The systemic risk embedded within ETFs

Author: Helen Thomas

Published: 14 June 2023

The most significant intervention in financial markets during the pandemic panic of 2020 came on March 23^rd when the Federal Reserve announced it would buy ETFs. Although it was only one measure buried deep amongst the alphabet soup of asset purchases and loan facilities, it deserves wider attention for arresting the vicious circle of financial instability that had taken hold. At the heart of an ETF lies a liquidity mismatch. With liquidity now leaving the markets once again, policymakers need to be prepared for the fallout from another dash for cash by gaining access to better data and performing more frequent checks of what lies under the hood of ETFs.

Photo by Anne Nygård on Unsplash

The popularity of ETFs is understandable. They are a cheap and easy way to access exposure to almost any sector in any asset class that an investor could desire. Concomitantly the size and scale of the industry has grown, roughly doubling every four years, with total global AUM hitting $10 trillion in 2021[1] which is roughly equivalent to 8% of assets under management in the global investment management industry[2]. The scale is relevant because ETFs are designed to be a liquid asset, traded frequently in the secondary market. As we have seen repeatedly in the past fifteen years, even deep liquid markets such as the S&P500 or US Treasuries can be vulnerable to a flash crash, precisely because so much money is invested in them. In the classic example of crowded cinema with only one exit, when someone yells “Fire” the devastation of the stampede is proportional to the number of bodies in the room.

The mechanism behind an ETF is deceptively simple. There is an ETF issuer, such as an asset manager, but the creation and redemption of units in the ETF is only permitted by an “Authorised Participant” (AP), often an investment bank. The price of the ETF tracks the NAV of the underlying assets because the AP is incentivised to create or redeem units by arbitraging the differential for profit.

Minimising tracking error is integral to the attractiveness of the product. An ETF needs to do what it says on the tin. Responsibility therefore falls onto the APs being able and willing to perform this arbitrage. But in the prospectus for an ETF there is no requirement for the APs to make markets under any and all circumstances. No bank or broker would sign up to such constraints. Profitable arbitrage isn’t always possible. The price of an asset can be volatile or market dysfunction can prevent buying or selling from even taking place. So the very mechanism which is designed to deliver the promise of an ETF – to ensure liquidity and the close tracking of an index – is also one that can disappear in a stressed situation. The ETF might be branded BlackRock or Vanguard but nobody knows who the AP might be on the day you want to redeem units in the product. Given we have lived through the failure of a number of large market making entities, from Lehman Brothers to MF Global, it is important for policymakers to gain transparency on where fragility in ETF market making might manifest.

The AP can also participate in the secondary market for ETFs where brokers and traders make the market. The price of an ETF only exists if the market makers determine it is profitable to do so. This is different to an index fund, where the fund manager purchases all the constituent parts for the investor. An AP isn’t required to own all the constituent parts[3]. They’re not even obligated to deliver all the securities to the investor. ETF prospectuses allow for a ‘representative basket of assets’ to be delivered, or cash. For a fixed income index that might contain 10,000 bonds (like the widely followed Bloomberg Aggregate Bond Index), it is far quicker, easier and cheaper to deliver only a proportion of these. In a normal market, that’s efficient. In a dysfunctional one, it is a catalyst for systemic risk.

A recently updated paper by academics at Columbia Business School, Chicago Booth and Wharton analysed the effect of such security selection. Or to put it more plainly, the hot potato problem. Nobody wants the bad bonds. “Steering a Ship in Illiquid Waters: Active Management of Passive Funds”[4] concludes that in a stressed situation, illiquidity in a bond begets illiquidity. They looked at the March 2020 market dysfunction and concluded: ‘Given their balance sheet constraints, APs became reluctant to purchase even more of the same bonds in their role as market makers. Bonds present in redemption baskets thus lost their most natural buyers. When its own market makers do not want to buy it, a security can become quite illiquid’.

The liquidity mismatch inherent in an ETF becomes an illiquidity accelerant when market dysfunction takes hold. There are two ways in which this propagates systemic risk:

1. Such “price discovery” is in fact the discovery of panic. Speed is not always a virtue for a financial system, as we have seen with the latest slew of bank runs. If the price of an asset suddenly collapses it can cause ripple effects that force the re-pricing of everything else in the financial system.
2. The disconnect causes market dysfunction such that there is no longer a price at which to sell the ETF. Liquidity evaporates and belief in the viability of the product itself comes under attack. The demand for cash triggers selling across the board.

Policymakers are aware of the threat. When Andrew Bailey was Chief Executive of the FCA he made a speech in April 2018 at the London Business School Annual Asset Management conference which warned “our job is to be watchful to the build-up of risks and vulnerabilities in the system… The secondary market liquidity of ETF shares is dependent on market makers and authorised participants… We know relatively little… about the capacity and willingness of APs to execute their function in stressed conditions where they may be under pressure to tighten their own risk limits. The result could be unexpectedly large discounts for ETF investors selling their holdings relative to the estimated value of the underlying assets, and possibly a need to suspend fund dealings… this could have the potential to amplify shocks to market conditions which are already under stress. We have no easy way of sizing this risk, but we cannot ignore its potential given the rapid growth of ETFs.…the global financial system is more resilient than it was… [the issue] lies in what level of continuous market liquidity conditions investors expect, and thus the liquidity transformation they accept”[5].

And yet we saw exactly these issues in March 2020. Fixed Income ETFs came under extreme selling pressure as the market re-priced the risk that bonds wouldn’t be repaid in a world that was shutting down. The ETF price traded far lower than that of its constituent bonds, not least because the bonds themselves weren’t trading. The discounts hit extremes. Chapter 3 of the IMF Financial Stability Report from October 2022 looked at “Asset Price Fragility at Times of Stress”[6] and noted in Box 3.1 that ‘ETF discounts reflect market liquidity costs… during the March 2020 stress episode, when liquidity conditions were poor, the discounts on ETFs increased dramatically, reaching more than 5 percent across all bond ETFs (up to 27 percent for high-yield bond ETFs and up to 13 percent for investment-grade bond ETFs)’.

This situation was only resolved because the Federal Reserve announced that it would step into the market and buy ETFs under its Section 13 (3) powers[7]. This created a massive short squeeze whereby the Fixed Income ETFs flipped from trading at a discount to NAV to a premium. You can see this more clearly in a chart of LQD, the BlackRock iShares iBoxx USD Investment Grade Corporate Bond ETF[8], which came top of the list of ETFs bought by the Federal Reserve in the first week of its ETF purchase programme[9]:

If it looks like a heartbeat, that’s because it is. The Fed effectively applied the defibrillator and shouted “Clear”, jump starting the market back into action and snapping it out of its downward spiral. Not that they needed to buy that many ETFs in practice. The sheer announcement itself did enough to turn the tide, with the LQD returning to its 9^th March 2020 levels within 14 trading days of the Fed’s decision, well before the purchases even took place from 12^th May onwards. In the end the Fed bought $8.56bn ETFs[10], never owning more than 5% of any single fixed income ETF, a far cry from the trillions of bonds purchased for its QE programme.

Those powers have now expired and it is not guaranteed that they would be revived, not least because of the political climate. Congress and the Presidency are under split control as we head into an ever-more acrimonious election cycle. The UK and Europe have never undertaken such asset purchases, suggesting that if they were required to step in, they would need time to set up the process from scratch. We saw during the response to the LDI debacle that the Bank of England moving from Gilt purchases to Index-Linked Gilt purchases took 18 days, at least in part because the Bank had never bought Linkers before. The Bank of England’s Executive Director for Markets, Andrew Hauser, warned in his speech of 3^rd March 2023 on “Lessons from the 2022 LDI Intervention” that “Without clearer guide rails… we are destined to proceed somewhat messily from crisis to crisis, building a framework by circumstance rather than by design”[11].

Before designing the guard rails, policymakers need to understand the mechanics of how ETFs operate. In particular, we must ask the question of why such a deeply desirable asset has become cheaper over time. The APs aren’t stepping up to create or redeem units from the goodness of their heart. They must be making money from the flows. Cheaper fees for the product suggests that more flows are required to make more profit.

Scale is always of use to a market maker as they take the spread between bid and offer. But ETFs offer something more. They are exposed to a wide range of assets but all contained within one price. This offers a cheap way to hedge other flows in a market maker’s book. If the market maker has a large position in Apple or Microsoft, they can offset this with prices on the S&P500 or Nasdaq ETF. They can even hedge out specific risk by trading in an ETF tracking one sector, such as a Semiconductor sub-index. They can trade correlation itself by dipping in and out of different ETFs.

Whether there is proprietary or market making risk has the opportunity to be blurred. They could, for example, purchase a chunk of Apple and Google stock as a way of apparently warehousing risk in order to show a good offer for the Nasdaq ETF.

They can even short-sell exposure to certain sectors or stocks by using the “Create to Lend” mechanism, whereby the AP creates units of the ETF precisely in order to lend them out as a short. They will be even more incentivised to do this when investor sentiment turns bearish. The AP can then charge a higher fee as investors become desperate to borrow stock in order to short-sell it. Take October 2008, when creations of SPY, the largest ETF tracking the S&P500, exploded even as the stock market collapsed, as ETF Trading Strategist Dave Lutz pointed out in “The Hidden Truth Behind ETF Inflows” for Forbes in June 2012[12]:

Many of these trading decisions are now taken by an algorithm, offering split-second risk management that has the potential to take place even before a human being, such as an institution’s risk manager, makes any decision.

None of this behaviour is inherently wrong. None of it is restricted to ETFs. Traders will dial up risk or hedge correlations as they see fit. It causes a problem only when the size of the positions is too large, too concentrated, too hidden and the product itself is vulnerable to a liquidity mismatch.

For policymakers charged with managing financial stability, a number of steps could be considered to improve oversight and gain an early warning before market dysfunction accelerates. Such policies would aim for more information, given at regular intervals, so that policymakers would have the precious commodity of time as they seek to stabilise the system. These policies could include the following:

1. Demanding that fund managers hold more capital so that they can meet redemption requests.
2. Checking at least annually an inventory from fund managers of where the assets are held.
3. Requiring fund managers to provide regular publication of bid-ask spreads in each ETF which the regulators could then compare with the bid-ask on the underlying assets.
4. Requiring each ETF to publish long run charts of premium/discount to NAV, including intra-day, to provide transparency over any incipient strain developing for the ETF.
5. Regular stress testing of liquidity demands on an ETF and a plan from the fund manager of how redemptions in a stressed scenario will be met, for example gating funds or slowing distributions.

These will place burdensome administrative requirements onto fund managers but a “race to the bottom” to avoid such constraints is exactly the sort of dynamic that raises correlation within the system and exacerbates redemption events. Hiding won’t work. Someone has to set the rules of the game or the game shouldn’t even be played.

Without transparency it is difficult for policymakers to monitor the risk. Supervisors will need more resources to scrutinise the information. It is welcome that the Bank of England’s FPC announced in March that it would undertake a “system-wide exploratory scenario” to look at the behaviour of banks and, more importantly, non-bank financial institutions “following a severe but plausible stress to financial markets”[13]. We don’t yet know what this will entail and given the growing tremors developing within the financial system we hope that they get on with it.

March 2020 was a wake-up call for policymakers to the systemic risk that lies within ETFs. Liquidity was pumped into financial markets but it is now being withdrawn. It has taken down a crypto exchange, a family office, several banks, the London Metal Exchange and the UK pension industry. We are in the midst of a rolling liquidity crisis. As ETFs are the embodiment of a liquidity mismatch they will be tested once again. Policymakers can take the time now so that they are prepared for the possible systemic risk ahead.