What Drives Commodity Returns? Market, Sector or Idiosyncratic Factors?
| DOI | http://doi.org/10.1111/obes.12334 |
| Published date | 01 April 2020 |
| Date | 01 April 2020 |
| Author | Mark E. Wohar,Jun Ma,Andrew Vivian |
311
©2019 The Department of Economics, University of Oxford and JohnWiley & Sons Ltd.
OXFORD BULLETIN OF ECONOMICSAND STATISTICS, 82, 2 (2020) 0305–9049
doi: 10.1111/obes.12334
What Drives Commodity Returns? Market, Sector or
Idiosyncratic Factors?
Jun Ma,*Andrew Vivian† and Mark E. Wohar‡
*Department of Economics, College of Social Sciences and Humanities, Northeastern
University, Boston, MA 02115, USA (e-mail: ju.ma@northeastern.edu)
†School of Business and Economics, Loughborough University, Leicestershire, LE11 3TU,
UK (e-mail: a.j.vivian@lboro.ac.uk)
‡Department of Economics, University of Nebraska at Omaha, Omaha, NE 68182, USA
(e-mail: mwohar@unomaha.edu)
Abstract
This paper examines the relationship between 43 commodity returns using a dynamic
factor model with time varying stochastic volatility.The dynamic factor model decomposes
each commodity return into a common (market), sector-specific and commodity-specific
component. It enables the variance attributed to each component to be estimated at each
point in time. We find the return variation explained by the common factor has increased
substantially for the recent period and is statistically significant for the vast majority of
commodities since 2004 (at each point in time) This phenomenon is the strongest for non-
perishable products. We link the amount of variation explained by the common factor to
economic variables.
I. Introduction
Economic shocks have an impact that often go far beyond the immediate asset, product or
country from which they emanate. For example, the impact of economic shocks will be
greater if commodity markets are integrated and share a common component. The extent to
which commodity returns are globally determined has receivedrelatively little attention and
little rigorous analysis has been conducted. This topic is of great importance though, given
that strong links between markets could result in global shocks having a dramatic impact
upon commodities as witnessed during the recent financial crisis of 2008–2010. Market-
wide transmission of shocks also have policy implications for commodity producers and
extractors who could find that diversification benefits are not as large as anticipated.While
a large literature examines commodity price co-movements (building on Pindyck and
Rotemberg, 1990), a much smaller body of literature examines commodity returns and
whether these are correlated with each other.1
JEL Classification numbers: C12, C32, E20, G12, G15.
1There is a large literature examining international co-movements of commodity prices that builds on the seminal
paper of Pindyck and Rotemberg (1990). This area of inquiry tends to focus upon the economic determinants of
312 Bulletin
In stark contrast to the literature on commodity price co-movement, studies using com-
modity returns generally suggest that the market is highly segmented and any common
component is very small (see, e.g. Erb and Harvey, 2006). However, there are numerous
channels which could lead to common movementin commodity returns. For example, Hong
andYogo(2012) provide a model and evidence that futures open interest predicts commod-
ity returns and other asset returns. Further, since the turn of the millennium there have been
dramatic developments within commodity markets.In par ticular,there has been an influx of
speculative investorswho view commodities purely as financial assets; there has been rapid
growth in the money under management in commodity investment funds. Such investment
inflows can induce return co-movement across traded commodities (Basak and Pavlova,
2016). Hence individual commodity returns can be at least partly determined by agents
who respond to factors other than those facing the commodity producer or commodity
consumer (Gorton and Rouwenhorst, 2006; Tang and Xiong, 2012). There is also evidence
of commonality in liquidity across commodities which would induce return co-movements
via both the demand channel (co-ordinated buying and selling by institutional investors)
and the supply channel (liquidity withdrawal during large market declines) [see Marshall,
Nguyen and Visaltanachoti, 2013]. Hence, it is plausible that the extent of co-movement
has changed substantially recently and that it also exhibits variation over time.
The main objective of this paper is to model the dynamic behaviour of commodity
returns in the context of a sample of 43 commodities drawn from eight sectors over the
period January 1984–June 2017.2This enables us to examine the role of market-wide and
sectoral factors across a wide spectrum of different commodities. Specifically, our dynamic
factor model decomposes commodity returns into three components: a common (or market)
factor, a sectoral factors and a commodity-specific factor. Note that the common factor
and sectoral factor are orthogonal (in the population). Importantly, the common factor is
the common empirical factor across these 43 commodity returns; thus it is not based on
value-weighted commodity indices, such as S&P GSCI, and hence is not driven by the
largest component (WTI and Brent Crude) or segment (Energy);3this contrasts with much
literature where the commodity market factor is based on a value-weighted index (see
e.g. Tang and Xiong, 2012).
Our innovative development of a dynamic factor model, which extends those imple-
mented in other contexts (e.g. Stock and Watson (2002a, b); Neely and Rapach, 2011),
enables variance contributions of each component of the commodity return to be estimated
at each point in time and statistical significance to be assessed at each point in time. These
time varying factor contributions are estimated in the dynamic factor model by allowing
for time-varying stochastic volatility.Therefore, a key contribution of the paper is to esti-
co-movementand tries to resolve the puzzle that commodity price co-movement is much greater than can be explained
byeconomic fundamentals. Saadi (2011) provides a review of commodity price co-movementin international markets.
Despite a vast literature, the empirical evidencehas not reached a consensus on the impor tant variablesin determining
commodity co-movements. Not surprisingly, each candidate variable (such as aggregate demand, inflation, interest
rates and exchange rates) influences the movementof commodity prices during at least one period, some more in one
period than other periods.
2Please note this model requires the series to be stationary and hence we cannot apply this approach to commodity
prices.
3The S&P GSCI index is drivenby Energy sector commodities which typically comprise at least 60% of the overall
index with Crude Oil (WTI and Brent Crude) typically accounting for at least 40% of the overall index.
©2019 The Department of Economics, University of Oxford and JohnWiley & Sons Ltd
To continue reading
Request your trial