How Market Liquidity Disappeared During The COVID-19 Crisis

The Essence of the Piece

The Fed conducted a study on Treasury market liquidity during the height of the COVID-19 crisis. They found that trading algos used by principal trading firms didn’t provide liquidity during the crisis as many had hoped. Instead, market depth disappeared, and bid-ask spreads surged and became more volatile. The Fed also found a big deterioration in the usually very liquid FX markets. This adds to the growing case for understanding how bouts of illiquidity are one the biggest risks for investors.

The Structure of Treasury Markets

The US Treasury market is one of the most liquid markets for government securities in the world. Within that market, the ‘dealer-to-dealer’ (interdealer) segment, rather than the ‘dealer-to-client’ segment, is thought to be the most liquid. Moreover, most of the interdealer flows occur over electronic trading platforms such as BrokerTec and Dealerweb. And with the growth of such platforms, there’s also been a big increase in non-bank dealers or principal trading firms (PTFs). They employ high-speed automated trading strategies and often account for over half the trading volumes on the platforms. In this context, the Fed researchers analysed Treasury market conditions in the interdealer market over March.

From Markets Behaving Well to Behaving Badly

On any given day, bid/offer quotes are posted on a Central Limit Order Book (CLOB), which lists quotes from the highest bid/lowest offer to lowest bid/highest offer. These quotes are then matched with incoming orders from market participants who generally wish to execute a low-cost strategy.

The maximum volume of posted quotes is referred to as ‘market depth’ or the ‘stack’. In essence, these are all the active quoted prices at various volumes on the order book, which may or may not be matched with a transaction. As such, the total trade volume or transaction flow (the amount of initiated trades) is usually lower than the market depth. Indeed, most trades take place near the top of the order book, avoiding multiple levels (that is, there is no need to exhaust the full string of quotes).

On a normal day, then, there is the market depth, and trades are executed within that depth. Chart 1 shows the ask depth (green) and the bid depth (red). The executed flows (black dots) are typically done at the top of the book, that is at the best bid or ask price. Such conditions are associated with a stable spread between the best ask and bid quotes (green dots, lower panel).

Source: Dobrev and Meldrum (2020)

Before the COVID-19 crisis, it was thought that PTFs would act as a liquidity provider even as bank dealers withdrew from the market, however this time was different. Using a similar analysis as on a normal day, the authors found that in the middle of the crisis (13 March), liquidity conditions in the Treasury market deteriorated substantially. As Chart 2 shows, market depth declined sharply (smaller green and red areas), trade flows often occurred outside of the maximum depth, and bid-ask spreads were wider and more volatile.

Source: Dobrev and Meldrum (2020)

Why Did This Happen?

From this, the authors conclude that the order book replenishment must have been too slow to cope with the flow of orders. Large trade flows quickly used up all of the posted quotes near the top of the book (that is, at the best prices) and were executed at substantially worse prices for a period until the book was replenished and bid-ask spreads narrowed again.

The scaling down of Treasury market-making activity, at least in aggregate, by dealers, PTFs and other high-speed trading entities meant development in mid-March led to a significant widening of spreads. The rationale behind the response could be the considerably elevated economic uncertainty; the exceptionally high volatility of Treasury yields; the unusual correlations between Treasury yields and the prices of other assets; and possible concerns about the settlement risks associated with trades with counterparties experiencing financial strains.

But Trading Algorithms Performed Better

The ability of PTFs to respond to real-time trading conditions meant that those with machine algorithms were able to wait a short period for spreads to narrow when the order book had been replenished. Chart 3 shows the daily average quoted spreads (red) and volume-weighted average spreads (green) for 5-, 7-, 10- and 30-year Treasuries. While spreads for all tenors rose sharply in mid-March (especially the 30-year), those that weighted their execution by volume – as algos often do – saw less widening in spreads.

Source: Dobrev and Meldrum (2020)

Cash Bonds and FX Markets Had the Worst Liquidity

The paper also presents the change in spreads of other markets. Within Treasury markets, they found futures had better liquidity. Outside of Treasuries, they found that foreign exchange markets saw a notable increase in quoted spreads in mid-March, and the gap between equally weighted and volume-weighted spreads widened (Chart 4). There was also some increase in quoted spreads in the equity futures market, albeit to a lesser extent. In contrast, quoted spreads in the oil futures market remained relatively stable throughout March.

Source: Dobrev and Meldrum (2020)

The authors conclude from this that FX futures markets appear to indicate potentially more severe strains in liquidity provision than energy futures markets, equity futures markets, and Treasury futures markets. Furthermore, the fact that volume-weighted spreads were consistently lower than equally weighted spreads in FX markets is again indicative of the role that sophisticated execution strategies played during the COVID-19 market turmoil.

Bottom Line

The COVID-19 crisis saw an unprecedented hit to US treasury market liquidity. The oft-touted principle trading firms (PTFs) did not provide liquidity as they had in previous bouts of volatility. It wasn’t just Treasury markets that suffered but also the usually very liquid FX markets, which saw worse conditions than either equities or commodities.

Appendix 1: The Data

The authors analyse granular high-frequency data on orders placed and trades executed based on transaction and order book data from the BrokerTec platform. They focus on the following variables at millisecond frequency:

• Prices of buyer- and seller-initiated trades,
• Corresponding bid/ask quotes,
• The volume of posted bid/ask quotes (measure of market depth),
• Sum of buyer- and sell-initiated trades (measure of transaction flow),
• Quoted spreads.

For the following instruments:

• 5yr, 7yr, 10yr T-Note,
• 30yr T-Bond,
• 10yr, 30yr T-Bond Futures,
• 10yr, 30yr Ultra T-Bond Futures,
• EUR, GBP FX Futures,
• E-Mini S&P 500 Futures,
• WTI Crude Oil Futures.

Appendix 2: More Details

Treasury Market

There are two parts of the Treasury market: ‘dealer-to-client’ and ‘interdealer broker’. Simply put, in both parts of the market, end investors trade primarily on an over-the-counter basis with dealers. These dealers, typically banks, can buy securities at a quoted bid price, which are added as assets to their balance sheet/inventories. To avoid taking an unnecessary market position, dealers then look to sell securities at an ask/offer price to either a client or another bank.

Electronic Trading

Electronic trading broadly covers the following: trades conducted in systems such as electronic quote requests, electronic communications networks or dealer platforms; alternative electronic platforms such as dark pools; the quotation of prices or the dissemination of trade requests electronically; and settlement and reporting mechanisms that are electronic (see the Joint Staff Report on the U.S. Treasury Market from 15 October 2014 for more details on ETPs).

The ‘electronification’ of all these aspects of fixed income trading has been steadily increasing. For some fixed income securities, electronification has reached a level similar to that in equity and foreign exchange markets, but for other instruments the take-up is lagging.

To view the full paper, please click here