You are on page 1of 21

Applied Economics

ISSN: 0003-6846 (Print) 1466-4283 (Online) Journal homepage: http://www.tandfonline.com/loi/raec20

Relative scarcity and convenience yield: evidence


from non-ferrous metals

Akihiro Omura, Richard Chung, Neda Todorova & Bin Li

To cite this article: Akihiro Omura, Richard Chung, Neda Todorova & Bin Li (2016): Relative
scarcity and convenience yield: evidence from non-ferrous metals, Applied Economics, DOI:
10.1080/00036846.2016.1181832

To link to this article: http://dx.doi.org/10.1080/00036846.2016.1181832

Published online: 28 May 2016.

Submit your article to this journal

Article views: 16

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at


http://www.tandfonline.com/action/journalInformation?journalCode=raec20

Download by: [University of California, San Diego] Date: 27 June 2016, At: 11:03
APPLIED ECONOMICS, 2016
http://dx.doi.org/10.1080/00036846.2016.1181832

Relative scarcity and convenience yield: evidence from non-ferrous metals


Akihiro Omura, Richard Chung, Neda Todorova and Bin Li
Department of Accounting, Finance and Economics, Griffith Business School, Griffith University, Brisbane, Queensland, Australia

ABSTRACT KEYWORDS
We study the relationship between convenience yield and relative scarcity in the non-ferrous Commodity markets;
metal market for the period January 2000–March 2015. We identify various sets of economic options; economic linkages;
relationships for six major base metals, namely, aluminium, copper, lead, nickel, tin and zinc. Our convenience yield; the
bivariate and multivariate VARs and associated Granger-causality test results generally support theory of storage;
the existence of a positive relationship between convenience yields of base metals and our forecasting
relative scarcity measure. Furthermore, the time-varying characteristics observed in the results, JEL CLASSIFICATION
especially during contango and backwardation periods, provide useful information to market
Downloaded by [University of California, San Diego] at 11:03 27 June 2016

G15; Q31; D51; D81; E20


players in developing inventory strategies.

I. Introduction relative scarcity should play an important role in


considering the convenience yield of a commodity.
Non-ferrous industrial/base metals are vital materi-
For instance, thermal coal, natural gas and crude
als for our modern technologies. Understanding how
oil are some of the most commonly traded energy-
spot and future prices are related is crucial for mar-
related commodities and the consumption of each
ket participants including producers, stockists and
commodity can be replaced by the other two. In this
manufacturers to secure stable supply and produc-
sense, the consumption decision on thermal coal is
tion of goods. Furthermore, although the price levels
also determined by the market condition of the other
of major base metals have dropped from their peak
two commodities. Casassus, Liu, and Tang (2013)
in the last several years, drastic price hikes that we
show that the economic linkages contain valuable
have witnessed in the first decade of this century
information useful to understand how the conveni-
have urged investors such as traders to pay more
ence yield of energy commodity behaves.
attention to the non-ferrous metal market. For this
Although there are some studies examining the
reason, disentangling the dynamic relationship
relationship among the prices of industrial metals,
between the futures/forward prices of a shorter
there is no extensive study which expands the theory
maturity and a distant contract is also of great inter-
of storage by applying the notion of relative scarcity.
est to those market participants. This study tests
While Adhikari and Putnam (2014) incorporate a
whether information regarding the supply/demand
notion of economic linkages into their study of the
condition of economically linked commodities helps
theory of storage, they do not directly address the
explain how the two futures/forward contracts with
impact of ‘relative’ scarcity on the convenience yield
different maturity dates are related.
of major industrial metals. To obtain a comprehen-
The theory of storage traditionally aims to explain
sive understanding of how the convenience yield of
the benefits of possessing a physical asset or a short-
base metals is related to the economic linkages
term futures contract, proxied by a notion of con-
among other metals, we have to identify each eco-
venience yield, based on its own scarcity level.
nomic relationship that exists among metals. For
However, the consumption decision of a commodity
example, both copper and aluminium are common
is also affected by the market conditions of other
electrical conductors for cables, lead and zinc are
economically related commodities. In other words,
often mined from the same deposit, and alloys such

CONTACT Akihiro Omura akihiro.omura@griffithuni.edu.au Department of Accounting, Finance and Economics, Griffith Business School, Griffith
University, Brisbane, Queensland 4111, Australia
© 2016 Informa UK Limited, trading as Taylor & Francis Group
2 A. OMURA ET AL.

as brass are a composite of a number of different readily accessible (Troszkiewicz and Kolesnikova
industrial metals. We argue that these economic 2013). In some cases, it takes up to 2 years to with-
relationships should play an important role in deter- draw a metal from a warehouse (Erheriene 2015). If
mining the benefits of possessing a physical metal or the prices of industrial metals are also determined by
a shorter term futures contract. In other words, the the supply/demand environment of other metals, the
convenience yield of a metal is determined not only withdrawal constraint imposed on the aluminium or
by its own scarcity but also by the dynamic relation- copper market may also destruct other base metals
ship with other closely related metals. By extending from settling at their equilibrium prices. Hence, our
the basic idea of Casassus, Liu, and Tang (2013) to study can provide further justification for the LME
the industrial metals market, we examine whether to impose this new rule.
the relative scarcity of metals against other metals Third, our study contributes to further general-
helps understand the behaviour of their perceived izing the options-based approach framework in the
convenience yields. In particular, we conduct a study context of convenience yield approximation. We
on six major base metals, namely, aluminium, cop- employ the options-based approach proposed and
Downloaded by [University of California, San Diego] at 11:03 27 June 2016

per, lead, nickel, tin and zinc. developed by Heinkel, Howe, and Hughes (1990)
Our examination of empirical convenience yields and Milonas and Thomadakis (1997) to estimate
and relative scarcity is important and useful for three the perceived convenience yield of industrial metals.
main reasons. First, it provides valuable information The options approach has received attention in
to market participants such as metal producers and recent years, for example, Dockner, Eksi, and
commodity traders whether they should also con- Rammerstorfer (2015), Hochradl and
cern the relative scarcity of a metal, in addition to Rammerstorfer (2012), Lin and Duan (2007),
the absolute scarcity measured by its own scarcity Omura et al. (2015), Omura and West (2015) and
level, when they establish their inventory and invest- West (2012).
ment strategies associated with non-ferrous metals. Through rationalizing the economic relationships
Based on the report of Metal Bulletin (2013), it can among base metals, we identify a number of eco-
be estimated that major copper smelters receive less nomic linkages. Our analysis uncovers statistically
than 10% of the metal’s value from mining firms as a significant economic linkages in determining the
processing fee (based on the copper price ranged convenience yield of non-ferrous metals. The
between 5000 and 10,000 USD/MT (metric tonne)). observed time-varying characteristics of these rela-
Hence, in order to implement effective inventory tionships have important practical implications for
strategies, it is crucial for producers to understand stockists, producers and traders. In particular, some
how perceived benefits of holding physical metals, as of our relative scarcity variables are more significant
opposed to selling immediately, are determined. during backwardation periods. This is probably
Second, our article also provides valuable infor- because as the perceived benefits of immediate pos-
mation to policymakers. The London Metal session of a physical metal are higher during back-
Exchange (LME), the largest industrial metal wardation, market participants pay more attention
exchange in the world, is imposing a new warehouse to the supply/demand environment of economically
rule related to the withdrawal time in order to related commodities. Our analysis can help market
improve the accessibility of inventories. It is sug- participants to choose which economic linkages they
gested to ask warehouses to increase net load-out should concern, in addition to the absolute scarcity,
of metals if thresholds queue at a warehouse exceed when they attempt to quantify the benefits of obtain-
50 days (LME 2013). The existing rule is 100 days. ing a physical base metal.1
The proposed rule reflects the requests from the The remainder of the study is structured as fol-
physical industry especially for aluminium and cop- lows. Sections 2 provides a background of the study
per. In particular, it is criticized by industrial users and the hypothesis developed through reviewing
that the inventories in some warehouses are not past relevant literature. Section 3 presents the data

1
Our study also shows that economic linkages between two metals can be important in determining the convenience yield of one metal but not for the
other metal. For instance, the relationship between aluminium and copper is shown to affect the convenience yield of copper but not of aluminium.
APPLIED ECONOMICS 3

and method of the study. Sections 4 and 5 report our 2013), especially when appropriate carrying charges
empirical results and discuss the findings. Section 6 including storage costs are not factored in.
concludes. This issue is avoided by following Heinkel, Howe,
and Hughes (1990) and Milonas and Thomadakis
(1997) who consider convenience yield as a value
II. Background and research hypothesis of contingent claim for a payoff from a call option.
In particular, Milonas and Thomadakis (1997)
Convenience yield, the theory of storage and
employ a method of Fischer (1978) to estimate the
economic linkages
convenience yield of three agricultural commodities
Kaldor (1939) introduces the notion of convenience and copper by regarding convenience yields as call
yield to explain the spread between spot and futures options. Kocagil (2004), Lin and Duan (2007) and
prices. He claims that possessing physical assets West (2012) apply a similar approach which is a
allows the holders to make use of them whenever modification of the Black and Scholes (1973) options
they wish to. From this reason, a ‘benefit’ of conve- price model.
Downloaded by [University of California, San Diego] at 11:03 27 June 2016

nience is awarded to spot prices. Moreover, the The theory of storage introduced and developed
possession of the asset can be a benefit as it mitigates by Working (1948, 1949) and Brennan (1958) links
the risk arising from a possible stock-out. A conve- the behaviour of convenience yield to the scarcity
nience yield represents all these benefits enjoyed by level of the commodity. In the context of this theory,
owning the actual asset. convenience yield is a negative and decreasing func-
One important requirement in eliminating arbit- tion of its own inventory level. This is because the
rage opportunities is that the convenience yield inventory absorbs demand and supply shocks on
should be non-negative (Liu and Tang 2010; spot prices but a decrease in the inventory level
Stepanek, Walter, and Rathgeber 2013; Geman and amplifies the risk of stock-out and the volatility of
Smith 2013). A negative value implies that the future spot prices as inventory is bounded by zero
futures contract of a commodity is priced higher (Deaton and Laroque 1992; Gorton, Hayashi, and
than the relevant spot contract after considering a Rouwenhorst 2013).
necessary cost of storing it. Hence, market partici- The theory of storage seeks to explain the benefit
pants can obtain a riskless profit by buying an asset of possessing a physical asset using the supply/
in the spot market and immediately selling it in the demand environment of the same asset (Casassus,
futures market. In addition, Routledge, Seppi, and Liu, and Tang 2013). However, as information of an
Spatt (2000) demonstrate that since inventory is economically linked commodity helps explain the
bound by zero, benefits of possessing storable com- supply/demand of other commodities, such a rela-
modities, which is represented by convenience yield, tionship should also provide valuable information to
equate to a timing option whose value should be understand how convenience yield of the commod-
non-negative. This non-negative requirement also ity behaves. In particular, many commodities are
implies that convenience yields (net of storage produced/extracted/refined from other commodities
costs) cannot be less than the inverse value of the (e.g. gasoline is extracted from crude oil), can be
cost of storage (Omura and West 2015). replaced by other commodities (natural gas is a
Since convenience yields cannot be directly substitute of coal in the electricity making), have
observed, the cost-of-carry approach is often used the same end users (the largest user of palladium
to quantify the convenience yield. The method esti- and rubber is the automobile industry), or together
mates convenience yield by taking the difference get transformed to a final product (e.g. alloys are
between the spot (shorter maturity futures contract) made up of multiple metals).
and futures (distant futures contract) prices after While some researchers such as Cortazar, Milla,
adjusting for necessary carrying charges such as and Severino (2008) and Chng (2009, 2010) concern
interest, storage costs and insurance premia. these economic linkages and examine how they are
However, because of the model structure, the cost- related to commodity prices, we attribute the initial
of-carry model can obtain economically implausible development of applying the notion of the relative
negative values (Stepanek, Walter, and Rathgeber scarcity in the context of the theory of storage to
4 A. OMURA ET AL.

Casassus, Liu, and Tang (2013). In theory, two com- commodity increases/decreases as the production of
modities are in (1) a production relationship if one is the other changes. A positive demand shock on
produced by processing the other commodity, (2) a gasoline causes its price to inflate but the price of
substitution relationship if one can be replaced by diesel should be suppressed. This results in an
the other commodity, (3) a complementary relation- increasing price-level ratio (gasoline over diesel).
ship if an increase in the production/consumption of To satisfy the demand, suppliers increase the pro-
one also increases production/consumption of the duction of gasoline. However, this will cause the
other commodity. The relative scarcity level is mea- projected inventory level to return back to the equi-
sured by the spot price-ratio between the economic- librium level. As a result, the expected price of gaso-
ally linked two commodities. Following Casassus, line will fall below the current level and thereby the
Liu, and Tang (2013), the following paragraphs convenience yield will increase.
explain, using energy commodities as an example,
how those three possible relationships affect the
Non-ferrous metals and economic linkages
convenience yield of a commodity.
Downloaded by [University of California, San Diego] at 11:03 27 June 2016

Crude oil is an input of gasoline, and the two are Adhikari and Putnam (2014) conduct a study by
linked through the production relationship. A applying the theory of storage on some major base
demand shock on gasoline stimulates its price to metals (aluminium, lead, nickel, tin and zinc) and
rise if the inventory does not sufficiently absorb also incorporating the idea of economic linkages.
this shock. Such a shock also stimulates the demand However, they do not (1) specifically identify which
of crude oil (in order to extract gasoline); however, if metal is economically linked with other metals, (2)
inventory is plentiful, the price of crude oil would directly address the impact of relative scarcity, or (3)
not change. This means that the relative scarcity of take into account of storage costs in estimating the
gasoline against crude oil is high and causes an convenience yield. Instead, they examine how con-
increase in the current price-level ratio (spot price venience yield of a metal is related to its own and
of gasoline divided by crude oil). Nevertheless, this other metals’ inventory levels, and the open interest
shock will be eventually absorbed by an increase in obtained from the CRB. Adhikari and Putnam
the production of gasoline, causing a decrease in the (2014), therefore, do not address which metals are
expected price of gasoline and thereby, increase in its economically related, nor do they comprehensively
current convenience yield. In addition, if an increase capture the impact of ‘relative’ scarcity of one metal
in the price-level ratio is caused by an easing supply/ against others. Through economic rationalization
demand environment of crude oil, the production stated in the following paragraphs, we identify
margin of gasoline will increase as the producers’ major economic linkages for each metal against
margin in the short term is mainly determined by other metals.
the difference between the prices of two. When this Firstly, non-ferrous metals are produced by smelt-
happens, producers will attempt to take the advan- ing and refining ores or recycled metals (also known
tage by storing a fraction of new production of gaso- as secondary or scrap metal), and therefore there is a
line. As a result, this causes an increase in the production relationship among those raw materials
projected inventory level of gasoline and fall in its and the refined metals. While reliable price series of
expected price. Thus, its convenience yield increases. ores are rarely available, information regarding the
Similarly, the relative scarcity of coal against nat- price of secondary metals is often publically avail-
ural gas which is its substitute can cause an increase able. In relation to the relative scarcity of metals
in the convenience yield of coal. When the price- arising from the production relationship, Xiarchos
level ratio (coal over natural gas) increases due to a and Fletcher (2009) find evidence that the primary
positive demand shock on coal, there will be more markets of copper, lead and zinc are related to the
incentives to substitute natural gas for coal. This will secondary market in the U.S.A.
cause projected coal inventory to increase and its Secondly, a number of non-ferrous metals can be
expected price to fall. Lastly, for the complementary substituted by each other. For example, for corrosion
relationship, both gasoline and diesel are extracted resistance, stainless steel or aluminium can be replaced
from crude oil and the production of either with galvanized steel in some cases (Nippon Steel &
APPLIED ECONOMICS 5

Sumitomo Metal Corporation 2012). In addition, cop- a metal can substitute and at the same time comple-
per cables used in the electricity power network, auto- ment the production/consumption of another metal,
mobiles and electronic equipment can be substituted, so we combine the categories of substitution and
although not completely, with aluminium cables and complementary into one to avoid misspecification.
vice versa. Furthermore, electrodes of batteries are In the case of batteries, nickel hydroxide, lithium
commonly made of alloys, and there exists a substitu- manganate and lithium nickel oxide are some of
tion relationship among metals used in those alloys (in the popular positive-electrode materials for high-
some cases also a complementary relationship). With capacity batteries. Hence, there are both substitution
utilization of the advantage of material, lead acid bat- and complementary relationships existing between
teries are widely used for motor starter in the auto- nickel and lithium. The price-level ratio is used in
mobile industry. Lithium-ion and nickel-ion batteries our study as a proxy for relative scarcity of a metal
are also commonly used in many industries, especially against another metal, which is in line with Casassus,
for applications where high-energy density is required. Liu, and Tang (2013).
For example, lithium-ion batteries are used to power The theory of storage, traditionally, aims to
Downloaded by [University of California, San Diego] at 11:03 27 June 2016

electronics such as smartphones and electric vehicles explain the convenience yield of a commodity
(Apple 2015; Nissan 2013). Also, as a recent trend, by using its own supply/demand condition.
hybrid vehicles, where nickel-ion cells have been a However, there are economic linkages among
popular electricity power source, are also starting to commodities and those relationships should also
shift to higher capacity lithium-ion batteries (Toyota directly impact the benefits that can be enjoyed
2014; Ford 2012). by owning a physical asset. In the case of the
Thirdly, complementary relationships also exist non-ferrous metal market, it is evident that pro-
among non-ferrous metals (Adhikari and Putnam duction, substitution and complementary rela-
2014). This is because final products of metals tionships exist. It is paramount for producers,
often take the form of alloys. For example, brass is stockists, manufacturers and traders to under-
an alloy made mainly of copper and zinc, and solder stand whether those economic linkages among
is an alloy of lead and tin. Stainless steel is also non-ferrous metals play an important role in
defined as a steel which contains more than 10.5% determining the perceived convenience yields of
of chromium; however, nickel is often added to major base metals.
increase functionalities of the metal such as a higher Our review of past literature identifies the pre-
corrosion resistance. From this sense, the two metals sence of a research gap that needs to be filled, and
are complements of each other. Furthermore, it is our study thus poses a research question: ‘Do
common that multiple metals are extracted from the economic linkages among non-ferrous metals
same ore. In particular, zinc and lead are often affect the convenience yield of those metals?’ To
mined from the same deposit and hence at this answer this question, we propose the following
type of mine, production of the two metals is com- hypothesis.
plement of each other.
Based on the explanations above, we identify var- H1. Relative scarcity of a non-ferrous metal has a
ious sets of economic linkages for the six major base positive relationship with its perceived conve-
metals, as presented in Table 1. In a number of cases, nience yield.

Table 1. Economic linkages among base metals.


Types of the economic linkage
Production Substitution/Complementary
Aluminium Al scrap (raw material) Cu (cable), Ni (alloy), Zn (alloy)
Copper Cu scrap (raw material) Al (cable), Zn (alloy)
Lead Ni (battery), Lithium (battery), Zn (by-product), Sn (alloy)
Nickel Al (alloy), Pb (battery), Cr (alloy), Li (battery), Zn (alloy)
Tin Pb (alloy)
Zinc Al (alloy), Cu (alloy), Pb (by-product)
Note: Al is aluminium, Cu is copper, Cr is chromium, Pb is lead, Li is lithium, Ni is nickel, Sn is tin and Zn is zinc.
6 A. OMURA ET AL.

III. Data and methodology Also, following Mensi et al. (2013), we use the S&P
500 index values to represent the equity market.
Daily spot and futures prices of metals including the
Both are obtained from Bloomberg.
secondary metals, and inventory data are obtained from
In addition, we incorporate the impact of hed-
DataStream and Bloomberg for the period January
ging, wherever data are available, on the convenience
2000–March 2015 (lithium price data are available
yield. Hedging pressure in futures markets is likely
from September 2008). Similarly to Stepanek, Walter,
to affect the price of commodity (Hirshleifer 1990;
and Rathgeber (2013), the warehouse rental prices
Roon, Nijman, and Veld 2000), thus we include such
obtained from the LME are used as proxies for the
a factor in this study. In many cases, hedging pres-
storage cost of each non-ferrous metal employed in
sure is measured by (net commercial open-interest
the analysis. We use 3-month U.S. Treasury bills to
positions)/(total number of hedge positions) and the
represent near-term interest rates. This is consistent
relevant up-to-date historical data for copper is pub-
with the past studies, for example, Stepanek, Walter,
lically available at the COMEX. By assuming that
and Rathgeber (2013) and Liu and Tang (2010).
hedgers in the different geographical locations have
Also, we employ three control variables in the
Downloaded by [University of California, San Diego] at 11:03 27 June 2016

a similar hedging need, we employ COMEX open-


study to isolate overall effects of commodity and
interest positions to measure the hedging pressure.
equity markets on the convenience yield, as a num-
We convert weekly open-interest positions into
ber of studies including Mensi et al. (2013) indicate
monthly averages to synchronize with other monthly
the presence of statistically significant relationships
average variables in the study.
among individual commodity markets, and overall
Table 2 presents descriptive statistics of monthly
commodity and equity markets. In particular, we use
changes in the data (except for the hedging pres-
the Thomson Reuters/Jefferies CRB index as a proxy
sure). It shows that the convenience yields and
for the overall commodity market development as
inventories, in general, are more volatile than the
this index comprises 19 major commodities includ-
spot prices. Also, the equity and the commodity
ing crude oil, gas, metals and live stocks and is one
market indices tend to be less volatile.
of the most traded commodity indexes in the world.

Table 2. Descriptive statistics.


Mean Median SD Kurtosis Skewness Minimum Maximum
CY Al 0.15 −0.25 21.86 2.76 0.13 −56.69 62.97
CY Cu 0.93 0.75 20.59 3.28 0.31 −48.97 67.12
CY Pb 0.93 1.76 16.87 2.63 −0.06 −42.98 39.84
CY Ni 0.32 −2.02 21.74 4.78 0.83 −46.96 94.11
CY Sn 0.33 −2.57 20.19 3.88 0.83 −45.37 71.47
CY Zn 0.29 −0.95 17.75 4.73 0.41 −64.39 62.93
Invent. Al 0.90 0.05 6.24 4.97 0.60 −13.63 23.76
Invent. Cu −0.48 −1.37 13.99 5.16 0.60 −39.20 64.79
Invent. Pb 0.12 −0.42 12.19 4.69 0.21 −36.27 47.35
Invent. Ni 1.24 2.32 16.26 6.99 −0.29 −76.20 60.04
Invent. Sn 0.02 0.11 13.74 5.68 −0.08 −55.05 52.45
Invent. Zn 0.35 0.54 7.77 5.18 0.10 −29.46 28.17
Spot Al 0.03 0.17 5.14 4.68 −0.58 −21.07 14.16
Spot Cr 0.58 0.07 7.51 64.60 −5.62 −76.98 21.87
Spot Cu 0.64 1.10 6.95 8.35 −0.84 −35.42 22.87
Spot Pb 0.73 1.05 7.84 5.06 −0.71 −29.24 23.18
Spot Li −0.21 0.00 3.31 11.40 −0.91 −16.01 12.52
Spot Ni 0.28 −0.15 9.02 4.13 −0.35 −38.13 24.06
Spot Sn 0.60 0.10 6.59 3.76 −0.21 −23.84 15.64
Spot Zn 0.30 0.45 6.71 5.01 −0.40 −29.18 23.37
Scrap Al 0.16 0.34 4.31 11.61 −1.59 −26.39 12.44
Scrap Cu 0.62 0.61 6.03 8.29 −1.06 −31.25 18.91
CRB 0.33 0.35 2.70 8.11 −1.15 −13.70 6.43
U.S. Equity 0.21 1.00 3.97 9.22 −1.60 −22.81 11.35
Hedge Cu −2.09 −1.65 19.19 2.45 −0.26 −53.45 36.18
Note: The table represents the descriptive statistics of the data used in the study. The numerical values are monthly changes in those variables (except
hedging pressure) and are in the percentage form. CY is the convenience yield, Invent. is the inventory at the LME, Spot indicates for the spot price, and
Scrap indicates for the spot scrap price of the relevant metal. Al is aluminium, Cu is copper, Pb is lead, Li is lithium, Cr is chromium, Ni is nickel, Sn is tin, Zn
is zinc, CRB is the CRB index and U.S. Equity is the S&P 500 index, Hedge Cu is the hedging pressure of copper futures calculated as net commercial
position/(commercial long position + commercial short position) at the COMEX. The sample period is January 2000–March 2015 (lithium price data starts
from September 2008).
APPLIED ECONOMICS 7

 
In theory, convenience yields should be bound by SðtÞ Max FðT Þ  1; 0 (4)
zero. However, the conventional cost-of-carry
approach used by researchers such as Fama and where FðT Þ ¼ SðtÞ =Fðt;T Þ . Using Ito’s lemma allows us
French (1988) is limited by the output of negative to establish following solution
convenience yields, though the model is elegant and CYi;t;T ¼ Sði;tÞ N ðd1 Þ  Fðt;T Þ N ðd2 Þ (5)
simple. The model aims to estimate convenience
yield by taking the difference between prices of two where
forward contracts with different maturities after  
ln FðT Þ þ σ 2c τ=2
adjusting for necessary costs associated with carrying d1 ¼ pffiffiffi
σc τ
the asset. It is evident from a recent study by
 
Stepanek, Walter, and Rathgeber (2013), that even pffiffiffi ln FðT Þ  σ 2c τ=2
after accounting for storage costs provided by the d2 ¼ d1  σ c τ ¼ pffiffiffi
σc τ
LME and the time value of money, negative conve-
and
nience yields can be still obtained when the cost-of-
Downloaded by [University of California, San Diego] at 11:03 27 June 2016

carry approach is used. σ 2c ¼ σ 2S ðtÞ þ 2σ S ðtÞ σ Fðt;T Þ ρSðtÞFðt;T Þ þ σ 2Fðt;TÞ (6)
As mentioned above, applying the options-based
approach of Heinkel, Howe, and Hughes (1990) and σ c is the volatility of the convenience yield, σ S ðtÞ and
Milonas and Thomadakis (1997) can avoid negative σ Fðt;TÞ are the volatilities of the spot and futures
convenience yield estimates. This study follows an prices, respectively, ρSðtÞFðt;T Þ is the correlation coef-
approach similar to Milonas and Thomadakis ficient between spot and futures contracts, and τ is
(1997), Kocagil (2004), Lin and Duan (2007), the time period until the maturity of the forward
Omura et al. (2015), Omura and West (2015) and contract. We assume that the spot price is the near-
West (2012), which is an extension of the Black– est available tenor and the forward price is repre-
Sholes option pricing model, to estimate conveni- sented by the 3-month futures contract price. The
ence yields. In particular, we express the conveni- volatility figures applied in the convenience yield
ence yield observed at time t as estimation are computed by combining the expo-
  nentially weighted moving average (EWMA) for
CYt;T ¼ Max SðtÞ  Fðt;TÞ ; 0 (1)
the Parkinson (1980) volatility estimator. Let σ 2tþ1jt
where CYt;T is the convenience yield from time t to be a volatility estimated by EWMA, λ be an expo-
T. SðtÞ ¼ SðtÞ þ Wðt;TÞ with SðtÞ as the spot price at nential weight coefficient,
time t, Wðt;T Þ the storage cost from time t to T and X
1
Fðt;TÞ is the price observed at time t of a futures σ 2tþ1jt ¼ ð1  λÞ λi Rti 0<λ<1 (7)
i¼0
contract which matures at time T. In this study, we
employ the three month futures contracts traded on where the subscript t þ 1jt illustrating a forecast at
the LME. This study assumes that both the spot time t + 1 is based on the available information up to
price and the forward price are stochastic and follow and including time t, and
standard diffusion processes (i.e. a geometric 2
Brownian motion), ðxHL Þ PHti
Rti ¼ ti t ¼ ln
and xHL (8)
4  ln2 PLti
dSðtÞ ¼ μS SðtÞ dt þ σ S SðtÞ dzS (2)
where PHt is the intraday high price, and PLt is the
intraday low price. We use λ ¼ 0:94 as commonly
dFðt;T Þ ¼ μF Fðt;TÞ dt þ σ F Fðt;TÞ dzF (3)
done in the literature and the validity of its empirical
where the subscripts S and F represent the spot and rule is justified by J.P. Morgan and Reuters (1996).
the futures tenor for each contract type, respectively. For the purpose of conducting regression analysis,
Furthermore, this study makes an assumption that we first estimate daily convenience yields and then
the diffusion terms dzS and dzF are uncorrelated, average them to obtain monthly values.
and that zS and zF follow a Wiener process. The We use well-established methods: bivariate and
associated boundary condition is also defined as multivariate vector autoregressive (VAR) models
8 A. OMURA ET AL.

and the associated Granger-causality tests to exam- position + commercial short position) of the copper
ine the relationship between convenience yield of a futures at the COMEX. As the levelled hedging
metal and the relative scarcity measure. A similar pressure dataset is found to be stationary, we use
method is used by Hameed (1997) and Rossi (2012) the levelled data for this variable. The number of lags
in their asset pricing-related studies. The general in the VAR is first selected based on the Akaike
form of the VAR system in this study is information criterion (AIC), Bayesian information
0 1 0 1 0 1 10 1 criterion and Hannan and Quinn information criter-
Δy1;t c1 β1;1 β11;h Δy1;t1 ion. However, as in a number of cases serial correla-
B : C B : C B :: C
B C B C B : : CB : C tion is observed in the residual series, we add extra
B C¼@ AþB B CBB
C
C
@ : A C
: A@ : A lags until this issue is solved.
: @ :
::
Δyh;t ch β1h;1 β1h;h Δyh;t1
0 k 10 1 IV. Results
β1;1 βk1;h Δy1;tk
B :: C
B : : CB : C The Augmented Dickey–Fuller test (Dickey and
þ ::: þ B CBB
C
B : C@ : C
Downloaded by [University of California, San Diego] at 11:03 27 June 2016

@ : A A Fuller 1979) is first applied to examine the stationar-


:: k ity of the considered variables (the results are avail-
βh;1 βh;h
k Δyh;tk
0 1 able upon request). The test statistics are estimated
ε1;t based on the modified AIC and all the variables
B : C
þB @
C
A
become stationary when converted into monthly
: change figures, apart from the hedging pressure
εh;t where the levelled data are found to be stationary.
(9) The remainder of this section reports our results of
the VAR and associated Granger-causality tests. We
where Δyi;t is the monthly log difference in the vari-
provide a comprehensive interpretation and discus-
able h (unless noted otherwise) at time t and εh;t is
sion of our results in the later discussion section.
an error term. We use monthly change figures to
overcome unit root issues inherent in the data
(except for the hedging pressure term as it is found Bivariate VAR results
to be stationary at the levelled form). In particular,
First, we conduct bivariate (convenience yield of a
for bivariate VAR estimation, the convenience yield
 metal and the relevant relative scarcity or inventory
variable for metal i is computed as ln CYi;t;T = variable) VAR tests. The results are reported in
CYi;t1;T1 , and the relative scarcity variable for Table 3 (we report the results for models which use
metal i in relation to metal j is computed as
  the convenience yield variable as a dependent vari-
ln Si;t =Sj;t ln½Si;t1 =Sj;t1 where i Þ j and St able. Detailed results are available upon request).
denotes the spot price at time t. We interchangeably The results indicate that many of the inventory vari-
describe this variable as the price-ratio in this study. ables have one or more significant negative lagged
Furthermore, for multivariate VAR model estima- coefficient/s at the 10% significance level. This is
tion, we incorporate four other control variables. In consistent with the conventional notion of the the-
particular, we employ (1) the inventory level variable ory of storage.
for metal i to control for own scarcity levels, com- It is found that fourteen relative scarcity variables
 
puted as ln Ii;t =Ii;t1 where I is the inventory prox- (price-ratios) have a significant positive lagged coef-
ied by the LME warehouse inventory, (2) the ficient/s at the 10% significance level. Similar results
Thomson Reuters/Jefferies CRB index variable to are obtained from the bivariate Granger-causality
partial out the effect of overall commodity market test displayed in Table 4. To be more specific, eleven
conditions, which is computed as ln½CRBt =CRBt1 , relative scarcity variables appear to Granger-cause
(3) the S&P 500 index to isolate the impact of the the convenience yield of the relevant metal at the
equity market condition, which is computed as 10% significance level. When it comes to the actual
ln½S Pt =S Pt1  and (4) the hedging pressure mea- impact of change in the level of relative scarcity on
sured by net commercial position/(commercial long the convenience yield, for those models with the
APPLIED ECONOMICS 9

Table 3. Bivariate VAR (results for the models which have the convenience yields as dependent variables).
Explanatory variable
Constant Lag 1 Lag 2 Lag 3 Lag 4 Lag 5
Coef. (p-value) Coef. (p-value) Coef. (p-value) Coef. (p-value) Coef. (p-value) Coef. (p-value)
(a)
CY Aluminium versus
Inventory 0.008 (0.61) −0.751* (0.06) −0.481 (0.35) 0.603 (0.13)
Relat. Cu 0.003 (0.85) −0.239 (0.50) 0.373 (0.30)
Relat. Ni 0.003 (0.83) 0.441*** (0.01) 0.110 (0.62)
Relat. Zn 0.001 (0.95) −0.329 (0.32) −0.079 (0.81)
Relat. Scrap 0.002 (0.88) 0.517 (0.30) −0.235 (0.64)
CY Copper versus
Inventory 0.012 (0.43) −0.169 (0.20) 0.096 (0.53) 0.048 (0.71)
Relat. Al 0.007 (0.65) 0.833** (0.01) 0.154 (0.64)
Relat. Zn 0.010 (0.52) 0.185 (0.54)
Relat. Scrap 0.012 (0.44) 0.774 (0.10) 1.103** (0.03) 0.715 (0.18) 0.502 (0.33) −0.086 (0.86)
Downloaded by [University of California, San Diego] at 11:03 27 June 2016

CY Lead versus
Inventory 0.012 (0.33) −0.200 (0.11) −0.196 (0.15) 0.310** (0.01)
Relat. Ni 0.009 (0.49) 0.258* (0.08)
Relat. Li −0.015 (0.37) −0.046 (0.84) 0.479** (0.03)
Relat. Zn 0.008 (0.51) 0.386* (0.07)
Relat. Sn 0.010 (0.41) 0.351* (0.06) −0.105 (0.57)
CY Nickel versus
Inventory 0.001 (0.86) −0.256** (0.04) 0.114 (0.39) −0.301** (0.03) 0.129 (0.33) 0.175 (0.14)
Relat. Al −0.002 (0.90) 0.565** (0.02) −0.016 (0.95)
Relat. Pb 0.004 (0.80) 0.565*** (0.00)
Relat. Li −0.013 (0.47) 0.680*** (0.01) −0.025 (0.93) −0.204 (0.43) 0.180 (0.38)
Relat. Cr 0.002 (0.91) 0.185 (0.24)
Relat. Zn 0.002 (0.92) 0.433* (0.07)
(b)
CY Tin versus
Inventory 0.003 (0.84) −0.234** (0.04)
Relat. Pb 0.003 (0.82) 0.448** (0.03)
CY Zinc versus
Inventory 0.006 (0.67) −0.364* (0.07) −0.086 (0.67)
Relat. Al 0.002 (0.88) 0.659** (0.01) −0.079 (0.77)
Relat. Cu 0.005 (0.73) 0.169 (0.52)
Relat. Pb 0.005 (0.71) 0.221 (0.29)
Relat. Ni 0.002 (0.87) 0.403** (0.03) 0.082 (0.68) −0.135 (0.46)
Note: The table reports the results of the bivariate VAR
! !
Δy1;t c1 β11;1 β11;2 Δy1;t1 β11;1 β11;2 Δy1;tk ε1;t
¼ þ þ ::: þ þ
Δy2;t ch β12;1 β12;2 Δy2;t1 β12;1 β12;2 Δy2;tk ε2;t

where Δy1 is the monthly change in the convenience yield of a metal in each model and Δy2 is the monthly change in the other variable in the model. The
table reports the results for the models having Δy1 as an explained variable and does not report the coefficients of autoregressive variables. Inventory is
the monthly change in the inventory figures from the LME for the explained variable. Relat. Al, Relat. Cr, Relat. Cu, Relat. Pb, Relat. Ni, Relat. Sn, Relat. Zn and
Relat. Scrap in rows below the subheading of CY . . . versus. indicate which metal is used to create a relative scarcity variable and each row reports the
result of the bivariate VAR having such relative scarcity variable. For example, Aluminium under the heading of CY Zinc versus. represents a relative scarcity
of zinc to aluminium and it is computed as the spot price of zinc divided by the spot price of aluminium. All variables are in the natural log monthly
change form. The standard errors in the results are adjusted for the small sample-size and degree-of-freedom in estimating the error variance–covariance
matrix. ***, **, * denote the significance level at 1%, 5% and 10%, respectively.

significant results, a 1 percentage point increases in Todorova (2013) in selecting the GFC period (July
the lagged price-ratio results in 0.3 to 1.1 percentage 2008 to December 2009). The results are shown in
points increase in the convenience yield. Table 4. Interestingly, we find that fewer relevant scar-
Since our sample incorporates the recent global city variables Granger-cause the relevant convenience
financial crisis (GFC) period, we also conduct subsam- yield variable, at the 10% significance level, after July
ple Granger-causality tests. We follow Souček and 2008. In contrast, we find a slightly larger number of
10 A. OMURA ET AL.

Table 4. Bivariate Granger-causality test (p-values).


Granger-causality test results
January 2000–June 2008 July 2008–March 2015 January 2010–March 2015 Full sample period
CY Aluminium versus
Relat. Copper 0.74 0.64 0.18 0.55
Relat. Nickel 0.07* 0.00*** 0.11 0.07*
Relat. Zinc 0.33 0.27 0.01*** 0.53
Relat. Scrap 0.70 0.77 0.68 0.49
CY Copper versus
Relat. Aluminium 0.00*** 0.31 0.48 0.01**
Relat. Zinc 0.75 0.39 0.16 0.54
Relat. Scrap 0.10* 0.16 0.10 0.25
CY Lead versus
Relat. Nickel 0.05** 0.22 0.13 0.08*
Relat. Lithium ‒ 0.08* 0.45 0.08*
Relat. Zinc 0.00*** 0.20 0.14 0.07*
Downloaded by [University of California, San Diego] at 11:03 27 June 2016

Relat. Tin 0.03** 0.15 0.42 0.17


CY Nickel versus
Relat. Aluminium 0.01*** 0.14 0.06* 0.06*
Relat. Lead 0.00*** 0.05* 0.29 0.00***
Relat. Lithium ‒ 0.08* 0.03** 0.08*
Relat. Chromium 0.01*** 0.02** 0.03** 0.24
Relatt Zinc 0.11 0.11 0.05** 0.07*
CY Tin versus
Relat. Lead 0.05** 0.37 0.26 0.03**
CY Zinc versus
Relat. Aluminium 0.01*** 0.87 0.01** 0.05**
Relat. Copper 0.17 0.60 0.42 0.52
Relat. Lead 0.40 0.92 0.13 0.28
Relat. Nickel 0.08* 0.37 0.08* 0.10
Note: The table reports the results of the Granger-causality test results based on the bivariate VAR
! !
Δy1;t c1 β11;1 β11;2 Δy1;t1 β11;1 β11;2 Δy1;tk ε
¼ þ þ ::: þ þ 1;t
Δy2;t ch β12;1 β12;2 Δy2;t1 β12;1 β12;2 Δy2;tk ε2;t

where Δy1 is the monthly change in the convenience yield of a metal in each model and Δy2 is the monthly change in the other variable in the model. The
table reports the test results which use Δy1 as an explained variable and the relative scarcity variables as an explanatory variable. For instance, Relat.
Copper under the subheading of CY Aluminium versus represents relative scarcity of aluminium to copper (computed as the spot price of aluminium
divided by the spot price of copper) and the numerical values in the subsequent columns indicate the p-values of the Granger-causality tests which use
the convenience yield of aluminium as the explained variable and its relative scarcity level to copper as the explanatory variable in the VAR model. CY is
the convenience yield of an indicated metal in the table. All the variables used in each model are in the natural log monthly change form. The standard
errors in the VAR results are adjusted for the small sample-size and degree-of-freedom in estimating the error variance–covariance matrix. The second row
in the table indicates the estimation period. ***, **, * denote the significance level at 1%, 5% and 10%, respectively.

significant causal relationships, compared with the and backwardation (determined by the interest-
full-sample period when the tests are conducted on adjusted spread between futures and spot contracts)
the period before the GFC. In addition, we observe a and reconduct the Granger-causality tests based on the
slightly fewer number of significant results (seven), for bivariate VAR. When the market is backwarded, the
the period after January 2010. The results may be relationship between the convenience yield of a metal
directly affected by the fact that limited number of and other economically linked metals may draw more
observations is available for the period after the GFC. attention of the market compared with the period of
contango. This is because, during backwardation per-
iods, perceived benefits of immediate possession of a
Effect of contango and backwardation physical metal are considered to be higher. The results
We also divide our sample based on the relative price are presented in Table 5. We display the test results
level of spot contract against futures contract. In parti- showing whether relative scarcity variables Granger-
cular, we separate the data into periods of contango cause convenience yield of a focused metal or not.
APPLIED ECONOMICS 11

Table 5. Bivariate Granger-causality test on contango and back- during contango periods. A similar trend is also
wardation periods (p-values). observed when contango and backwardation periods
Contango Backwardation are judged based on the simple spread without consid-
CY Aluminium versus ering the time value of money. In addition, while no
Relat. Copper 0.26 0.97
economically significant Granger-causality test results
Relat. Nickel 0.10* 0.36
Relat. Zinc 0.78 0.16 are obtained in the previous bivariate VAR tests for
Relat. Scrap 0.88 0.15 relative scarcity of zinc against either of copper or lead,
CY Copper versus when the test is conducted for backwardation periods,
Relat. Aluminium 0.16 0.05* both price-ratios are found to have a significant impact
Relat. Zinc 0.06* 0.65
on the convenience yield of zinc. This is probably
Relat. Scrap 0.55 0.19
because the coefficients of those variables are negative
CY Lead versus
Relat. Nickel 0.45 0.16 during contango periods, offsetting the significant posi-
Relat. Lithium 0.11 ‒ tive impacts during backwardation times.
Relat. Zinc 0.40 0.04**
Downloaded by [University of California, San Diego] at 11:03 27 June 2016

Relat. Tin 0.35 0.02**


CY Nickel versus Multivariate VAR results
Relat. Aluminium 0.26 0.01***
Relat. Lead 0.18 0.00*** Next, we further check the robustness of our result by
Relat. Lithium 0.09* ‒ adding control variables into the bivariate VAR mod-
Relat. Chromium 0.06* 0.01***
Relat. Zinc 0.24 0.03**
els. The control variables that we select in this study
CY Tin versus
are (1) the own inventory level to take into account of
Relat. Lead 0.20 0.17 the effect of absolute scarcity level, (2) the S&P 500
CY Zinc versus index to control for the influence of the equity mar-
Relat. Aluminium 0.77 0.03** ket, (3) the CRB index to control the effect of the
Relat. Copper 0.30 0.07* overall commodity market behaviour and (4) the
Relat. Lead 0.10* 0.04**
Relat. Nickel 0.26 0.4)
hedging pressure whenever relevant data are available.
Note: The table reports the results of the Granger-causality tests based on The multivariate VAR results (Table 6) show that
the bivariate VAR
!
there are thirteen relative scarcity variables which
Δy1;t c1 β11;1 β11;2 Δy1;t1 have statistically significant lagged coefficient esti-
¼ þ þ :::
Δy2;t ch β12;1 β12;2 Δy2;t1
! mates at the 10% level. However, of those, three
β11;1 β11;2 Δy1;tk ε1;t models (aluminium-to-zinc, lead-to-zinc, nickel-to-
þ þ
β12;1 β12;2 Δy2;tk ε2;t
lithium) contain a negative lagged coefficient esti-
where Δy1 is the monthly change in the convenience yield of a metal in each
model and Δy2 is the monthly change in the other variable in the model. The mate. Similarly to the previous bivariate tests, we
table reports the test results which use Δy1 as an explained variable and the conduct the Granger-causality tests based on the
relative scarcity variables as an explanatory variable. For instance, Relat.
Copper under the subheading of CY Aluminium versus represents relative results obtained in the multivariate VAR. Table 7
scarcity of aluminium to copper (computed as the spot price of aluminium displays the results for the tests with a null-hypothesis
divided by the spot price of copper) and the numerical values in the
subsequent columns indicate the p-values of the Granger-causality tests ‘relative scarcity variable does not Granger-cause con-
which use the convenience yield of aluminium as the explained variable venience yield of a focused metal’. We find ten sig-
and its relative scarcity level to copper as the explanatory variable in the VAR
model. CY is the convenience yield of an indicated metal in the table. All the nificant causal relationships at the 10% level.2 We will
variables used in each model are in the natural log monthly change form.
Data are divided into contango and backwardation periods, determined by
further discuss our results in the following section.
interest-adjusted futures price–spot price at time t. The standard errors in the
results are adjusted for the small sample-size and degree-of-freedom in
estimating the error variance–covariance matrix. ***, **, * denote the sig-
nificance level at 1%, 5% and 10%, respectively. V. Discussion
Our results broadly support the presence of relative
There is a total of ten causal relationships, significant scarcity that affects the perceived convenience yield
at the 10% level, found during backwardation, but only of major base metals. The results are summarized in
half of the number of significant results are obtained Table 8. Nevertheless, no strong pattern among the
2
We also test whether the significant relationships can be observed in recent years by controlling for various factors; however, the model does not converge.
This is possibly because the sample sizes are not large enough.
12 A. OMURA ET AL.

Table 6. Multivariate VAR tests with one relative scarcity variable in each model.
(a)
CY Al Coef. (p-val.) CY Al Coef. (p-val.) CY Al Coef. (p-val.)
Con. 0.013 (0.44) Con. 0.014 (0.40) Con. 0.009 (0.56)
Own Var. Own Var. Own Var.
Lag1 −0.185** (0.03) Lag1 −0.161* (0.05) Lag1 −0.180** (0.01)
Lag2 −0.161* (0.06) Lag2 −0.154* (0.08) Lag2 ‒ ‒
Lag3 −0.009 (0.91) Lag3 −0.017 (0.85) Lag3 ‒ ‒
Lag4 ‒ ‒ Lag4 −0.053 (0.53) Lag4 ‒ ‒
Invent. Invent. Invent.
Lag1 −0.819* (0.05) Lag1 −0.867** (0.05) Lag1 −0.880** (0.01)
Lag2 −0.631 (0.24) Lag2 −0.754 (0.16) Lag2 ‒ ‒
Lag3 0.690 (0.11) Lag3 1.022* (0.06) Lag3 ‒ ‒
Lag4 ‒ ‒ Lag4 −0.418 (0.34) Lag4 ‒ ‒
Relat. Cu Relat. Zn Relat. Ni
Lag1 −0.274 (0.48) Lag1 −0.614* (0.09) Lag1 0.380* (0.08)
Lag2 0.477 (0.25) Lag2 0.066 (0.86) Lag2 ‒ ‒
Downloaded by [University of California, San Diego] at 11:03 27 June 2016

Lag3 0.086 (0.82) Lag3 −0.274 (0.45) Lag3 ‒ ‒


Lag4 ‒ ‒ Lag4 0.283 (0.40) Lag4 ‒ ‒
CRB CRB CRB
Lag1 0.392 (0.63) Lag1 0.178 (0.82) Lag1 0.461 (0.47)
Lag2 −0.116 (0.90) Lag2 −0.146 (0.87) Lag2 ‒ ‒
Lag3 −0.218 (0.78) Lag3 0.375 (0.67) Lag3 ‒ ‒
Lag4 ‒ ‒ Lag4 −0.157 (0.84) Lag4 ‒ ‒
U.S. U.S. U.S.
Lag1 −0.762* (0.10) Lag1 −0.641 (0.18) Lag1 −0.676 (0.13)
Lag2 −0.187 (0.69) Lag2 −0.256 (0.59) Lag2 ‒ ‒
Lag3 0.161 (0.72) Lag3 0.173 (0.72) Lag3 ‒ ‒
Lag4 ‒ ‒ Lag4 −0.791* (0.09) Lag4 ‒ ‒
CY Al Coef. (p-val.) CY Pb Coef. (p-val.) CY Pb Coef. (p-val.)
Con. 0.012 (0.44) Con. 0.007 (0.58) Con. −0.024 (0.19)
Own Var. Own Var. Own Var.
Lag1 −0.234*** (0.01) Lag1 −0.150* (0.07) Lag1 −0.237* (0.05)
Lag2 −0.166* (0.06) Lag2 ‒ ‒ Lag2 −0.225* (0.06)
Invent. Invent. Invent.
Lag1 −0.680* (0.09) Lag1 −0.192* (0.08) Lag1 −0.037 (0.89)
Lag2 −0.251 (0.55) Lag2 ‒ ‒ Lag2 −0.449 (0.87)
Relat. Scrap Relat. Ni Relat. Li
Lag1 0.533 (0.30) Lag1 0.260* (0.08) Lag1 −0.225 (0.46)
Lag2 −0.158 (0.75) Lag2 ‒ ‒ Lag2 0.136 (0.64)
CRB CRB CRB
Lag1 0.339 (0.66) Lag1 0.376 (0.47) Lag1 −0.078 (0.93)
Lag2 −0.158 (0.83) Lag2 ‒ ‒ Lag2 0.664 (0.46)
U.S. U.S. U.S.
Lag1 −0.747* (0.10) Lag1 0.475 (0.17) Lag1 0.843 (0.16)
Lag2 −0.264 (0.56) Lag2 ‒ ‒ Lag2 0.451 (0.40)
(b)
CY Pb Coef. (p-val.) CY Pb Coef. (p-val.) CY Ni Coef. (p-val.)
Con. 0.008 (0.51) Con. 0.007 (0.55) Con. 0.003 (0.88)
Own Var. Own Var. Own Var.
Lag1 −0.235*** (0.01) Lag1 −0.231*** (0.01) Lag1 −0.144* (0.09)
Lag2 −0.167** (0.04) Lag2 −0.199** (0.02) Lag2 ‒ ‒
Invent. Invent. Invent.
Lag1 −0.343*** (0.01) Lag1 −0.343*** (0.01) Lag1 −0.173 (0.10)
Lag2 −0.072 (0.54) Lag2 −0.050 (0.66) Lag2 ‒ ‒

(Continued )
APPLIED ECONOMICS 13

Table6. (Continued).
Relat. Zn Relat. Sn Relat. Al
Lag1 0.453** (0.04) Lag1 0.318* (0.08) Lag1 0.583** (0.02)
Lag2 −0.477** (0.03) Lag2 −0.155 (0.39) Lag2 ‒ ‒
CRB CRB CRB
Lag1 −0.704 (0.22) Lag1 −0.539 (0.35) Lag1 0.222 (0.73)
Lag2 1.464** (0.01) Lag2 1.495** (0.01) Lag2 ‒ ‒
U.S.A. U.S.A. U.S.A.
Lag1 0.553 (0.11) Lag1 0.471 (0.17) Lag1 -0.461 (0.30)
Lag2 0.344 (0.32) Lag2 0.346 (0.32) Lag2 ‒ ‒
CY Ni Coef. (p-val.) CY Ni Coef. (p-val.) CY Ni Coef. (p-val.)
Con. −0.003 (0.87) Con. −0.005 (0.82) Con. 0.001 (0.98)
Own Var. Own Var. Own Var.
Lag1 −0.107 (0.25) Lag1 −0.353*** (0.01) Lag1 −0.165* (0.06)
Lag2 0.002 (0.98) Lag2 0.087 (0.53) Lag2 −0.046 (0.59)
Lag3 −0.100 (0.28) Lag3 0.151 (0.30) Lag3 −0.148* (0.08)
Lag4 −0.204** (0.03) Lag4 -0.202 (0.13) Lag4 −0.214** (0.01)
Downloaded by [University of California, San Diego] at 11:03 27 June 2016

Lag5 0.018 (0.85) Lag5 ‒ ‒ Lag5 −0.088 (0.31)


Invent. Invent. Invent.
Lag1 −0.215* (0.10) Lag1 −0.419 (0.46) Lag1 −0.196 (0.13)
Lag2 0.143 (0.31) Lag2 −0.027 (0.97) Lag2 0.157 (0.26)
Lag3 −0.340** (0.02) Lag3 −1.885** (0.01) Lag3 −0.288** (0.04)
Lag4 0.116 (0.40) Lag4 2.305*** (0.00) Lag4 0.052 (0.70)
Lag5 0.164 (0.20) Lag5 ‒ ‒ Lag5 0.211* (0.08)
Relat. Zn Relat. Li Relat. Cr
Lag1 0.180 (0.53) Lag1 0.578 (0.18) Lag1 0.305 (0.11)
Lag2 −0.106 (0.72) Lag2 0.470 (0.22) Lag2 0.207 (0.27)
Lag3 −0.433 (0.12) Lag3 −0.895** (0.01) Lag3 −0.068 (0.71)
Lag4 0.360 (0.21) Lag4 0.070 (0.84) Lag4 0.994 (0.57)
Lag5 −0.297 (0.28) Lag5 ‒ ‒ Lag5 0.312* (0.07)
CRB CRB CRB
Lag1 0.289 (0.72) Lag1 −0.373 (0.72) Lag1 −0.180 (0.82)
Lag2 0.072 (0.93) Lag2 1.335 (0.17) Lag2 −0.273 (0.75)
Lag3 0.106 (0.90) Lag3 0.965 (0.36) Lag3 0.247 (0.78)
Lag4 1.370 (0.12) Lag4 −0.487 (0.58) Lag4 1.444 (0.11)
Lag5 −0.972 (0.21) Lag5 ‒ ‒ Lag5 −0.692 (0.40)
U.S. U.S.A. U.S.A.
Lag1 −0.142 (0.77) Lag1 −0.113 (0.88) Lag1 −0.264 (0.58)
Lag2 −0.113 (0.82) Lag2 −1.406** (0.04) Lag2 −0.294 (0.54)
Lag3 0.398 (0.40) Lag3 0.192 (0.76) Lag3 0.337 (0.48)
Lag4 −0.694 (0.14) Lag4 0.333 (0.58) Lag4 −0.497 (0.29)
Lag5 0.706 (0.13) Lag5 ‒ ‒ Lag5 0.655 (0.17)
(c)
CY Ni Coef. (p-val.) CY Sn Coef. (p-val.) CY Zn Coef. (p-val.)
Con. 0.005 (0.74) Con. 0.000 (0.99) Con. 0.004 (0.76)
Own Var. Own Var. Own Var.
Lag1 −0.142* (0.09) Lag1 −0.048 (0.54) Lag1 −0.122 (0.11)
Lag2 ‒ ‒ Lag2 ‒ ‒ Lag2 −0.095 (0.21)
Invent. Invent. Invent.
Lag1 −0.154 (0.15) Lag1 −0.192 (0.10) Lag1 −0.146 (0.51)
Lag2 ‒ ‒ Lag2 ‒ ‒ Lag2 −0.203 (0.37)
Relat. Pb Relat. Pb Relat. Al
Lag1 0.524*** (0.01) Lag1 0.324 (0.13) Lag1 0.583* (0.06)
Lag2 ‒ ‒ Lag2 ‒ ‒ Lag2 −0.263 (0.36)

(Continued )
14 A. OMURA ET AL.

Table6. (Continued).
CRB CRB CRB
Lag1 0.410 (0.52) Lag1 1.109* (0.07) Lag1 0.036 (0.96)
Lag2 ‒ ‒ Lag2 ‒ ‒ Lag2 −0.274 (0.67)
U.S. U.S.A. U.S.A.
Lag1 −0.433 (0.33) Lag1 −0.028 (0.95) Lag1 −0.084 (0.83)
Lag2 ‒ ‒ Lag2 ‒ ‒ Lag2 0.668* (0.08)
CY Zn Coef. (p-val.) CY Zn Coef. (p-val.) CY Zn Coef. (p-val.)
Con. 0.005 (0.69) Con. 0.007 (0.60) Con. 0.004 (0.75)
Own Var. Own Var. Own Var.
Lag1 −0.092 (0.22) Lag1 −0.122 (0.12) Lag1 −0.105 (0.16)
Lag2 ‒ ‒ Lag2 −0.105 (0.17) Lag2 ‒ ‒
Invent. Invent. Invent.
Lag1 −0.386** (0.04) Lag1 −0.324 (0.11) Lag1 −0.289 (0.12)
Lag2 ‒ ‒ Lag2 −0.047 (0.82) Lag2 ‒ ‒
Relat. Cu Relat. Pb Relat. Ni
−0.015
Downloaded by [University of California, San Diego] at 11:03 27 June 2016

Lag1 (0.96) Lag1 0.164 (0.46) Lag1 0.315* (0.08)


Lag2 ‒ ‒ Lag2 0.190 (0.39) Lag2 ‒ ‒
CRB CRB CRB
Lag1 0.032 (0.96) Lag1 0.087 (0.89) Lag1 0.162 (0.76)
Lag2 ‒ ‒ Lag2 −0.482 (0.44) Lag2 ‒ ‒
U.S.A. U.S.A. U.S.A.
Lag1 0.113 (0.76) Lag1 −0.007 (0.99) Lag1 0.201 (0.59)
Lag2 ‒ ‒ Lag2 0.671* (0.08) Lag2 ‒ ‒
(d)
CY Cu Coef. (p-val.) CY Zn Coef. (p-val.) CY Cu Coef. (p-val.)
Con. 0.006 (0.70) Con. 0.004 (0.79) Con. 0.006 (0.69)
Own Var. Own Var. Own Var.
Lag1 −0.377*** (0.00) Lag1 −0.322*** (0.00) Lag1 −0.227*** (0.00)
Lag2 −0.260*** (0.01) Lag2 −0.171** (0.05) Lag2 ‒ ‒
Lag3 −0.063 (0.49) Lag3 0.008 (0.92) Lag3 ‒ ‒
Lag4 −0.018 (0.83) Lag4 ‒ ‒ Lag4 ‒ ‒
Invent. Invent. Invent.
Lag1 −0.128 (0.38) Lag1 −0.066 (0.63) Lag1 0.008 (0.95)
Lag2 0.162 (0.32) Lag2 0.091 (0.55) Lag2 ‒ ‒
Lag3 0.020 (0.90) Lag3 0.087 (0.52) Lag3 ‒ ‒
Lag4 0.213 (0.14) Lag4 ‒ ‒ Lag4 ‒ ‒
Hedge Hedge Hedge
Lag1 0.236 (0.27) Lag1 0.047 (0.81) Lag1 −0.107 (0.21)
Lag2 −0.362 (0.23) Lag2 −0.138 (0.62) Lag2 ‒ ‒
Lag3 0.210 (0.49) Lag3 −0.071 (0.72) Lag3 ‒ ‒
Lag4 −0.320 (0.13) Lag4 ‒ ‒ Lag4 ‒ ‒
Relat. Scrap Relat. Al Relat. Zn
Lag1 1.038** (0.02) Lag1 0.797** (0.03) Lag1 0.062 (0.85)
Lag2 1.377** (0.02) Lag2 0.122 (0.75) Lag2 ‒ ‒
Lag3 1.216** (0.04) Lag3 −0.509 (0.16) Lag3 ‒ ‒
Lag4 0.723 (0.15) Lag4 ‒ ‒ Lag4 ‒ ‒
CRB CRB CRB
Lag1 −0.662 (0.43) Lag1 −0.551 (0.48) Lag1 0.427 (0.51)
Lag2 0.800 (0.42) Lag2 1.004 (0.25) Lag2 ‒ ‒
Lag3 0.108 (0.91) Lag3 0.328 (0.66) Lag3 ‒ ‒
Lag4 0.357 (0.67) Lag4 ‒ ‒ Lag4 ‒ ‒
U.S.A. U.S.A. U.S.A.
Lag1 0.078 (0.87) Lag1 −0.504*** (0.01) Lag1 0.165 (0.70)
Lag2 0.281 (0.54) Lag2 0.130 (0.48) Lag2 ‒ ‒

(Continued )
APPLIED ECONOMICS 15

Table6. (Continued).
Lag3 −0.194 (0.67) Lag3 −0.294 (0.11) Lag3 ‒ ‒
Lag4 0.276 (0.55) Lag4 ‒ ‒ Lag4 ‒ ‒
Note: The tables report the results of the multivariate VAR
0 1 0 1 0 1 1 0 k 1
Δy1;t c1 β1;1 β11;h 0 Δy1;t1 1 β1;1 βk1;h 0 Δy1;tk 1 0 ε1;t 1
B : C B : C B : :: : CB : C B : :: : CB : C B : C
B C B C B C@ A þ ::: þ B C@
@ : A¼@ : Aþ@ : : A : @ : : A :
Aþ@
:
A
:: Δy :: Δy ε
Δyh;t ch βh;1 βh;h
1 1
h;t1 βh;1 βh;h
k k
h;tk h;t

where Δyh is the monthly change in the variable h. The table reports the regression results which use the monthly change in the convenience yield
of a metal that the model focuses as an explained variable, and a relative scarcity variable and other control variables as the explanatory
variables. The multivariate VAR models used in the table are developed from the above bivariate VAR by incorporating the control variables. CY is
the monthly change in the convenience yield of an indicated metal. Inventory is the monthly change in the inventory figures from the LME for
the explained variable, CRB is the monthly change in the CRB index and U.S.A. is the monthly change in the S&P 500 index. A relative scarcity
variable is indicated by Relat. and the subsequent word in the table. For instance, Relat. Al under the subheading of CY Zinc is the relative
scarcity of zinc to aluminium (computed as the spot price of zinc divided by the spot price of aluminium). Own Var. indicates for the
autoregressive variable. Con. is the constant term. Hedge is the hedging pressure of copper futures proxied by net commercial position/
(commercial long position + commercial short position) at the COMEX. All the variables, except the hedging pressure, are in monthly log change
figures. The standard errors in the results are adjusted for the small sample-size and degree-of-freedom in estimating the error variance–
Downloaded by [University of California, San Diego] at 11:03 27 June 2016

covariance matrix. ***, **, * denote the significance level at 1%, 5% and 10%, respectively.

results in our study is found when the tests are behaviour of convenience yield of copper dur-
conducted for different time periods. This indicates ing backwardation periods. However, during
that the relationships can be time varying as the contango periods, the convenience yield of
importance of other metals to an industrial metal copper is better explained by the price-ratio
changes over time. Furthermore, unlike energy com- against zinc.
modities such as crude oil, thermal coal and natural (3) For lead, many of the economic linkages
gas, it requires some time for one metal to be sub- are shown to be less significant after the
stituted by other metals. Hence, our results especially GFC. However, our backwardation/con-
for the substitution relationship are likely to be dri- tango subsample test shows that it is
ven by the expectations of market participants. recommended to observe the relative scar-
These expectations depend heavily on which indus- city against tin and zinc during the back-
try the market participants are paying attention to, wardation period, which represents about
further justifying why our significant results change 40% of our sample.
over time. (4) When it comes to nickel, market participants
are required to observe the behaviour of the
price-ratios for most of the economically
Economic implications linked metals; however, during contango,
The time-varying characteristic of the relation- information about lithium and chromium is
ships allows us to derive several economic impli- more effective than others. In relation to the
cations in determining the convenience yield of a results obtained for convenience yield of
metal. nickel, as the market size of this metal is
relatively small compared to aluminium, cop-
(1) In order to quantify the benefits of possessing per, lead and zinc, this may have caused the
physical aluminium especially during contango nickel market to be easily affected by the
periods (which represent 70% of our sample economic linkages to other metals. In parti-
dataset period), it is recommended for market cular, trading volume of the nickel futures at
participants including producers, stockists, man- the LME (2014) is merely a few per cent of the
ufacturers and traders to pay close attention to overall market.
the relative market condition against nickel. (5) The price-ratios against aluminium and
(2) The economic linkage with aluminium offers nickel, in general, have been playing an
meaningful information regarding the important role in determining the
16 A. OMURA ET AL.

Table 7. Multivariate Granger-causality test (p-values). metals (Fedorinova and Khrennikov 2012; Maverick
Full sample period 2014). The price ratio of aluminium over copper (at
CY Aluminium versus the LME) has gone up from slightly above 1 to more
Relat. Copper 0.59 than 4 over the last one and a half decades. This has
Relat. Nickel 0.08*
prompted products like high voltage cables for the
Relat. Zinc 0.42
Relat. Scrap 0.54 electricity supply industry to substitute copper with
CY Copper versus aluminium. Furthermore, industries including the
Relat. Aluminium 0.04** automobile industry where reducing weight of the
Relat. Zinc 0.85 products is becoming crucial are also replacing the
Relat. Scrap 0.06*
cables made of red metal with aluminium. Hence,
CY Lead versus
since the replacement is proceeding from copper to
Relat. Nickel 0.08*
Relat. Lithium 0.72
aluminium, perceived importance of the economic
Relat. Tin 0.19 linkage in determining the benefits of owning a
Relat. Zinc 0.03** physical metal differs among the two metals.
Downloaded by [University of California, San Diego] at 11:03 27 June 2016

CY Nickel versus
Relat. Aluminium 0.02**
Relat. Lead 0.01*** Battery-related metals
Relat. Lithium 0.03**
Relat. Chromium 0.10* Our study incorporates three battery-related economic
Relat. Zinc 0.48 linkages (lead, lithium and nickel) and it is shown that
CY Tin versus those relationships are important in determining the
Relat. Lead 0.13 convenience yield of lead and nickel. Interestingly,
CY Zinc versus while the relative scarcity of lead against both nickel
Relat. Aluminium 0.16
Relat. Copper 0.96
and lithium obtain significant results in explaining the
Relat. Lead 0.39 convenience yield of lead in the post-GFC period, the
Relat. Nickel 0.08* price-ratio against lithium is shown to be important in
Note: The table reports the results of the Granger-causality tests conducted understanding the behaviour of convenience yield of
based on the multivariate VAR in Table 6. Rows with a subheading Relat.
. . . report the p-values of Granger-causality tests with a null-hypothesis of nickel during the same period. This is possibly because
‘the relative scarcity variable of indicated metal does not Granger-cause (1) as a recent trend, nickel-ion batteries in some
convenience yield of a focused metal’.
industries are being replaced by lithium-ion batteries,
(2) nickel and lithium are interrelated by both substi-
convenience yield of zinc but during con- tution and complementary relationships.
tango, it is recommended to concern the rela-
tionship with lead.
Economic linkages between zinc and other
We further explore why the identified economic industrial metals
linkages yield different results. We focus on the results The convenience yield of zinc is found to be affected
obtained from the bi- and multivariate full-sample tests by the economic linkage against aluminium in the
and our rationale is mainly based on the difference in bivariate study and against nickel in the multivariate
the economic importance of a linkage to a metal. study. On the other hand, our full-sample period test
does not provide a significant result for relative
scarcity against copper or lead.
Aluminium and copper
We argue that this is possibly due to the fact that
The results indicate that the economic linkage galvanizing steel, the largest consuming industry for
between aluminium and copper is important in zinc, can be substituted by aluminium alloys and stain-
determining the perceived convenience yield of cop- less steel in some cases. From this reason, the market is
per but not of aluminium. In recent years for the paying more attention to the relative scarcity against
cable industry, replacement of copper with alumi- those metals. On the other hand, the relationship
nium is proceeding in various industries in response between zinc and copper is mainly driven by the brass
to the widening in the price gap between the two and bronze industry which only represents one-third of
APPLIED ECONOMICS 17

Table 8. Summary of results.


Bivariate Multivariate
- July 2008 July 2008 - January 2010 - Full Cont Back Full
CY Aluminium versus
Relat. Cu ✗ ✗ ✗ ✗ ✗ ✗ ✗
Relat. Ni ✓ ✓ ✗ ✓ ✓ ✗ ✓
Relat. Zn ✗ ✗ ✓ ✗ ✗ ✗ ✗
Relat. Scrap ✗ ✗ ✗ ✗ ✗ ✗ ✗
CY Copper versus
Relat. Al ✓ ✗ ✗ ✓ ✗ ✓ ✓
Relat. Zn ✗ ✗ ✗ ✗ ✓ ✗ ✗
Relat. Scrap ✓ ✗ ✗ ✗ ✗ ✗ ✓
CY Lead versus
Relat. Ni ✓ ✗ ✗ ✓ ✗ ✗ ✓
Relat. Li ‒ ✓ ✗ ✓ ✗ ‒ ✗
Relat. Zn ✓ ✗ ✗ ✓ ✗ ✓ ✗
Downloaded by [University of California, San Diego] at 11:03 27 June 2016

Relat. Sn ✓ ✗ ✗ ✗ ✗ ✓ ✓
CY Nickel versus
Relat. Al ✓ ✗ ✓ ✓ ✗ ✓ ✓
Relat. Pb ✓ ✓ ✗ ✓ ✗ ✓ ✓
Relat. Li ‒ ✓ ✓ ✓ ✓ ‒ ✓
Relat. Cr ✓ ✓ ✓ ✗ ✓ ✓ ✓
Relat. Zn ✗ ✗ ✓ ✓ ✗ ✓ ✗
CY Tin versus
Relat. Pb ✓ ✗ ✗ ✓ ✗ ✗ ✗
CY Zinc versus
Relat. Al ✓ ✗ ✓ ✓ ✗ ✓ ✗
Relat. Cu ✗ ✗ ✗ ✗ ✗ ✓ ✗
Relat. Pb ✗ ✗ ✗ ✗ ✓ ✓ ✗
Relat. Ni ✓ ✗ ✓ ✗ ✗ ✗ ✓
Note: The table reports the summary of the Granger-causality test results. CY . . . versus indicates that the following rows employ convenience yield of
indicated metal. Rows with a subheading Relat. . . . report the p-values of Granger-causality tests with a null-hypothesis of ‘the relative scarcity variable of
indicated metal does not Granger-cause convenience yield of a focused metal’. Cont stands for contango and Back stands for backwardation. A tick
indicates that the result is significant at the 10% level.

the galvanising demand (ILZSG 2014). A similar result relationship. We use the producer price index (PPI) in
is obtained when relative scarcity of copper against zinc the U.S.A. to proxy the secondary metal market.
is used to explain the convenience yield of copper. However, countries like Japan and China also actively
Furthermore, as zinc and lead can be extracted recycle non-ferrous metals and therefore the PPI in the
from the same deposit, the relationship between the U.S.A. may not well represent the overall scrap market.
two is likely to be complementary. Although no com-
prehensive information on the proportion of zinc
produced from the ores which also contain lead is VI. Concluding remarks
available, the importance of such a relationship is
expected to have a direct impact on our results. It is evident that possessing a physical asset offers
various benefits that cannot be enjoyed through
entering a futures contract. Studies related to the
theory of storage, traditionally, focus on explaining
Economic linkages between primary and secondary
those benefits of a commodity by observing the
metals
behaviour of its own inventory levels, in other
Lastly, we incorporate the economic linkages between words, the absolute scarcity level. However, the sup-
primary and secondary metals of copper and alumi- ply/demand environment of commodities is often
nium. It is found that, especially for aluminium, many economically related to the production and con-
of our test results do not support the significance of this sumption behaviour of other commodities. For this
18 A. OMURA ET AL.

reason, the convenience yield of a commodity addition to alerting them regarding the general
should be affected by the dynamic relationship with importance of the relative scarcity level, by utilis-
those related commodities. ing the time-varying characteristics of the relation-
We successfully identify thirteen sets of economic ships between convenience yield and the price-
linkages for six major base metals, namely, alumi- ratio, especially during contango and backwarda-
nium, copper, lead, nickel, tin and zinc. We examine tion periods, market participants would know
whether those economic linkages play an important which economic linkage they should refer to in
role in determining the perceived level of conveni- order to measure the benefits of possessing a phy-
ence yield. We employ a well-established VAR sical metal.
model and Granger-causality test approaches, both Secondly, as the benefits of physically owning a
bivariately and multivariately, to verify the signifi- metal are found to be also affected by the scarcity
cance of the relationships. Our study obtains mixed level against other metals, a withdrawal constraint at
but interesting results which generally support the warehouses for some metals that the LME concerns
presence of statistically positive economic linkages can also cause the deviation of the prices of other
Downloaded by [University of California, San Diego] at 11:03 27 June 2016

among base metals. economically related metals from their equilibrium


Overall, in addition to this general finding, a prices. Thus, our study can potentially help the LME
number of important implications can be derived to justify its proposal to implement a new withdra-
from our results. First, the significance of the rela- wal rule. A future study direction could be to further
tionship between metals can change over time. There generalize our findings and apply similar tests to
are sets of relationships existing during a certain other regional commodities such as agricultural
subsample period, but not occurring in other sub- products.
sample periods. This is reasonable because the appli-
cations that market participants consider as
important can change over time. These results offer Disclosure statement
crucial and highly practical implications to market No potential conflict of interest was reported by the
participants who are seeking to value the benefits of author(s).
obtaining a physical metal.
Second, it is found that the intensity of the influ-
ence of the price-ratio on the convenience yield References
defers depending on how important the linkage is Adhikari, K. R., and J. K. Putnam. 2014. Cross-Market
to the metal under consideration. In particular, our Dependence in the Commodity Futures Market.
results show that a linkage between two metals, A WorkingPaper. New Orleans, LA: University of New
and B, has a significant explanatory power over the Orleans.
Apple. 2015. Why Lithium-ion? Accessed April 6 2015.
convenience yield of metal A but not over metal B.
https://www.apple.com/au/batteries/why-lithium-ion/
For example, the null-hypothesis of no significant Black, F., and M. Scholes. 1973. “The Pricing of Options and
relationship cannot be rejected when the price ratio Corporate Liabilities.” Journal of Political Economy 81 (3):
between aluminium and copper is tested against the 637–654. doi:10.1086/260062.
convenience yield of aluminium, but can be rejected Brennan, M. J. 1958. “The Supply of Storage.” The American
in a number of models when the ratio is tested Economic Review 48 (1): 50–72. doi:10.2307/1812340.
Casassus, J., P. Liu, and K. Tang. 2013. “Economic Linkages,
against the convenience yield of copper. This differ-
Relative Scarcity, and Commodity Futures Returns.”
ence in the results can be attributed to a recent trend Review of Financial Studies 26 (5): 1324–1362.
of replacing copper with aluminium in some indus- doi:10.1093/rfs/hhs127.
tries thanks to the widening in the price gap between Chng, M. T. 2009. “Economic Linkages across Commodity
the two metals. Futures: Hedging and Trading Implications.” Journal of
Our findings have practical implications to var- Banking & Finance 33 (5): 958–970. doi:10.1016/j.
jbankfin.2008.10.006.
ious stakeholders. Firstly, producers, stockists,
Chng, M. T. 2010. “Comparing Different Economic Linkages
manufacturers and traders can improve the preci- Among Commodity Futures.” Journal of Business Finance
sion of estimating the perceived convenience yield & Accounting 37 (9–10): 1348–1389. doi:10.1111/
of major base metals using our findings. In jbfa.2010.37.issue-9-10.
APPLIED ECONOMICS 19

Cortazar, G., C. Milla, and F. Severino. 2008. “A Hub Trading.” Journal of Futures Markets 32 (5): 459–
Multicommodity Model of Futures Prices: Using Futures 479. doi:10.1002/fut.20523.
Prices of One Commodity to Estimate the Stochastic ILZSG. 2014. “End Uses of Lead and Zinc.” International
Process of Another.” Journal of Futures Markets 28 (6): Lead and Zinc Study Group, 2014. Accessed July 31 2014.
537–560. doi:10.1002/fut.20322. http://www.ilzsg.org/static/enduses.aspx?from=1
Deaton, A., and G. Laroque. 1992. “On the Behaviour of J.P. Morgan, and Reuters. 1996. Riskmetrics(TM): Technical
Commodity Prices.” The Review of Economic Studies 59 Document. New York: Morgan Guaranty Trust Company
(1): 1–23. doi:10.2307/2297923. of New York.
Dickey, D. A., and W. A. Fuller. 1979. “Distribution of the Kaldor, N. 1939. “Speculation and Economic Stability.” The
Estimators for Autoregressive Time Series with a Unit Review of Economic Studies 7 (1): 1–27. doi:10.2307/
Root.” Journal of the American Statistical Association 74 2967593.
(366a): 427–431. doi:10.1080/01621459.1979.10482531. Kocagil, A. E. 2004. “Optionality and Daily Dynamics of
Dockner, E. J., Z. Eksi, and M. Rammerstorfer. 2015. “A Convenience Yield Behavior: An Empirical Analysis.”
Convenience Yield Approximation Model for Mean- Journal of Financial Research 27 (1): 143–158.
Reverting Commodities.” Journal of Futures Markets 35 doi:10.1111/j.1475-6803.2004.00082.x.
(7): 625–654. doi:10.1002/fut.21670. Lin, W. T., and C.-W. Duan. 2007. “Oil Convenience Yields
Erheriene, E. 2015. “LME Launches New Round of Estimated Under Demand/Supply Shock.” Review of
Downloaded by [University of California, San Diego] at 11:03 27 June 2016

Warehousing Rules Proposals.” The Wall Street Journal, Quantitative Finance and Accounting 28 (2): 203–225.
July 1 2015. doi:10.1007/s11156-006-0008-5.
Fama, E. F., and K. R. French. 1988. “Business Cycles and the Liu, P., and K. Tang. 2010. “No-Arbitrage Conditions for
Behavior of Metals Prices.” The Journal of Finance 43 (5): Storable Commodities and the Modeling of Futures Term
1075–1093. doi:10.1111/j.1540-6261.1988.tb03957.x. Structures.” Journal of Banking & Finance 34 (7): 1675–
Fedorinova, Y., and I. Khrennikov. 2012. “Aluminum Over 1687. doi:10.1016/j.jbankfin.2010.03.013.
Copper for Cables Helps Rusal Alcoa: Commodities.” London Metal Exchange. 2013. FAQs-LME Warehousing
Bloomberg, February 8 2012. Consultation 2013. Accessed April 20 2013. http://www.
Fischer, S. 1978. “Call Option Pricing When the Exercise lme.com/~/media/Files/Warehousing/Warehouse%20con
Price is Uncertain, and the Valuation of Index Bonds.” sultation/FAQs.pdf
The Journal of Finance 33 (1): 169–176. doi:10.1111/ London Metal Exchange. 2014. Trading Volumes 2014.
j.1540-6261.1978.tb03396.x. Accessed July 31 2014. http://www.lme.com/metals/
Ford. 2012. Ford’s New Li-Ion Batteries Reduce Use of Rare reports/monthly-volumes/annual/2013/
Earth Metals, Enable Superior Fuel Economy for Fusion, Maverick, T. 2014. “Is Aluminum the New Copper?” Wall
C-Max Hybrids 2012. Accessed June 4 2015. https://media. Street Daily, August 26 2014.
ford.com/content/fordmedia/fna/us/en/news/2012/09/13/ Mensi, W., M. Beljid, A. Boubaker, and S. Managi. 2013.
ford_s-new-li-ion-batteries-reduce-use-of-rare-earth- “Correlations and Volatility Spillovers across Commodity
metals–ena.html and Stock Markets: Linking Energies, Food, and Gold.”
Geman, H., and W. O. Smith. 2013. ““Theory of Storage, Economic Modelling 32: 15–22. doi:10.1016/j.
Inventory and Volatility in the LME Base Metals.” econmod.2013.01.023.
Resources Policy 38 (1): 18–28. doi:10.1016/j. Metal Bulletin. 2013. “Jiangxi Copper Agrees TC/RCs at $70/
resourpol.2012.06.014. 7c; Impasse with BHP.” Metal Bulletin Weekly. London:
Gorton, G. B., F. Hayashi, and K. Geert Rouwenhorst. Euromoney Institutional Investor PLC.
2013. “The Fundamentals of Commodity Futures Milonas, N. T., and S. B. Thomadakis. 1997. “Convenience
Returns.” Review of Finance 17 (1): 35–105. Yields as Call Options: An Empirical Analysis.” The
doi:10.1093/rof/rfs019. Journal of Futures Markets 17 (1): 1–15. doi:10.1002/
Hameed, A. 1997. “Time-Varying Factors and Cross- (SICI)1096-9934(199702)17:1<1::AID-FUT1>3.0.CO;2-P.
Autocorrelations in Short-Horizon Stock Returns.” Nippon Steel & Sumitomo Metal Corporation. 2012.
Journal of Financial Research 20 (4): 435–458. SuperDyma. http://www.nssmc.com/product/catalog_
doi:10.1111/jfir.1997.20.issue-4. download/pdf/U030.pdf
Heinkel, R., M. E. Howe, and J. S. Hughes. 1990. Nissan. 2013. Nissan Begins U.S. Assembly of 2013 LEAF
“Commodity Convenience Yields as an Option Profit.” Electric Vehicle and Batteries 2013 Accessed June 4 2015.
The Journal of Futures Markets 10 (5): 519–533. http://nissannews.com/en-US/nissan/usa/releases/nissan-
doi:10.1002/(ISSN)1096-9934. begins-u-s-assembly-of-2013-leaf-electric-vehicle-and-
Hirshleifer, D. 1990. “Hedging Pressure and Futures Price batteries
Movements in a General Equilibrium Model.” Omura, A., N. Todorova, L. Bin, and R. Chung. 2015.
Econometrica 58 (2): 411–428. doi:10.2307/2938209. “Convenience Yield and Inventory Accessibility: Impact
Hochradl, M., and M. Rammerstorfer. 2012. “The of Regional Market Conditions.” Resources Policy 44 (0):
Convenience Yield Implied in European Natural Gas 1–11. doi:10.1016/j.resourpol.2014.12.002.
20 A. OMURA ET AL.

Omura, A., and J. West. 2015. “Convenience Yield and the Supply Risk?” Resources Policy 38 (3): 395–405.
Theory of Storage: Applying an Option-Based Approach.” doi:10.1016/j.resourpol.2013.06.001.
Australian Journal of Agricultural and Resource Economics Toyota. 2014. 2014 Toyota Prius Plug-in Hybrid is
59 (3): 355–374. doi:10.1111/1467-8489.12092. Evolutionary - and a Little Bit Revolutionary 2014.
Parkinson, M. 1980. “The Extreme Value Method for Accessed June 4 2015. http://toyotanews.pressroom.
Estimating the Variance of the Rate of Return.” The toyota.com/releases/2015+toyota+prius+plug+in+hybrid
Journal of Business 53 (1): 61–65. doi:10.1086/296071. +revolutionary.htm
Roon, F. A., T. E. Nijman, and C. Veld. 2000. “Hedging Troszkiewicz, A., and M. Kolesnikova. 2013. “LME Alters
Pressure Effects in Futures Markets.” The Journal of Warehousing Rules to Shorten Withdrawal Times.”
Finance 55 (3): 1437–1456. doi:10.1111/0022-1082.00253. Bloomberg, November 8 2013.
Rossi, B. 2012. “The Changing Relationship Between West, J. 2012. “Convenience Yields in Bulk Commodities:
Commodity Prices and Equity Prices in Commodity The Case of Thermal Coal.” The International Journal of
Exporting Countries.” IMF Economic Review 60 (4): 533– Business and Finance Research 6 (4): 33–44.
569. doi:10.1057/imfer.2012.20. Working, H. 1948. “Theory of the Inverse Carrying Charge
Routledge, B. R., D. J. Seppi, and C. S. Spatt. 2000. “Equilibrium in Futures Markets.” Journal of Farm Economics 30 (1): 1–
Forward Curves for Commodities.” The Journal of Finance 28. doi:10.2307/1232678.
55 (3): 1297–1338. doi:10.1111/0022-1082.00248. Working, H. 1949. “The Theory of Price of Storage.” The
Downloaded by [University of California, San Diego] at 11:03 27 June 2016

Souček, M., and N. Todorova. 2013. “Realized Volatility American Economic Review 39 (6): 1254–1262.
Transmission between Crude Oil and Equity Futures doi:10.2307/1816601.
Markets: A Multivariate HAR Approach.” Energy Xiarchos, I. M., and J. J. Fletcher. 2009. “Price and Volatility
Economics 40 (0): 586–597. doi:10.1016/j.eneco.2013.08.011. Transmission between Primary and Scrap Metal Markets.”
Stepanek, C., M. Walter, and A. Rathgeber. 2013. “Is the Resources, Conservation and Recycling 53 (12): 664–673.
Convenience Yield a Good Indicator of a Commodity’s doi:10.1016/j.resconrec.2009.04.020.

You might also like