t simply invest 100% in gilts to create the matching they wanted as they tended to have deficits, and this would have left corporate sponsors on the hook for much larger contributions.

So, LDI worked by investing, say 35% of a pension scheme’s actual assets into the strategy, but obtaining leverage, through derivatives like swaps and gilt repo to increase the matching coverage to the majority of their liabilities.

In the early days, the pitch was often being made by bankers to corporate management, concerned about the pension deficit on their balance sheet. LDI performed well during the financial crisis of 2008, and during the subsequent decade where it helped stabilise funding levels as yields crashed lower.

LDI grew to a £1.5trn+ asset class by 2019, from around £0.5tr in 2013.

Pooled funds (mutual funds) became a feature of the marketplace to allow smaller pension funds to invest and hedge the liabilities. Built-into the design of these mutual funds were specific permitted leverage levels and tolerance buffers to yield rises, as well as a process for refreshing the collateral in the fund if it fell too low. As with other asset classes larger schemes typically invested via segregated mandates (where a fund manager manages the assets directly on behalf of one pension scheme), these tended to operate with a little more flexibility.

Coming into 2022 the UK DB market was in very good health and has perhaps enjoyed its best ever funded position in the weeks and months before September 2022[2]. Rather than the deficits that had been a persistent feature for many years, lots of DB pension funds were actually beginning to see surpluses for the first time in many years. This meant that many DB pension schemes were turning their thoughts to the possibility of getting to a situation where they could carry out an insurance transaction to shift both their assets and liabilities to an insurance company.

At the same time LCP estimate that liabilities were c85% hedged to interest rates, using bonds and LDI[3] by early 2022.

What happens to LDI when interest rates rise?

Just as LDI paid out as rates fell to compensate increases in liability present values, the industry knew it would see negative returns when rates inevitably rose. As LDI is a hedge these losses would be exactly mirrored by falls in the present value of liabilities. But, because LDI is implemented with derivatives these losses still needed to be carefully planned for so as not to fully exhaust the collateral assets that were set aside and invested into LDI (which would result in counterparties withdrawing access to the derivatives and the schemes losing their hedge).

The level of collateral is crucial when using derivatives. This is because, if you own an asset outright you can continue owning it even if its value goes all the way to zero, at which point you have no asset, and also no exposures. But with derivatives, because you have exposure to assets greater than the value of your collateral, there is a point where movements in the underlying assets reduce your overall position. The collateral is there to satisfy your counterparties that you will make good on any losses incurred. At a certain point your counterparties will start to get worried about your ability to make good on the losses. If your collateral falls to zero then it is quite likely that your counterparties will close you out of the position, leaving you with no hedge. If the move in the underlying then reverses you’ll be at risk of suffering losses that you are no longer hedged against. In this instance the worry would be that you as the pension scheme would lose your hedge against interest rates, which might then fall, raising the present value of liabilities but with no corresponding offset on the assets.

Programs were designed to take rising interest rates into account, for example by having notifications in place as yields rose to alert pension funds to transfer more assets into their LDI program. A typical collateral buffer against yield rises coming in to 2022 might have been enough to cover a 150bps rise in long-dated gilt yields. This was usually calibrated using a 95 or 99%, 1-year event based on historical data. Prior to the ending of the Bank of England’s intervention on 14th October 2022, these buffer levels had been significantly increased across the market.

Typically pension schemes had planned in advance for yields rising and exhausting these collateral buffers. And by early summer 2022 the aggregate of the yield rises we saw between January and July had already caused one series of collateral calls early in the summer, putting into effect the plans that had already been made. In general, these progressed in an orderly manner.

What happens in these cases is that pension funds will move additional assets into their LDI program (whether that’s a pooled fund or a segregated mandate). In many situations the pension fund had arranged to run with readily-realisable assets such as cash or short-dated credit sitting alongside the LDI mandate that could be quickly realised and added to the LDI collateral pool. In other cases they might have requested a redemption from an equity market fund, or another growth fund that might have daily, or weekly dealing points.

What happened following the 23 September

Following the mini-budget of 23 September we saw a second very large increase in yields toward the end of September, but over a very short space of time – just 4 trading days. And here it was the sheer speed and scale of the moves caught funds by surprise.

The letter written by John Cunliffe of the Bank of England [4] puts the moves into historical context:

“As set out above, in the period immediately prior to Wednesday 28 September, the speed and scale of the moves in gilt yields was unprecedented. That period saw two daily increases in 30 year gilt yields of more than 35 basis points. The biggest daily increase before last week in the data that goes back to 2000 was 29 basis points. Measured over a four day period, the increase in 30 year gilt yields was more than twice as large as the largest move since 2000, which occurred during the ‘dash for cash’ in 2020. It was more than three times larger than any other historical move. Gilt market functioning was severely stretched, particularly at the long end of the curve (20 year maturities and above). These moves in gilt yields placed particular pressure on LDI funds.”

It was this dysfunctionality of the gilt market in the last week of September that placed such a strain on the mechanisms that pension funds and their managers had put in place to move around collateral in the event that yields rose. The danger was (explained well in the Bank’s letter) that some pension schemes would not be able to move assets quickly enough to meet the collateral call. This would result in some of their derivative exposures being closed out. But this would further push yields higher, putting yet more pension schemes under pressure, potentially creating a liquidation spiral that could have had the effect of “stopping out” a significant number of the LDI programs using leverage, as well as creating serious knock-on effects for the economy from higher gilt yields such as mortgage rates and financial conditions.

By “stopping out” I mean an LDI pooled fund reaching a sufficiently low level of underlying assets in the fund that all the counterparties would immediately close out the fund’s derivative positions, leaving the fund with little or no hedging unless further collateral could be placed immediately.

Lessons from the events

Some folks looking with the benefit of  hindsight will use words like “always” and “never” when it comes to anything involving derivatives. I prefer not to use those words when it comes to markets, I don’t think anything is inevitable. Yes, good ideas can be taken too far, and with derivatives it’s always a question of degree: 1% leverage makes no difference, 100x leverage on anything is a disaster waiting to happen. So what’s the right level? That’s the key question.

I think the industry needs to have a measure of humility and introspection about the necessary analysis of what happened and how things should work in the future. Everyone involved needs to recognise that the world has changed.

As is so often the case, many of the key questions to ask relate to our own behavioural tendencies.

Joe Wiggins provides a useful behavioural framework for thinking through a lot of this. Quoting some of Joe’s key points, which are all important principles worth remembering:

Both financial and mental models can break under stress

Predicting when things will happen is close to impossible

Markets are about the behaviour of other investors

Small sparks can lead to great damage in complex systems

In hindsight you can say we’re now in a different interest rate environment to the last 10 years. The availability heuristic says that we tend to base our judgements on situations that can be readily brought to mind and this is a key behavioural issue in a changing world.

When you’ve had a period of stability, a lot of the features of that get very familiar, investors become conditioned and it can get built quite deeply into how the industry operates, and even unconsciously make their way into the beliefs of those designing programs.

We don’t deal well with evaluating low probability, high-impact events. We’ll either tend to completely disregard them or attach far too much probability. This means that scenario tests of extreme things, even if they exist, are often not internalised properly or aren’t actionable in any useful way.

One reaction to an extreme event is that we should add it to our scenario tests in the future. That’s sensible enough. But how do scenarios actually get used and are they helpful? What if I come up with a scenario that shows my investments lose 50% of their value (which surely is possible) – what do I do? Hedge? Take less risk? Nothing? For single scenarios it’s not obvious, you can’t hedge every single one or you’ll just end up in a very weird place. That’s why you get pointed toward more probabilistic models, which can weigh scenarios appropriately.

But probabilistic models also only get you so far.

My view is that the fundamentals of LDI remain sound, it’s the particulars of the implementation details that need to be looked at. Gilts will remain very important to pension funds, one reason is that if you want to insure your liabilities, that’s the currency used in such transactions.

Derivative programs need to have margins of safety calibrated, and it’s common to use historical returns as a guide to the future for this purpose. Of course, as the well-known disclaimer goes, past performance isn’t a guide to the future, but it’s what we have, and it is generally a pretty decent guide to the broad parameters of the future, until it isn’t.

Taking the past as a guide has helped set expectations for things like how much rates might rise in a year, which is a key input into the design of an LDI program. Let’s look at what they were telling us. Coming into this year a commonly used risk model would have given a 1 year rise in yield at a 95% confidence level of 100bps or less, over a year. And to a 99% confidence this would have been about 150bps. We got that in just 3 days, on the back of 200bps already during the year. You can start to see how much of an outlier this was.

Bigger picture, DB pension funds in the UK in the majority of cases are in good health, while facing liquidity and asset allocation challenges in the immediate term. The better funding position is, if anything going to drive more rather than less demand for these sort of hedging strategies.

Trustees and their advisers will probably have to wrestle with a new trade off on risk and return. The ability to hedge 100% of liabilities while investing 25-30% in LDI, for so long a core tenet of DB investment strategy might well no longer be available.

At a system level it’s clear that LDI is systemically important and needs to be seen as such. The aggregate of individual risk management decisions can add up to something very important for the overall system, even in a market as large as UK government bonds. This is a matter for authorities and regulators, but individual pension funds will also want reassurance their programs are robust and not at risk of being caught up in further market dislocations.

The LDI community has been unexpectedly thrust into the spotlight and recent events seem far removed from the core purpose of buying government bond to reduce risk. There are a few important tasks at this time: Help clients take well-informed decisions in the short term, contribute responsibly to wider understanding and perception of the issue and finally to review operating practices openly and with humility. LDI has served pension funds well for the last 10 years, let’s make it work for the next decade.