How foreign central banks can affect liquidity in the Government of Canada bond market

Introduction

Imagine it’s Friday night and you’re out for dinner at a busy restaurant with some friends. After an hour, you’ve all finished eating but decide to stay at your table for a bit longer to continue catching up.

Remaining at your table for longer presents a trade-off. Of course, you and your friends are allowed to stay at your table after your meal. In fact, you may even order drinks or desserts, generating more revenue for the restaurant. But staying reduces the number of tables available for other customers. This can lead to lengthier wait times or some guests potentially being turned away.

In a sense, financial assets are like tables at a restaurant. When investors hold their assets for long periods—often referred to as buy-and-hold investors—fewer of these assets are available for other investors. This can potentially worsen liquidity in secondary markets.

We examine the trading behaviour of foreign central banks and its potential impact on liquidity in the Government of Canada (GoC) bond market.1 Our analysis builds on recent projects from staff at the Bank of Canada aimed at better understanding how asset managers can affect liquidity in Canadian fixed-income markets (Sandhu and Vala 2023; Bédard-Pagé et al. 2021). The Bank monitors liquidity in secondary markets and has different tools to support liquidity in certain circumstances. This is because liquid and well-functioning fixed-income markets are important for the transmission of monetary policy and the stability of the financial system.

We estimate that foreign central banks own at least 12% to 21% of the float of GoC bonds, which is the amount of GoC bonds outstanding minus the Bank’s holdings. Foreign central banks also tend to hold their GoC bonds for longer than other market participants do.

This buy-and-hold behaviour can promote market stability because foreign central banks’ trading decisions may be less sensitive to market conditions than those of other types of asset managers. In fact, during the peak period of market illiquidity in March 2020, foreign central banks had negligible trading flows, while many other types of asset managers sold GoC bonds, contributing to liquidity strains (Sandhu and Vala 2023).

In contrast, buy-and-hold behaviour could reduce the availability of GoC bonds in secondary markets, potentially reducing liquidity. We find that foreign central banks lend out only a small share of their bonds in the repo market—another sign of buy-and-hold behaviour. We also find that higher levels of foreign central bank holdings of GoC bonds are related to illiquidity. This is because their holdings have a positive and statistically significant relationship with borrowing volumes from primary dealers in the Bank’s Securities Repo Operation (SRO), a program designed to support liquidity in securities-financing markets.

While our results show that foreign central banks’ behaviour could promote illiquidity, that is only one part of the picture. Foreign central banks’ participation in the GoC bond market offers many benefits beyond promoting stability in periods of turmoil. For example, their steady demand for GoC bonds can reduce the cost of funding for the federal government (Lane 2019; Pomorski, Rivadeneyra and Wolfe 2014).2

Central banks manage portfolios of foreign exchange reserves

Central banks—and, to a lesser extent, other authorities like governments—are portfolio managers for foreign exchange reserves. Foreign exchange reserves are assets held in currencies other than a country’s domestic currency. These reserves allow central banks to intervene in their currency markets, potentially helping:

• promote a country’s exports
• reduce capital outflows
• manage market volatility

Reserves can also provide a source of liquidity that can be leveraged to meet international financial obligations or to provide foreign currency to domestic institutions, among other reasons.

Given that reserves are held for precautionary purposes, central banks and governments typically manage reserve portfolios with the objectives of preserving capital and maintaining liquidity. Generating returns is a common tertiary objective conditional on meeting the first two goals. As a result, foreign exchange reserves are typically invested in securities that are highly liquid and of high credit quality, like highly rated government bonds and bills as well as securities issued by government agencies or sub-sovereign levels of governments (Pomorski, Rivadeneyra and Wolfe 2014).

Globally, foreign exchange reserves that are held or denominated in Canadian dollars have increased in recent years and are 2.39% of total reserve assets, which is small relative to currencies like the US dollar or euro (International Monetary Fund 2024). However, this share can be meaningful given that Canadian financial markets are smaller than markets in some other jurisdictions.

Foreign central banks are large players in the Government of Canada bond market

We learn about foreign central banks’ presence in the market by examining their holdings of GoC bonds. Given the lack of a comprehensive dataset of foreign central banks’ holdings, we estimate their holdings of GoC bonds by applying three approaches and using two different datasets. Together, the different approaches give us an idea of the possible levels of foreign central banks’ holdings.

• First, we use data on the holdings of foreign central banks that have the Bank as their custodian. However, these data exclude foreign central banks that have other custodians for their Canadian investments.
• Second, we use data from the Market Trade Reporting System (MTRS) to accumulate the buy and sell transactions of GoC bonds in secondary markets where foreign central banks were counterparties and remove bonds that have matured to approximate their holdings (see Sandhu and Vala 2023 for a description of these data). These data include a larger set of foreign central banks. However, the accumulation starts only in October 2019, so we exclude any bonds purchased and held before then from the approximation. Additionally, we do not capture transactions in the primary market or those with dealers that are not required to report their transactions to MTRS.
• Third, we consider the second approach using MTRS but include the initial holdings as at October 2019 for those central banks that use the Bank as their custodian.

Chart 1 shows the range of estimated holdings of foreign central banks using our three approaches. The holdings of foreign central banks could be at least 12% to 21% of the GoC bond float, showing that foreign central banks own a significant share of GoC bonds, regardless of the estimation approach.3

However, these estimates could understate the holdings of foreign central banks. Even when we combine the two datasets, we miss the initial holdings of those foreign central banks that use custodians other than the Bank of Canada.

Chart 1: Foreign central banks hold a significant share of the float of Government of Canada bonds

Foreign central banks have long holding periods

Buy-and-hold behaviour presents a trade-off for liquidity conditions:

• On the one hand, buy-and-hold investors can promote stability in markets during periods of turmoil, unlike other investors who may trade quickly in response to market movements and amplify market strains. For instance, Sandhu and Vala (2023) show how hedge funds and other clients of dealers sold GoC bonds during the peak of market turmoil in March 2020, adding to strains on market liquidity. However, we find that foreign central banks had negligible trading flows during that period.
• On the other hand, buy-and-hold investors may also reduce liquidity because they effectively reduce the amount of securities available for other market participants. The potential impacts for market liquidity from the trading behaviour of foreign central banks may be greater than other types of market participants given their large presence in the GoC bond market.

Using MTRS data, we calculate the period between the purchase of a bond by each foreign central bank and the date that same bond was sold or had matured. For this estimated holding period, we consider only bonds that were both bought and sold or had matured during our sample period. Some bonds may have been purchased before the start of our sample period or sold after its end. We exclude these transactions from our calculation. For each foreign central bank, we use volume-weighting to aggregate the holding periods across different bonds. We then take the median holding period across different foreign central banks to arrive at an estimate for the entire sector.

Chart 2 compares the estimated holding periods for foreign central banks and for other types of market participants. Foreign central banks have the longest estimated holding period, around four months, suggesting that they exhibit the strongest buy-and-hold behaviour. This could be related to central banks placing less importance on generating returns because their investment objectives differ from other types of asset managers. In turn, this could lead to less trading activity and, consequently, longer holding periods

Chart 2: Foreign central banks have longer estimated holding periods than other asset managers

Foreign central banks lend out a small share of their holdings in the repo market

Buy-and-hold investors could offset any negative impacts their behaviour has on market liquidity if they lend their securities to other participants in securities-financing markets. To see the extent to which foreign central banks do this, we use repo transaction data from MTRS to calculate the amount of GoC bonds that foreign central banks lend out in the repo market, also known as their gross repo position.

Chart 3 shows that foreign central banks lent, on average, around $6 billion of their GoC bonds on any given day from June 2023 to June 2024. The gross repo position of foreign central banks is significantly smaller than that of pension funds and hedge funds, and slightly smaller than that of wealth managers. Insurers and other types of market participants have minimal positions. For foreign central banks, their gross repo position roughly amounts to only between 0% and 4% of their total GoC bond holdings, depending on the estimation approach used.

Chart 3: Foreign central banks’ gross repo position is smaller than some other asset managers

Some foreign central banks may lend their GoC bonds in the securities lending market rather than through the repo market. Foreign central banks that use the Bank of Canada as their custodian likely do not lend their securities. We are, however, unable to measure the securities lending activity of foreign central banks that use other custodians. One reason foreign central banks may have limited securities financing activity is that Canadian dollar assets represent a minimal share of their total portfolios. As a result, the incremental returns from greater securities-financing activity in Canadian markets may not be worth the operating costs for these foreign central banks relative to the returns they could earn from lending out assets denominated in other currencies. Combined with longer holding periods, their limited activity in securities-financing markets observed in our available data suggests that foreign central banks could cause liquidity to worsen in secondary markets for GoC bonds.

Bonds held by foreign central banks tend to be more illiquid

We identify cases of illiquidity in the GoC bond market from the Bank’s SRO program, which aimed to support liquidity in securities-financing markets when it was active from July 2020 until October 2024. Through the SRO program, the Bank was able to lend its GoC securities to primary dealers at a slight backstop rate. Given certain terms and restrictions of the program, primary dealers would have been most incentivized to borrow bonds if they were especially illiquid in secondary markets.

Using daily bond-level data, we run a simple linear regression to assess how different factors relate to SRO volumes. We consider a short list of factors that could explain volumes in the SRO program. These factors include bond characteristics like their time to maturity or benchmark status, and program parameters like pricing and bidding limits. We also consider the GoC bond holdings of foreign central banks using custody data because the time series is more accurate than the MTRS data, where transactions are accumulated over time to estimate holdings.

Table 1 reports the coefficients and significance of each factor in the model. Our results indicate the following:

• In line with the expected relationship between the factors and liquidity, bonds with higher SRO volumes tend to:
  • have lower floats
  • be benchmark bonds
  • have longer maturities
  • have shorter tenors
• The holdings of foreign central banks have a positive and statistically significant relationship with SRO volumes. That is, an increase of Can$1 billion in foreign central banks’ holdings of a given GoC bond is, on average, associated with an increase of Can$220 million in SRO volumes for that bond.

Table 1: Regression model on Securities Repo Operation volumes

Note: All regression variables are statistically significant at the 1% level (***).

While we cannot conclude whether foreign central banks are causing illiquidity or simply purchasing bonds that are already more illiquid, our results show that their holdings are related to low liquidity. If foreign central banks were to increasingly lend their GoC bonds in securities-financing markets, it may reduce the need for programs like the SRO and improve liquidity conditions independent of the Bank’s programs. As an example, the Bank has a securities-lending program that makes US Treasury securities available for lending from Canada’s foreign exchange reserves to enhance returns, while also supporting market liquidity.

Conclusion

We estimate that foreign central banks own a significant share of the float of GoC bonds, ranging from at least 12% to 21%. They tend to also hold their GoC bonds for longer than other market participants. Although this buy-and-hold behaviour can promote stability in periods of turmoil, it could reduce the availability of GoC bonds in secondary markets, potentially lowering liquidity. Indeed, foreign central banks lend only a small share of their bonds in the repo market—another sign of buy-and-hold behaviour.

We also find that higher holdings of GoC bonds by foreign central banks are related to illiquidity in the GoC bond market. This is because the level of these holdings has a positive and statistically significant relationship with volumes in the Bank’s SRO program. Nevertheless, foreign central banks’ participation in the GoC bond market brings several important benefits, such as promoting stability in periods of turmoil.

Appendix

Robustness of holding period results

To calculate holding periods, we consider only bonds that were bought and sold or had matured between August 1, 2020, and June 10, 2024. We exclude any unmatched transactions, meaning bonds that were sold but not bought or that were bought but not sold or matured. Compared with other types of market participants, foreign central banks had a significantly larger share of transactions that were unmatched and therefore excluded when calculating the holding period. Unmatched transactions indicate holding periods that may extend beyond our sample period, meaning that foreign central banks may have even longer holding periods than our calculations suggest.

Robustness of regression results

We consider two alternative methods to validate the approach we use to measure the relationship between foreign central banks’ holdings of GoC bonds and liquidity. Results of these methods show that our overall conclusions hold. The two alternative methods are the following:

• We run the same regression on SRO volumes but include fixed effects for each bond and for each day. These fixed effects offer stricter ways to control for any differences between bonds and across days compared with our original model. Using this alternative regression model, we find that foreign central banks’ GoC bond holdings continue to have a positive and statistically significant relationship with SRO volumes.
• We also consider a regression model where we use a different metric of illiquidity instead of SRO volumes. In this model, the dependent variable is the repo spread, which is the difference between the Canadian Overnight Repo Rate Average and the volume-weighted average repo rate for each bond on each day. Higher repo spreads indicate that a bond borrower in a repo agreement is foregoing interest income to borrow the bond, suggesting that the bond may be scarce or illiquid. Our results generally remain consistent in this alternative model.

References

Bédard-Pagé, G., D. Bolduc-Zuluaga, A. Demers, J.-P. Dion, M. Pandey, L. Berger-Soucy and Adrian Walton. 2021. “COVID-19 crisis: Liquidity Management at Canada’s Largest Public Pension Funds.” Bank of Canada Staff Analytical Note No. 2021-11.

Feunou, B., J.-S. Fontaine, J. Kyeong and J. Sierra. 2015. "Foreign Flows and Their Effects on Government of Canada Yields.” Bank of Canada Staff Analytical Note No. 2015-1.

International Monetary Fund. 2024. “Currency Composition of Official Foreign Exchange Reserves.”

Lane, T. 2019. “Taking Precautions: The Canadian Approach to Foreign Reserves Management.” Speech to the Peterson Institute for International Economics, Washington, D.C., February 6.

Pomorski, L., F. Rivadeneyra and E. Wolfe. 2014. “The Canadian Dollar as a Reserve Currency.” Bank of Canada Review (Spring): 1–11.

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

Acknowledgements

We thank Cassie Davies, Jean-Philippe Dion, Jeffrey Gao, Alexandra Lai, Stéphane Lavoie, Philippe Muller, Andreas Uthemann, Adrian Walton and Jonathan Witmer for helpful comments. We are also grateful to Rishi Vala for analytical support and to Christina Harvey and Manu Pandey for data support.

Endnotes