An analysis of sectoral equity and CDS spreads

https://doi.org/10.1016/j.intfin.2014.10.004Get rights and content

Highlights

  • CDS shocks explain the forecast error variance (FEV) of sectoral equity returns.

  • CDS shocks have different effects on equity returns and return volatility.

  • In the post-Lehman crisis period CDS return shocks are the most dominant.

  • A spillover index explains a larger share of total FEV.

Abstract

In this paper, we find that CDS return shocks are important in explaining the forecast error variance of sectoral equity returns for the USA. The CDS return shocks have different effects on equity returns and return volatility in the pre-crisis and crisis periods. It is the post-Lehman crisis period in which the effects of CDS return shocks are the most dominant. Finally, we construct a spillover index and find that it is time-varying and explains a larger share of total forecast error variance of sectoral equity and CDS returns for some sectors than for others.

Introduction

The relationship between equity markets and credit default swap (CDS) spreads is one of the most traditional topics in financial economics, dating back to the work of Merton (1974).1 This relationship becomes relatively more important when markets are exposed to crises. Therefore, unsurprisingly, following the most recent global financial crisis, a number of studies emerged, re-examining the relationship between equity markets and CDS spreads from different perspectives; see, inter alia, Forte and Pena (2009) and Trutwein et al. (2011). Among those works more closely related to our proposed idea, consider first the study by Eichengreen et al. (2012). They examined how the financial crisis which started in the US affected the entire global banking system and extracted common factors in the movement of banks’ CDS. From the common risk factor they showed that credit risk was increasing and it increased significantly between the start of the crisis in July 2007 and the demise of Lehman brothers in September 2008. Second, Grammatikos and Vermeulen (2012) examined the transmission of the 2007–2010 financial and sovereign debt crises to 15 European countries. They found strong evidence that the crisis transmitted from US non-financials to European non-financials but the transmission was not found for the European financial sector. In a related study, Trutwein and Schiereck (2011) found that equity and credit markets (CDS) become more integrated during times of heightened stress; Berndt and Obreja (2010) found that CDS spreads were relatively more important during the financial crisis; and, in a relatively recent study, Avino et al. (2013) examined price discovery of CDS and found its dominance strong in the pre-crisis period.

It is helpful to be clear about what we do in this paper that is missing from the literature. But before we do this, it is equally helpful to remind ourselves about what we have learned so far from the literature with respect to the recent global financial crisis. The literature has documented that credit risk increased significantly, at least from the time of the advent of the crisis in July 2007 to the collapse of Lehman brothers in September 2008, as documented in the work of Eichengreen et al. (2012). Therefore, the resulting questions, which we believe are new, can be summarized as follows: how did this rise in risk (as captured by the CDS spread) affect the equity market? At the sectoral-level (on the S&P500), even though stocks are highly correlated, the covariance of CDS spread returns differs significantly from sector-to-sector, as we show in the next section; therefore, did the CDS spread affect sector returns differently? Then, what about the equity return volatility? Did CDS spreads also affect equity return volatilities differently? And finally: what about spillover of shocks? Did shocks from the CDS spread and sectoral returns have a strong spillover effect; and, if so, has this spillover been time-varying? In this paper, we attempt to answer all of these questions.

We are not first to test for spillover effects. Eichengreen et al. (2012) also do this. However, their approach and objective is completely different. For example, they regressed changes in the CDS spread of a bank on its own lags, common factors, and lags and current changes in CDS spreads of another bank. They found limited evidence of bank CDS spillover effects. In fact, they report that in less than 3% of the 2025 regression models, CDS spread of another bank has spillover effects. Therefore, four differences between the literature and our study are obvious. First, our concern is not banks. We consider sectoral CDS. Second, we consider how CDS-equity return and CDS-equity return volatility spillovers evolve at the sectoral level for stocks of S&P500. Third, we consider how these relationships have changed, if at all, from the pre-crisis to crisis periods. Finally, our approach, because it leads to the construction of a spillover index, allows us to gauge whether spillover is time-varying.

Four key results emerge from answering the proposed questions. First, we find that CDS return shocks are more important in explaining the forecast error variance of sectoral equity returns than sectoral return volatility. CDS return shocks explain between 10 and 20% of the forecast error variance of equity returns of financial, consumer discretion, materials, and energy sectors. Second, we find that CDS shocks have different effects on equity returns and return volatility in the pre-crisis and crisis periods. Third, we show while CDS return shocks are the most instrumental in explaining sectoral equity returns in the crisis period, it is the post-Lehman crisis period in which the effect of CDS returns is the most dominant. In the post-Lehman crisis period, CDS return shocks explain between 22 and 28% of the forecast error variance of equity returns of the financial, industrial, and materials sectors. Fourth, we construct a spillover index and find that it explains a large share of total forecast error variance of sectoral equity and CDS returns for some sectors (materials, industrial, financial, and consumer discretion) but not for all sectors. This evidence is stronger during the post-Lehman crisis period. Moreover, we document sectoral spillover indices that are time-varying, characterized by multiple cycles and peaks during the recent global financial crisis.

The rest of the paper is organized as follows. In the next section, we discuss the estimation technique. In particular, we explain how we compute the forecast error variance and the spillover effects from a vector autoregressive (VAR) model. We conclude the section with a brief description of how we compute the volatility of stock returns using the proposal of Schwert (1989a). Section 3 discusses the data and the results. A robustness test is undertaken in Section 4, while Section 5 interprets the results. The final section concludes the paper.

Section snippets

Forecast error variance and spillover effects

The forecast error variance and spillover effects, based on the vector autoregressive (VAR) model, are proposed by Diebold and Yilmaz (2009), therefore, this section draws heavily on their work. We have a two-variable VAR model. The two variables are sectoral equity return (Rt) and CDS spread return (CDSRt). Let us denote by Zt the vector, Zt=[Rt,CDSRt]. Then, the first-order VAR model for Zt is: Zt=AZt1+kt, where A is a 2 × 2 parameter matrix. The moving average representation of Zt is: Zt=B(L)η

Data

The equity price and CDS spread data are obtained from the Bloomberg system.3 For the CDS spread, we only consider the 5-year tenor series contracts as these instruments are known to have adequate

A robustness test

In this section, we test the robustness of our results on two fronts. The two issues that have direct relevance for our empirical results in this paper are: (i) the forecasting horizon at which the forecast error variance and spillover index are computed, and (ii) the measure of stock return volatility. Let us consider the forecasting horizon first. So far, we have considered a long horizon forecasting framework by setting h = 30 days. We believe that it is imperative to test whether results are

A discussion of the results

The literature has produced mixed results on whether it is the CDS returns that lead equity returns, or the equity returns that lead CDS returns. Acharya and Johnson (2007) show that information flow from the CDS market to the stock market is greater for sub-samples where insider trading is more prominent. Blanco et al. (2005) use a sample of high-grade credits and find mixed results in that, in some cases, they find greater price discovery in the CDS market than in the bond market while, in

Concluding remarks

In this paper, motivated by the rise in risk as measured by the CDS spread, we study the dynamic relationship between (a) sectoral equity returns and CDS spread returns, and (b) sectoral equity volatility and CDS spread returns using a VAR model. We consider 10 sectors of the S&P500 for which long time-series data on CDS spreads are available. We notice that CDS returns, like equity returns, differ significantly from one sector to the other, suggesting that sectoral riskiness varies. We,

References (35)

  • D.B. Nelson

    Filtering and forecasting with mis-specified ARCH models I

    J. Economet.

    (1992)
  • H.H. Pesaran et al.

    Generalised impulse response analysis in linear multivariate models

    Econ. Lett.

    (1998)
  • X. Pu et al.

    Correlation in credit risk changes

    J. Bank. Finance

    (2012)
  • J. Qiu et al.

    Endogenous liquidity in credit derivatives

    J. Financ. Econ.

    (2012)
  • P. Trutwein et al.

    The fast and the furious – stock returns and CDS of financial institutions under stress

    J. Int. Financ. Markets Inst. Money

    (2011)
  • A. Berndt et al.

    Decomposing European CDS returns

    Rev. Finance

    (2010)
  • A. Berndt et al.

    Do equity markets favour credit market news over options market news?

    Working Paper

    (2008)
  • Cited by (0)

    View full text