Borrowing Costs for Government of Canada Treasury Bills

Borrowing Government of Canada securities is like renting an apartment

A Government of Canada (GoC) security can be borrowed in exchange for a cash loan. To understand what influences borrowing costs for GoC securities, let’s consider the apartment rental market for comparison.

When landlords rent out their apartments, they consider housing market conditions in the area and the amenities and location of the building when deciding how much to charge their tenants. In addition, they might also consider whether a tenant is reliable and intends to stay long term. In other words, the landlord-tenant relationship may affect the cost of rent.

Similarly, lenders of GoC securities consider market conditions and other factors when pricing their securities. For GoC bonds borrowed in a repurchase agreement (repo), bond characteristics such as liquidity and amount outstanding, as well as market conditions, are the most important factors for explaining borrowing costs. For GoC treasury bills (t-bills), in contrast, the relationship between the parties involved in a repo, like the relationship between landlords and tenants, is the most important factor for explaining borrowing costs.

We find that certain pairs of parties conduct most of their repos for t-bills rather than for comparable bonds. In these pairs, the t-bill borrower tends to face higher borrowing costs than parties in other pairs involved in a repo for the same t-bill. We speculate that certain pairs have formed mutually beneficial relationships—one party receives a service in which it can consistently lend its cash in exchange for borrowing t-bills. Apart from earning interest on its cash, this party benefits because it consistently receives t-bills, which are often treated more favourably than bonds for managing risk. On the other side of the trade, the t-bill lender benefits because it receives a cash loan at a relatively low interest rate.

Staff at the Bank of Canada monitor indicators of conditions in GoC securities markets, including borrowing costs in the repo market. Our results are important because they help us interpret the significance of high borrowing costs for GoC securities. While high borrowing costs for bonds are interpreted as scarcity or high demand for a particular bond in the repo market, this is not necessarily the case for t-bills. Our findings suggest that high borrowing costs for t-bills could reflect the benefits of other services exchanged between the parties involved in the transaction, rather than the scarcity of or high demand for the particular t-bill used as collateral.

Many t-bills are expensive to borrow

Borrowing costs for t-bills can vary. We measure the cost of borrowing these securities as the difference between the target policy rate and the interest rate earned on the cash lent. Chart 1 shows that 51 percent and 33 percent of the average daily repo volume for t-bills and bonds, respectively, are conducted at borrowing costs ranging from 5 basis points (bps) to greater than 20 bps during the sample period from January 1, 2016, to December 31, 2018. Note that the share of daily repo volume with these relatively high borrowing costs is slightly larger for t-bills than it is for bonds.

Chart 1: Average daily repo volume at different ranges of borrowing costs

The most important factor of t-bill borrowing costs is the parties involved in the transaction

We identify the factors that are most important for explaining the costs of borrowing t-bills. We use a regression framework to evaluate the relative importance of the factors typically thought to contribute to the borrowing costs of fixed-income securities (see the Appendix).

Chart 2 shows the importance of each of these factors for both t-bills and bonds. For a bond, its characteristics and trading conditions, such as its liquidity, are the most important. For a t-bill, the most important factor is who is trading with whom. This suggests that t-bill borrowers accept earning a lower rate on their cash because they find some value in these repos that is not related to the characteristics or trading conditions of the t-bills, or general market conditions, but rather is related to their relationship with the t-bill lender.

Chart 2: Factors contributing to borrowing costs

Note: Security characteristics and trading conditions include security identifiers, time to maturity, amount outstanding, settlement fails, and measures of liquidity and relative value. The impact of who is trading with whom is measured using fixed effects for each pair of parties involved in a repo. General market conditions include the volatility index, bank funding costs and the timing of policy interest rate announcements. See the Appendix for details.

Repo relationships focused on t-bills transact at high borrowing costs

To explore how repo relationships can influence the cost of borrowing t-bills, we calculate the average borrowing costs for pairs of parties with ongoing repo relationships. We define pairs with ongoing repo relationships as those that conducted at least 100 repo transactions during the sample period, though our findings do not materially change with different minimum transaction thresholds. The pairs that meet this criterion conducted more than 90 percent of the total repo volume for t-bills and for bonds with time to maturity of one year or less. We focus our analysis on these bonds because they are close substitutes for t-bills.

For each pair, we calculate the volume of repo transactions for t-bills as a share of their total repo transaction volume for both t-bills and bonds of similar maturity. Chart 3 shows the pairs sorted into five groups based on that proportion. We find that pairs that conducted more than 80 percent of their repos for t-bills were the largest group, both in terms of number of pairs and transaction volume. The t-bill borrowers in this group had the highest average borrowing costs. For the same t-bills, on the same days, these parties’ borrowing costs were on average about 5 bps higher than the borrowing costs of parties that conducted less than 80 percent of their repos for t-bills.

Chart 3: Statistics of repo relationships based on their usage of t-bills

Chart 3: Statistics of repo relationships based on their usage of t-bills

Chart 3: Statistics of repo relationships based on their usage of t-bills

Usage of t-bills (as % of total repo volume for GoC securities with maturities of 1 year or less)

Other services may offset high borrowing costs for t-bills

We offer two plausible reasons why certain parties consistently borrow t-bills at high costs in the repo market.

First, we speculate that they may value the ability to lend their cash for interest consistently to the same counterparty. By having a relationship with a counterparty, they avoid the cost and time associated with repeatedly searching for someone to lend to.

Second, we speculate that they may also value the ongoing access to t-bills that their repo relationship offers them because securities with such short maturity and low market risk can be treated more favourably than bonds when they are used to manage risk. Take, for instance, securities pledged as collateral in a loan or a derivatives transaction. The percentage of the security’s market value that will be recognized as collateral value is often higher for GoC securities with less than one year to maturity, such as t-bills (see, for example, the collateral schedules of the Canadian Derivatives Clearing Corporation [2019] or the Bank of Canada [2018] Standing Liquidity Facility). In addition, the Canadian Derivatives Clearing Corporation accepts only cash or t-bills, and not bonds, as collateral to safeguard the system in the event of a member’s default. Rather than purchase the t-bills outright, these parties may prefer borrowing them through repos to avoid exposure to the t-bills’ price fluctuations.

These two benefits could offset the higher borrowing costs incurred by the t-bill borrower. This repo relationship is beneficial to the t-bill lender as well because the lender receives cash at a relatively low interest rate. Alternatively, these repo transactions could be part of a broader set of services that these parties exchange, such as other cash management services.

Conclusion

The factors that are most important for explaining the cost of borrowing GoC t-bills in the repo market are different than those for bonds. The most important factor for t-bills is the trading relationship between the parties involved in the transaction. Parties that tend to borrow predominantly t-bills in a repo relationship face higher borrowing costs. We speculate that these pairs have formed a mutually beneficial service relationship in which one party receives a service where it can consistently earn interest on its cash and borrow t-bills, and the other party receives a relatively cheap cash loan.

Our results are important because they suggest that unlike high borrowing costs for bonds, high borrowing costs for t-bills mostly reflect the benefits of other services exchanged between the pairs of repo market participants.

Appendix

Data

We use repo transaction data reported by GoC securities dealers to the Investment Industry Regulatory Organization of Canada’s Market Trade Reporting System (MTRS). Our sample period is from January 1, 2016, to December 31, 2018. We restrict our sample to trades between dealers and clients by removing interdealer trades. We exclude interdealer trades because the interdealer market can serve as an intermediary for client orders. Including these intermediary trades in our sample could overstate our results.

Factors related to borrowing costs

Table 1: List of factors used in the regression framework

Regression framework

To determine how much of the variation in borrowing costs we can explain, we first regress all the factors listed in Table 1 on borrowing costs. Next, we evaluate how much of the variation explained by this original regression model is associated with each of the individual factors. We calculate the incremental variation for each factor as the difference between the variation explained by the original regression model and that explained by a new regression in which we remove that specific factor. We report this incremental variation as a share of the variation explained by the original regression model. This share represents each factor’s relative importance for borrowing costs. We sum the relative importance for three categories of related factors: security characteristics and trading conditions, relationships, and general market conditions.

References

1. Bank of Canada. 2018. “Historical Assets Eligible as Collateral under the Bank of Canada’s Standing Liquidity Facility—July 23, 2018 to July 28, 2019.”
2. Canadian Derivatives Clearing Corporation. 2019. “Collateral.”
3. Fan, C., S. Gungor, G. Nolin and J. Yang. 2018. “Have Liquidity and Trading Activity in the Canadian Provincial Bond Market Deteriorated?” Bank of Canada Staff Analytical Note No. 2018-30.
4. Fontaine, J.-S. and G. Nolin. Forthcoming. “Measuring Limits of Arbitrage in Fixed-Income Markets.” Journal of Financial Research.
5. Fontaine, J.-S., J. Pinnington and A. Walton. 2017. “What Drives Episodes of Settlement Fails in the Government of Canada Bond Market?” Bank of Canada Staff Working Paper No. 2017-54.
6. Gungor, S. and J. Yang. 2017. “Has Liquidity in Canadian Government Bond Markets Deteriorated?” Bank of Canada Staff Analytical Note No. 2017-10.

Acknowledgements

We thank Joshua Fernandes for excellent research assistance and Nicole van de Wolfshaar for being our editorial superhero. We are also grateful to Rohan Arora, Danny Auger, Jean-Philippe Dion, Jean-Sébastien Fontaine and Virginie Traclet for helpful comments and suggestions.