On the time‐varying links between oil and gold: New insights from the rolling and recursive rolling approaches

• Author: Zeynel Abidin Ozdemir,Muhammad Shahbaz,Mehmet Balcilar
• Published date: 01 July 2019
• Date: 01 July 2019
• DOI: http://doi.org/10.1002/ijfe.1704

RESEARCH ARTICLE

On the time‐varying links between oil and gold: New

insights from the rolling and recursive rolling approaches

Mehmet Balcilar

1,2,3

| Zeynel Abidin Ozdemir

4

| Muhammad Shahbaz

2

1

Department of Economics, Eastern

Mediterranean University, Famagusta,

Turkey

2

Energy and Sustainable Development,

Montpelier Business School, Montpelier,

France

3

Department of Economics, University of

Pretoria, Pretoria, South Africa

4

Department of Economics, Gazi

University, Ankara, Turkey

Correspondence

Zeynel Abidin Ozdemir, Gazi University,

Ankara, Turkey.

Email: zabidin@gazi.edu.tr

JEL Classification: C22; E31; Q20

Abstract

This study analyses the dynamic linkages between oil and gold prices for the

spot and 1‐to 12‐month futures markets using monthly data over the period

1983–2016. To do this, we use the rolling and recursive rolling Granger causal-

ity approaches. The distinguishing feature of this study from the previous stud-

ies is that this is the first study investigating the causal links between oil and

gold using time‐varying causality tests. The findings show that the causality

links between oil and gold display strong time variation. Although causal links

are not detected for most of the study period, strong bidirectional or unidirec-

tional causality is found in several subsamples. The duration of the periods

with causality links varies from a few months to 3 years, whereas the duration

for the noncausality periods might be 15 years long. By date stamping the cau-

sality links between oil and gold, our paper discovers that causality from oil to

gold is related to large oil price changes, whereas causality from gold to oil is

related to large financial crises. The evidence obtained in the paper points

out the dangers of assuming a constant causality link between oil and gold

markets because these links might break down unexpectedly. Our findings

point out to the dangers of assuming noncausality between oil and gold partic-

ularly in hedging oil price risk using gold.

KEYWORDS

gold and oil prices, recursive rolling estimation, rolling estimation,time‐varying Granger causality

1|INTRODUCTION

Oil and gold, which are frequently tradable and have high

liquidity and synchronization in their movements, are the

two commodities with an irreplaceable role in an econ-

omy. Oil is the most traded commodity in the world

and has a highly volatile price. Gold, on the other hand,

with the lowest volatility, is the leader of all precious

metals. When stock market risks are considered, gold is

known as a safe haven and used as an efficient hedge

instrument (Junttila, Pesonen, & Raatikainen, 2018). As

oil has been increasingly becoming financialized,

whether gold does also function as an efficient hedge

instrument against oil price risk is interest to financial

industry. A long‐term analysis of the behavioural ten-

dency of oil and gold demonstrates that both commodi-

ties historically tend to move simultaneously upward

and downward in price, and a change in the price of

one appears to induce a change in the price of the other.

An observation of the two commodities for the last

50 years, for instance, has shown that they tend to behave

in similar ways with respect to price, with a positive

correlation of 80% (see Tiwari & Sahadudheen, 2015)—a

view that is also supported by a large number of studies.

Accordingly, Sari, Hammoudeh, and Soytas (2010) con-

tend that changes in the price of gold mainly stem from

Received: 21 February 2018 Revised: 5 July 2018 Accepted: 9 September 2018

DOI: 10.1002/ijfe.1704

Int J Fin Econ. 2019;24:1047–1065. © 2018 John Wiley & Sons, Ltd.wileyonlinelibrary.com/journal/ijfe 1047

fluctuations in the price of oil. Hence, this study empiri-

cally investigates the time‐varying Granger causality

between the oil and gold markets.

The relationship between the oil price and gold price

can arise from several channels. Most oil‐importing coun-

tries' preference to pay for their oil supplies in gold and

the investment of a large part of oil‐exporting countries'

revenues in gold are just a few of the examples to support

this suggestion. Still another example is that an increase

in oil price means an increase in the cost of gold extrac-

tion, and this results in a reduction in the profit margin.

It may be said that oil prices are inversely proportional

to the share prices of gold mining companies.

Another channel—indeed, the best one according to

Narayan, Narayan, and Zheng (2010)—used to explain

the link between oil and gold stems from inflation.

Accordingly, an increase in the international crude oil

price leads to an increase in general price levels due to

an increase in transportation and production costs. This

causes negative effects on oil importing countries. There-

fore, there is a positive relationship between oil price and

inflation. Because gold is a unique instrument of inflation

hedging in the long term, investors usually invest in gold

during periods of increased inflation to balance their

portfolio (Ghosh, 2011). The gold price will rise in these

periods of high inflation; thus, positive relationships will

be observed between oil and gold. The World Gold Coun-

cil holds the view that gold has always held itself against

inflation throughout history. A higher oil price, on the

other hand, is likely to fit the bill of countries—net oil‐

importing countries in particular—and this causes an

increase in import costs, which causes a high trade

deficit; this in turn influences the value of the domestic

currency and the money in circulation, and all these

factors lead to inflation—a chain reaction. In this case,

gold will increase in response to inflation, which is

caused by rising oil prices. Gold, on the other hand, might

increase when the goods and financial markets are

strong, proxying the market conditions. Rising markets

will also lead to higher oil demand, inducing an upward

movement in oil prices. As financialized commodities,

such as gold, respond faster than goods markets, this

may generate predictive power from gold to oil. However,

the most significant causality from gold to oil should be

observed when investors shift large amounts of funds

between the gold and oil markets. Such shifts occur dur-

ing large financial crises.

Few studies examine oil and gold prices and their rela-

tionship with macroeconomic and financial variables in

the literature. However, researchers were more interested

in studying the interrelations between oil and gold follow-

ing the new economic crisis due to the increases in the

prices of these commodities and the common use of gold

as a safe haven for their investments. The pioneering

study in this respect was the one conducted by Melvin

and Sultan (1990), in which the researchers found a

strong correlation between oil and gold due to the export

revenue channel. Another study was performed by Kim

and Dilts (2011), where they had similar observations.

Other studies, including Soytas, Sari, Hammoudeh, and

Hacıhasanoglu (2009), Liao and Chen (2008), Sari,

Hammoudeh, and Ewing (2007), Hammoudeh and Yuan

(2008), Narayan et al. (2010), Šimáková (2011), Le and

Chang (2011b), and Lee, Huang, and Yang (2012), how-

ever, did not find evidence of the relationship between the

movement of the prices of oil and gold. Shahbaz, Balcilar,

and Ozdemir (2017), using a nonparametric causality‐in‐

quantiles test, showed that the oil price has weak predic-

tive power for the gold price, and the causality‐in‐variance

tests found strong support for the predictive capacity of oil

for gold market volatility. According to others, the prices

of oil and gold act simultaneously because they are corre-

lated with the movement of their long‐term driving fac-

tors, such as volatility in U.S. dollars and the turmoil in

the international politics (e.g., Bampinas & Panagiotidis,

2015; Le & Chang, 2011a). This simultaneous movement

can generate dynamic causality links between the oil

and gold markets.

With respect to literature's main conclusions

concerning on the investigation of the dynamic nexus

between gold market and oil market, the existing studies

employing the methods fall into two groups: those that

use a linear Granger causality and non‐linear Granger

causality methods and those that use causality‐in‐quantile

method. First group studies examining the causality

between gold market and oil market using these methods

have been conducted by Zhang and Wei (2010), Jain and

Ghosh (2013), Bildirici and Turkmen (2015), Jain and

Biswal (2016), Kumar (2017), Kanjilal and Ghosh (2017),

Gil‐Alana, Yaya, and Awe (2017), Bilgin, Gogolin, Lau,

and Vigne (2018), Sephton and Mann (2018), and Mei‐Se,

Shu‐Jung, and Chien‐Chiang (2018). For instance,

Bildirici and Turkmen (2015), Jain and Biswal (2016),

Kumar (2017), Kanjilal and Ghosh (2017), Mei‐Se et al.

(2018), and Sephton and Mann (2018) show that there is

a bidirectional causality between crude oil prices and gold

prices. But Zhang and Wei (2010) indicate that the crude

oil price change linearly Granger causes the volatility of

gold price, but not vice versa. Second group studies investi-

gating the causality between both markets using causality‐

in‐quantile method have been showed by Shahbaz et al.

(2017) and Das, Kumar, Tiwari, Shahbaz, and Hasim

(2018). Shahbaz et al. show that oil price has a weak pre-

dictive power for the gold price using the nonparametric

causality‐in‐quantiles test. Furthermore, the study indi-

cates strong support for the predictive capacity of oil for

1048 BALCILAR ET AL.