Mutual Fund Holdings of Credit Default Swaps: Liquidity, Yield, and Risk

• Date: 01 April 2021
• Author: JITAO OU,WEI JIANG,ZHONGYAN ZHU
• DOI: http://doi.org/10.1111/jofi.12996
• Published date: 01 April 2021

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

Mutual Fund Holdings of Credit Default Swaps:

Liquidity, Yield, and Risk

WEI JIANG, JITAO OU, and ZHONGYAN ZHU*

ABSTRACT

This study analyzes the motivations for and consequences of funds’ credit default

swap (CDS) investments using mutual funds’ quarterly holdings from pre- to postf‌i-

nancial crisis. Funds invest in CDS when facing unpredictable liquidity needs. Funds

sell more in reference entities when the CDS is liquid relative to the underlying bonds

and buy more when the CDS-bond basis is more negative. To enhance yield, funds en-

gage in negative basis trading and sell CDS with the highest spreads within rating

categories, and with spreads higher than those of their bond portfolios. Funds with

superior portfolio returns also demonstrate more skill in CDS trading.

CREDIT DEFAULT SWAPS (CDS) GREW from a small market in the early 2000s

to a $62 trillion global market (in terms of notional amount) by 2007, the year

prior to the f‌inancial crisis (Jarrow (2011)). Despite the popularity of CDS as

a synthetic security to gain or hedge credit exposure, there has been little

empirical research on how institutional investors, an important class of end

users, employ and benef‌it from CDS contracts. Not only do mutual funds act

as both long- and short-term investors, but their trades arise from a variety of

motives, including liquidity management, speculating on private information,

and searching for yield. Mutual funds therefore provide a rich setting in which

*Wei Jiang is at Finance and Economics Division, Columbia Business School. Jitao Ou is at

the Department of Mathematics, Hong Kong Baptist University. Zhongyan Zhu is at the Depart-

ment of Banking and Finance, Monash Business School. We thank TaylorBegley, Martin Oehmke,

George Aragon, Rich Evans, Tim Riley, Ronald Sverdlove, Lorne N. Switze, Suresh Sundaresan,

Adam Zawadowski, two anonymous referees, the Associate Editor, and the Editor for valuable

comments and suggestions. We benef‌ited from discussions with seminar and conference partici-

pants at AIM Investment Conference (at Austin), Australian National, Columbia, Drexel, Monash,

Lehigh, University of Technology Sydney, the Second Annual Conference on Regulation of Finan-

cial Markets (at the SEC), FIRS, MFA,WFA, and the Financial Institutions, Regulation and Corpo-

rate Governance Conference. Zhu acknowledges the generous f‌inancial support from the Research

Grants Council (RGC490712) Hong Kong and Monash University. Zhu also thanks the excellent

research assistance of a group of over 20 college and graduate students at CUHK, especially Wu

Rui, Maggie Jingwei Lai, and Silvia Mei Ngan Wong. The authors do not have potential conf‌licts

of interest to disclose as def‌ined in The Journal of Finance’s disclosure policy.

Correspondence: Zhongyan Zhu, Department of Banking and Finance, Monash Business

School, Monash University,900 Dandenong Road, Caulf‌ield East Victoria 3145, Australia; e-mail:

zhongyan.zhu@monash.edu.

DOI: 10.1111/jof‌i.12996

© 2020 the American Finance Association

537

538 The Journal of Finance®

to test the interaction between the bond and CDS markets and the derivatives

investment strategies of institutional investors more broadly.

Using a detailed and comprehensive data set of mutual fund quarterly posi-

tions in CDS from 2006 to 2012, a period spanning the pre- and postf‌inancial

crisis eras, in this study we present a comprehensive set of empirical tests

on the motivations for and consequences of CDS investment by mutual funds.

A data overview shows that mutual funds are signif‌icant end users of CDS,

with total CDS positions outstanding ranging from 10.7 to 84.0 billion dollars

annually. As a group, mutual funds are net sellers of CDS, which the total sell-

ing notional amount during the sample period exceeded the buying notional

amount by about 50%.

Such involvement of mutual funds in the CDS market may appear curious

at f‌irst glance because CDS contracts represent “redundant” securities in that

their payoffs could be replicated by underlying securities in a perfect capital

market. For example, the credit exposure and payoff from selling CDS could be

replicated by a long position in the underlying bond. We show that with f‌inan-

cial market frictions and existing incentive structures, CDS contracts serve

multiple objectives that would be diff‌icult to achieve by trading in the under-

lying securities alone.

The f‌irst distinct motive for using CDS that we study is liquidity manage-

ment. CDS contracts require less capital and afford more liquidity to create

the same return prof‌ile as buying or selling a comparable bond from the same

issuer. Recent theory (notably Oehmke and Zawadowski (2015)) indicates that

these features make CDS contracts particularly appealing to investors with

short-horizon capital and a directional view of the credit risk of the reference

entity. Moreover, investors with long-horizon capital but no strong directional

view can employ a “negative basis trade,” whereby the investor takes a long

position in both the CDS and the underlying bonds, to capture the typically

negative spread between those positions that compensates the investor for the

illiquidity of bonds.

Our results support the theoretical predictions above. Though CDS and their

underlying bonds could be viewed as substitutes in taking long or short posi-

tions in credit exposure, mutual funds with less predictable funding f‌lows (and

hence more liquidity needs) and with longer-duration bond holdings (and hence

less liquidity realized from naturally maturing securities) are more likely to in-

vest in the CDS market. The economic magnitude of this relation is sizable: a

one-standard-deviation increase in the volatility of funding f‌lows increases the

propensity of CDS selling or buying by mutual funds by about 38% to 39%.

The relative liquidity of bonds and CDS contracts also plays a signif‌icant role.

While more liquidity on the CDS side (proxied by the number of distinct CDS

contracts on the same reference entity) encourages CDS buying and selling, it

has a larger impact on selling: a one-standard-deviation increase in CDS liq-

uidity more than doubles the likelihood of CDS selling and increases the odds

of CDS buying by 26%.

The second CDS motive we analyze is yield enhancement. CDS contracts al-

low funds to take levered risk in high-spread reference entities and thus juice

Mutual Fund Holdings of Credit Default Swaps 539

up portfolio yields without increasing the perceived risk of the portfolio com-

mensurately, due to the generally lower transparency of derivative positions

compared to long portfolios in the underlying securities.1By selling CDS con-

tracts on reference entities with credit spreads that are signif‌icantly higher

than the credit spreads of either the bond portfolio or the CDS buying posi-

tions, mutual funds gain higher yields relative to the yield of their portfolio of

bonds, in the absence of major credit events. Building on the f‌inding in prior

literature (e.g., Becker and Ivashina (2015)) that institutional investors favor

bonds with the highest yields within a given credit rating category, we expect

a similar pattern in CDS selling by mutual funds.

Consistent with this prediction, we f‌ind that mutual funds generally sell

(buy) CDS with high (low) credit spreads. The average difference peaked at

over 1,000 bps during the f‌inancial crisis and remained around 150 bps postcri-

sis. The average CDS spread which a mutual fund sells is close to 100 basis

points higher than the average credit spreads of the rest of the same fund’s

portfolio. Without taking the full credit risk in a “negative basis trade,” funds

that take long positions in both bonds and CDS earn the premium in the bond

yield in excess of the CDS spread due to the relative illiquidity of the bond

(Bongaerts, de Jong, and Driessen (2011), Oehmke and Zawadowski (2015)).

A one-standard-deviation decrease in the CDS-bond basis almost triples the

odds of a negative basis trading. Finally, mutual funds’ CDS selling is heavily

skewed toward reference entities that command the highest yields within a

given credit range, such as investment grade. The empirical frequencies of all

CDS sell positions that fall into the f‌irst, second, third, and fourth quartiles

are 9%, 12%, 25%, and 54%, respectively, rejecting the null hypothesis of no

yield-chasing, which would result in an even distribution across all quartiles.

Given that investing in derivatives is a hallmark of sophisticated strate-

gies, it is natural to ask whether the CDS positions are informed, and to what

extent they contribute to fund performance. Our analyses show that, overall,

CDS-using funds tend to have better performance, as measured by the sim-

ple market-adjusted alphas or as benchmarked against a full 10-factor (Bai,

Bali, and Wen (2019)) model. However, the intensity of CDS investment does

not seem to contribute positively to portfolio performance. While evidence is

weak that CDS trading is informed in the full sample, CDS selling by “skilled”

funds (funds that fall into the top quartile in terms of 10-factor alpha) exhibits

superior timing during and after the f‌inancial crisis.

The incremental yield earned from selling CDS comes at the cost of “hidden

tail risk” that is similar to selling disaster insurance and that is usually not

fully captured by the benchmark in real time (Rajan (2006)). Since the true

performance can only be assessed over a period that is much longer than the

1Several features of CDS make proper risk assessment by outsiders diff‌icult. First, derivatives,

including CDS, are not included in most processed holdings databases. Second, before the SEC

specif‌ied the rules for cash collateral segregation for CDS short positions in 2012, there was no

clear metric to gauge the leverage gained by CDS. Finally, because most CDS positions are initi-

ated at close to zero market value, they do not immediately affect risk metrics that are weighted

by market value.