Disentangling types of liquidity and testing limits‐to‐arbitrage theories in the CDS–bond basis

• Published date: 01 January 2021
• Author: Patrick Augustin,Jan Schnitzler
• Date: 01 January 2021
• DOI: http://doi.org/10.1111/eufm.12278

Eur Financ Manag. 2021;27:120–146.wileyonlinelibrary.com/journal/eufm120

|

© 2020 John Wiley & Sons Ltd.

DOI: 10.1111/eufm.12278

ORIGINAL ARTICLE

Disentangling types of liquidity and testing

limits‐to‐arbitrage theories in the CDS–bond

basis

Patrick Augustin

1

|Jan Schnitzler

2

1

Desautels Faculty of Management &

Canadian Derivatives Institute, McGill

University, Montreal, Quebec, Canada

2

VU Finance Department, VU Amsterdam,

Amsterdam, The Netherlands

Correspondence

Patrick Augustin, Desautels Faculty of

Management & Canadian Derivatives

Institute, McGill University,

1001 Sherbrooke St. West, Montreal,

Quebec H3A 1G5, Canada.

Email: Patrick.Augustin@mcgill.ca

Abstract

We disentangle asset‐specific, market, and funding

liquidity in the CDS–bond basis outside and

during the 2007–9 global financial crisis. Our findings

stress the importance of separating different types

of liquidity, since all three measures have in-

dependently negative impacts on the basis. Funding

liquidity emerges as the economically most important

liquidity metric. While asset‐specific liquidity is

cross‐correlated in both the cash and derivative

markets, funding and market liquidity only matter for

the cash market. We exploit the decomposition of the

basis to test predictions of limits‐to‐arbitrage theories.

We find strong evidence in favor of margin‐based

asset pricing and flight‐to‐quality effects.

KEYWORDS

arbitrage, basis, credit default swaps, corporate bonds, credit risk,

counterparty risk, liquidity

JEL CLASSIFICATION

C1; C23; G01; G12; G14

EUROPEAN

FINANCIAL MANAGEMENT

We are grateful to the editor, John Doukas, and an anonymous referee forinvaluable feedback. We would like to thank

Jan Ericsson,David Lando, and seminar participantsat the New York University and the StockholmSchool of Economics

for helpful commentsand discussions. All remaining errors areour own. The present project has been supportedby the

LuxembourgNational Research Fund and by the OMX‐NasdaqNordic Foundation. The paper circulatedpreviously under

the title “Squeezed Everywhere —Disentangling Types of Liquidity and Testing Limits‐to‐Arbitrage.”

1|INTRODUCTION

Credit default swaps (CDSs) and bonds are the primary building blocks of the CDS–bond basis

trade. That popular leveraged convergence arbitrage strategy exploits small price discrepancies

between cash bonds and their synthetic counterparts, which , in the absence of frictions, should be

priced equally in both markets (Duffie, 1999). During the 2007–9 global financial crisis (GFC),

textbook arbitrage opportunities in fixed income markets were violated and the CDS–bond basis

became persistently negative (Bai & Collin‐Dufresne, 2019; Choi, Shachar, & Shin, 2019; Fonta-

na, 2010), implying that bonds were priced at a relative discount. Similar dislocations have emerged

with the onset of the Covid‐19 pandemic (Haddad, Moreira, & Muir, 2020).

In this paper, we study the dynamics of the CDS–bond basis (henceforth the basis). During the

GFC, market liquidity dried up, margin calls increased, and funding constraints became binding.

Hence, many investors were forced to delever and unwind their basis trades, which amplified losses

for arbitrageurs and financial institutions already plagued be severe write‐downs (D. E. Shaw

Group, 2009). Thus, the observed price dislocations in fixed income markets are likely to be

intimately linked to asset‐specific liquidity characteristics, the credit deterioration of insurance

counterparties, runs in repo markets, and heightened funding constraints during times of distress.

Our objective is to disentangle how the basis is affected by the idiosyncratic liquidity component

independently from common market and funding liquidity. We exploit this decomposition to test

various predictions of limits to arbitrage theories. The dynamics of the basis provide a useful

laboratory to better understand such major fixed income anomalies (Gromb & Vayanos, 2009).

Improving our understanding of these dynamics is especially important in market venues as large

as the US corporate bond and single‐name CDS markets, which were estimated to have notional

amounts outstanding of approximately $6.7 trillion (Securities Industry and Financial Markets

Association, 2019) and $18.1 trillion (Bank for International Settlements, 2011) in 2010, respectively.

We measure the basis nonparametrically as the difference between an entity's CDS spread and

the spread of a synthetic constant‐maturity bond yield over the swap curve. After our decom-

position of the basis, we examine the relation between the basis, the liquidity components, and

other factors, such as counterparty risk, using a dynamic panel regression framework. Our sample

covers the period between January 2004 and September 2010. However, our results are likely to

provide insights for better understanding the dynamics of the negative CDS–bond basis observed

during the 2020 coronavirus pandemic as well (Haddad et al., 2020).

Our results suggest that all of asset‐specific illiquidity, market illiquidity, and funding

illiquidity bear a statistically significant negative association with the basis. Among these

factors, funding liquidity displays the economically strongest effect, which becomes even more

pronounced during the crisis period. Similarly, we find stronger effects for asset‐specific and

market liquidity during the crisis for our sample of non‐financial companies. For financial

companies, however, these two factors become statistically insignificant during the crisis, even

though their point estimates do not change very much. Finally, we show that asset‐specific

liquidity is a statistically and economically significant determinant of cash and synthetic

fixed income markets, consistent with the evidence of liquidity effects in CDS contracts

(e.g. Bongaerts, De Jong, & Driessen, 2011; Qiu & Yu, 2012; Tang & Yan, 2007). In contrast, we

show that market and funding liquidity are significant determinants only for the bond market.

Our findings are consistent with the existence of liquidity spirals in the spirit of Brunnermeier

and Pedersen (2009), who stress that funding shocks and binding financing constraints raise the

shadow cost of capital for arbitrageurs and decrease market‐wide liquidity (e.g. Bai & Collin‐

Dufresne, 2019; Fontana, 2010). Another possible interpretation of the evidence is one of loss spirals

AUGUSTIN AND SCHNITZLER EUROPEAN

FINANCIAL MANAGEMENT

|

121