Modeling and forecasting commodity market volatility with long‐term economic and financial variables

• DOI: http://doi.org/10.1002/for.2617
• Date: 01 March 2020
• Author: Thomas Walther,Duc Khuong Nguyen
• Published date: 01 March 2020

Received: 9 February 2018 Revised: 3 December 2018 Accepted: 12 June 2019

DOI: 10.1002/for.2617

SPECIAL ISSUE ARTICLE

Modeling and forecasting commodity market volatility with

long-term economic and financial variables

Duc Khuong Nguyen1,2 Thomas Walther3,4

1IPAGLab, IPAG Business School, 184

Boulevard Saint-Germain, 75006, Paris,

France

2School of Public and Environmental

Affairs, Indiana University, 107 S Indiana

Ave, Bloomington, IN 47405, USA

3Institute for Operations Research and

Computational Finance, University of St.

Gallen, St. Gallen, 9000, Switzerland

4Faculty of Business and Economics,

echnische Universität Dresden, Dresden,

01062, Germany

Correspondence

Thomas Walther,Institute for Operations

Research and Computational Finance,

University of St. Gallen, 9000 St. Gallen,

Switzerland.

Email: thomas.walther@tu-dresden.de

Funding information

Innosuisse, Grant/AwardNumber:

SCCER CREST

Abstract

This paper investigates the time-varying volatility patterns of some major

commodities as well as the potential factors that drive their long-term volatility

component. For this purpose, we make use of a recently proposed generalized

autoregressive conditional heteroskedasticity–mixed data sampling approach,

which typically allows us to examine the role of economic and financial variables

of different frequencies. Using commodity futures for Crude Oil (WTI and

Brent), Gold, Silver and Platinum, as well as a commodity index, our results

show the necessity for disentangling the short-term and long-term compo-

nents in modeling and forecasting commodity volatility. They also indicate

that the long-term volatility of most commodity futures is significantly driven

by the level of global real economic activity as well as changes in consumer

sentiment, industrial production, and economic policy uncertainty. However,

the forecasting results are not alike acrosscommodity futures as no single model

fits all commodities.

KEYWORDS

commodity futures, GARCH, long-term volatility, macroeconomic effects, mixed data sampling

1INTRODUCTION

Earlier studies on commodity markets have shown that

commodity futures can be a valuable source of diver-

sification benefits for investors and portfolio managers,

given their distinct risk–return characteristics as com-

pared to traditional assets like bonds and stocks.Bodie and

Rosansky (1980) noted, for example, that their benchmark

portfolio of commodity futures performed as well as the

portfolio of common stocks in terms of average returns

over the period 1950–1976. More importantly, a diversi-

fied portfolio of 60% stocks and 40% commodity futures

led to a return variability reduction of about one-third rel-

ative to the 100% stock portfolio, while having the same

level of return. The hedging ability against inflation is

another interesting feature of commodity futures (Lucey

et al., 2017). Similarly, Lintner (1983) found that the vari-

ability of portfolios of stocks and bonds was consistently

lower when they were combined with managed com-

modity futures. More recent studies such as (Gorton &

Rouwenhorst, 2006), Arouri, Jouini, and Nguyen (2011),

(Narayan et al., 2013), and (Klein, 2017) also found evi-

dence to confirm this diversifying potential of commodity

futures through the use of various data sets and evaluation

methods. The specific drivers of commodity returns as well

as their low correlations with stocks and bonds can thus be

viewed as the key factors that explain the increasing role of

commodity futures in portfolio investments and diversifi-

cation strategies (Bekiros et al., 2017; Domanski & Heath,

2007; Dwyer et al., 2011).

With the intensification of their financialization since

2004, commodity markets are exposed to some structural

changes in the distributional characteristics of returns

and dependence with other asset classes. Commodity

futures returns now behave more like stock returns, and

their correlation with stocks has become positive and

Journal of Forecasting. 2020;39:126–142.wileyonlinelibrary.com/journal/for© 2019 John Wiley & Sons, Ltd.126

increased in recent years, particularly after the collapse

of Lehman Brothers (Adams & Glück, 2015; Büyük¸sahin

& Robe, 2011; Daskalaki & Skiadopoulos, 2011; Tang &

Xiong, 2012) As a result of this increasing equity-like

behavior, researchers find evidence of lower diversifica-

tion benefits associated with the inclusion of commodity

futures in diversified portfolios and a higher level of their

shock transmission and volatility spillovers with stocks

(Baur & McDermott, 2010; Daskalaki & Skiadopoulos,

2011; Filis et al., 2011; Narayan & Sharma, 2011; Silven-

noinen & Thorp, 2013)

The large fluctuations in commodity prices over recent

years have also generated concerns for macroeconomic

stability and overall economic performance. The standard

deviation of the IMF All Commodity Price Index over the

period 2005M1–2017M6 is 36.45%. The same price index

also reached the highest value of 220.03 index points in

July 2008 (base index of 100 points in 2005), or an increase

of 120%. Since information concerning volatility is a criti-

cal input for portfolio design, risk management and policy

decisions (i.e., volatility directly affects cross-asset corre-

lation and portfolio risk level), an important strand of

the commodity finance literature has devoted attention to

commodity volatility modeling and the identification of its

determinants. A general consensus from the majority of

past studies is that main volatility drivers tend to differ

across different classes of commodities.

For instance,(Daskalaki et al., 2014) attempted to iden-

tify common factors for the pricing of commodities. They

concluded that neither macroeconomic, equity-related,

nor commodity-specific factors could explain the pric-

ing over all commodity classes. Batten et al. (2010) ana-

lyzed the macroeconomic drivers of monthly precious

metal volatility and documented that monetary (e.g., infla-

tion) and financial (e.g., S&P 500 returns) variables could

explain the volatility block-wise, but their results did not

hold for Silver. Moreover, the drivers of volatility within

the group of precious metals are not alike. Silvennoinen

and Thorp (2013) analyzed the correlation of commodities

and found lagged VIX to have a positive impact on weekly

energy volatility, but no impact on precious metals.

Regarding the energy market volatility, Pindyck (2004)

documented that macroeconomic variables such as Trea-

sury Bill yields or effective exchange-weighted dollar rate

did not affect oil price volatility using weekly data.Kilian

and Vega (2011) found evidence that WTI oil price returns

were not sensitive to macroeconomic news. Karali and

Ramirez (2014) used macroeconomic variables, political

and weather events to identify drivers of crude oil, heating

oil, and natural gas futures volatility. Their results indi-

cated that only crude oil's volatility increased following

political, financial, and natural events, whereas macroe-

conomic variables had no significant impact on oil price

volatility. A recent study by Yin and Zhou (2016) showed

that economic policy uncertainty spilled over to oil price

spot and futures volatility.

Nevertheless, several studies empirically uncover com-

mon volatility links among commodity classes. The work

of (Verma, 2012) shows, for example, the negative influ-

ence of sentiment on the volatility of energy and precious

metal futures. Considering a sample of agricultural,

energy, and metal commodities, (Karali & Power, 2013)

found evidence of significant influences of inflation and

industrial production on commodity markets’ long-term

volatility. (Smales, 2017) documented that the volatility

of commodity markets, represented by the Commodity

Research Bureau Index and the S&P Goldman Sachs Com-

modity Index, reacted to both US and Chinese macroe-

conomic news, including US employment and economic

output as well as the purchasing intentions of Chinese

manufacturers. Lastly, (Prokopczuk et al., 2017) investi-

gated the comovement of commodity market volatility

and economic uncertainty via regression with realized

volatility and found that certain macroeconomic and

financial variables (i.e., inflation volatility, VIX, default

return spread and TED spread) drove the monthly com-

modity volatility. The authors suggested scrutinizing the

issue further through the framework proposed by Engle

et al. (2013), which combines generalized autoregres-

sive conditional heteroskedasticity (GARCH; Bollerslev,

1986; Engle 1982) models with the mixed data sampling

(MIDAS; Ghysels et al. 2004; Ghysels et al. 2007) tech-

nique. This combination particularly allows one to use

macroeconomic variables, usually available at monthly

or quarterly frequency, as explanatory variables of daily

volatility.

The GARCH-MIDAS model has been mostly used to

examine the macroeconomic effects of equity (Asghar-

ian et al., 2013; Conrad & Loch, 2015; Opschoor et al.,

2014) and bond markets (Nieto et al., 2015). Some stud-

ies have also employed this methodology to examine the

volatility in commodity markets. Dönmez and Magrini

(2013) investigated possible drivers of long-term volatility

of agricultural commodities (wheat, corn, and soybean).

For oil prices, Yin and Zhou (2016) and Pan et al. (2017)

used GARCH-MIDAS with demand and supply shocks as

explanatory variables for the volatility.Conrad et al. (2014)

used macroeconomic variables to explain the dynamic cor-

relations of stock markets and oil prices. Regarding com-

modities, Wei et al. (2017) and Fang et al. (2018) showed

that economic policy uncertainty was positively associ-

ated with WTI spot returns and Gold futures variance and

improved forecasts. Moreover, Liu et al. (2018) used news

implied volatility indices to explain the long-term volatil-

ity of commodities. The authors presented evidence that

stock-market-related news affected energy and nonenergy

NGUYEN AND WALTHER 127