Speculative Activity and Returns Volatility of Chinese Major
Agricultural Commodity Futures*
Martin T. Bohl1, Pierre L. Siklos 2 and Claudia Wellenreuther 3
November 16, 2016
* An earlier version of the paper was presented at the 2016 Econometric Research in Finance
(ERFIN) Conference in Warsaw. The authors are grateful for the comments by the participants.
1 Corresponding author: Martin T. Bohl, Department of Economics, Westfälische Wilhelms-University Münster, Am Stadtgraben 9, 48143 Münster, Germany, phone: +49 251 83 25005, fax: +49 251 83 22846, e-mail: [email protected]
2 Pierre L. Siklos, Lazaridis School of Business & Economics, Wilfrid Laurier University, 75 University Avenue West, Waterloo, ON, N2L 3C5, Canada, e-mail: [email protected]
3 Claudia Wellenreuther, Department of Economics, Westfälische Wilhelms-University Münster, Am Stadtgraben 9, 48143 Münster, Germany, phone: +49 251 83 25003, fax: +49 251 83 22846, e-mail: [email protected]
Speculative Activity and Returns Volatility of Chinese Major
Agricultural Commodity Futures
Abstract
We empirically investigate whether speculative activity in Chinese futures markets for
agricultural commodities destabilizes futures returns. To capture speculative activity a
speculation and a hedging ratio is used. Applying GARCH models we first analyse the
influence of both ratios on the conditional volatility of three heavily traded Chinese futures
contracts, namely soybeans, soybean meal, and sugar. Additionally, VAR models in
conjunction with Granger causality tests, impulse-response analyses and variance
decompositions are used to get insight into the lead-lag relationship between speculative
activity and returns volatility.
Keywords: Speculation Ratio, Returns Volatility, Chinese Futures Markets, Agricultural
Commodities
1
1. Introduction
Since the mid-2000s, commodity markets have witnessed turbulent times, including both
dramatic price peaks in 2007-2008, and again in 2010-2011, and a surge in returns volatility.
Furthermore, a sharp rise in the popularity of commodity investing has triggered a large inflow of
investment capital into commodity futures markets. This phenomenon, known as the
“financialization” of commodity markets, has encouraged a heated public and an extensive
academic debate (e.g., see Cheng and Xiong 2014).
In particular, commodity index traders, who represent a new player in commodity futures
markets, have become the centre of public attention. Hedge fund manager Michael W. Masters is a
leading supporter of the claim that the spikes in commodity futures prices in 2007-2008 were
mainly driven by long-only index investment. Masters argues that the index investment created
massive buying pressure, which thus led to a bubble in commodity prices with prices far away from
their fundamental values (Master 2008, Master and White 2008). Nevertheless, the academic
literature that has generated thorough empirical analyses has failed to find compelling evidence for
the Masters hypothesis (Irwin et al. 2009, Stoll and Whaley 2009, Gilbert and Morgan 2010).
Discussing several empirical findings on the influence of index traders, Irwin and Sanders (2012)
conclude that index trading is unrelated to the recent price peaks.
While the academic debate about the effects of long-only index investment seems to be settled,
the role of traditional speculators on commodity futures markets, the so called long-short investors,
still remains an empirical issue. Our research builds upon this debate and aims to investigate
whether long-short speculators contribute to the observed price changes. Studies by Till (2009) and
Sanders et al. (2010) come to the conclusion that long-short speculators are not to blame for the
excessive price impact in 2007-2008 because the rise in speculation was only a response to a rise
in hedging demand. Brunetti et al. (2011) use Granger causality tests to analyse the relationship
between changes in the net positions of hedge funds in three commodities, namely corn, crude oil
and natural gas, and volatility. The authors find that such funds actually stabilize prices by
decreasing volatility. Miffre and Brooks (2013) also investigate the influence of long-short
2
speculators and conclude that speculators have no significant impact on volatility or cross-market
correlation.
Only a few studies investigate the influence of futures speculation on spot returns volatility.
Bohl et al. (2012) analyse how expected and unexpected speculative volume and open interest of
six heavily traded futures contracts impact on conditional spot returns volatility. After applying
their tests to two sub periods, which differ by the size of the market shares of speculators, they
conclude that the financialization of commodity futures markets does not increase volatility of spot
returns. Furthermore, Kim (2015) shows that especially during the recent period, when
commodities have become financial assets attracting diverse types of speculators, speculation in
futures markets can even contribute to reducing spot returns volatility.
The literature review indicates that most of the studies show either no effect or even a
stabilizing effect of speculation on returns volatility. However, the studies above only investigate
futures markets in the US and are all based on Commitments of Traders (COT) reports, provided
by the US Commodity Futures Trading Commission (CFTC). The original COT report, which
separates solely traders into commercial (hedgers) and non-commercial traders (speculators), has
been put into question many times from diverse perspectives (Peck 1982, Ederington and Lee
2002). To deal with these concerns, the CFTC published two variations to the COT reports, the
Disaggregated Commitments of Traders (DCOT) report, which further disaggregates the
commercial and non-commercial trader categories and the Supplemental Commitments of Traders
(SCOT).1 Nevertheless, CFTC data are publicly available only at a weekly level and therefore
unsuited for analyses, which aim to examine the short-run dynamics of commodity prices. To
investigate effects of speculative activity on return volatility, empirical analyses should be based
on data at the daily frequency. Furthermore, the CFTC publishes only data for specific futures
contracts traded on markets in the US. Hence, to investigate other markets apart from the US,
different methods to separate hedging from speculative activity must be applied.
1 For more details about the CFTC database see Stoll and Whaley (2009) as well as Irwin and Sanders (2012).
3
Therefore, we use two ratios, namely ones proposed by Garcia et al. (1986) and Lucia and
Pardo (2010), that combine trading volume and open interest data to measure the relative
dominance of speculative activity and hedging activity on a market. The extant literature on
commodity futures markets has generally accepted that volume contains information about
speculative activity while open interest reflects hedging activity (Rutledge 1979, Leuthold 1983,
Bessembinder and Seguin 1993).
Anecdotal evidence suggests that trading behaviour in Chinese financial markets is highly
speculative. For example, the Wall Street Journal even compares China’s stock markets to casinos,
which are driven by fast money flows (Wall Street Journal, August 22, 2001). Due to strengthening
stock market regulation, drawn by the collapse in Chinese stock markets in 2015, futures markets
for commodities have also become very attractive to speculators lately. Recently, the Financial
Times states: “In the past month near mania has gripped China’s commodity futures markets with
day traders and yield-hungry wealth managers pouring into a lightly regulated sector, often with
astonishing results.” (Financial Times, April 27, 2016). In a similar vein, a report published by
Citigroup Research describes Chinese investors as perhaps prone to being the most speculative in
the world. Furthermore, the report points out that speculative trading volume on Chinese
commodity futures markets has exploded in the last years and has created high returns volatility
(Liao et al. 2016). Using market activity data enables us to examine the speculative content in
China’s futures markets.
Another argument in favour of examining Chinese commodity markets, is that Chinese futures
markets for commodities have grown rapidly in recent years. A loosening of regulations also
permits foreign investors to participate in Chinese futures markets and trading volume has increased
substantially. According to trading volume, commodity futures markets in China already belong to
the most active ones in the world. For instance, Dalian Commodity Exchange (DCE) soybeans
futures market is the second largest soybeans futures market in the world, right behind the one of
the Chicago Board of Trade (CBOT). Therefore Chinese futures markets gain more and more global
importance and Chinese prices have begun affecting global prices for commodities (Wang et al.
4
2016, Wang and Ke 2005). Additional to the increased trading volume, Chinese futures markets
have also seen turbulent events in the last decade, including price spikes and high returns volatility.
Keeping in mind all these facts it is quite surprising that studies on futures markets in China
are rare. Due to its global importance and the mentioned characteristics, it is important to investigate
speculation in Chinese futures markets. To measure speculative activity we use the speculation ratio
as well as the hedging ratio. The empirical analysis includes GARCH models and Granger causality
tests to examine both contemporaneous and lead-lag relationships between speculation activity and
returns volatility in three heavily traded agricultural commodities, namely soybeans, soybean meal
and sugar.
In contrast to the available literature we find a positive influence of the speculation ratio on
returns volatility and a negative influence of the hedging ratio on conditional volatility for all
commodities examined. These empirical results indicate that a rise in speculative activity leads to
an increase in returns volatility and a rise in hedging activity stabilizes the returns. Moreover, we
show that for soybeans, soybean meal and cotton, the speculation ratio and the hedging ratio
Granger causes conditional returns volatility and vice versa. This relationship is positive in the case
of the speculation ratio and negative in the case of the hedging ratio. These results imply that the
amount of speculative activity in relation to hedging activity contains information about changes in
futures returns volatility.
The remainder of this paper is structured as follows: In Section 2 we introduce the speculation
measures and their computation. After presenting the econometric methods in section 3 and the
data in section 4, we discuss the empirical results in section 5. Section 6 summarizes our findings
and concludes.
2. Measures Construction
To analyse the character of trading activity on a specific trading day, we compute two ratios,
both of which combine daily figures of volume and open interest. Daily trading volume captures
5
all trades of a particular contract, which are executed during a specified day. Open interest describes
all positions of that contract that are still open at the end of that trading day, meaning that the
position has neither been equalized by an opposite futures position nor been fulfilled by the physical
delivery of the commodity or by cash settlement. The first ratio used in this study captures
speculative activity and is defined as daily trading volume (𝑇𝑇𝑇𝑇𝑡𝑡) divided by end-of-day open
interest (𝑂𝑂𝑂𝑂𝑡𝑡):
𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑡𝑡𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 =
𝑇𝑇𝑇𝑇𝑡𝑡𝑂𝑂𝑂𝑂𝑡𝑡
(1)
The speculation ratio measures the relative dominance of speculative activity in the contract
analysed in comparison to the hedging activity. A high (low) speculation ratio indicates high (low)
speculative activity with respect to hedging activity. Therefore, a rise in the speculation ratio
reflects a rise in the dominance of speculators in the market.
The idea behind the speculation ratio lies in the assumption that hedgers hold their positions
for longer periods due to their underlying positions, whereas speculators mainly try to avoid holding
their positions over night. Based on their different trading behaviours, speculators and hedgers
influence the amount of trading volume and open interest in a different way. Speculators mostly
impact on trading volume instead of open interest because they buy and sell contracts during the
day and close their positions before trading ends. Thus outstanding contracts at the end of a trading
day are mainly hold by hedgers (Rutledge 1979, Leuthold 1983, Bessembinder and Seguin 1993).
Obviously, the ability of the ratio to measure the dominance of speculative activity depends on the
assumption that hedgers and speculators sit on their trading position for different time periods.
There is empirical evidence that seems to approve the assumption that hedgers tend to hold their
position for longer periods than speculators (Ederington and Lee 2002, Wiley and Daigler 1998).
We use a second ratio to provide supportive results of the first one. The second ratio is also
based on the different trading behaviour of speculators and hedgers, but relates daily trading volume
to open interest in a different way. The ratio gauges the relative importance of hedging activity
6
instead of speculative activity on a specific trading day and is defined as the daily change in open
interest ( ΔOIt = OIt – OIt−1) divided by daily trading volume:
Ratio𝑡𝑡𝐻𝐻𝑆𝑆𝐻𝐻𝐻𝐻𝑆𝑆 =
∆OIt𝑇𝑇𝑇𝑇𝑅𝑅
(2)
The change in open interest during period t is a measure of net positions being opened or closed
each day and held overnight and is used to capture the hedging activity. The hedging ratio can take
any value in the range of [1 and -1], where a high hedging ratio, with a value close to one, indicates
low speculative activity in the contract examined. The correlation between the two ratios used in
this study should be negative.
Based on the speculation ratio (1) we are able to investigate the role of short term speculators
on commodity futures markets. In a few studies, short term speculation in US futures markets is
explored by using the speculation ratio. For agricultural commodities Streeter and Tomek (1992)
find a positive influence of the speculation ratio on returns volatility for soybeans. Robles et al.
(2009) investigate speculative activity in four agricultural future markets and find a Granger
causality relationship between the speculation ratio and prices for wheat and rice. More recently
Chan et al. (2015) examine the role of speculators on oil futures markets by using the speculation
ratio to proxy speculative activity and conclude that the oil futures market is dominated by
uniformed speculators in the post-financialization period. 2 Only Lucia et al. (2015) apply both the
speculation (1) and hedging ratios (2) to explore the relative importance of speculative activity
versus hedging activity in the European carbon futures market. The authors show the different
dynamics of speculative behaviour during three phases of the European Union Emission Trading
Scheme.
2 The speculation ratio has not only be used to investigate commodity markets. Chatrath et al. (1996), for instance, apply the speculation ratio to examine the influence of speculation on the volatility of exchange rates.
7
3. Econometric Methodology
To analyse the impact of speculative activity, proxied by the speculation and the hedging ratio,
on returns volatility, a generalized autoregressive conditional heteroscedasticity (GARCH) model
(Bollerslev 1986), is used. Our AR(1)-GARCH(1,1) model reads as follows:
𝑟𝑟𝑡𝑡 = 𝑅𝑅 + 𝑏𝑏1𝑟𝑟𝑡𝑡−1 + 𝜀𝜀𝑡𝑡 (3)
𝜎𝜎𝑡𝑡2 = 𝛼𝛼0 + 𝛼𝛼1𝜀𝜀𝑡𝑡−12 + 𝛽𝛽1𝜎𝜎𝑡𝑡−12 + 𝛾𝛾𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑡𝑡 (4)
where 𝑟𝑟𝑡𝑡 = (𝑙𝑙𝑙𝑙(𝑃𝑃𝑅𝑅) − 𝑙𝑙𝑙𝑙(𝑃𝑃𝑅𝑅−1)) ∗ 100 is the return on day t, 𝜎𝜎𝑡𝑡2 is the conditional volatility on day t
and 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑡𝑡𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆,𝐻𝐻𝑆𝑆𝐻𝐻𝐻𝐻𝑆𝑆 describes the speculation ratio on day t in the first run and hedging ratio on
day t in the second run.3 The mean equation (3) models the returns as a first-order autoregressive
(AR) process. The relationship between conditional volatility and speculative activity has been
modelled by modifying the volatility equation (4). The parameter 𝛼𝛼1 captures the ARCH effect,
which measures the reaction of conditional volatility to new information (shocks), whereas 𝛽𝛽1
describes the GARCH effect, which displays the duration of a shock to die out.
The influence of speculative activity is captured by the parameter 𝛾𝛾 . In the case of the
speculation ratio a positive sign of the parameter 𝛾𝛾 implies a destabilizing effect of speculative
activity on returns volatility, whereas a negative sign would imply a stabilizing effect. Regarding
the hedging ratio a negative influence indicates that speculation drives volatility, while a positive
influence points out that speculation stabilizes the market. Furthermore, the GARCH (1,1) model
has a number of restrictions to ensure a positive conditional variance, i.e., 𝛼𝛼0 > 0, 𝛼𝛼1 ≥ 0, 𝛽𝛽1 ≥ 0
and 𝛼𝛼1 + 𝛽𝛽1 ≤ 1.
The previously introduced GARCH model only measures the possible influence of speculative
activity on conditional volatility and not vice versa. Since not only speculation can drive returns
volatility, but high volatility also can attract speculators attention and thus lead to speculative
activity, we are also interested in the lead-lag relationship between the two variables. To investigate
3 We apply a GARCH model of order p = 1 and q = 1, since a number of researchers have frequently demonstrated the suitability of GARCH (1,1) models to represent the majority of financial time series (Bera and Higgins 1993).
8
the dynamic relationship (lead-lag) of returns volatility and the speculative activity, we use a vector
autoregressive (VAR) model that is expressed by the two following equations:
𝜎𝜎𝑡𝑡2 = 𝑅𝑅0 + �𝛼𝛼1,𝑡𝑡𝜎𝜎𝑡𝑡−𝑖𝑖2𝑘𝑘
𝑖𝑖=1
+ �𝛽𝛽1,𝑡𝑡𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑡𝑡−𝑖𝑖
𝑘𝑘
𝑖𝑖=1
+ 𝜖𝜖𝑡𝑡 (5)
𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑡𝑡 = 𝑅𝑅0 +�𝛼𝛼1,𝑡𝑡𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑡𝑡−𝑖𝑖
𝑘𝑘
𝑖𝑖=1
+ �𝛽𝛽1,𝑡𝑡𝜎𝜎𝑡𝑡−𝑖𝑖2𝑘𝑘
𝑖𝑖=1
+ 𝑢𝑢𝑡𝑡 (6)
The VAR equations show that the endogenous variables as conditional volatility (𝜎𝜎𝑡𝑡2), that is
estimated from a simple GARCH(1,1) model, and speculation ratio (𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑡𝑡𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆) in the first run and
hedging ratio (𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑡𝑡𝐻𝐻𝑆𝑆𝐻𝐻𝐻𝐻𝑆𝑆) in the second run are dependent on their own lagged values and on the
lagged values of the respective other variable. Optimal lag length (k) for each variable of the VAR
model is determined by minimizing the (Schwarz 1978) information criterion. Parameters ϵt and
ut represent the residuals of the regression, which are assumed to be mutually independent and
individually i.i.d. with zero mean and constant variance.
Based on the VAR model (5) and (6), we perform three further analyses. They are: Granger
causality testing, variance decompositions and impulse response function estimation. Granger
causality tests (Granger 1969) are applied to gain information about the lead-lag relationship
between returns volatility and the speculation ratio and returns volatility and the hedging ratio
respectively. The test will help to answer the question, whether speculative activity causes
conditional volatility in a forecasting sense and/or vice-versa. To test for Granger causality we
estimate a standard F-test and test the null hypothesis, which states that the speculative activity
(conditional volatility) does not Granger causes conditional volatility (speculative activity). The
hypothesis is rejected when coefficients of the lagged values are jointly significantly different from
zero (𝛽𝛽1 ≠ 𝛽𝛽2 ≠ … ≠ 𝛽𝛽𝑘𝑘 ≠ 0).
Next analysis, based on our VAR model, we obtain the variance decompositions. These
measure the percentage of the forecast error of a variable that is explained by another variable. It
indicates the conditional impact that one variable has upon another variable within the VAR system.
Therefore the variance decomposition mirrors the economic significance of these impacts as the
9
percentages of the forecast error for a variable sum to one (Fung and Patterson 1999). To find out
whether the causal relationships are positive or negative we compute impulse response functions,
which explain the impact of an exogenous shock in one variable on the other variables of the VAR
system. Finally, we generate the impulse responses to visually represent and analyse the behaviour
of volatility on simulated shocks in the speculation ratio or in the hedging ratio respectively and
vice versa.
4. Data
To examine China’s agricultural commodity markets, we analyse three heavily traded
commodity futures contracts for soybeans, soybean meal and sugar. Currently, there are four futures
exchanges in China, namely, the Dalian Commodity Exchange (DCE), the Zhengzhou Commodity
Exchange (ZCE), the Shanghai Futures Exchange and the China Financial Futures Exchange.
Agricultural commodities are mainly traded on DCE and ZCE. Therefore, our analyses focuses on
these two futures exchanges. The contracts for soybeans and soybean meal are traded on the DCE,
whereas the sugar futures contract is traded on the ZCE.
We have selected the most active contracts according to their trading volume. For all three
contracts, daily prices (settlement prices) and daily figures of trading volume and open interest (end
of day) are obtained from Thomson Reuters Datastream. Prices of contracts are quoted in Chinese
Yuan Renminbi per 10 metric ton (MT), daily trading volume represents the number of contracts
traded during a day and open interest reflects the number of contracts outstanding at the end of a
trading day. The sample period extends from 1 July, 2002 to 29 July, 2016 for soybeans, from 1
January, 2001 to 29 July, 2016 for soybean meal, and from 3 March, 2006 to 29 July, 2016 for
sugar. Table 1 provides the key specifications for each futures contract.
[Table 1 about here]
Table 2 displays summary statistics for returns 𝑟𝑟𝑡𝑡, open interest 𝑂𝑂𝑂𝑂𝑡𝑡, trading volume 𝑇𝑇𝑇𝑇𝑡𝑡, the
speculation ratio 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑡𝑡𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆 and the hedging ratio 𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑅𝑡𝑡
𝐻𝐻𝑆𝑆𝐻𝐻𝐻𝐻𝑆𝑆 for all three commodities examined.
10
It gives the mean, maximum, minimum, standard deviation (Std. Dev.), skewness, kurtosis and
Jarque–Bera statistics (JB) with corresponding probability values in parenthesis.
[Table 2 about here]
Several interesting findings are reported in table 2. The speculation ratios for sugar and soybean
meal futures have the highest means with 1.44 and 1.04. The ratio for soybeans futures shows the
lowest mean with 0.71. Note that a high ratio implicates a high amount of speculative activity
compared to hedging activity. In addition, the speculation ratio of soybean meal futures appears to
be most volatile as indicated by the largest maximum and minimum values and the high standard
deviation of the ratios.
The mean values of the hedging ratios are close to zero and negative for all contracts. The
distance of the extreme values of the hedging ratios is the highest for soybeans. A ratio close to
minus one indicates high speculative activity in respect with hedging activity. Soybean meal shows
the highest speculation with a hedging ratio of -0.99.
Mean returns are close to zero and positive for all the time series examined. According to the
distance of the extreme values (minimum, maximum) and the standard deviation, the market for
soybeans reveals the highest volatility. Skewness and kurtosis parameters indicate that none of the
three time series follows a normal distribution. This can be confirmed by the Jarque-Bera statistics
and its corresponding p-values. Regarding the results of Jarque-Bera tests the null hypothesis of
normal distribution is rejected for all series at the 1% significance level.
[Figure 1 about here]
Figure 1 shows futures prices and calculated speculation ratios for the three commodity
contracts examined. Futures prices for soybeans increased up to early 2008 but then crashed down
during the world financial crisis. Interestingly, the speculation ratio for soybeans has its maximum
also in the beginning of 2008. Time series for soybean meal also indicate co-movements in prices
and the speculation ratio. Peaks in soybean meal prices are also observable in the speculation ratio
for soybean meal. According to these graphs it seems like times of high prices coincide with a high
11
dominance of speculative activity compared to hedging activity. After reaching a peak in 2006,
futures prices for sugar fall rapidly till the beginning of 2009. After financial crisis sugar prices rise
up again and reach their maximum in the end of 2011.
We apply augmented Dickey and Fuller (1979) (ADF) unit root tests on prices, returns, trading
volume, open interest, speculation ratio and hedging ratio. The numbers of lags are selected in
accordance with the Akaike information criterion. Results of ADF tests are presented in Table 3.
The results show that prices of all contracts and some time series of trading volume and open
interest contain a unit root, whereas the ADF test clearly rejects the unit root for returns and both
ratios for all three contracts. Thus, each of the time series that are used in empirical tests are
stationary. To test for conditional heteroscedasticity we perform Engle’s Lagrange Multiplier (LM)
test (Engle 1982) on returns. The test results, also displayed in table 3, show that GARCH effects,
implying volatility clusters, are present in all series. The results of LM tests confirm our assumption
and motivate the usage of the GARCH model.
[Table 3 about here]
5. Empirical Results
In this section results of both GARCH and VAR analyses are discussed. We start with the
GARCH results, which are presented in table 4. The GARCH(1,1) model is used to measure the
influence of speculative activity on volatility. First, we run the GARCH model extended by the
speculation ratio. Some findings are similar across all three commodities examined. GARCH and
ARCH parameters and the ratio have a highly statistically significant and positive influence in each
case and stationarity requirements implying shocks die out in finite time are met for all contracts.
The constant, which represents the time-invariant level of conditional volatility, is positive and
highly significant for soybeans and soybean meal. The significantly positive influence of the
speculation ratio implicates that conditional volatility is driven by speculative activity in each case.
12
Results of the second run, when the hedging ratio is used as an explanatory variable, support
the results of the first run. For all contracts examined, the GARCH as well as ARCH parameters
are highly significant and positive. Additionally, all stationarity requirements are met. The hedging
ratio has a significantly negative influence on conditional volatility in each case, indicating a
stabilizing influence of hedging activity.4
[Table 4 about here]
Further, we discuss the results of the tests that are based on the VAR model. Optimal lag length
(k) for each variable of the VAR model is determined by minimizing the Schwarz information
criterion. We set maximum lag length at kmax=20 (4 trading weeks). For this purpose, all possible
combinations between 1 and 40 lags of the variables are considered. In the case of the speculation
ratio, for soybeans an optimal lag length of k=3 is chosen and optimal lag length of k=4 is selected
for the remaining commodities. In the case of the hedging ratio an optimal lag length of k=1 is used
for each of the commodities.
Table 5 reports results of Granger causality tests between speculation ratio (hedging ratio) and
conditional volatility for all three commodities examined. Number of observation, F-values,
probability values and the number of lags of Granger causality relations are displayed for each of
the contracts.
[Table 5 about here]
Testing for Granger causality between the speculation ratio and conditional volatility, the
results are as follows: for soybeans, soybean meal and sugar the null hypothesis can be rejected in
both cases. Therefore, speculation ratio Granger causes conditional volatility and volatility causes
speculation ratio also in the Granger sense. These results imply that the amount of speculative
activity in relation to hedging activity contains information about changes of volatility in the future.
Also, current volatility involves information about futures speculative activity. In the case of the
4 GARCH-in-Mean (GARCH-M) tests are also applied to the data but GARCH terms in the mean equations are not significant. Higher AR terms added in the mean equation are either insignificant or do not change the conclusions.
13
hedging ratio, the results indicate that the hedging ratio Granger causes conditional volatility for all
contracts but not vice versa.5
The VAR estimation results are also used to perform a variance decomposition for all
commodities, for that we were able to rejects the null hypothesis of no‐causality. Results of the
variance decomposition for volatility and speculation ratio as well as the hedging ratio are presented
in table 6. The table presents results in percent for trading day 1, 5, 15 and 20. Across all contracts
examined, we observe similar results. Variations in volatility are mostly caused by their own lagged
values, while the speculation ratio appears to play only a minor role in explaining return volatility.
Own lagged values of the speculation (hedging) ratio are also mostly responsible for its own
variation. Thus lagged volatility only explains a small effect of the variation of the two ratios.
[Table 6 about here]
In a next step we examine the impact of a one standard deviation shock to the VAR system.
Therefore we create impulse response functions for all commodities. Impulse respond functions
displays the response of volatility to simulated shocks to the speculation (hedging) ratio and vice
versa.
[Figures 2, 3 and 4 about here]
Figure 2, 3 and 4 display impulse response functions for all commodities examined. Figure 2
shows the response of conditional volatility to shocks in the speculation ratio, whereas figure 3
displays the response of the speculation ratio to volatility shocks. Since we were only able to find
unidirectional Granger causality in the case of the hedging ratio, only the significant response of
conditional volatility to shocks in the hedging ratio is presented in figure 4. Shocks are simulated
as one standard deviation and 20 trading days are regarded.
Regarding the speculation ratio, for all commodities the response of conditional volatility to
shocks in the speculation ratio is positive. This implicates that a rise in speculative activity leads to
5 Granger causality tests applied on samples only consisting of data exclusive since 2007 do not change the conclusions.
14
a rise in price volatility. The rise of volatility persists up to 5 days for soybeans, up to 9 days for
soybean meal and up to 12 days for sugar and afterwards volatility converged to its mean. All
responses are significant. One exception is the response of soybean meal volatility to shocks in the
ratio. The response becomes significant positive after 4 trading days. Volatility shocks also produce
only positive responses in speculation ratio for all commodities. Therefore, speculative activity is
driven by inclines in volatility. Figure 4 displays the response of volatility to shocks in the hedging
ratio. The response is significant negative for all three commodities, which implies that a rise in the
dominance of hedging activity in a contract leads to a reduction in conditional volatility. All results
of the VAR model support the results of the previous GARCH model.
6. Conclusion
Motivated by periods of high returns volatility as well as the ongoing finalization of
agricultural commodity futures markets, we investigate the impact of speculative activity on returns
volatility of Chinese commodity futures markets. We focus on Chinese futures markets because
these markets are believed to be highly speculative. Additionally, China’s futures markets for
commodities have grown rapidly in the last years and their global importance is increasing.
However, studies investigating Chinese futures markets, especially empirical ones, are rare. On that
account, we use a speculation ratio defined as trading volume divided by open interest to capture
the relative dominance of speculative activity in China’s futures markets. To verify our results we
use a second ratio, which captures the relative importance of the hedging behaviour instead of
speculative behaviour, by combining volume and open interest data in a different way.
To estimate the influence of speculative activity, proxied by the two ratios, on futures returns
volatility, we estimate both GARCH and VAR models. The empirical tests enable us to get insight
into the contemporaneous and the lead-lag relationship between speculative activity and returns
volatility of three heavily traded Chinese futures contracts, namely soybeans, soybean meal and
sugar. The contracts are traded on the Dalian Commodity Exchange and on the Zhengzhou
15
Commodity Exchange. We find a positive influence of the speculation ratio on volatility for all
commodities examined, using the GARCH model. These empirical results indicate that a rise in
speculative activity can lead to an increase in returns volatility. This deduction is supported by a
negative influence of the hedging ratio on returns volatility. Moreover, we show that for soybeans,
soybean meal and sugar both ratios Granger cause conditional volatility and vice versa. The
influence is positive in the case of the speculation ratio and negative in the case of the hedging ratio.
These results imply that the amount of speculative activity in relation to hedging activity contains
information about changes in futures volatility.
Interestingly, our results seem to be inconsistent with the results of the current literature, which
finds a stabilizing influence of speculation on returns volatility. The stabilizing hypothesis is quite
reasonable since speculation provides liquidity to the market, helps hedgers to find their
counterparties to hedge their risks, improves price discovery and therefore stabilizes prices.
So, why are our results contrary to previous findings? In contrast to our research, most of the
studies that have found stabilizing effects concentrate on US markets and use weekly data of CFTC
reports to measure speculation. Our research, however, investigates Chinese commodity futures
markets, which appear to be characterized by trading behaviour that is extremely speculative and
where presumable speculative activity often exceeds the hedging demand. In that context, our
results imply that speculation is not harmful in general, but excessive speculation, which is above
hedging needs drives returns volatility. Finally, our study suggests useful policy implications. First,
it is important to distinguish between speculation and excessive speculation and therefore policy
decisions should concentrate on an adequate level of speculation, in relation to the level of hedging
activity. Furthermore, regulation should aim to curb only “harmful” speculation, which is
disproportionately high to hedging activity.
16
7. References
Bera AK, Higgins ML (1993) ARCH MODELS: PROPERTIES, ESTIMATION AND TESTING. J Economic Surveys 7(4):305–366. doi: 10.1111/j.1467-6419.1993.tb00170.x
Bessembinder H, Seguin PJ (1993) Price Volatility, Trading Volume, and Market Depth: Evidence from Futures Markets. The Journal of Financial and Quantitative Analysis 28(1):21. doi: 10.2307/2331149
Bohl MT, Javed F, Stephan PM (2012) Do Commodity Index Traders Destabilize Agricultural Futures Prices? SSRN Journal. doi: 10.2139/ssrn.2088426
Bollerslev T (1986) Generalized autoregressive conditional heteroskedasticity. Journal of Econometrics 31(3):307–327. doi: 10.1016/0304-4076(86)90063-1
Brunetti C, Buyuksahin B, Harris JH (2011) Speculators, Prices and Market Volatility. SSRN Journal. doi: 10.2139/ssrn.1736737
Chan LH, Nguyen CM, Chan KC (2015) A new approach to measure speculation in the oil futures market and some policy implications. Energy Policy 86:133–141. doi: 10.1016/j.enpol.2015.06.034
Chatrath A, Ramchander S, Song F (1996) The role of futures trading activity in exchange rate volatility. J. Fut. Mark. 16(5):561–584. doi: 10.1002/(SICI)1096-9934(199608)16:5<561:AID-FUT4>3.0.CO;2-E
Cheng I-H, Xiong W (2014) Financialization of Commodity Markets. Annu. Rev. Financ. Econ. 6(1):419–441. doi: 10.1146/annurev-financial-110613-034432
Dickey DA, Fuller WA (1979) Distribution of the Estimators for Autoregressive Time Series With a Unit Root. Journal of the American Statistical Association 74(366):427. doi: 10.2307/2286348
Ederington L, Lee JH (2002) Who Trades Futures and How: Evidence from the Heating Oil Futures Market. J BUS 75(2):353–373. doi: 10.1086/338706
Fung H-G, Patterson GA (1999) The dynamic relationship of volatility, volume, and market depth in currency futures markets. Journal of International Financial Markets, Institutions and Money 9(1):33–59. doi: 10.1016/S1042-4431(98)00035-3
Garcia P, Leuthold RM, Zapata H (1986) Lead-lag relationships between trading volume and price variability: New evidence. J. Fut. Mark. 6(1):1–10. doi: 10.1002/fut.3990060102
Gilbert CL, Morgan CW (2010) Food price volatility. Philos Trans R Soc Lond B Biol Sci 365(1554):3023–3034. doi: 10.1098/rstb.2010.0139
Granger CWJ (1969) Investigating Causal Relations by Econometric Models and Cross-spectral Methods. Econometrica 37(3):424. doi: 10.2307/1912791
Irwin SH, Sanders DR (2012) Testing the Masters Hypothesis in commodity futures markets. Energy Economics 34(1):256–269. doi: 10.1016/j.eneco.2011.10.008
Irwin SH, Sanders DR, Merrin RP (2009) Devil or Angel?: The Role of Speculation in the Recent Commodity Price Boom (and Bust). J. agric. appl. econ. 41(02):377–391. doi: 10.1017/S1074070800002856
17
Kim A (2015) Does Futures Speculation Destabilize Commodity Markets? Journal of Futures Markets 35(8):696–714. doi: 10.1002/fut.21716
Leuthold RM (1983) Commercial use and speculative measures of the livestock commodity futures markets. J. Fut. Mark. 3(2):113–135. doi: 10.1002/fut.3990030202
Liao TX, Morse EL, Doshi A, Wilson DB, Yuen A, Lee EG, Kleinman SM, Jansen HR (2016) China Commodities: Hold Onto Your Hats - Explosion in Chinese Commodities Futures Brings Unprecedented Liquidity, Untested Volatility. Citi Research.
Lucia JJ, Mansanet-Bataller M, Pardo Á (2015) Speculative and hedging activities in the European carbon market. Energy Policy 82:342–351. doi: 10.1016/j.enpol.2014.11.014
Lucia JJ, Pardo A (2010) On measuring speculative and hedging activities in futures markets from volume and open interest data. Applied Economics 42(12):1549–1557. doi: 10.1080/00036840701721489
Master MW (2008) Testimony before the Committee on Homeland Security and Governmental Affairs, U.S. Senate. May 20.
Master MW, White AK (2008) The accidental Hunt brothers: how institutional investors are driving up food and energy prices. Special report.
Miffre J, Brooks C (2013) Did Long-Short Investors Destabilize Commodity Markets? SSRN Journal. doi: 10.2139/ssrn.2244526
Peck AE (1982) Estimation of hedging and speculative positions in futures markets revisited. Food Research Institute Studies(18):181–195.
Robles M, Torero M, Braun J von (2009) When speculation matters. International Food Policy Research Institute, Policy Brief 57.
Rutledge DJ (1979) Trading volume and price variability: New evidence on the price effects of speculation. Futures markets: Their establishment and performance:137–156.
Sanders DR, Irwin SH, Merrin RP (2010) The Adequacy of Speculation in Agricultural Futures Markets: Too Much of a Good Thing? Applied Economic Perspectives and Policy 32(1):77–94. doi: 10.1093/aepp/ppp006
Schwarz G (1978) Estimating the Dimension of a Model. Ann. Statist. 6(2):461–464. doi: 10.1214/aos/1176344136
Stoll HR, Whaley RE (2009) Commodity Index Investing and Commodity Futures Prices. SSRN Journal. doi: 10.2139/ssrn.1478195
Streeter DH, Tomek WG (1992) Variability in soybean futures prices: An integrated framework. J. Fut. Mark. 12(6):705–728. doi: 10.1002/fut.3990120609
Till H (2009) Has There Been Excessive Speculation in the U.S. Oil Futures Markets?: What Can We (Carefully) Conclude from New CFTC Data? SSRN Journal. doi: 10.2139/ssrn.2608027
Wang HH, Ke B (2005) Efficiency tests of agricultural commodity futures markets in China. Aust J Agric Res Econ 49(2):125–141. doi: 10.1111/j.1467-8489.2005.00283.x
Wang X, Ye Q, Zhao F (2016) Trading activity and price behavior in Chinese agricultural futures markets. Finance Research Letters. doi: 10.1016/j.frl.2016.03.028
18
Wiley MK, Daigler RT (1998) Volume relationships among types of traders in the financial futures markets. J. Fut. Mark. 18(1):91–113. doi: 10.1002/(SICI)1096-9934(199802)18:1<91:AID-FUT5>3.0.CO;2-6
19
Table 1: Contract specifications
Contract Futures Exchange
Contract Size Currency Sample Number of Obs.
No. 1 Soybeans Dalian Commodity exchange (DCE)
10 MT Chinese Yuan Renminbi
7/01/2002 -7/29/2016 (daily) 3227
Soybean Meal Dalian Commodity exchange (DCE)
10 MT Chinese Yuan Renminbi
01/05/2001 - 07/29/2016 (daily) 3475
White Sugar Zhengzhou Commodity exchange (ZCE)
10 MT Chinese Yuan Renminbi
3/03/2006 -7/29/2016 (daily) 2487
Table 2: Descriptive Statistic
Note: This table presents descriptive statistics of the investigated time series of the three futures contracts. 𝐫𝐫𝐭𝐭, 𝐎𝐎𝐎𝐎𝐭𝐭 and 𝐓𝐓𝐓𝐓𝐭𝐭 describe the returns, end- of-day open interest and daily trading volume on day t. The speculation ratio is represented by 𝐑𝐑𝐑𝐑𝐭𝐭𝐑𝐑𝐑𝐑𝐭𝐭
𝐒𝐒𝐒𝐒𝐒𝐒𝐒𝐒and the hedging Ratio by 𝐑𝐑𝐑𝐑𝐭𝐭𝐑𝐑𝐑𝐑𝐭𝐭𝐇𝐇𝐒𝐒𝐇𝐇𝐇𝐇𝐒𝐒. JB stands for Jarque-Bera statistics and significance at
the 1% level is represented by ***. Both open interest and trading volume are presented in units of 1000. All data is taken from Thomson Reuters Datastream.
Variable Mean Maximum Minimum Std.dev. Skewness Kurtosis JB
Soybeans 𝒓𝒓𝒕𝒕 0.016 6.189 -9.594 1.057 -0.570 11.965 10980.468***
𝑶𝑶𝑶𝑶𝒕𝒕 461.467 1205.210 87.022 210.408 0.918 3.840 548.340***
𝑻𝑻𝑻𝑻𝒕𝒕 338.442 2677.400 2.796 317.679 2.041 9.035 7136.405***
𝑹𝑹𝑹𝑹𝒕𝒕𝑹𝑹𝑹𝑹𝒕𝒕𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺
0.708 7.099 0.015 0.588 2.591 15.221 23693.053***
𝑹𝑹𝑹𝑹𝒕𝒕𝑹𝑹𝑹𝑹𝒕𝒕𝑯𝑯𝑺𝑺𝑯𝑯𝑯𝑯𝑺𝑺
-0.010 0.844 -0.990 0.094 -2.097 24.542 64759.520***
Soybean Meal 𝒓𝒓𝒕𝒕 0.017 8.792 -14.644 1.423 -0.963 13.974 17974.620***
𝑶𝑶𝑶𝑶𝒕𝒕 1240.684 5837.670 7.682 1319.471 0.973 2.826 552.577***
𝑻𝑻𝑻𝑻𝒕𝒕 1066.920 11868.480 1.800 1313.051 2.336 10.652 11638.600***
𝑹𝑹𝑹𝑹𝒕𝒕𝑹𝑹𝑹𝑹𝒕𝒕𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺 1.104 8.379 0.010 0.803 2.229 11.626 13653.288***
𝑹𝑹𝑹𝑹𝒕𝒕𝑹𝑹𝑹𝑹𝒕𝒕𝑯𝑯𝑺𝑺𝑯𝑯𝑯𝑯𝑺𝑺
-0.004 0.591 -0.999 0.065 -3.626 48.854 312055.157***
Sugar 𝒓𝒓𝒕𝒕 0.006 10.796 -10.370 1.201 -0.092 15.805 16993.448***
𝑶𝑶𝑶𝑶𝒕𝒕 801.586 1556.438 7.116 364.161 -0.338 2.495 73.648***
𝑻𝑻𝑻𝑻𝒕𝒕 1112.418 5438.290 1.436 836.960 1.389 5.428 1410.622***
𝑹𝑹𝑹𝑹𝒕𝒕𝑹𝑹𝑹𝑹𝒕𝒕𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺 1.443 7.594 0.029 0.897 1.651 7.779 3496.266***
𝑹𝑹𝑹𝑹𝒕𝒕𝑹𝑹𝑹𝑹𝒕𝒕𝑯𝑯𝑺𝑺𝑯𝑯𝑯𝑯𝑺𝑺 -0.001 0.527 -0.774 0.047 -0.989 48.293 212987.313***
20
Figure 1: Futures prices and speculation ratios for soybeans, soybean meal, and sugar.
2000
2500
3000
3500
4000
4500
5000
5500
6000
2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016
CN
Y/1
0 M
TSoybeans Futures Price
0
1
2
3
4
5
6
7
8
2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016
Rat
io (T
V/O
I)
Soybeans Speculation Ratio
21
1000
1500
2000
2500
3000
3500
4000
4500
5000C
NY
/10
MT
Soybean Meal Futures Price
0
1
2
3
4
5
6
7
8
9
Rat
io (T
V/O
I)
Soybeans Meal Speculation Ratio
22
Note: The graphs show daily prices quoted in Chinese Yuan Renminbi (CNY) per 10 MT and daily speculation ratios, computed as daily trading volume divided by end of day open interest for soybeans, soybean meal and sugar. All data is taken from Thomson Reuters Datastream.
2000
2500
3000
3500
4000
4500
5000
5500
6000
2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016
CN
Y/1
0 M
TSugar Futures Price
0
1
2
3
4
5
6
7
8
2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016
Rat
io (T
V/O
I)
Sugar Speculation Ratio
23
Table 3: Augmented Dickey Fuller (ADF) Test and Lagrange Multiplier (LM) test
Price Returns Volume Open Interest 𝑹𝑹𝑹𝑹𝒕𝒕𝑹𝑹𝑹𝑹𝒕𝒕
𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺 𝑹𝑹𝑹𝑹𝒕𝒕𝑹𝑹𝑹𝑹𝒕𝒕
𝑯𝑯𝑺𝑺𝑯𝑯𝑯𝑯𝑺𝑺
Soybeans -2.21 (7) -20.27***(6) -4.63***(20) -3.66***(17) -5.68***(17) -11.77***(13)
Soybean Meal -2,64*(6) -19,39***(6) -2,5(29) -1,56(29) -4,26***(29) -16,02***(10)
Sugar -1.11(6) -21.52**(5) -4.2***(11) -2,58*(21) -5.64***(10) -44,19***(0)
LM(1) LM(5) LM(10) LM(15) LM(20)
Soybeans 44,19*** 74,78*** 91,87*** 92,01*** 93,22***
Soybean Meal 27.61*** 49.88*** 55.39*** 60.31*** 123.04***
Sugar 13.00*** 27.17*** 73.83*** 76.34*** 77.99***
Notes: First rows show results of the ADF test and the lower rows show results of the LM tests. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% level, respectively. Numbers of Lags for each ADF- and LM-test are given in parenthesis. Lags are chosen according the AIC.
Table 4: GARCH(1,1)
Speculation Ratio
Soybeans Soybean Meal Sugar
Constant 0.19*** 0.19*** -0.07***
Resid² 0.26*** 0.19*** 0.18***
Volatility 0.38*** 0.53*** 0.51***
𝑹𝑹𝑹𝑹𝒕𝒕𝑹𝑹𝑹𝑹𝒕𝒕𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺
0.37*** 0.37*** 0.39***
Hedging Ratio
Soybeans Soybean Meal Sugar
Constant 0.28*** 0.42*** 0.58***
Resid² 0.3*** 0.25*** 0.03***
Volatility 0.49*** 0.57*** 0.58***
𝑹𝑹𝑹𝑹𝒕𝒕𝑹𝑹𝑹𝑹𝒕𝒕𝑯𝑯𝑺𝑺𝑯𝑯𝑯𝑯𝑺𝑺 -0.67*** -1.24*** -1.90***
Notes: Table 4 shows results for volatility equation (eq. 4) using normal distribution. Resid²(-1) and Volatility(-1) represent squared residuals from mean equation (eq. 3) and conditional volatility. Ratiot
Specstands for the computed speculative ratio and captures speculative activity. Ratiot
Hedgerepresents the hedging ratio. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% level, respectively.
24
Table 5: Results Granger-Causality test
Null Hypothesis: Obs F-Statistic Prob.
Soybeans
𝑹𝑹𝑹𝑹𝒕𝒕𝑹𝑹𝑹𝑹𝒕𝒕𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺 does not Granger Cause
Conditional Volatility 3224 13.0781*** 0.00000002
Conditional Volatility does not Granger Cause 𝑹𝑹𝑹𝑹𝒕𝒕𝑹𝑹𝑹𝑹𝒕𝒕
𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺 Lags = 3 7.57535*** 0.00005
𝑹𝑹𝑹𝑹𝒕𝒕𝑹𝑹𝑹𝑹𝒕𝒕𝑯𝑯𝑺𝑺𝑯𝑯𝑯𝑯𝑺𝑺 does not Granger Cause
Conditional Volatility 3226 13.8497*** 0.0002
Conditional Volatility does not Granger Cause 𝑹𝑹𝑹𝑹𝒕𝒕𝑹𝑹𝑹𝑹𝒕𝒕
𝑯𝑯𝑺𝑺𝑯𝑯𝑯𝑯𝑺𝑺 Lags = 1 2.83604* 0.0923
Soybean Meal
𝑹𝑹𝑹𝑹𝒕𝒕𝑹𝑹𝑹𝑹𝒕𝒕𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺
. does not Granger Cause Conditional Volatility
3470 9.53301*** 1.E-07
Conditional Volatility does not Granger Cause 𝑹𝑹𝑹𝑹𝒕𝒕𝑹𝑹𝑹𝑹𝒕𝒕
𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺 Lags 4 6.13641*** 6.E-05
𝑹𝑹𝑹𝑹𝒕𝒕𝑹𝑹𝑹𝑹𝒕𝒕𝑯𝑯𝑺𝑺𝑯𝑯𝑯𝑯𝑺𝑺 does not Granger Cause
Conditional Volatility 3473 11.0778*** 0.0009
Conditional Volatility does not Granger Cause 𝑹𝑹𝑹𝑹𝒕𝒕𝑹𝑹𝑹𝑹𝒕𝒕
𝑯𝑯𝑺𝑺𝑯𝑯𝑯𝑯𝑺𝑺 Lags = 1 1.86200 0.1725
Sugar
𝑹𝑹𝑹𝑹𝒕𝒕𝑹𝑹𝑹𝑹𝒕𝒕𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺does not Granger Cause
Conditional Volatility 2482 10.1834*** 4.E-08
Conditional Volatility does not Granger Cause 𝑹𝑹𝑹𝑹𝒕𝒕𝑹𝑹𝑹𝑹𝒕𝒕
𝑺𝑺𝑺𝑺𝑺𝑺𝑺𝑺 Lags = 4 2.79944** 0.0246
𝑹𝑹𝑹𝑹𝒕𝒕𝑹𝑹𝑹𝑹𝒕𝒕𝑯𝑯𝑺𝑺𝑯𝑯𝑯𝑯𝑺𝑺does not Granger Cause
Conditional Volatility 2485 8.91242*** 0.0029
Conditional Volatility does not Granger Cause 𝑹𝑹𝑹𝑹𝒕𝒕𝑹𝑹𝑹𝑹𝒕𝒕
𝑯𝑯𝑺𝑺𝑯𝑯𝑯𝑯𝑺𝑺 Lags = 1 1.22867 0.2678
Note: F-values of the Granger causal relations for each variable within the VAR model are presented. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% level, respectively. The lag lengths are based on the SIC.
25
Table 6: Variance Decomposition
Speculation Ratio Soybeans Soybean meal Sugar
Explained Variable Day Volatility Ratiot
Spec Volatility RatiotSpec Volatility Ratiot
Spec
Volatility 1 100.00 0.00 100.00 0.00 100.00 0.00 5 98.78 1.22 99.72 0.28 98.88 1.12 10 97.08 2.92 98.47 1.53 96.21 3.79 15 96.29 3.71 97.38 2.62 93.29 6.71 20 95.97 4.03 96.62 3.38 90.80 9.2
RatiotSpec 1 1.10 98.90 1.21 98.79 0.20 99.79
5 3.39 96.61 2.91 97.09 1.61 98.39 10 4.25 95.75 3.84 96.16 2.49 97.51 15 4.53 95.47 4.28 95.72 3.17 96.83 20 4.64 95.36 4.51 95.49 3.66 96.34
Hedging Ratio Soybeans Soybean meal Sugar
Explained Variable Day Volatility Ratiot
Hedge Volatility RatiotHedge Volatility Ratiot
Hedge
Volatility 1 100.00 0.00 100.00 0.00 100.00 0.00 5 99.50 0.50 99.64 0.36 99.63 0.37 10 99.49 0.51 99.63 0.37 99.57 0.43 15 99.49 0.51 99.63 0.37 99.56 0.44 20 99.49 0.51 99.63 0.37 99.56 0.44
RatiotHedge 1 0.05 99.95 0.01 99.99 0.03 99.97 5 0.13 99.87 0.08 99.92 0.07 99.93 10 0.14 99.86 0.08 99.92 0.08 99.92 15 0.14 99.86 0.08 99.92 0.09 99.91 20 0.14 99.86 0.08 99.92 0.09 99.91
Note: The table presents variance decompositions based on the two VAR models with endogenous variables volatility and 𝐑𝐑𝐑𝐑𝐭𝐭𝐑𝐑𝐑𝐑𝐭𝐭
𝐒𝐒𝐒𝐒𝐒𝐒𝐒𝐒( 𝐑𝐑𝐑𝐑𝐭𝐭𝐑𝐑𝐑𝐑𝐭𝐭𝐇𝐇𝐒𝐒𝐇𝐇𝐇𝐇𝐒𝐒).
Figure 2: Impulse Response Functions – Response of Volatility to Speculation Ratio
26
Note: Impulse response functions are displayed along with corresponding plus and minus 2 standard error bands (dashed lines), used to determine statistical significance. The impulse response functions show responses to Cholesky one standard deviation innovations. The horizontal axis shows the number of days after the shock.
27
Figure 3: Impulse Response Functions – Response of Speculation Ratio to Volatility.
28
Note: Impulse response functions are displayed along with corresponding plus and minus 2 standard error bands (dashed lines), used to determine statistical significance. The impulse response functions show responses to Cholesky one standard deviation innovations. The horizontal axis shows the number of days after the shock.
Figure 4: Impulse Response Functions – Response Volatility to Hedging ratio.
29
Note: Impulse response functions are displayed along with corresponding plus and minus 2 standard error bands (dashed lines), used to determine statistical significance. The impulse response functions show responses to Cholesky one standard deviation innovations. The horizontal axis shows the number of days after the shock.