You are on page 1of 13

Resources Policy xxx (xxxx) xxx–xxx

Contents lists available at ScienceDirect

Resources Policy
journal homepage: www.elsevier.com/locate/resourpol

The role of trading volume, open interest and trader positions on volatility
transmission between spot and futures markets

Beyza Mina Ordu-Akkayaa, , Ecenur Ugurlu-Yildirimb,c, Ugur Soytasc,d
a
Social Sciences University of Ankara, Department of Business Administration, Hükümet Meydanı No: 2 06030, Altındağ, Ankara, Turkey
b
Ankara Yıldırım Beyazıt University, Department of International Trade and Business, Esenboğa Külliyesi, 06970 Esenboğa, Ankara, Turkey
c
Middle East Technical University, Department of Business Administration, 06531 Ankara, Turkey
d
Middle East Technical University, Department of Earth System Science, 06531 Ankara, Turkey

A R T I C LE I N FO A B S T R A C T

JEL codes: In this paper, we investigate the role of open interest, trading volume and trading positions of trader groups on
Q02 volatility spillover between futures and spot markets of two major commodities; oil and gold during the last two
Q40 decades. The initial analysis including only spot and futures markets imply that the relationship is bi-directional
G12 for crude oil, and uni-directional for gold. Though, including open interest and trading volume enrich our results
Keywords: indicating open interest and spot markets are closely connected and trading volume provide cross-market in-
Commodity formation, which might suggest investors investing in both commodities make these markets informationally
Spillover connected. Given the increasing presence of institutional investors in commodity markets during the sample
Open interest
period, we also check whether speculators lead to excess volatility in futures market as financialization pro-
Trading volume
ponents argue. Findings depict that actually the spillover is from futures market to speculators’ positions im-
Speculators
Financialization plying volatility in commodity markets is not attributable to speculators in the last two decades.

1. Introduction and a review of literature Glück, 2015). Financialization proponents statistically show specific
type of traders’ positions (money managers, swap dealer, producer etc.)
The debate on the role of futures market leading to excess volatility Granger-cause futures prices and volatilities (e.g. Stoll and Whaley,
in spot markets, dates back to 19th century with the introduction of 2010) and Hong and Yogo (2012) show trading volume and open in-
futures market in major commodities. Therefore studies examining the terest are highly informative on commodity returns, as well. Therefore,
linkage between futures and spot prices have been highly crucial in the we investigate the role of trading volume, open interest and different
last century (Jacks, 2007). This discussion has got even deeper after the types of traders’ positions on the link between spot - futures prices of two
heavy influx of institutional investors to commodity markets, which are major commodities, gold and oil.
referred to as financialization of commodity markets in the literature Our research paper mainly has three stages of analysis. In the first
(Cheng and Xiong, 2014). US Commodity Futures Trading Commission stage, we include solely spot and futures markets in our analysis and
(CFTC) (2008) present that institutional holdings in oil futures dis- examine the interaction between them. Cheng and Xiong (2014) state
played a dramatic increase from USD 13 billion in 2003 to over USD commodity futures market affects commodity prices through three
200 billion in 2008. This increase can be attributed to two main ad- mechanisms. The first and major mechanism asserts that there is an
vantages of commodities; low correlation with traditional financial intrinsic storage or consumption decision in commodities. If market
assets and hedging opportunities against inflation (Ordu et al., 2017). players opt for storing, believing prices would get higher in the future, a
Even though some argue financialization is a passing trend which link between futures and commodity market is generated (Kaldor,
temporarily increases connections between markets (e.g., Fattouh et al., 1939; Working, 1949; Brennan, 1958). The second mechanism is risk
2013), others contend it is a permanent phenomenon (Adams and sharing, which argues that market players share the commodity price


Corresponding author.
E-mail address: beyza.akkaya@asbu.edu.tr (B.M. Ordu-Akkaya).

https://doi.org/10.1016/j.resourpol.2018.02.005
Received 1 December 2017; Received in revised form 29 January 2018; Accepted 7 February 2018
0301-4207/ © 2018 Elsevier Ltd. All rights reserved.

Please cite this article as: Ordu-Akkaya, B.M., Resources Policy (2018), https://doi.org/10.1016/j.resourpol.2018.02.005
B.M. Ordu-Akkaya et al. Resources Policy xxx (xxxx) xxx–xxx

Fig. 1. Crude oil and gold prices. Fig. 1 presents the historical spot prices of crude oil (CSPOT) and gold (GSPOT). Crude oil prices are represented by the left and gold by the right axis.

risk via commodity futures market (Keynes, 1923; Hicks, 1939). The positions2 for crude oil and gold and investigate the role of these po-
last mechanism is information discovery which argues futures markets sitions on volatility spillover between spot and futures markets. Basak
aggregate information (weather, demand/ supply, inventory, economic and Pavlova (2016) theoretically show that the presence of institutional
cycles, and business conditions etc.) and incorporate this information in investors increase the volatility of storable commodities. Under absence
futures prices of each commodity. Hence, first of all, we investigate the of institutions, prices are mainly determined through supply and de-
volatility spillover between futures and spot markets of two highly mand risks, whereas with the inflow of institutions a new type of risk
traded commodities; crude oil and gold markets. These commodities are emerges; which is ‘falling behind the index’. This risk contends that
the main representatives of the large commodity markets (Zhang and institutional investors strive to match the performance of the index,
Wei, 2010).1 Therefore, information transmission between the spot and since inflow to their institution is dependent upon how much they
futures markets of these major commodities is of utmost importance to surpass index levels. And this risk increases volatility for commodity
traders and investors. To address this area, we do not only check the markets. However, recent studies point out that results depend on the
spillover between gold and crude oil's own spot and futures markets, commodity market in question (Gozgor et al., 2016; Adams and Glück,
but also cross-market spillovers. As one can note from the historical 2015). Hence, we check whether this is the case for crude oil and gold
price plot of these commodities in Fig. 1, they display similar fluctua- during our sample period.
tions over time. Even though several studies check the spillover be- The contribution of our paper is mainly two-folds. First of all, we
tween gold and crude oil (e.g., Baffes, 2007; Narayan et al., 2010; utilize a rather new but intuitive methodology developed by Diebold
Zhang and Wei, 2010; Ewing and Malik, 2013), they do not consider and Yilmaz (2012). This method does not only take the dynamic aspect
cyclical changes in the economic environment. Although some papers of spillovers into account, but also shows the “net” direction of spil-
find significant spillovers (Adams and Glück, 2015) and some do not lover. Even if statistically there might be spillover in both directions, we
(Brunetti et al., 2016), our methodology will shed more light on this can identify which asset is the net transmitter of volatility. A market
issue by considering time-varying characteristics of the last two dec- may switch roles from being a net volatility transmitter/receiver to a
ades. net receiver/transmitter over time. We also attempt to capture this role
In the second stage, following Hong and Yogo (2012) we include switching behavior. Secondly, we do not solely check the spillover
open interest and trading volume information of both commodities. between futures and spot markets, which is already a very rich litera-
Traders should consider not only heterogeneous expectations and nu- ture; but we include new phenomenon of financialization into our
merous information, but also trading activities of new groups of in- study. Moreover, the idea that open interest and trading volume in-
vestors. Trading activities, especially volume and open interest has corporate more information on future prices than any other variable is
been on the radar of researchers for a long while. Several studies check rather new (Hong and Yogo, 2012). Therefore, through this way, we
spillovers between selected energy asset classes (e.g.,Gormus et al. aim to enrich established literature on spot-futures market nexus and
(2014)) or geographical markets (e.g., Kocaarslan et al., 2017). How- newly growing literature on financialization of commodities.
ever, not many include trading volume or open interest in commodity The major empirical findings of our study are as follows. The primary
market analysis. Since, Hong and Yogo (2012) show that open interest analysis between futures and spot markets show that the relationship is
helps to predict not only commodity returns, but also business cycles bi-directional for crude oil, and uni-directional for gold. Moreover, gold is
and macroeconomic environment; we also include such variables. a more internally oriented commodity and seems to have its own dy-
Therefore, in the next step of our analysis we examine the role of namics. We also find that open interest and spot markets are significantly
trading volume and open interest on leading to volatility in futures and connected. Furthermore, trading volume provides cross-market informa-
spot markets. Given the increasing interest of institutional investors in tion, which might be implying investors investing in both crude oil and
commodity markets, these liquidity variables may contain important gold markets make these markets informationally connected. Finally, we
information regarding the price and risk dynamics in these markets. find that positions of all traders and commodity markets are closely in-
The heavy flood of institutional investors in commodity markets terconnected. Speculators, particularly, do not lead to increased volati-
further heated the debate on the impact of transaction quantities by lities in futures market. In fact, spillover seems to work in opposite di-
different trader groups on the futures/spot markets. Hence, in the third rection. Similarly, gold money manager positions are net volatility
stage of our analysis we employ weekly disaggregated data on trader receivers throughout the sample period, whereas crude oil money man-
ager positions sometimes become a net transmitter. The change in roles
could be attributable to the change in the price of commodities.
1
Gold is an important precious metal that is viewed as safe haven during the times of
crises by both individual and institutional investors and gold stock is worth around USD 9
trillion (Erb and Harvey, 2013). Similarly, crude oil is a very significant commodity not
2
only as an energy input but also as an investment tool. The investment in commodity Following the public commentary (CFTC, 2006), CFTC decided to publish
trading strategies grew from USD 13 billion in 2003 to USD 317 billion in July 2008 Disaggregated Commitments of Traders (DCOT) reports, which provide weekly positions
(Masters, 2008). of swap dealers, money managers and producers.

2
B.M. Ordu-Akkaya et al. Resources Policy xxx (xxxx) xxx–xxx

Table 1
Descriptive statistics and correlation matrix for daily analysis.

Panel A

L-CFUT L-CFUTOI L-CFUTV L-CSPOT L-GFUT L-GFUTOI L-GFUTV L-GSPOT

Mean 57 1116,403 453,797 57 817 337,348 114,378 817


Median 51 1194,475 390,132 51 718 362,935 94,628 720
Maximum 145 2488,716 2530,530 145 1889 657,776 897,219 1898
Minimum 11 0 0 11 253 4843 0 253
Std. Dev. 30 533,456 343,641 30 480 136,901 90,350 481
Skewness 0.38 0.17 1.28 0.38 0.32 −0.01 1.24 0.32
Kurtosis 2.07 1.87 4.89 2.08 1.68 1.93 5.45 1.68
Observations 5261 5261 5261 5261 5261 5261 5261 5261

Panel B

CFUT CFUTOI CFUTV CSPOT GFUT GFUTOI GFUTV GSPOT

Mean 0.02 0.01 0.32 0.02 0.01 0.04 0.46 0.01


Median 0.02 0.01 0.26 0.02 0.01 0.03 0.39 0.01
Maximum 0.07 0.05 1.54 0.07 0.03 12.47 2.54 0.03
Minimum 0.01 0.01 0.18 0.01 0.00 0.00 0.26 0.00
Std. Dev. 0.01 0.00 0.16 0.01 0.00 0.22 0.20 0.00
Skewness 2.01 3.49 3.00 1.86 1.89 48.77 3.69 1.87
Kurtosis 8.82 22.57 13.35 7.73 7.52 2533.38 27.17 7.15
Observations 5259 5259 5259 5259 5259 5259 5259 5259

Panel C

CFUT CFUTOI CFUTV CSPOT GFUT GFUTOI GFUTV GSPOT

CFUT 1.00
CFUTOI 0.11 1.00
CFUTV 0.09 0.21 1.00
CSPOT 0.97 0.11 0.10 1.00
GFUT 0.34 −0.01 −0.16 0.32 1.00
GFUTOI −0.01 0.01 0.02 −0.01 0.00 1.00
GFUTV 0.01 0.24 0.54 0.02 −0.15 0.22 1.00
GSPOT 0.32 −0.04 −0.20 0.29 0.96 −0.01 −0.21 1.00

This table gives descriptive statistics for the first part of the analysis (Daily). Data is available in daily frequency and for the period between 25 August 1997 and 17 October 2017 for all
variables. Panel A presents statistics of variables in levels. Panel B shows GARCH (1,1) volatilities of same variables. CFUT and GFUT are continuous futures prices of crude oil and gold,
respectively. CSPOT is the spot US dollar prices of WTI Cushing Oklahoma per barrel and GSPOT is the spot gold bullion price per ounce. CFUTOI (GFUTOI) and CFUTV (GFUTV) are open
interest positions of crude oil (gold) and trading volume, respectively. Trading volume is the total quantity of contracts traded and open interest is the number of contracts outstanding at
the end of each trading day. Panel C presents correlation matrix of selected variables.

Our results provide important policy-making implications. First of autoregression method of Sims (1980) and forecast error variance de-
all, we show that futures markets do not lead to excess volatility in spot composition (FEVD). FEVD investigates the contribution of each vari-
markets. Therefore, unlike the common perception in public about fu- able on forecasting the H -step ahead variance on itself or another
tures market (Jacks, 2007), it is not an evil market and thus policy- variable. So, FEVD provides one to analyze the percentage share of
makers do not necessarily be worried about the impact of futures on variance of gold spot prices on crude oil prices. One should note that DY
spot markets. Second, open interest and trading volume might in- employs generalized frameworks (Koop et al., 1996; Pesaran and Shin,
formation about crude oil and gold. There is no harm for policy-makers 1998) but not the traditional method using Cholesky-decomposition.
to share the information on trading volume and open interest, since Cholesky-decomposition urges the user to order variables upon the
although these two provide information about the market, they do not importance of the impact on other variables. However, one might not
increase the volatility. Last but not least, our results imply that futures always be equipped with this information. Hence, generalized frame-
market do not harm small investors via speculations, since speculators work covers this issue and allows the user to include variables in-
do not increase volatility. This result is in line with Brunetti et al. dependent from ordering.
(2016) who claim that speculators do not lead to increased volatilities First, the analysis runs a 4-lag3 vector autoregression based on SIC
in futures market; but the opposite is true. Therefore, policy-makers do criterion with N -number of variables as in Eq. (1). Based on the number
not need to interfere specific markets or market-players as unlike al- of endogenous variables, coefficient matrices are [N *N ]; and εt is the
leges in literature. In fact, sharing such information increases the error term and is a vector of independent and identically distributed
transparency and would benefit public, even more. disturbances as (0, Σ).
The paper is organized as follows. Second and third sections present
4
the empirical method and data, respectively. Results are displayed in
the following section. The final section concludes.
Yt = ∑ θi Yt−i + εt
i=1 (1)

2. Method The moving average representation of Eq. (1) is;

We mainly employ Diebold and Yilmaz (2012) (DY) framework for


3
our analysis. This framework is built upon the very well-known vector The results are very similar when we use lags between 2 and 8.

3
B.M. Ordu-Akkaya et al. Resources Policy xxx (xxxx) xxx–xxx

Table 2
Descriptive statistics and correlation matrix for weekly analysis.

Panel A

L-CFUT L-CMM L-CPU L-CSD L-CSPOT L-GFUT L-GMM L-GPU L-GSD L-GSPOT

Mean 76 790,939 619,876 870,813 76 1183 177,999 215,109 159,975 1183


Median 77 772,503 580,186 875,439 77 1221 176,712 205,701 150,838 1223
Maximum 141 1466,645 1094,206 1095,025 141 1871 333,083 380,866 301,934 1888
Minimum 28 402,537 325,917 587,303 28 564 78,066 77,910 80,704 569
Std. Dev. 24 237,594 162,872 88,866 24 312 45,664 74,561 50,125 312
Skewness 0.04 0.42 1.09 −0.36 0.03 −0.09 0.50 0.19 0.71 −0.09
Kurtosis 2.10 2.39 3.79 3.63 2.10 2.39 3.69 2.06 2.87 2.39
Observations 593 593 593 593 593 593 593 593 593 593

Panel B

CFUT CMM CPU CSD CSPOT GFUT GMM GPU GSD GSPOT

Mean 0.05 0.04 0.06 0.03 0.05 0.03 0.06 0.06 0.05 0.02
Median 0.04 0.04 0.06 0.03 0.05 0.02 0.06 0.06 0.05 0.02
Maximum 0.14 0.11 0.07 0.03 0.19 0.04 0.16 0.11 0.10 0.05
Minimum 0.02 0.04 0.05 0.02 0.03 0.02 0.05 0.03 0.04 0.02
Std. Dev. 0.02 0.01 0.00 0.00 0.02 0.00 0.01 0.02 0.01 0.00
Skewness 1.56 5.21 1.32 −6.08 2.81 1.70 2.96 0.58 0.98 1.76
Kurtosis 6.12 43.94 5.66 67.64 14.71 5.86 15.30 2.93 3.84 5.96
Observations 591 591 591 591 591 591 591 591 591 591

Panel C

CFUT CMM CPU CSD CSPOT GFUT GMM GPU GSD GSPOT

CFUT 1.00
CMM 0.07 1.00
CPU 0.09 0.36 1.00
CSD −0.05 −0.03 0.05 1.00
CSPOT 0.95 0.05 0.08 −0.03 1.00
GFUT 0.39 0.07 0.09 −0.18 0.45 1.00
GMM 0.09 0.46 0.16 0.06 0.08 0.07 1.00
GPU 0.39 −0.05 0.11 −0.01 0.31 −0.12 0.02 1.00
GSD 0.13 0.11 −0.03 0.00 0.06 0.05 0.22 0.25 1.00
GSPOT 0.37 0.07 0.08 −0.24 0.41 0.99 0.07 −0.13 0.04 1.00

This table gives descriptive statistics for the second part of the analysis (Weekly). Data is available in weekly frequency and for the period between 27 June 2006 and 17 October 2017 for
all variables. Panel A presents statistics of variables in levels. Panel B shows GARCH (1,1) volatilities of same variables. Panel C presents correlation matrix. CFUT and GFUT are
continuous futures prices of crude oil and gold, respectively. CSPOT is the spot US dollar prices of WTI Cushing Oklahoma per barrel and GSPOT is the spot gold bullion price per ounce.
CMM (GMM), CSD (GSD) and CPU (GPU) are weekly open interest positions of Money Managers, Swap Dealers and Producers, respectively. Since all these traders have both long and
short positions, we compute open interest via (long +short +2*spreading) for the first two groups and via (long +short) for producers. Spread positions are simultaneous buying and
selling and are valid for speculators.


to 1 and thus each entry is normalized by the row sum as follows;
Yt = ∑ Ai εt−i
i=0 (2) ∼ δij
δij = N
or more simply Yt = A (L) ut . Moving average coefficients denoted by ∑ j = 1 δij (4)
A (L) is of utmost importance in understanding the dynamic links be- N N
By construction, = 1 and moreover
∑ j = 1 δij = N . Now, we
∑i, j = 1 δij
tween variables. These coefficients allow dividing the H -step-ahead
are equipped with available measures to calculate gross and net spil-
forecast error variances of each variable into parts attributable to the
lovers. The gross connectedness is basically the off-diagonal entry and is
various system shocks. There are hundreds of moving average coeffi-
as follows for the spillover from i to j and j to i , respectively.
cients to interpret (i.e., A (L) = A0 + A1 L+A2 L + …,) , but variance de-
N ∼ N ∼
composition framework transforms them to a readable format. ∑ j = 1 δij ∑ j = 1 δji
Then the FEVD for H-step ahead is as follows: GSi ← j = *100; GSi → j = *100
N N (5)
H −1
σjj−1 ∑h = 0 (e′i Ah Σej )2 Next, one can compute the net spillover, which is;
δij = H −1
∑h = 0 (e′i Ah ΣA′h ei )2 (3) NSij = GSj ← i − GSi ← j. (6)

where σjj stands for the standard deviation of εj , Σ is the covariance Diebold and Yilmaz (2012) also compute a spillover index, which is
matrix for the error vector, ej is a selection N *1 vector with j -th element basically a measure of the extent of spillover. To compute the spillover
unity and zeroes otherwise. As a result, a N *N matrix is produced, of index, we sum all off-diagonal elements and divide it by total column or
which each entry provides the contribution of variable j on the forecast row sums including the diagonal elements. Therefore the denominator
error variance of variable i . In the matrix, diagonals and off-diagonals is always N *100 .
present own contributions (variable i to itself) and pairwise-contribu- N ∼
∑i, j = 1, i ≠ j δij
tions (variable i to variable j ), respectively. However, row sums in TSi ← j =
N *100 (7)
generalized variance decomposition matrices are not necessarily equal

4
B.M. Ordu-Akkaya et al. Resources Policy xxx (xxxx) xxx–xxx

Table 3
Volatility spillover between futures and spot markets.

Panel A – Spillover table (H = 10 days)


CFUT CSPOT GFUT GSPOT From others

CFUT 54.4 42.6 2.0 1.0 45.6


CSPOT 43.1 54.9 1.5 0.6 45.1
GFUT 1.6 1.1 60.0 37.3 40.0
GSPOT 0.5 0.4 39.2 60.0 40.1
Contribution to others 45.2 44.1 42.7 38.8 170.7
Contribution including own 99.6 99.0 102.6 98.8 42.7%

Panel A – Spillover table (H=1 days)


CFUT 56.0 42.1 1.4 0.5 44.0
CSPOT 42.3 56.3 1.0 0.5 43.8
GFUT 1.7 1.2 66.4 30.8 33.6
GSPOT 0.6 0.6 31.3 67.5 32.5
Contribution to others 44.6 43.8 33.7 31.8 153.9
Contribution including own 100.6 100.1 100.1 99.3 38.5%

Panel B – Net spillovers (H= 10 days)

CFUT CSPOT GFUT GSPOT

CFUT
CSPOT 0.5
GFUT −0.4 −0.3
GSPOT −0.5 −0.2 1.9

Panel B – Net spillovers (H = 1 days)

CFUT CSPOT GFUTOI GSPOT

CFUT
CSPOT 0.2
GFUT 0.3 0.2
GSPOT 0.1 0.1 0.5

This table presents the volatility spillover tests utilizing Diebold and Yilmaz (2012) framework. Panel A shows the gross spillover findings. The percentage in the below right hand corner
of the table is the spillover index and shows the share of spillovers in explaining forecast error variance of all assets, on average. It is the grand off-diagonal sum divided to the grand sum
including diagonals, expressed as a percentage. Contribution to others and from others figures are off-diagonal column and row sums, respectively. Panel B depicts the net (pairwise)
spillovers between each pair. One can find the directional spillover via deducting same pairs’ gross spillovers from each other. For instance the net spillover is [43.1–42.6 = 0.5] between
oil futures and spot markets. Positive (Negative) figure shows spillover from column (row) to row (column). CFUT and GFUT are continuous futures prices of crude oil and gold,
respectively. CSPOT is the spot US dollar prices of WTI Cushing Oklahoma per barrel and GSPOT is the spot gold bullion price per ounce.

of futures markets is believed to be this information (Cheng and Xiong,


2014). Hence, we check whether there is an information spillover be-
tween futures and spot markets of crude oil and gold and amongst each
other's markets. We utilize daily spot and futures prices of gold and
Fig. 2. Gross and net volatility network between oil and gold markets. We present crude oil for the period between 25 August 1997 and 17 October 2017
the volatility network between oil and gold markets based upon Diebold-Yilmaz output.
obtained from Datastream. Our data is on a trading-day basis and takes
First we compute gross spillovers between Asset i and Asset j and take an average of these
log returns of all variables via the conventional method of
spillovers. If the average spillover is greater than 1 (which means contribution of shocks
in one of the assets to the others’ FEV is at least 1%), there is a significant spillover rt = log(Pt ) − log(Pt − 1)*100 . We proxy volatility via GARCH (1,1) since
between these assets. If the net spillover is greater than 1 (NSij = GSj ← i − GSi ← j .), then this method is found to be no inferior compared to other 330 ARCH-
we plot a uni-directional arrow, otherwise bi-directional arrow. Dashed line refers to the type models (Hansen et al., 2005). Since Narayan and Narayan (2007)
cross-commodity spillover and straight line refers to spillovers between same commodity note that leverage effect should be taken into account in commodities,
markets. CFUT and GFUT are continuous futures prices of crude oil and gold, respectively. we also run the model utilizing EGARCH (1,1). However, findings were
CSPOT is the spot US dollar prices of WTI Cushing Oklahoma per barrel and GSPOT is the
highly similar with GARCH (1,1) and hence we continue with this
spot gold bullion price per ounce.
method. Please also note that we apply the Ljung–Box and ARCH tests
to examine whether there is any remaining autocorrelation or condi-
One of the major advantages of DY framework is providing a spil- tional heteroscedasticity.
lover plot via rolling windows estimation. Through this way, one can Hong and Yogo (2012) show open interest could be informative on
understand not only the direction of spillover but also the time-varying asset prices and therefore we check the volatility spillover between
characteristics between selected variables. Hence, we estimate the trading volume, open interest and futures/spot markets. We employ
model in Eq. (1) using 200-day rolling windows and calculate the FEVD daily trading volume and open interest of the nearest continuous fu-
and spillover indices and present in Section 5. tures4 contract of both crude oil and gold, which are also available for
the same period mentioned above.
Moreover, the financialization literature argues that institutional
3. Data
investor presence creates a volatility spillover between commodities

As we previously mention, market efficiency hypothesis argues,


futures prices should not be leading spot prices, otherwise there will be 4
We take Type 0 rollover, which rolls over on the first business day of the new notional
an arbitrage opportunity. On the other hand, one of the major benefits contract month. No price adjustment is made (Datastream, 2016).

5
B.M. Ordu-Akkaya et al. Resources Policy xxx (xxxx) xxx–xxx

and also increase prices and volatilities. Hence, we also investigate the
volatility transmission mechanism between open interest positions of
trader types and commodity prices. We obtain open interest positions
for sub-groups of traders; which are Swap Dealers, Money Managers
and Producers starting from 2006. As its name suggests, DCOT report
provides weekly disaggregated data on trader positions beginning from
2006 for almost all commodities.
This report divides trading positions mainly into four as; Swap
Dealers, Processors, Managed Money and Other Reportables. Swap
dealers are often affiliated with large financial institutions and are the
counterparty of swap transactions (CFTC, 2008). Money Managers in-
clude fundamentally hedge funds, commodity trading advisors and
commodity pool operators which are included in futures trading on
behalf of their clients. Therefore they manage investments of either
Fig. 3. Volatility index for oil and gold spot and futures markets. The figure presents institutional or individual clients. Processors and Merchants, on the
the volatility spillover index between CFUT, GFUT, CSPOT and GSPOT for the period other hand, is the core commercial group participating in producing,
between 25 August 1997 and 17 October 2017. Since Diebold and Yilmaz (2012) advise it processing or handling of commodities. Swap dealers and Money
to be calculated via 200-day rolling windows. Managers are generally on the long side of the market, whereas

Table 4
Volatility spillover including trading volume and open interest.

Panel A – Spillover table

CFUT CFUTOI CFUTV CSPOT GFUT GFUTOI GFUTV GSPOT From others

CFUT 53.1 1.1 0.4 41.7 2.2 0.4 0.3 1.1 46.9
CFUTOI 1.6 95.2 0.1 1.0 0.5 0.3 1.2 0.1 4.8
CFUTV 1.2 1.0 94.1 0.7 0.0 0.1 2.7 0.3 5.9
CSPOT 42.0 1.1 0.5 53.7 1.6 0.3 0.2 0.6 46.3
GFUT 1.5 0.1 0.0 1.1 57.3 2.7 1.8 35.7 42.7
GFUTOI 0.3 0.3 0.2 0.3 5.3 87.0 2.6 4.1 13.0
GFUTV 0.0 0.1 3.9 0.0 1.8 1.3 92.2 0.6 7.8
GSPOT 0.5 0.0 0.1 0.4 38.0 2.6 1.1 57.3 42.7
Contribution to others 47.2 3.6 5.1 45.1 49.4 7.6 9.7 42.4 210.2
Contribution including own 100.3 98.9 99.2 98.8 106.6 94.6 102.0 99.7 26.3%

Panel B – Net pairwise spillovers

CFUT CFUTOI CFUTV CSPOT GFUT GFUTOI GFUTV GSPOT

CFUT
CFUTOI 0.57
CFUTV 0.84 0.83
CSPOT 0.38 0.06 −0.24
GFUT −0.62 −0.44 0.03 −0.51
GFUTOI −0.07 0.06 0.07 −0.03 2.62
GFUTV −0.22 −1.02 1.15 −0.21 0.08 −1.27
GSPOT −0.59 −0.06 −0.16 −0.26 2.37 −1.48 0.49

Panel C – Net spillovers


CFUT 0.3
CFUTOI −1.2
CFUTV −0.8
CSPOT −1.2
GFUT 6.7
GFUTOI −5.4
GFUTV 1.9
GSPOT −0.3

This table gives the spillovers including both crude oil and gold futures/ spot markets and trading volume/ open interest. The
percentage in the below right hand corner of the table is the spillover index and shows the share of spillovers in explaining forecast
error variance of all assets, on average. It is the grand off-diagonal sum divided to the grand sum including diagonals, expressed as a
percentage. Contribution to others and from others figures are off-diagonal column and row sums, respectively. Panel B depicts the net
(pairwise) spillovers between each pair. One can find the directional spillover via deducting same pairs gross spillovers from each
other. For instance the net spillover is [42.0–41.7 = 0.4] for oil futures and spot markets. Positive (Negative) figure shows spillover
from column (row) to row (column). CFUT and GFUT are continuous futures prices of crude oil and gold, respectively. CSPOT is the
spot US dollar prices of WTI Cushing Oklahoma per barrel and GSPOT is the spot gold bullion price per ounce. CFUTOI (GFUTOI) and
CFUTV (GFUTV) are open interest positions of crude oil (gold) and trading volume, respectively. Trading volume is the total quantity
of contracts traded and open interest is the number of contracts outstanding at the end of each trading day.

6
B.M. Ordu-Akkaya et al. Resources Policy xxx (xxxx) xxx–xxx

4.1. Futures – spot and oil – gold linkages

Table 3 depicts the DY results between futures and spot markets of


crude oil and gold and a graphical illustration of volatility network is
also summarized in Fig. 2. Results suggest that shocks in one market
have similar percentage contributions on the other, which implies fu-
tures and spot markets seem to be integrated and almost equally in-
formative on each other.
Findings also depict that the share of oil futures (spot) on spot
(futures) market’ FEV is 43.1% (43.2%) and hence, new market in-
formation spreads into both markets within a 10 day horizon. Gold
markets are slightly more self-oriented such that the share of gold fu-
tures (spot) on forecast error variance on spot (futures) market is 39.2%
(37.3%). However, it is of no surprise that each market's explanatory
power on itself FEV is highest, since some news may be related with
futures but not spot prices, or vice versa (Antonakakis et al., 2016).
We also investigate the spillover within a day (H = 1 days) and
check whether there is a significant discrepancy between horizons.
Although the pairwise spillovers (from futures to spot or vice versa) are
not considerably lower for the oil. This finding is somewhat in line with
the argument that spot and futures market impound information right
away, and no market is more dominant on the other (e.g., Alquist and
Kilian, 2010; Nicolau and Palomba, 2015). On the other hand, the in-
terconnectedness of gold's future and spot markets become weaker
suggesting these markets have their own dynamics.
Table 3 also presents the net volatility spillover between futures and
spot markets, which is calculated via deducting off-diagonals from each
other as in Eq. (6). For instance the net spillover is [43.1–42.6 = 0.5]
for oil markets. This indicates that spot markets receive volatility from
futures markets as it is discussed in several papers (e.g. Gil-Alana and
Tripathy, 2014). However, one should note that 0.5% is somewhat low
Fig. 4. Volatility network including trading volume and open interest. Panel A – in terms of 10 day ahead FEVD (0.2% at H = 1 days). Therefore, we
Gross spillovers. Panel B – Net spillovers. We present the volatility network between oil can conclude there is a bi-directional spillover between futures and spot
and gold markets based upon Diebold-Yilmaz output. First we compute gross spillovers markets. The net spillover for gold market, on the other hand, is rela-
between Asset i and Asset j and take an average of these spillovers. If the average spil-
tively higher [1.9], depicting that gold futures are a major transmitter
lover is greater than 1 (which means contribution of shocks in one of the assets to the
of volatility.
others’ FEV is at least 1%), there is a significant spillover between these assets. If the net
spillover is greater than 1 (NSij = GSj ← i − GSi ← j .), then we plot a uni-directional arrow,
Our results in Table 3 suggest that there is a bi-directional re-
otherwise bi-directional arrow. Dashed line refers to the cross-commodity spillover and lationship between the two commodities over futures markets; but not
straight line refers to spillovers between same commodity markets. CFUT and GFUT are over spot markets. This might imply the spillover impact of speculators,
continuous futures prices of crude oil and gold, respectively. CSPOT is the spot US dollar since they invest only in futures markets. Also one should note that
prices of WTI Cushing Oklahoma per barrel and GSPOT is the spot gold bullion price per there is an indirect relationship from crude futures to gold spot market
ounce. CFUTOI (GFUTOI) and CFUTV (GFUTV) are open interest positions of crude oil and this link could be due to speculators.
(gold) and trading volume, respectively. Trading volume is the total quantity of contracts
The spillover index which is located at the right hand corner of
traded and open interest is the number of contracts outstanding at the end of each trading
day.
Table 3 Panel A, shows the average inter-connectedness between assets.
This figure is 38.5%, which means the contribution of all these 4 assets
on average to each other's FEV is around 40%. Fig. 3 depicts the time
Producers are on the short side. However, all these traders can either varying behavior of the index which is around 30–45% range during the
take short and long positions depending upon the market environment. period between 1998 and 2008. On the other hand, after the burst of
Therefore, following Bhardwaj et al. (2015), we compute open interest the crisis, the index suddenly jumps to 63% by September 2008. And up
positions as follows: [long + short +2*spreading] for Money Managers until 2015, spillover index has been highly volatile and displayed ca-
and Swap Dealers and via (long +short) for producers. Spread positions pricious behavior. Even though the index is somewhat stable in the last
refer to simultaneous buying and selling and are valid for speculators. two years, as a future direction of research, researchers could look for
Descriptive statistics for all variables included in the analysis as well underlying forces behind the spillover index.
as the correlation matrix are provided in Tables 1, 2.
4.2. Informative power of trading volume and open interest
4. Results
Given the previous findings on the interconnectedness between
Utilizing DY framework, we calculate total, gross (directional) and net volatility, trading volume and open interest, we extend the previous
(pairwise) volatility spillovers. Although results are provided based on the analysis by including daily trading volume and open interest of both
lag = 4 model, they are very similar for the lag lengths between 2 and 8. futures markets. Table 4 presents the DY output and Fig. 4 shows the
DY framework depends on the selection of horizon days (H ) and selection volatility network in a summarized illustration.
of H is highly arbitrary but authors propose to select an H dependent upon Fig. 4 Panel A and B presents gross and net volatility spillovers,
the aim of the study. We select H = 10 for our daily analysis (Diebold and respectively. First we compute gross spillovers between Asset i and
Yilmaz, 2012) and H = 2 for weekly analysis. Authors indicate con- Asset j and take the average. If the average spillover is greater than 1
nectedness can increase with higher horizon days, and hence results are (which means contribution of shocks in one of the assets to the others’
not less but more significant under longer horizon days. FEV is at least 1%), there is a significant spillover between these assets.

7
B.M. Ordu-Akkaya et al. Resources Policy xxx (xxxx) xxx–xxx

Fig. 5. Time-varying net spillovers. This figure gives time-varying net spillover of each variable against other variables. We calculate time-varying spillover via 200-day rolling
windows. CFUT and GFUT are continuous futures prices of crude oil and gold, respectively. CSPOT is the spot US dollar prices of WTI Cushing Oklahoma per barrel and GSPOT is the spot
gold bullion price per ounce. CFUTOI (GFUTOI) and CFUTV (GFUTV) are open interest positions of crude oil (gold) and trading volume, respectively. Trading volume is the total quantity
of contracts traded and open interest is the number of contracts outstanding at the end of each trading day.

8
B.M. Ordu-Akkaya et al. Resources Policy xxx (xxxx) xxx–xxx

throughout our sample period. We plot time-varying characteristics of


the plot in Fig. 6, which suggests spillover sways between 30 and 50
range and attains its historical high in 2008. Hence, based on this
episode, open interest and trading volume holds significant information
about both commodity markets.

4.3. Trader types and futures markets

Financialization theory argues that institutional investors increase


volatilities and create spillovers between commodity markets (Basak
and Pavlova, 2016). However, empirical papers do not reach a con-
sensus. We next investigate the spillover between trader types and
commodity markets utilizing the data from DCOT reports.
Results show that spillover index is 28.7%, which is considerably
Fig. 6. Volatility index including trading volume and open interest. The figure high and provides that all traders’ positions and commodity markets are
presents the volatility spillover index between CFUT, CFUTOI, CFUTV, CSPOT, GFUT, closely interconnected. However, we also present time-varying spil-
GFUTOI, GFUTV, and GSPOT for the period between 25 August 1997 and 17 October lovers to examine how market conditions affect the spillover index. As
2017. Since Diebold and Yilmaz (2012) advise it to be calculated via 200-day rolling Fig. 7 presents, the spillover index displays quite a volatile figure
windows.
throughout the sample period. Especially up until 2013, we see very
high figures, which might be attributable to financialization. The spil-
lover index gradually declines, but after 2016 we see a quite significant
incline, again.
Table 5 and Fig. 8 suggest that crude oil and gold markets are clo-
sely related and their traders spill volatility to the other market, sig-
nificantly. Particularly, Fig. 8 shows that money managers and swap
dealers are bi-directionally connected. We also show that speculators
receive volatility from both spot and futures markets. Therefore, our
results are in line with Brunetti et al. (2016) who argue that speculators
do not lead to increasing volatilities in futures market; but the opposite
is true..
Table 5 also shows that oil money managers explain 14.8% of
producers’ FEV; whereas gold money managers explain 3.3% of gold
Fig. 7. Time-varying spillover index. The figure presents the volatility spillover index producers and 4.0% of gold swap dealers’ FEV. However, the net spil-
between CFUT, CMM, CPU, CSD, CSPOT, GFUT, GMM, GPU, GSD, GSPOT for the period lover in Table 5 Panel C shows that gold money managers are receiving
between 18 May 2010 and 17 October 2017. Since Diebold and Yilmaz (2012) advise it to volatility more than they transmit. This is further supported by time-
be calculated via 200-week rolling windows, the period between 2008 and 2010 are not varying spillover figures. As Fig. 9 presents, gold money manager po-
presented. sitions are consistently volatility receivers throughout the sample
period. On the other hand, crude oil money managers do not exhibit a
If the net spillover is greater than 1 (NSij = GSj ← i − GSi ← j .), then it is consistent receiver/ transmitter role. This finding is in line with Adams
depicted by a uni-directional arrow, otherwise a bi-directional arrow is and Glück (2015) arguing financialization is dependent upon the per-
drawn. formance of the individual commodity compared to the S&P GSCI ag-
We observe that open interest and spot markets are closely related gregated index. They show crude oil is a very well performer against the
and direction is bi-directional for oil and uni-directional for the gold index and thus it is financialized and there is a significant volatility
market. Furthermore, trading volume provides cross-market informa- spillover from index trader positions to futures prices. On the other
tion, which might be implying investors investing in both crude oil and hand, gold does not present likewise results, it is a poor performer and
gold markets spill information to one another. Furthermore, gold open there is no significant spillover from index traders to gold futures.
interest receives volatility from futures market as well as the trading The disparity between oil money manager volatility roles could also
volume. Therefore, our results partially support findings of Antonakakis be attributable to the changing motives of money managers depending
et al. (2016), but we should note that trading volume and open interest upon the prices of oil. Since they are perceived to be speculators in the
do not spill volatility to futures or spot markets as it is argued in Hong market, their sole motive is to chase prices. To have a clear view of this,
and Yogo (2012). we plot crude oil and gold spot prices against money manager positions
Although net connectedness figures of open interest and trading of both commodities in Fig. 10. This figure shows a visibly significant
volume shows both are net volatility receivers, time varying spillover connection between the price and volatility transmission of money
figure (Fig. 5) indicates that both variables play significant volatility managers. When the price of the commodity increases (decreases),
transmitting roles in selected time episodes. Therefore, a visual analysis volatility receiver (transmitter) role of money managers becomes
could hint that open interest and trading volume are clearly important stronger. What drives this trend? Speculators generally are on the long
indicators during specific times; and even more important than futures side of the market and undertake the commodity price risk. However,
prices. they enter into contract on the proviso expected spot price is higher
The total spillover index shows average contribution of un- than settled futures price. Therefore, when the market is on back-
anticipated changes to volatilities of dependent variables, and is 26.3% wardation, they have a tendency to enter into that particular market
and as they start to enter their positions may be displaying a volatility

9
B.M. Ordu-Akkaya et al. Resources Policy xxx (xxxx) xxx–xxx

Table 5
Volatility spillover based on trader types.

Panel A – Gross spillover

CFUT CMM CPU CSD CSPOT GFUT GMM GPU GSD GSPOT From others

CFUT 49.5 0.0 0.5 0.4 44.2 2.2 0.0 0.1 0.5 2.7 50.5
CMM 0.6 83.8 12.2 2.6 0.4 0.1 0.1 0.1 0.1 0.0 16.2
CPU 0.3 14.8 83.6 0.9 0.2 0.1 0.0 0.1 0.1 0.0 16.4
CSD 1.7 1.9 0.9 91.8 1.3 0.1 0.0 1.5 0.1 0.6 8.2
CSPOT 43.6 0.1 0.5 0.3 50.2 2.2 0.0 0.1 0.5 2.7 49.8
GFUT 2.5 0.1 0.0 0.0 2.2 47.2 2.7 0.5 1.2 43.8 52.8
GMM 0.2 0.1 0.0 0.0 0.1 5.5 81.1 3.7 4.0 5.2 18.9
GPU 0.5 0.2 0.1 1.7 0.6 1.1 3.3 90.0 1.1 1.6 10.0
GSD 0.9 0.1 0.1 0.1 0.7 2.4 4.0 1.4 88.1 2.3 11.9
GSPOT 2.7 0.1 0.0 0.2 2.3 42.9 2.3 0.6 1.1 47.7 52.3
Contribution to others 52.9 17.3 14.3 6.1 52.0 56.6 12.4 8.0 8.5 58.9 286.9
Contribution including own 102.4 101.1 97.9 98.0 102.2 103.7 93.5 98.0 96.6 106.5 28.70%

Panel B – Net pairwise spillover

CFUT CMM CPU CSD CSPOT GFUT GMM GPU GSD GSPOT

CFUT
CMM 0.6
CPU −0.2 2.5
CSD 1.4 −0.6 0.0
CSPOT −0.6 −0.3 0.3 −1.0
GFUT 0.3 0.0 −0.1 −0.1 0.0
GMM 0.2 0.0 0.0 0.0 0.1 2.9
GPU 0.4 0.1 −0.1 0.2 0.5 0.6 −0.5
GSD 0.4 0.0 0.1 0.0 0.3 1.2 0.0 0.3
GSPOT 0.0 0.1 0.0 −0.4 −0.4 −0.9 −2.9 −1.0 −1.1

Panel C – Net spillovers


CFUT 2.4
CMM 1.1
CPU −2.1
CSD −2.1
CSPOT 2.2
GFUT 3.8
GMM −6.5
GPU −2.0
GSD 6.2
GSPOT 6.6

This table presents the volatility spillover between crude oil and gold futures and spot markets and their trader types. The
percentage in the below right hand corner of the table is the spillover index and shows the share of spillovers in explaining
forecast error variance of all assets, on average. It is the grand off-diagonal sum divided to the grand sum including diagonals,
expressed as a percentage. Contribution to others and from others figures are off-diagonal column and row sums, respectively.
Panel B depicts the net (pairwise) spillovers between each pair. CFUT and GFUT are continuous futures prices of crude oil and
gold, respectively. CSPOT is the spot US dollar prices of WTI Cushing Oklahoma per barrel and GSPOT is the spot gold bullion
price per ounce. CMM (GMM), CSD (GSD) and CPU (GPU) are weekly open interest positions of Money Managers, Swap Dealers
and Producers, respectively.

receiver role from futures market prices and other traders’ positions. 2016). Moreover, the increasing presence of institutional investors in
However, when the market is on contango, they might prefer to exit commodity markets changed the commodity market fundamentals such
from the market since they are not rewarded for the risk they are as prices, volatility significantly. Therefore, the link between spot and
bearing. After they commence to exit, this behavior could be leading futures markets could structurally change following the financializa-
the volatility in futures market prices and in positions of other traders. tion. To examine the role of trading volume, open interest and trader
positions on volatility spillover, we mainly employ Diebold and Yilmaz
(2012) method. This technique is critically advantageous over other
5. Conclusions spillover tests since it helps one to understand the direction and time-
varying characteristics of spillover.
The goal of our paper is to investigate the role of trading volume, Our results can be summarized as follows. First of all, our findings
open interest and trader positions on volatility spillover between fu- show that futures and spot markets have almost equal spillover impact
tures and spot markets of two commodities; crude oil and gold. on the other, and hence no market is more dominant on the other. This
Although the linkage between futures and spot markets is a rich lit- result is in line with the assertion that spot and futures market impound
erature, including liquidity variables such as trading volume and open information immediately, and no market is more dominant on the other
interest as well as traders’ positions are rather new (Antonakakis et al.,

10
B.M. Ordu-Akkaya et al. Resources Policy xxx (xxxx) xxx–xxx

Fig. 8. Volatility network including trader types. Panel A – Gross spillovers. Panel B – Net spillovers. We present the volatility network between oil and gold markets based upon
Diebold-Yilmaz output. First we compute gross spillovers between Asset i and Asset j and take an average of these spillovers. If the average spillover is greater than 1 (which means
contribution of shocks in one of the assets to the others’ FEV is at least 1%), there is a significant spillover between these assets. If the net spillover is greater than 1
(NSij = GSj ← i − GSi ← j .), then we plot a uni-directional arrow, otherwise bi-directional arrow. Dashed line refers to the cross-commodity spillover and straight line refers to spillovers
between same commodity markets. CFUT and GFUT are continuous futures prices of crude oil and gold, respectively. CSPOT is the spot US dollar prices of WTI Cushing Oklahoma per
barrel and GSPOT is the spot gold bullion price per ounce. CMM (GMM), CSD (GSD) and CPU (GPU) are weekly open interest positions of Money Managers, Swap Dealers and Producers,
respectively.

(e.g., Alquist and Kilian, 2010; Nicolau and Palomba, 2015). Even positions have always been volatility receivers during the sample
though, there are various studies showing gold and oil markets are period, crude oil money managers do not present such behavior. This
closely connected, results are dependent upon the sample period, result resonates well with the related literature suggesting that fi-
methodology employed etc. Our analyses indicate that while the re- nancialization of the commodity is dependent on the performance of
lationship between futures markets is bi-directional in oil and gold that specific commodity with respect to the S&P GSCI (e.g., Adams and
markets, for spot market it is not, possibly due to the spillover effect of Glück, 2015).
speculators who invest only in futures market. Such bi-directional re- These findings suggest that policy-makers do not need to put addi-
lationship is further proven by Ewing and Malik (2013). tional control mechanisms over speculators since they do not increase
Extending the analysis via daily trading volume and open interest of volatilities in futures market. Moreover, trading volume and open in-
both futures markets, we find that there is a close relationship between terest disclose information about the market, but they do not lead to an
open interest and spot markets. Furthermore, trading volume provides excess volatility. As a matter of fact, sharing such information would
cross-market information, which might be implying investors investing increase the transparency of the market. As a future direction of re-
in crude oil and gold markets spill information to one another. This search, researchers could examine cointegrating relationship between
finding partially resonates with Antonakakis et al. (2016); however we abovementioned variables and for other selected group of commodities.
should note that trading volume and open interest do not spill volatility Since previous studies show financialization is dependent upon the
to futures or spot markets as it is argued in Hong and Yogo (2012). commodity (Gozgor et al., 2016; Ordu et al., 2017), results could sig-
Lastly, we conduct tests to examine the role of hedge funds, swap nificantly change for other commodities. Moreover, we do not parti-
dealers and producers’ positions on future-spot markets nexus. We find cularly examine underlying behavior for time-varying spillover indices
evidence that positions of all traders and commodity markets are clo- in our paper. One could look whether particular economic conditions,
sely interconnected as financialization proponents argue (Basak and business cycle or changes in trader behavior affects spillover char-
Pavlova, 2016). Money manager positions of oil and gold are closely acteristics.
linked, and their traders significantly spill volatility to the other market.
Time-varying spillover findings display that while gold money manager

11
B.M. Ordu-Akkaya et al. Resources Policy xxx (xxxx) xxx–xxx

Fig. 9. Time varying net spillovers This figure gives time-varying net spillover of each variable against other variables. We calculate time-varying spillover via 200 rolling windows.
CFUT and GFUT are continuous futures prices of crude oil and gold, respectively. CSPOT is the spot US dollar prices of WTI Cushing Oklahoma per barrel and GSPOT is the spot gold
bullion price per ounce. CMM (GMM), CSD (GSD) and CPU (GPU) are weekly open interest positions of Money Managers, Swap Dealers and Producers, respectively.

12
B.M. Ordu-Akkaya et al. Resources Policy xxx (xxxx) xxx–xxx

J. Financ. Quant. Anal. 51 (5), 1545–1574.


Cheng, I.H., Xiong, W., 2014. Financialization of commodity markets. Annu. Rev. Financ.
Econ. 6, 419–441.
Datastream, T.R. (2016). Futures continuous series: Methodology and definitions.
Retrieved from http://extranet.datastream.com/data/Futures/Documents/
Datastream%20Product%20Futures%20Continuous%20Series.pdf [accessed 4 March
2016.
Diebold, F.X., Yilmaz, K., 2012. Better to give than to receive: predictive directional
measurement of volatility spillovers. Int. J. Forecast. 28 (1), 57–66.
Erb, C.B., Harvey, C.R., 2013. The golden dilemma. Financ. Anal. J. 69 (4), 10–42.
Ewing, B.T., Malik, F., 2013. Volatility transmission between gold and oil futures under
structural breaks. Int. Rev. Econ. Financ. 25, 113–121.
Fattouh, B., Mahadeva, L., Kilian, L., 2013. The role of speculation in oil markets: what
have we learned so far? Energy J. 34, 7–33.
Gil-Alana, L.A., Tripathy, T., 2014. Modelling volatility persistence and asymmetry: a
study on selected indian non-ferrous metals markets. Resour. Policy 41, 31–39.
Gormus, N.A., Soytas, U., Dlitz, J.D., 2014. Volatility trasnmission between energy-re-
lated asset classes. Glob. Financ. J. 25 (3), 246–259.
Gozgor, G., Lau, C.K.M., Bilgin, M.H., 2016. Commodity markets volatility transmission:
roles of risk perceptions and uncertainty in financial markets. J. Int. Financ. Mark.
Inst. Money 44, 35–45.
Hansen, P.R., Lunde, A., 2005. A forecast comparison of volatility models: does anything
beat a GARCH (1,1)? J. Appl. Econ. 20 (7), 873–889.
Hicks, J.R., 1939. Value and Capital: An Inquiry into Some Fundamental Principles of
Economic Theory. Clarendon Press, Oxford.
Hong, H., Yogo, M., 2012. What does futures market interest tell us about the macro-
economy and asset prices? J. Financ. Econ. 105, 473–490.
Jacks, D.S., 2007. Populists versus Theorists: futures markets and the volatility of prices.
Explor. Econ. Hist. 44, 342–362.
Kaldor, N., 1939. Speculation and economic stability. Rev. Econ. Stud. 7 (1), 1–27.
Keynes, J.M., 1923. A Track on Monetary Reform, Publisher: London MacMillan and Co.
Kocaarslan, B., Sari, R., Soytas, U., 2017. Are there any diversification benefits among
global finance center candidates in Eurasia? Emerg. Mark. Financ. Trade 52,
357–374.
Koop, G., Pesaran, M.H., Potter, S.M., 1996. Impulse response analysis in nonlinear
multivariate models. J. Econ. 74, 119–147.
Fig. 10. Money manager positions with respect to crude oil/gold prices. This figure Masters,M.W., 2008. Testimony before the Committee on Homeland Security and
presents money manager open interest positions with respect to spot US dollar prices of Governmental Affairs. United States Senate. May 20. Retrieved from 〈http://www.
WTI Cushing Oklahoma per barrel and spot gold bullion price per ounce. LHS refers to the hsgac.senate.gov//imo/media/doc/052008Masters.pdf?Attempt=2〉 (Accessed 22
left hand side. February 2017).
Narayan, P.K., Narayan, S., 2007. Modelling oil price volatility. Energy Policy 35 (12),
6549–6553.
References Narayan, P.K., Narayan, S., Zheng, X., 2010. Gold and oil futures markets: are markets
efficient? Appl. Energy 87, 3299–3303.
Nicolau, M., Palomba, G., 2015. Dynamic relationships between spot and futures prices.
Adams, Z., Glück, T., 2015. Financialization in commodity markets: a passing trend or the the case of energy and gold commodities. Resour. Policy 45, 130–143.
new normal? J. Bank. Financ. no.60, 93–111. Ordu, B.M., Oran, A., Soytas, U., 2017. Is Food Financialized? Yes, but Only When
Alquist, R., Kilian, L., 2010. What do we learn from the price of crude oil futures? J. Appl. Liquidity Is Abundant, Journal of Banking and Finance, .10.1016/j.jbankfin.2017.06.
Econ. 25 (4), 539–573. 001.
Antonakakis, N., Floros, C., Kizys, R., 2016. Dynamic spillover effects in futures markets: Pesaran, H.H., Shin, Y., 1998. Generalized impulse response analysis in linear multi-
UK and US evidence. Int. Rev. Financ. Anal. 48, 406–418. variate models. Econ. Lett. 58 (1), 17–29.
Baffes, J., 2007. Oil spills on other commodities. Resour. Policy 32 (3), 126–134. Sims, C.A., 1980. Macroeconomics and reality. Econometrica 48, 1–48.
Basak, S., Pavlova, A., 2016. A model of financialization of commodities. J. Financ. 71 Stoll, H.R., Whaley, R.E., 2010. Commodity index investing and commodity futures
(4), 1511–1556. prices. J. Appl. Financ. 20, 7–46.
Bhardwaj, G., Gorton, G., Rouwenhorst, G., 2015. Facts and fantasies about commodity Working, H., 1949. The theory of the price of storage. Am. Econ. Rev. 39 (6), 1254–1262.
futures ten years later (No. w21243), National Bureau of Economic Research. Zhang, Y.J., Wei, Y.M., 2010. Crude oil market and the gold market: evidence for coin-
Brennan, M.J., 1958. Economics and the theory of social systems. Am. J. Econ. Sociol. 17 tegration, causality and price discovery. Resour. Policy 35, 168–177.
(2), 113–122.
Brunetti, C., Buyuksahin, B., Harris, J.H., 2016. Speculators, prices, and market volatility.

13

You might also like