You are on page 1of 5

Applied Economics Letters, 2009, 16, 15551558

Price discovery for copper futures in


informationally linked markets
Xindan Li* and Bing Zhang
School of Management and Engineering, Najing University, P.R. China

The purpose of this article is to contribute to the research


on informationally linked markets by investigating the relationships
between the Chinese copper futures market and its London counterparts.
There is a long run relationship between the Shanghai Futures Exchanges
(SFE) with London Metals Exchange (LME) copper futures prices.
Furthermore, we find that three regime Markov switching model with
changing intercept and variance turns out to be good description of the
data. The influence of LME on SFE is bigger than that of SFE on LME.

I. Introduction
A significant number of studies have examined the
efficiency of copper futures markets using different
methodological
techniques.
For
example,
MacDonald and Taylor (1988a) test the EMH for
four metals in the London Metals ExchangesLME
covering the period 19761987 and find that the
copper and lead futures markets can be considered as
efficient, whilst the EMH is rejected for tin and zinc.
MacDonald and Taylor (1988b) support the EMH
for the same metals in the LME for the period
19761985, using cointegration method. MacDonald
and Taylor (1989) find weak evidence for the presence
of time varying premium. However, only few studies
have explored the price discovery in the domestic
futures and foreign futures markets simultaneously.
For example, Booth et al. (1998) have found
a cointegration relationship between the prices
of wheat futures contracts traded in the Chicago
Board of Trade (CBOT) and the Winnipeg
Commodities Exchange (WCE) of Canada.
During the last ten years, the Chinese copper
consumption has grown at about 2.4 times to the
world average. The Shanghai Futures Exchange

(SFE) has become the second largest copper futures


market in the world and its trading volume in terms
of tonnage had exceeded that of the New York
Mercantile Exchange. Given the rapid development
of Chinese copper futures market and the process of
globalization of trading and competition among
futures exchanges, it is very important to understand
the international linkage between the Chinese and
world copper futures markets. However, little
research has been performed on Chinese copper
futures markets, not alone its linkage with world
leading market.
This article is one of the first investigations to find
the price discovery for the copper futures in two
informationally-linked markets, the SFE and the
LME. Our study differs from the above in the
following ways: we examine the dynamic relationship
between the two variables using not only the linear
VECM but also using the Markov switching-VECM
framework. The traditional linear models do not
allow the parameters to adjust for the structural
changes, which is inappropriate, given that the
number and composition of participants in
the futures markets are likely to change. In fact, the
price of copper futures increased over 50% in 2005

*Corresponding author. E-mail: zhangbing@nju.edu.cn


Applied Economics Letters ISSN 13504851 print/ISSN 14664291 online 2009 Taylor & Francis
http://www.informaworld.com
DOI: 10.1080/13504850701578801

1555

X. Li and B. Zhang

1556
and continued to rise rapidly in the first half of 2006.
The rest of the article is arranged as follows.
Section II describes the data and the test framework.
Section III presents empirical results. Section IV
concludes.

Table 1. Johansen cointegration test


Statistics

Copper

5% Critical value

Null
hypothesis

trace

max

trace

max

r0
r1

53.04
8.52

44.51
8.52

20.26
9.16

15.82
9.16

II. Data and Methodology


Our sample covers from 4 January 2000 to 17 March
2006. We deleted nonmatching data caused by
holidays and nontrading dates in order to make the
data of the Shanghai and London copper futures
prices comparable. In doing so, we obtained 1410
observations. We also consolidated the quotation
units for our data, for consistency, we converted
LME quotation unit into RMB per ton.
Following Krolzig (1997), the estimation tests were
done with two-stage maximum likelihood estimation.
Initially the VAR is approached with a finite order
and then the Johansen procedure is used to know the
number of cointegration vectors. After, the other
parameters of the Markov switching-VAR representation are estimated conditionally to the cointegration matrix already estimated in the first stage.
Finally, the Equation 1 presents an error correction
model, shaped with the form MS-VECM, with a
change in the drift and in the long run equilibrium:
xt  st  0 xt1  st

p1
X

i xtk

informational linked futures exchanges are integrated


of order 1.1
Cointegration test
Since the time series of copper futures prices follow
I(1) process, we can test the cointegration relationship
between the copper futures prices in Shanghai
exchange and London futures exchange. Table 1
reports the cointegration results given by the
Johansen tests (Johansen, 1988). Both the trace and
max statistics indicate that the Chinese copper futures
prices are cointegrated with their London counterparts, even though copper futures prices are nonstationary and therefore, in the short run the futures
prices in the Shanghai and London exchanges can
diverge; there is a long-term equilibrium relationship
that binds the movement of the copper futures prices
in this informational linked markets.
Error correction model

 st ut

k1

where ut  NID0, ; xt is the vectors of the


variables; st is the drift regime dependent term; st
is the long-run equilibrium;  is the speed of
adjustment coefficients; st is governed by a Markov
channel with a finite number of states defined by the
probabilities of transition pij.

Since copper futures in SFE and LME are cointegrated, the error correction term can be represented
by the following model:
e S  0:967240 L  0:512385
0:0140

0:1374

SFE

Here, S and L represent the logarithm of copper


futures prices in SFE and LME, respectively. The
numbers in parentheses are standard deviations of
coefficients.
MS-VECM model

III. Empirical Results


In order to perform the cointegration test, we first
need to determine the order of integration of
each price series. We use the standard Augmented
DickeyFuller (ADF) test to determine if the unit
root existed in each time series. In all cases, with or
without a deterministic trend, the copper prices in the

The cointegration results from above are now used


in the second stage of our analysis. We adopt
an MSIH-VECM with 3 regimes and 1 lag
(MSIH(3)-VECM(1)) model, with shifts in the intercept st and the error variance.
The LR linearity test (LR 674.33) significantly
rejects the linearity hypothesis even when considering the upper bound suggested by Davies (1977).

To save space, we have not provided the results here, they are available upon request.

Copper futures prices in informationally linked markets

1557

Probabilities of regime 1
1.0

Smoothed

0.5

2000

2001

2002

2003

2004

2005

2006

2002

2003

2004

2005

2006

2003

2004

2005

2006

Probabilities of Regime 2
1.0

Smoothed

0.5

2000

2001
Probabilities of regime 3

1.0

Smoothed

0.5

2000

2001

2002

Fig. 1.

Smoothed regime probabilities

Table 2. ML estimates of the MSIH(3)VECM(1)

Intercept st 1
Intercept st 2
Intercept st 3
Rs(1)
Rl(1)
Error correction term (1)
SE (Regime 1)
SE (Regime 2)
SE (Regime 3)

MSIH-VAR for SFE

MSIH-VAR for LME

0.299019
0.002945
0.152382
0.223626
0.462333
4.411114
2.053092
0.540778
1.033495

0.392786 (1.0410)
0.000912
(0.0251)
0.265117
(3.9251)
0.024519 (0.7405)
0.053909 (1.8195)
1.142235
(0.8018)
2.688960
0.92976
1.283657

(1.1780)
(0.1383)
(2.8386)
(9.7287)
(22.7343)
(4.5105)

Note: The numbers in parentheses are t-values.

This allows us to evaluate the difference in the


regime-dependent speed of adjustment parameters.
Testing for the number of regimes in an MSIHVECM model, we have used the tests based on
information criteria (AIC/HQ). We choose 3 regime
for the AIC (with 5.50 vs. 5.56) and the HQ (with
5.54 vs. 5.58) are in favour of it. Besides, MS(2)
model exhibit two symmetric regimes, which is not
appropriate.
The resulting regime probabilities are given
in Fig. 1. Regime 3 characterizes high-growth
episodes, note that it mainly happened since 2004,
which might indicate a structural change in the phase
structure. Regime 3 can be named as speculative
market of copper market, the average duration time
is 11.6 days.
While Regime 1 depicts very clearly the
suddenly drop, the average duration time is 2.7

days, it often happen shortly after speculative phase.


Regime 2 represents by default normal rise.
It appeared mainly before 2004 and the probability
of staying in this regime is 50%, it lasts 45 days
on average.
The estimated parameters of the MSIH(3)VECM(1) model are presented in Table 2. The
transition matrix is given by


 0:6338 0:0008 0:365 





p  0:0083 0:9779 0:0137 


 0:0595 0:0268 0:9137 
where,
pij Pst jjst1 i

LME

X. Li and B. Zhang

1558
Rs(1) and Rl(1) are the return of copper futures
in SFE and LME at time t1. The coefficients of
the error correction terms measure the speed with
which deviations from the long-run relationship are
corrected by changes in the futures prices of the two
markets.
At a 5% level of significance, the coefficients of the
error correction term, , for SFE is significant. This
implies that the error correction term is an important
factor in influencing the SFE copper futures prices.
However, both the coefficients of Rs(1) and 
coefficient for LME are not significant, these results
imply that the SFE copper futures market do not
influence the LSE copper futures market.
Summarizing the above findings, we conclude that
LME have stronger influence to SFE.

further test time varying relations of copper and other


metal futures in SFE and LME.

Acknowledgements
This research was supported by the fund for Study
on the Evolution of Complex Economic System and
Behaviour Finance at Innovation Center of
Economic Transition and Development of Nanjing
University and China National Science Fund NSFC
70671053. This research was also supported by the
China National Social Science Fund 07CJL014.

References

IV. Conclusion
This article is one of the first researches on
informationally linked markets by investigating the
relationships between the Chinese copper futures and
its London counterparts. Studying such a relationship
could shed light on the nature of cross-market
information transmission. There is a long run
relationship between the SFE and LME copper
futures prices. The results dovetail the casual
observations that the copper markets in China are
becoming more open, as China has abolished import
quotas and reduced the import tariffs greatly.
Furthermore, we found that three regime Markov
switching model with changing intercept and variance
turns out to be good description of the data. The
influence of LME on SFE is bigger than that of SFE
on LME. In future work it would be interesting to

Booth, G. G., Brockman, P. and Tse, Y. (1998) The


relationship between US and Canadian wheat futures,
Applied Financial Economics, 8, 7380.
Davies, R. B. (1977) Hypothesis testing when a nuisance
parameter is only present under the alternative,
Biometrica, 74, 24754.
Johansen, S. (1988) Statistical analysis of cointegration
vector, Journal of Economic Dynamics and Control, 12,
23154.
Krolzig, H. M. (1997) Markov-Switching Vector
Autoregressions: Modelling, Statistical Inference and
Application to Business Cycle Analysis, Springer,
Berlin.
MacDonald, R. and Taylor, M. P. (1988a) Testing rational
expectations and efficiency in the London metal
exchange, Oxford Bulletin of Economics and
Statistics, 50, 4152.
MacDonald, R. and Taylor, M. P. (1988b) Metals prices,
efficiency and cointegration: some evidence from the
London metal exchange, Bulletin of Economic
Research, 40, 23539.
MacDonald, R. and Taylor, M. P. (1989) Rational
expectations, risk and efficiency in the London metal
exchange: an empirical analysis, Applied Economics,
21, 14353.

You might also like