Forecasting high‐yield equity and CDS index returns: Does observed cross‐market informational flow have predictive power?

• Published date: 01 August 2022
• Author: William J. Procasky,Anwen Yin
• Date: 01 August 2022
• DOI: http://doi.org/10.1002/fut.22342

Received: 30 September 2021

|

Accepted: 29 April 2022

DOI: 10.1002/fut.22342

RESEARCH ARTICLE

Forecasting high‐yield equity and CDS index returns: Does

observed cross‐market informational flow have predictive

power?

William J. Procasky

1

|Anwen Yin

2

1

College of Business, Texas A&M

University Kingsville, Kingsville,

Texas, USA

2

WHT 217B, A.R. Sanchez, Jr. School of

Business, Texas A&M International

University, Laredo, Texas, USA

Correspondence

William Procasky, College of Business,

Texas A&M University Kingsville,

Kingsville, TX 78363, USA.

Email: william.procasky@tamuk.edu

Funding information

College of Business Administration,

Texas A&M University‐Kingsville,

Grant/Award Number: Research Grant

Abstract

We examine the predictive power of cross‐market informational flow in the

systematic high‐yield credit default swap (CDS) and equity markets from 2004

to 2019. Overall, we find both markets useful in forecasting future values of the

other, indicating each is more efficient in pricing certain types of information.

However, the CDS market has an informational advantage over the high‐yield

and broader equity market, something not previously documented in the closely

related literature, although the advantage and forecasting ability of these

markets have decreased with time due to lower volatility. These results have

implications for high‐yield investors and stakeholders who monitor Markit's

CDX North American High‐Yield Index for informational content.

KEYWORDS

CDS indexes, credit derivatives, forecasting, market efficiency

JEL CLASSIFICATION

G11, G12, G14, G17, G23

1|INTRODUCTION

A significant body of historical literature has investigated whether credit default swap (CDS) or equity markets have an

informational advantage over the other. While under the model proposed in Merton (1974), this should not occur since

the same factors impact prices in both markets, that is, they should both react the same way to the same news, in reality

frictions and other market‐related factors give rise to inefficiencies (see Duffie, 1999). As a result, it is possible for one

market to have an informational advantage over the other. Longstaff et al. (2004), Acharya and Johnson (2007), Han

and Zhou (2011), Marsh and Wagner (2012), Narayan et al. (2014), Hilscher et al. (2015), and Acharya and Johnson

(2017) broadly studied the North American market while Norden and Weber (2004), Norden and Weber (2009), and

Forte and Pena (2009) examined international markets. Ni and Pan (2004) and Rodriguez‐Moreno and Pena (2013)

focused on the financial sector while Bystrom (2006), Fung et al. (2008), and Procasky (2021) used indexes to

investigate systematic flow. In this paper we contribute to the literature by providing further empirical evidence on this

issue.

While most historical studies such as Narayan et al. (2014) have focused on nonsystematic, or single name CDS and

generally observed that the equity market has an overall advantage over the CDS (although results have been

somewhat mixed), as stated, Procasky (2021) uses CDS indexes and matched equity portfolios, and observes default risk

based heterogeneity in the efficiency and cross‐market informational flow of systematic markets. (The reasons

J Futures Markets. 2022;42:1466–1490.wileyonlinelibrary.com/journal/fut1466

|

© 2022 Wiley Periodicals LLC.

supporting the use of CDS indexes are thoroughly discussed in Procasky, 2021 and Fung et al., 2008, for example,

these indexes are efficiently bundled packages of diversified credit risk, thus resulting in lower transaction costs than

firm‐specific swaps). Specifically, while neither market has an advantage in the investment‐grade sector, suggesting

they are equally efficient, a strong two‐way interactive effect is observed in the high‐yield market. Building on this

result, we extend the research by examining whether the reported in‐sample evidence of significant two‐way flow

documented in Procasky (2021) can be translated into meaningful predictive power. As is well‐known in the financial

time‐series forecasting literature, in‐sample evidence of one variable's (X) statistical relationship with futures values

of another (Y) does not necessarily mean that Xis useful in forecasting futures values of Y. Therefore, evidence

of predictive content embedded in the observed cross‐market informational flow would complement and build on

in‐sample results reported in prior studies, such as Procasky (2021).

1

In addition, our research is motivated by the fact

that, while the vector autoregressive regression (VAR) models used in such studies are effective in detecting cross‐

market informational flow, conventional statistical tests are not very useful in quantifying the relative strength of that

flow. This is because the coefficients on lagged price variables in the endogenous system of equations do not

correspond to the magnitude of subsequent price movements. As a result, researchers are left to categorize observed

cross‐market flow in broad statistical terms, that is, either as “strong”or “weak,”based solely on the pvalues associated

with these lagged terms. Thus, while the prior results suggested that the observed two‐way informational flow in the

high‐yield sector is strong, there is no way to tell whether flow from one subject market to the other may be more

persistent and/or significant.

Our idea is simple: If the observed cross‐market informational flow in the high‐yield sector is significant, then we

would expect the model which takes into account such flow to deliver greater predictive gains than models which ignore

it. The stronger the cross‐market informational flow is, the greater the forecasting gains should be. Using daily systematic

high‐yield CDS index return data defined as the percentage change in the index, and closely matched equity portfolio

returns which we painstakingly constructed, as well as S&P 500 index return data, we perform a forecasting analysis with

VAR models to determine the true predictive power, if any, of the observed cross‐market informational flow. While

different methods have been used in the literature to calculate CDS returns, with innovation in this area ongoing, for

example, Augustin et al. (2020), we follow Hilscher et al. (2015) for historical comparison purposes. In doing so, we pull

from the literature on VAR model forecasting and forecast evaluation for the first time in an analysis of cross‐market

informational flow. West (2006) and Clark and McCracken (2013a) provide a comprehensive overview of forecast

construction, estimation, inference, and evaluation. Such methods have been widely utilized in empirical finance, in

particular the field of forecasting market equity returns, for example by Goyal and Welch (2008), Rapach et al. (2010),

Pettenuzzo et al. (2014), J. Li and Tsiakas (2017), and Yin (2019). Forecasting ability analysis has also been used

extensively in examining the predictive performance of various economic fundamentals‐based models in the foreign

exchange (FX)markets, for example, byJ. Li et al. (2015), Beckmann and Schussler (2016), and Jamali andYamani (2019).

Moreover, we use a battery of classic and contemporary methods to analyze the robustness and relative strength of

any predictive flow and to ascertain whether one market has a greater informational advantage than the other.

Specifically, we first assess forecast rationality by using the forecast unbiasedness test originally proposed in Mincer

and Zarnowitz (1969). Recent applications and extensions of the forecast rationality tests can be found in studies, such

as Gurkaynak et al. (2013). We then judge the overall quality of forecasts using a robust set of statistics which draw on

insights from seminal methodological works by Diebold and Mariano (1995), Giacomini and Rossi (2009), Giacomini

and Rossi (2016), Diebold (2015), and Rossi and Sekhposyan (2016). To gain further insights, we also investigate

separately cross‐market informational flow under different investor sentiment regimes, as the importance of regime‐

dependent evaluation in empirical works is emphasized in studies, such as Baltas and Karyampas (2018). Finally, given

that the traditional statistical significance test does not provide a direct measure to quantify the observed causality in

the sense of Granger causality, using the innovative methods proposed in Dufour et al. (2012) we measure and report

the strength of cross‐market informational flow in the forecasting framework.

Interestingly, our analysis reveals that lagged values in both markets are useful in forecasting future values in the

other. In addition, we document that the high‐yield CDS market has an informational advantage over the equity in the

strength of this predictive power. Viewed intandem, these results have significant implications forinvestors, arbitrageurs,

and risk managers who monitor systematic markets for informational content.

1

The issue of disconnection between in‐sample causality and predictability is discussed and illustrated in works, such as Goyal and Welch (2008) and

Rapach and Zhou (2013).

PROCASKY AND YIN

|

1467