Macroeconomic Conditions and Credit Default Swap Spread Changes

• Published date: 01 August 2017
• Date: 01 August 2017
• Author: Jae Won Park,Tong Suk Kim,Yuen Jung Park
• DOI: http://doi.org/10.1002/fut.21836

Macroeconomic Conditions and Credit

Default Swap Spread Changes

Tong Suk Kim, Jae Won Park, and Yuen Jung Park*

This study investigates the importance of the business cycle in explaining credit default swap

spread changes by utilizing ex ante proxies. It uses portfolio regression and finds the structural

variables, including the business cycle, explain approximately 65% of the spread differences.

Furthermore, the business cycle variable enhances explanatory power more during the pre- and

post-crisis periods than during the crisis period and shows greater improvement for investment-

grade than for non-investment-grade firms. These results suggest that macroeconomic

conditions play a critical role when the underlying asset value is likely to have greater distance

from the default barrier. © 2017 Wiley Periodicals, Inc. Jrl Fut Mark 37:766–802, 2017

1. INTRODUCTION

Even with the high explanat ory power of structural mode l variables for credit defa ult swap

(CDS) premium levels shown in previous studies,

1

changes in CDS premia are not well

explained by theoretical variables, consistent wi th the findings of Collin-D ufresne,

Goldstein, and Martin (200 1) for bond credit spread chan ges.

2

Zhang, Zhou, and Zhu

(2009) find that the explanator y power of structural factors in C DS spread changes is

around 5%, while Ericsson, Jacob s, and Oviedo-Helfenberger (200 9) and Greatrex (2009)

find that the variables suggest ed by structural models explai n approximately 30% of the

Tong Suk Kim is at Graduate School of Finance and Accounting,Korea Advanced Institute of Science and

Technology,87 Heogiro, Dongdaemoon-gu, Seoul, Korea. JaeWon Park is at Derivatives Trading Department,

Daishin Securities, 34-8 Yeongdeungpo-gu, Seoul, Korea. Yuen Jung Park is at Department of Finance,

College of Business, Hallym University, 1 Hallymdaehak-gil,Chuncheon, Gangwon-Do, Korea. The authors

would like to thank the editor, Robert Webb, Jea Ha Lee, Dohyun Park, Byung Jin Kang, Brad Goldie, Jun

Wang, Chanatip Kitwiwattanachai, and seminar participants at the 2012 annual meeting of the Financial

Management Association, the 2014 Annual Meeting of the Multinational Finance Society, and the 2016

Conference of the Asia-PacificAssociation of Derivatives for their helpful comments and suggestions.We also

thank Byoung-KyuMin for providing us with the Matlab code for the estimation. All remaining errors are ours.

JEL Classification: G12

*Correspondence author, Department of Finance, College of Business, Hallym University, 1 Hallymdaehak-gil,

Chuncheon, Gangwon-Do, Korea. Tel: 82-33-248-1855, Fax: 82-33-248-1804, e-mail: yjpark@hallym.ac.kr

Received December 2016; Accepted December 2016

1

A body of literature has confirmed that structural variables inspired by theory are important determinants of the

variation in CDS spreads. Ericsson et al. (2009) show that the leverage ratio, volatility, and the risk-free rate explain

about 60% of CDS spread levels. Zhang et al. (2009) find that the realized volatility from high-frequency equity prices

predicts around 50% of the variations in CDS premia, while jump risk characteristics such as jump intensity, jump

mean, jump volatility, and jump size forecast 19%. Cao et al. (2010) also provide evidence that credit-related firm-

specific and macro variables, including option-implied volatility, explain approximately 84% of CDS spread

variations.

The Journal of Futures Markets, Vol. 37, No. 8, 766–802 (2017)

© 2017 Wiley Periodicals, Inc.

Published online 2 March 2017 in Wiley Online Library (wileyonlinelibrary.com).

DOI: 10.1002/fut.21836

variation in CDS spread cha nges. Han and Zhou (2015), usin g panel regression, also

report that structural mode l variables explain only 10% of the d ifferences in premia.

Moreover, the results determ ining a common risk factor in CDS p remium changes are

mixed. By utilizing a datase t of 107 firms’CDS spreads from CreditT rade from 1999 to

2002, Ericsson et al. (2009) sh ow that principal component a nalysis of the three factors’

(leverage ratio, equity vol atility, and risk-free rate ) regression residuals pres ents only weak

evidence of a common factor. On the contra ry, by using a dataset of 167 firms’CDS

spreads from January 2002 to March 2009 obtain ed from Bloomberg, Cesare and

Guazzarotti (2010) repor t that a common risk factor drove CDS spread change s during the

crisis period and that pro xies for economic activity, uncertainty, and risk aversion cann ot

explain this systematic risk f actor.

This study, motivated by the empirically insufficient usefulness of structural models in

addition to the mixed findings for CDS spread changes, focuses on the sources of the low

explanatory power of the structural variables for changes in CDS spreads and conjectures

that both the usage of inappropriate proxies for macroeconomic conditions and the pricing

errors of the structural variables can be culprits.

To examine the first source, continuous and ex ante proxies for the business cycle are

used, based on Petkova and Zhang’s (2005) investigation into the relation between time-

varying market risks and value firms’portfolio returns. The authors insist that ex post realized

market returns or gross domestic product (GDP) growth can be a noisy measure for marginal

utility or the business cycle and that the ex ante expected market risk premium should be used

to capture business states. Thus, by using the ex ante proxy instead of real economic variables,

this study investigates whether the business cycle is an important determinant of CDS spread

changes and hypothesizes that the ex ante business cycle variable is related to a common or

systematic risk factor of changes in CDS spreads.

This research hypothesis is based on recent theoretical stu dies of the relation between

credit spreads and the business cycle. Chen (2010) insists it is necessary to endogenously

consider a cyclical market price of risk that increases with the default probability or default

loss in structural models to explain the observed corporate credit spreads. In addition, the

author states that these co-movements require higher credi t risk premia for investment-

grade firms, which could explain the credit spread puzzle, where the proportion of

theoretically estimated credit spreads to observed spreads tends to be much smalle r for

investment-grade than for non-investment-grade firms. Kim and Kim (2005) introduce a

tractable model for the default probability depending on the business cycle, an extension of

Merton’s (1974) model, reflecting the assumption that expected asset returns will be higher

in a bullish market if instantaneous asset returns are proportion al to the production growth

rate. Tang and Yan (2006) construct a theoretical model that expli citly incorporates

equilibrium macroeconomic dynamics into a firm’s cash flow process. They find that firm

characteristics such as cash flow volatility, the current firm-specific growth rate, a nd cash

flow beta have significant effects on credit spreads and that these effects change depending

on economic conditions.

2

The credit spread puzzle that “observed corporate spreads are much larger than what would be predicted by

historical rates of default and recovery rates”(Amato & Remolona, 2003, p. 1) has led to a series of investigations of

the factors impacting the credit spread that cannot be explained by theoretical variables (e.g., Amato & Remolona,

2003; Chen, 2010; Chen, Collin-Dufresne, & Goldstein, 2009; Elton, Gruber, Agrawal, & Mann, 2001). The

determinants of credit spreads have been explored by regressing corporate bond credit spreads on the proxies for

structural variables. A representative study by Collin-Dufresne et al. (2001) finds that the theoretical variables of

structural models have limited explanatory power for credit spread changes. In addition, the authors’principal

component analysis of the regression residuals suggests that monthly credit spread changes are mostly driven by a

single common risk factor, one that cannot be explained by their macroeconomic and financial variables.

Credit Default Swap Spread Changes 767

To examine the second source, that is pricing errors, portfolio-level regressions are

conducted. Eom, Helwege, and Huang (2004) document that the pricing errors of structural

models are systematically relevant to several firm characteristics, such as the leverage ratio and

asset volatility. Furthermore, the authors insist that the leverage ratio has the most significant

effect on pricing errors. Therefore, in this study, portfolios are constructed that are grouped by

leverage ratio and equity volatility ranges to eliminate the errors generated from idiosyncratic risk.

The main empirical results are summarized as follows. First, contrary to the conclusion

of Ericsson et al. (2009) that only three factors are important determinants of CDS premia,

the coefficients of the business cycle variables orthogonal to the three factors are found to

be strongly significant and robust. The business cycle variables are also discovered to

explain a greater part of the variation in CDS spread changes for investment-grade than for

non-investment-grade firms.

Second, through portfolio-level regressions, the structural model variables are found to

explain approximately 65% of spread changes, which is almost twice the explanatory power of

the variable sets of Ericsson et al. (2009). In addition, even after the removal of idiosyncratic

risk, the coefficients of the business cycle variables are substantially significant.

Finally, the business cycle variables—the expected market risk premium, financial

conditions index, and industrial price index—are found to be strongly significant, increasing

explanatory power considerably for the pre-crisis period (e.g., by approximately 3.4%, 9.7%,

and 8.4%, respectively) and the post-crisis period (e.g., by approximately 10%, 11.2%, and

4.6%, respectively). Furthermore, they are robust over the full sample as well as for the pre-

and post-crisis periods, independent of the number of portfolios. However, it should be noted

that, during the crisis period, the three factors of Merton’s (1974) model explain 67% of the

differences in CDS spreads, while the coefficients of the market risk premium, financial

conditions index, and industrial price index show relatively weak significance levels, with

slight incremental explanatory power as well (about 0.5%, 0.1%, and 2.1%, respectively).

Given these results, if CDSs are considered a kind of put option

3

on the asset value, it is

inferred that the main factors affecting CDS price depend on its current moneyness, which is

a mapping to the distance between the current underlying asset value and the default barrier.

In particular, the distance tends to become severely tight during a crisis period and

investment-grade firms tend to have a relatively long distance from default. Therefore, it is

concluded that the factor representing macroeconomic conditions could play a critical role in

pricing CDSs when the underlying their asset value is likely to be farther from the default

barrier, whereas the three factors of Merton’s (1974) model fare well only when the distance

to default is very short.

Most empirical studies of the impacts of macroeconomic conditions on CDS spreads

use real economic variables. For instance, Baum and Wan (2010) find that macroeconomic

uncertainties such as the predicted conditional volatilities of the GDP growth rate, the index

of industrial production, and Standard & Poor’s (S&P) index returns have greater explanatory

power for CDS spread levels than traditional macroeconomic factors such as the risk-free

rate and term spread do. Tang and Yan (2010) find that the average credit spread increases

with a declining GDP growth rate but decreases with an increase in its volatility. In addition,

firms with higher cash flow beta exhibit lower credit spreads than firms with lower cash flow

beta, but this tendency disappears during recessions.

The previous studies mentioned above examine the associations between macroeco-

nomic conditions and CDS spread levels, not CDS spread changes, and generally use GDP

growth, GDP volatility, and investor sentiment as proxies for macroeconomic conditions.

3

Carr and Wu (2011) and Kim et al. (2013) demonstrate the strong linkage between CDSs and deep-out-of-money

put options based on the cash flow replication concept.

768 Kim, Park, and Park