Uncovering predictability in the evolution of the WTI oil futures curve

Date01 January 2020
DOIhttp://doi.org/10.1111/eufm.12212
Published date01 January 2020
DOI: 10.1111/eufm.12212
ORIGINAL ARTICLE
Uncovering predictability in the evolution of the
WTI oil futures curve
Fearghal Kearney1Han Lin Shang2
1Queen’s Management School, Queen’s
University Belfast, Belfast BT9 5EE,
UK. Email: f.kearney@qub.ac.uk
2Research School of Finance, Actuarial
Studies and Statistics, Level 4, Building
26C, Australian National University,
Kingsley Street, Canberra ACT 2601,
Australia. Email:
hanlin.shang@anu.edu.au
Abstract
Accurately forecasting the price of oil, the world’s most
actively traded commodity, is of great importance to both
academics and practitioners. We contribute by proposing
a functional time series based method to model and fore-
cast oil futures. Our approach boasts a number of theoretical
and practical advantages, including effectively exploiting
underlying process dynamics missed by classical discrete
approaches. We evaluate the finite-sample performance a-
gainst established benchmarks using a model confidence set
test. A realistic out-of-sample exercise provides strong sup-
port for the adoption of our approach, which resides in the
superior set of models in all considered instances.
KEYWORDS
crude oil, forecasting, functional time series, futures contracts, futures
markets
JEL CLASSIFICATIONS
G10, G15, C53
1INTRODUCTION
Traditionally, statistical frameworks adopted in both the academic literature and practice address the
modeling of low-frequency regularly spaced data sets. However, the recent availability of higher-
resolution time series brings with it unique challenges and opportunities (Engle,2000). Functional
time series (FTS) is a convenient modeling and forecasting framework boasting the ability to tack-
le such problems. Tsay (2016) demonstrates that FTS analysis is widely applicable to many areas of
big data analytics and business statistics. We take up this mantle and propose a novel exponential
The authors wish to thank an anonymous referee and John A. Doukas (the Editor) for their suggestions which greatly improved
the quality of our paper. Furthermore, the authors are grateful for the insightful comments of the participants at the 1st Irish
Statistical Association Workshop on Frontiers in Functional Data Analysis in Dublin, and the 25th Forecasting Financial Markets
Conference in Oxford. This work was supported through a research grant from the College of Business and Economics at the
Australian National University.
2
s
m
o
c
k
w
T
i
t
t
f
p
p
P
C
r
p
o
a
i
B
t
w
l
p
i
l
m
a
s
s
o
e
a
D
t
s
1
p
e
i
H
K
238 © 2019 John Wiley & Sons, Ltd. wileyonlinelibrary.com/journal/eufm Eur Financ Manag. 2020;26:238–257.
e
r
-
l
k
-
f
l
d
h
s
e
2SHANG
smoothing state space FTS framework to address a critical problem in empirical finance, namely the
modeling and forecasting of crude oil futures prices.
Alquist, Kilian, and Vigfusson (2013) outline three critical areas for which reliable forecasts of
oil prices are essential. First, airlines, automobile firms, utilities and even homeowners make pricing,
consumption and investment decisions based on forecasts of oil prices. Second, oil price forecasts are a
key component of central bank macroeconomic projections and market practitioners’ risk assessments,
with more accurate forecasts potentially improving policy responses and risk management strategies.
Third, oil price forecasts inform energy usage and carbon emission projections with regulatory policy
implications such as fuel excise adjustments and climate change interventions. More specifically, in
this paper, we focus on oil futures prices. In general, commodity futures are far more actively traded
than their related spot contracts (see, for example, Ng & Pirrong, 1996). Outlining the importance of
futures markets, Laws and Thompson (2004) explicitly identify four functions they serve: efficient
price discovery, resource allocation, risk management, and financing. Furthermore, energy futures
price forecasts can accurately inform the pricing of related securities and projects (Kyriakou, Nomikos,
Papapostolou, & Pouliasis, 2016).
The use of such financial derivatives in Europe has been widely documented (see, for example,
Carroll, O’Brien, & Ryan, 2017). More specifically for oil derivatives, Brent crude was traditionally
regarded as the leading indicator of prices within Europe. However, as the United States recently re-
pealed its 40-year-old ban on exporting West Texas Intermediate (WTI) crude, forecasting WTI crude
oil futures has become a vital issue for European audiences. Since the repeal of this ban, there has been
a surge in WTI futures trading (Kemp,2018). This phenomenon has led to a trend of increased trad-
ing volumes for WTI oil futures coupled with reduced trading volumes for the traditionally European
Brent oil futures (Meyer,2017).
Furthermore, recent production of oil in the European North Sea has seen a structural decline at a
time when US WTI crude has observed record production levels. This has led to reports that the firm
which sets the price of Brent, S&P Global Platts, is now considering using WTI prices to calculate its
latest European index (Hurst, 2018). These factors have culminated in WTI arguably becoming the
price discovery leader in global crude oil markets and overtaking Brent as the leading oil benchmark
in Europe (Brusstar, 2018).
Two main classes of models are outlined in the commodity futures modeling literature: discrete
latent factor and fundamental macroeconomic-based models. Latent factor approaches to model com-
modity futures curves include those presented by Casassus and Collin-Dufresne (2005), Casassus, Liu,
and Tang (2013), Chantziara and Skiadopoulos (2008), and Trolle and Schwartz (2009), with a separate
strand of literature concluding that fundamentals are important in explaining energy prices (Andreas-
son, Bekiros, Nguyen, & Uddin, 2016; Basak & Pavlova, 2016; Kilian, 2009; Singleton, 2013) . Based
on this literature, we adopt the fundamental commodity futures forecasting framework of Andreasson
et al. (2016), as well as the principal component approach of Chantziara and Skiadopoulos (2008)
as two benchmarks for our proposed FTS model. This aligns with the approach taken by Cummins,
Dowling, and Kearney (2016) who adopt both frameworks and find no discernible difference between
the performance of the latent and fundamental models.
To the best of our knowledge, this paper constitutes the first application of FTS methods to de-
scribe, model, and forecast commodity futures curves.1In the FTS domain, the item of interest is
1A number of previous empirical studies have benefited from being cast in a functional data environment. Prominent exam-
ples include credit card transactions (Laukaitis, 2008), online auction price dynamics (Wang, Jank, & Shmueli, 2008) and
electricity price curves (Chen & Li, 2017). More specifically within financial markets, applications span equity index volatil-
ity (Müller, Sen, & Stadtmüller, 2011), stock returns (Horváth, Kokoszka, & Rice, 2014; Horváth & Rice, 2015; Kokosz-
HEATON 1151
239
KEARNEY AND SHANG
7.
EUROPEAN
|

To continue reading

Request your trial

VLEX uses login cookies to provide you with a better browsing experience. If you click on 'Accept' or continue browsing this site we consider that you accept our cookie policy. ACCEPT