China's effect onworld soybean prices

China's

China's

China's effect

China's

effect

effect

effect on

on

on

on world

world soybean

world

world

soybean

soybean

soybean prices

prices

prices

prices

MSc Economics track International Economics and Globalization

Name: Jingke Li Supervisor: Dr. N. Leefmans

Student number: 10828095 Second reader: Dr. Kostas Mavromatis

Email: lijingke102683@163.com

Abstract Abstract Abstract Abstract

There is often a debate about the effects of rising food demand from China and India on world food inflation, but most papers use descriptive analysis rather than empirical methodology. In this thesis, a vector error correction model (VECM) has been applied to illustrate China's impact on world soybean prices with annual data from 1980-2010 and 1996-2010. Explanatory variables are China's GDP, its population, soybean imports volume and the exchange rate of Chinese RMB per dollar. The results indicate that China's population for sample 1980-2010 and its soybean imports volume and exchange rate in 1996-2010 have a long-run influence on the international soybean price.

Statement

Statement

Statement

Statement of

of

of

of Originality

Originality

Originality

Originality

This document is written by Student Jingke Li who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

Table Table

TableTable ofofofof contentscontentscontentscontents

1. 1.

1.1. IntroductionIntroductionIntroductionIntroduction ···· 4

2. 2. 2.2. LiteratureLiteratureLiteratureLiterature reviewreviewreviewreview ···· 7

2.1 Introduction ··· 7

2.2 Demand-side drivers of world food price inflation ··· 8

2.3 Circumstances of world soybean price and China's effect ··· 16

2.4 Why China's soybean imports increase dramatically ··· 19

3. 3. 3.3. DDDDataataataata andandandand methodologymethodologymethodologymethodology ··· 25

3.1 Data description ···25

3.2 Sample description ··· 27

3.3 Methodology ··· 27

1).Lag Selection Criterion ··· 29

2).Augmented Dicker-Fuller Tests (ADF) ··· 30

3).Johansen and Juselius Cointegration Test ··· 32

4).Vector Error Correction Model (VECM) Estimation ··· 35

4. 4. 4.4. ResultsResultsResultsResults andandandand assessmentassessmentassessmentassessment ··· 37

4.1 Sample 1980-2010 ···38

4.2 Sample 1996-2010 ···39

4.3 Discussion: limitation ··· 41

5. 5. 5.5. ConclusionsConclusionsConclusionsConclusions ··· 43

6. 6. 6.6. ReferenceReferenceReferenceReference ··· 45

Lists Lists

ListsLists ofofofof figuresfiguresfiguresfigures

1.1 The volume of main food imports of China ··· 5

1.2 Soybean prices: US dollars per metric ton ··· 6

2.1 China GDP per Capita at Constant Prices (National currency k) ··· 10

2.2 Population, Total - China (Billion) ··· 10

2.3 Factors contributing to higher food commodity food prices ··· 12

2.4 China all oils: production/consumption: (1000 mt) ··· 13

2.5 China and India consumption growth rates per capita and population growth ··· 14

2.6 China's imports of food and agricultural products ··· 17

2.7 China's per capita meat consumption projection ··· 21

2.8 China's soybean imports projection ··· 21

3.1 Selection-order criteria for sample 1980-2010 ··· 30

3.2 Selection-order criteria for sample 1996-2010 ··· 30

3.3 ADF results: 1980-2010 ··· 31

1.

1.

1.

1. Introduction

Introduction

Introduction

Introduction

During 2006-2007, the world food price surged, which drew plenty of economists' attention. They began to identify what factors result in this fastly increased price and to give a large number of explanations. Many of them cited the rising demand from emerging countries such as China and India as key drivers of the high food price. Wolf (2008) stated that strong income growth by China, India, and other emerging economies, which boosted demand for food commodities, was the key factor behind the post-2007 increases in food prices. Based on the facts that China's economic growth is fast and stable, and that even with the One-Child Policy since they have a large population base, their population still increases faster than other countries', these conclusions seem plausible. Moreover, since 1979, China began economic reforms, liberalized foreign trade and eliminated trade barriers step by step, which enlarged China's impact on the world commodity market. With the industrialization and urbanization, most of the citizens now try to improve their living standards and shift their food consumption patterns, which exactly increases the food demand because more and more people move to modern cities and crops producers become less and less. However, several economists (C. Abbott 2008, G. Tadesse 2013) did empirical analyses of many factors including rising food demand from developing countries, and pointed out that the latter has nothing to do with the surging food price. Chinese National Bureau of Statistics provided China's imports data of four major grains (figure 1.1) from 1990 to 2014: cereals and cereals flour, wheat, paddy and rice, and soybeans. It is clear that only the volume of soybean imports continue to rise since the 1990s, and after 2005, China became the largest soybean importing country. Therefore, this thesis will not identify the China's effect on the whole world food prices but concentrate on its effects on the world soybean price.

Figure 1.1. The volume of main food imports of China

Source: National Bureau of Statistics

To begin with, it is necessary to find what factors drive up food demand of China and then why its soybean imports have increased dramatically. China's rising real GDP, its population, and the Chinese RMB appreciation are usually regarded as the drivers of its rising food demand. In China, with the development of the economy, living standards of citizens gradually improved, which boosted the demand for food, including soybeans. However, due to an agricultural self-sufficiency policy, the imports of cereals, wheat and rice showed weak growth, but the circumstance of soybeans is distinct. There are a large number of papers that illustrate the reasons why the imports of soybeans have risen for these years in China. Conclusively, the rising imports of soybeans might result from four reasons: growth of demand, shortage of supply, policy reforms and the exchange rate effect. Firstly, the increase of GDP and population leads to a higher standard of living, which upgrades the diet pattern and raises the demand of animal protein. Soybeans can be crushed to soybean oil for daily cooking and the residual soybean meal is the main feed for animals. Secondly, in China, the acreage of soybeans is small and technology of planting is outdated, while the fast process of urbanization and uneven economic developments worsen this situation since many young adults leave the countryside forcedly or willingly which leads to the lack of rural labor force and thus soybean production. Thirdly, China's

0

1000

2000

3000

4000

5000

6000

7000

19

9

0

19

9

2

19

9

4

19

9

6

19

9

8

20

0

0

20

0

2

20

0

4

20

0

6

20

0

8

20

1

0

20

1

2

Year

1

0

0

0

0

t

o

n

s

Volume of Imports,

Cereals and Cereals

Flour(10000 tons)

Volume of Imports,

Wheat(10000 tons)

Volume of Imports,

Paddy and Rice

(10000 tons)

Volume of Imports,

Soybean(10000 tons)

government began to carry out the Reform and Opening Policy since 1979, which liberalized foreign trade, and after China joined the WTO in 2002, soybean tariff quotas were canceled. Finally, as soybean prices are dollar-denominated on the world market, the depreciation of US dollars due to the financial crisis and the appreciation of Chinese RMB resulting from the exchange rate system reforms, make the cost of importing soybean much lower than that of domestic production, which fuels China's demand for overseas soybeans.

Many economists try their best to find factors that drive up world food prices, and rising food demand from emerging countries, such as China and India, is often seen as an important factor. But as mentioned before, both in China and India, the government implements an agricultural self-sufficiency policy, so imports of most major grains did not grow strongly. China's soybean import is an exception. Figure 1.2 shows that the world soybean prices rose dramatically like world food prices since 2006. Therefore, the thesis focuses on soybean as one of the major food crops.

Figure 1.2 Soybean prices: US dollars per metric ton

Source: IMF cross country macroeconomic statistics

In my thesis, I use a Vector Error Correction Model (VECM) as the methodology to test for a causal long-term and short-term relationship between world soybean prices and explanatory variables, which include China's GDP, its population, its soybean imports volume and the exchange rate of Chinese RMB per dollar. My dependent variable is world soybean prices from IMF cross country macroeconomic statistics.

Owing to data availability, the period will cover 1980 to 2010. Moreover, the data are divided into two samples: sample 1980-2010 and sample 1996-2010, in both of which I will use all four explanatory variables. For full sample 1980-2010, because before 1996, China's soybean imports remained at a stable and low level, and the exchange rate system was fixed, these two variables are expected to have no impact on the world soybean price, while for the sample 1996-2010, China's soybean imports and exchange rate are expected to affect the world soybean prices.

The remainder of this thesis is structured as follows: Chapter 2 summarizes relevant results of recent literature that identifies demand-side drivers of world food inflation, that illustrates world soybean market and that explains the reasons why China's soybean imports has increased dramatically. Chapter 3 introduces the data and methodology that will be used in the empirical analysis and Chapter 4 presents the results of vector error correction model (VECM). Chapter 5 draws a conclusion and gives implications of obtained results.

2.

2.

2.

2. Literature

Literature

Literature

Literature review

review

review

review

2.1

2.1

2.1

2.1 Introduction

Introduction

Introduction

Introduction

Plenty of economists investigated the factors that drive up the world food price, mainly for major crops and concluded that a large number of factors, such as economic growth of emerging countries (rising food demand), the oil price, the use of biofuel and world population growth, affect the world food price. There has always been a debate about the effects of increasing food demand from emerging countries on the world food price. The volume of imports can reflect a country's demand for food commodities. As mentioned previously, soybean is one of the major food crops. The purpose of this section is to provide an overview of the literature that deals with the

factors that affect world food prices and the topics of the world soybean market. Articles that analyze both the drivers of overall food prices and soybean circumstances will be selected. The academic articles selected in this chapter can be divided into three categories. The first category includes what factors lead to the growing food demand of emerging countries, especially in China and whether they might raise the world food price. The second category presents papers that illustrate the circumstances of the world soybean price. The third category concentrates on explaining why China's soybean imports increased dramatically since 1996. The number of empirical analyses of China's effect is very small, most academic articles are based on descriptive analyses and only a few of them use empirical methodology. It is worthwhile noticing that the main drivers of food price inflation are divided into demand-side- , supply-side- and other factors. In general, demand-side-factors include economic and population growth, oil prices, biofuel, and speculation, supply-side-factors include weather shocks, food reserves, trade restrictions and investment, and other factors include the US dollar depreciation, the interest rate, money supply, and volatility. In this thesis, I focus more on demand-side factors because China's soybean demand has risen dramatically, which might result from China's GDP and population growth and lead to world soybean price inflation. Meanwhile, since China's government changed the fixed exchange rate system to a managed floating exchange rate system since 1994, and China and America are the most essential traders on the world soybean market, the exchange rate of US dollar against China's RMB is taken into consideration.

2.2 2.2

2.22.2 Demand-sideDemand-sideDemand-sideDemand-side driversdriversdriversdrivers ofofofof worldworld foodworldworldfoodfoodfood pricepricepriceprice inflationinflationinflationinflation

Food demand is consumption of food that is either purchased in the market or home-produced. According to economic theory, the demand growth of food results in higher food prices due to the supply-and-demand model. As my explanatory variables are China's soybean imports, China's GDP and population growth, and the exchange rate, the relationship between those factors and the rising food demand from China are

studied. Then, whether China's food demand growth resulted in world food price inflation will be discussed.

The first factor that raises China's food demand is economic growth, GDP per capita growth enhances purchasing the power of households and thus drives up food demand. However, there are multiple channels between GDP growth and food price inflation, which makes it difficult to use OLS or TSLS to capture price effects. For example, economic growth of emerging countries might result in food price inflation through increased demand for crude oil, as their economy is growing and thus they are expected to use more energy to expand. This crude oil prices increase might affect food price, through increased demand for agricultural crops used for energy. On the other hand, GDP growth may also stimulate the development of technologies, and increase agricultural capital stock through the improvements of farm machinery, which might be positively correlated with food production efficiency and thus decrease food prices. Von Braun (2012) identified the endogenous and exogenous drivers of world food prices, and supported that the rising income in China and India, as an intermediate cause, led to growing food demand because the remarkable income growth (figure 2.1) increased the consumption per capita and changed the consumption patterns in emerging countries. Moreover, the effects of income distribution have been taken into consideration. Xavier Cirera (2010) compared food demand forecasting models and used Engal's law to illustrate the relationship between income distribution and food demand and confirmed that the income and income distribution affect food demand.

Figure 2.1. China GDP per Capita at Constant Prices (National currency k)

Source: IMF Cross Country Macroeconomic statistics

The second factor is population growth, which has been regarded as a key driver of rising food demand in emerging countries. It is obvious that food demand rises with population growth, and in some circumstances, GDP growth can also transmit to food price inflation through population growth. For instance, between 1970 and 2010, China's population grew from 0.8 billion to 1.37 billion (Figure 2.2). Even though the government carried out birth-control policies, the one-child policy particularly, which led to a downward trend of population growth rate, the population base remains large, which results in rising food demand (Patrick Hamshere 2014, A. Schneider 2011).

Figure 2.2 Population, Total - China (Billion)

Source: United Nations database

reforms of the exchange rate system and the US dollar depreciation. Most commodity prices are dominated in US dollars, especially in world soybean market. Countries with a floating exchange rate will increase their demand for commodities change everywhere in US dollars due to the dollar depreciation, which may contribute to world food prices inflation. Hua (1998) made use of a vector error correction model (VECM) using quarterly data from 1970/2 to 1993/3 and confirmed that there is a causal long-term relationship between food prices and the exchange rate. Awokuse (2005) chose vector autoregressive (VAR) model and found that the exchange rate was the primary macroeconomic factor of commodity price changes.

Several economists regard the increasing food demand from emerging countries as a reason that leads to surging food prices since 2006. But few of them did empirical analyses that discuss the quantitative impacts of rising food demand on price inflation, which might be because it is difficult to find a variable to measure the food demand of different countries and to model the interconnectedness with a large number of variables. Therefore, only descriptive analysis are given. It is worthwhile noticing that, in these descriptive analyses, they mostly discussed the situations of major crops, such as cereals, wheat, rice, maize and soybeans, not for all food.

The number of papers that only focus on effects of rising food demand from emerging countries is small as food price inflation is caused by several factors. Many economists identified many factors that raised the world food price and the increasing food demand from China and India is only a small part of their topics. Trostle (2008) explored the factors contributing to the most recent rise in food commodity prices, divided them into short- and long-term drivers from both demand and supply side, and stated that, the growing demand from emerging countries due to their population and income growth, affected the world food prices since the 1990s (Figure 2.3).

Figure 2.3. Factors contributing to higher food commodity food prices.

Source: Trostle (2008)

C.Abbot ed al. (2008 &2011) also identified the drivers of world food prices, and is cited many times. They reviewed 25 recent studies and reports, and provided as an complete an assessment as possible of the major forces behind the dramatic increases in commodity prices. One focus is whether growing food demand and dietary transition in China and India result in global food consumption increasing faster than production and they chose stock-to-use ratio1 _{to explain, which is a convenient}

measure of supply and demand interrelationships of commodities. Comparing the stock-to-use ratio of corn, wheat, and rice of China and the amount of rice consumption and production in China, the authors concluded that although the China's stock-to-use ratio decreased dramatically, the stock changes were internal decisions and China is not an important trader on world markets due to its agricultural policies of self-sufficiency. Therefore, growing food demand from China did not affect world food prices for most agricultural commodities. But there is one exception: the consumption of food oil (Figure 2.4), which is mostly used during Chinese daily life, rose sharply since mid-1980s and the amount of imports showed an upward trend.

1 _{Stock-to-use ratio: this ratio indicates the level of carryover stock for any given commodity as a percentage of}

the total demand or use. The mathematical formula is (Beginning stock + Total Production - Total Use) / Total use. After consolidating the upper portion of the relationship, this can be simplified to: Carryover / Total use = stock-to-use ration. The beginning stocks represent the previous year's ending or carryout inventories. The total production represents the total grain produced in a given year, while the total usage is the sum of all end uses in which the stock of grain has been consumed.

More specific soybean circumstances will be discussed in the next section.

Figure 2.4. China all food oils: production/consumption: ( 1000 metric tons)

Source: PSD, FAS, USDA

The EU commission (2011) also considered whether there has been an acceleration in demand growth for agricultural products in their analysis to assess the extent to which developments in agricultural prices are demand driven, and conducted more detailed analysis of China and India. After calculating China's and India's consumption growth rates per capita for consecutive periods (Figure 2.4), it is obvious that although absolute levels of demand are higher for all products analyzed, only dairy products and vegetable oils show an upward trend since 1985, while the demand growth for feed grains and meat in China fell sharply since 1985. The EC stated that the global consumption growth slows down due to lower world population growth, even in emerging countries like China and India, so the consumption growth from developing countries might not be a major driver of food price inflation. They used consumption data to describe the food demand from China and India, and according to the above conclusion, only dairy and vegetable oil prices should be affected by rising demand growth from emerging countries, while there is no direct effect on

other agricultural commodities price.

Figure 2.5: China and India consumption growth rates per capita and population growth

Source: Authors' calculation based on USDA, FAO and UN data

Headey and Fan (2008) supported C.Abbot's conclusion that China and India are self-sufficient in major grains and have not increased imports of staple foods, which has no direct effect on food price inflation. Similarly, Baffes and Haniotis (2010) compared the demand in China for major agricultural products between 1997-2002 and 2003-08 and concluded that rising demand for food commodities in China did not affect world food prices. The demand in 2003-08 for meat, palm oil, soybeans, rice and maize decreased, while only wheat demand increased by 0.9 percentage, compared to 1997-2002.

Even though it is hard to measure China's impact on the world food price, several economists did impressive empirical analyses to capture the quantitative general price effects. Usually they identified a large number of drivers of food prices, in which several factors such as income growth or stock-to-use ratio would be used indirectly to reflect growing food demand from emerging countries. Others tried to find a method directly to describe the relationship between increasing food demand from China and world food prices through gravity models developed by Anderson and van

Wincoop, which focused on all food products.

Baffes and Dennis (2013) applied a reduced-form econometric model and examined the relative contribution of various sector level and macroeconomic drivers to price changes of five food commodities (maize, wheat, rice, soybeans, and palm oil). Those drivers include one supply-side variable (energy prices), three macroeconomic indicators (exchange rate, interest rate, and inflation), income2 _of

the global, low and middle income countries and low income countries on the demand side, and a driver reflecting market fundamentals (stocks and consumption expressed as a single stocks-to-use ratio variable). They chose the income of China and India as a variable to reflect the food demand from emerging countries, especially from China and India, because most economic theories support that demand increases with income. They applied the following reduced-form price-determination model, which is based on equating aggregate demand to the supply of a commodity, and then expressing the equilibrium price as a function of sectoral and macroeconomic fundamentals:

log(Pti

) =β0+β1log(S/Ut-1) +β2log(PtOIL

) +β3log(XRt) +β4 log(Rt) +β5log(GDPt) +β6log(MUVt) +εt

where Pti _{denotes the nominal price of commodity i (i = maize, wheat, rice, soybeans,}

and palm oil). S/Ut-1 denotes the lagged stock-to-use ratio (use includes human, animal, and industrial consumption), PtOIL_{is the price of crude oil, XRtis the exchange}

rate, Rt denotes the interest rate, GDPt denotes gross domestic product (GDP) of the global, low and middle income countries and low income countries respectively, and MUVtrepresent a price index of manufacturing exports. The βs are parameters to be estimated, andεtis the error term.

2 _{They used a total of six measures of GDP: Global, low and middle income countries, low income countries, each}

Using 1960-2012 annual data from world bank's database, and the ADF (Augmented Dickey-Fuller) and the PP (Phillips-Perron) tests, Baffes concluded that there is no contemporaneous effect of income of the world on food prices (maize is an exception, but the parameter estimate is negative) because the estimated coefficients of income are not significantly different from zero. But income might affect prices indirectly through the lagged stock-to-use ratio, as higher income results in higher consumption. So the insignificant parameter estimate implies that, if income affect prices, it does so with lags. Therefore, they found that the high income growth, which led to higher food demand from China and India, has no significant direct impact on the inflationary world food prices. Moreover, the grain consumption by emerging economies has not experienced high growth rates that are comparable to their income growth rates since the 1990s, which proved above conclusions. The results on the effect of the exchange rate are mixed while they are consistent with expectations. Exchange rate movement affects rice the most, and has a moderate impact on soybean and wheat, and no impact on palm oil and maize, which means food commodity prices respond to exchange rate movement in a mixed manner.

2.3

2.3

2.3

2.3 Circumstances

Circumstances

Circumstances

Circumstances of

of

of

of world

world

world

world soybean

soybean price

soybean

soybean

price

price

price and

and

and

and China's

China's

China's

China's effect

effect

effect

effect

Soybeans are the world's largest source of animal protein feed and the second largest source of vegetable oil. Soybean imports growth of China is an important symptom of rising food demand. Gale and Huang (2015) analyzed China's recent situations as a major agricultural importer. With China's gradual liberalization, rising living standards, and changing consumption patterns, its commodity imports have risen dramatically in recent years. In particular, soybeans and other oilseeds represent almost half of China's agricultural import value (Figure 2.6). Currently, according to CME Group (2010), the United States is the largest producer of soybean oil as its production accounts for more than 23 percent of the world's production, while south American production, especially in Argentina and Brazil, is growing quickly. China

and India are the two largest importers since their importing volumes are over 45 percent of the world imports. The authors identified the fundamental price drivers of soybeans and pointed out that the primary driver has been growing demand over the past decades. To be more specific, world population and GDP growth led to increased food demand. Exchange rate exposure also has been a key factor in the demand. If a country has a strong currency, they will consume more because they can buy these commodities at a lower price.

Figure 2.5: China's imports of food and agricultural products

Source: USDA, Economic Research Service analysis of China’s customs statistics reported by the Global Trade Atlas (2014).

Soybean is one of the major crops so academic articles that illustrate the drivers of the whole food price inflation include soybeans. C.Abbot (2011) pointed out the importance of China's importing effect on the world soybean market. One of key issues behind high agricultural commodity prices is big and persistent demand shocks: Chinese soybeans. The continued surge in China's soybean use for livestock feed and human vegetable oil consumption, led to world oilseed growth. After 1997-1998,

China abandoned a soybean self-sufficiency policy, with continued increases in consumption, which have doubled Chinese soybean import purchases since 2005-2006. Headey and Fan (2008) discussed the strengths and weaknesses of many explanations of the 2005-2008 global food crisis, including the above conclusions. They stated that the oil consumption growth from China could partly explain China's rising demand for oilseeds, especially the surging import amount of soybeans since 1990.

Similarly, several economists made use of econometric methodology to identify the determinants of international soybean prices. The mentioned possible causes of soybean price inflation are: the increased crude oil price, dollar depreciation, economic growth of China and India, and speculation bubble. Wiston A. Risso (2011) focused on China's effect and applied the Johansen and Joselius cointegration methodology to test a long-run relationship among international soybean prices, the Yuan-dollar exchange rate, and Chinese real GDP. Based on the fact that the US is the main producer of soybean and China is the main soybean importer, they play an important role in the formation of the international soybean prices. Therefore, a simple model is developed, in which the soybean supply is produced in the US while the demand is generated in China. Their Vector Auto Regressive (VAR) model is as follows:

)

,

,

(

p

e

y

X =

Where p is the US soybean prices, e is the exchange rate yuan per dollar, and y is

Chinese real GDP, all variables are expressed in logs.

As a cointegration test only can find a long-run relationship among these three variables, a Vector Error Correction model (VECM) is necessary to model the short-run dynamic. The form of the model is:

t i t k i i i t t

X

X

X

=

µ

+

π

+

Γ

∆

+

ε

∆

∑

Where X is the vector defined in the previous model, and µ is a constant term vector.

The matrix π implies the long-run relationship between the X variables, while ε are assumed to be independent error terms.

Data comprise monthly US soybean prices, the exchange rate Chinese Yuan per dollar, and Chinese real GDP from January 1981 to November 2010. At first, they applied unit root tests to confirm the stationarity of the time series. After confirming the non-stationarity, they used the Johansen cointegration test and got cointegration equation: [14

.

52

0

.

57

0

.

44

4

e

y

P

=

−

+

They find a cointegration relationship among these variables: the US soybean price depends negatively on the exchange rate Chinese Yuan per dollar and positively on Chinese real GDP. From the results of the ARIMA model, the weakness of the dollar with respect to yuan has the largest effect even though Chinese economic growth is essential to explaining world soybean price inflation.

2.4

2.4

2.4

2.4 Why

Why

Why

Why China's

China's

China's

China's soybean

soybean

soybean imports

soybean

imports

imports

imports increase

increase

increase

increase dramatically

dramatically

dramatically

dramatically

As mentioned previously, China's volume of soybean imports has increased dramatically since 1996. To be more specific, between 1995 and 2009, China's imports of soybeans grew by 1.459 billion bushels (from 0.029 to 1.488 billion). Before using econometric methodology to identify the cointegration relationship among world soybean prices and these explanatory variables, it is necessary to

explain why soybean imports of China continued to rise since 1996. The number of academic articles is large and in general, there are four reasons: rise of demand, shortage of supply, policy reforms and the exchange rate effect, which are highly interconnected.

Firstly, according to most research conclusions, the surge of soybean imports results from China's fast economic growth, which belongs to demand-side factor. In China, importing soybeans are used for soybean crushing industry to get soybean oil and soybean meal. For Chinese, soybean oil is important in daily cooking and soybean meal is main feed of livestock industry. As their income rises, Chinese consumers are improving their living standard and changing their diets from major grains to more meat and dairy products. Past studies have indicated that the demand for poultry, meat, fish and dairy products is responsive to income growth. Therefore, rapid income growth has changed the structure of Chinese food expenditure and their consumption pattern, which reflects the livestock boom. J. Hansen and F.Gale (2014) stated that in the next decade, China will continue raising meat demand and imports of feed due to its large population, fast economic growth and anticipated dietary change. China is moving into a new stage of economic development, which will promote continued urbanization and living standards of the whole population. China produces nearly all of its own meat. They pointed out that since about three kilograms of feed are needed to produce one kilogram of meat, with rising population of animals, feeding might be a large challenge. Meanwhile, with the process of urbanization, both the number of urban citizens and the consumption for meat show an upward trend, which boost the demand of feed grains, especially, of soybeans. The per capita meat consumption is expected to rise (Figure 2.7). USDA anticipates that China's soybean imports for animal feed will continue to increase in pace with its robust economic growth and meat consumption, and the imports volume might reach over 70 percent of global soybean imports by 2023/24 (Figure 2.8), which means China will still be the dominant global soybean importer.

Figure 2.7. China's per capital meat consumption projection

Source: USDA production, supply and distribution database and projection

Source: USDA production, supply and distribution database and projections

The second driver is the shortage of supply, which is caused by plenty of factors. The reason why China's soybean crushing industry choose importing soybeans is that international soybean price is lower than domestic price. Gale and Huang (2007) illustrated China's food market structure and demand for food quantity and quality. They found that China's soybean acreage is becoming less and less with the process of urbanization and soybean production is relied on small-scale farmers with outdated technology. In China, plains that can be used for agricultural planting are becoming small, so it is difficult to apply machinery for production. Thus, there remains small-peasant economy in most rural areas, with low productivity and no economies of scale, which made raising per unit area yield challenging. Moreover, rapid economic growth and urbanization worsen the soybean supply. Economic developments in China are uneven owing to government policies. The southeast of China is much more developed than other areas while agricultural areas are located in the southwest and northeast of China. Hansen and Gale (2014) pointed out that in the last ten years, economic growth has absorbed rural labor and labor scarcity in the countryside plays an essential role of domestic soybean supply shortage. In order to earn more and get a higher living standard, a large number of young adults choose to leave the countryside and work in big cities. In many villages, only the old and children are left and engaged in agricultural production, which lowers soybean productivity and drives up the costs of production. Meanwhile, because lots of agricultural lands are converted to constructions of cities and agricultural production cannot be mechanized, domestic soybean yield is less likely to satisfy the rising demand. It is necessary to pay attention to another crop - corn- as Chinese farmers use both corn and soybean meal as animal feed. China's soybeans are typically grown in the North East provinces, accounting for forty percentage of total national production in 2010 but several farmers there switched from soybean to corn production because corn could bring more value than beans. Considering climate change, the environment might be more favorable to growing corn. Hoffman and Griffin (2012) studied this

issue and showed that while corn expansion in the North East of China reduced.

The third driver is Chinese government policies. Since 1996, Chinese government started to reduce tariff on soybeans. At that time, the tariff rate was 3 percent. Guo and Wang (2011) identified what factors led to China's soybean imports growth, making use of Liunemann's trade gravity model and collected time series data that covered the period 1991 to 2010. The regression equation is:

ijt t it it it jt ijt it it ijt

e

F

F

R

SOC

SMC

AY

AY

D

POP

GDP

Y

+

+

+

+

+

+

+

+

=

)

*

ln(

)

/

ln(

)

ln(

)

ln(

)

ln(

)

ln(

α

α

α

α

α

α

α

α

α

Where i is China and j are countries that export soybeans to China, which are America,

Brazil and Argentina respectively. Yijtis soybean trade flow between country i and j. F1 is dummy variable that equals to one if China's domestic market is open to the world, and to zero otherwise. F2 is another dummy variable that equals to one if Chinese government carried out GMO policy, and to zero otherwise.eijtis error term.

After regressing all data describe above with Eviews, they concluded that China's GDP, China's population and exchange rate could affect the soybean imports of China. There is a relationship between China's policy that canceled tariff quota of soybeans and opened domestic market to the world and China's rising soybean imports. They explained that after 2000, China's government abandoned tariff quota and instead imposed a common external tariff of soybeans, only 3 percent, which encouraged soybean imports because the costs of importing is lower than domestic production costs. Moreover, domestic demand continue to rise with the development of economy but domestic production increases very slowly. This huge demand gap can only be satisfied by imports.

The final factor is exchange rate changes, which resulted from two reasons. On the one hand, with rising foreign exchange rate reserves and rapid growth of economy,

China's government decided to carry out exchange rate reforms. In 1996, RMB exchange rate was no longer pegged to the dollar, and instead became managed floating based on market supply and demand. This regime was improved more flexibly in 2005. Due to China's exchange rate regime reforms, the RMB has started and continued to appreciate seriously. On the other hand, the US dollar has depreciated because of the pressure from international market, such as appreciation of RMB and growing overseas US dollar reserves held by developing countries, and from domestic market, such as rising financial deficits. According to basic economic theory, the appreciation of the RMB increases China's purchasing power because exchange variations are expected to change the relative prices of importable goods. Since international soybean prices are dollar-denominated, the appreciation of the RMB makes the costs of importing soybean much lower for the domestic soybean crushing industry. Mao and Liu (2012) used econometric methodology to analyze the effect of RMB appreciation on China's soybean importing trade and constructed following model: t t t t C REER Y u M = + ln + ln + ln α β

Where Mt is the importing volume of soybeans, REERt is China's real exchange rate,

Ytis China's real GDP growth and C is a constant.

The monthly data covered 2005 July to 2010 December. After using ADF unit root test to confirm non-stationarity, Johansen and Juselius Co-integration test, and Granger's Vector Error Correction Model (VECM), they got a conclusion that there is a positive relationship between China's RMB appreciation and the volume of importing soybeans. If RMB appreciate one percent, the volume of soybean imports will increase 1.63%.

3.

3.

3.

3. Data

Data

Data

Data and

and

and

and methodology

methodology

methodology

methodology

This thesis tries to identify China's effect on international soybean prices, thus this chapter will use empirical methodology to test the relationship between the world soybean prices, and the explanatory variables: China's real GDP, China's population, China's soybean imports, and exchange rate of RMB per dollar. The research includes three parts: data description, sample description, and empirical methodology, Vector Error Correction Model specifically. In the following sections, each will be described in turn.

3.1 3.1

3.13.1 DataDataDataData descriptiondescriptiondescriptiondescription

The empirical analysis is going to examine co-integration relationship between explanatory variables and international soybean prices, using annual data from 1980-2010 and one sub-sample 1996-2010. The sub-sample has been specified in accordance to the fact that China's soybean imports started to increase dramatically since 1996 and that in 1995 China's government abandoned fixed exchange rate system and carried out a managed floating regime. Therefore, since 1996, soybean imports and exchange rate experienced a huge change, which might have greater effects on international soybean prices. The annual data from both 1980-2010 and 1996-2010 both include China's real GDP and its population, because in 1979, China's government decided to put reform and opening-up policy thus its GDP and trade imports began to increase. Moreover, in 1982, China's government implemented another state policy, family planning, which is also called one-child policy, and China's population growth showed a downward trend. Although two samples also include China's soybean imports and the exchange rate of the RMB per dollar as explanatory variables, the sample 1996-2010 will illustrate more about effects of two factors on the world soybean price.

Existing academic articles usually identify plenty of drivers of world food price inflation and concentrate on specific staple food commodities, such as wheat. However, this thesis is focusing on China's effect on world soybean prices. On the other hand, conclusions might not be able to fully explain all drivers of world soybean price inflation, which means that the object of this thesis cannot paint the whole picture.

The dependent variable included in the analysis is SOYP (soybean price index: US

dollar per metric ton) from IMF Cross Country Macroeconomic Statistics. Monthly data are considered better than annual data, but it is difficult to find monthly data of other variables especially for exchange rate and soybean imports. As a result, annual data are used.

Chapter 2 illustrated factors of China that might affect international soybean prices. Th frist factor is real GDP GDP, which comes from Measuring Worth Dataset, in

thousands of 2005 Chinese RMB. It seems that most economists choose real GDP rather than nominal GDP as a variable while they estimate its effect on world food prices, for example, Wiston A. Risso (2011) mentioned in previous chapter. Since real GDP is a macroeconomic measure of the value of economy output, adjusted for price changes, it might be more proper for estimating. The second factor is its population

POP, and original data also come from the Measuring Worth Dataset. The third factor

is China's soybean import volume IMP, whose annual data are from Earth Policy

Institute, million tons. They are complied by Earth Policy Institute from US Department of Agriculture, Production, Supply and Distribution, electronic database. The forth factor is the exchange rate of Chinese RMB per dollar, and it is from Federal Reserve Economic Data.

All variables above are used in the form of logs so the regression coefficients need to be interpreted as estimates of elasticities, instead of changes in units across time. For instance, gross domestic product (GDP), exhibits growth that is approximately

exponential, which means the series is likely to grow by a certain percentage per year on average over the long run. If so, the logarithm of the series grows approximately linearly. This adjustment could make comparing the results more conveniently. Besides, in a VECM, since the dataset of these variables proved to be non-stationary, in VECM, they are converted to their first differences.

3.2 3.2

3.23.2 SampleSampleSampleSample descriptiondescriptiondescriptiondescription

This section gives a description of two samples. As China's GDP and population continue to increase since 1980, they are included in both samples. Although China's soybean imports remain at a low level and China's government implemented fixed exchange rate regime from 1980 to 1995, they are included in both samples.

The first sample, from 1980 to 2010, concentrates on effects of China's real GDP

GDP and its population POPG on international soybean price SOYP. The continuous

rise of its GDP and population is expected to have a significant effect on the world soybean prices, while China's soybean imports and the exchange rate of Chinese RMB is expected to have no impact on the world soybean. The focus of subsample, from 1996 to 2010, is not only impact of China's GDP and population, but also the impact of its exchange rate EXrate and soybean imports volume IMP on international

soybean price SOYP.

3.3 3.3

3.33.3 MethodologyMethodologyMethodologyMethodology

As described before, the relevant articles on this topic are likely to choose four different approaches to measure China's effect on world food prices: 1. Descriptive analysis. Because it is hard to find a variable directly to reflect China's rising food demand, most of economists use descriptive analysis, for example, C.Abbot (2008 &2011) and EU commission (2011). 2. Least squares estimation method, specifically, two-stage least squares (TSLS) or three-stage least squares (3SLS). They carefully

make a choice of instrumental variables to eliminate endogeneity problem. 3. Granger causality tests. These tests can examine a causal relationship between variables, see J. Baffes (2013). 4. Vector Error Correction Model (VECM). According to discussion in Chapter 2, this method is the most popular and widely used, for example, T. Guo and Y. Wang (2011).

The estimation methodology of TSLS or 3SLS model, depends on the chosen instruments thus if instrumental variable cannot capture all interdependencies between these explanatory variables, there might be a risk of inconsistency. Based on the discussion presented in Chapter 2, there is a broad range of interdependencies among China's real GDP, its population, exchange rate of RMB per dollar, and soybean imports volume so it is very difficult to find instrumental variables to capture all of those interdependencies. Granger causality analysis can be useful only when referring to the existence of a relationship between two variables, because it checks for relevant bivariate information. In other words, it does not establish a direct causal relationship between the dependent variable y and explanatory variable x. Conclusively, a VECM

seems to be the best approach to examine China's effect on world soybean price. In principle, VECM is a theoretically-driven method that can be used to estimate not only long-term effects, but also short-run effects of one time series on another, with the error correction term. After deviations that are resulted from shocks, error correction term measures short-run convergency of the explanatory variables to the long-term equilibrium mean. Since the explanatory variables that reflect China's impact are partly linked to each other, VECM eliminates risk of endogeneity and corrects for spurious correlation.

In order to find long-term and short-term relationships between explanatory variables and international soybean price, the following empirical analysis makes use of VEC model estimation methodology. The pattern of the two samples that this analysis follows is the same, which can be divided into four steps. Firstly, lag selection criterion is performed to choose the most proper lags for the analysis.

Secondly, augmented Dickey-Fuller (ADF) tests are used to check for stationarity of all variables because the logarithmic level values should be non-stationary and stationary after differenced once. Thirdly, Johansen cointegration method tests existence and number of relationships between co-integrated variables. Finally, VECM estimations are employed in order to not only specify the variables that have a statistically significant long-term impact on international soybean prices, but also quantify the influence of these variables on international soybean price inflation. All steps described above will be elaborated in the following.

1) 1)

1)1) LagLagLagLag SelectionSelectionSelectionSelection CriterionCriterionCriterionCriterion

Lags are widely used in time series data. Special terminology and notation are used to indicate future and past values of dependent variable, Y. The value of Y in the

previous period is called its first lag, and is denotedYt-1. Its jth_{lag is its value j periods}

ago, which is Yt-j. Lag length selection is important in auto-regression model, because

if the order of an estimated auto-regression is too low, some variable information contained in the more distant lagged values might be omitted. On the other hand, if it is too high, there may be more estimating coefficients than necessary, which leads to additional estimation error.

Usually, experience, knowledge and theory are the best way to determine the number of lags. However, there are information criterion procedures to help to get a proper number. To be more specific, three commonly used criterion are: the Akaike's information criterion (AIC), Schwarz's Bayesian information criterion (SBIC) and Hannan and Quinn information criterion (HQIC). Therefore, when all three agree, the lag selection is clear. But in general, it is likely to get conflicting results. When that happens, according to a paper from Center for Economic and Policy Research (CEPR), in VAR models, AIC tends to be more accurate with monthly data, SBIC works fine with any sample, and HQIC is better for quarterly data on samples over 120. However, since these are not absolute for every sample, I will combine and compare all results

to get a better one.

As reported in table 3.1 and 3.2, it is clear that for sample 1996-2010, the best lag

order is the third lag because all three, AIC, HQIC and SBIC, agree. For sample 1980-2010, the best lag order is the forth lag as all supports

Table 3.1 Selection-order criteria for sample 1980-2010

E xo g e n ou s : _c o n s E n do g e n ou s : ln S O Y P l n P O P l n G D P l n I M P E x r at e 4 33 4 . 5 7 5 1 3 6 .2 5 * 2 5 0 .0 0 0 3. 9 e - 13 * - 1 7. 0 0 5 6 * - 1 5 .5 0 7 1 * - 1 1 . 9 66 2 * 3 26 6 . 4 5 1 5 2 . 05 5 2 5 0 .0 0 1 1. 7 e - 12 - 1 3. 8 1 1 2 - 1 2 .6 6 9 5 - 9 . 9 7 16 6 2 24 0 . 4 2 3 1 1 2 .8 4 2 5 0 .0 0 0 9. 6 e - 13 - 1 3. 7 3 5 1 - 1 2 .9 5 0 2 - 1 1 . 0 95 4 1 18 4 . 0 0 1 4 1 6 .5 2 2 5 0 .0 0 0 7. 9 e - 12 - 1 1. 4 0 7 5 - 1 0 .9 7 9 4 - 9 . 9 6 76 9 0 - 24 . 2 6 0 3 6. 0 e - 06 2 .1 6 7 4 3 2 . 23 8 7 9 2 . 4 07 4 l a g L L L R d f p F P E A I C H QI C S B I C S a mp l e : 1 9 8 4 - 2 0 1 0 N u mb e r o f o b s = 2 7 S e le c t i on - o r d er c r i te r i a

Table 3.2 Selection-order criteria for sample 1996-2010

E x o g e n o u s : _ c o n s E n d o g e n o u s : l n S O Y P l n I M P l n P O P l n G D P E x r a t e 4 . . 2 5 . . . . . 3 1 6 9 8 . 9 5 1 8 9 . 9 4 * 2 5 0 . 0 0 0 . - 2 9 8 . 9 * - 3 0 0 . 1 5 4 * - 2 9 6 . 9 1 1 * 2 1 6 0 3 . 9 8 2 8 7 3 . 2 2 5 0 . 0 0 0 . - 2 8 1 . 6 3 3 - 2 8 2 . 8 8 7 - 2 7 9 . 6 4 4 1 1 6 7 . 3 9 9 1 9 4 . 9 8 2 5 0 . 0 0 0 1 . 9 e - 1 7 * - 2 4 . 9 8 1 6 - 2 5 . 6 6 5 7 - 2 3 . 8 9 6 5 0 6 9 . 9 0 7 7 5 . 2 e - 1 2 - 1 1 . 8 0 1 4 - 1 1 . 9 1 5 4 - 1 1 . 6 2 0 5 l a g L L L R d f p F P E A I C H Q I C S B I C S a m p l e : 2 0 0 0 - 2 0 1 0 N u m b e r o f o b s = 1 1 S e l e c t i o n - o r d e r c r i t e r i a 2) 2)

2)2) AugmentedAugmentedAugmentedAugmented Dicker-FullerDicker-FullerDicker-FullerDicker-Fuller TestsTestsTestsTests (ADF)(ADF)(ADF)(ADF)

The Augmented Dicker-Fuller test checks for stationarity, which is the most commonly used test in practice and is one the most reliable. To begin with, Dickey-Fuller test can be applied in the autoregressive model. At first, a simple AR(1) - model is considered as an example:

t t t

t

X

Y

Y

=

'

δ

+

β

+

ε

where Xt' is an optional exogenous regressor, including a constant, or a constant and a

trend. δ and β are parameters to be estimated. The error term εt is assumed to be a white noise. H0: |β| = 1 vs. H1: |β| < 1. If |β| = 1, Yt is nonstationary and contains a stochastic trend. Thus AR(1) has an autoregressive root of 1, which impies that the null hypothesis H0 is that the AR(1) has a unit root and the alternative is that it is stationary. In order to rule out the deterministic trend, the ADF test is implemented by a modified version of Equation 3.1 by subtractingYt-1 from both sides:

t t t t

X

Y

Y

=

δ

+

η

+

ε

∆

'

where η = β - 1 and the test hypotheses now become H0: η = 0 and H1: η < 0. The

t-statistics follow the MacKinnon critical values.

As reported in table 3.3, for sample 1980-2010, their logarithmic levels are

non-stationary since all t-values are less negative than Dickey-Fuller critical values and differences are stationary at different significance levels, thus these variables can be applied in the cointegration model, and might have a long-term interdependency.

Table 3.3 ADF results: 1980-2010 1980-2010

lnSOYP ADF(4) trend -0.939 ΔlnSOYP PP -4.819 lnPOP ADF(4) trend -1.718 ΔlnPOP ADF(4) -3.619 **

lnGDP ADF(4) 0.72 ΔlnGDP ADF(4) -3.874

lnIMP ADF(4) trend -2.67 ΔlnIMP ADF(4) -2.185 Exrate ADF(4) trend -0.482 ΔExrate ADF(4) -1.118

Note: The best lag order has been chose previously, and for sample 1980-2010 is the second length so the model is ADF(4). two * show that for ΔlnPOP, at 5% and 10% significance level, the

differences of lnPOPare stationary. The differences oflnIMPand Exrate are both non-stationary so they will not be used in a VECM. More details in data appendix table A.1.

The circumstances of sample 1996-2010 (Table 3.4), is similar to those of sample 1980-2010. Most variables' logarithmic levels are non-stationary and differences are stationary. However, there is one exception: the difference of lnGDP is non-stationary

because its t-statistic is less negative than Dickey-Fuller critical values, which impies that we cannot reject null hypothesis. Therefore, I will rule it out in the sample 1996-2010 when analyzing these variables in next section. Namely, there is no cointegration relationship between China's GDP and international soybean prices.

Table 3.4 ADF results: 1996-2010 1996-2010

lnSOYP ADF(3) trend -1.812 ΔlnSOYP PP -2.687 * lnPOP ADF(3) -1.288 ΔlnPOP ADF(3) -18.306 lnGDP ADF(3) trend -1.68 ΔlnGDP ADF(3) -1.161 lnIMP ADF(3) -1.568 ΔlnIMP ADF(3) -10.93 Exrate ADF(3) -1.734 ΔExrate ADF(3) -4.453

Note: the explanation is similar to table 3.3's. For the differences of lnSOYP, the method is Phillips-Perron test rather than Augmented Dicker-Fuller test and the results are presented in data appendix table A.2. The differences oflnSOYPare stationary at 10% significance level thus there is one *, while the differences oflnGDPare non-stationary all the time.

3) 3)

3)3) JohansenJohansenJohansenJohansen andandandand JuseliusJuselius CointegrationJuseliusJuseliusCointegrationCointegrationCointegration TestTestTestTest

Johansen and Juselius (1990) developed a method to test whether there is a relationship between co-integrated variables. They demonstrated this method of maximum likelihood and applied it to two examples. This maximum likelihood test for cointegration can analyze multivariate system, considering a VAR with p lags as

follows: t p t p t t t

v

A

y

A

y

A

y

y

=

+

+

+

⋯

+

+

ε

where yt is a K×1 vector of variables, v is a K×1 vector of parameters, A1- Ap are

K×K matrices of parameters, and εt is a K×1 vector of disturbances. εt is independent and identically distributed (i,i,d) normal over time, and has covariance matrix ∑ and mean 0. With some algebra, Equation 3.3 can be rewritten in following VECM form:

t i t p i i t t

v

y

y

y

=

+

∏

+

Γ

∆

+

ε

∆

∑

with the difference operator Δ, yt denotes a K×1 vector of co-integrated variables of

order 1. Thev and εtare the same as Equation 3.3. Now we have coefficients matrix:

∑

∑

According to Engle and Granger (1987), if the variables are co-integrated and 0 <r

<K, where r is the number of linearly independent co-integrating vectors, Equation

3.4 indicates that a VAR in first differences is mis-specified as it omits the lagged level term Пyt-1. Alternatively, if the variablesytare IIII(1), the matrix П in Equation 3.4 will rank 0 ≤r < K.

Johansen and Juselius assumed that П has reduced rank 0 <r < K, thus it could be

expressed as П = αβ', where α and β are both r×K matrices of rank r. Each β is the

co-integrating vector with rankr, which means in П = αβ', β'ytis integrated of order 0, I(0), i.e. stationary. However, if there are no further restrictions, the co-integrating vectors cannot be identified: we are not able to distinguish the parameters (α, β) from

the parameters (αQ, βQ-1_{' ) for any r×r non-singular matrix Q. Since only the rank of}

П is identified, the VECM can be said to identify the number of cointegration vectors. In order to check whether these restrictions can be rejected, they estimated the coefficients matrix П from an unrestricted model. After estimating parameters П and Γi, they found that VECM could provide information about short- and long-term

adjustments in yt , because Пyt-1 is the error correction term. α is the adjustment

parameter of the VECM andr give the number of co-integrating relations.

Two sample sets of variables have been tested for the number of co-integrating relations. The results of Johansen's test will give the lag-length. But the lag-length in next section is chosen by the information criteria as discussed in section 3.1, so I present the Johansen's test results here just to check co-integrating relations. If the log likelihood of unrestricted model that includes the co-integrating equations, is significantly different from the log likelihood of restricted model that does not include the co-integrating equations, null hypothesis of no integration can be rejected,

Sample Sample

SampleSample 1980-20101980-20101980-20101980-2010

Johansen's method is employed to test the number of cointegration between international soybean price and China's population, and between the price and its GDP respectively. The results are presented in data appendix Table B.1. The header indicates that the test statistics are based on a model with two lags and a constant trend. The null hypothesis is that the number of co integrating relationships is equal to

r, given in the "maximum rank" column of the output and is rejected if trace statistic is

greater than the critical value.

Based on the results, for circumstance of world soybean price and China's population, H0: r = 0 is rejected at the 5% level as 19.27 > 15.41. Same for another H0: r = 1. It shows there might be more than 2 co-integrating relationships between

hypothesis, H0: r = 0, cannot rejected at 5% level (8.01 < 15.41). It fails to reject the

null hypothesis of no co-integrating equation, thus there is no co-integrating relationship between international soybean prices and China's GDP.

Sample Sample

SampleSample 1996-20101996-20101996-20101996-2010

Based on the theories described above, Johansen's test results indicates that there is no co-integrating relationship between international soybean price and China's GDP, and between the price and China's population. There is one co-integrating relationship between world soybean price and China's soybean imports, while more than one co-integrating relationships exist between soybean price and exchange rate of Chinese RMB per dollar. ( More details in Appendix Table B.2)

The findings above might result from plenty of reasons, such as the lack of observations. For sample 1980-2010, important factor is China's GDP, while for sample 1996-2010, focus should be China's soybean imports and its exchange rate of RMB per dollar, which will be discussed in next section.

4) 4)

4)4) VectorVectorVectorVector ErrorErrorErrorError CorrectionCorrectionCorrectionCorrection ModelModelModelModel (VECM)(VECM)(VECM)(VECM) EstimationEstimationEstimationEstimation

According to Engle and Granger (1987), if variables are co-integrated, we can use a valid error correction model to illustrate. In previous section, several variables were found to be co-integrated, and thus have a long-run relationship. In this section, the structure of VECM will be given firstly, and then the analysis will try to find out what factors of China affect international soybean price and quantify their estimated results.

The vector error correction (VEC) model is a special case of VAR for variables, whose differences are stationary (i.e., I(1)). One obvious advantage of VECM is that this model take into any co-integrating relationships among variables and prevent these endogenous variables converging to their long-run co-integrating relationships.

Now, an example from Engle and Granger (1987) is used. Define yt and xt as follows:

,

x

y

+

β

=

ε

ε

=

ε

+

λ

,

x

v

y

+

α

=

where λtandσt are i.i.d. disturbances over time that are correlated with each other. As

εt is assumed to be covariance stationary, both xt and yt are I(1) implied by Equation 3.6 and 3.7. The condition that |ρ| < 1 shows thatvt and yt+ αxtare I(0). Therefore, yt

and xt has a co-integrating relationship and (1,α) is the co-integrating vector. Then,

Equation 3.6 and 3.7 are rewritten as:

t t t

z

y

=

βδ

+

η

∆

z

x

=

−

δ

+

η

∆

where δ = (1-ρ) / (α-β), Zt= yt + αxt, andη1t andη2t are distinct, stationary and linear combinations of λtand σt. This representation is known as the vector error correction model.

To make this model easy to understand, a VAR(1) is used to illustrate. Dynamic relationship to two interrelated variables,xtand yt, yields a system of equations:

t t t t

y

x

v

y

=

β

+

β

+

β

+

y

x

v

x

=

β

+

β

+

β

+

If xtandytare not stationary in their levels, but stationary in their first differences (i.e., I(1)), now we take the differences and get:

t t t t t t

y

x

Y

X

u

y

=

∆

+

∆

+

(

−

)

+

∆

β

β

α

θ

y

x

Y

X

u

x

∆

+

∆

+

(

−

)

+

=

∆

β

α

θ

β

As xt and yt are I(1) and co-integrated, VECM can be applied. In this thesis, in accordance with the results of Johansen test, for sample 1980-2010, world soybean prices are integrated with China's GDP, while for sample 1996-2010, soybean prices are integrated with China's soybean imports and exchange rate of Chinese RMB per dollar. My focus will be the effects of China's soybean imports on international soybean prices, considering previous results and world soybean trade situation nowadays. Because the specification of the VECM for China's GDP and exchange rate is similar to China's imports, only the final specification for China's soybean imports is showed: t t t t t t t t t u IMP SOYP IMP IMP IMP SOYP SOYP SOYP SOYP 1 1 1 1 3 13 2 12 1 11 3 13 2 12 1 11 10 ) ln (ln ln ln ln ln ln ln ln + − + ∆ + ∆ + ∆ + ∆ + ∆ + ∆ + = ∆ − − − − − − − − θ α γ γ γ β β β β t t t t t t t t t u IMP SOYP IMP IMP IMP SOYP SOYP SOYP IMP 2 1 1 2 3 23 2 22 1 21 3 23 2 22 1 21 20 ) ln (ln ln ln ln ln ln ln ln + − + ∆ + ∆ + ∆ + ∆ + ∆ + ∆ + = ∆ − − − − − − − − θ α γ γ γ β β β β

where α(lnSOYPt-1 - θlnIMPt-1) is called error correction term (J.H. Stock and M.M.

Watson, 2012). Results will be discussed in next chapter.

4.

4.

4.

4. Results

Results

Results

Results and

and

and

and discussion

discussion

discussion

discussion

China's soybean imports has risen dramatically over the last 15 years and world food prices including the international soybean price experienced a rapid growth. As these rising might be caused by China's increasing food demand, this thesis tries to