The Performance of Commodity Futures Markets A Comparison of China and US Wheat Futures

1
The Performance of Commodity Futures MarketsA
Comparison of China and US Wheat Futures

• August 2, 2003
• Wen Du
• H. Holly Wang
• Washington State University

2
Background

• China the world biggest wheat production and
  consumption country
• Wheat futures markets
• China Zhengzhou Commodity Exchange (CZCE)
• founded in 1993
• the only exchange trading wheat futures in China
• Chicago Board of Trade (CBOT)
• founded in 1848
• the world largest and most developed commodity
  market

3
Objectives

• To study and compare price behaviors of Chinas
  and US wheat futures using time-series methods
• To analyze the integration of Chinas market to
  the world market

4
Data

• Daily settlement prices of September contract
• Jan 4, 2000 Sept 13, 2002
• Properties
• unit root
• time-varying volatility
• excess kurtosis

5
Wheat Daily Settlement Price of CZCE September
Contract
Unit Yuan/Ton
6
Methods

• Univariate/multivariate time series analysis
• Conditional heteroskedasticity modeling
• Engle (1982) Autoregressive Conditional
  Heteroskedasticity (ARCH)
• Bollerslev (1986) Generalized ARCH (GARCH)

7
Univariate Model

• GARCH(p, q) model
• where
  
• dependent variable (first
  difference of prices)
• explanatory variable
  vector (constant dummy variables)
• time period
  error component
• information set at t-1
  conditional variance
• p, q numbers of lags
• , , , parameters
  to be estimated

8
Multivariate Model

• M-variate GARCH(P, Q) model
• where
• dependent variable
  vector (first difference of prices)
• conditional
  variance-covariance matrix
• constant term matrix
• , coefficient matrix
  in the variance equation
• P, Q numbers of lags

9
Univariate Model Fitting

• Criteria
• Akaike Information Criterion (AIC) and Bayesian
  Information Criterion (BIC)
• significance
• Most fitted models
• CZCE ARCH(1)
• CBOT ARCH(1)
• GARCH(1,1)
• Goodness-of-fit improved under t distribution

10
Univariate Estimation Results - N

• no evident drifts in both price changes
• contract switching effect significant and
  positive in CZCE but insignificant in CBOT
• ARCH effect is relatively small compared to the
  constant in the variance equation for CZCE, while
  it is stronger for CBOT

11
Univariate Estimation Results - t

• Similar pattern as under normality
• ARCH/GARCH effect enhanced
• For CBOT, GARCH part more influential than ARCH
  part on current volatility

12
Bivariate Model Fitting

• Most fitted models
• DVEC GARCH(1,1)
• ,
• BEKK ARCH(1)
• Goodness-of-fit improved under t distribution

13
Bivariate Estimation Results DVEC-GARCH(1,1)

• Normal conditional variance
• contract switching in CBOT significant for both
  own- and cross-market effects while that in CZCE
  only significant for own effect
• GARCH part more influential than ARCH part on
  current volatility
• t conditional variance
• cross-market effects of contract switching
  insignificant in the two markets

14
Own- Cross-market effects BEKK-ARCH(1)
15
Bivariate Estimation Results BEKK-ARCH(1)

• Normal conditional variance
• cross-market effects of contract switching
  insignificant
• in price volatility in CBOT, own-market effect
  dominant and effect from CZCE very small
• in price volatility in CZCE, own-market effect
  dominant but effect from CBOT stronger
• t conditional variance
• cross-market effects weaker
• the gap between the two markets impacts on each
  other even larger

16
Conclusions

• ARCH/GARCH price models
• Wheat futures prices in CZCE and CBOT can be
  modeled by ARCH(1) or GARCH(1,1).
• Models have better fit when conditional variance
  is t distributed.
• The interrelations between CZCE and CBOT are
  significant but asymmetric.
• CZCE closely correlated with world market
• CBOT leader and CZCE follower

17
Questions?