Liquidity Supply in the Corporate Bond Market

• Published date: 01 April 2021
• Date: 01 April 2021
• Author: JONATHAN GOLDBERG,YOSHIO NOZAWA
• DOI: http://doi.org/10.1111/jofi.12991

THE JOURNAL OF FINANCE •VOL. LXXVI, NO. 2 •APRIL 2021

Liquidity Supply in the Corporate Bond Market

JONATHAN GOLDBERG and YOSHIO NOZAWA∗

ABSTRACT

This paper examines dealer inventory capacity, or liquidity supply, as a driver of

liquidity and expected returns in the corporate bond market. We identify shocks to

aggregate liquidity supply using data on corporate bond yields and dealer positions.

Liquidity supply shocks lead to persistent changes in market liquidity,are correlated

with proxies for dealer f‌inancial constraints, and have signif‌icant explanatory power

for cross-sectional and time-series variation in expected returns, beyond standard

risk factors. Our f‌indings point to liquidity supply by f‌inancially constrained inter-

mediaries as a main driver of market liquidity and asset prices.

LIQUIDITY RISK IS PRICED IN the cross section of asset returns: average re-

turns are higher for assets more sensitive to aggregate shocks to market liq-

uidity. The liquidity risk premium has been documented in equities (Amihud

(2002), Pástor and Stambaugh (2003)), corporate bonds (Lin, Wang, and Wu

(2011)), and other asset classes (Sadka (2010)). The mechanisms underlying

the liquidity risk premium and aggregate illiquidity, however, are the subject

of debate.1

∗Jonathan Goldberg is at the Federal Reserve Board. Yoshio Nozawa (corresponding author,

nozawa@ust.hk) is at the Hong Kong University of Science and Technology (HKUST). We thank

Hui Chen; Darrell Duff‌ie; Giovanni Favara; Valentin Haddad; Grace Hu; Yesol Huh; Maureen

O’Hara; Or Shachar; Kairong Xiao; Alex Zhou; participants at the Asian FA, Baltimore Area Fi-

nance, CICF, Erasmus Liquidity, HEC-McGill Winter Finance,ITAM Finance, and SAFE Market

Microstructure conferences; and seminar participants at Penn State, the Federal Reserve Board,

the Securities and Exchange Commission, and the University of California at San Diego. For com-

ments that improved the paper, we are grateful to Stefan Nagel (Editor), two anonymous referees,

an associate editor, and discussants Patrick Augustin, Jack Bao, Frank de Jong, Xiaoxia Lou,

Andreas Rapp, Dragon Tang, and Quan Wen. We thank FINRA for providing transaction data.

FINRA screening was limited to whether there is suff‌icient aggregation such that no particular

dealer is identif‌ied. The authors received, but were not required to incorporate, comments from

Federal Reserve economists. The views expressed here are those of the authors and do not nec-

essarily represent the views of the Federal Reserve Board or its staff. The authors have no other

material f‌inancial or nonf‌inancial interests related to this research per The Journal of Finance

disclosure policy.

Correspondence: Yoshio Nozawa, Hong Kong University of Science and Technology, Lee Shau

Kee Business Building, Clearwater Bay, NT,Hong K ong; e-mail: nozawa@ust.hk

1Vayanos and Wang (2013) review the literature.

DOI: 10.1111/jof‌i.12991

© 2020 American Finance Association. This article has been contributed to by US Government

employees and their work is in the public domain in the USA

755

756 The Journal of Finance®

This paper provides evidence connecting market liquidity and expected re-

turns to intermediaries’ inventory-absorption capacity, or liquidity supply. We

propose a new method to identify liquidity supply using the sign restriction

that higher liquidity supply leads to a lower price and a higher quantity of liq-

uidity. We apply this method to the corporate bond market. Our liquidity price

ref‌lects “noise” in corporate bond yields, or deviations from f‌itted issuer-level

yield curves. Our liquidity quantity indicates how dealers use their balance

sheets to provide immediacy. We f‌ind that liquidity supply shocks have persis-

tent effects on noise and dealer positions and signif‌icant explanatory power for

cross-sectional and time-series variation in expected returns.

The corporate bond market is nearly ideal for our purposes. Corporate bonds

are traded in an over-the-counter market in which trading costs are substan-

tial and f‌luctuate signif‌icantly (Bao, Pan, and Wang(2011), Dick-Nielsen, Feld-

hütter, and Lando (2012)). In addition, corporate bonds are a signif‌icant source

of f‌irm f‌inancing. The expected return on corporate bonds has been shown to

have predictive power for f‌inancing f‌lows and real activity (Gilchrist and Za-

krajšek (2012), Ma (2019)). Understanding the determinants of corporate bond

liquidity and returns is thus of foremost importance.

Studying the links between liquidity and asset prices is challenging be-

cause standard liquidity measures are driven not only by dealers’ inventory-

absorption capacity (Grossman and Miller (1988)), but also by investors’ liquid-

ity demand and asymmetric information about the assets traded. To identify

shocks to liquidity supply in the corporate bond market, we focus on dealers’

absorption of demand imbalances for similar bonds. Investor demand to buy

bonds even if they are priced richly (or to sell bonds even if they are priced

inexpensively) generates noise in corporate bond yields. Noise compensates

dealers for absorbing these demand imbalances. We therefore use noise as our

liquidity price.

We measure noise in corporate bond yields using weekly bond-level yield

data from 2002 to 2016. Each week, we f‌it a smooth yield curve for each large

issuer of corporate bonds. Our noise measure is the average, across all bonds

in our sample, of the divergence between a bond’s market yield and the yield

curve of its issuer. Because noise captures deviations from issuer-level yield

curves, it is unaffected by issuers’ credit risk and asymmetries in information

about issuers.

Our focus on noise builds on Fontaine and Garcia (2012) and Hu, Pan, and

Wan g (2013), who study Treasury noise and assume that it is driven by liq-

uidity supply. An increase in noise, however, could ref‌lect a deterioration in

liquidity supply or an increase in investors’ liquidity demand. To address this

concern, we study noise together with the quantity of liquidity provided. To ab-

sorb the demand imbalances that give rise to noise, dealers take long positions

in bonds that investors want to sell and short positions in bonds that investors

want to buy. Thus, as the quantity of liquidity, we use gross positions, the sum

of dealers’ gross long and gross short positions.

Liquidity Supply in the Corporate Bond Market 757

We construct this liquidity quantity using detailed transaction data that al-

low us to identify dealers.2To construct each dealer’s position in each bond,

we cumulate transactions. We def‌ine dealer gross positions as the sum, across

bonds and dealers, of the absolute value of each dealer’s position in each bond.

Next, we estimate a structural vector autoregression (VAR) model of noise

and dealer gross positions. We identify supply shocks as leading contemporane-

ously to opposite-sign changes in price and quantity, whereas demand shocks

lead to same-sign changes in price and quantity.

Our method has several advantages. First, we infer liquidity supply directly

from the price and quantity of corporate bond liquidity,rather than using prox-

ies for f‌irm-wide constraints such as dealer f‌irms’ leverage. Second, our method

also identif‌ies aggregate shocks to investor demand imbalances for similar

bonds, or liquidity demand. We examine whether liquidity demand shocks have

different properties than liquidity supply shocks. Third, our method provides

estimates of the persistence of liquidity supply and demand shocks. A shock

with persistent effects on liquidity is expected to be more important for asset

pricing (Acharya and Pedersen (2005)).

We f‌ind that a positive liquidity supply shock is associated with a decrease

in noise and an increase in positions that are quite persistent. Liquidity sup-

ply shocks are correlated with proxies for dealer inventory capacity such as

dealer capital, corroborating our interpretation of the estimated supply shocks.

Moreover, liquidity supply shocks capture well episodes of intermediary stress,

including key events of the 2007 to 2009 f‌inancial crisis. In contrast, liquidity

demand shocks are associated with transitory changes in noise and positions,

are correlated with lagged corporate bond mutual fund f‌lows and issuance,

and do not follow a consistent pattern around stress events. These differences

in the properties characterizing liquidity supply and demand shocks suggest

that liquidity supply and demand should have different asset pricing implica-

tions.

Using a regression approach similar to that of Fama and MacBeth (1973),

we f‌ind that corporate bonds with returns that are more sensitive to liquidity

supply shocks earn higher expected returns, even when controlling for bond-

level liquidity and other characteristics.3The liquidity supply risk premium

remains signif‌icant when controlling for the sensitivity of each bond to well-

known liquidity measures (Amihud (2002), Pástor and Stambaugh (2003)),

which is consistent with the view that these liquidity measures are driven

by factors beyond just liquidity supply. In contrast, we do not f‌ind evidence for

a liquidity demand risk premium.

2We focus on primary dealers, the main intermediaries in the corporate bond market, and

designated counterparties of the Federal Reserve.We list the primary dealers in Internet Appendix

Section I. The Internet Appendix is available in the online version of the article on The Journal of

Finance website.

3Internet Appendix Section II presents a dynamic equilibrium model that relates liquidity sup-

ply and demand shocks to noise, gross positions, aggregate excess returns, and the cross section of

excess returns.