Could all-to-all trading improve liquidity in the Government of Canada bond market?

Introduction

Government bond markets are typically considered to be among the most-liquid fixed-income markets. Even so, liquidity conditions in government bond markets can deteriorate sharply when a sudden imbalance occurs between the demand for and supply of liquidity. In fact, in recent extreme episodes of market turmoil like the onset of the COVID‑19 crisis in March 2020 and the UK gilt crisis in September 2022, market liquidity deteriorated so much that central banks intervened by buying government bonds to stabilize and restore orderly functioning of these markets.

Some policy-makers, academics and financial market participants believe that one reason market liquidity of government bonds is vulnerable to periods of market turmoil is because the structure of these markets is centred around dealers. In this type of market structure, dealers use their own balance sheets or act as a broker to match offsetting transactions between their different clients. In periods of turmoil, however, client demands for liquidity can surge, and trading can become more one-sided. In other words, many clients trade bonds in the same direction, by, for example, buying government bonds in a flight to safety or selling government bonds in a dash for cash.

Dealers may charge a higher fee to intermediate such trades as compensation for the additional risks that they bear on their own balance sheets when market conditions are volatile.1 In extreme cases, dealers may stop intermediating completely, either due to constraints on their balance sheets or their inability to hedge risks. This could prevent transactions from occurring among clients who would otherwise be willing to transact with one another.

Adopting an all-to-all market structure is one of several proposals for improving the resilience of government bond markets during periods of turmoil (Duffie 2023). In this structure, market participants would be able to transact directly with each other on an electronic trading platform and therefore avoid any limits in dealers’ ability to intermediate. An all-to-all structure could also attract new market participants through increased transparency of executable and executed prices on all-to-all trading platforms. This increased transparency could improve the bargaining power of market participants and reduce barriers to entry that may exist in a dealer-intermediated structure. Supporters of all-to-all trading argue that a more diverse set of market participants could improve market liquidity. This is because it could increase the likelihood and amount of bond transactions occurring in opposite directions, even during periods of turmoil. These offsetting transactions could reduce pressures on market liquidity.

To examine these considerations of all-to-all trading, we use granular transaction-level data to assess how much client-to-client trading could be possible in the Government of Canada (GoC) bond market. We find, on average, almost half of the GoC bond transactions of dealers’ clients could potentially be offset with those of other clients over the trading day. This share is stable over time, including during the COVID-19 crisis in March 2020. This demonstrates that clients were indeed transacting in the opposite direction of other clients’ transactions in a period of market turmoil. For GoC bond futures—instruments that are like GoC bonds but trade on an all-to-all platform—we find that almost all clients’ transactions can be offset by those of other clients and that this high offsetting share is also stable.

So would all-to-all trading support liquidity in the GoC bond market? The answer remains unclear. On the one hand, our results shed some light on the potential for clients’ transactions to offset each other. On the other hand, our methodology overlooks important considerations for the sake of simplicity. For instance, we do not account for differences in the prices when offsetting client transactions or for the influence that client-dealer relationships may have on trading behaviours. These considerations make it challenging to understand whether our estimated extent of client offsetting would take place if GoC bonds were traded entirely on an all-to-all platform. In addition, several other aspects of all-to-all trading merit further investigation.

All-to-all trading presents a range of risks and considerations

While all-to-all trading may bring benefits, it presents a range of risks and considerations. Critics argue that the potential for matching transactions across different participants could be minimal in periods of market turmoil. They argue that even new entrants could have demands for liquidity that lead them to transact government bonds in the same direction as other clients, amplifying one-sided trading rather than promoting matching. Some critics also argue that new entrants and existing clients would have fewer incentives to maintain intermediation services in periods of turmoil. They cite the lower importance on relationships with clients on an all-to-all platform compared with a dealer-intermediated structure as the reason for the potential reduction in intermediation services.

Other considerations that require further analysis include:

• the impact of greater price transparency on an all-to-all platform can affect the behaviour of market participants
• the impacts to settlement and clearing risks as clients’ transactions are executed on an all-to-all platform
• the potential negative impacts to other services provided by dealers, such as access to primary bond markets
• the implications for markets, such as repurchase (repo) or derivatives markets, that trade in conjunction with GoC bonds
• the potential reduction of capital that dealers allocate to their intermediation businesses

Almost half of transactions could potentially be offset on any given trading day

Using data from October 2019 to November 2023 on GoC bond transactions from the Market Trade Reporting System, we calculate the share of transactions of clients of dealers that could have potentially been offset by transactions with other clients within a trading day, following Chaboud et al. (2022).2 In other words, for each purchase of a specific bond that a client makes, what share of the amount bought was sold by other clients on the same trading day and vice versa. The methodology allows buys and sells to offset each other despite potential differences in the prices of trades or the exact times within the day that trade trades occurred.3 It also does not match exact trade amounts; instead, it looks at total buys and sells for a specific bond. Despite these limitations, our methodology helps establish estimates of how many client transactions in a trading day could potentially be matched.

Chart 1 reports the average daily amount of GoC bond transactions that can be offset by clients as a share of the amount of transactions offset plus the residual amount that is not offset. Across all types of GoC bonds, 49% of client transactions could be offset on an average trading day. Across bond tenors, the offsetting shares range from 57% to 65% for benchmark bonds and from 23% to 46% for non-benchmark bonds.

The higher offsetting shares for benchmark bonds is likely due to their greater liquidity, which leads to higher turnover of trading volume compared with that of non-benchmark bonds (Gungor and Yang 2017). The relatively lower offsetting share for non-benchmark bonds suggests that dealers play an important role in using their balance sheets to intermediate these less-liquid bonds. Overall, high offsetting transactions among clients promote two-sided markets. This can help dealers more easily find the bonds or cash they need to fulfill their clients’ transactions, which could promote market liquidity (Sandhu and Vala 2023a).

Chart 1: Benchmark bonds have higher average daily shares of offsetting client transactions than those of non-benchmark bonds

The share of offsetting transactions was stable during the COVID‑19 crisis

In this section, we examine how the share of clients’ offsetting GoC bond transactions changes over time, including in periods of market turmoil. Client-to-client trading could be most beneficial in periods of market turmoil when a larger share of clients, like asset managers, may experience an increased need for liquidity. This type of client-to-client trading would also reduce the volume of transactions that dealers would need to intermediate using their own balance sheets, which they may be more reluctant to do in periods of turmoil.

Chart 2 shows the 21-day moving average of the total daily share of clients’ offsetting GoC bond transactions with the offsetting and residual amounts. The share of clients’ offsetting transactions is fairly stable, at around 50%, from October 2019 to November 2023. The offsetting share was also stable during the COVID‑19 crisis in March 2020, indicating that many clients were transacting GoC bonds in the opposite direction despite significant market turmoil.

Chart 2: The share of offsetting client transactions was stable during the COVID-19 crisis

While our analysis sheds light on the share of client transactions that could potentially be offset, our estimates of clients’ offsetting transactions could have been different if clients had actually traded on an all-to-all platform. This difference arises because our estimates are based on trading behaviour that is influenced by the relationships between dealers and their clients. These relationships are typically formed in dealer-intermediated market structures, like the repo market for GoC treasury bills (Sandhu, Walton and Lee 2019). For instance, dealers can encourage their clients to conduct a trade that they otherwise may not have entered. They could do this by offering price concessions or packaging a GoC bond with other services or products offered to their clients.

As a result, our estimates may be higher than what we would observe on an all-to-all platform given the influence of relationships between dealers and their clients.

Almost all bond futures transactions could potentially offset each other

To understand what client-to-client trading could look like without the influence of dealer-to-client relationships on trading behaviour, we apply our methodology to the GoC bond futures market. In other words, we calculate the share of client transactions that could be offset by other clients for each bond futures contract each day, ignoring individual transaction amounts and any difference in prices or time of the trade over the course of the trading day.4 Like our analysis of GoC bonds, this exercise is useful for understanding whether trading among clients is two-sided at the daily level for GoC bond futures.

GoC bond futures can be a substitute for GoC bonds because they offer similar exposure to interest rates. GoC bond futures, however, are transacted entirely on a futures exchange, which is akin to an all-to-all platform, where clients can trade directly with other clients or dealers anonymously. In this market structure, relationships between dealers and clients are less likely to form because dealers do not directly benefit from their clients’ transactions.

Our comparison is imperfect due to several differences between GoC bonds and bond futures. These differences include the set of clients and dealers, standards for settling transactions and the amount of cash required to enter into positions. As well, futures trades are much smaller than those in the GoC bond market. Moreover, trading is more concentrated in bond futures because there are fewer active bond futures than there are different GoC bonds, possibly increasing their liquidity.

Chart 3 presents the 21-day moving average of the total daily share of offsetting client transactions in the GoC bond futures market as well as the offsetting and residual amounts. The offsetting share is quite stable over our entire sample, from January 2020 to June 2023, and transactions among clients that could potentially offset each other is high, at around 90% on average. In other words, approved participants—who are comparable to dealers in the GoC bond market—are on the other side of only 10% of client transactions over the trading day. These results demonstrate that a large amount of client-to-client trading indeed takes place on an all-to-all platform for an asset class that is like GoC bonds and is less influenced by the relationship between dealers and clients.

Chart 3: The share of offsetting client transactions in Government of Canada bond futures was stable during the COVID-19 crisis

Conclusion

We find that almost half of the GoC bond transactions from clients could be potentially offset by the transactions of other clients over the trading day. This share remains stable in periods of market turmoil. These results help address concerns with all-to-all trading around the potential for overwhelmingly one-sided transactions from clients in periods of turmoil. In the GoC bond futures market, we find that almost all of client transactions are ultimately offset by transactions of other clients. This demonstrates that clients are willing to trade a security similar to a GoC bond but traded on an all-to-all platform. Our results are an important first step in understanding the relevance of broader all-to-all trading in the GoC bond market. However, it remains unclear whether broader all-to-all trading would improve liquidity given the simplicity of our methodology as well as several other risks and considerations that merit further investigation.

Appendix: Robustness of clients’ offsetting transactions

To ensure the approach of measuring clients’ offsetting transactions used in our analysis is robust, we consider alternative methods and find that our overall conclusions hold. One caveat, however, is that we do not account for differences in the prices of buy and sell transactions in our offsetting methodology. Despite the robustness exercises discussed below, ignoring price differences overstates our offsetting potential results.

We calculate the same measure, but instead of offsetting buy and sell transactions for individual GoC bonds, we offset transactions according to broader categories based on the term to maturity of bonds. Specifically, we allow buys and sells of different bonds to offset each other if their term to maturity is between 1 and 3 years; between 3 and 8 years; between 8 and 12 years; and greater than 12 years.

Matching transactions within these broader buckets is plausible because bonds of similar maturity are close substitutes. Clients trading on an all-to-all platform may adjust their behaviour to trade substitutable bonds because the platform may foster greater price transparency and lower costs to locate available bonds for trade. Under this approach, the average daily clients’ offsetting transaction for non-benchmark bonds, across all term-to-maturity buckets, increases by between 14 and 16 percentage points, depending on the term-to-maturity bucket.

For our analysis, we allow buy and sell transactions to offset at the daily level. This is a useful theoretical exercise to help understand whether client trading is one-sided during the day. In reality, clients may not be willing to be exposed to the risks of holding or being short a GoC bond or GoC bond future over the course of the trading day and may want to offset their trades immediately.

To understand the extent of client two-sided trading at shorter intervals, we apply our same offsetting approach, but at 1-hour and 15-minute intervals. For GoC bonds, this approach is only a rough approximation because the time stamps in the Market Trade Reporting System may be inaccurate. As expected, the share of clients’ offsetting transactions declines at shorter intervals.5 Despite these lower levels of client offsetting, the results remain indicative of two-sided trading among clients at shorter intervals, albeit to a lesser extent.

References

Chaboud, A., E. Correia Golay, C. Cox, M. Fleming, Y. Huh, F. Keane, K. Lee, K. Schwarz, C. Vega and C. Windover. 2022. “All-to-All Trading in the U.S. Treasury Market.” Federal Reserve Bank of New York Staff Report No. 1036.

Duffie, D. 2023. “Resilience Redux in the US Treasury Market.” Stanford University Graduate School of Business Research Paper No. 4552735. Presented at the Jackson Hole Symposium, Federal Reserve Bank of Kansas City.

Gungor, S. and J. Yang. 2017. “Has Liquidity in Canadian Government Bond Markets Deteriorated?” Bank of Canada Staff Analytical Note No. 2017-10.

Sandhu, J. and R. Vala. 2023a. “Do Hedge Funds Support Liquidity in the Government of Canada Bond Market?” Bank of Canada Staff Analytical Note No. 2023-11.

Sandhu, J. and R. Vala. 2023b. “The Government of Canada Bond Market: Discussion on Market Structure.” Presentation to the Canadian Fixed-Income Forum, Ottawa, Ontario, November 21.

Sandhu, J., A. Walton and J. Lee. 2019. “Borrowing Costs for Government of Canada Treasury Bills.” Bank of Canada Staff Analytical Note No. 2019-28.

Acknowledgements

We thank David Cimon, Jean-Philippe Dion, Toni Gravelle, Stéphane Lavoie and Adrian Walton for their comments. We are grateful to Harri Vikstedt and Corey Garriott for their advice and suggestions and to members of the Canadian Fixed-Income Forum for sharing their perspectives at their meeting in November 2023. We also thank Alain Chaboud, Ellen Correia Golay, Michael Fleming and Frank Keane for helpful discussions comparing our analysis to their work on all-to-all trading in the US Treasury Market. Finally, we are grateful to Robert Tasca and members of his team at the Montréal Exchange for providing data for our analysis.

Endnotes