Price Discovery on the International Soybean Futures Markets: A Threshold Co‐Integration Approach

• DOI: http://doi.org/10.1002/fut.21794
• Date: 01 January 2017
• Published date: 01 January 2017

Price Discovery on the International

Soybean Futures Markets: A Threshold

Co-Integration Approach

Chao Li* and Dermot J. Hayes

This paper investigates the lead-lag relationships among soybean prices in United States,

Brazilian, and Chinese futures markets. We focus on both long-run price co-movements and on

short-run price relationships. Various co-integration methodologies and causality tests are

applied to examine the changes in price relationships over time. The empirical results indicate

the following: (a) the soybean futures market in the U.S. is still the most important and

influential market, and the U.S. price, in the long-term, leads price changes in Brazil and

China; (b) in the short-term, the overnight return of U.S. soybean futures and the daytime

return of Chinese No. 1 soybean futures contemporaneously affect each other, but there is no

significant causality between U.S. overnight return and the daytime return of Chinese No. 2

soybean futures; and, (c) a weak temporal seasonal causality between U.S. and Brazilian

soybean futures price exists and more often than not Brazilian futures lead U.S. futures during

the Brazilian growing season. © 2016 Wiley Periodicals, Inc. Jrl Fut Mark 37:52–70, 2017

1. INTRODUCTION

The United States, Brazil, and Argentina account for over 90% of the world’s soybean exports.

China, which imported 71.4 million tons of soybeans in 2014, is by far the largest importer

and gets approximately 50% of its soybeans from the United States and 40% from Brazil.

The United States, Brazil, and China all have active soybean futures. China has two

different markets, one for non-GMO soybeans and the other for imported GMO soybeans. In

the Dalian Commodity Exchange (DCE), the No. 1 contact is for non-GMO soybeans that

are used for human consumption, and the No. 2 contract allows delivery of imported GMO

soybean crops, which are used for soy oil and animal feed. Figure 1 shows soybean futures

prices in all four markets. There is visual evidence of strong co-movements among these

prices, and we investigate whether this co-movement is due to a stable long-run price

relationship and examine the price lead-lag relationship across the four markets.

This paper is the first to investigate the long-run lead-lag relationship among the United

States, Brazilian, and Chinese markets, and explores the seasonal relationship between U.S.

and Brazilian futures markets and the influence the Globex overnight trading platform in

Chao Li is a PhD candidate at Department of Economics, Iowa State University, Ames, Iowa. Dermot J. Hayes

is a Pioneer Chair in Agribusiness at Department of Economics, Iowa State University, Ames, Iowa. We thank

the editor and an anonymous referee for suggestions that improved the analysis and presentation of the paper.

*Correspondence author, Department of Economics, Iowa State University, Heady Hall, Office 80-B, Ames,

IA 50011. Tel: 515-294-8288, Fax: 515-294-0221, e-mail: chaoli@iastate.edu

Received November 2015; Accepted April 2016

The Journal of Futures Markets, Vol. 37, No. 1, 52–70 (2017)

© 2016 Wiley Periodicals, Inc.

Published online 3 June 2016 in Wiley Online Library (wileyonlinelibrary.com).

DOI: 10.1002/fut.21794

the U.S. has on the daytime return of soybean futures in China.

1

Overnight trading in

the U.S. and daytime trading in China occur contemporaneously, so we apply an

autoregressive distributed-lag model to address this problem.

The paper is organized as follows: Section 1 provides a brief literature review. Section 2

describes the methodology that characterizes the price lead-lag relationships using linear and

non-linear co-integration. Section 3 describes the data. Section 4 exhibits and explains

empirical results of co-integration and demonstrates two short-run causality relationships.

Section 5 presents conclusions.

1.1. Previous Work

Granger (1981) introduced the most widely used methodology to study long-run price

causality co-integration. He showed that two variables may have a long-run equilibrium

relationship even if they are non-stationary. Engle and Granger (1987) extended this concept

and showed that co-integrated variables can be represented by a vector error correction model

(VECM) and provided test methodology for this framework. Balke and Fomby (1997)

introduced the threshold concept to explain possible non-linear long-run equilibrium

relationships. Hansen and Seo (2002) and Seo (2006) provided two methods to test a

threshold and a method to estimate the parameters of a threshold vector error correction

model (TVECM).

Wahab and Lashgari (1993) and Ghosh (1993) investigated the forecasting power of the

S&P 500 index spot and futures prices changes using co-integration. Their results indicated a

stable long-run equilibrium relationship between the index and its futures price. Chu, Hsieh,

and Tse (1999) investigated the price discovery function in three S&P 500 index markets: the

spot index, index futures and S&P Depositary Receipts markets. They found that the three

price series are a co-integrated system with one long-run stochastic trend and the futures

market serves the dominant price discovery function when the common stochastic trend is

decomposed. Martens, Kofman, and Vorst (1998) applied a threshold error correction model

to study index-futures arbitrage and found that the impact of futures market on the spot

market is larger when the mispricing error is negative and that the impact of the mispricing

error increases with the magnitude of that error.

FIGURE 1

Soybean Futures Prices in the U.S., Brazil, and China from 2005 to 2015

1

Globex is the electronic trading system in the United States.

A Threshold Co-Integration Approach 53