The impact of food safety standards on rice exports from vietnam

Thus, developing countries face many challenges to export food to developed countries as well as meet the requirements of food safety standards imposed by importing countries.. There is

VIETNAM – NETHERLANDS PROGRAMME FOR M.A IN DEVELOPMENT ECONOMICS

THE IMPACT OF FOOD SAFETY STANDARD

ON RICE EXPORT FROM VIETNAM

By BUI MINH KHOI

MASTER OF ARTS IN DEVELOPMENT ECONOMICS

Ho Chi Minh City, November 2015

ABSTRACT

This paper investigates the impact of food safety standard on rice export from Vietnam The study covers 154 countries in the UN Comtrade data website from 2003 to 2013 and uses aflatoxin B1 to present for food safety standard The findings confirm that there is a negative impact of food safety standard on rice export from Vietnam Moreover, there is an existence of self – selection bias in the data Although Vietnam has exported rice to 154 countries in all over the world, there are some countries that do not import rice of Vietnam In other words, Vietnam does not export to those countries The reasons for this problem could be that those countries set low maximum allowable of aflatoxin B1 or the citizens of those countries do not eat low – value rice from Vietnam Besides, high-income countries will import less rice compared to lower income countries The reason for this issue could be that rice is a staple goods and the demand of rice depends on culture and habit of people in importing countries Another finding

is that joining WTO or signing FTA could lead Vietnam to export more rice to other countries However, this is only a signal, this issue should be considered in depth in future researches by using higher econometric methodologies

Keywords: food safety standard, rice export, Vietnam

ACKNOWLEDGEMENT

Foremost, I would like to sincerely and gratefully thank Dr Tran Tien Khai, my supervisor, for his great support, crucial advice and precious during my thesis finish Without his guidance, I am unable to finish this thesis

Besides, I would like to thank Dr Truong Dang Thuy, who also gave me useful advice when I got stuck in doing thesis More special thanks to the Vietnam – Netherlands Programme, especially professors and staffs for their help during my thesis process

Furthermore, I would like to thank all my friends, including my high school friends, my university friends, my group and some brothers in K20 All of them always beside me to encourage, help me when I got stuck in doing thesis and want to give up Without them, I cannot finish this thesis

Last but not least, I would like to thank my family for their sacrifices for supporting me not only in doing thesis but also in my life

ABBREVIATIONS

CEPII The Centre d’Études Prospectives et d’Informations Internationales

UN Comtrade The United Nations Commodity Trade Statistics Database

TABLE OF CONTENTS

ABSTRACT i

ACKNOWLEDGEMENT ii

ABBREVIATIONS iii

TABLE OF CONTENTS iv

LIST OF TABLES vi

LIST OF FIGURES vii

CHAPTER 1: INTRODUCTION 8

1.1 Problem Statement 8

1.2 Research Objectives 9

1.3 Research Questions 9

1.4 Data and Methodology 9

1.5 Thesis Structure 10

CHAPTER 2: LITERATURE REVIEW 11

2.1 Some Concepts 11

2.1.1 Food Safety Standards 11

2.1.2 Aflatoxin B1 11

2.2 Theoretical Literature 12

2.2.1 Food safety standards 12

2.2.2 Theory of Gravity Model 17

2.3 Empirical Literature 18

2.4 Hypothesis Testing 23

CHAPTER 3: DATA AND METHODOLOGY 24

3.1 Data Source 24

3.2 Model Specification 24

3.2.1 Constructing Dependent Variable 25

3.2.2 Model 25

3.3 Estimation Strategy and Correction Model 29

3.3.1 Panel Data Method 29

3.3.2 Heckman Sample Selection Bias Method 32

CHAPTER 4: OVERVIEW OF VIETNAMESE RICE EXPORT, FOOD SAFETY STANDARD FOR RICE, AND EXPORT’S DETERMINANTS 36

4.1 Overview about the Posture of Vietnamese Rice Export 36

4.2 Food Safety Standard Applied for Rice 40

4.3 Determinants of Vietnamese Rice Export 43

4.3.1 Regression Results of OLS, FE, RE, and Driscoll – Kraay Standard Errors 43

4.3.2 Regression Results of Heckman Sample Selection Bias Model 48

CHAPTER 5: CONCLUSION 54

5.1 Main Findings 54

5.2 Policy Implications 55

5.3 Limitations and Future Research 55

REFERENCES 57

APPENDICES 61

LIST OF TABLES

Table 4.1: Regression Results of OLS, FE, RE, Driscoll – Kraay Standard Errors for OLS and

FE 46Table 4.2: Regression Results of Heckman Maximum Likelihood, Heckman Two – Steps and Heckman Maximum Likelihood with Clustered 52

LIST OF FIGURES

Figure 2.1: Tariff and NTBs 14

Figure 2.2: Analytical Framework of Trade and Determinant Factors 22

Figure 4.1: Average Export Quantity of Vietnamese rice 37

Figure 4.2: Average Export Value of Vietnamese Rice 38

Figure 4.3: Average Export Quantity of Vietnamese Rice to 5 Continents in 2003 – 2013 39

Figure 4.4: Average Allowable Level of Aflatoxin B1 of 5 Continents in 2003 – 2013 40

Figure 4.5: Food Safety Standard and Export Quantity of Vietnamese Rice to 5 Continents 41 Figure 4.6: Average GDP per capita and Export Quantity of Vietnamese Rice 42

CHAPTER 1: INTRODUCTION

1.1 Problem Statement

Food safety standard is an emerging issue in recent years Especially, imposing food safety standards could be a signal for consumers that product is safe, but in parallel, if this is a protectionist way of importing countries (developed countries), it could impede trade (Disdier et al., 2008) The latter is more concerned regarding developing countries compared to the former Most of importing countries (developed countries) use food safety standard as a tool to protect their domestic producers Thus, developing countries face many challenges to export food to developed countries as well as meet the requirements of food safety standards imposed by importing countries In the perspective of global trade, goods of a country have to compete with many other countries’ Therefore, consumers are faced with importing foods that are supplied by unclear suppliers These problems are real, but in parallel, this could lead to an impediment in trade (van Veen, 2005) The weakness in physical infrastructure and high possibility to be infected

by disease could make developing countries that face more challenges than developed countries (Henson & Jaffee, 2008) Thus, developing countries should view food safety standards not just

as a tool for impeding imports of developed countries but as an opportunity to enhance the quality

of foods (Jongwanich, 2009) However, he also argues that developing countries are hard to meet the requirements of food safety standard imposed by developed countries compared to developed countries There are many papers that measure the effect of food safety standards on food export from developing countries such as Jongwanich (2009), Ferro et al (2013) or from African countries like Otsuki et al (2001a, 2001b) and Wilson and Otsuki (2001) However, doing research about food safety standards for Asia, ASEAN are very rare, especially for Viet Nam Thus, it is necessary to analyze this problem in the case of Vietnam when Vietnam is an agricultural country

In this study, the impact of food safety standard on rice export from Vietnam is the main research objective Although there are several papers that studied the food safety standard for agricultural goods of Vietnam, those studies are for fish products such as Henson et al (2000) There is practically no research that measure the impact of food safety standard on rice export from Vietnam Many papers such as Otsuki et al (2001a, 2001b), Wilson and Otsuki (2001), Disdier et

al (2008), Jongwanich (2009) and Ferro et al (2013) that find a negative relationship between food safety standard and export from developing countries However, those studies have never

applied for Vietnamese case Hence, by measuring the impact of food safety standard on rice export from Vietnam, this study could be a shed light to the situation of Vietnamese rice export

By drawing the results, this research aims to give a better understanding about the food safety standard and export from developing countries like Vietnam From there, future studies could do research for a different type of Vietnamese agricultural products to help Vietnamese agricultural goods could export more to high quality as well as stickiness markets such as EU, USA or Japan

1.4 Data and Methodology

This study uses the manually collected panel data of 154 countries on UN Comtrade from

2003 to 2013 The quantity of rice is collected from UN Comtrade website, rice is coded by Harmonized System - HS 1006 Besides, the data of food safety standard is obtained from an FAO publication: “Worldwide Regulations for Mycotoxins (2003): A Compendium” Moreover, GDP per capita and openness are collected annually from World Bank Further, CEPII and FAO websites provide the data of distance and domestic consumption of rice, respectively The data of tariff is gathered from TRAINS database WTO and FTA are referred from World Trade Organization and Asia Regional Integration Center websites The OLS, Fixed Effects (FE), Random Effects (RE), Driscoll – Kraay Standard Errors for OLS and FE, Heckman Maximum Likelihood and Heckman Two – Steps will be used to estimate the impact of food safety standard

on rice export from Vietnam The results of Heckman Maximum Likelihood and Heckman Two – Steps are considered the main results While the OLS, FE, RE, and Driscoll – Kraay Standard Errors for OLS and FE will be used as the comparative results Cluster – Robustness Check will

be used to strengthen the findings

1.5 Thesis Structure

The following section is the literature review The third section will be data and methodology The fourth one is an overview of Vietnamese rice export, food safety standard for rice and export’s determinants The last one is the conclusion, policy implications, and limitations

CHAPTER 2: LITERATURE REVIEW

2.1 Some Concepts

2.1.1 Food Safety Standards

“Food safety standard is a scientific discipline describing handling, preparation, and storage of food in ways that prevent foodborne illness This includes a number of routines that should be followed to avoid potentially severe health hazards” (Department of Agriculture, 2009) Recent years, there is a reduction in tariffs and quantitative restrictions in developed and developing countries Instead, NTBs have been growing to impede trade from developing countries One of the NTBs is SPS Agreement under the WTO which could hamper trade in agricultural and food products Essentially, developed countries impose food safety standard to control the potential risks of food export from developing countries However, the main purposes

of imposing food safety standard by developed countries are to protect their domestic producers and to impede trade from developing countries Especially, imposing food safety standards could

be a signal for consumers that product is safe, but in parallel, if this is a protectionist way of importing countries (developed countries), it could impede trade (Disdier et al., 2008) The latter

is more concerned regarding developing countries compared to the former Most of importing countries (developed countries) use food safety standard as a tool to protect their domestic producers Thus, developing countries face many challenges to export food to developed countries

as well as meet the requirements of food safety standards imposed by importing countries

2.1.2 Aflatoxin B1

Aflatoxins are a group of toxic combinations that contaminate in foods and cause cancer

in human body There are some kinds of aflatoxins: B1, B2, G1 and G2 that could be found in foodstuffs In four categories, aflatoxin B1 is the most toxic Aflatoxins are commonly in cereals, nuts, fruits and vegetables According to van Egmond et al (2003), a serious effect on humans and animals health by mycotoxins, particularly aflatoxins will lead to many countries to impose regulations on aflatoxins in foodstuffs and feedstuffs to protect human health In the late 1960s, the first limit for aflatoxins was set Until 2003, there are approximate 100 countries that set the regulations on aflatoxins on foodstuffs and feedstuffs

2.2 Theoretical Literature

2.2.1 Food safety standards

In the perspective of global trade, goods of a country have to compete with many other countries’ Therefore, consumers are faced with importing foods that are supplied by unclear suppliers These problems are real, but in parallel, this could lead to an impediment in trade (van Veen, 2005) The weakness in physical infrastructure and high possibility to be infected by disease could make developing countries that face more challenges than developed countries (Henson & Jaffee, 2008) Thus, developing countries should view food safety standards not just as a tool for impeding imports of developed countries but as an opportunity to enhance the quality of foods (Jongwanich, 2009) However, he also argues that developing countries are hard to meet the requirements of food safety standard imposed by developed countries compared to developed countries Moreover, standards that are imposed by importing countries lead to an increase in compliance cost and limit or prevent market access (Henson & Jaffee, 2008) Henson et al (2000) also find that SPS requirements, especially in EU, become one of the most impediments to trade

in agricultural and food products under the considering of a number of developing countries Vietnamese fish export to EU decreases remarkably since there was a new sanitary regulatory announced by EU in 1991 and a compliant requirement in 1997 (Henson et al., 2000) Roberts et

al (1876) claim that technical barriers to trade are used to protect the domestic producers in international competition Hillman (1996) and Thornsbury (1999) argue particularly that NTBs are seen as an evolution in protecting domestic production, in other words, also known as “new protectionism” However, countries have the right to establish their own standards as long as these standards based on science and do not violate GATT, SPS agreement

In order to control food export from developing countries, many governments establish a food standard as known as HACCP system HACCP is known as a food system which is established in food industries to produce safe foods Unnevehr and Jensen (1999) claim that standards could impede trade from developing countries that do not meet the requirements of HACCP According to Unnevehr and Jensen (1999), although establishing HACCP brings benefits that could reduce food foodborne illness, there are some difficulties In practice, mandating HACCP does not mean reducing the risk of foodborne illness Besides, a standard that could balance marginal costs and marginal benefits is very difficult to find Henson et al (2000) also

argue that SPS agreement is one of the impediments to trade in agricultural and food products from developing countries to developed countries, especially in the case of European countries Most of developing countries are poor and they have to face the requirements of food standards imposed

by developed countries Thus, problems of complying with SPS standards are difficult to have experienced scientists in food field and technical scientific skills

Besides, Thornsbury et al (1997) state that technical barriers cause barriers in respect to international trade in agricultural products These barriers have existed in most of the industries, but it is especially important in primary agricultural goods and processed goods (Roberts et al., 1876) Importing countries commonly require exporters that not only prove their products do not cause dangers to human health but also meet their requirements of standards about nutrition, package, and label Developed countries normally set up the requirements to exporters such as enhancing food standards and protecting resources in the Earth However, policy makers also realize that except above reasons, using technical barriers is not transparent The purpose of establishing these barriers is to protect domestic producers of importing countries Sum up, although the purpose of setting up a food safety regulation is to protect the human health, to reduce the risks of morbidity and death, the main aim is to establish a shield to protect domestic producers

By establishing the standards of contaminants, the risks of morbidity and death could be decreased, these are the benefits of food safety regulations (Antle, 1999) Furthermore, Hillman (1966) and Thornsbury et al (1997) have the same point of view that a country should not create an SPS safety standard if it is not based on empirical studies about science Roberts and Orden (1995) also argue that food safety standard could be used as a shield for domestic producers in international competition

Figure 2.1: Tariff and NTBs

Source: Thornsbury et al (1997)

According to Beghin and Bureau (2001), there are several methods to measure the NTBs

to trade such as Price – Wedge Approaches, Inventory – Based Approaches, Survey – Based Approaches, Gravity – Based Approaches, Risk – Assessment – Based Cost – Benefit Measures, Stylized Microeconomic Approaches, and Quantification Using Sectoral or Multi – market Models Each of method has advantages and disadvantages, and could be applied in specific cases

Trade Barriers Tariffs

Quotas Voluntary export constraints

Subsidies and Taxes

Products Inputs

Trade Restricting Regulations

procurement

Domestic Content Regulations

Foreign Direct Investment Regulations

Price – Wedge Approaches: this method is the method that is used to compare the domestic

price to importing price in order to measure NTBs Campbell and Gossette (1994) uses this method for a large number of sectors, including food and agriculture Besides, Calvin and Krissoff (1998) studied a research which estimates tariff equivalents to quantify the phytosanitary barriers to the U.S apple exports to Japan They used domestic wholesale price compared to CIF prices (landed prices plus insurance and freight costs) Although there are several studies that use this method to measure the impact of NTBs on trade, this method remains some limitations First, we can measure the effect of NTBs on trade, but we could not know exactly what those NTBs are Second, this method is only valid when the imported goods are perfect substitutes Third, if there is a price difference in exporting firms, not only NTBs will be reflected, but also rents are reflected more than NTBs by the price – wedge method Finally, the main limitation of this method is data lacking

Inventory – Based Approaches: this method is used for both quantitative and qualitative

analysis We could collect data from three sources: data on regulations, data on the frequency of detentions and data on complaints from the industries that are imposed regulatory Swann et al (1996) used a variable which is frequency indicators of standards to present for a number of voluntary national and international standards of United Kingdom and Germany in the period 1985 – 1991 Analyzes of border detentions in the U.S were applied to measure the impact of import procedures and border inspection on EU exports and the impact of detained shipments on processed food exports from developing countries (Lux & Henson, 2000; Jongwanich, 2009) However, there are some limitations of this method For example, a number of regulations will be

a poor proxy to reflect the effect on trade because different standards will have different effects Although actual detention is a good proxy for measurement, the data is limited, except the case of the United States

Survey – Based Approaches: the scope of the analysis will be narrowed down as well as

the relevant issues could be concentrated on are the aims of this method Thornsbury et al (1997) and Thornsbury (1999) estimated the impact of questionable technical barriers on the U.S agricultural exports Although researchers use this method to measure NTBs that are difficult to estimate, the survey result could be biased One of the reasons causes bias result is the information being used for policy purposes

Gravity – Based Approaches: Gravity model is based on the Newton’s “Law of Universal

Gravitation” The Law says that attractive force has a positive relationship with the masses of two

objectives and negative relationship with their distances Similarly, economists introduced gravity model in economic in the 1960s The first studies that used gravity model are Tinbergen (1962), Pöyhönen (1963) and Aitken (1973) Year after year, gravity model is used to measuring the bilateral trade but lack of theoretical foundations Until 1979, Anderson presented a theoretical – based gravity model Several papers have results consistent with the monopolistic competition when using gravity model (Bergstrand, 1989; Deardoff & Stein, 1998) Otsuki et al (2001a, 2001b) and Wilson and Otsuki (2001) used a gravity model to estimate the impact of food safety standards

on trade In the main, gravity – based approaches, combined with inventory – or survey – based methods, could be a good method to measure the bilateral trade

Risk – Assessment – Based Cost – Benefit Measures: Although NTBs are hard to measure

by risk – assessment analysis, this method is combined with costs – benefits analysis and could explain partly the effect of regulations or NTBs In the case of pest infestation, Bigsby et al (2001) introduced a methodology to measure both of economic effects and probability aspects of risks This method is a promising method in the measurement of the effect of NTBs The only limitation

of this method is that it is hard to quantify exactly the level of risks and economic consequences

Stylized Microeconomic Approaches: The basic idea of this method is measuring the effect

of regulations on supply and demand by using cross – sectional or panel data with the cost or profit functions econometrically Marette, Bureau and Gozlan (2000) find that whether consumers evaluate the quality of products or the time period when they assess are to determine differently the welfare effects of regulations Due to regulations, trade, as well as the preferences of consumers,

could be measured by the econometric method

Quantification Using Sectoral or Multi – Market Models: In respect to partial equilibrium

models’ property, compared to gravity models, instead of estimating the effect of regulations on trade flows, it could also measure the impact of regulations on welfare On the other hand, it provides more quantitative effects in comparison with stylized micro-economic approaches which only concentrate on qualitative results The effect of regulations on production, consumption, trade, and welfare could be estimated in large – scale models, which focus on both tariff and regulations Calvin and Krissoff (1998) presents a study that combined price – wedge method with partial equilibrium model Another combining research between risk – based approach and partial equilibrium model is Paarlberg and Lee (1998) Their study investigates the impact of the U.S

tariff on beef exports from countries may have FMD This method is one of the promising methods

which narrows down the gap between stylized micro-economic method and its own

After considering all potential methods that could measure the effect of regulations on bilateral trade and referring consistent literature studies (Otsuki et al., 2001a; Otsuki et al., 2001b; Wilson & Otsuki, 2001; Mangelsdorf et al., 2012; Ferro et al., 2013), gravity model is the most appropriate model to estimate the impact of food safety standard on bilateral trade

2.2.2 Theory of Gravity Model

The gravity model is a famous and successful model over decades According to Frankel and Wei (1997), the gravity model of bilateral trade is a functional form that measures trade between two countries by including a directly proportional per capita income of each country and negatively in respect to the distance between them into the model Gravity model was introduced firstly by Tinbergen (1962), which estimated the bilateral trade among countries Furthermore, one

of the first papers which mentioned gravity model is the research of Pöyhönen (1963) Pöyhönen (1963) observed the bilateral trade among ten European countries In his research, he included national income of both exporting and importing countries, transport cost is presented by the distance between exporting and importing country Besides, Aitken (1973) also based on the Tinbergen’s trade flow model, or in other words, gravity model to estimate a cross – sectional trade flows among countries He also used main variables of gravity model such as national income, population, distance Furthermore, he added dummy variables such as EEC and EFTA into the model in order to analyze whether the effect of EEC and EFTA on trade flows if countries join EEC and EFTA

Although gravity model was developed and used successfully to explain bilateral trade, a lack of theoretical foundation should be considered Anderson (1979) mentioned a problem of gravity model that policy variable such as border taxes is put into the model, there is no theoretical justification Thus, his paper’s aim is to provide a theoretical background for gravity model which applied for all commodities The difference of his research compares to previous studies is that he included tariff into the gravity model since tariff could hamper trade Besides, although income which has a positive effect on trade is an important foundation, Helpman (1987) and some other authors do not see distance as a vital role which could affect trade Thus, this could be seen a lack

of theoretical for a full gravity model Until 1985, Bergstrand (1985) argues that incomes, distances

are variables made gravity model However, price and exchange rate also play an important role and impact trade flows The model which included price terms was called generalized gravity model

An analytical framework for realizing the gravity model which was provided by Bergstrand (1989) It is consistent with theories of inter – industry and intra – industry Units of capital and capital – labor endowment ratios of the country could be seen as national income through exporter income and per capita income, respectively Besides, the taste preferences and the change of expenditure capacity could be presented by per capita income and importer income Bergstrand (1990) uses a gravity model to measure the relationship between the share of intra – industry trade and inequality in national incomes, per capita incomes, capital – labor endowment ratios and tariffs

of countries He found that lower inequalities of two countries’ per capita income would lead to a more intra – industry trade between them Although three types of research’ objectives of Bergstrand are different, per capita income, price, tariff, national income and exchange rate were included in gravity model

Although the gravity model is mainly based on three trade theories: Heckscher – Ohlin, monopolistic competition and Ricardian technologies, three of them have the same functional form (De Frahan & Vancauteren, 2006) In addition, Anderson and van Wincoop (2003) introduce a theory – based gravity model that includes income, distance, and price as main variables In sum, all papers based on theory gravity model that also included national income or per capita income, distance These variables are the main variable of the gravity model Higher average national income or per capita income shows the higher level of economics development, this could lead to

an increase in demand for products Besides, a higher capital – labor endowment ratio is caused

by higher average national income or per capita income (Helpman & Krugman, 1985) In addition, the higher distance between exporting and importing countries, the trade flow between two countries will be less Furthermore, in theory – based gravity model, tariff and price play a prominent role in influencing trade flows and they should be included into the gravity model (Bergstrand, 1989; Bergstrand, 1990)

2.3 Empirical Literature

Calvin and Krissoff (1998) studied a research that measure the impact of SPS on the US – Japanese apple trade They estimated descriptive statistics results for the US – Japanese apple trade

in three periods: 1994 – 1995, 1995 – 1996 and 1996 – 1997 In their research, they only use descriptive statistics to measure the short – run and long – run change in Japan’s Fuji Apple Imports with the elimination of trade barriers They did not use econometric to estimate the impact

of SPS on the US – Japanese apple trade The results of their research show that when eliminating the trade barriers, the Japanese imports will increase

Using foods that are contaminated could increase risks of morbidity and death The benefit

of food safety regulation is a decline of those (Antle, 1999) Besides, Mangelsdorf et al (2012) and Disdier et al (2008) believe that food safety standards could reduce asymmetric information and transparency could be improved by signaling to consumers that products are safe However, Antle (1999) also points out that the costs of food safety regulation are compliance cost of industry and administration cost A gravity model was applied by Otsuki et al (2001a) to estimate that the stringent requirements for the maximum allowable contaminant in groundnuts imposed by the European Union lead to a decrease in exports from African countries In that paper, they use the data of trade between Europe and Africa in the period 1989 – 1998 By applying fixed effects method, Otsuki et al (2001a) find that a 10% increase in the maximum allowable level of aflatoxins B1 will lead to an 11% increase in edible groundnuts imports of European from Africa Besides, with the same applying method of gravity model and fixed effects, Otsuki et al (2001b) finds that a high restriction of new EU standards will lead to a decrease in African exports of cereals, dried nuts and nuts to Europe Nevertheless, those are cases that measure the standards imposed by importing countries

In addition, Wilson and Otsuki (2001) finds the results that aflatoxin B1 standard has an adverse effect on trade in cereals and nuts They use the data of 15 importing countries and 31 exporting countries Besides, due to the lack of data, Disdier et al (2008) have to use cross-sectional method for 2004 to estimate the impact of SPS and TBT on agricultural trade Nonetheless, a difference of Disdier et al (2008) from the studies of Otsuki et al (2001a, 2001b) and Wilson and Otsuki (2001) is that they use tariff barrier as an independent variable The reason

of including this variable in the gravity model is that we could separate the different effect between NTBs, i.e SPS and TBT, and tariff barriers on agricultural trade

Furthermore, De Frahan and Vancauteren (2006) used a gravity model to measure the effect of harmonization of EU food regulations on intra – EU in the period 1990 – 2001 They claim that the stricter in EU harmonization of food safety regulations, the more trade in intra – EU

Besides, Jongwanich (2009) defines the food safety standard through SPS by a number of detained shipments Although the way to measure the food safety standard is different from other papers, the result is still the same That means the standard imposed by importing countries will hamper processed food export from developing countries He also includes openness variable into the model, the result shows that openness has the positive impact on the processed food export from developing countries

Commonly, most of the papers studied about food safety standard use a data that have a pair of exporting countries and a pair of importing countries Besides, researchers ordinarily use gravity model and estimate the food safety standard imposed by importing countries Nevertheless, there is one research which is the research of Mangelsdorf et al (2012) measures the food safety standard of exporting country, China Although they use a gravity model to estimate the impact of food standard on export from China, they do not include distance into the model This could be seen as a limitation of this research However, there are some differences compared to other studies: they use four types of standards: voluntary, the mandatory domestic standard of China and voluntary, mandatory international standard This is a new way to present the food safety standard They also use a new method to estimate the relationship between food standards and export from China Instead of using only fixed effects method, they use Heckman two – steps and threshold Tobit to measure the impact of food standards on Chinese export Furthermore, the finding result

is interesting in comparison with other papers’, i.e they find that the standards have the positive effect on Chinese export Besides, using domestic consumption of a product as an independent variable is remarkable

Another research of Ferro et al (2013), which has a new way to introduce food safety standard They use the database of MRLs to present the food safety standard This is a new database which could collect from Agrobase – Loligram’s Homologa website However, this database is not free to access According to Ferro et al (2013), the advantages of this database which compare to other previous studies are more products and more contaminants that are used

In comparison with some other papers such as Otsuki et al (2001a, 2001b) and Wilson and Otsuki (2001), those studies only use one contaminant to measure the impact of food safety standard on food export Besides, a new method which is Heckman sample selection bias was applied to estimate the effect of food safety standard on agricultural exports from developing countries The result of this research shows that there is a negative relationship between MRLs and agricultural

exports from developing countries, i.e more MRLs or a higher number of pesticides imposed by importing countries, less agricultural exports from developing countries

An analytical framework will be presented below in order to show the determinants that impact the trade, i.e the rice export from Vietnam in the scope of this research

Figure 2.2: Analytical Framework of Trade and Determinant Factors

Gravity variables

Food Safety Standard

Tariff Regulations variables

2.4 Hypothesis Testing

Previous studies such as Otsuki et al (2001a, 2001b), Wilson and Otsuki (2001), De Frahan and Vancauteren (2006), Disdier et al (2008), Jongwanich (2009) and Ferro et al (2013) provide argument that food safety standards imposed by importing countries have a negative impact on export from developing countries

H1: Food safety standard has a negative effect on rice export from Vietnam

CHAPTER 3: DATA AND METHODOLOGY

3.1 Data Source

The data covers 154 listed countries on UN Comtrade from 2003 to 2013 The quantity of rice is collected from UN Comtrade website, rice is coded by Harmonized System - HS 1006 Besides, the data of food safety standard is obtained from an FAO publication: “Worldwide Regulations for Mycotoxins (2003): A Compendium” Moreover, GDP per capita and openness are collected annually from World Bank Further, CEPII and FAO websites provide the data of distance and domestic consumption of rice, respectively The data of tariff is gathered from TRAINS database WTO and FTA are referred from World Trade Organization and Asia Regional Integration Center websites Finally, the data sample contains 1069 observations for 154 countries

- Heckman sample selection bias model: this method is used to measure observable

factors, while not measure unobservable factors

- Treatment effects: this method is used to measure both observable factors and unobservable factors

Since in this thesis’s data, only the data of Vietnamese exporting rice to 154 countries is available, whereas there is no data observation of importing countries that do not import rice from Vietnam In other words, the quantity of exporting rice from Vietnam to those countries is zero

 The sample selection bias in data sample is whether exists is measured by two equations: regression model and selection model

Regression model:

ln⁡(𝑞𝑢𝑎𝑛𝑡𝑖𝑡𝑦)𝑖𝑡 = 𝛼1+ 𝛼2𝑙𝑛(𝑝𝑟𝑖𝑐𝑒 ∗ 𝑑𝑖𝑠𝑡)𝑖𝑡+ 𝛼3ln⁡(𝑓𝑠𝑠)𝑖𝑡+ 𝛼4ln⁡(𝑔𝑑𝑝𝑝𝑐)𝑖𝑡+

𝛼5ln⁡(𝑐𝑜𝑛𝑠)𝑖𝑡+ 𝛼6tariff𝑖𝑡+ 𝛼7𝑜𝑝𝑒𝑛𝑖𝑡+ 𝛼8𝑑𝑢𝑚_𝑤𝑡𝑜 +

𝛼9𝑑𝑢𝑚_𝑓𝑡𝑎 + 𝜀1 Selection model:

γ1+ 𝛾2𝑎𝑣𝑒_𝑜𝑑𝑒𝑠𝑡𝑖𝑡+ 𝛾3ln(𝑓𝑠𝑠)𝑖𝑡+ 𝛾4𝑑𝑖𝑠𝑡𝑖 + 𝑢2 > 0 Details of variables information will be discussed in the next section

3.2.1 Constructing Dependent Variable

The quantity of Vietnamese rice export or saying other words, the quantity of importing rice countries from Vietnam is collected annually from UN Comtrade website in the period 2003 – 2013

3.2.2 Model

 Panel Data Model

Based on the theory of gravity model, GDP per capita and distance are the main variables

that must include in the model Thus, in this thesis, GDP per capita and distance are also included

ln⁡(𝑞𝑢𝑎𝑛𝑡𝑖𝑡𝑦)𝑖𝑡 = 𝛼1+ 𝛼2⁡ln(𝑝𝑟𝑖𝑐𝑒 ∗ 𝑑𝑖𝑠𝑡)𝑖𝑡+ 𝛼3ln⁡(𝑓𝑠𝑠)𝑖𝑡+ 𝛼4ln⁡(𝑔𝑑𝑝𝑝𝑐)𝑖𝑡+

𝛼5ln⁡(𝑐𝑜𝑛𝑠)𝑖𝑡+ 𝛼6tariff𝑖𝑡+ 𝛼7𝑜𝑝𝑒𝑛𝑖𝑡+ 𝛼8𝑑𝑢𝑚_𝑤𝑡𝑜 +

𝛼9𝑑𝑢𝑚_𝑓𝑡𝑎 + 𝜀𝑖𝑡

Whereas:

Quantity is the total quantity of rice exporting from Vietnam to country i in time t (kg), collected from UN Comtrade website

Dependent variables:

Price is the price of rice export from Vietnam to country i in time t (USD/kg), calculated

by the value of trade over quantity Dist is the geographical distance between Vietnam and country

i (thousand kilometers), collected from CEPII website Since distances do not change over time,

so when running fixed effects regression, this variable will be omitted Thus, the interaction term between price and distance should be considered The meaning of this interaction term is that the higher distance, the greater the effect of price on quantity and otherwise Essentially, distance represents for transport costs, i.e the higher price could offset the higher transport costs and this will make Vietnam, which exports more rice to other countries Thus, the expected sign of this variable is positive

Otsuki et al (2001a, 2001b), Wilson and Otsuki (2001) introduce food safety standard which is the maximum allowable level of aflatoxin B1 Thus, this research will present FSS is food safety standard which is maximum allowable level of aflatoxin B1 in rice (𝜇g/kg), obtained from FAO publication: “Worldwide Regulations for Mycotoxins (2003): A Compendium” Higher maximum allowable level of aflatoxin B1 imposed by importing countries will lead to a higher export quantity of Vietnam rice Thus, the expected sign of this variable is positive

GDPPC is GDP per capita of country i in time t (constant 2005 US$), gathered from World Bank database The richer countries, the more rice import of importing countries Therefore, the expected sign is positive

Mangelsdorf et al (2012) find that domestic consumption of a product positively impacts

on Chinese exports Therefore, this thesis will include this variable to measure the impact of it on the quantity of rice export from Vietnam Cons is domestic consumption quantity of rice of country

i in time t (kg)1, computed by the author Thus, the sign of this variable is expected positively

of rice (measured by kg), S is domestic supply quantity of rice (measured by kg), M is import quantity of rice (measured by kg) and E is export quantity of rice (measured by kg) S, M and E are collected from Food Agricultural Organization website

According to Disdier et al (2008), the tariff should be included in the model to clarify the impact of food safety standard and tariff on trade Thus, the tariff will be put into the model Tariff

is the average weighted tariff rates of included HS 6-digit subheading rice (in percentage)2 The more tariff rate imposed by importing countries, the less rice exporting quantity of Vietnam Hence, the expected sign is negative

Jongwanich (2009) presents trade policy regime in his research and also introduces two ways to measure trade policy regime: trade to GDP and implied tariff rate However, to simplify the way to measure, this research will choose trade to GDP to measure the trade policy regime or openness Open is the openness, trade to GDP (%), collected from World Bank database The more openness, the more import Thus, the expected sign of this variable is positive

WTO is referred from World Trade Organization websites This variable is dummy variable Since this research studies for Vietnam, so Vietnam will be the country which is referred That means, because Vietnam joined the WTO in 2007, thus, before 2007 and in respect to a country which have not joined WTO, WTO dummy will take the value 0 In respect to countries that joined WTO before Vietnam, from 2003 to 2007, the dummy variable of WTO will take the value of 1 Besides, in respect to countries that joined WTO after Vietnam, the value of 1 will be taken in the year which that country joined WTO Countries that join WTO could tend to trade more with each other Hence, the expected sign of WTO is positive

FTA is free trade agreements that Vietnam signed with other countries This is a dummy variable which is referred from Asia Regional Integration Center website It is similar to the way constructing WTO dummy, Vietnam is the country which is referred However, in Asia Regional Integration Center website, there are 15 FTAs included Vietnam Thus, the value of 1 will be taken

in the year which Vietnam and a country or a group of countries sign a bilateral trade agreement

or free trade agreements This dummy variable will take the value of 0 if Vietnam and a country

do not sign a bilateral trade agreement or free trade agreements as well as the years before Vietnam

included HS 6-digit subheading products weighted by the country's own imports from the world in the same or the nearest available year as tariff The tariff rate for each HS 6-digit product is itself a simple average rate of included tariff lines”

and a country sign a bilateral trade agreement or free trade agreements Countries that sign FTA would tend to trade more with each other Thus, the expected sign of FTA is positive

 Heckman Sample Selection Bias Model

at least one exogenous variable Thus, there are three exogenous variables: ave_odest, ln(fss) and dist, which are put into selection model in this thesis Detail information of these variables will be mentioned below

Ave_odest, which is presented in the research of Ferro et al (2013) In their research, this variable is “the average restrictiveness index an exporter faces in all possible destinations other than i” (i presents for country i) That means, in respect to a particular country, e.g USA, this variable will present the average restrictiveness index an exporter faces in all possible destinations other than the USA Thus, similarly, this variable will be included in this thesis However, it will

be redefined, i.e ave_odest is the average level of aflatoxin B1 Vietnam faces in all possible destinations other than country i in time t

FSS is food safety standard which is maximum allowable level of aflatoxin B1 in rice (𝜇g/kg), obtained from FAO publication: “Worldwide Regulations for Mycotoxins (2003): A Compendium” This variable could affect the probability of Vietnamese rice export to other countries

Dist is the geographical distance between Vietnam and country i (thousand kilometers) Ferro et al (2013) introduce this variable in selection model in their research Thus, this thesis follows their empirical research and put this variable into selection model

3.3 Estimation Strategy and Correction Model

3.3.1 Panel Data Method

3.3.1.1 The Method of Ordinary Least Squares (OLS)

Based on Gujarati (2003), the method of OLS starts with the two-variable PRF:

𝑌𝑖 = 𝛽1+ 𝛽2𝑋𝑖 + 𝑢𝑖Nevertheless, the PRF could not be directly observable Thus, the SRF should be considered:

𝑌𝑖 = 𝛽̂1+ 𝛽̂2𝑋𝑖 + 𝑢̂𝑖

Where 𝑌̂𝑖 is the estimated value of 𝑌𝑖

𝑢̂𝑖 = 𝑌𝑖 − 𝑌̂𝑖 = 𝑌𝑖 − 𝛽̂1− 𝛽̂2𝑋𝑖

As we can see, the 𝑢̂𝑖 is the difference between the actual value 𝑌𝑖 and the estimated value 𝑌̂𝑖 The idea of this method is to square the residuals (𝑢̂𝑖) which is as small as possible Thus,

𝛽̂1 =∑ 𝑋𝑖2∑ 𝑌𝑖 − ∑ 𝑋𝑖∑ 𝑋𝑖𝑌𝑖

𝑛 ∑ 𝑋𝑖2− (∑ 𝑋𝑖)2

= 𝑌̅ − 𝛽̂2𝑋̅

The estimators 𝛽̂1 and 𝛽̂2 are the least-squares estimators Besides, the OLS will be efficient if assumptions are met There are seven assumptions:

Assumption 1: Linear regression model, it means linear in the parameters

Assumption 2: Fixed X values or X values independent of the Error Term

where var stands for variance

Assumption 5: No autocorrelation between the disturbances

𝑐𝑜𝑣(𝑢𝑖, 𝑢𝑗|𝑋𝑖, 𝑋𝑗) = 0 𝑐𝑜𝑣(𝑢𝑖|𝑢𝑗) = 0, if X is nonstochastic where cov means covariance and where i and j are two different observations

Assumption 6: The number of observations n must be greater than the number of parameters to be estimated

Assumption 7: The nature of X variables, it means there are no outliers in the value of X variables

3.3.1.2 The Method of Fixed Effects (FE)

Based on Greene (2012), because of the assumption that the omitted effects, c i, the fixed effects model should be considered, in the general case,

𝑦𝑖𝑡 = 𝑥𝑖′𝛽 + 𝑐𝑖+ 𝜀𝑖𝑡, are correlated with the variables that included in the model In a general form,

𝐸(𝑐𝑖|𝑋𝑖) = ℎ(𝑋𝑖) Since the conditional mean does not change in every period, we can rewrite the model:

With 𝑣𝑎𝑟(𝑐𝑖|𝑋𝑖) is constant as an assumption The model will become a classical linear regression model The fixed effects means the differences across observations are the differences

in the constant term An unknown parameter is each 𝛼𝑖 to be estimated The time-invariant error term 𝑢𝑖 is not only a part of the intercept, but also could be allowed to correlate with the explanatory variables There are two ways to estimate the FE model: using least square dummy variables (LSDV) or within effect estimation method The variation of unexpected events to the outcome could be controlled by the FE model when including time dummy variable in the model

3.3.1.3 The Method of Random Effects (RE)

The RE model is different from the FE model Based on Greene (2012), in the FE model, the unobservable effects are allowed to correlate with the regressors On the other hand, if the unobservable effects do not correlate with the included variables, the FE model will be inappropriate Thus, the RE model should be considered,

𝑦𝑖𝑡 = 𝑥𝑖𝑡′𝛽 + (𝛼 + 𝑢𝑖) + 𝜀𝑖𝑡

Or we could write: 𝑦𝑖𝑡 = 𝑥𝑖𝑡′𝛽 + 𝛼 + (𝜀𝑖𝑡+ 𝑢𝑖)⁡

As we can see, the intercept, 𝛼, are constant across individuals and the term 𝑢𝑖 does not correlate with the regressors The RE model could include time invariant variables such as gender, skin, etc in the model, while the effects of these variables are included in the intercept The RE model could be estimated by many different methods such as Generalized Least Squares (GLS), Feasible Generalized Least Squares (FGLS), etc

3.3.1.4 Choosing between OLS, Fixed Effects, and Random Effects

According to Greene (2012), the F test and the LM test are the ways to choose which method is appropriate between OLS and FE, OLS and RE, respectively Besides, to choose between FE and RE, the Hausman test will be applied

Thus, firstly, we will consider the F test The F test is used to consider OLS or FE is a method which is more appropriate for the model All the individual intercepts that are zero except one is the null hypothesis of the F test Or saying simplicity, there is no differences among individuals The FE model is appropriate when the null hypothesis is rejected Otherwise, OLS is the best one

Secondly, OLS and RE methods could be compared by using the LM test If the p-value of the LM test after running RE model is less than 5%, then RE method is more appropriate than OLS method

Finally, the Hausman test is used to test between FE and RE methods The null hypothesis

of the Hausman test is that there are differences in coefficients of FE and RE are not systematic

If the p-value of the Hausman test is less than 5%, or in other words, the null hypothesis is rejected,

FE method is the best one Otherwise, RE method is more appropriate

3.3.1.5 Cluster – Robustness Check by Driscoll – Kraay Standard Errors Method

Driscoll – Kraay Standard Errors method is used to estimate OLS/WLS and fixed effects (FE) regression model The purpose of this method is to solve the problem of existence of autocorrelation and heteroskedasticity in the model According to Hoechle (2007), one of the reasons causes autocorrelation is cross – sectional dependence between individuals If there is an existence of autocorrelation in the data, the Driscoll – Kraay standard errors method is more efficient than other methods, such as OLS, White, Rogers, etc Thus, this thesis applies the method

of Driscoll – Kraay standard errors to deal with the problem of the presence of autocorrelation and heteroskedasticity that could bias the standard errors estimates

3.3.2 Heckman Sample Selection Bias Method

A problem which arises when analyzing the impact of food safety standard on rice export from Vietnam is: the sample data tends to not to be selected randomly since Vietnam could decide the destinations where it exports to That means the quantity of Vietnamese exporting rice does

not present for all observations This could be seen as “selection bias” problem This issue is referred in the study of Mangelsdorf et al (2012), which estimates the trade between China and other countries; and Ferro et al (2013), which studies for a group of exporting countries and a group of importing countries Both of these papers mention the self-selection bias problem and use Heckman sample selection model to deal with this issue Guo and Fraser (2014) based on the model of Heckman (1978, 1979) to present two ways to solve this problem: Heckman sample selection bias and treatment effects However, as mentioned in section 3.2, Heckman sample selection bias model is more appropriate to this thesis’s model than the treatment effects Thus, this thesis will choose Heckman sample selection bias model which is the main method to measure the self-selection bias in the data

According to Heckman (1978, 1979), a sample selection bias model derives from two models: outcome regression model to estimate the dependent variable y, and selection model with output value takes from a part of the sample Based on Guo and Fraser (2014), we have:

Regression model: 𝑦𝑖 = 𝑥𝑖𝛽 + 𝜀𝑖, observed only if 𝑤𝑖 = 1,

Selection equation: 𝑤𝑖∗ = 𝑧𝑖𝛾 + 𝑢𝑖, 𝑤𝑖 = 1 if 𝑤𝑖∗ > 0, and 𝑤𝑖 = 0 otherwise

𝑃𝑟𝑜𝑏(𝑤𝑖 = 1|𝑧𝑖) = Φ(𝑧𝑖𝛾) and

𝑃𝑟𝑜𝑏(𝑤𝑖 = 0|𝑧𝑖) = 1 − Φ(𝑧𝑖𝛾), where

- 𝑥𝑖 is a group of exogenous variables that determine 𝑦𝑖

- 𝑤𝑖∗ is a latent endogenous variable

- 𝑧𝑖 is a vector of exogenous variables determining the selection process or the outcome of 𝑤𝑖∗

- Φ( ) is the standard normal cumulative distribution function

- 𝜀𝑖 and 𝑢𝑖 are the error terms of regression model and selection model, respectively

𝜀𝑖~𝑁(0, 𝜎)

𝑢𝑖~𝑁(0,1) 𝑐𝑜𝑟𝑟(𝜀𝑖, 𝑢𝑖) = 𝜌, 𝜌 ∈ (−1,1)

If 𝜌 ≠ 0, the regression results are shown by regression model are biased

Then, the coefficient 𝛽 will be estimated by regressing observable variables (y, x, z and w) Particularly, the estimation of above equations could be estimated by two ways: Heckman Maximum Likelihood method (ML) and Heckman Two – Steps More details will be discussed below

3.3.2.1 Heckman Maximum Likelihood (ML)

In the Heckman maximum likelihood, the correlation of 𝜀𝑖⁡and⁡𝑢𝑖, 𝜌, is not directly estimated and it is estimated through:

atanh 𝜌 =1

2ln⁡(

1 + 𝜌

1 − 𝜌) Delta method is used to calculate the standard error for 𝜌 Similarly, the standard error of

𝜀𝑖, 𝜎, is also not directly estimated, but it is estimated by 𝑙𝑛𝜎 through Heckman Finally, the selectivity effect is not represented by 𝜌 but by 𝜆 = 𝜌𝜎

Essentially, the Heckman maximum likelihood method uses iteration to measure the 𝜃̃ =(𝛽̃, 𝛾̃, 𝜎̃, 𝜌̃)′ to measure the true parameter vector 𝜃0 Since presentation of mathematic equations

of this method is very complicated, thus, this thesis only presents the simple understanding about the Heckman maximum likelihood method (if being interested in more details, could see Bierens (2007))

3.3.2.2 Heckman Two – Steps

Heckman (1979) introduced a Heckman two – steps model to solve the problem of selection bias Particularly,

Step 1: Regress Probit model to obtain the probability of w* event, each of sample will be chosen by an inverse of Mill’s ratio:

𝜆 = 𝜙(𝑧𝑖𝑦)

1 − Φ(𝑧𝑖𝑦)where

- 𝜙(𝑧𝑖𝑦) is a density function for a standard normal distribution

- Φ(𝑧𝑖𝑦) is a distribution function for a standard normal distribution