Oil and stock markets before and after financial crises: A local Gaussian correlation approach

• DOI: http://doi.org/10.1002/fut.21860
• Author: Theodore Panagiotidis,Georgios Bampinas
• Published date: 01 December 2017
• Date: 01 December 2017

Received: 16 December 2016

|

Accepted: 13 April 2017

DOI: 10.1002/fut.21860

RESEARCH ARTICLE

Oil and stock markets before and after financial crises: A local

Gaussian correlation approach

Georgios Bampinas

|

Theodore Panagiotidis

Department of Economics, University of

Macedonia, Thessaloniki, Greece

Correspondence

Theodore Panagiotidis, Department of

Economics, University of Macedonia, 156

Egnatia Street. 54006, Thessaloniki, Greece.

Email: tpanag@uom.edu.gr

The effect of financial shocks on the cross-market linkages between oil prices (spot and

futures) and stock markets is examined for four major crises. We employ the local

Gaussian correlation approach and find that the two markets were regionalized for most

of the 1990s and the early 2000s. Flights from stocks to oil occur in all crisis episodes,

except the recent global financial crisis. The view that stock and oil markets behave like

“a market of one”after the financialization of commodities is further supported by the

presence of contagion between US stock markets and all the benchmark oil markets.

JEL CLASSIFICATION

F3, G01, G10

1

|

INTRODUCTION

The rapid process of financialization of commodity markets, has made crude oil prices more susceptible to financial market

conditions and unforeseen events(Büyükşahin & Robe, 2014; Tang & Xiong, 2012).

1

Although financial crises are not a new

phenomenon, the global financial crisis of 2007–2009 has emphasized how highly exposed energy markets are to financial

shocks. As a consequence, investors have been questioning previously held beliefs about the risk associated with investing in oil

and the expected benefits of diversification (Adams & Glück, 2015; Daskalaki & Skiadopoulos, 2011). Most of the empirical

studies thus far have focused on the effect of an oil shock on stock returns (see, inter alia, Huang, Masulis, & Stoll, 1996; Jones &

Kaul, 1996; Kilian, 2008; Kling, 1985; Wei, 2003), though less attention has been given to the changes in cross-market linkages

after a market-specific shock. This paper focuses on four major financial crises and aims to address the question of whether the

cross-market linkages between oil and stock markets increase after an adverse financial shock.

Of particular interest is whet her those linkages can be attrib uted to contagion. Tang and Xio ng (2012) argue that the vast

inflows of institutional investo rs over the last decades led to a process of integration of com modity futures markets

with financial markets in wh ich portfolio rebalancin g of index investors can caus e volatility spillovers fr om external to

commodity markets. Moreo ver, the risk transfer from the f inancial to the commodity sec tor and the role of speculati on in

futures markets with respect to crude oil and ene rgy prices are additional reasons to expe ct higher interconnectedness

between stocks and crude oil.

2

An intuitive starting poi nt to measure this potenti al increase in interdepend ence could be

looking at simple correlation s between the two markets. Some of th e first studies on testing for contag ion focused on the

cross-market correlatio ns before and after a shock (see , for instance, Bertero & May er, 1989; King & Wadhwani, 199 0). A

key methodological contribu tion is Forbes and Rigobon (2002), whe re it is shown that contagion is over-acce pted, if the

1

The term financialization is defined as the increasing dominance of the finance sector and the expanding role of financial motives in the overall economy.

During the 2000s, flows in commodity investments increased from $15 billion in 2003 to $250 billion in 2009 (Irwin & Sanders, 2011).

2

See for instance the United States Congress and Senate, Committee on Homeland Security and Governmental Affairs. Permanent Subcommittee on

Investigations (2006).

J Futures Markets. 2017;37:1179–1204. wileyonlinelibrary.com/journal/fut © 2017 Wiley Periodicals, Inc.

|

1179

heteroscedasticity bias is ignored (when testing fo r changes in correlation).

3

In addition, changes in (inter-) depend ence in a

crisis often show that the depe ndence of assets or markets is no nlinear and asymmetric (Bau r, 2013). Bodart and Candelon

(2009) argue that contagion is intrinsically asymm etric, whereas, Bae, Karoly i, and Stulz (2003) assert that linear measures

such as correlation are inappr opriate if contagion is nonli near in nature.

In this study, we focus on four major financial crises and investigate the effect that financial shocks have on the degree of

dependence between stock markets and the spot and futures oil markets. This is done by using a local measure of correlation that

can account for the nonlinear nature of the relationship. More specifically, we employ the local Gaussian correlation approach

introduced by Tjøstheim and Hufthammer (2013) to examine the nonlinear dependence between stock and oil markets during (i)

the Mexican “Tequila”crisis; (ii) the Asian “flu”crisis; (iii) the dot.com bubble; and (iv) the 2007–2009 financial crisis. This

framework is in conceptual proximity to the traditional correlation analysis (see, e.g., Baig & Goldfajn, 1998; Calvo & Reinhart,

1995; King & Wadhwani, 1990) but differs in several respects.

The local Gaussian correlation is a dependence measure that is localized and nonlinear. Theoretically, it gives a much more

complete description of dependence relationships. In contrast to linear dependence measures, like the Pearson correlation

coefficient, it avoids the bias of the conditional correlation, and also provides a way of describing any nonlinear structure in

dependence and the departure from global normality. Moreover, the local Gaussian correlation does not suffer from the bias

problem which stems from the increased volatility observed in unstable periods. It can also capture the asymmetric response that

is derived from the effect of financial shocks of different magnitudes on oil prices. It measures lower, average, and upper tail

dependence and as a result enhances our understanding of the dependence structure between stock and oil markets in the different

segments of the distribution. Contagion could also be tested by comparing the local correlation of the stable and the crisis period.

By using the traditional correlation concept as our preferred contagion metric, this is done within a bootstrap procedure (for an

application in financial markets see Støve, Tjøstheim, & Hufthammer, 2014). Conceptually, testing for contagion implies that

cross-market linkages increase because of a crisis in the financial sector of a specific country that is the origin.

Our paper is also related to the portfolio management literature, where it is well documented that the dependence or correlation

between various asset classes may be distribution and/or market condition specific.

4

For example, the recent subprime crisis

exacerbated the shortage of real assets in the world economy, which triggered a bubble in commodit ies and oil markets (Caballero,

Farhi, & Gourinchas, 2008). The idea of using crude oil as a diversification tool for financial assets has attracted the attention of

investors mainly because of the remarkable rise in crude-oil prices over the past few years and the low (or negative) correlation

between crude oil and stock markets (Büyükşahin, Haigh, & Robe, 2010; Fratzscher, Schneider, & Van Robays, 2014; Geman &

Kharoubi, 2008; Gorton & Rouwenhorst, 2006).

5

A higher level of dependence implies that benefits from market-portfolio

diversification diminish.

6

If oil prices increase in crisis periods on the other hand, the losses incurred with global stock market

investments are partially compensated. The flight-to-alternative investments from stocks to oil can lead to such compensation. If

investors flee from stocks into oil then we observe a flight-to-alternative assets. If the co-movement of stocks and oil turns negative in

crisis periods, investors that hold both stocks and oil will suffer less as one asset class reduces the losses by providing positive returns.

This effect is important for risk management and the performance of international portfolios. The present work complements the

literature by offering further evidence on the benefits of diversification between oil and equities, conditional on the degree of

dependence in the different segments of the joint distribution, and under different market circumstances (stable vs. turmoil period).

7

Our empirical findings show that flights-to-alternative assets from stocks to oil are a common feature during three crisis periods

(the Mexican crisis, the Asian crisis, and the dot.com bubble). They occur in extreme market conditions and are more profound in a

market booming scheme (right tail). A different picture emerges during the global financial crisis of 2007–2009. We observe

important increases in correlations throughout the conditional distribution, with those correlations being more profound when

markets are in a state of financial distress (left tail). Moreover, after a common shock in the US stock market, we demonstrate the

existence of contagion—in the Forbes and Rigobon (2002) sense—between the S&P500 and the crude oil markets (both spot and

3

Similar studies that identify contagion through increased bivariate correlations during stress periods include Favero and Giavazzi (2002) and Corsetti,

Pericoli, and Massimo Sbracia (2005).

4

Støve and Tjøstheim (2013) demonstrated that in the case of financial data there is stronger dependence in the left tail.

5

For related studies that consider commodities in portfolio diversification and risk management see, for example, Cheung and Miu (2010), Daskalaki,

Kostakis, and Skiadopoulos (2014), Silvennoinen and Thorp (2013), Creti, Joets, and Mignon (2013), Hammoudeh, Nguyen, Reboredo, and Wen (2014),

Cheung and Miu (2010).

6

Correlation between financial assets increases as markets fall. “Correlation breakdown”is the phenomenon where correlation increases during a crisis.

7

Büyükşahin et al. (2010) employ data from 1991:01 to 2008:11 and provide evidence of weak comovements between commodities and equities.

However, these do not lead to diversification benefits.

1180

|

BAMPINAS AND PANAGIOTIDIS