# Using Exchange-Traded Funds to Measure Liquidity in the Canadian Corporate Bond Market

## Introduction

Market participants in Canada have suggested that the liquidity of corporate bonds worsened in recent years. Yet, previous analysis by Bank of Canada staff (Fan et al. 2018b) shows that corporate bond market liquidity has generally improved since 2010. That conclusion relies on two commonly used liquidity proxies that are computed using data from transactions. However, most corporate bonds trade infrequently; in fact, only about 200 bonds transact on any given day. Consequently, infrequently traded bonds are missing from the existing liquidity proxies. This raises the concern that these proxies do not provide a complete picture of the liquidity conditions of the overall market.

We construct a new liquidity proxy using the price of exchange-traded funds (ETFs), the ETF-based proxy of Canadian corporate bond market liquidity (ECML). ECML measures the average liquidity of about 900 corporate bonds each day. Many of these bonds transact infrequently and are consequently missing from the existing liquidity proxies. Nonetheless, using ECML leads us to the same conclusion as Fan et al. (2018b):

• corporate bond market liquidity has generally improved since 2010

But, the longer sample available with ECML also shows that liquidity has remained relatively stable since 2014.

Staff at the Bank of Canada have also computed the same liquidity proxies using data from transactions of federal and provincial government bonds (Gungor and Yang 2017; Fan et al. 2018a). These proxies show that liquidity in both markets has likewise improved since 2010.

This note is part of a broad effort to subject the previous findings to stronger tests or robustness checks. Fontaine et al. (2017) show that the existing proxies are reliable for measuring liquidity of federal government bonds that are less frequently traded. Our results further strengthen confidence that the proxies used by staff reflect underlying liquidity conditions in Canadian fixed-income markets.

## Dynamics are similar across liquidity proxies

We compare the dynamics of ECML with those of the existing liquidity proxies—i.e., a bid-ask spread proxy and a price-impact proxy. ECML shows that corporate bond liquidity deteriorated significantly during the 2008 global financial crisis. It also shows that liquidity has been recovering since, interrupted only occasionally by episodes of stress. Moreover, ECML and the other two liquidity proxies paint a similar picture of improving liquidity conditions since 2010. We therefore confirm the previous findings using information from a much larger number of corporate bonds. Nonetheless, the longer sample of ECML shows that market liquidity has remained relatively stable since the oil price shock in 2014 (Chart 1).

We measure how closely these proxies move over time using the correlation coefficient. We find coefficients of 0.43 and 0.41, respectively, for the bid-ask spread proxy and the price-impact proxy with ECML. The positive correlations remain when we examine their relationship in weekly changes (see Appendix A.2 for the robustness result). Our findings confirm the reliability of the two existing proxies for measuring aggregate corporate bond liquidity.

### Chart 1: Liquidity of Canadian corporate bonds has improved since 2010

Sources: Bloomberg, Canadian Depository for Securities and Bank of Canada calculations Last observation: December 31, 2017

## What is the ETF-based proxy of Canadian corporate bond market liquidity?

An ETF issues shares to investors that can be traded on an exchange throughout a given day. The ETF then invests the money from investors in a basket of securities. Consequently, two prices exist for an ETF: one can be obtained from transactions on the exchange, while the other is the reported net asset value (NAV) of the ETF’s security holdings. In theory, these two prices should be the same because they reflect the cash flows of the same underlying securities. Moreover, the presence of authorized participants (APs), who can create and redeem ETF shares, helps keep the two prices close (See Box 1 in Foucher and Gray [2014] for more details on the creation and redemption process).

Nonetheless, these two prices can sometimes differ significantly, especially if the underlying securities are less liquid, such as in the case of corporate bonds. When corporate bonds become less liquid, the costs of buying or selling corporate bonds increase. This makes it harder for APs to create and redeem ETF shares and close the gap between ETF prices and NAVs. Therefore, we can use this difference as a proxy for the liquidity of the corporate bonds. Inspired by Chacko, Das and Fan (2016), ECML averages the differences between the price that we observe on the exchange and the reported NAV for eight funds. (See Appendix A.1 for data details.)

## ECML has advantages and limitations

Table 1 summarizes the key features of ECML and the other two liquidity proxies. Empirically, ECML has several advantages:

• ECML uses information from a broad set of corporate bonds and thus provides a consistent assessment of liquidity conditions of the overall market.
• ECML exploits NAV estimates that capture how traders assess the underlying prices for every bond that the ETF holds. We do not have better estimates of bond prices in the absence of recent transactions.
• ECML can be computed for a longer history, which gives us additional insights about liquidity dynamics around the 2008 global financial crisis.

ECML also has limitations:

• ECML is based on the difference between ETF prices and their NAVs; it thus is an indirect measure of corporate bond liquidity.
• ECML could be affected by factors that are not related to bond liquidity. For example, transactions of ETF shares by short-term noise traders may push ETF prices away from their NAVs.
• ECML relies on publicly available end-of-day NAVs that introduce measurement error. This is because traders use internal estimates of tradable bond prices to assess arbitrage opportunities.

However, the close correlation between ECML and the two liquidity proxies suggests that the impact of other factors on ECML may be small.

### Table 1: Key features of liquidity proxies

 Description Bid-ask spread / price-impact ECML Empirical implication Rely on transaction data Overrepresented by liquid corporate bonds Available since 2009 Relies on transacted exchange-traded fund prices and reported net asset values Incorporates information from a broader set of corporate bonds Available since 2006 Theoretical foundation Based on theoretical model constructed to study market liquidity Direct measures of market liquidity Based on the difference between exchange-traded fund prices and their net asset values Indirect measure of market liquidity May be affected by non-liquidity related factors

## Conclusion

We construct a new proxy for the liquidity of Canadian corporate bonds using the price of ETFs. This new proxy measures the liquidity of about 900 bonds every day and addresses the limitations faced by the existing proxies. The new proxy moves with the existing proxies and confirms the previous findings that corporate bond liquidity has improved since 2010. ECML will be an important addition to the staff’s tool box to measure market liquidity.

## Appendix

### A.1. Data

Table A.1 lists the eight ETFs used in this study.

### A.2. Robustness results

To examine whether changes in ECML are also positively related to weekly changes in the other proxies, we estimate the following regression:

$$ECML_{t}$$ $$=\,α$$ $$+\,\displaystyle\sum_{i=1}^{2}β_{i}*ECML_{t-i}$$ $$+\,γ*x_{t}$$ $$+\,time\,\,trend$$ $$+\,ε_{t}.$$

The coefficient of interest is $$γ$$ because it captures the effects of the other liquidity proxies on ECML. We also account for time trend exhibited by both ECML and the other liquidity proxies in our sample. Table A.2 reports estimate of $$γ$$ over different sample periods.

### Table A.2: Relationship of ECML to the other two liquidity proxies

 2009–17 2009–13 2014–17 Bid-ask proxy 0.170*** (3.85) 0.197*** (2.82) 0.068** (2.11) Price-impact proxy 0.061*** (3.20) 0.080** (2.51) 0.019 (1.56)

Note: Absolute values of t statistics are in parentheses. *** indicates 1% (statistical) significance, ** 5% significance, *10% significance.

## References

Chacko, G., S. Das and R. Fan. 2016. “An Index-Based Measure of Liquidity.” Journal of Banking and Finance 68 (July): 162–178.

Fan, C., S. Gungor, G. Nolin and J. Yang. 2018a. “Have Liquidity and Trading Activity in the Canadian Provincial Bond Market Deteriorated?” Staff Analytical Note No. 2018-30.

Fan, C, S. Gungor, G. Nolin and J. Yang. 2018b. “Have Liquidity and Trading Activity in the Canadian Corporate Bond Market Deteriorated?” Staff Analytical Note No. 2018-31.

Fontaine, J.-S., J. Gao, J. Sandhu and K. Wu. 2017. “Do Liquidity Proxies Measure Liquidity in Canadian Bond Market?” Staff Analytical Note No. 2017-23.

Foucher, I. and K. Gray. 2014. “Exchange-Traded Funds: Evolution of Benefits, Vulnerabilities and Risks.Bank of Canada Financial System Review (December): 37–46.

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

## Acknowledgments

We thank Jason Allen, Guillaume Bédard-Pagé, Narayan Bulusu, Jean-Sébastien Fontaine, Virginie Traclet and Nicole van de Wolfshaar for helpful comments and suggestions. We are also grateful to Ryan Shotlander for excellent research assistance.

## Avis d’exonération de responsabilité

Les notes analytiques du personnel de la Banque du Canada sont de brefs articles qui portent sur des sujets liés à la situation économique et financière du moment. Rédigées en toute indépendance du Conseil de direction, elles peuvent étayer ou remettre en question les orientations et idées établies. Les opinions exprimées dans le présent document sont celles des auteurs uniquement. Par conséquent, elles ne traduisent pas forcément le point de vue officiel de la Banque du Canada et n’engagent aucunement cette dernière.

DOI : https://doi.org/10.34989/san-2019-25