universidade estadual de campinas instituto de … · tese de doutorado paulo ricardo da silva...

UNIVERSIDADE ESTADUAL DE CAMPINAS

INSTITUTO DE ECONOMIA

PAULO RICARDO DA SILVA OLIVEIRA

TECHNOLOGICAL GAP, DEMAND LAG AND TRADE: A CASE STUDY ON GM-SOYBEANS

HIATO TECNOLÓGICO, LAG DA DEMANDA E COMÉRCIO: UM ESTUDO DE CASO DA SOJA

TRANSGÊNICA

CAMPINAS 2016

UNIVERSIDADE ESTADUAL DE CAMPINAS INSTITUTO DE ECONOMIA



HIATO TECNOLÓGICO, LAG DA DEMANDA E COMÉRCIO: UM ESTUDO DE CASO DA SOJA TRANSGÊNICA

Prof. Dr. José Maria Ferreira Jardim da Silveira – orientador Prof. Dr. David Streed Bullock – co-orientador

Tese apresentada ao Instituto de Economia da Universidade Estadual de Campinas como parte dos requisitos exigidos para a obtenção do título de Doutor em Desenvolvimento Econômico, na área de Desenvolvimento Econômico, Espaço e Meio Ambiente. ESTE EXEMPLAR CORRESPONDE À VERSÃO FINAL DA TESE DEFENDIDA PELO ALUNO PAULO RICARDO DA SILVA OLIVEIRAI E ORIENTADA PELO PROF. DR. JOSÉ MARIA FERREIRA JARDIM DA SILVEIRA.

_______________________________________________ Orientador

CAMPINAS 2016

TESE DE DOUTORADO



HIATO TECNOLÓGICO, LAG DA DEMANDA E COMÉRCIO: UM ESTUDO DE CASO DA SOJA TRANSGÊNICA

Defendida em 23/03/2016

COMISSÃO JULGADORA

A Ata de Defesa, assinada pelos membros da Comissão Examinadora, consta no processo de vida acadêmica do aluno.

DEDICATION

To all those people

whom made the last four years seem too short…

ACKNOWLEDGMENT

I would like to express the deepest appreciation to my committee chair Professor

José Maria who has the attitude and the substance of a genius: he continually and

convincingly conveyed a spirit of adventure in regard to research and scholarship, and an

excitement in regard to teaching. Without his guidance this dissertation would not have been

possible.

I would like to thank you Prof. Bullock for the hospitality and priceless

contribution to improve the dissertation and guide through my fruitful visit to the University

of Illinois.

In addition, thank very much for all my teachers and professors whose dedication

can be seen as very important foundation to this work be built upon. I still believe in a better

world, with better people and education is a reasonable way top get there.

EPIGRAPH

Surely, nothing can be more plain or even more trite common sense than a proposition that innovation […] is at the

center of practically all the phenomena, difficulties, and problems of economic life in capitalist society

(Schumpeter, 1939: 62).

ABSTRACT

The last three decades was marked by several disputes and debates over the international

trade of genetically modified organisms (GMOs). As national regulatory frameworks were

built upon unilateral basis, many conflicts emerged from trading, opening a room for

questioning the adverse effects of technology innovation and adoption on trade and

wellbeing. Therefore, the central aim of this dissertation is to investigate the role of

technological gap and demand lag on trade, in the context of high levels of technology hatred.

The technology gap is the difference or technological distance of techniques employed by

late-movers when compared with technology used by leaders. Likewise, the demand lag may

be understood as the difference or technological distance of techniques employed by

producers in exporting countries and level of acceptance or compatibility in destination

markets. The GM-soybean case is interesting since it comprises all the relevant features to

answer the key questions raised in this dissertation. The concentrate international market – in

terms of both producing and consuming markets – along with distinct technological and

regulatory postures across countries enables the analysis in spite of absence of disaggregated

data on exports of conventional and genetically modified grains. By means of a gravity

equation we empirically estimated the effects of the technology-gap and the demand-lag on

bilateral trade flows of soybeans. In order to find theoretical basis for our analysis, we also

carried out a concise literature review on related trade theories. From the theoretical

perspective this dissertation points to the need of developing models to deal with different

tastes among consumers from different countries and explicitly consider adverse

technological effects in trade – i.e. effects beyond the common relation between innovation,

efficiency and gains of market shares. The results confirm that both technological gap and

demand lag had important impacts on bilateral trade of soybeans. Furthermore, results make

clear that we need better theories to consider cases in which taste differences across countries

play a role in bilateral trade.

Keywords: Bilateral Trade, Technology Gap, Demand Lag, Gravity Equation,

Genetically Modified Organisms (GMO), Agricultural Economics.

RESUMO

As últimas três décadas foram marcadas por inúmeros debates sobre o comércio internacional

de organismos geneticamente modificados (OGM). Como os quadros regulatórios pertinentes

foram desenhados de maneira unilateral, muitos conflitos surgiram no âmbito comercial

abrindo espaço para se questionar o efeito adverso da inovação e adoção tecnológica no

comércio e no bem-estar. Dado isto, o objeto central desta tese é investigar o papel do gap

tecnológico e do lag da demanda no comércio sob o contexto de forte rejeição da demanda. O

gap tecnológico pode ser definido como a diferença ou distância entre a tecnologia utilizada

para produção em países atrasados (late-movers) quando comparada com a tecnologia

adotada pelos país líderes. De forma similar, o lag da demanda pode ser entendido como a

distância ou diferença da tecnologia adotada pelos países exportadores e o nível de aceitação

ou compatibilidade nos mercados de destino. O caso da soja contem todas as características

relevantes para o tratamento das questões levantadas neste trabalho. A concentração da oferta

e da demanda nos mercados internacionais e padrões tecnológicos e regulatórios distintos

entre os países possibilita a análise mesmo sem dados desagregados para exportação de grãos

convencionais e transgênicos. Por meio de um modelo gravitacional nós estimamos de forma

empírica os efeitos do gap tecnológico e do lag da demanda no comércio. Buscando-se bases

teóricas, uma breve revisão da literatura sobre as teorias de comércio foi realizada. Da

perspectiva teórica, a tese aponta para a necessidade de desenvolver-se modelos capazes de

tratar preferências distintas entre os países e considerar explicitamente a possibilidade de

efeitos adversos da tecnologia – isto é, considerar efeitos para além da relação usual entre

inovação, eficiência e ganhos de mercado. Os resultados confirmam que tanto o gap

tecnológico, como a o lag da demanda tiveram impactos importantes no fluxo bilateral de

comércio da soja. Além disto, os resultados apontam para a necessidade de desenvolvimentos

teóricos capazes de tratar de forma mais recorrente casos onde a diferença nas preferências

sejam importantes.

Palavras-chaves: Comércio Internacional, Hiato Tecnológico, Rejeição da Demanda,

Modelo Gravitacional, Organismos Geneticamente Modificados, Economia Agrícola.

LIST OF ILLUSTRATIONS

TABLE 1 – EXPORTS OF SOYBEAN SEEDS 2014 19 TABLE 2 -THE WORLD'S TOP 10 SEED COMPANIES 2010 20 TABLE 3 – WORLD EXPORTS OF SOY PRODUCTS BY COUNTRY (2014) 22 TABLE 4. HECTARE AREA PLANTED WITH GM-SEEDS BY COUNTRY (2013) 24 TABLE 5 – WORLD IMPORTS OF SOYA PRODUCT BY COUNTRY (2014) 29 TABLE 6 – UNITED STATES’ APPROVED GM SOYBEAN (2015) 35 TABLE 7 – ARGENTINA’S APPROVED VARIETIES OF GM SOYBEAN (2015) 39 TABLE 8 – BRAZIL’S APPROVED GM EVENTS OF SOYBEAN 41 TABLE 9- GM SOYBEAN APPROVED IN THE EU FOR FOOD AND FEED 47 TABLE 10 - CHINA’S APPROVED GM SOYBEANS 52 TABLE 11. UNEP-GEF COUNTRIES 56 TABLE 12 – TEST FOR SAMPLE SELECTION AND FIRM HETEROGENEITY BIASES 108 TABLE 13 – DESCRIPTIVE STATISTICS 115 TABLE 14 – ESTIMATES RESULTS 116

FIGURE 1 – EUROPEAN UNION APPROVAL PROCESS OF NEW GMOS FOR FOOD AND FEED 45

CHART 1 – EUROPEAN IMPORTS BY SOURCE (BILLIONS OF USD FROM 1990-2014) 63 CHART 2 – EUROPEAN COUNTRIES’ IMPORTS BY CLUSTER 67 CHART 3 – CHINA’S IMPORTS BY SOURCE (TONES 1990-2014) 69

BOX 1 – SUMMARY OF MAJOR ESTIMATES STEPS ............................................................................................................. 109 BOX 2– MODEL’S VARIABLES DESCRIPTIONS ...................................................................................................................... 111

TABLE OF CONTENTS

Dedication ......................................................................................................................................... 5

Acknowledgment ............................................................................................................................ 6

Epigraph ............................................................................................................................................. 7

Abstract .............................................................................................................................................. 8

Resumo ............................................................................................................................................... 9

List of Illustrations ....................................................................................................................... 10

Introduction .................................................................................................................................... 12

Chapter I - The Private and Public Agents and Controversies Around GMOs .......... 16 1.1 Soybean Industry Organization .............................................................................................. 16 1.2 Countries’ Regulatory Frameworks and Public Opinion Towards GMOs ................ 31

1.2.1 Producing Countries: Regulation, Adoption and Consumers’ Perception ........................ 34 1.2.2 Importing Countries: Regulation, Adoption and Consumers’ Perception ......................... 43

1.3 Remarks .............................................................................................................................................. 58

Chapter II – Technological Effects and Trade Theories ................................................... 60 2.1 Evidences of Technological Effect on International Trade of GMOs .............................. 61 2.2 Trade theories and Dual-market System ................................................................................ 79

2.2.1 The Ricardian Models of Trade and underlying role of technology .................................... 80 2.2.2 Firm Heterogeneity Models .................................................................................................................. 87 2.2.3 Technological Gap and Trade .............................................................................................................. 93

2.3 Theories of Trade and the Case of GMOs ................................................................................. 96

Chapter III - Empirical Estimation .......................................................................................... 99 3.1 Method ............................................................................................................................................. 99 3.2 Data ................................................................................................................................................. 110 3.3 Results and Discussion ............................................................................................................. 114

IV. Conclusions......................................................................................................................... 127

References .................................................................................................................................... 131

12

INTRODUCTION

The genetically modified organisms (GMOs) have been produced and exported

since 1996, when the combination of scientific developments and genetic appropriability

mechanisms enabled the first commercial production of GM-soybeans in the United States

(US). The technology became rapidly available to other producing countries via trade in

technology headed by large multinational seed companies. But, some important consuming

markets have taken contrary positions to the production and consumption of genetically

modified food, arguing mainly about high health, environmental and economic involved

risks.

Together, Brazil, United States and Argentina accounts for approximately 87% of

world exports of soybeans, and the European Union (EU) and China accounts for more than

83% of world imports. Noteworthy, it was possible for growers in United States and

Argentina adopt GM seeds already in 1996. Policymakers in Brazil, on the other hand, took

almost a decade to legalize cultivation from GM seeds. But, on the demand side, many

European countries have been contrary to use of GM seeds in agriculture, encouraging the

raising of trade barriers or even fully banning importation of food or contents deriving from

genetic modified plants. In China, however, in spite of some few limitations to free trade of

GM food, policymakers have passed no rules preventing the country of being a certain

destination for GMOs, partially because of large amounts demanded by the internal

processing and livestock industry.

In sum, the absence of multilateral bodies powerful enough to enforce a

compromise, national policymakers ended up taking unilateral positions, in terms of

approval, coexistence, labeling, and other issues related to GMOs production and trade. As

expected, technology became a new source of trade conflicts lasting until today, as pointed

by many applied studies.

Nonetheless technology has been an issue in trade models at least since the rise of

Ricardo’s model. The baseline model assumes that countries make use of different

technologies – or different production functions –, which become a source of comparative

advantage (CA) leading to different degrees of specialization. More recent works have

advanced in technology and trade mainly under the umbrella of firm heterogeneity and

technology-gap models of trade. Current interpretations of Ricardo and firm heterogeneity

model are similar to the extent they resort to the neoclassical tools. On the other hand,

technology-gap models are based on the building blocs of evolutionary economics.

13

At first, innovation and adoption could be treated as a shock in neoclassical

models, but the problem of different tastes leading to different trade patterns cannot be

addressed straightly. Preferences has been treated as identical and homothetic in the form of

Constant Elasticity of Substitution (CES) utility functions in these models.

However, we have enough evidences to believe that the case studied is not only

impacted by relative productivity changes, but also by differentiated consumers’ perceptions

of the technology across countries. In other words, consumers in different countries can have

different tastes. Also important, they can demand more than goods, i.e., they can choose

among different technologies based not only on price or efficiency criteria. Empirics show

that this can be especially true if they are from high-income countries. That is important from

both the empirical and theoretical perspective.

Many interesting questions arise from this simple case. First, how the

technological change can impact trade flows? Will the first-movers have some advantage in

the presence of technology hatred? Is it possible for a late-mover rip some benefits of late-

adoption? What are the key variables impacting trade in the case of backward effects of

technology? We believe that current theoretical frameworks cannot provide reasonable

answers to these and other related questions.

Our central hypothesis to be assessed in this dissertation is that of technological

innovation leading to a double effect in trade in the presence of asymmetrical adoption and

acceptance of a particular technology. The first effect is the technology-gap in relation to the

most advanced countries, and the second is the demand-lag1 in relation to consuming

markets. The concepts of technology gap, i.e., how advanced or efficient a technology

employed by producers in a country is when compared to most productive technologies

available, and demand lag, i.e., how accepted a technology employed by producers in a

country is by consumers in destination markets in a point of time, are very insightful to the

purposes of this study. To the best of our knowledge these concepts were discussed firstly by

Posner (1961), but have been neglected in the new developments of trade theory, perhaps

because of the lack of cases in which these two effects play such a clear and opposing role.

One of the main goals of this dissertation, therefore, is to evaluate how

technological gap and demand lag have been impacting on the bilateral trade in presence of

unequal technology adoption and significant levels of technology rejection. We also seek to

bring out the bottleneck of the inexistence of treatment for consumer preferences beyond the

1

As it will be discussed latter on this dissertation, we can treat this distance or proximity as the other

geographical costs relating to bilateral trade.

14

“corollary” of identical tastes within and across countries. Specific objectives include

estimating trade elasticities for technology gap and demand lag for the case of soybeans from

1996 and 2012. Also, we will review literature on Ricardian, firm heterogeneity and

technology-gap models discussing their overlays, divergences and contributions to explain

the case of soybean trade. It is important to say this study is primary focused on identifying

possible stylized facts, instead of developing a new theoretical approach to deal with the

flaws in demand side modeling approaches as presented in neoclassical models 2.

We consider this particular experiment very important for several reasons. First,

this seemingly unique case is very passible of recurrence with other agricultural commodities

intend to be used as food and feed components. An example of analogous case, that is high-

income countries revealing preferences for production means, is the increased demand for

certified organic crops, decent work – e.g. no child labor, slavery, etc. – and more

environment-friendly farming activities.

Likewise, technology is becoming more complex also outside the farm gates; at

the same pace consumers are becoming more and more aware of production means used to

manufacturing the goods they acquire. Complex technologies involve social, economic,

ethical, ethnical, religious, environmental and health issues, which can potentially be new

sources of trade conflicts between countries. Biotechnology itself has many others

applications in different industries such as genetic improvement of animal and humans,

development of organic materials and new pharmaceuticals.

Hence, this study can be valuable for both policymakers and private actors since

better understanding the relation across innovation, technology adoption, market rejection

and trade can improve decision-making and consequently wellbeing – adverse trade impacts

will lead to unequal gains and losses across countries and to different actors.

This study contributes to current literature by raising issues about the role of

consumer preferences, and the impacts of these preferences towards certain technologies in

trade. We also advance by estimating the effects of demand lag and technology gap in the

same model. In spite of the importance of these concepts for the study of agricultural

production and trade, the technological-gap hasn’t been explicitly considered in empirical

models. The combination of technology-gap theories, Ricardian models, and gravity

equations are innovative to studies in this area as far as we know.

This dissertation is divided into four more chapters besides this introduction. In

2 This is a very complex subject to be treated in a future research project.

15

chapter I, we bring out figures on GM-food production and outline the trade controversy

foundations. The underlying goal of Chapter I is to show how the soybean global chain

operates, how different agents see the new developments and how country authorities acted

to mitigate increased commercial risks of adoption.

Chapter II introduces the major evidences of changes in the trade patterns from

1996-2012, and how theoretical developments can contribute to shed some light in the case of

trade of GMOs. We present a discussion on technology and trade having as background

models based on Ricardian and technology-gap theorists’ ideas. The main goal of this chapter

is to provide the reader with the grounds to understand what occurred in soybeans market,

and show how theorists that considered technology impacts on trade contributes to answering

some of the questions we have raised so far. Additionally, the inadequacies or absence of

treatment of particular points of our case are indicated for future treatment.

In chapter III we introduce and discuss the empirical results of the gravity

equation highlighting how technological variables – technology-gap and demand-lag –

impacted on GM soybeans trade from 1996-2012. The adopted estimating strategy is also

described in Chapter III, including the gravity equation foundations, limitations and

approaches we have employed in this particular study.

Finally, Chapter VI concludes the dissertation underlining the major findings

from the theoretical and empirical perspectives and retaking points that can be of interest of

policymakers and others decision makers. Also, considerations about open questions to be

addressed in future research programs are outlined.

16

CHAPTER I - THE PRIVATE AND PUBLIC AGENTS AND CONTROVERSIES

AROUND GMOS

In this chapter, we show how important players in international markets of

soybeans held different views of the technology innovation in seed industry. In addition, the

complexity of biotechnology developments and the resulting lack of compromise in terms of

principles and regulation among countries are discussed.

This chapter is split into three more sections. Section 1.1 brings some figures on

main global players and major characteristics of the commercial relations in soybeans

markets. In Section 1.2, we discuss how complexity of biotechnology developments led to

unilateral regulations on labeling and other issues. Section 1.2 explicit the role of

policymakers in managing the commercial risk. Finally, section 1.3 concludes Chapter I.

1.1 Soybean Industry Organization

Soybean production has faced an upsurge in the last decades, mainly after the

1970s. The broad usage of soybean and its by-products – meal and oil – in several industrial

processes is surprising, as long as soybean production in large scale is a relatively new

activity. Historians usually consider the plant gained US farmers interest after 1940s and

South Americans only by the 1980s (HighQuest & Soytech, 2011).

Global consumption has been pushed by high economic growth of developing

world – especially East Asia – and the emergence of new uses, such as feedstock for

biodiesel fuel production. Increased demand along with institutional speculation has been

raising soybean prices remarkably in the last decades.

The primary product of soybean is soybean meal, whereas the RBD (refined,

bleached, deodorized) soybean oil is a secondary product. Indeed, soybean oil is a residual

product of soybean meal manufacturing after solvent extraction process is employed3. Over

the past five decades, soybean meal has become the most available and preferred source of

protein for animal feed manufacturers. The high level of protein (up to 50%), as well as low

fiber content, make it especially good for poultry, swine, aquaculture and dairy and livestock

cattle leading to rapid gain of muscle mass and weight.

Other efficient sources of protein such as fish, meat and bones in spite of high

protein contents have considerable drawbacks. Fishmeal, for example, is significantly more

3 A mechanical press of oil produces high quality oil but is less efficient in protein meal production.

17

expensive and supply is unsteady. Moreover, it is often claimed that poultry acquire fish taste

when feed with fishmeal. Meat and bone meals were the main source of protein for feeding

before soybean meal came into scene. However, many countries have prohibited the use of

these products in breeding especially after the so-called “mad cow” disease. In this way, we

can say that steady supply, some intrinsic characteristics and changes in institutional rules

related to food safety are the major factors behind sustained growth of consumption of

soybean meal. There is no a perfect substitute in meal markets worldwide4.

However, soybean is usually considered an inferior source of oil because of its

low content compared to other major oilseeds. Typically, the low prices and more reliable

supply drives food processors and food service operators to use oil as an ingredient for baked

and fried food, or for cooking oil production. However, drawbacks as trans fat content and

availability of other better sources of vegetal oil (sunflower, canola and rapeseed) make this

market less attractive than market for soybean meal. Noteworthy, developments in biodiesel

industry have attracted more attention to soybean oil.

Although competition for soybean is weak among other vegetal protein sources,

the crop compete for acreage with corn as they convey very similar growing conditions – hot,

wet and humid climate with fertile lands for the highest yields. But, intense competition

between corn and soybeans is more likely to be seen in the U.S. and Argentina, because of

climate issues, rather than other markets such as Brazil.

Agents in the supply chain are settled out all over the world according to the

nature of their activities. They are in general closer to large consuming and producing

markets to rip benefits of reduced transportation costs. However, different processes of

soybean production – such as seed production, growing, processing and manufacturing – can

occur in different countries.

Major soybean producing countries are the United States, Brazil and Argentina.

The larger crushers, feed and food manufacturers and a vast industry for animal breeding is

also based in these countries. However, China and the European Union are the largest

importers of soybean and have a small internal production when compared to volume

demanded by their processing industries.

In general, an input sector (seed and crop protection industries overall), growers,

logistic operators (elevators, crushers, trading companies, etc.) and consumers (feed millers,

food processors and other consumers such as biodegradable plastics and biodiesel producers)

4 The lack of substitutes will be reflected in our empirical exercise showing that imported quantity of other

goods to produce meals is positively related with imports of soybean.

18

are the agents in the soybean value chain. In the case of new conflicting technologies, such as

GMOs innovation, policymakers will also play a strategic role by establishing rules for

national production and consumption.

The combination of high specialization or international labor division, clashing

views of technology by different agents in the value chain and lack of multilateral regulation

are the roots of trade impacts in the case of GMOs.

Seed Industry

The modern and multinational seed industry can be seen as the centerpiece of the

GMO innovation. Actually, very often the history of GM-seeds in agriculture can overlap the

history of global seed industry itself. This industry passed through a drastic transformation in

terms of product portfolios and technology during the last decades.

Scientific and legal developments occurring in the 1980s brought forth a new

technological trajectory – from common methods of seed selection and reproductions to the

hybrid and the genetically modified techniques to selection of desirable intrinsic

characteristics for the seeds (see Wilkinson & Castelli, 2000). The new technology and

appropriability mechanisms required a completely new base of knowledge to the new

developments. Specialists usually agree that we migrated from chemistry to biology-based

innovation trajectories.

Unsurprisingly, the new challenges called for a general movement of merging and

acquisitions, as part of companies’ effort to be competitive in face of the new scientific and

economic requirements. Before 1970s, the seed industry was primary regional with small-

scale firms replicating seeds developed under public domain. A large fraction of famers used

to save their own seeds, since legal framework for appropriability or substantial scientific

innovations weren’t significant at the time. In the 1970s, after transgenia and appropriability

came into scene, seed industry turned out to be more concentrated and international, pursuing

profits not only from their day-to-day replication activities but also from royalties of specialty

varieties (Wilkinson & Castelli, 2000). However, the big clash will come to international

trade chains in 1996, with the first commercial release of Monsanto GM-soybean. Table 1

shows the major exporting countries of soybeans seed in 2014.

19

Table 1 – Exports of soybean seeds 2014

Exporter Trade Value (US) % of world

exports.

Accumulated %

USA 51,120,057.00 30.21 30.21

Argentina 28,041,752.00 16.57 46.78

Canada 26,145,221.00 15.45 62.23

Paraguay 12,242,793.00 7.24 69.47

Malaysia 7,269,400.00 4.30 73.77

Chile 6,242,947.00 3.69 77.46

Uruguay 4,696,756.00 2.78 80.23

Brazil 4,315,733.00 2.55 82.78

India 4,175,584.00 2.47 85.25

Zambia 2,891,612.00 1.71 86.96

Note: Code HS12 120110 was selected to return these data. It was only possible to segregate seed and grain data

internationally after the adoption of HS12 classification system.

Source: Elaborated by the authors based on Comtrade database (2015).

Top 5 countries exporting seeds accounted for 73.77% of world exports in 2014.

As it can be seen, seed companies with higher international presence are based in the United

States and they accounted for 30% of global exports of seed in 2014. The United States are

followed by Argentina (16.57%), Canada (15.45%), Paraguay (7.24%) and others (30%). To

a certain extent the global importance of seed industry in each country can be seen as a

reflection of how long countries took to allow production of GMOs in their territories. Brazil,

for instance, authorized the growth of GMOs a decade later the first commercial release and

stand for the 8th

position in the global seed exporters rank, accounting for only 2.55% of

world trade.

It is worth noting that seed exports are not expressive when compared to soybean

trade because much of the seeds grew in major producer countries are nationally produced.

From a broader perspective, the commercial seed market can be divided into

proprietary and non-proprietary. Simplistically, the first comprehend seeds owned and

marketed by brand companies while the second comprehends seeds traded by local farmers or

plant breeders – the common situation prior to the 1970s. Accordingly to estimates from

International Seed Organization (ISO), 85% of commercial seed market is proprietary

nowadays, a sign of the economic power of these companies in global food chains. Table 2

shows the Top ten firms operating in the seed industry, their annual sales and shares of global

proprietary seed market in 2010.

20

Table 2 -The World's Top 10 Seed Companies 2010

Company – 2010 Seed sales (USD billions) % of global proprietary seed market

Monsanto (US) 4.964 23%

DuPont (US) 3.300 15%

Syngenta (Switzerland) 2.018 9%

Groupe Limagrain (France) 1.226 6%

Land O' Lakes (US) 917 4%

KWS AG (Germany) 702 3%

Bayer Crop Science (Germany) 524 2%

Sakata (Japan) 396 <2%

DLF-Trifolium (Denmark) 391 <2%

Takii (Japan) 347 <2%

Top 10 Total 14,785m 67%

[of global proprietary seed market] Source: ETC Group

As it can be seen, the top 10 firms already accounted for more than 67% percent

of the global proprietary seed market in 2010. This structure has direct impacts on the pace of

technological diffusion, as trade in technology became globally available very quickly and

these companies could control the availability of conventional seeds in the largest agricultural

producing countries as large-scale farmers have been increasingly dependent on the

proprietary seed market. Moreover, they could put together a higher amount of resources to

lobby other agents including policymakers.

The predominant strategy used by giant seed companies has been both heavy

investments in R&D and merger and acquisitions of companies with know-how or large

market-share in areas of interest (Howard, 2009). Innovations have pursued to boost

production yields, cut down production costs and deliver nutritional profiles and value-added

traits desired by consumers – industrial and individual ones.

Monsanto, which had more than 23% of global seed markets in 2010, turns out to

be a key player only from the 1980s. Besides patenting technologies the company acquired

over 50 seed companies between 1996-2013. Some important acquisitions comprehend Delta

& Pine land (1.5 USD billions), Cargill’s5 International Seed Division (1.4 USD billions) and

Holdens’s Fondation Seeds (1.02 USD billions). DuPont major acquisition was Pionner Hi-

Bred, the world’s largest seed company at the time. However, DuPont strategy has involved

more customized agreements with some of the largest remaining independent seed companies

to share germplasm. The other companies adopted very similar strategies to gain market

shares and keep sustained increasing returns (Howard, 2009).

5 Cargill has ownership of companies operating in logistic, crushing and also in feed manufacturing.

21

The economic power of seed companies come not only from economic

concentration but also from their ability to coordinate activities between themselves (mainly

R&D) and vertical integration downstream and upstream to farming activities, such as

partnerships with processors, which allows easy identification of their needs and increased

proximity with farmers worldwide. Monsanto and Cargill partnership is only an example of

seed companies engaging in partnerships with processors. GreenLeaf Genetics is an example

of a partnership between two seed companies to sell foundation seeds6.

These companies also sell or develop partnerships with sellers of crop protection

inputs. Indeed, usually they don’t sell a particular seed but a technological package, meaning

herbicides and insecticides that work along with plants with specific traits.

But, even though seed companies have bargaining power to influence important

actors determining most of the direction and pace of technological innovation, they cannot

fully control end-consumers aversion to the technology and the design of the national

regulatory frameworks.

In sum, this partial power to coordinate agents all over the value chain and the

absence of effective multilateral regulation will be constant source of increased commercial

risk and/or opportunity costs for producers that serve markets with certain levels of

technology hate. In other words decisions taken by the seed companies in terms of the pace

and scope of the new developments, as well their ability to coordinate innovation and

adoption will impact the entire value chain in many ways.

Growers

The major producing countries – namely United States, Brazil7 and Argentina –

accounted for respectively 35.02%, 28.13% and 17.31% of world total soybean production in

2014. These three leading producers were followed by China (3.96%), India (3.41%)

Paraguay (3.23%), Canada (1.96%), Ukraine (1.06%) and Uruguay (1.03%) (FAOSTAT,

2015). Data on shares of world exports are provided in Table 3.

Besides the US, Brazil and Argentina (the G3 hereinafter) a sub-group of

countries also play considerable role in international markets of soya products, given their

expressive exporting and production volumes. This group is made up of Canada, Paraguay,

6 Foundation seed is seed so designated by an agriculture experiment station. Its production must be carefully

supervised or approved by representatives of an agricultural experiment station. It is the source of all other

certified seed classes, either directly or through registered seeds (Rice Knowledge Bank, 2015). 7 Note that Brazil is the only giant producer of soybeans that has no expressive role in international markets of

soybean seeds. As we have argued before, this may be a result of later approval for GMOs cultivation in the

country.

22

India, Ukraine, China, Bolivia and Uruguay - the G7 hereinafter. These countries together

accounted for 20.30% of world production of soybeans in 2013 (FAO, 2015).

Not surprisingly, besides being leaders in soybean exports, the G3 are also the

largest exporters of the two major by-products, soybean meal and oil. Argentina has a

noteworthy position in meal and oil markets, which can be partially explained by the country

tax policy, which favors exports of meal and oil instead of soybeans. On the other hand,

Brazil and US export more soybeans than oil and meal, supplying markets with large

crushing capacity such as European Union (EU) and China.

Table 3 – World Exports of Soy Products by Country (2014)

Soybeans (HS12 – 120190)

Exporter Trade Value (USD trillion) % of world share Accumulated

USA 24.206 41.0% 41.0%

Brazil 23.273 39.0% 80.0%

Argentina 3.748 6.0% 87.0%

Paraguay 2.292 4.0% 91.0%

Canada 1.756 3.0% 94.0%

Uruguay 1.616 3.0% 96.0%

Ukraine 0.701 1.0% 98.0%

Flour and Meal (HS12- 230400 and 120810)

Argentina 11.853 35.61% 35.61%

Brazil 7.000 21.03% 56.65%

USA 5.476 16.45% 73.10%

Netherlands* 2.151 6.46% 79.57%

India 1.292 3.88% 83.45%

China 1.181 3.55% 87.00%

Paraguay 1.109 3.33% 90.34%

Oil (HS12 - 150710 and 150790)

Argentina 3.467 41% 41%

Brazil 1.129 13% 54%

USA 0.806 9% 63%

Paraguay 0.481 6% 69%

Spain* 0.415 5% 74%

Netherlands* 0.375 4% 78%

Germany* 0.328 4% 82%

Bolivia 0.293 3% 85%

Notes: Both codes 230400 and 120810 are used in HS for meal and flour. Code 120190 was adopted after HS12 version and exclude

soybean seeds (120110), previously classified in a single code 120100. Oil different codes only discriminate between refined (150790) and crude (150710). Countries marked with “*” are processors or distribution ports as they don’t produce meaningful amounts of soybean.

Source: Elaborated by the authors based on Comtrade database (accessed in 2015).

23

Specialists usually agree that soybean production in Brazil and Argentina has cost

advantages when compared to production in the United States. That would be a reason behind

the increasing market share of Brazil and Argentina over the US shares (HighQuest &

Soytech, 2011). However, there are other differences explaining how importers (crushers and

processors) decide upon where to source their soybeans such as the level of foreign material,

moisture, beans integrity, protein and oil content.

The discrimination between GM and conventional seeds are the additional setback

the seed technology industry put to growers’ decision making. Actually, the modern version

of this problem is set in term of which GM variety is being produced and approved for

importation in different countries. Considering low variability of prices and traditional

quality standards across major producing countries, such a discrimination seems to have been

played an underlying role in deciding from where to source soybean in the last decades – as

we are arguing in this Ph.D. dissertation.

Brazil was the only country from the G3 group that has taken almost a decade to

approve production of GMOs internally. There are two equally valid explanations for that.

First, the legal moratorium was effective in prohibiting farmers to grow GM-seeds until 2005.

We know it can be only partially true on account of growers in south of Brazil have started

planting GM-seeds by the late-1990s, making use of pirate seeds smuggled from Argentina.

Second, the first GM-seeds weren’t good enough for tropical regions, performing poorly in

terms of yield. This explanation seems very plausible if we consider the lobby of large

farmers in Brazil. As soon as these farmers could see any significant rewards in growing

genetically modified seeds in Brazilian Middle West, they would fight for GMOs

liberalization – as happened in the early-2000s.

On the other hand, U.S. and Argentina growers are ripping the benefits and paying

the costs of planting GM-seeds since 1996. U.S. has notably higher average yields per hectare

than Brazil, as a result of more capital-intensive production means. The existence of a large

producer supplying large amounts of conventional soybean along with adverse markets

demanding GM-free products created a dual-market system – i.e. a market for GMOs and

another for conventional crops. However, with the Brazilian Biosafety Law of 2005, dual-

market basis will be significantly weakened by a severe reduction on supply of conventional

soybean. Adoption in Brazil was rapid in pace, making the country the second largest country

in terms of planted area with GM-crops already in 2013, as can been in Table 4.

24

Table 4. Hectare area planted with GM-seeds by country (2013)

Rank Country Area (millions of hectares 2013)

Area (millions of hectares 2013)

Crops

1 United States

70.1

73.1 Maize, soybeans, cotton, canola, sugar beet,

alfalfa, papaya, squash

2 Brazil 40.3 42.2 Soybean, maize, cotton

3 Argentina 22.9 24.3 Soybean, maize, cotton

4 India 11 11.6 Cotton

5 Canada 10.8 11.6 Canola, maize, soybean, sugar beet

6 China 4.2 3.9 Cotton, papaya, poplar, tomato, sweet pepper

7 Paraguay 3.6 3.9 Soybean, maize, cotton

8 Pakistan 2.8 2.9 Cotton

9 South Africa 2.9 2.7 Maize, soybeans, cotton

10 Uruguay 1.5 1.6 Soybean, maize

11 Bolivia 1 1 Soybean

12 Australia 0.6 0.5 Cotton, canola

13 Philippines 0.8 0.8 Maize

14 Myanmar 0.3 0.3 Cotton

15 Burkina

Faso 0.5

0.5 Cotton

16 Spain 0.1 0.1 Maize

17 México 0.1 0.2 Cotton, soybean

18 Colombia 0.1 0.1 Cotton, Maize

19 Chile <0.1 <0.1 Maize, soybean, canola

20 Honduras <0.1 <0.1 Maize

21 Portugal <0.1 <0.1 Maize

22 Czech

Republic <0.1 <0.1 Maize

23 Poland <0.1 <0.1 Maize

24 Egypt <0.1 <0.1 Maize

25 Slovakia <0.1 <0.1 Maize

26 Costa Rica <0.1 <0.1 Cotton, soybean

27 Romania <0.1 <0.1 Maize

Source: James (2014)

Table 4 shows that the three largest soybean producers also figures among the

largest world producers of GMOs. However, countries adopted technology in different times,

for different crops and varieties as we are going to discuss in more details in the next section.

25

Considering their strategic position in the supply chain, farmers acquire seeds

from input industries to carry out their core activities. Usually, along with seeds farmers are

also choosing part of other technologies. Although farmers can be substantially different

worldwide in terms of size or technology-level, the new shape of soybean value chain has

certainly decreased their scope of decision, especially in terms of choosing across different

varieties of the same crop.

Decisions are made considering legal approvals at home, costs – including the

overall costs of technological package – yield delivered given the environment conditions,

and crusher’s acceptance of available varieties. In other words, seed industry, crushers, food

and feed processors, and policymakers have been playing a more central role in technology

development, adoption and acceptance. By the end of the day these agents determine the

premiums and penalties related to different seeds adoption, and growers respond accordingly

to their rationale of higher returns. We can sum up the role of growers in global trade as to

produce as many as soybeans as possible.

Grain elevators, Domestic Crushers and Trading Companies

Grain elevators are usually complete receival points comprehending activities like

receiving, testing, weighting and storing grains until selling them to crushers or other

elevators. They are called this way because they scoop up grains from lower level into silos

or other storage facility. There are many types of elevators, but from a broader perspective

they can be classified as country and exporting elevators – also called exporting terminal.

Country elevators sell to other larger elevators in the country or processors

whereas export terminals sell beans to trading companies, international processors or end-

consumers in international markets. As this study deals with soybean trade we are mainly

concerned about GM-soybeans being exported to be processed in countries averse to the

technology.

Domestic processors or crushers are companies carrying out all activities from

handling and elevating to extraction of soybean meal and crude degummed oil. As cited

before, hexane extraction is the most common method employed to produce soybean meal

and have oil as a residue8. Domestic processors buy soybeans from country elevators and sell

soybean meal and oil to end-consumers in the country or abroad – directly or through trading

companies.

8 Other extraction methods like mechanical press can delivery a higher quality oil but affects productivity of

soybean meal.

26

Export terminals, in turn, acquire soybeans from country elevators or growers to

sell to trading companies or direct to countries with high crushing capacity – such as the EU

and China. Very often country elevators, crushers, exporting terminals and trading companies

are controlled by the same corporations making difficult to breakdown their activities and

commercial relationships. ADM, Bunge, Cargill, Louis Dreyfus, AGP and CHS are important

players controlling grain distribution activities all over the world.

Countries differ substantially in terms of presence and size of these agents. Brazil

and Argentina, for instance, are more likely to sell the grains to domestic processors or export

terminals as these countries have less complex logistic operations. Otherwise, U.S. growers

usually sell to country elevators, making soybean pass through at least 3 or 4 operators before

entering the exporting market.

The better infrastructure in the U.S. cuts down logistic costs but has some quality

drawbacks because of the higher levels of foreign materials and breakage of beans, resulting

from over handling. On the other hand, U.S. growers gain as they can sell FOB9 to elevators

operating nearby their farmers, which requires much less expertise in logistic issues. Many

specialists consider logistic advantages partially pay down lower production costs in South

America – due to relatively lower wages and land prices.

Just to provide a better picture of the relative importance of these agents, let’s take

into account some estimates from the American Soybean Association (ASA). According to

them, U.S. processors purchased on average 55% of all soybeans domestically produced from

2006 to 2010. Export terminals purchased 36% and cattle breeders 5%10

. Processors, likewise

country elevators, are mainly sited closer to large producing or consuming areas around the

world.

Export terminals, otherwise, are located in ports of easy access to growers or

elevators in major producing regions. They seek to minimize overall costs of transportation

and storage at the same time they mitigate the transaction costs related to moving soybean

from a country to another. Thus, these companies have offices, facilities and others branches

in both major producing and consuming countries and often sell to their own affiliates.

The concentration and hierarchical control in distribution can be partially

explained by high risks involved in these activities. Besides soybean is normally considered a

commodity, transactions in the segment can be often considered complex. Indeed,

9 Free on board price.

10 The remaining 2% was kept as elevators carryovers.

27

coordination in agro-food chains in general has been considered complex mainly because of

high risks involved in farming.

Considering the risks from the standpoint of transaction costs, the more

agriculture assets become specific the more complex transactions will be. Frequency and

uncertainty are also other two issues to be considered. Even before the GM-seeds came on

the scene, other qualities criteria, the significant economies of scale – calling for operating at

full capacity – overall level of uncertainty, and market concentration were far enough to

make soybean chains complex. Nonetheless, it is very clear that GM-seeds created one more

source of asset specificity. For each different variety of soybeans emerging a result of

innovation in seed industry, we have an increased range of similar but different products in

the marketplace11

.

Therefore, logistic operators and soybean processors – often the same agent – are

managing significant part of the risks related to production and trade of soybean. They

depend on many independent growers to keep their activities at full capacity. They usually

play a very important role in assuring or achieving quality standards and in IP (identity

preservation). They have a strategic position not only between end-consumer (domestic and

international ones) but also between input industry and farmers – being common to

processors and seed companies commit themselves in partnerships. Last but not least, they

are subjected to different national regulatory frameworks to carry their grains from producing

to processing countries.

In sum, these corporations are in charge of managing national and international

growers decisions to guarantee their supply of soybeans to operate at full capacity. In the

relationship with seed industry they signalize what their needs are, in terms of output traits –

higher oil or protein content for instance – and also input traits demanded by growers given

their closeness to them. Other important role is to align premiums they pay to growers that

are producing varieties they demand and royalties paid by growers to seed companies.

Moreover, they are also a key agent in monitoring royalties’ payment as they can easily test

varieties they acquire and request documents to proof that seeds were obtained by legal

means.

11

This idea is not only realistic but convenient for theoretical and empirical analysis, since monopolistic

competition is a good representation for markets like this.

28

External Processors, Feed millers and Food Processors

After been grown and/or processed soybean and by-products follow to internal

consumption or exports. As we have seen, soybeans are mostly used to produce feed and food

contents. Feed is produced from soybean meal and used as protein source especially for

poultry, swine, aquaculture and diary cattle breeding. Food ends include consumption of

beans, flours, oil, among others, and alternative source of protein. Many food products have

soybean products in their composition, for example the chocolate with soy lecithin.

More recently, soybean has been used as feedstock for biodiesel production and

raw material in biodegradable plastic. Noteworthy, sometimes companies operating elevators,

processing and trading activities also controls or have ownership of feed manufacturing

facilities and other end-consumers – e.g. the Cargill corporation operating in feed sector.

In general, countries with crushing capacity and low production of soybeans will

be largest consuming markets of soybeans. That is the case of the European Union and China.

Countries with large presence of feed processors, as a result of large animal breeding sectors,

but without crushing capacity will be key markets to soybean meal. Lastly, large presence of

food processors, biodiesel industry and low crushing capacity will determine the major

importers of soybean oil. Very often countries are important players for both soybean and by-

products markets given general gains of scope. Table 5 shows world largest importers, value

of trade, percentage and accumulated percentage of world imports for soybeans, soybean

meal and oil.

29

Table 5 – World Imports of Soya Product by Country (2014)

Soybeans (HS12 – 120190)

Exporter Trade Value (USD

billions)

% of world share Accumulated

China 40.265 69.13% 69.13%

EU-27 8.277 14.21% 83.34%

Mexico 2.071 3.56% 86.89%

Japan 1.831 3.14% 90.04%

USA 1.149 1.97% 92.01%

Turkey 1.119 1.92% 93.93%

Thailand 1.076 1.85% 95.78%

Egypt 1.057 1.82% 97.60%

Flour and Meal (HS12- 230400 and 120810)

EU-27 13.311

51.92% 51.92%

Thailand 1.676

6.54% 58.46%

Japan 1.057

4.12% 62.58%

Philippines 0.974

3.80% 66.38%

Mexico 0.826

3.22% 69.60%

Algeria 0.822

3.21% 72.81%

Malaysia 0.790

3.08% 75.89%

Egypt 0.644

2.51% 78.41%

Peru 0.602

2.35% 80.76%

Canada 0.515

2.01% 82.77%

Oil (HS12 - 150710 and 150790)

India 1.985

29.10% 29.10%

China 1.092

16.01% 45.11%

EU-27 0.999

14.65% 59.76%

Algeria 0.566

8.30% 68.06%

Peru 0.335

4.92% 72.98%

Mexico 0.190

2.79% 75.77%

South Africa 0.158

2.33% 78.10%

Dominican Rep. 0.129

1.90% 80.00%

Pakistan 0.119

1.76% 81.76%

Ecuador 0.119

1.75% 83.50%

Notes: Codes 230400 and 120810 have been used in international classification for meal and flour. Code 120190 was

adopted after HS12 version and exclude soybean seeds (120110), previously classified in a single code 120100. Oil

different codes only discriminate between refined (150790) and crude (150710)

Source: Elaborated by the authors based on Comtrade database (accessed in 2015).

30

The top 5 destinations together for each product accounted for 92.01% of

soybeans, 69.60% of meal and flour and 72.98% of oil world imports in 2014. This level of

concentration makes decisions in major markets a big deal for production and exporting

decisions taken in sourcing countries – what we are naming as commercial risk. In this way,

the EU large importing volumes combined with high levels of technology “hatred” is

undoubtedly a source of trade conflicts (see Anderson & Jackson, 2004).

China (69.13%), Europe Union (14.21%), Mexico (3.56%) and Japan (3.14%) are

currently the world major importers of soybeans. These countries have minor levels of

soybean production, excluded China that is the fourth largest producer in the world, but still a

net major importer. The European Union is also the major destination for soybean meal and

flour (51.92% of world’s imports in 2014). Major oil importers are India, China and

European Union, respectively.

Crushers in these countries are highly dependent on imports of soybean to supply

soybean meal to feed and food manufacturers operating internally. Usually, these countries

are also dependent on imports of soybean meal and oil, as national crushing capacity is not

enough to fulfill internal demand by feed and food manufacturers.

From the perspective of commercial risks, end-consumers opposing to the

technology in some of the major importing countries are the agents creating additional risks

for the supply chain given the presence of genetic engineered seeds in marketplace, even in

the absence of legal barriers. Again, because of contrary position of manufacturing industry

– a result consumers’ aversion of soy products along with strict government regulations –

logistic operators and processors in sourcing countries have to plan their production strategies

also based on technological rejection worldwide.

As we have been arguing consumers’ aversion to products of GMOs and

government policies towards the technology adoption will play an underlying role in the level

of commercial risks. Indeed, we defend that these two agents sustain many of the controversy

established in the production chain. If not by the commercial risk related to the law and the

end-consumer aversion to the technology we believe that processors and manufacturers

would have no reasons to keep skeptical to the technology. These two additional burdens will

be treated in the next section.

31

1.2 Countries’ Regulatory Frameworks and Public Opinion Towards GMOs

In this section we introduce the way in which the countries have been regulating

GM production and trade during the past three decades of agricultural biotechnology

developments. At first, regulation can be thought from the unilateral and multilateral

perspectives. The latter is especially necessary when norms placed by national authorities in

one country can potentially affect business and citizens as a whole in another country.

History has shown that many conflicts can emerge from international affairs,

including those propelled by environmental, social, economic, ethical matters. The nature of

the conflict should be taken into account, as the drivers and remedies for each type of dispute

will depend on it.

Trade agreements determining sanitary and other technical standards for example,

have mostly employed scientific-based approaches, whereas environment-based agreements

have employed the precautionary principle to establish international standards and norms

(Winham, 2009).

The precautionary principle states that the introduction of a new product or

process, whose ultimate effects are disputed or unknown, should be restrict. In the absence of

precautionary principle the matter can be treated under more pragmatic approaches, such as

the substantial equivalence principle. Substantial equivalence states that a product containing

comparable amounts of a few basic components, such as proteins, fats, and carbohydrates as

its counterpart, should be considered as safe as the comparable one.

Modern biotechnology is particularly complex because its multi-issue profile.

Ethical, environmental, economic and health issues have basically the same weight in terms

of pros and cons. Consequently, regulation was spread out in few different agreements based

on distinctive matters relating to this same technology.

These agreements, however, in spite of incomplete congruence in some specific

points, are equally valid and their principles and guidelines can be claimed accordingly to the

parts’ particular views of the technology. The contrasting scopes can partially explain the

mismatches between rules enacted in World Trade Organization (WTO) agreements and the

Cartagena Protocol for Biosafety (CPB). These are considered the main agreements on

international regulation of modern agricultural biotechnology.

The Cartagena Protocol, opened for signatures in May/2000, is primarily focused

on environmental risks. So, it is based on the precautionary approach and seeks to

32

“(…) contribute to ensuring an adequate level of protection in the field of

safe transfer, handling and use of living modified organisms resulting from

modern biotechnology that may have adverse effects on the conservation

and sustainable use of biological diversity (…)”(Secretariat of the

Convention on Biological Diversity, 2000).

Even though risks to human health are taken into consideration – as declared in

other parts of the protocol – focus is kept on environment issues and monitoring of living

modified organisms. Living modified organisms means that the organisms are capable of

transferring or replicating genetic material. Currently 170 countries have signed the protocol,

being worth noting that the United States and Argentina have not ratified it yet. This fact

alone is enough to bring out the weakness of the Protocol to reach compromise among big

players when it comes to commercial disputes.

On the other hand, the World Trade Organization (WTO) agreements and their

reference to the Codex Alimentarius are clearly more scientific-based. Unsurprisingly, they

are primarily focused on assuring harmonious trade of products involving sanitary,

phytosanitary and other technical standards instead of potential risks involving environment

and ethical issues.

Of course, these differences in scope and principles alone will be an obstacle to

achieve harmonization of norms and procedures. Some countries can base their regulatory

framework in a more scientific-based approach at the same time others can claim for the right

of enacting precautionary measures.

One example of these incongruences is the establishment of minimum and

maximum levels of protection and the possibility of claiming the so-called safeguard

measures. The scientific-based approaches under substantial equivalence principle tend to

offer a very limited space for safeguards and set up standard ceilings whereas precautionary

approach, by taking into account uncertainty, set up standard floors.

Both Technical Barriers to Trade (TBT agreement of WTO) and Sanitary and

Phytosanitary Measures (SPS agreement of WTO) came into force in 1995, at the very

beginning of commercial production of GMOs. Time mismatch of regulation and conflicts

may partially explains why GMOs are not straightly treated in the scope of these agreements.

The SPS establishes a “…multilateral framework of rules and disciplines to guide

the development, adoption and enforcement of sanitary and phytosanitary measures in order

to minimize their negatives effects on trade (SPS agreement text, emphasis mine) ”.

Furthermore, it is stated that relevant international organizations – such as the Codex

33

Alimentarius Commission – will provide the bases to guide the building of regulatory

frameworks. The TBT seeks to ensure that

…technical regulations and standards, including packaging, marking and

labeling requirements, and procedures for assessment of conformity with

technical regulations and standards do not create unnecessary obstacles to

international trade (TBT agreement text, emphasis mine).

The UN Food Agriculture Organization (FAO) and the World Health

Organization (WHO) established the Codex Alimentarius Commission or “Food Code” in

1962. Membership in Codex is open to all member nations of the United Nations (UN) and

currently 165 countries participate. It has been developing a series of guidelines, including

labeling rules to GMOs that may bring more harmonization to the field.

However, the current scenario makes clear that multilateral regulation failed in

fully organizing the production and trade of GMOs. The complexity involving biotechnology,

the lack of clear general standards for dealing with GMOs production and trade, the existence

of a couple of agreements not always in congruence with one another, as well as the lack of

more efficient sanction mechanisms, made country-level regulation the relevant part for

understanding the state of affairs of GMOs.

National regulations differs in restrictiveness degree, type of approach or principle

(scientific or precaution-based), type of basis (scientific or political), and liability (private or

public sector)(Josling, Orden, & Roberts, 2004). A range of questions is treated by national

regulations, such as approval process, coexistence rules, labeling regimes, traceability,

liability schemes, and others. For our purposes approval process in producing and importing

countries and labeling rules in importing countries in particular will be central questions.

We see countries regulatory profile as a result of internal disputes across different

groups of interest, such as firms, consumers, Non Governmental Organizations (NGOs) and

the government itself. That is why the largest agricultural countries tend to take more

pragmatic regulatory posture whereas net importers of agricultural goods with high-income

levels tend to implement more strict rules (P. R. S. Oliveira, Silveira, Magalhães, & Souza,

2013).

Accordingly, policymakers took two important dimensions into account when

considering commercial risks of producing GMOs. First, the approval of many varieties can

raise risks of producing events not approved in every destination countries. Second, even

34

with just a few approved varieties, if adoption rates are too high and logistic capacity for IP is

poor the country can have problems in exporting to countries that are strongly averse to

technology. Indeed, in the 1990s the pertinent dilemma was between the GM-Free vs GM as

many manufacturers preferred to import conventional products as a precautionary measure in

face of large supply amounts. After 2005, when Brazil finally allowed farmers to legally

grown GM crops, the dilemma became one of asymmetrical approval, i.e. a risk of approving

varieties not allowed for consumption in destination markets.

Needless to say, to keep IP of different GM varieties is a huge challenge

considering current transportation structure based on large gains of scale. From a

policymaking perspective, it is simpler to deny approval of varieties not accepted in the most

important destination markets.

These polices implications show how policymakers became very important to

keep commercial risks at a manageable level to middlemen – export elevators, processors

and trading companies being very often owned by the same corporations. In other words, the

absence of certain level of regulation and a less concentrate industry would make risk

management much more complicated than it actually was.

1.2.1 Producing Countries: Regulation, Adoption and Consumers’ Perception

As we have been arguing the aftermaths of ongoing debates in each country

resulted in different regulatory frameworks for global production and trade of GMOs. Only

by considering the G3 countries we can see differences in regulation that will shape the new

pattern of trade. The U.S. has approved a larger number of soybean varieties for cultivation

and other processing uses when compared to Brazil and Argentina. The country also accounts

for the highest global shares of GMOs production over the total agricultural outcome – i.e.

considering all crops to which GM varieties are available (see table 4).

Table 6 shows the events, traits, developing company, purpose and year of

approval for soybean varieties in the U.S..

35

Table 6 – United States’ Approved GM Soybean (2015)

Event Trait Company Authorized

For

Year

Food and

Feed

Year

Cultivation

ACS-GMØØ2-9*** Glufosinate herbicide

tolerance

Bayer CropScience

(including fully and partly

owned companies)

Cultivation n/a 1996

DD-Ø26ØØ5-3

Modified oil/fatty

acid, Antibiotic

resistance, Visual

marker

DuPont Pioneer

Food and

Feed/Cultiva

tion

1997 1997

A5547-127 Glufosinate herbicide

tolerance Bayer CropScience

Food and

Feed/Cultiva

tion

1998 1998

GU262***

Glufosinate herbicide

tolerance, Antibiotic

resistance

Bayer CropScience

(including fully and partly

owned companies)

Food and

Feed/Cultiva

tion

1998 1998

MON 89788 Herbicide Tolerant Monsanto

Food and

Feed/Cultiva

tion

2007 2007

DP-3Ø5423-1

Sulfonylurea

herbicide tolerance,

Modified oil/fatty

acid

DuPont Pioneer

Food and

Feed/Cultiva

tion

2009 2009

MON 87705 *

Herbicide Tolerance

+ Modified Product

Quality

Monsanto

Food and

Feed/Cultiva

tion

2011 2011

SYHTØH2 ***


tolerance, Mesotrione

Herbicide Tolerance

Bayer CropScience and

Syngenta Cultivation n/a 2014

DAS-44406-6 *


tolerance, Glyphosate


2,4-D herbicide

tolerance

Dow AgroSciences LLC

Food and

Feed/Cultiva

tion

2014 2014

DAS-81419-2 *


tolerance,

Lepidopteran insect

resistance


Food and

Feed/Cultiva

tion

2014 2014

40-3-2 Glyphosate herbicide

tolerance Monsanto

Food and

Feed/Cultiva

tion

1995 1993



Food and

Feed/Cultiva

tion

1996 1998

MON 87701 Lepidopteran insect

resistance Monsanto

Food and

Feed/Cultiva

tion

2010 2011

Continued….

36

Event Trait Company Authorized

For

Year

Food and

Feed

Year

Cultivation

MON 87708* Dicamba herbicide

tolerance Monsanto

Food and

Feed/Cultiva

tion

2011 2015

MON 87769*

Glyphosate herbicide

tolerance Modified

oil/fatty acid

Monsanto

Food and

Feed/Cultiva

tion

2012 2011

MST-FGØ72-3***


tolerance, Isoxaflutole

herbicide tolerance

Bayer CropScience and

MS Technologies LLC

Food and

Feed/Cultiva

tion

2012 2013

DAS-68416-4*


tolerance, 2,4-D

herbicide tolerance


Food and

Feed/Cultiva

tion

2014 2011

DAS-68416-4 X

MON 89788*


tolerance, Glyphosate


2,4-D herbicide

tolerance


Refer to

Individual

Event Status

n/a n/a

DAS-81419-2 X

DAS-44406-6 *


tolerance ,

Lepidopteran insect

resistance


Refer to

Individual

Event Status

n/a n/a

DP-3Ø5423-1 X

GTS 40-3-2


tolerance,

Sulfonylurea


Modified oil/fatty

acid

DuPont Pioneer

Food and

Feed/Cultiva

tion

n/a n/a

MON 87701 X

MON 89788 **


tolerance Modified

oil/fatty acid

Monsanto

Refer to

Individual

Event Status

n/a n/a

MON 87705 x MON

89788


tolerance

Modified oil/fatty

acid

Monsanto

Refer to

Individual

Event Status

n/a n/a

MON 87708 X

MON 89788 *


tolerance

Dicamba herbicide

tolerance

Monsanto

Refer to

Individual

Event Status

n/a n/a

MON 87769 X

MON 89788


tolerance

Modified oil/fatty

acid

Monsanto

Refer to

Individual

Event Status

n/a n/a

Note: * not commercialized ** imported ***status not reported

Source: Prepared by the authors based on Cera (2015), CropLife International (2015), ISAAA (2015)

and GMO Compass (2015)

From a total of 24 approved varieties, major seed companies have been reporting

that at least 10 are actually been produced in the United States12

. The U.S. is often

acknowledged as the major advocate in favor of increased marketing liberalization for

GMOs.

U.S. regulatory framework has a low degree of restrictiveness, reflecting the

adoption of substantial equivalence approach combined with well-aligned semi-autonomous

12 Commercial status is reported by seed companies participating in the CropLife International website – BASF

Plant Science, Bayer CropScience, Dow AgroScience LLC, DuPont Pioneer, Monsanto and Syngenta. There are cases in

which approved varieties are not updated in the website, and so, it is not possible to obtain the commercial status of them.

http://www.isaaa.org/gmapprovaldatabase/gmtrait/default.asp?TraitID=1&GMTrait=Glufosinate%20herbicide%20tolerance

http://www.isaaa.org/gmapprovaldatabase/gmtrait/default.asp?TraitID=1&GMTrait=Glufosinate%20herbicide%20tolerance

http://www.isaaa.org/gmapprovaldatabase/gmtrait/default.asp?TraitID=8&GMTrait=Modified%20oil/fatty%20acid

http://www.isaaa.org/gmapprovaldatabase/gmtrait/default.asp?TraitID=8&GMTrait=Modified%20oil/fatty%20acid

37

government bodies. An important characteristic of U.S. regulation is the basis on the nature

of the products, rather than the process in which they were produced –a natural result of

substantial equivalence principle.

Government regulates biotech plants through the Coordinated Framework for

Regulation of Biotechnology, established as a formal police in 1986. The government

agencies responsible for oversight of the products of modern biotechnology are the USDA’s

Animal and Plant Health Inspection Service (USDA-APHIS), the U.S. Environmental

Protection Agency (EPA) and the Department of Health and Human Services’ Food and Drug

Administration (FDA). Depending on its characteristics a product may be subject to the

jurisdiction of one or more of these agencies.

USDA-APHIS is primary concerned with potential plant diseases resulting from

genetically engineered plants and environmental impacts related to growing new varieties.

The Plant Protection Act is the underlying framework in which USDA decisions are based.

FDA is responsible for assuring food safety overall, including GM-food. The regulatory

framework is based upon the Federal Food, Drug, and Cosmetic Act and the Public Health

Service Act. EPA controls the use of all crop protection products regardless the form they are

presented. As GMOs comprises insect resistance or/and herbicide tolerance traits, the agency

also carry out mandatory analysis of new developments, under the Federal Insecticide,

Fungicide and Rodenticide Act and the Toxic Substances Control Act.

With regards to adoption, 54% of total soybean produced in 2000 was genetically

engineered according to data from the National Agricultural Statistics Service (NASS). This

share reached 87% in 2005 and 94% in 2015. High adoption shares and number of approved

varieties suggest that commercial risks didn’t prevent U.S. policymakers of keeping a

favoring scenario for GMOs in the country.

U.S. citizens’ opinion on GMOs has been considered mixed, but studies

considering polls carried out between 2001 and 2006 commonly state that:

i. Public knowledge and understanding of biotechnology remains relatively low;

ii. Consumers know little about the extent to which their foods include genetically

modified ingredients;

iii. While support for GM foods has been stable, opposition has softened and opinions

on safety remain split;

iv. Animal cloning evinces much stronger opposition than does modifications of

plants;

38

v. Consumers look to those closest to them as trusted sources of information on GM

foods and biotechnology;

vi. Religious belief has some impact but it is not a key source of variation in public

attitudes towards biotechnology (The Mellman Group, 2006).

According to these polls only 41% of U.S. citizens claimed to have heard a “great

deal” or “some” about genetically modified food in 2006. The remaining 59% said they heard

“not too much” or “nothing at all” about it – the share of people responding “nothing at all”

was on average 25% of the sample. Just 26% of Americans favored the introduction of

genetically modified foods into the U.S. food supply, and 58% opposed. In 2006, opposition

declined to 46% although support continued almost at the same level (27%)(The Mellman

Group, 2006).

On the other hand, farmers and scholars have been reporting positive views of

modern biotechnology advances. Groups in the US opposed to GMOs include some

organizations related to environmental, organic farming and consumers’ rights (LOC, 2014).

The question of mandatory labeling has been central accordingly to a recent poll

showing that 93% of consumers in U.S. advocates in favor of mandatory labeling (Kopicki,

2013). However, authorities are keeping the regulation in place, which allows industry to

voluntarily use labels indicating whether foods have or have not been derived from

genetically engineered plants. FDA provides guidance for voluntary labeling.

Argentina, in spite of passing by a rapid adoption of GM seeds in agriculture, has

approved just a few varieties of soybean for cultivation, decreasing the commercial risks in

terms of asymmetrical approval – i.e. a lag between cultivation approval at home and

food/feed approvals abroad. Table 7 presents the list of Argentina’s approved soybean

varieties for different purposes.

39

Table 7 – Argentina’s Approved Varieties of GM Soybean (2015)

Event Trait Company Authorized For

Year for

food and

feed

Year for

cultivation

40-3-2 Glyphosate herbicide

tolerance Monsanto

Food and

feed/Cultivation 1996 1996

A5547-127* Glufosinate herbicide


Food and


A2704-12* Glufosinate herbicide


Food and


MON 87701 X

MON 89788


tolerance Monsanto

Food and


BPS-CV127-9 ** Imidazolinone herbicide

tolerance. BASF Inc.

Food and


DAS-44406-6*


tolerance, 2,4-D herbicide

tolerance

Dow AgroSciences

LLC

Food and


DP-3Ø5423-1

Sulfonylurea herbicide

tolerance, Modified

oil/fatty acid

DuPont Pioneer Food and

Feed/Cultivation 2015 2015

DP-3Ø5423-1 X

GTS 40-3-2*


tolerance, Sulfonylurea


Modified oil/fatty acid

DuPont Pioneer Food and


Note: * not commercialized ** status not reported

Source: Prepared by the authors based on Cera (2015), Biotechnology Industry Organization (2015),

ISAAA (2015) and GMO Compass (2015)

In spite of approving the Roundup Ready TM

(40-3-2) soybean from Monsanto in

1996, Argentina authorities took 15 years to approve cultivation of a second different variety,

the Liberty Link TM

(A5547-127 and A2704-12) from Bayer. This kept the asymmetry

between approved varieties in source and destination quite low, as Roundup Ready TM

was

broadly accepted worldwide, at least from the legal standpoint. Indeed, according to major

seed companies Argentina has been producing only the stacked variety MON 87701 X MON

89788, since 2003.

However, the GM soybean share reached 90% of the total produced in the first

seven seasons after 1996. Planted area with GM soybean was already 98% of total area with

the crop in 2004/2005 seasons. The great adoption rate is usually credited to weak

mechanisms to guarantee property rights and agricultural benefits perceived by Argentinean

farmers as a whole (Finger & Hartmann, 2009).

The Law on Seeds and Phytogenetic Creations (Ley de Semillas y Creaciones

Fitogéneticas), the Law on the Promotion of the Development and Production of Modern

Biotechnology (Ley de Promoción del Desarrollo y Producción de la Biotecnologia

Moderna) and administrative acts issued by the Secretary of Agriculture, Livestock, Fisheries

and Food (SAGPA) regulate GMOs issues in Argentina.

40

The SAGPA is responsible for oversight the release and commercialization of new

GMOs. Approval process count with assistance of an expert advisory committee, and

comprehends assessments of accomplishment with biosafety standards related to farming and

environment protection, food safety and commercial risks. Farming, environmental and food

safety risks are assessed under scientific-bases, but commercial risk analysis brings a political

perspective into the approval process.

The Biotechnology Directorate (BD) made up of experts from public and private

sectors are in charge of farming and environmental risk assessment. The National Service on

Safety and Quality of Farming Products carries out the food safety analyses. The Agriculture

Market Board assesses commercial risks of approving new varieties not approved in

destination markets.

Public opinion polls show that awareness level is also very low in Argentina. On

average 39% of people voiced they knew about genetically modified soybean being grown in

the country. On the other hand, 51% said they are willing to pay more to consume non-GMO

food, only 12% believe genetically modified crops can be beneficial for society as a whole

and 51% believe large multinational companies will rip benefits of technology adoption

(Diamante & Izquierdo, 2004).

Labeling is not mandatory in Argentina, and country authorities often argue that

the EU mandatory labeling is detrimental to exports of agricultural products from Argentina.

Brazilian government, instead, implemented strict rules for the commercialization

of GMOs. The policymakers nearly banned commercial growth of any GMO in the country

for almost a decade after 1996. Table 8 put together information on soybean events approved

for cultivation by Brazilian government.

41

Table 8 – Brazil’s Approved GM Events of Soybean

Event Trait Company Authorized For Year of

approval

40-3-2 Glyphosate herbicide tolerance Monsanto Food and

feed/Cultivation 1998

BPS-CV127-9 *** Imidazolinone herbicide tolerance. BASF Inc. Food and


A2704-12** Glufosinate herbicide tolerance Bayer

CropScience

Food and

feed/Cultivation** 2010

A5547-127** Glufosinate herbicide tolerance Bayer

CropScience

Food and

feed/Cultivation** 2010

MON 87701 X MON

89788 Glyphosate herbicide tolerance Monsanto

Food and


DAS-68416-4* Glufosinate herbicide tolerance, 2,4-

D herbicide tolerance

Dow

AgroSciences

LLC

Food and

feed/Cultivation* 2015

Note: * not commercialized, ** imported, *** status not reported



As it can be seen, Brazil was the last country from the G3 to approve the

cultivation of GMOs. It is worth noting that in spite of approval for cultivation of Roundup

Ready TM

(40-3-2) being granted in 1998, legal production would be actually possible only

after 2005, when the National Biosafety Law (no. 11 105 of March 24th

2005) came into

force.

However, as commonly cited GM soybean has been “informally” produced in the

south of Brazil since the mid-1990s. Technically, this prior production cannot be considered

illegal, since a prohibitive act will come into force only few years later due to a court claim

started by the Brazilian Institute for Consumer Defense (IDEC) in 1998. The moratorium

lasted officially until 2003, but remaining conflicts made adoption of GM seeds just take over

after 2005.

The Biosafety law established the creation of the National Technical Committee

(CTNBio), as a bureau under the Ministry of Science, Technology and Innovation, which is

responsible for oversight of all GMOs issues in the country. The CTNBio is made up of

experts and representatives from Brazilian Ministries, such as the Ministry of Agriculture,

Livestock and Food Supply (MAPA), Ministry of Health, Ministry of Environment, Ministry

of Agrarian Development (MDA), Ministry of Development, Industry and Trade (MDIC),

Ministry of Defense, Ministry of International Affairs (MRE) and Ministry of Fishing and

Aquaculture (MPA). This hybrid governance design ensures that political matters

(commercial risks included) are going to be considered in each approval process.

42

While in 2003/04 4.7 million of hectares was planted with genetically modified

seeds, in 2014/15 this share was of 29.1 million of hectares – about 93.2% of total area

planted with soybeans (Galvão, 2014). In spite of the rapid adoption, Brazilian authorities

approved the second variety of GM soybean (BPS-CV127-9) only by 2009. Also, in spite of

having approved 6 varieties of soybean, according to major seed companies’ declaration the

country actually grows only three genetically engineered varieties.

Unlike Argentina and US, Brazil hasn’t produced a meaningful amount of GM

soybean (Roundup ReadyTM

in this case) until the second half of the 2000s. This peculiarity

will play a very important role in the first decade of commercialization, when the major

distinction was made between GM and GM-free products.

Consumer awareness about GM food is surprisingly high in Brazil as indicated by

a recent poll. A survey with 1439 respondents revealed that 54.8% of respondents was aware

about the use of “transgenic plants” to produce medicines and 94.7 was aware of their use in

food production (Capalbo, Arantes, Maia, Borges, & Silveira, 2015).

One should consider that the above-mentioned poll was released online attracting

high proportion of people with college degree and somehow connected to farming activities.

However, even if these numbers are expected to be lower if more diverse respondents were

considered, awareness level may be expected to be higher when compared to the U.S. and

Argentina because of the existence of a mandatory labeling regime in place since 2003. Over

50% of respondents declared they believe transgenic plants are harmful to environment,

42.8% believe they are harmful to human health and 37.2% believe the new developments

are unethical13

.

Last but not least, let’s briefly consider other countries with smaller but still

noticeable presence in global markets, such as Paraguay, Canada, Uruguay and Ukraine.

Paraguay has approved and grown two GM soybean varieties – the Roundup

Ready TM

(40-3-2) and INTACTA RR2 PRO TM

from Monsanto (BIO, 2015; Yankelevich,

2011). Approval for Roundup Ready TM

was granted in 2004 (CERA, 2015). Although the

Ministry of Agriculture granted late approval for the first GMOs, there are strong evidences

of growers using GM seeds since the mid-1990s. Current GM-soybean adoption share of total

output is estimated in 97%, in Paraguay.

13

The survey also indicated that consumers have different perceptions of the words “biotechnology”,

“biosafety”, “genetic engineering”, “GMO” and “transgenic plants”. Perception tends to be more positive

towards biotechnology than biosafety, and transgenic plants have more negative connotation than GMO and

genetic engineering.

43

Canada, instead, have approved a large number of GMOs event for various crops.

The country has approved 23 varieties of GM soybean and at least 3 varieties have been

actually grown – namely Roundup Ready TM

(40-3-2 approved in 1995) LibertyLinkTM

(A2704-12 approved in 1999) and Genuity Roundup Ready 2 Yield (MON 89788 approved

in 2007). Other approved varieties comprehend “not commercialized” (12), “import”(2),

“closed loop cultivation” (2) and varieties not reported in the Biotradestatus database (4).

About 62% of soybean produced in Canada is genetically modified according to data from

Statistics Canada. A 2002 survey reported that only 31% of Canadians viewed genetically

modified fruits and vegetables as good, whereas 63% thought these products were bad (Pew

Reserch Center, 2003).

Uruguay has approved 4 varieties, from which only one is currently cultivated, the

INTACTA RR2 PRO TM

from Monsanto. Authorities have approved the first genetic

modified variety at the beginning of commercial production of GMOs. Planted hectares with

GM soybean are estimated in 99% of total harvested to this crop. Likewise Brazil, Uruguay

accounts with a national commission for biosafety (GNBio) operating with representatives of

industry, civil society and government. Public perception is mixed, but consumers have been

claiming for mandatory labeling. Today the country has a voluntary labeling regime in place

(Markley & Yankelevich, 2012).

Ukraine is the only noticeable exporter that has not approved cultivation of

GMOs. There are some rumors about pirate production of GM soybean in the country, but at

least for the exported amount it is very unlike the country would be able to access external

markets without any further official control – especially the EU markets.

Only recently Ukraine government approved importation of GM soybean meal to

be used by local feed manufacturers. Many specialists agree that the country continues to be

a challenging market for biotechnology promotion. The major drivers of this situation are the

generally negative public opinion, paper work and the gap between testing and approving

systems. Moreover, the industry (individual producers and traders) has not been very active

in supporting GMOs, unlike in other agricultural countries (Hager & Dubinyuk, 2014).

1.2.2 Importing Countries: Regulation, Adoption and Consumers’ Perception

European Markets

The European Union is one of the largest destinations of soybean and by-products

as we have seen previously. It accounted for 14.21% of soybean, 51.92% soybean meal and

44

14.65% of oil world imports in 2014 (FAO, 2015). Nevertheless, the bloc conjunct regulatory

framework is often considered one of the most restrictive in the world. The combination of

these two factors – a large market with a high degree of technology hatred – will become a

big deal for producing and processing industry as whole worldwide.

It is worth noting that in spite of the EU being often treated as a homogenous bloc

of countries, Member States (MS) can differ considerably in their views and attitudes towards

GMOs. From a broad perspective, the MS could be clustered into adopters, conflicted and

opposed countries (FAS-USDA, 2015).

Adopters usually account with a supportive industry and farming sectors with

little opposition from consumer organizations, which enable domestic cultivation of GMOs

once approved by the EU, significant R&D activities and importation of high volumes of

GMOs. Spain, Portugal, Czech Republic, Slovakia and Romania are part of this group.

Indeed, the only approved variety for cultivation in the EU is the Bt Corn (MON810) –

grown mainly in Spain.

Conflicted group has R&D activities, imports large amounts but has little or none

area planted with GMOs. Consumers’ aversion tends to be higher when compared to

“Adopters” prompting policymakers to put more restrictive rules into force. Industry and

retailers avoid using or selling GMOs fearing boycotts or even depredation of shops and

facilities by radical groups lead by some NGOs. Countries in this group are France, Germany,

Poland, Southern Belgium, Bulgaria, Ireland, Sweden and Lithuania.

Opposed group usually imports GMOs only when they need to fulfill domestic

demand, and private sector (farmers and industry), government and consumers intensely

oppose to GMOs. They tend to be more supportive to organic and traditional agricultural

sector marked by geographical indication products. Countries such as Austria, Croatia,

Cyprus, Greece, Hungary, Italy, Malta, Slovenia and Latvia make part of this more averse

group.

Notwithstanding contrasting views across MS, the Bloc has a common regulation

framework for approvals, labeling and traceability procedures for commercialization of

GMOs. Broadly based on precautionary principle, the EU regulatory framework has left

substantial room for most averse countries to ban production and importation of GMOs based

on the argument of lack of sufficient scientific evidences to assure safety in terms of human

and animal health and environment protection.

Indeed, many MS have raised bans against cultivation and importation of GMOs

during the past two decades of commercialization. None of these bans affected directly

45

soybean products, but the production and importation of corn varieties among other GM

crops.

The EU regulates GMOs through two major directives. Directive EC 1819/200314

establishes rules for importing, distributing and processing GMOs – including labeling rules

– and directive 2001/18/EC rules cultivation issues. Thus, approval process discriminates

between cultivation and importation (for food and feed). Taking into account the impacts on

trade, we are primarily concerned with approvals for importation of GMOs to be used as

food, feed and ingredients. A scheme of approvals for food and feed is provided in Figure 1.

In the EU, the approval process is the same for GMOs or products containing GM

ingredients for any purpose, such as processing, food and/or feed. First an interested part

submits an application containing all the relevant and available information to the MS

authority appreciation. Importers and seed companies are typically the applicants for this

type of approval.

Once submission is received, the MS authorities must send a dossier to be

assessed by the European Food Safety Authority (EFSA). EFSA core task is to independently

assess any possible risk of GM plants to human and animal health and the environment. The

European Community takes EFSA opinion as scientific advice and not as a final decision.

EFSA has being often considered supportive to technology adoption, since the analyses are

much more scientific-based than other phases in approval process.

14

It replaced the Directive 90/220/ECC – on the deliberate release into the environment of genetically modified

organisms – was based on the precautionary principle recommended by OECD (1986) to regulate the releases of

genetically modified organisms.

46

Figure 1 – European Union Approval Process of New GMOs for Food and Feed

Source: FAS-USDA (2015)

In the following, the European Community send in a draft decision to the

Standing Committee on Plants, Animals, Food and Feed Meetings (PAFF) reflecting EFSA

opinion. PAFF committee must vote in favor or against the draft within 3 months. For

applications submitted after 2011, the verdict can be only valid if qualified majority were

achieved. In case of deadline expiration without any valid decision, European Commission

can send forward the draft to Appeal Committee of Member States. In case of lack of

qualified majority to take a decision within the deadline, the Commission can finally

deliberate upon the matter.

Although timeline for taking decisions should be no longer than 12 months it has

taken 47 months on average. Even if the entire process comes into a conclusion with an

approval grant, MS can invoke safeguard clauses arguing lack of needed scientific evidence

to discard risks to human and animal health and the environment. Approvals last for 10 years

and need to be renewed after expiration.

In recent years, a number of approvals of new soybean varieties for food and feed

and renewals have been made. Although not formally confirmed, there are indications that

the EU approved new varieties because of the increasing scarcity of conventional crops is

47

creating a production risk for the domestic livestock and processing industry (see Stein &

Rodriguez-Cerezo, 2009). Table 1 shows GM soybean currently approved in the EU.

Table 9- GM Soybean Approved in the EU for food and feed

Event Trait Company Approval Type Year

40-3-2 Herbicide Tolerant Monsanto Food and feed 1996

A2704-12 Herbicide Tolerant Bayer Food, Feed and

Processing 2008

MON89788 Herbicide Tolerant Monsanto Food, Feed and

Processing 2008

A5547-127 Herbicide Tolerant Bayer

CropScience

Food, Feed and

Processing 2012

DP3560423 Herbicide Tolerant

(stacked gene) DuPont

Food, Feed and

Processing 2012

MON87701* Insect Resistant

(Lepidoptera) Monsanto

Food, Feed and

Processing 2012

MON 87705*

Herbicide Tolerance

+ Modified Product

Quality

Monsanto Food, Feed and

Processing 2012

MON87701xMON89788 Insect Resistant and

herbicide tolerant Monsanto

Food, Feed and

Processing 2012

MON 87708* Glyphosate herbicide

tolerance

Dicamba

herbicide

tolerance

Food, Feed and

Processing 2015

MON 87769* Glyphosate herbicide

tolerance

Modified

oil/fatty acid

Food, Feed and

Processing 2015

MON87708 x MON89788* Herbicide tolerant Monsanto

Food, Feed and

Processing 2015

BPS-CV127-9** Imidazolinone

herbicide tolerance. BASF Inc. Food and feed 2015

Note: * not commercialized ** status not reported



Note that approval of GMOs, Roundup ReadyTM

aside, took more than 10 years to

occur as a result of de facto moratorium. Roundup ReadyTM

from Monsanto received its

second approval in 2005, after the company has submitted an application to do so. That was

the only approved soybean event prior to the unofficial moratorium on new approvals from

1998 to 2003.

Despite having granted approval to 12 different varieties of soybean, only 6 have

been actually imported by the EU. Compared to U.S. and Canada there is a lag of more than

48

10 varieties. On the other hand, Brazil, Argentina, Paraguay, Ukraine and Uruguay approved

has mostly approved varieties already approved by EU authorities.

Taking into account approval year and actually grown varieties in sourcing

countries, however, the controversy in the mid-1990s can be clearly see around the Roundup

Ready TM

soybean approvals. The EU approved the Roundup ReadyTM

in 1996, whereas

Argentina has approved in 1996, United States in 1994 and Canada in 1995. Brazil, unlikely,

will approve this event only by 1998, and intensive adoption – due to agronomic and

regulatory issues – will take over only after 2005. In spite of official approval, it is known

that the EU feed and food manufactures continued skeptical about the technology due to

strongly negative views of consumers.

In 2004, when regulation claims on GMO importation intensified, the labeling and

traceability of GMOs was implemented under EC 1829/2003, which replaced the previous

regulation. Traceability and labeling of GMOs are mandatory. Labeling rules define which

products must be labeled as "contains GMOs". The product must be labeled regardless of the

degree of processing, and when sold without packaging, for example, in restaurants, the

information "contains GMOs" should be visible. Any product with an adventitious level of

GM ingredients greater than allowed should be labeled as "GMO". Animal protein – meat,

diary products, and eggs - feed with GM products is the only exception to mandatory labeling

norms. However, there are signs that voluntary labeling with the information “GM-Free” has

been opening marketplaces for IP grains in feed sector.

There is a threshold of adventitious presence of 0.9% of GMOs to a product be

considered conventional. When adventitious presence involves an event that is not authorized

for consumption in the EU, the product cannot be placed on the market, despite being

labeled. It is worth noting that the level of adventitious presence tolerated by the EU is one of

the world’s lowest. For products not approved the tolerance level is virtually zero (0.1)15

.

Polls on consumer opinions provide indications of strong level of opposition to

GMOs in Europe as a whole. In November 2000, for example, the "Nordic Industrial Fund"

conducted a survey in Denmark, Finland, Norway and Sweden on consumers' opinion about

genetically modified foods and their derivatives. The survey found that conventional foods

have benefits simply because they are not genetically modified. A series of negative

associations, such as "unhealthy products" and "great uncertainty about the risks" were

15 This level can be a great problem to modern trade basis. The current logistic system for international transportation and

storage of grains is based on scale gains, complicating identity preservation and maintenance of low levels of adventitious

presence.

49

attributed to GMOs (Grimsrud, 2004). The same author pointed out that the Norwegians, who

the authors take as a sample for the European market, would be willing to buy a GM product

at an average discount of 50% in price.

The results of another survey revealed that highest level of formal education

positively affects the consumer’s acceptance in the same country, while women and older

respondents reacted negatively. Curiously, self-declared awareness about the GM technology

negatively affects the will to consume GM-food (Mccluskey, Grimsrud, & Wahl, 2006). It

could be evidencing that the way in which media has conveyed the issue contributed to the

negative perception among European consumers (see Brossard, Shanahan, & Nesbitt, 2007).

Surveys on the EU27 revealed that Europeans are quite aware about GM issues.

Results show that 46% of respondents talked about or searched for information occasionally,

9% talked about or searched for information frequently and 27% at least heard about GM-

food. The same study also revealed that levels of technology support for food production in

2010 were low and decreasing when compared to surveys carried out in 2005. In 2010, only

5% of respondents declared “totally agree” with technology and 18% declared that they “tend

to agree”. On the other hand, 33% declared “tend to disagree” and 16% “totally

disagree”(Gaskell, Stares, Allansdottir, & Allum, 2010).

However, empirical researches often point to a certain degree of inconsistences

between public opinion and effective decreases in total intakes of GM-food/feed. These

inconsistences may be related to the fact that consumers do not properly recognize labels

(Noussair, Robin, & Ruffieux, 2002). Finally, the role of retailers must be considered in the

EU case.

The other European countries not subjected to the official EU standards, have also

hold a number of rules contrary to GMOs. Actually, in some cases, they are even more

restrictive than EU countries in terms of national biosafety regulations. Switzerland, for

example, has passed several decrees establishing moratoriums on production of any GMOs in

the country. The two largest retailers in the country (Coop and Migros) representing around

70% of the whole market, advertises they only commercialize GM-free products – assuring

that even the livestock and diary products were not feed with GMOs. Coop published in a

press release in 2007 results of an in-house poll indicating that 87% of respondents told to be

unwilling to eat GM-food (Strossman, 2009).

To sum up, in European countries most opposed to technology, producers see

organic production as more promising and believe that the rules of coexistence will not be

able to ensure good farming practices to maintain identity preserved. The presence of

50

organizations such as Greenpeace and Friends of the Earth contributes to the formulation of

the concept of "Frankenstein foods". In other words, politicians, policy makers, agricultural

cooperatives and consumers share the same opinion, to say that, agricultural biotechnology

creates uncertainties, offers no clear benefits, and therefore is not required (Gaskell et al.,

2010).

China

China is also one of the largest markets for soybean and by-products as we have

seen in the first section of this chapter. The country accounted for 69.17% of global imports

of soybeans and 16.01% of soybean oil in 2014 (FAO, 2015). Currently, the country is not a

major importer of soybean meal. One central characteristic of country is the increasing

crushing capacity, which put Chinas in the first position of major importers rank. The history

of GMOs would be certainly different if China was not a big player, or if it had enacted more

strict regulation to these products.

The basis of the regulation in China is noticeably different from that observed in

the EU and other countries declared opposed to GM technology. According to the National

Biosafety Law, enacted in 1993, China aims to “promote research in biotechnology, consider

the adequate control of biosafety, ensure the maintenance of public health, prevent

contamination of the environment and maintain the balance of biodiversity” (Decree

304/2001).

Regulatory framework is outlined by the State Council and implemented by the

Ministry of Agriculture (MOA). The regulation of production, importing, processing among

other issues are broadly covered by Decrees 8, 9 and 10 of MOA. All the approval process

for cultivation and importation are virtually controlled by the Ministry of Agriculture and

bureaus under its coordination such as the National Biosafety Committee (NBC) formed by

(Lagos & Jie, 2013).

For the importation of product to be processed internally, a foreign seed developer

must apply for an agricultural biosafety certificate from Administrative Examination and

Approval Office of MOA. Applicants have to provide information and test results for the

varieties intended to be imported.

After receiving all paper work, MOA bureaus will carry out a number of tests in

different environments to testify safety issues. Once the test results are obtained MOA

bureaus send them forward to NBC for final opinion on safety matters. Based on the entire

process MOA issues the biosafety certificate, enough for importing products to be processed

but not for cultivation.

51

Indeed, China has not approved importation of any GM seed for domestic

cultivation in spite of being one of the largest GM producers in the world. Government has

invested large amounts of money in R&D while private research is very limited and foreign

capital for R&D is legally prohibited. This may be a strategy of infant industry protection

when the favoring position towards GMOs by Government is declared. The approval process

in China was often criticized in the past for lacking transparency. No written regulation or

guideline for approval procedures existed before May 2003. Currently three GM soybeans

have received market approval in China, as seen in table 10.

52

Table 10 - China’s approved GM soybeans

Event Trait Company Authorized for Year

GTS-40-3-2 Glyphosate herbicide

tolerance Monsanto Processing, Food and Feed 2002


tolerance

Bayer

CropScience Processing, Food and Feed 2007

MON89788 Glyphosate herbicide

tolerance Monsanto Processing, Food and Feed 2008

DP356043



herbicide tolerance

DuPont (Pioneer

Hi-Bred

International Inc.)

Processing, Food and Feed 2010

DP305423


tolerance, Modified

oil/fatty acid

DuPont Processing, Food and Feed 2011

BPS -CV127-

9**


tolerance Basf Processing and Feed 2013

MON87701 Lepidopteran insect

resistance Monsanto Processing and Feed 2013

MON87701 x

MON89788


tolerance, Lepidopteran

insect resistance

Monsanto Processing and Feed 2013


tolerance

Bayer

CropScience

Importing processing

material 2014

DP305423 x

GTS 40-3-2




Modified oil/fatty acid

DuPont (Pioneer

Hi-Bred

International Inc.)

Importing processing

material 2014

Note: ** market status not reported



Note that the approvals of GM soybeans in China were delayed even for the

Roundup ReadyTM

from Monsanto. The lack of effective monitoring system in place is the

only explanation for large amounts of soybean coming from United States and Argentina that

have entered the Chinese marketplace. Indeed, many analysts consider that regulation of

53

imports was minimal in the 1990s. China has a system of temporary approvals, so granted

approvals expires each three years requiring renewals.

Although positive view and acceptance by Chinese government, it is worthy

noting that China’s import approval process takes on average 2-3 years and can only

commence when a submitter for import approval has already received full regulatory

approval in their country of origin, intensifying negative impacts of asynchronous approval (J

Huang, Yang, & Yang, 2012).

Labeling is mandatory in China since 2007, although policymakers have

established several exceptions and the criteria for exception are not clear. Government voices

that criteria to inclusion or removal of products from the list include socioeconomic factors,

political goals and national biosafety issues. The regulation is defined by Decree 10

(CH7053) and the list includes soybeans, soybean meal and soybean oil (Jikun Huang, Qiu,

Bai, & Pray, 2006).

Some issues related to labeling, and more specific procedures for export and

import, are decided by an inter-ministerial council formed by the State Council. Also in 2001,

the Ministry of Public Health promulgated the first guideline to GM food safety. These laws

came into force in June of 2002, bringing a number of important deliberations, such as the

requirement of field tests before market release, mandatory labeling, new standards for

importation and exportation of GM food and rules for regional monitoring (J Huang et al.,

2012).

Consumer participation in shaping the legal apparatus is often considered very

incipient, making it easier for the government to implement standards from the top down, and

so, reducing the importance of consumer opinion.

There are several studies focused on consumer opinion in China. They present

significant divergence in consumer acceptance in the country. On the one hand, a study

carried out by Greenpeace (2004) claimed that GM foods were generally not accepted by

Chinese consumers. On the other hand, some studies identified that Chinese consumers

would be willing to pay a premium for GM foods.

Mccluskey et al. (2006), for instance, analyzed the differences between the views

of consumers in both China and Japan. According to the authors, Chinese consumers would

be willing to pay a premium of 38% to consume a variety of GM rice (Golden Rice) and

16.4% for GM soybean oil. Although factors such as formal education and self-reported

knowledge have negative impacts on their willingness to consume GM products, the values

are always less significant than in Japan.

54

The explanation for divergences may be a change in opinion across the years.

There was a trend in studies pre-2008 indicate that consumers were very opened and

accepting biotechnology products. More recent studies, however, have revealed certain level

of concerns among Chinese consumers (Lagos & Jie, 2013). According to these authors

media and NGOs have constantly broadcasted more suspicious views towards GMOs and

some scientists have voiced concerns about long-term risks surrounding the technology.

Other importing countries

Finally, we could consider other importing countries to conclude our analysis of

country regulatory frameworks for GMOs. Mexico, Japan, Philippines, India and Algeria are

secondary markets, which could be considered briefly. We are also going to present a list of

countries taking part in the UNEP-GEF project to outline biosafety regulations.

Mexico has not being a challenge for exporting countries of GM soybeans.

Country legal framework, however, has a negative impact on importation of GM seeds not

approved for production in the country. Society as a whole doesn’t seem very participative in

GMOs debate and label isn’t mandatory for food and feed being produced with GM

ingredients (Otero, 2015). About 20 GM soybeans have been granted approval, but only 9

have been actually imported. Mexico country is a large importer from United States.

Japan’s regulatory framework has been considered very pragmatic. When

compared to other countries, Japan has one of the largest numbers of approvals for soybean

varieties, including those that are nutritionally modified. There are 19 different varieties

approved, being 8 actually imported. The approval process depends on the product’s purpose,

so not all events undergo the same process. Although almost all the varieties have been

approved for cultivation, according to data from the Biotechnology Industry Organization

(BIO), so far Japan has not produced any GM soybean. As in Europe, the food industry and

retailers have avoided selling products labeled as GM - but processed foods end up being

excluded from the list of mandatory labeling.

Japanese consumers’ attitude toward GM technology is closer to the attitude of

European consumers than their Asian neighbors. Consumers were willing to buy a GM

product if the discount on the final price of the product was, on average, 50%(Curtis &

Mccluskey, 2004). Variables such as knowledge of the subject, level of formal education, if

the individual is female or in an older age group, directly affect the rejection of the product at

an increasingly rate when compared with a sample of U.S. consumers.

Philippines, together other agricultural countries in Asia, has been more open to

55

biotechnology products. The country has the highest level of consumer awareness in Asia,

according to a poll (Asian Food Information Center, 2008). A total of 8 GM soybean varieties

is approved for food and feed uses in the country, being only 2 of them not commercialized

(BIO, 2015). India is also an agricultural country, but has approved only two GM varieties of

GM soybean (Roundup ReadyTM

and Genuity Roundup Ready 2 YieldTM

). Consumers in

both countries are more favorable towards the technology and a majority of consumers

believe that food biotechnology will bring benefits in the next few years. In Philippines 70%

of consumers believe in benefits, and in India this share is of 70% of consumers(Asian Food

Information Center, 2008). Expected benefits included better quality, improved yield,

healthier products and higher levels of food security. There is not much information on

situation of GMOs in Algeria, but in December 2000, the country banned the importation,

distribution, commercialization and utilization of any GM plant material (Moola & Munnik,

2007). National regulatory framework is being prepared with the cooperation of international

organizations such as the United Nations Environment Programme of the Global

Environment Facility Coordination (UNEP-GEF). The UNEP-GEF project was created to

support developing and underdeveloped countries in building national biosafety laws.

Participant countries can potentially impact trade flows in the near future when they fully

implement regulatory frameworks. The project has been inspired mainly by precautionary

approach being the regulations based on the case-by-case procedure. Table 11 shows

countries participating in the project.

56

Table 11. UNEP-GEF Countries

Africa Asia and the Pacific Central and Eastern Europe Latin America and the

Caribbean

Algeria Azerbaijan Albania Antigua and Barbuda

Benin Bangladesh Armenia Argentina

Botswana Bhutan Belarus Argentina

Burkina Faso Cambodia Croatia Bahamas

Burundi Cook Islands Czech Republic Barbados

Cape Verde Indonesia Estonia Belize

Central African Republic Iran, Islamic Republic of Georgia Chile

Chad Jordan Latvia Costa Rica

Comoros Kazakhstan Lithuania Dominica

Congo Kiribati Malta Dominican Republic

Congo, Democratic

Republic of the

Korea, Democratic People's

Republic of Moldova, Republic of Ecuador

Côte d'Ivoire Korea, Republic of Romania El Salvador

Djibouti Kyrgyzstan Serbia Grenada

Eritrea

Lao People's Democratic

Republic Slovakia Guatemala

Ethiopia Lebanon Slovenia Guyana

Gabon Micronesia

Macedonia, The Former

Yugoslav Republic of Haiti

Gambia Maldives Turkey Honduras

Ghana Marshall Islands Ukraine Jamaica

Guinea Mongolia Nicaragua

Guinea Bissau Myanmar Panama

Lesotho Nepal Paraguay

Liberia Niue Peru

Continued…

57

Africa Asia and the Pacific Central and Eastern Europe Latin America and the

Caribbean

Libya Arabic Palau

Saint Kitts and Nevis

Madagascar Papua New Guinea Saint Lucia

Mali Philippines

Saint Vincent and the

Grenadines

Mozambique Samoa Suriname

Morocco French Solomon Islands Uruguay

Niger Sri Lanka Venezuela

Nigeria Syrian Arab Republic

Trinidad and Tobago

Rwanda Tajikistan

Sao Tome and Principe Thailand

Senegal Tonga

Seychelles Tuvalu

Sierra Leone Vanuatu

Sudan Viet Nam

Swaziland Yemen

Tanzania, United

Republic of

Togo

Zimbabwe

Source: Unep-GEF (2014)

The quantity of countries establishing their regulatory frameworks reveals not

only potential risks related to strict regulations, but that regulations based on risk assessment

prior to the first importation or cultivation of a certain variety is a practice of few countries –

not disregarding the fact that these few countries accounts for more than 90% of global trade

of soybeans.

58

1.3 Remarks

In sum, the history of production and trade of GMOs can be told as an innovation

in the seed industry that increases operation risks across the whole value chain. Commercial

risk arises when end-consumers from important markets develop considerable levels of

“hatred” against the technology. When growers and other grain handlers sell to end-

consumers in other countries, the conflict became a problem of international trade. In this

context, processor and logistic operators will become central agents to manage the

commercial risks related to adoption of new developments. That is because they are

strategically between growers, willing to adopt the technology, and adverse feed and food

manufacturers demanding conventional grains in foreigners’ countries. Thus, the

concentrated ownership in grain handling and processing, and partnerships with the largest

seed corporations, is appropriate to manage gains of scale and mitigate commercial risks.

The government through it regulatory agencies ended up mitigating risks, as

private coordination of produced varieties in producing country would only by a chance

minimize all the production and commercial risks involved in growing varieties not approved

in destination markets. However, regulation responded more to technology constrains and

levels of hatred than to commercial risk itself. Anyway, if no control in variety approvals was

taken a large range of different varieties being produced in one country, and given the high

costs of IP, to segregate varieties by destination would be very costly.

The most known examples of international conflicts related to GMOs are the

technology rejection by the EU – both in terms of end-consumers and regulatory frameworks

– and high rates of technology adoption and large number of approval of GMOs varieties in

the U.S.. On the other hand, China was an alternative destination for GMOs produced in

United State. Brazil and Argentina instead, took more cautious decisions upon the production

of GMOs.

It is worth noting that first clear discrimination in international markets was

between conventional and genetically modified products. Along with decreasing availability

of conventional seeds, traits differences are setting the ground for differentiation within

different GM-seeds instead the former antagonism between GM and conventional crops.

From a broader perspective, when one looks back in history of agro-industry

innovations, we can see a lot of important developments on production techniques leading to

huge gains of productivity perceived by consumers only in terms of price cut-downs. But,

nothing ever made consumers believe that a soybean could actually be an imperfect substitute

59

for a soybean, not because they taste or look different but because they were produced from

two different seeds. By no means a consumer was concerned about seed used to produce

grains processed to produce food.

This point is noteworthy because increasing consumer power is an international

reality and it is creating a series of conflicts in agricultural markets. This is very important

issue to think future impacts on agricultural trade, as it seems a structural change in

consumption drove by increases in world income.

60

CHAPTER II – TECHNOLOGICAL EFFECTS AND TRADE THEORIES

In this chapter we pursue two related objectives. We start by seeking for empirical

evidences technological impacts on trade. Then, we seek to assess how existing theories of

trade and technology change can shed some lights on the framework drawn from systematic

observation of real data. This organization of the ideas is justified by relatively new set of

challenges we are dealing with, and the absence of a theoretical framework that fully applies

to a case marked by backwards technology effects on trade.

As we have seen in Chapter I, the EU and China figure among the major

importers of soybean and soybean by-products and they have taken different positions

regarding the regulation of technology as well as consumers in these countries tend to have

opposing views towards benefits of GMOs. On the other hand, Brazil, Argentina and the U.S.

also enacted divergent measures to regulate cultivation of GMOs in their farmlands.

We believe that the combination of giant markets rejecting the technology and

asynchronous international diffusion led to noteworthy changes in bilateral trade pattern.

Thus, a look at the trade flows between (non) adopters and (non)“haters” after 1996 seems a

good place to start from. In addition, a look at studies pointing out to the effects of

technology adoption and strict regulations on trade reinforces our thesis. These points are

addressed in the first part of this chapter ( section 2.1).

We dedicate the second part of this chapter to the theoretical investigation (section

2.2). We describe the multi-country Ricardian model presented by Eaton and Kortum (2002),

the firm heterogeneity by Helpman, Melitz and Rubinstein (2008) and, last but not least, the

set of main ideas from technological gap theorists, such as Posner (1961), Maggi (1993) and

Dosi et. al. (1998; 2015)

It is possible to note that an initial graphical analysis reveals the most outstanding

impacts of technology in terms of replacement effects in trade flows. Also, as expected, none

of the theoretical frameworks can fully addresses the case of trade in GMOs. Yet the models

provide several helpful insights and highlight the need for developments to consider a case of

adverse technological effects on trade. These and other findings are discussed in section 2.3.

61

2.1 Evidences of Technological Effect on International Trade of GMOs

At first, let’s focus on the big picture. As broadly known, the EU and other

Europeans countries, huge importers of soybean and by-products and hereinafter only the

EU, have raised several worries and trade barriers against the free trade of GMOs. Brazil, one

of the three giant producing countries, will delay cultivation approval of GMOs for almost a

decade. Argentina, instead, will approve just a few varieties for production, which is also

approved in the EU, but growers will rapid adopt the new technology. The U.S. will approve

more varieties of soybean and face a rapid internal adoption process during the first years of

commercial release. Our last protagonist will be China, a giant importer of soybean, which

makes lesser distinction between the “old” and “new” technology used to produce soybean

and by-products.

As we are dealing with big players and very marked positions towards the

technology, at least some flows rearrangement should be seen in bilateral trade series. Of

course, we are not saying that technological effects can explain alone trade patterns after

1996. Instead, we are arguing that besides traditional production and trade costs, the GMOs

rate of adoption should also be considered as one of the key drivers of preference for certain

sources of soybean.

Accordingly, the big picture depicted at the charts below reflects a twofold reason

for the same problem. On the one side, most opposing processors or consumers will demand

a GM-free product (IP) making minor distinction across different genetically modified

varieties. IP capability and low level of adoption will be key to avoid related commercial

losses. This first source of conflicts is primarily related to end-consumer perception seeing

GMOs as unsafe regardless of the type of gene inserted or technique used to produce them.

Policymakers played a key role here by enacting mandatory labeling rules that allow

discrimination among conventional and GM products, but consumers will keep denying the

product even if it can be legally imported and used in the manufacturing of food and feed.

On the other hand, the case-by-case approach for risk assessment will lead to

asymmetric and/or asynchronous approval, which will affect the commercialization even if a

processor is indifferent or in favor of using the technology in food and feed production – and

here is the second source of conflict. Put in another way, ships containing a non-approved

event will be simply refused at arrival port no matter what grains handlers and processors

think about the technology. Policymakers play a more conspicuous role in this type of

conflict as they dictate the pace of approval assessment and can ban entry of unauthorized

62

GMOs into their countries, no matter the general public perception of GMOs. An overlook at

the volumes and origin of imports of European countries16

from 1990 to 2014 (Chart 1) can

reveal some interesting trends. First of all, it is important to remember that much of the

imported soybean is intended to be crushed and sold domestically as soybean meal and/or oil

to feed and food manufacturers and biodiesel millers. As these end-consumers are differently

affected by labeling regulation in the EU, they also have different incentives to show the flag

in favor or against the technology depending on the general perception of society as a role of

the technology.

As we have mentioned before, animal protein feed with GMOs are exempt from

mandatory labeling in the EU. Also, there are likely much less worries about “transgenics”

when there are not intended to human consumption – having no records of public pressures

against the use of GMOs as feedstock, for instance.

Thus, both the strict regulation and end-consumer rejection is expected to impact

on trade. In presence of certain level of hatred, the domestic crushers will try to mitigate risks

by decreasing the amounts of GMOs in their facilities, and obviously, even if they desire,

they cannot purchase varieties not approved for importation in their countries – unless

monitoring systems are poor.

16

European countries comprise not only the European Union (EU) states members but also other economies in

the continent, such as Switzerland, Belarus, Rep. of Moldova, Russia Federation, Iceland, Norway, Ukraine,

Albania, Bosnia Herzegovina, Serbia, Montenegro and Macedonia (FYROM). European countries out of the

EU, in general adopted very similar frameworks when compared to EU members (P. R. S. Oliveira et al., 2013).

By adding all the countries in the continent we can disregard the year of entry into the bloc to the purposes of

this study.

63

Chart 1 – EU’s Soybeans Imports by Source (Billions of USD from 1990-2014)

Source: Elaborated by the authors based on COMTRADE and BACI data17.

Cutting down imports of GMOs to a possible extent is rational if we assume that

indifferent customers wouldn’t mind about buying conventional varieties at the same prices.

As we are going to see also in this section, relative prices equal to 1, considering

conventional and GM grains as imperfect substitutes, is expected under some simulation

conditions as well as it is predicted by some empirical works.

17

Data from 1995 to 2012—underlying years for this case study—were gathered from Baci database. This is a

reconciled world trade database developed by the CEPII at a high level of product disaggregation. The original

data come from United Nations Statistical Division (Comtrade database). CEPII uses original procedures to

reconcile declaration of exporters and importers (for details see Guillaume & Zignago, 2010). Additional years,

1990-1994 and 2013-2014, come from the original Comtrade database, since Baci do not comprise years before

1995 and post 2012.

64

Argentina’s shares of soybean markets are reduced not only in the EU, especially

because of the national strategy favoring exports of soybean meal and oil – using export

taxation for soybeans18

. Yet, data in Chart I show that Argentina’s exports to the EU felt

mainly after 1996. They dropped from a half billion of USD to virtually zero between 1996

and 2005.

Nonetheless, its clear that the U.S. lost market shares of European markets,

whereas Brazil increased its exports into the continent. As it can be seen, Brazil’s exports

have been increasing since the early 1990s, and continued growing after 1996. Exports spiked

from approximately 1 to 3 billion of USD between 1996 and 2005. Conversely, the U.S.

exports to Europe began falling away right by 1996, going from almost 3 to less than 1

billion of USD between 1996 and 2005.

Taking into account that the U.S. is the major global exporter of soybeans this

consistent decrease along with sustained growths of Brazilian market shares is a strong

indication of backward technological effects on trade – more precisely a negative effect of

demand rejection affecting the U.S. soybean exports.

Naturally, one should also consider other drivers for export growth or decrease,

such as the relatively lower costs of production in Brazil – as pointed in chapter I. In this

particular case, variable trade costs related to distance seem to play a less important role as

both countries have access to the Atlantic Ocean, and U.S. distance from Europe is smaller.

Exchange rates can also play a role in favor of South American traders as production costs

are expressed in local currencies and soybeans are traded in USD19

. But, in spite of these

other effects, we cannot deny that the replacement of market shares is very timely marked to

this case in particular.

In addition, some changes in data after 2005 – the first year of legal growth of

GMOs in Brazil – will also back up the idea of strong technological effects occurring in

trading with Europe. Data shows that at the same time the adoption of GMOs had increased

abruptly after 2005 in Brazil, other producing countries have increased their market shares in

Europe.

18

Mr. Macri, the new president elected in 2015 is lifting some taxes on grains’ exports that are expected to have

impacts on international grains markets. 19

Indeed, in spite of other analysis pointing to negative effects in U.S. exports related to currency devaluations

in South America, when we consider all the existing and potential flows of trade this effect changes sign and

decrease in significance.

65

Canada, Ukraine and Paraguay20

are the most noticeable examples of this trend.

This new replacement of sources seems to have a link with the technology adoption as these

countries have approved just a few or none GM varieties for cultivation and/or have

relatively low rate of technology adoption along with good IP capacity. Aggregated soybean

exports of these three countries summed up nearly 4 billions of USD in 2013, more than the

value of Brazilian exports at the same year.

On the other hand, the same clear replacement pattern does not hold for soybean

meal and oil trade, especially because the U.S. plays a secondary role in these markets –

making difficult to have counterfactual for large number of approved varieties from the

supply side.

Another reason for different pattern is that end-consumers in this case are mainly

feed manufacturers. The exemption of mandatory labeling for animal protein feed with

GMOs makes agents from the meat, poultry and diary industry less afraid of boycotts. Only a

small fraction of breeders demands GM-free ration, then the impact of importing meal

deriving from GM-soybean does not seem a “big deal”. IP soybean demand is estimated at

20% of all soybean consumption in Europe (FAS-USDA, 2015).

These facts together indicate that trade of soybean meal is especially impacted by

asymmetric approval instead of end-consumer rejection – especially by the absence of

mandatory labeling requirements.

Notwithstanding, some few indications of technological effects can still be drawn.

First, as there was no mismatch of approved varieties between Europe, Brazil and Argentina,

feed manufacturers were taking reduced risks by importing large amounts of soybean meal

from these countries (more than 8 billions of USD in 2014). Second, instead of a cut down in

market shares of major GM producers during the second half of the 1990s, Brazil lost market

shares after 1997, and lost the leading position in this market in 2005. On the other hand,

Argentina and United States enlarged their market shares of European markets after 1996.

This arrangement open a room for considering a second type of technological effect, which

we have been calling “technological gap” – or the positive effect of relative efficiency gains

or losses related to technology adoption.

20

As we have seen in chapter 1, Paraguay has approved only one variety of GM-soybean for cultivation as well

as has low levels of adoption. Canada has approved a plenty but has improved capacity to IP and adoption rate

is around 50%. Ukraine is considered a GM-Free country.

66

In the case of soybean oil, technological effects are even more ambiguous. The

relatively small Europe’s imports of soybean oil21

and minor participation of the U.S. in

global markets may be an explanation for such ambiguity. The EU experienced an oil-

importing boom during the 2000s, so data were split into two charts for clarity.

Argentina and the United States hold larger shares of European markets in the first

half of the 1990s. However, in spite of continued exports of the US and Argentina, Brazil was

the unique country of the G3 that enlarged its market share from 1996 to 1997.

Noteworthy, soybean oil demand will take off in the continent from 2002 to 2014,

and international prices will peak. Increased demand and prices are a result of new uses –

such as feedstock for biodiesel – and the outbreak of the international financial crisis. During

these years, Brazil and Argentina will take over the European markets while U.S. shares will

keep relatively low. Biodiesel production in Europe spiked from the worth of 2 to 10 billion

tons between 2004 and 2013 (European Biodiesel Board, 2015). It was not by chance that

soybean oil imports boomed at this period.

As we have discussed in Chapter I, actors in European countries differ in terms of

their views of technology benefits. From that, a look at how agents in the most opposing or

supporting countries choose their soybean suppliers can contribute to make some trends

clearer. Chart 2 shows importing patterns for different groups of countries in Europe.

21

In Chapter I we have seen how soybean oil is considered a residual product of soybean meal extraction and

how inferior it is considered for cooking purposes when compared to other vegetal sources of oil.

67

Chart 2 – European Countries’ Imports by Cluster

Source: Elaborated by the authors based on COMTRADE and BACI data

Although all the groups somehow replicate the pattern of aggregated data, i.e.

decreases in shares of U.S. soybean along with increases of Brazilian shares, the impacts

were different in terms of pace of replacement and differences between market shares of one

or another country.

During the second half of the 1990s, the Adopters reduced less their imports from

the U.S. and kept them up with imports from Brazil. After 2005, however, there was a clear

preference for the Brazilian soybean – remember that the second half of the 2000s was the

period of rapid adoption in Brazil. Imports from Canada and Ukraine have been increasing in

the last years, but smaller in pace and proportion when compared to most adverse countries in

Europe.

68

For the Conflicted group, however, decreases in imports from U.S. was sharper

from 1997 to 2000, in spite of the imports from Brazil and the U.S. have varied at the same

pace during the 2000s. After 2009, Paraguay and Canada gained expressive market shares in

these markets, and in greater proportion when compared to Adopters. This indicates that

more recent replacement of Brazilian imports for soybean coming from countries with higher

proportion of conventional soybean or reduced number of approved varieties is more

remarkable in this group than in the Adopters one.

For the most opposed countries two points outstand. First, as expected there is a

sharpen decrease of imports from the U.S. along with sustained increases of Brazil’s exports.

Second, there is a marked fall of Brazilian market shares after 2005 – excluded results for

2008 – at the same time Ukraine imports spike. This replacement may be related to high

levels of adoption in Brazil and the image of GM-free country of Ukraine – but here factors

as proximity and other trade barrier may be playing an important role as well.

Lastly, other European countries also reduced their U.S. imports replacing them

initially by Canadian soybean, but imports from Brazil spiked after 1999. More recently the

trend of importing more soybean from Canada and Paraguay emerged. Conversely, by

looking at trade between China and the major global suppliers we can see a different pattern

– see Chart 3.

69

Chart 3 – China’s Imports by source (tones 1990-2014)

Source: Elaborated by the authors based on COMTRADE and BACI data

China is much larger than Europe in terms of amounts of imported soybean, but

the high national crushing capacity reduces considerably the needs of importing soybean

meal and oil. China imported more than 40 billion of USD in soybeans in 2014, whereas

Europe imported 8.3 billion of USD.

In spite of differences in production and trade costs Brazil and U.S. exports into

China grew as long as the country expanded its internal crushing capacity. Noteworthy,

Argentina also played an important role in the Chinese market. As it can be seen, the

financial crisis affected less the processing industry in China in comparison with Europe22

.

22

It is beyond the focus of this study, but this consideration raises concern about specialization of Brazil in

serving a mature instead of a booming market and effects on sustained growth of exports.

70

Thus, it is not possible to drawn any visible parallel between technology adoption

and marked falls in imports from big adopters. Note that there is no replacement of sources of

soybean in 1996. Moreover, Brazil never took the leading position in soybean markets in

China. It didn’t happen because the U.S. not only has yield advantages over Brazil as well as

trade costs to export to China are in general reduced because of exit to the Pacific Ocean.

Therefore, the absence of major trade barriers, such as a demand rejection,

guaranteed the relatively common effect of “technological gap” – i.e. countries innovating or

adopting new and better technologies will have increased shares of international markets.

Another difference of China is the negligible shares of minor producers, such as

Paraguay, Canada and Ukraine. This is a result of large imported amounts and also from the

lack of a substantial market demand for conventional soybeans.

From 1995 to 1999 imports of soybean meal boomed in China. Brazil and

Argentina were the major suppliers during this period as well as in the following years. The

shares of Chinese markets, once more simply reflected the global participation of major

producing countries in world markets. Shares of soybean oil market in China also kept the

same proportion of major countries shares of global markets, i.e. Argentina kept the leading

position followed by Brazil and, then the U.S.

A simple Constant Market Share23

analysis carried out by Oliveira et al. (2012)

can give out some interesting issues about the drivers of the growth of soybean exports of

Brazil, the U.S. and Argentina across the years.

In the partial adoption period (1995-1997) competiveness effect was the key

driver of Brazilian exports growth whereas the world imports was the key driver of U.S.

exports growth. In the dual-market period (2000-2002) U.S. global exports of soybean grew

only by 11% while Brazil faced a growth of 178% and Argentina of 246%. While

competiveness effects can explain up to 69% of Brazilian exports growth, it can explain up to

(-389%) of U.S. exports growth in the same period. It means that if not by the strong growth

of global imports of soybean, which explains up to 442% of U.S. growth, the country could

have lost meaningful shares in global markets during this period because of the lack of

“competitiveness”. This can be seen as a strong sign of high levels of technology hatred

negatively affecting U.S. competitiveness. It is worth noting that the competitiveness effect is

23

The CMS technique assumes that a country keeps constant it market shares being any change in the trade-

flows a result of three basic effects: growth of world trade, destination market and competitiveness.

Competitiveness is a residual effect and can have a number of explanations such as reduced production or trade

costs or, in the case studied, certain level of hate against GMOs, see (Carvalho, 1995; Tomich & Leite, 1999).

71

the amount of growth variation that cannot be explained neither by growth of world imports

nor by growth of destination markets of a particular exporter.

However, the competiveness effect will change in favor of the U.S. after 2005. In

the post adoption period (2007-2009) competiveness effect will be positive to the U.S. (31%)

and negative to both Brazil (-39%) and Argentina (-47%).

Most of the empirical studies on trade of GMOs focused on regulation

mismatches between trade partners leading to changes in trade flows in terms of volume and

prices – as they analyzed the problem from the partner conflict perspectives. Indeed, the

standardization versus differentiation of international trade basis, including regulatory issues,

has been a particularly important starting point to address changes in trade flows since they

impact the overall transaction costs. It can be assumed that the goal of standardization is to

realize scale effects of world product mandates (Feinberg, 2000), reducing transaction costs

and risks or maximizing profits (III & Kashlak, 1999; Isaac, Perdikis, & Kerr, 2004; Meyer,

2001; Rugman, 1976). Achieving these benefits requires the centripetal forces of

international convergence, including production and process standardization (Griffith, Hu, &

Ryans, 2000). Recent research has focused on the general challenges facing the international

standardization of technology, as well as the challenges facing the international

standardization of biotechnology (Madhok & Osegowitsch, 2000).

From a neoclassical perspective, it can be said that competing with the centripetal

forces of standardization are the countervailing centrifugal forces of public and private

policies that threaten the scale benefits by fragmenting foreign markets (Isaac et al., 2004).

The idea of attracting and repealing forces of trade is key from the empirical and

theoretical perspective. In empirical works they are considered in the form of iceberg costs

and cultural and regulatory differences among trade partners. In theoretical models they are

usually referred as geographical costs – as we are going to see latter on this chapter. Many

empirical studies have been pointing to considerable impacts of regulatory heterogeneity on

bilateral trade. These studies also built index and techniques to better measure this

heterogeneity. In agriculture, in particular, authors usually consider these differences as non-

tariff barriers (NTB) increasing trade costs for most dissimilar partners (Burnquist, Shutes,

Rau, Souza, & Faria, 2011; Vigani & Olper, 2013; Winchester et al., 2012).

Vigani et al. (2012) performed a gravity model to analyze the impact of

technology on bilateral trade in 2005, 2006 and 2007. The gravity variable was the gap

between the regulatory frameworks of trade partners measured by an index estimated by the

authors.

72

The magnitude of the estimated coefficient implies that one standard deviation

decrease in the GMO dissimilarity index (=0.188) increases exports by 33%, all else

remaining equal. Thus, the effect is not only statistically significant but appears also relevant

from an economic point of view. Moreover, the results suggest that labeling is the most

detrimental dimension to trade followed by slow and complex approval process and

traceability. Yet, it is important to say that some studies point to low-level of allowed

adventitious presence and asynchronous24

approval as key drivers of conflicts in international

markets (Faria & Wieck, 2015; Gruere, 2011; Kalaitzandonakes, Kaufman, & Miller, 2014;

Stein & Rodriguez-Cerezo, 2009).

Although very aligned with their purpose of analyzing “regulatory distances” and

trade, Vigani et.al. (2012)25

estimated a cross-section model and their index do not vary

across the time. Thus, they lost the dynamical effect of a change in regulation and the impact

on trade. We are primary interested in these effects of a dynamical increase or decrease of

proximity, which we only can assess with panel data – as we are going to see in Chapter 3.

A more complete index, taking into account not only biosafety regulatory

dissimilarity index but also the gap between approved varieties were built in a recent study,

also pointing to significant impacts on trade. In this study, however, the index varies across

the time, capturing the continued tension between importer and exporters. Considering the

years of 2000, 2009 and 2012 it is possible to see that restrictiveness index is higher for

European countries and South America, and lower for the U.S. (Faria & Wieck, 2015).

Disdie & Fontagné (2010), in turn, studied the impact of the EU de facto

moratorium and bans of other European countries on the exports of complainants (Canada,

Argentina and US) and non-complainants in the WTO dispute from 1995-2005. For all

agricultural products considered26

, estimated coefficients on the “EU moratorium and/or

product-specific measures” variable are negative and statistically significant. As a result, this

econometric specification shows that EU measures on GMOs reduced Argentina, Canada and

US exports of maize seeds by 89.4% on average. Regarding national bans, it appears that

only the Austrian ones on maize (seeds and other) and the Italian one on maize seeds do not

have a significant impact. All other national safeguard measures affected Argentinean,

24

Asynchronous approval means the short-term gap between years of approval across different countries. Long-

term gaps may be seen as asymmetric approval. 25

A more recent work found that regulatory strictness seems to be endogenous. Lack of comparative advantage

in agriculture, strong presence of rural population, stringent environmental laws and spread media ownership in

rich countries led to strict regulation of GMOs (Vigani & Olper, 2013). 26

Maize seeds, maize, oilseed rape, cottonseeds, starch residues and other preparations of a kind used in animal

feeding.

73

Canadian and US exports. Noteworthy, recent studies show that policymakers from different

member states have kept their positions regarding the technology by voting in a favor or

against new approvals in a steady way (Smart, Blum, & Wesseler, 2015).

However Disdier & Fontagné (2010) did not analyzed the soybean trade because

they focused on potential regulatory effects, and soybean was the only crop that was

approved before the de facto moratorium initiated in 199827

. Our study goes to another

direction, showing that impacts beyond the differences in regulatory positions can give out

many interesting stylized effects of technology adoption under certain levels of hatred.

Moreover, the soybean case seems a unique opportunity to highlight how asymmetries in

adoption and acceptance can impact bilateral trade given the international concentration of

markets.

Anderson & Jackson (2004), by using a GTAP model with neoclassical closure,

pointed out that since 1998 when the EU implemented the moratorium, GM adopting

countries have lost EU market shares to GM free suppliers, particularly Brazil for maize and

soybean and Australia and Central Europe for rapeseed.

On the other hand, there are evidences that Canada’s rapeseed and US corn sales

to the EU were successfully shifted to other markets. Market losses occurred but only over a

short period, and globalization quickly offered new export opportunities to GM producers

avoiding exporters incurring in major economic losses via demand diversification (Smyth,

Kerr, & Davey, 2006; Stein & Rodriguez-Cerezo, 2009). This shift to less adverse markets

seems to be the case of soybean markets when we consider the replacement of U.S. by

Brazilian exports in the second half of 1990s, along with growing imports of China.

There are also special concerns in literature about bans and technology diffusion

especially for the case of developing countries. The general idea is that GM technology

diffusion was hampered to a certain extent by major markets rejection (K. Anderson, 2010).

Our description of regulatory frameworks outlining risk assessment procedures in Brazil and

Argentina corroborates this assertion to the extent that these countries deliberately used

political issues to approve new varieties of GMOs afraid of high commercial risks. However,

consider that only the commercial risk was driving the approval process is a very simplistic

assumption.

27 The EU pressured by national interest groups did not approve any new event between 1998 and 2003. This period was

defined by literature as the de facto moratorium. The controversies and conflicts that arose from this period were discussed

at the DSB (Dispute Settlement Body) under WTO.

74

Besides effects on bilateral market attraction and repealing one can think of

premium prices for IP grains. Foster (2010), Parcell & Kalaitzandonakes (2004) and Bullock

& Desquilbet (2002) carried out empirical analyses on prices and dual-market system28

.

Foster (2010) points out that apart from consumer attitudes, the key driver of price

premiums are mandatory labeling of GM products in some key grain consuming countries

(particularly high-income countries) higher production costs for non-GM crops and the cost

of IP. The author examines whether premiums exist for some crops and countries, assuming

European Union and Japan as major markets for certified non-GM soybeans, while Brazil,

United States and Canada are the major suppliers. But accordingly to data presented by the

author, Brazil certifies only a small amount of total conventional soybean internally

produced29

. Taking into account the increases of Canada shares in Europe, it is possible that

averse importers are choosing more reliable and with higher certifying capacity sources for IP

soybeans.

Based on data for premiums paid for Illinois growers, EU import prices of

soybean meals and Tokyo Grain Exchange (TGE) future markets, Foster (2010) argues that

there is enough evidence to assume that premiums were paid for non-GM grains. Illinois

growers traded their grains over and above normal cash prices at harvest time in autumn

between 2004 and 2008. Moreover, the author argues, based on United Nations (UN) data,

that imports price of Brazilian soybean meal into the EU had averaged 4 to 9 per cent higher

than soybean meal imports from Argentina between 1996 and 2008. Considering that from

February 2001 to August 2009 the future prices for IP soybeans had exceeded 30% of the

GM-soybean, author concludes that demand for IP product was increasing at that time.

Although there are no many studies looking for actual demand size for non-GM, the

persistent voting position of the EU towards technology can be seen as an indication that

conflicts are not going to be solved in the near future.

Bullock & Desquilbet (2002) analysis of TGE data also found similar results.

According to the authors, conventional soybean prices per ton averaged $27.5 higher than

GM-soybeans price between May 2000 and September 2001 – calculated as the difference

between monthly prices of non-GM and GM-soybeans. Noteworthy, in accordance with our

discussion in Chapter I, the authors found that $7.50 on average was the premium paid to

28

Unfortunately, to the best of our knowledge there is publication of recent study looking at empirical evidences

of price premiums. This is likely because of lack of macro-level evidences of premiums as highlighted by an

unpublished work by Oliveira et al. (2013). 29

Brazil certified only 2.5% of all non-GMO grain domestically produced in 2008, according to data presented

by Foster (2010).

75

contracted farmers while $20 remained with grain handlers. This can be seen as a reflection

of the higher complexity involved in intermediary activities, as we have showed in chapter I.

Parcell & Kalaitzandonakes (2004) carried out a slightly different analysis. They

studied shifts on prices by analyzing responses of Chicago Board of Trade (CBT) and TGE

future prices of non-GMO soybeans to large food manufacturers and retailers announcements

intentions to remove bioengineered ingredients from their branded products. They agree upon

the thesis that small demand shifts in niche markets with limited size would result in only a

small price impact on the conventional commodity. However, if the demand shift is

significantly large, then price impact may be noteworthy. Three models were estimated

following a GARCH (1, 1) framework – which is more indicated for periods of varying

volatility. In Models I and II the dependent variable is the percentage rate of return of CBOT

futures price between open on day t+1 and settlement on day t-1. Model three has as

dependent variable the percentage rate of return of TGE non-GMO soybean futures price

between settlement on day t+1 and settlement on day t-1. Empirical results from “Model I”

suggest that soybean futures prices did not respond to ban announcements. The joint F-test on

the summation of the coefficients for the five days prior to and five days after the

announcement is statistically significant; however, the summation of price changes around

the announcement is not statistically significant. This further finding suggests that while there

is some evidence of a soybeans future price reaction, the market quickly filtered out the

information.

In model II, when each announcement is analyzed separately, they found that

there were no significant differences in the impact of individual bans and so no individual

effects of firms can be seen. In model III, estimation returned significant positive coefficients

(at p<0,01) to TGE conventional rate of return and futures contracts rollover. Neither the firm

ban announcement coefficient nor the summation of coefficients accounting for the rate of

return the five days prior to and the five days after the announcement are statistically

significant. This indicates that the impact of ban announcements by key food companies, as a

proxy for the size of the non-bioengineered soybean market, was not considered large enough

by the market to matter.

Besides empirical studies dealing with the problem of trade of GMOs, there are

also some simulation models.

76

Moschini (2004)30

developed a partial-equilibrium model to analyze implications

from the introduction of genetically modified products into international markets. Results

show that by imposing mandatory labeling regimes, GMOs exports into Europe decrease. In

other words, labeling could become a ban on imports depending on the level of segregation

costs.

Choi (2010), Lence & Hayes (2001) and Desquilbet & Bullock (2009)

investigated the international trade of genetically modified products, modeling a market for

close substitutes under market cleaning and rational agent assumptions. Choi (2010) and

Lence & Hayes (2001) made use of comparative statics while Desquilbet & Bullock (2009)

estimated a simulation model allowing for multiple equilibriums.

Choi (2010) set the United States as a monopolist GM producer exporting into

Europe, which is an importer of GM food, and a producer of conventional crops. Even

though the author is mainly concerned with the effect of a ban on the land rental prices31

,

some interesting intermediate propositions arise from the paper. According to author, GM

crops require extensive R&D and are not easily copied by others. On the other hand, many

firms grow traditional crops given that there is not entry barrier for this market. However, due

to the close substitutability between the two goods, the U.S. has only a weak monopoly

power in GMOs market.

The author argues that a restrictive quota imposed by European markets on GMOs

imports makes the price of GM food higher and decreases consumer surplus in Europe. Since

goods are close substitutes, quota on GM product also makes the price of traditional food

higher – via cross price elasticity of demand. Thus, the quota increases the producers’

surpluses in the USA, via GM-food increased price, but doesn’t increase the surpluses for

traditional producers in Europe given the perfect competition in traditional farming. This

framework can be very close to what happen with crops grown in Europe, but that is not the

case for soybean.

Lence & Hayes (2001)32

, also considering goods as imperfect substitutes, state

that for certified grains and fixed supply, i.e. short run, the relative prices adjust for market

cleaning. They structured a market with two types of consumers – one is indifferent and the

30 In the model, innovators hold property rights, farmers are competitive, and some consumers believe that GM food is

inferior in quality when compared with traditional food. Identity preservation generates additional costs to the whole market

(GM and traditional) via segregation costs. 31

The author concludes that given perfect competition in the market for traditional crops only land rental prices

can be maintained high in the long run. 32

They also assume that consumer preferences categorize consumers into broad homogenous groups, for

example, feed industry is indifferent to GMO and non-GMO and food industry prefer non-GMO grains.

77

other is willing to buy non-GM grain at any relative prices – and two types of firms – one

producing GMO and the other producing non-GMO. Authors also explicitly bring IP costs

into the model.

They found that premiums only exist when the GM output is relatively large when

compared to non-GM output and demand for Non-GM grains are also relatively higher.

When the conventional supply is relatively large, the equilibrium conditions call for relative

prices to be equal to 1. Premium prices arise as a required incentive to sustain conventional

production under higher costs – i.e. IP and production costs – when there are consumers with

strong preference for Non-GM grains and supply is relatively small. Moreover, IP costs may

lead to part of non-GM product being commercialized without certification.

Desquilbet & Bullock (2009), by exploring who pays the costs and who reaps the

benefits of maintaining a dual-market system, estimated a simulation model in which both

type of grains are produced as well as a third good (alternative good) is also produced. The

model allowed for six equilibriums classes differing in which type of goods are produced and

if premiums are positive or zero. They explicitly considered directed and indirect externality

costs33

of transportation in the model as well as the endogenous production costs of each type

of grain. By externality costs they mean the scale diseconomies emerging from higher

segregation.

According to authors, producers take into account production costs, externality

costs of transportation, direct IP (identity preservation) costs, the technology fee and prices of

GM and non-GM products to make their decisions on production levels. Net prices, instead,

strongly depend on the level of hatred and IP costs, since it is the market price less total IP

costs.

If GMO technology is already being commercialized, the introduction of a small

amount of “hatred” causes the IP demand curve to “appear” and the regular demand curve

(i.e. the curve representing indifferent consumers) to shift-in. If there are no costs of IP, the

IP grain price and regular grain price – i.e. non-segregated grains - remain equal to the

regular price brought about in the equilibrium with GMO technology and without hatred.

Thus, given that GMO technology exists, and there are no costs to identity preservation, the

economy moves from a state with no hatred to a state with a small amount of hatred without

affecting prices, producer welfare, or consumer welfare (Desquilbet & Bullock, 2009b) .

33 See Oliveira, Silveira, & Alvin (2012)for further information on segregation and logistic costs effects

resulting from dual-market systems.

78

The high IP costs, instead, allow multiple competitive equilibriums. Generally, the

equilibrium depends on the size of the channels and premium prices may be positive or zero,

depending on the total costs (IP + technology fee + endogenous costs) and the level of hatred.

High level of hatred in comparison to total costs may lead to equilibria with both regular

producers – i.e. GM producers and Non-GM producers whose don’t segregate grains – and IP

producers and price premiums.

When externality costs are too high, only very high levels of hatred could allow

for a dual-market system, being equilibrium only possible with premium prices regime to pay

off high IP and opportunity costs. Otherwise, too high opportunity costs for non-GM

producers may lead dual-market to fade in spite of the level of hatred. Equilibria with zero

price premiums – i.e. relative prices equal to 1 – may occur when seed market is a monopoly

and IP costs are the same for IP producers and regular producers34

– given the significant

output of IP grains.

In terms of benefits of planting GMOs there are many studies corroborating

economic gains ( see Bärwald Bohm et al., 2014; Brookes & Barfoot, 2014; Chavas, Shi, &

Lauer, 2014; Qaim & Zilberman, 2003; Sturges et al., 2003). Authors usually point to less

expansive and easy control of weed, higher yields and reduction of adoption of tillage

systems.

Yield gains are the most questionable benefit being possible to find evidences of

negligible or negative effects of technology. However, many studies points to higher yields

especially for developing countries in which prior pest controls were poor(Qaim &

Zilberman, 2003).

A recent studied estimated that economic gains reached 116.6 billion of USD

from 1996 to 2012. For the soybean case, there was a cut down in production costs, mainly

through reduced expenditure on weed control (herbicides). In South America, additionally,

there were gains associated with the adoption of no tillage production systems, shortening the

production cycle, so enabling famers to rip benefits of growing a second crop in the interval

of two seasons. They estimate that gains for farm incomes amounted 4.8 billion in 2012

(Brookes & Barfoot, 2014).

It is important to note that technology costs vary across countries and so cost

savings also differ across countries. In Argentina technology costs vary from 2-4 dollars per

34

Authors also discuss the different equilibria when seed market is monopolist or competitive. Zero price

premiums equilibrium is possible only when seed industry is a monopoly and technology fee is set at profit-

maximizing value. They found numerically, that monopoly maximizes when it avoid equilibriums with price

premiums – when IP costs for regular producers are too high.

79

hectare, whereas in Brazil it is 11-25 and in US 15-39. Yield gains are more likely to be seen

in Brazil and Argentina where insect resistant varieties improved considerably pest control

(Bärwald Bohm et al., 2014; Brookes & Barfoot, 2014).

In sum, on the one hand we have major producing and exporting countries

regulating technology in different ways and, so altering the market forces of international

technological diffusion. On the other hand, we have major importers taking different

regulatory positions towards this same technology. In addition, consumers all over the word

will have different views towards the consumption of products deriving from GMOs.

The combination of this initial scenario will become a unique experiment for

studying the interactions of technical change and trade. Overall, a set of empirical and

theoretical analysis has been pointing to negative effects of regulatory heterogeneity on trade

– mainly through asynchronous approval, mandatory labeling and LLP of unauthorized event.

However, empirical analyses have often not considered the effects of technological gap on

trade, and this effect is important once not all markets developed levels of hatred against the

GMOs technology.

It is the same of saying that for each approval of a new variety, the countries are

facing not only a commercial risk but also an opportunity costs defined as the distance a

country is taking from the most innovative markets.

At first, the aftermath of these interactions can be drawn from a dual-market

system for closer substitutes. Explicitly defining the market structure, as usually done by

authors carrying out simulations, is important since premiums will highly depend on the

preference of a good over the other. Last but not least, we are considering a period of

constant technical change meaning that innovation and adoption is also constant. This is

important since from our perspective the differences in approval and changes in regulatory

framework is the key drivers of negative or positive effects on trade – not the general level of

technology available in a country, although this also affect trade volumes.

Thus, technical change instead of overall level of technology seems to be the

underlying forces determining bilateral trade in this model. Next section brings how trade

theories, developed based in most general frameworks can shed some light on this case study.

2.2 Trade theories and Dual-market System

In this section we bring together some theoretical foundations of trade economics

and technology and how they contribute to better understanding a case of adverse

80

technological effects on trade. Noteworthy, we don’t intend to exhaust the topic but bring

some elements to move on with our empirical analysis in Chapter III. Instead of formalizing a

model to deal with major stylized facts identified above, we advance in identifying some

parts of theory that can contribute to explaining the case of GMOs and others that can be

treated in future research.

We depart from a review of how neoclassical35

models, especially the Ricardo’s

one, analyzing how them treat the relationship between technology and trade. In the

following, we introduce the most recent developments, which brought out issues such as

increasing returns, under the scope of firm heterogeneity models. Third, we focus on less

conventional developments presenting the ideas of technology-gap approach – which focus

mainly on impacts of technical change on bilateral trade. Although gravity is discussed along

with theories presentation, further discussion on this topic will be provided in the last Chapter

of this dissertation along with the final results.

2.2.1 The Ricardian Models of Trade and underlying role of technology

Mainstream economics often see trade as general equilibrium model with

hierarchical differences between factors – usually immobile across countries – and goods that

are perfectly mobile across countries. In trade, theory was mainly developed to answer

questions about the sources of gains countries could rip by trading.

In terms of gains, models can return very different results depending on structural

assumptions36

. But, more related to our goals, classical models of trade are aligned in saying

that trade comes from comparative advantage. By comparative advantage they mean

differences between autarky and integrated economy prices, which drive trade specialization

patterns (see Deardorff, 1980). Nonetheless, from a broader perspective there are two major

roots for modern theory of trade with different views about the source of comparative

advantage, namely Ricardian and Heckscher-Ohlin (HO) model.

HO model assumes that the proportion of factors is key for CA, i.e. countries

have different endowments of factors, which lead to different input prices. As countries differ

35

By Neoclassical model we mean models in which producers in country 𝑖 maximize revenue, usually by

choosing optimal level of output given the prices – perfect competition – and representative household

maximize utility by optimally allocating income. 36

Literature is plenty of studies about the gains and possible losses from trade. One classical example is the

debate over specialization in agricultural goods and the impacts on innovation rates and terms of trade discussed

by Latin America economists from CEPAL. Other questions as adjustment mechanism, as the wage decrease as

a necessary adjustment for deficit in balance of payments are commonly a natural result from some general

equilibrium models. As we are focused on the relationship between trade and technology we explicitly ignore

welfare and growth issues in this analysis.

81

only by factors proportion, technology is the same all over the world37

. In other words,

difference in technologies cannot be a source of trade as we have been arguing in the case of

soybean trade. On the other hand, the Ricardian model is strongly based on the assumption

that technology differences across countries are the bases for comparative advantages. As it

stands to reason Ricardian model is more appropriated for our purpose of studying

technological effects in trade when compared to HO-based models.

The model presented here to illustrate the underlying Ricardo’s ideas is the

seminal paper by Eaton & Kortum (2002) – a multi-country model based in the two-country

version by Dornbusch, Fischer, & Samuelson (1977). The model, as usual, is focused on

issues of supply side, simply assuming homothetic preferences in the form of a CES utility

function38

. This functional form is the most used by international economists because of the

operational treatment provided by Dixit-Stiglitz (1977) allowing for treatment to the

preference for variety under monopolistic competition. Preference for variety is a good

explanation for intra-industry trade. The general idea of “love for variety” is based on

Argmington assumption.

Eaton and Kortum (2002) depart from a world with 𝑁 countries {𝑖 = 1, … , 𝑁}

producing a continuum of goods 𝑗 ∈ [0,1]. As in Ricardo technology is country-specific.

Thus, authors denote country 𝑖′𝑠 efficiency in producing good 𝑗 as 𝑧𝑖(𝑗).

Unit input costs 𝑐𝑖 differ across countries39

, but are the same within a country

because they are mobile across activities and activities do not differ in terms of inputs shares.

Considering the productivity level of firms in country 𝑖 and input costs 𝑐𝑖 , the cost of

producing a unit good 𝑗 in country 𝑖 is then 𝑐𝑖/𝑧𝑖(𝑗).

As common in the most recent developments in the field, authors also consider

geographical barriers, which they claim to be a new development in Ricardian tradition at

that time. As usual, the operational grounds for geographical barriers are provided by

Samuelson’s standard and convenient “iceberg” assumption. According to this, delivering a

unit of a good from country 𝑖 to country 𝑛 requires producing 𝑑𝑛𝑖 units in 𝑖. As there is no

trade cost to serving domestic markets, 𝑑𝑛𝑛 = 1 and 𝑑𝑛𝑖 > 1 for any 𝑛 ≠ 𝑖. So,

37

Noteworthy but beyond the scope of this study one should consider some developments to plug technological

differences into HO models (see Fisher, 2011). 38

The CES mathematical form was developed by Hardy, Littlewood and Polya (1934). It was introduced in

economics by Arrow, Minhas and Solow (1961). In the field o f international trade, it was used in the form of

monopolistic competition by Dixit-Stiglitz (1977) and Spence (1977). 39

Authors also advance and break 𝑐𝑖 into intermediate inputs and cost of labor.

82

𝑝𝑛𝑖(𝑗) = (𝑐𝑖

𝑧𝑖(𝑗)) 𝑑𝑛𝑖

(1.a)

This would be the price country 𝑛 would pay if it choose to buy a good from 𝑖,

but considering perfect competition, it will be a unique price for good 𝑗 in country 𝑛, and this

will be the minimum international price for this good, represented as

𝑝𝑛(𝑗) = min{𝑝𝑛𝑖(𝑗); 𝑖 = 1, … , 𝑁}. (2.a)

As consumers have a CES utility functions with elasticity 𝜎 > 0, they purchase

individual goods in amounts 𝑄(𝑗) to maximize their utilities

𝑈𝑖 = (∫ 𝑞𝑖(𝑢)𝜎−1

𝜎

1

0

𝑑𝑢 )

𝜎𝜎−1

.

(3.a)

Maximization is subject to a budget constraint 𝑋𝑛, which is the country 𝑛′𝑠 total

spending. Note that consumers all over the world are represented by a representative

consumer, letting no room for difference in tastes within and across countries. In this case,

consumers have full access to information about all prices and the quality of all goods. As the

reader may have noticed this is important because in the soybean case empirical literature

have emphasized that consumers took different positions towards the product of the

innovation.

To make the model operational and coherent with underlying assumptions in

Ricardo, the authors pursue a probabilistic representation of technologies that can relate trade

flows to underlying parameters for an arbitrary number of countries across the continuum of

goods.

In doing so, they take country 𝑖′𝑠 efficiency in producing good 𝑗 as the realization

of a random variable 𝑍𝑖 (drawn independently for each 𝑗 ) from its country-specific

probability distribution 𝐹𝑖(𝑧) = Pr [𝑍𝑖 ≤ 𝑧]. Taking into account the law of large numbers,

this probability can also be considered the proportion of goods for which country 𝑖′𝑠

efficiency is bellow a cutoff 𝑧.40

Note that productivity is product and country-specific and

not related to firms as we are going to see later.

40

Eaton & Kortum (2002) do not explicitly consider the existence of a cutoff. However, the idea is pretty much

the same developed by HMR (2008) as we are going to see latter in this chapter.

83

By assuming that all markets are perfectly competitive, firm heterogeneity is

wiped out. This only makes sense if technology diffusion rate is instantaneous within

countries. This is also equivalent to saying that all firms producing a homogenous good

within country 𝑗 will export, as goods can only be differentiated by productivities reflected on

prices.

From the model, the cost of purchasing a particular good from country 𝑖 in

country 𝑛 will be 𝑃𝑛𝑖 = 𝑐𝑖𝑑𝑛𝑖/𝑍 – it is, the trade price conditional to efficiency probability –

and the existence of a lowest price 𝑃𝑛 resulting from perfect competition, the likelihood that 𝑖

will serve 𝑛 is the joint probability 𝜋𝑛𝑖 that 𝑖′𝑠 price turns out to be the lowest. Thus, the

Fréchet – or inverse Weibull distribution – is a good representation to the distribution of

efficiencies and prices.

𝐹𝑖(𝑧) = 𝑒−𝑇𝑖𝑧−𝜃

,

(4.a)

where 𝑇𝑖 > 0 and 𝜃 > 1.41

The country-specific parameter 𝑇𝑖 represents country 𝑖′𝑠 general

technology state. In terms of the shape this parameter regulates the location of the

distribution. Thus, the higher this parameter the greater is the chance of getting a higher

efficiency level for any good 𝑗 produced in country 𝑖 . From Ricardo’s perspective the

parameter 𝑇𝑖 is the absolute advantage of this country across the continuum of goods.

Parameter 𝜃, in turn, reflects heterogeneity across goods in countries’ relative

efficiencies – or the spread of the distribution of efficiencies. The bigger this parameter the

smaller is the variability across efficiency of different goods produced within the same

country. In a trade context 𝜃 governs comparative advantage within the continuum of goods

𝑗 ∈ [0,1]. In this particular model authors set this parameter as common to all countries.

Moreover, given the assumptions about the efficiency distribution, the authors

have drawn some interesting results about price distributions. Substituting the equation (1.a)

into the equation (4.a) we have the trade prices (costs) at which country 𝑖 could export goods

41

This restriction is important to make cross-elasticity of demand elastic in the CES.

84

into country 𝑛 – 𝐺𝑛𝑖(𝑝) = Pr[𝑃𝑛𝑖 ≤ 𝑝] = 1 − 𝐹𝑖(𝑐𝑖𝑑𝑛𝑖/𝑝) or

𝐺𝑛𝑖(𝑝) = 1 − 𝑒−[𝑇𝑖(𝑐𝑖𝑑𝑛𝑖)−𝜃]𝑝𝜃

.

(5.a)

However, country 𝑛 will actually by only the range of goods that country 𝑖 can

supply with a price lower than 𝑝. So, the amount of goods that country 𝑛 actually buys from

abroad, can be expressed by

𝐺𝑛(𝑝) = 1 − ∏[1 − 𝐺𝑛𝑖(𝑝)].

𝑁

𝑖=1

(6.a)

That is, by shopping around the globe, the good’s price in 𝑛 will have a domestic

distribution representing the lowest prices for each good internationally and domestically

supplied. We can get a more general formulation by inserting equation (5.a) into equation

(6.a).

𝐺𝑛(𝑝) = 1 − 𝑒−𝛷𝑛𝑝𝜃

,

(7.a)

where the parameter 𝛷𝑛 of country 𝑛′𝑠 price distribution is

𝛷𝑛 = ∑ 𝑇𝑖(𝑐𝑖𝑑𝑛𝑖)

−𝜃𝑁𝑖=1 .

(8.a)

Note that this price parameter summarizes the world state of technology, input

costs and geographical costs – being a measure for multilateral trade resistance (MTR) as we

are going to see in chapter III. The state of technology comprehends not only the gap between

general stocks of innovation across countries but also the gap across efficiencies within the

continuum of goods. This version of the model doesn’t advance on international technology

diffusion.42

Importantly, at the end of the day, countries trade technologies – in the form of

goods – discounted by input and geographic/trade costs. The aftermath is an increased access

to technology through trade goods perceived only by productivity differences. At this point it

42

Eaton and Kortum (1999) showed how a process of innovation and diffusion can give rise to a Fréchet

distribution.

85

must be clear that technology in this context is the set of techniques that a country use to

produce goods. These techniques or means of production are different in terms of quality,

resulting in different productivities.

Finally, three underlying properties can be drawn of the price distributions. First,

the probability that country 𝑖 provides a good at the lowest price in country 𝑛 is simply

𝜋𝑛𝑖 = 𝑇𝑖(𝑐𝑖𝑑𝑛𝑖)−𝜃/𝛷𝑛 . This is the 𝑖′𝑠 contribution to country 𝑛′𝑠 price parameter and the

fraction of goods that country 𝑛 actually buys from country 𝑖 - remember that we are dealing

with a continuum of goods and perfect competition.

Second, for goods that are actually purchased by country 𝑛 the source has no

bearing on the good’s price. Countries with better technology, reduced production and trade

costs will trade a wider range of goods, exactly to the point at which the distribution of prices

for what it sells in 𝑛 is the same as 𝑛′𝑠 overall price distribution. Note that the relation of

country 𝑖′𝑠 technology and input and trade costs relatively to an overall state of technology

and costs around the world will be the driver of country 𝑖′𝑠 exports.

Although authors haven’t developed this point, the idea of an international

technological frontier is somehow represented here. Coeteris Paribus, as much as the

countries get far from it more they will become inexpressive in international markets as a

result of low technological dynamism.

Third, the exact price index for the CES utility function, assuming 𝜎 < 1 + 𝜃 is

𝑃𝑛 = 𝛾𝛷𝑛

−1/𝜃.

(9.a)

Here

𝛾 = [𝛤 (

𝜃+1−𝜎

𝜃)]

1/(1−𝜎)

,

(10.a)

where 𝛤 is the Gamma function. This expression shows how geographic barriers lead to

deviations in the purchasing power parity. Note that 𝜎 < 1 + 𝜃 is necessary to have a well

defined .

It is possible to solve the model to equilibrium by considering labor as the unique

86

factor of production, as in Ricardo, and trade balances.43

More importantly, the gravity

equation can also be derived from the model under basic assumptions of equilibrium of the

balance of payments. Let 𝑌𝑖 = ∑ 𝑋𝑛𝑖𝑛 be country 𝑖′𝑠 total exports, then

𝑌𝑖 = ∑ 𝑇𝑖(𝑐𝑖𝑑𝑛𝑖)−𝜃𝑋𝑛

𝑛

= 𝑇𝑖𝑐𝑖−𝜃𝛺𝑖

−𝜃

(11.a)

where

𝛺𝑖−𝜃 ≡ ∑

𝑑𝑛𝑖−𝜃𝑋𝑛

𝛷𝑛𝑛

Solving equation (11.a) to 𝑇𝑖𝑐𝑖−𝜃 and plugging into 𝑋𝑛𝑖 = 𝑇𝑖(𝑐𝑖𝑑𝑛𝑖)

−𝜃𝑋𝑛

and applying clearing market condition, we get

(12.a)

𝑋𝑛𝑖 =𝑋𝑛𝑌𝑖𝑑𝑛𝑖

−𝜃𝛺𝑖𝜃

𝛷𝑛 .

(13.a)

Using 𝑝𝑛 = 𝛾𝛷𝑛−1/𝜃

from consumer assumptions, we finally have

𝑋𝑛𝑖 = 𝛾−𝜃𝑋𝑛𝑌𝑖𝑑𝑛𝑖

−𝜃(𝑝𝑛𝛺𝑖)𝜃.

(14.a)

This can be considered a standard gravity equation since bilateral resistance 𝑑𝑛𝑖

and multilateral resistance terms 𝑝𝑛 and 𝛺𝑖 were considered. From this, it is possible to see

that the model seeks to predict the role of comparative advantage meaning the differences in

productivity across goods within countries – i.e. parameter 𝜃. 44

43

Solving to equilibrium we should consider 𝑋𝑛𝑖 as the total spending in country 𝑛 on goods from country 𝑖.

Income or total spending of country 𝑛, therefore, will be 𝑋𝑛𝑖 ≡ ∑ 𝑋𝑛𝑖𝑖 . We know that 𝑋𝑛𝑖

𝑋𝑛= 𝜋𝑛𝑖 , so 𝑋𝑛𝑖 =

𝑇𝑖(𝑐𝑖𝑑𝑛𝑖)−𝜃𝑋𝑛. Considering the simplest case of no intermediate goods we can assume that 𝑐𝑖 = 𝜔𝑖, or the cost

of a unit of labor. In equilibrium total income in country 𝑖 must be equal to total spending on goods from

country 𝑖 . 𝜔𝑖𝐿𝑖 = ∑ 𝑋𝑛𝑖𝑛 Similarly, trade balance requires 𝑋𝑛 = 𝜔𝑛𝐿𝑛 so that 𝜔𝑖𝐿𝑖 = ∑𝑇𝑖(𝜔𝑖𝑑𝑛𝑖)−𝜃

∑ 𝑇𝑗(𝜔𝑗𝑑𝑛𝑗)−𝜃

𝑗𝑛 𝜔𝑛𝐿𝑛 .

This is like an exchange economy, where countries trade labor units. By choosing a numeraire we can solve for

wages applying the Walras’ Law. Fréchet distributions imply that labor demands are iso-elastic. 44

Authors used relative prices of a range of commodities data to obtain 𝑑𝑛𝑖 , since it couldn’t be estimated

simply as the distance between importer and exporter country as usual. So, they obtained trade elasticity of

“comparative advantage”, parameter 𝜃 for their sample of countries. Their main point is to estimate how much

trade depend on Ricardo’s “comparative advantage”(Jonathan Eaton & Kortum, 2002).

87

It is worth noting that perfect competition will wipe out the taste for varieties by

making goods in a category homogenous with their counterparts. The only distinction

consumers only tell the goods apart by the degree of efficiency at which the goods were

produced. In the case of GMOs, such assumption wouldn’t leave to the situation of a dual-

market system, since products should be considered the same – innovations are strictly

changing processes instead of products.

Before advancing in a more comprehensive analysis of our case under this

framework let’s see how firm heterogeneity theories and technological–gap framework

contributes to our case.

2.2.2 Firm Heterogeneity Models

From the early-2000s we have seen the rise of influential developments in the

field of international trade. The firm heterogeneity models have advanced from many

developments of the New Trade Theory, as increasing returns to scale (IRTS) and

monopolistic competition. It is not by chance that many specialists have usually named this

family of models as the New New Trade Theory (NNTT).

The central contribution of these models is the deliberate consideration of firm-

specific factors as underlying for trade analysis – assuming that countries don’t trade firms

do.

Studies have pointing to the fact that exporting is extremely rare as well as

exporters are larger, more productive, they use factors differently and they pay higher wages

(see Aw, Roberts, & Xu, 2008; Bernard, Eaton, Jensen, & Kortum, 2003; Kugler &

Verhoogen, 2008).

Here, we are going to present and discuss works by Melitz (2003) and Helpman,

Melitz, & Rubestein (2008) – hereinafter HMR(2008) – which will bring some additional

considerations to the Ricardian model discussed in the last section. We are going to see how

models differ in terms of type of competition (monopolistic vs. perfect competition), returns

to scale (increasing vs. constant) and trade costs (variable vs. fixed and variable).

On the other hand, they also resorted to tools like the homothetic preferences

(derived from a CES utility function), identical tastes across countries and distribution

functions to represent heterogeneity in efficiency levels. But, in this matter, HMR (2008)

88

used a Pareto distribution to represent efficiency across firms instead of Fréchet distribution

across countries.

Beforehand, we can say monopolistic competition provides us with a better

framework to think of a market for imperfect substitutes goods, fixed cost of trade are closer

to the type of regulation costs we are dealing with and IRTS can be a reason for country and

firm-level concentration in the supply chain of grains.

Let’s start with the assumptions about production and consumption. For clarity

we adopt the same notation used in the HMR (2008) paper with minor modification.

Consider a word with 𝐽 countries, indexed by 𝑗 = [1,2, … , 𝐽]. As usual, consider

that every country produces and consume a continuum of goods 𝐵𝑗 = [𝑙, … , 𝑙𝑛], where 𝐵𝑗 is

the set of goods available for consumption in country 𝑗 . Utility function of a world

representative consumer will be

𝑢𝑗 = [∫ 𝑥𝑗(𝑙)𝛼 𝑑𝑙𝑙∈𝐵𝑗

]

1𝛼

, 0 < 𝜎 < 1.

(1.b)

Here, 𝑥𝑗(𝑙) is the consumption of product 𝑙 and the parameter 𝛼 is cross elasticity

of demand, defined by the authors as 휀 = 1/(1 − 𝛼). Unlike in Eaton and Kortum (2002)

here this parameter will determine the mark-up each monopolist firm will charge. Goods will

differ not only by the techniques employed to manufacture them but also by consumers’

perception about substitutability of goods being supplied.

Noteworthy, homothetic preferences and identical tastes will make 𝛼 unique for

all firms worldwide. If 𝑌𝑗 is the income of country 𝑗 and consumers maximize utility, product

𝑙’s demand in 𝑗 will be

𝑥𝑗(𝑙) =

��𝑗(𝑙)− 𝑌𝑗

𝑃𝑗1− . 45

(2.b)

Here, ��𝑗(𝑙) is the price for a particular variety 𝑙 in country 𝑗 and 𝑃𝑗 is the

country’s ideal price given by

45

Although authors haven’t derived demand function step-by-step, demand functions from standard CES utility

functions with a continuous of good can be obtained by taking some standard procedures. Most of the grounds

for deriving it are in the Dixit-Stiglitz seminal paper. A simple solution however can be obtained from taking

the ratio of Frisch demands for two varieties, then getting Marshallian demand functions. Some further algebraic

manipulation will return equations (2.b) and (3.b).

89

𝑃𝑗 = [∫ ��𝑗(𝑙)1− 𝑑𝑙𝑙∈𝐵𝑗

]

1/(1− )

.

(3.b)

As we have mentioned above, firms are different within and across countries and

they produce a unique product. Thus, in country 𝑗 exists 𝑁𝑗 firms, and in the world economy

we have a set of ∑ 𝑁𝑗𝐽𝑗=1 firms. Monopolistic competition assumption will make the number

of firms equal to the number of products. Differently from the perfect competition assumed

by Kortum and Eaton (2002), here products are essentially different from one another. In

other words, goods are substitutes to a greater or lesser extent depending on the cross

elasticity of demand and not the degree of efficiency of production.

Each good is produced with a different combination of inputs 𝑎. Note that 𝑎 is

firm specific and varies between the lowest and the highest amount of inputs used in

production [𝑎𝐿 , … , 𝑎𝐻]. Thus, 1/𝑎 is the individual firms’ productivity level. This differences

in technology can be represented by a distribution function 𝐺(𝑎) with support [𝑎𝐿 , 𝑎𝐻] where

𝑎𝐻 > 𝑎𝐿 > 0. For simplicity and without loss of generality authors consider that spread,

location and shape of this distribution are the same for every 𝑗. Differences across firms are

captured by 𝑎′𝑠 and aggregate differences across countries are subsumed in the different

input costs considered for each 𝑗.

With respect to that, let 𝑐𝑗 be the unit cost of a bundle of inputs. In a world with

free mobility of inputs within the countries but not across countries 𝑐𝑖 will be country-

specific. Noteworthy, although authors do not extol this insight, 𝑐𝑗 is capturing all the sources

of country comparative advantage as reported by classical theory. It can account for

differences in the state of technology – 𝑇𝑗 parameter by Eaton and Kortum (2002) – along

with other sources of input costs differences such as factors’ endowments. Thus, 𝑐𝑗 here, by

reflecting all the country-level heterogeneity covers a wider range of sources of comparative

advantage when compared to Eaton and Kortum (2002).

In autarky the combination of heterogeneous firms and input prices under

monopolistic competition and CES preferences will determine prices in country 𝑗 as being

𝑝𝑗(𝑙) = 𝑐𝑗𝑎/𝛼 . This is the standard markup pricing equation, with a smaller markup

associated to a large cross elasticity of demand, or the Mill’s price. The markup is a result of

preference for variety inexistent in Eaton and Kortum (2002).

Trading, however, will bring additional costs to exporting firms, such as tariffs,

regulation, and transportation, among others. These costs can be breakdown into fixed (𝑐𝑗𝑓𝑖𝑗)

90

and variable (𝜏𝑖𝑗) costs. Transportation costs, for instance are expected to vary accordingly to

the amount of trade, whereas regulation as approval of a variety in the importing country,

expenses with labeling requirements, among others, can be considered as fixed cost. The

intuition behind the form of the fixed costs 𝑐𝑗𝑓𝑖𝑗 is that these costs are fueled by internal costs

of exporting country.

For the variable costs will be convenient and simple also make use of the

“melting iceberg” costs formulation. Therefore,

��𝑖𝑗(𝑙) = 𝜏𝑖𝑗

𝑐𝑗𝑎

𝛼 . 46

(4.b)

However, exporters from country 𝑗 will also incur fixed costs of trade. Fixed costs

will be key to determine the profitability of a particular exporter. Given its efficiency level

and country-specific input prices a firm can have no capability to serve a certain market by

trading.

Taking into account the demand as in equation (2b) and variables and fixed costs

of trade we can express the operating profits from any firm in 𝑗 to serve 𝑖 as

𝜋𝑖𝑗(𝑎) = (1 − 𝛼) (𝜏𝑖𝑗𝑐𝑗𝑎

𝛼𝑃𝑖)

1−

𝑌𝑖 − 𝑐𝑗𝑓𝑖𝑗. (5.b)

From that, to carry out a profitable sale to any market 𝑖 a firm must export a

minimum amount of goods to at least pay off the fixed costs of trade. Revenue, however, will

depend on the efficiency level – or technology – employed in manufacturing. In other words,

there is a minimum level of productivity required to serve a market 𝑖 defined as 𝑎𝑖𝑗. Exports

will be profitable only if 𝑎 ≤ 𝑎𝑖𝑗. This cutoff can be derived from the zero profit condition.

(1 − 𝛼) (𝜏𝑖𝑗𝑐𝑗𝑎𝑖𝑗

𝛼𝑃𝑖)

1−

𝑌𝑖 = 𝑐𝑗𝑓𝑖𝑗. (6.b)

Note that both type of trade costs are exogenous to the firm. By considering

variable and fixed costs of trade we also bring into the model the influence of IRTS, since

certain degree of specialization will turn out to be a gain of trade (see Krugman, 1985).

Noteworthy, domestic sells have no further costs from trade, so 𝑓𝑗𝑗 = 0, ∀ 𝑗 and 𝑓𝑖𝑗 > 0, ∀ 𝑖 ≠

𝑗. Analogously 𝜏𝑗𝑗 = 0, ∀ 𝑗 and 𝜏𝑖𝑗 > 0, ∀ 𝑖 ≠ 𝑗. The Ricardian model presented above, by

assuming only variable costs of trade, can neither explicitly treat IRTS as a source of trade

nor consider regulatory costs as fixed – as the empirical literature usually does. These two

46

HMR (2008) have not used the subscript 𝑖 in ��𝑖𝑗 , but as this exporting price will depend on trade-level

variables it is more accurate to include it.

91

considerations put this model closer to the underlying features of our case.

As a result, not all firms will entry into the game of international trade, just those

with efficiency high enough to overcome trade costs. This relatively simple assertion and

natural result from the model is an important feature of HMR (2008). As we mentioned

above, many studies in this field have been pointing to differences between exporting and

non-exporting firms, and high presence of zeroes in bilateral trade data. Indeed, zeroes have

been a challenge not only from the theoretical perspective but also to estimate the gravity

equation, as we are going to see in detail in Chapter III.

All that considered, the bilateral trade volumes can be characterized as

𝑉𝑖𝑗 = {∫ 𝑎1− 𝑑𝐺(𝑎)

𝑎𝐻

𝑎𝐿

for 𝑎𝑖𝑗 ≥ 𝑎𝐿

0 otherwise.

(7.b)

The demand function (2.b) and pricing equation (4.b) determine the value of trade

between 𝑖 and 𝑗.

𝑀𝑖𝑗 = (𝜏𝑖𝑗𝑐𝑗

𝛼𝑃𝑖)

1−

𝑌𝑖𝑁𝑗𝑉𝑖𝑗.

(8.b)

Note that 𝑉𝑖𝑗 = 0 will result in 𝑀𝑖𝑗 = 0. That is, if no firm in country 𝑗 is efficient

enough to export into 𝑖 bilateral trade will be zero. Taking into account the price index and

the definition of equation (7.b) authors define the ideal price index in country 𝑖.

𝑃𝑖1− = ∑ (

𝜏𝑖𝑗𝑐𝑗

𝛼)

1−

𝑁𝑗𝑉𝑖𝑗𝐽𝑗=1 .

(9.b)

The model provides a mapping from the income levels 𝑌𝑖, the number of firms 𝑁𝑖,

the unit costs 𝑐𝑖 , the fixed costs 𝑓𝑖𝑗 and the transport costs 𝜏𝑖𝑗 to the bilateral trade 𝑀𝑖𝑗 .

Authors do not solve to equilibrium. To have a closed form solution further assumptions from

Melitz (2003) will be needed. The author basically resorted to a labor market formulation as

in Krugman (2008), the entry and exit model by Hopenhayn (1992) and common market

clearing conditions to obtain autarky and trade equilibrium. We are not going to advance on

this here, since we are mostly concerned with the general results of the model and the

empirical results we can drawn from the relations between trade and technology innovation47

.

However, gravity can be derived from the model by assuming a Pareto

47

As we mentioned in somewhere dissertation, a model embodying most of the empirical stylized facts should

be developed in the near future.

92

distribution to represent firm heterogeneity 𝐺(𝑎) = (𝑎𝑘 − 𝑎𝐿𝑘)/(𝑎𝐻

𝑘 − 𝑎𝐿𝐾), 𝑘 > (휀 − 1) .

Again, this framework is interesting since it explicitly allows non-positive trade flows and

asymmetric flows between countries.

An overlook at data for soybean trade will return a lot of flows with these

characteristics. If firm heterogeneity is represented by a Pareto distribution, trade volume 𝑉𝑖𝑗

can be expressed as

𝑉𝑖𝑗 =𝑘𝑎𝐿

𝑘− +1

(𝑘 − 휀 + 1)(𝑎𝐻𝑘 − 𝑎𝐿

𝑘)𝑊𝑖𝑗,

(10.b)

where

𝑊𝑖𝑗 = max {(𝑎𝑖𝑗

𝑎𝐿)

𝑘− +1

− 1, 0} . (11.b)

Note that 𝑎𝑖𝑗 is determined by the zero profit condition and free entry, and

equations (10.b) and (11.b) are monotonic functions. Taking this into account and expressing

𝑀𝑖𝑗 in its log-linear form, we have the following estimating equation

𝑚𝑖𝑗 = 𝛽0 + 𝜆𝑗 + 𝜒𝑖 − 𝛾𝑑𝑖𝑗 + 𝑤𝑖𝑗 + 𝑢𝑖𝑗,

(12.b)

where 𝜒𝑖 = (휀 − 1)𝑝𝑖 + 𝑦𝑖 is a fixed effect of the importing country, 𝜆𝑗 = −(휀 − 1) ln 𝑐𝑗 +

𝑛𝑗 is a fixed effect of the exporting country, 𝑑𝑖𝑗 are the variable costs of trade usually

measured as the symmetric distance between country pairs and, finally, 𝑤𝑖𝑗 is the proportion

of firms from 𝑗 exporting to 𝑖. By relating importer and exporter variables controlling for

sizes and trade costs to determine trade flows this can be considered a standard gravity

equation.

Noteworthy authors developed a technique to estimate 𝑤𝑖𝑗 based on the

observable variable 𝑚𝑖𝑗. The intuition behind this is that trade is only observed if at least one

firm is productive enough to do so. Also, trade increases as much as firms are relatively more

productive within a specific country. As we are going to use these insights to estimate most

of our empirical model, we are going to advance in this topic in Chapter III.

In sum, the main contributions of HMR(2008) are the monopolistic competition

assumption and treatment for firm heterogeneity and fixed costs of trade. In this framework,

we could consider GMOs trade as a process and product innovation and the existence of a

dual-market system as a consequence of love for variety. But we still cannot considerer

different tastes across countries.

93

2.2.3 Technological Gap and Trade

This strand of literature was developed under the umbrella of the evolutionary

economics48

. The most influential paper in the field of international trade and the

technological gap was written by Posner (1961), which brought forward the general idea of

continued technology differences among countries as a persistent source of trade.

Given that the developments in evolutionary economics are relatively new it is

often a hard task to outline a pattern or defined focus for the field49

. Therefore, we are going

to focus on some stylized facts which have being discussed in the literature and is related to

the issues we are developing in this study. As these models explicitly abandon the ties of the

equilibrium approach presented in neoclassical models, they bring forth a number of new

ideas, which often contradicts adjustment mechanisms from neoclassical models.

The underlying assumption in these models is the centrality and endogeneity of

technical change as a driver of economic activity, including trade patterns. In an evolutionary

world, firms continuously innovate causing instability and uncertainty in the economic

system. Technological gap theorists are more concerned with the relations between technical

change and trade than with trade equilibrium after a technological shock. The distinction

between technology and technical change is not merely a question of word’s choice.

Technical or technology change has the intrinsic meaning of a economic system marked by

continued breakups arising from innovation, as it may be clear from Rosenberg words:

…in a world where rapid technological change is taking place we may need

an analytical apparatus with focuses in a central way upon the process of

technological change itself, rather than treating it simply as an exogenous

force which leads to disturbances from equilibrium situations and thereby

sets in motion an adjustment process leading to a new equilibrium

(Rosemberg, N. (1970) apud Giovanni Dosi & Soete, 1988 pag. 402).

As uncertainty is present in addition to risk and agents are not fully rational,

instead of maximizing profits, firms adopt and develop routines according to their

experiences in the marketplace, investments in R&D, access to credit markets and random

effects inherent to economic activity. Routines are selected according to their “fitness” to the

48

See Nelson and Winter (1985) for a rich introduction of the field. 49

A detailed description of the modeling approach and concepts of these theories is beyond the scope of this

study, but we will shortly introduce the underlying ideas of these models, especially those related to the case we

are considering. Such an introduction is deeply based on papers by Dosi & Soete (1983), Dosi, (1982);

Giovanni Dosi, Grazzi, & Moschella (2015) and Maggi (1993).

94

actual market conditions. In this way, firms not fitting in will be dropped out of marketplace.

Although departing from very distinct and more realistic assumptions, note that this

framework allows for firms’ selection coming from technological heterogeneity. This idea is

central to understand the pattern of trade after GMOs came into scene and some markets

selected sources based on technological criteria.

Posner (1961) by not assuming identical preferences – an important assumption

for neoclassical model closures50

– explicitly considered the likelihood of a demand lag

impacting on the time needed for a technology adoption rate became critical to the

maintenance of market shares. However, theorists in the field somehow forgot this concept

during these years, perhaps because of the lack of a case in which this lag had such a sharp

effect on trade.

At this point, it is possible to see that these concepts, even if not fully formalized,

are insightful for our analysis of the soybean trade. Asymmetric and asynchronous approval

of new varieties is a persistent and institutionalized source of demand lag and innovation is a

continuous process led by seed companies. Even if we considerer this process trough a set of

partial equilibrium analysis suffering successive technological shocks, as we are more

interested in determining a relationship between technology gaps and trade patterns, the idea

of a technical change seems a more fruitful start point.

To prove this point Posner (1961) departs from very limiting assumptions such as

identical endowment of factors and zero costs of trade. Yet, central and pioneer contributions

lay on the concepts of international imitation lag.

When a firm in country 𝑗 innovates this technology won’t be readily available or

of concern by firms in country 𝑖 , creating a lapse between innovation and imitation or

adoption – or in HMR (2008) terms an increased level of firms’ heterogeneity. The time

elapsed until other firms could fully enjoy the technological benefits will be a result of four

effects. Note that if Eaton and Kortum are implicitly dealing with a world where industry

specialization is expected, being inter-industry trade more relevant, Posner (1961) by

considering that exporting firms can compete with other exporting firms from other countries

wipe out specialization leading to greater relevance of inter-industry trade. Naturally, by

assuming the other extreme of firm specialization in producing a unique product, we also are

assuming that all firms in everywhere are competing with one another.

50

Noteworthy, some neoclassical models have been considering non-homothetic preferences mainly after

Krugman (1979). Usually these models consider that demand in rich countries differs from that of poor

countries see (Dalgin, Mitra, & Trindade, 2004; Fieler, 2011; Hallak, 2006; Hunter, 1991).

95

Anyway, the first effect impacting on the imitation lag is the domestic reaction

lag 𝑙1, defined as the time required to innovation become central in dropping later-mover out

of marketplace51

. Second, the foreign reaction lag 𝑙2 , defined as the time needed to country

𝑖′𝑠 producers feel threaten by importation of country 𝑗′𝑠 products in domestic markets will

play a role. Third, one should also consider that imitation is not a “plug-and-play” process,

thus, agents will need some time to fully enjoy technology benefits as they faces a learning

curve. In other words, there is a learning to learn effect 𝑙3 that also affects the imitation lag.

Fourth and most important, the author considers that it is possible that some consuming

markets took some time to see any advantage in a new product or process, the demand lag 𝑙4.

Thus, this framework allows us to consider a case in which demand lag is strong enough to

make the innovation less destructive in terms of trade restructuration. From our perspective,

that is precisely what evidences are pointing to in the case of GMOs.

Some developments will also consider differences in “endowments” across

different countries (Maggi, 1993). This literature usually calls differences not related to the

rate of innovation across countries as the forces of Ricardian comparative advantage. That is,

these forces are related to an overall level of technology driving domestic factors and inputs

prices instead of technical change itself. We see no reason to not consider immobile factors

“endowment” as an additional driver of domestic prices. Anyway, the general conclusion

about countervailing forces of technical change holds for any source of trade not related to

technical change.

In other words, trade advantages coming from innovation and adoption can be

offset by advantages not related to technical changes – as the reduced cost of labor or land in

other producing countries. Breaking down these two different sources of trade has been a

current concern of literature, to prove the argument of prevalence of technical change as a

source of trade in detriment of labor costs adjustment – a natural result of Ricardian model

under equilibrium conditions (see Giovanni Dosi, Grazzi, & Moschella, 2015).

Another force acting over the advantages of the technical change is the rate of

international diffusion of technology. The intuition behind this is that the benefits of an

innovation made by country 𝑗, can be offset if the international diffusion and adoption by

other competitors is rapid enough.

This framework can shed some lights on questions related to the asymmetric

adoption and absence of strong decreases of market-shares of late-adopters. Countervailing

51

The idea of creative destruction by Schumpeter, first published in 1943, predicts similar patterns of

technological impacts on the economic system (Schumpeter, 2003).

96

forces such as non-technological advantages and pace of adoption itself can explain the

sustainability of late-movers in international markets.

Last but not least, this framework by considering that technical progress will

impact on countries’ market shares, open a room for considering international competition

between firms and countries in supplying destination markets. In this way, it is possible to

consider multi-specialization and trade as a result of international competition across

countries in common markets, instead of thinking of patterns based only on trade pair

characteristics. In other words, if countries 𝑗 and 𝑘 are serving a market 𝑖 with a similar

product, differences between 𝑗 and 𝑘 will also be a source of trade and determine the market

shares.

2.3 Theories of Trade and the Case of GMOs

Finally, we can draw a more specific parallel between the three frameworks

presented and the case of GMOs. We can think this question from three perspectives, i.e. the

market structures, trade and technical change impacts.

In terms of market structure, the monopolistic competition makes sense, since we

can think of a continuum of similar goods to describe soybean varieties. That will be true

especially after the asymmetric technology change, which will increase the range of products

available from the consumer’s perspective. Also, we cannot consider that differences are

enough to allow pure monopoly behavior, being the varieties at the best close substitutes of

one another. Of course, in the real word will be difficult to think of a markup equal to every

country, especially because varieties can enter utility asymmetrically, and preferences can

vary across consumers and countries. Indeed, the weakest argument from the theories

considered, in terms of pursing our goals, is the assumption of identical tastes across

countries.

In this sense, HMR (2008) by implicitly assuming that innovation may affect both

the productivity and cross elasticity of substitution, fits better to answer some questions

raised in this study. In addition, the imitation lag predicted by Posner (1961) also allows for

firm heterogeneity and differences in preferences for “technology”, providing foundation for

our two interest variables, namely the demand lag and technology gap.

In terms of trade, we can think that our case is marked by multi-specialization

with high concentration of exporting and importing countries. Firm, country and trade

specific variables affect the total volume of bilateral trade. Countries with higher levels of

97

technology stock (𝑇𝑖 from Ricardian Model), more efficient firms (𝛼 from HMR), reduced

input costs (𝑐𝑖 from HMR and Ricardian models) and competitive trade costs (𝑓𝑖𝑗 and 𝜏𝑖𝑗

from HMR) will trade more and higher volumes – all else being equal.

The technology change however, will have a double effect in the case of soybean

trade. It will affect firms’ relative levels of productivity (𝐺(𝑎) from HMR and 𝐹(𝑧) from

EK) through the innovation and adoption processes, and consumers’ perceptions through a

kind of preference for “old” production techniques. By being two-way asymmetric, the

innovation will increase firm and good heterogeneity (𝑎) , impacting also on the cross

elasticity of demand (휀) as the new product isn’t considered a better or equivalent substitute

for the old one in everywhere.

It’s also important to consider that regulatory costs were established at the

country-level. So, policymakers will also impact on countries general state of technology,

firms’ heterogeneity, but mainly on fixed costs of trade between adopters and opposing

countries. Considering HMR model, we could consider that 𝑐𝑖𝑓𝑖𝑗 could capture this effect of

increased regulatory costs.

In addition, during the period of analysis new events and asymmetric approval

across countries created a continued technological gap between the countries. Also, that

demand lag was persistent during the time leading to a dual market system accommodating

old and new technologies based on the existence of more than one consuming market.

Adoption process and other non-technological variables were important to keep late adopters

in international markets. This continuous interaction of technological change and trade under

the context of technological diffusion and other factors creating comparative advantages can

be only considered within the technology-gap framework.

These points can be considered into an empirical exercise by assuming that

technical change impacted trade mainly through two variables. First, the non-adoption

created a group of firms producing with an inferior technology, incurring in opportunity costs

of non-adoption, the technology-gap. Second, adoption created a problem for exporter in

markets under demand-lag (or technology hatred).

As most of the asymmetries of trade were created at county-level, with

uncoordinated case-by-case approach to approve new varieties, uncovered approvals – i.e.

varieties being approved in country 𝑗 but not in country 𝑖 – is a good proxy for measuring the

demand-lag. Opportunity costs of delaying approvals, in terms of distance of a technological

frontier given by world rate of innovation and adoption, can be proxied by the difference of

98

approved varieties in country 𝑗 and general state of world’s approvals for production – the

technological-gap. Developing a theoretical model to deal with the major stylized facts of our

case, although important, is a long-term task to be developed in future research.

99

CHAPTER III - EMPIRICAL ESTIMATION

Based primary on the technological-gap and firm heterogeneity frameworks, in

this chapter we estimate a gravity model to analyze the effects of innovation adoption in

terms of both the technology-gap and demand-lag from 1996 to 2012. This chapter is divided

into 4 more sections. In section 3.1 we discuss the method employed and the relationship

with theoretical points presented in Chapter II. In section 3.2 we introduce data used in this

experiment. In Section 3.3 we introduce and discuss the results. In section 5 we conclude and

put together the key findings.

3.1 Method

The gravity equation is a workhorse of international trade analysis. The method

has been used in empirical international economics at least since the seminal work of

Tinbergen (1962). However, it was needed many years before the model could have a

theoretical foundation. This lack of grounds, therefore, made the model a target for several

criticisms. For many authors, the paper by Anderson & Wincoop (2003) filled the gap

between an empirically stable relationship of bilateral trade, sizes and distances, and the

theory of international trade.

In terms of classical theory, it is important to have in mind that HO models is an

attempt to explain the existence of trade flows, which establishes a relationship between

factors proportions (endowment) and trade, whereas Ricardo resorted to technological

differences to explain why do countries trade. In both approaches countries size and trade

costs, made a minor contribution to understanding the stability of trade between similar

countries, the so-called intra-industry trade. On the contrary, this literature was dedicated to

prove that trade exists because of significant effects of complementarities and efficiency

gains of specialization.

A general formulation of the gravity equation can be written as

𝑀𝑖𝑗 = 𝑋𝜆𝑗𝜒𝑖𝜙𝑖𝑗 (1.c)

where, 𝑀𝑖𝑗 is the nominal value of country 𝑗′𝑠 exports to country 𝑖 , 𝜆𝑖 is the importer-effects

affecting trade flows, 𝜒𝑗 are the exporter effects and 𝜙𝑖𝑗 variables defined at the trade level

increasing or decreasing economic costs of trade. 𝑋 represents a general state of international

markets which is unrelated with countries characteristic, or the constant.

100

Importer and exporter effects have been commonly proxied by countries’ nominal

GDPs in aggregated data studies. More recently, many works have advanced on the topic

seeking to add other country-level variables into the model such as the state of technology,

rate of technical progress, regulatory issues, and other factors creating comparative

advantages for a particular country (Gómez-Herrera, 2012; Teh & Piermartini, 2009;

Winchester et al., 2012; WTO & UNCTAD, 2012). Very often, when the research objectives

allow, analysts make use of countries’ fixed effects52

to control for country heterogeneity. In

the particular case of estimating effects for a particular industry or good, it is recommended

to use the gross production in country 𝑗 of this particular good and the consumption

(imported and produced amount) in country 𝑖.

Additional to the general formulation in equation (1.c), many authors consider the

need of controls for multilateral trade resistance (MTR) in order to have a so-called

structured model53

. Actually, neglecting the MTR effects on trade is defined as the gold

medal of gravity misspecification (J. E. Anderson & Wincoop, 2003; Fieler, 2011; WTO &

UNCTAD, 2012). The intuitive concept behind it is that countries with higher MTR or

difficult to access the world markets, will trade less than countries with easier access to them.

Thus, increased trade activity can be a simple result of being closer to the most relevant

markets, for example.

From a most micro-founded perspective, MTR can also been seen as a result of a

model with CES utility function and monopolistic competition. In these models trading

around the world will result in an internal price index. The price index formation, in this case,

is the proxy for MTR in the theoretical model, as it represents an average price of all

suppliers of country 𝑗 weighted by their economic position. If not considered, the MTR

coefficient will be correlated with trade costs, so the others coefficients will be biased.

Virtually all the relevant works after Anderson and Wincoop (2003) have somehow

considered controls for MRT.

A common proxy for MRT has been the “remoteness” index described by Head &

Mayer (2013)54

. The calculation consists of creating a spatial weighted distance by dividing

geographical distance by the countries GDP shares of total world production – or the

52

By country-fixed effects we mean the use of dummy variables to estimate models on cross-sectional data, thus

controlling for country heterogeneity. The term is very common among trade economists but can lead to

misinterpretation if one thinks of fixed effect models in panel data. Importantly, the use of fixed effects for

country avoids the gold medal of misspecification since it controls for remoteness. 53

By structured we mean a model with micro-foundations. 54

Indeed, this index was developed in Head (2003).

101

summation over GPDs of countries in sample. That is, 𝑅𝑒𝑚𝑖 = ∑ 𝑑𝑖𝑗/(𝐺𝐷𝑃𝑗/𝐺𝐷𝑃𝑤)𝑗 , where

𝑑𝑖𝑗 is the geographical distance between countries 𝑖 and 𝑗 , 𝐺𝐷𝑃𝑗 is the country 𝑗′𝑠 gross

domestic product and 𝐺𝐷𝑃𝑤 is the sum of all GDPs of the countries considered (WTO &

UNCTAD, 2012). We have used this procedure to calculate or remoteness index of importers

and exporters.

The trade costs 𝜙𝑖𝑗 have been usually computed as the well-known Samuelson’s

“iceberg costs” formulation. The general idea is that the costs of sending a product from 𝑗 to 𝑖

increase alongside distance55

between this pair of countries. Thus, to a unit of good 𝑙 to arrive

it is necessary to send 𝜏𝑖𝑗 units to country 𝑖, as ad valorem tax. Usually, other variables are

used along with distance to capture other trade costs such as tariffs, cultural and regulatory

differences, and others (Burnquist et al., 2011; Samuelson, 1952; Vigani et al., 2012).

In this particular study, we are going to estimate a gravity equation to analyze

trade of one product and breakdown effects of technological gap and demand lag, as we are

primary concerned with the relationship between trade and over the time, in the context of

certain levels of technology “hatred”.

Thus, our challenge is to estimate a model of trade for one good – or more

precisely a short continuum of goods – and include our technology variables to drawn

conclusions about this relationship.

In doing so, we cannot resort to the convenience of using country fixed effects, as

it would hamper the estimation of technological gap, a country-level variable. But we can

alternatively use country 𝑖′𝑠 production and country 𝑗′𝑠 consumption of these commodities to

control for size. This procedure is not only recommended but also desired when working with

industrial data (Head & Mayer, 2013). The advantage of this procedure is to avoid the

problems of aggregation bias, getting more straightforward coefficients in terms of

interpretation. Noteworthy, without country’s fixed-effect, we need to control for RMT in

order to have a structured gravity equation.

One of the major drawbacks of using industry or one good data is the higher

shares of zero-valued flows, making the standard estimate of log-linear equation tricky.

Simply cutting out zero-valued flows from the sample is not the best solution since it can

potentially create a problem of strong sample selection bias. Yet, zeroes can be meaningful in

some situations as when impeditive fixed costs are playing a role in the chance of a country 𝑖

55

Commonly, distance is calculated by means of the great circle formula, which uses latitudes and longitudes of

the most important cities/agglomerations (in terms of population(Mayer & Zignago, 2011).

102

export to country 𝑗 – what may happen in our case if levels of “hatred” are big enough to

cripple some trade flows.

Econometric tests show that censored or truncated regressions and replacement of

zeroes by arbitrary numbers are biased and also not preferred to two-stage selection models

(Linders & De Groot, 2006). That is the reason why many economists have been using

Heckman (1979) two-stage model to correct for sample bias. In addition, the problem of firm

heterogeneity and impeditive costs was formally treated only recently by HMR (2008).

Also, we want to explicitly evaluate the effects of technology change in trade,

instead of considering only the state of technology in each country. As seen in chapter II, the

technology gap and the demand lag will change over time, and will impact differently on

trade for each period 𝑡. As we have panel-structured data, at first, it is not a problem to get

coefficients adjusted by the changes in these variables. But two others difficulties arise from

using panel data with gravity equation for one product.

First, Heckman type correction for selection bias is not straightforwardly

applicable for panel data. Second and related, HMR (2008) developed a whole model to

assess cross-sectional or pooled data, making the calculation of the controls for firm

heterogeneity and selection bias not ready to go with panel data analysis as well. Several

papers attempted to provide a final solution for Heckman type correction in panels, but until

the present there is no optimal solving for this puzzle (Charbonneau, 2014; Gómez-Herrera,

2012; Martínez-zarzoso, Vidovic, & Voicu, 2014).

All things considered, we decided to go further with HMR (2008) two-stage

adapted approach on pooled data and a FE model on panel data. With this, we can also assess

the stability of the coefficients in both models.

Our final models can account for the potential problems of sample selection,

omitted firm heterogeneity effects and dynamic effects56

of technology change predicted by

Posner (1961). In addition we could also include variables to test for other interesting effects,

such as technology state (Eaton and Kortum 2002) – measured by the countries’ average

productivity – and differences in land availability, since factor endowments can be a

significant source of agricultural goods.

56

By dynamic we mean an effect changing over time.

103

Original model by HMR (2008) can be obtained from the simple linearization of

equation 8.b57

.

𝑚𝑖𝑗 = 𝛽𝑜 + 𝜆𝑖 + 𝜒𝑗 − 𝛾𝑑𝑖𝑗 + 𝑤𝑖𝑗 + 𝑢𝑖𝑗 . (1.d)

For this specification, with lowercase letters representing the natural logarithms of

original variables, 𝜒𝑗 = −(휀 − 1) ln 𝑐𝑗 + 𝑛𝑗 is a fixed effect of the exporting country,

𝜆𝑖 = (휀 − 1)𝑝𝑖 + 𝑦𝑖 is a fixed effect of the importing country and 𝑑𝑖𝑗 is the symmetric

distance between 𝑖 and 𝑗 - with 𝜏𝑖𝑗−1 ≡ 𝐷𝑖𝑗

𝛾휀−𝑢𝑖𝑗 . The new variable 𝑤𝑖𝑗 controls for the

fraction of firms that exports from 𝑗 to 𝑖, possibly zero.

Taking into account our objectives and estimation strategy we adjusted the model

to

𝑚𝑖𝑗𝑡 = 𝛽𝑜 + 𝜑𝑐𝑜𝑛𝑠𝑖𝑡 + 𝛿𝑝𝑖𝑡 + 𝜓𝑝𝑟𝑜𝑑𝑗𝑡 − 𝜚𝑐𝑗𝑡 − 𝛾𝑑𝑖𝑗𝑡 + 𝑤𝑖𝑗𝑡 + 𝑢𝑖𝑗𝑡 , (2.d)

where, we directly control for country 𝑖′𝑠 consumption (𝑐𝑜𝑛𝑠𝑖𝑡) and variables

impacting on price index or remoteness (𝑝𝑖𝑡), and for country 𝑗′𝑠 outcome (𝑝𝑟𝑜𝑑𝑗𝑡 )and

production costs variables (𝑐𝑖𝑗). The control 𝑤𝑖𝑗𝑡 is defined as in HMR (2008) except for the

fact we are controlling firms’ fraction for each period 𝑡, explicitly assuming that the number

of exporting firms can change over time. Note that with this formulation we can compute the

technological gap as a country 𝑗′𝑠 specific-cost affecting the overall production costs. The

demand lag, in turn, will be computed as a type of variable cost of trade, making 𝑑𝑖𝑗 change

over the time58

.

As in HMR (2008), our first-stage consists of estimating a Probit model to

calculate both, the sample selection and the firm heterogeneity controls to be added into the

gravity equation at the second-stage. In addition to returning the controls for firm

heterogeneity and sample bias, by the means of the Probit model we can breakdown the

effects on trade into extensive and intensive margins. That is, how much each variable

impacts on the probability of trade and how much it increases the volume traded between 𝑖

and 𝑗.

57

Equation 8.b delivers the gravity equation from HMR (2008) model.

𝑀𝑖𝑗 = (

𝜏𝑖𝑗𝑐𝑗

𝛼𝑃𝑖)

1−

𝑌𝑖𝑁𝑗𝑉𝑖𝑗. (8.b)

58 For clarity, we are not saying that geographical distance changes over the time, but we are saying that some

variable costs, such as “demand lag” can vary across the t’s.

104

Thus, if firms’ productivity differs within countries in an interval (𝑎𝐿 , . . , 𝑎𝐻), in

which 𝑎𝐿 is the most productive and 𝑎𝐻 the least productive firm. Assume that productivities

variation can be represented by a Pareto distribution 𝐺(𝑎) = (𝑎𝑘 − 𝑎𝐿𝑘)/(𝑎𝐻

𝑘 − 𝑎𝐿𝐾), 𝑘 >

(휀 − 1). Though, only a share of firms will export – those with productivity high enough to

serve a market 𝑗 and breakeven fixed costs of trade.

The selection of firms into exporting markets is determined by a cutoff 𝑎𝑖𝑗, which

is implicitly defined by the zero profit condition – see equation (6.b)59

. Thus, we can define a

latent variable 𝑍𝑖𝑗 as (omitting time subscript for simplicity)

𝑍𝑖𝑗 =

(1 − 𝛼) (𝑃𝑖𝛼

𝑐𝑗𝜏𝑖𝑗)

−1

𝐶𝑂𝑁𝑆𝑖𝑎𝐿(1− )

𝑐𝑗𝑓𝑖𝑗 .

(3.d)

Note that 𝑍𝑖𝑗 is the ratio of variable export profits to the fixed costs for exports

from 𝑗 to 𝑖 by the most productive firm in country j. Positive exports are observed if and only

if 𝑍𝑖𝑗 > 1. Otherwise, if the most productive firm in 𝑗 cannot export to 𝑖, then no other firm

can. As a result, trade will be 0.

In this case, 𝑊𝑖𝑗60 is a monotonic function of 𝑍𝑖𝑗, that is, 𝑊𝑖𝑗 = 𝑍𝑖𝑗

(𝑘− +1)/( −1)−

1. 61

Fixed export costs are stochastic due to unmeasured trade frictions 𝑣𝑖𝑗 that are i.i.d., but

may be correlated with the residuals of the second stage estimation 𝑢𝑖𝑗′𝑠 . Let 𝑓𝑖𝑗 ≡

exp (𝑘𝜙𝑖𝑗 − 𝑣𝑖𝑗), where 𝑣𝑖𝑗~𝑁(0, 𝜎𝑣2), and 𝑘𝜙𝑖𝑗 is an observed measure of any country-pair

specific fixed trade costs. Note that unlike HMR (2008) we are assuming that fixed trade

costs only exist because of the interaction between 𝑖 and 𝑗. This assumption is appropriate for

our purposes since we are assuming that fixed costs will be a result of differences in

approved varieties for commercialization in 𝑖 and for production in 𝑗.

59

Remember we are assuming a Pareto distribution for firm heterogeneity with productivities varying from 𝑎𝐿

to 𝑎𝐻. Zero profit condition was defined as

(1 − 𝛼) (

𝜏𝑖𝑗𝑐𝑗𝑎𝑖𝑗

𝛼𝑃𝑖)

1−

= 𝑐𝑗𝑓𝑖𝑗. (6.b)

60

From the zero profit condition (6.b), and equation (11.b) presented below:

Wij = max {(aij

aL

)k−ε+1

− 1, 0} . (11.b)

61 See equations 4 and 8 in HMR(2008) or the theoretical model presentation in Chapter 2.

105

The number of approved varieties between different 𝑗𝑠 , that is the exporting

countries, may have an impact on average costs and yield, but is mainly a proxy for

innovation or imitation capacity. In other words, approvals are a proxy for countries 𝑗′𝑠

technological gap in relation to a technological frontier defined by technologies available and

adopted by other important exporters operating in the same industry. It is a new concept in

the literature, and future theoretical treatment can consider it as type of “technological

remoteness” determining countries’ importance given the level of technology and technical

progress.

Using this specification together with (휀 − 1) ln 𝜋𝑖𝑗 ≡ 𝛾𝑑𝑖𝑗 − 𝑢𝑖𝑗 , the latent

variable 𝑧𝑖𝑗𝑡 ≡ ln 𝑍𝑖𝑗𝑡 can be expressed as

𝑧𝑖𝑗𝑡 = 𝛾0 + 𝜑𝑐𝑜𝑛𝑠𝑖𝑡 + 𝜍𝑝𝑖𝑡 + 𝜓𝑝𝑟𝑜𝑑𝑗𝑡 − 𝜚𝑐𝑗𝑡 − 𝛾𝑑𝑖𝑗𝑡 − 𝑘𝜙𝑖𝑗𝑡 + 𝜂𝑖𝑗𝑡 . (4.d)

where 𝜂𝑖𝑗𝑡 ≡ 𝑢𝑖𝑗𝑡 + 𝑣𝑖𝑗𝑡 ~ 𝑁(0, 𝜎𝑢𝑡2 + 𝜎𝑣𝑡

2 ) is i.i.d. (yet correlated with the error term 𝑢𝑖𝑗𝑡 in

the gravity equation).

We know that 𝑧𝑖𝑗𝑡 > 0 when 𝑗 exports to 𝑖 , and 𝑧𝑖𝑗𝑡 = 0 when it does not.

Moreover, the value of 𝑧𝑖𝑗𝑡 affects the export volume. Thus, let’s define the indicator variable

𝑇𝑖𝑗𝑡 to equal 1 when country 𝑗 exports to 𝑖 and 0 when it doesn’t. Let 𝜌𝑖𝑗𝑡 be the probability

that 𝑗 exports to 𝑖, conditional on the observed variables62

. Thus we can specify the Probit

model as

𝜌𝑖𝑗𝑡 = Pr(𝑇𝑖𝑗𝑡 = 1|𝑜𝑏𝑠𝑒𝑟𝑣𝑒𝑑 𝑣𝑎𝑟𝑖𝑎𝑏𝑙𝑒𝑠)

= 𝛷(𝛾0 + 𝜑𝑐𝑜𝑛𝑠𝑖𝑡 + 𝜍𝑝𝑖𝑡 + 𝜓𝑝𝑟𝑜𝑑𝑗𝑡 − 𝜚𝑐𝑗𝑡 − 𝛾𝑑𝑖𝑗𝑡 − 𝑘𝜙𝑖𝑗𝑡)

(5.d)

where 𝛷(. ) is the cdf of the unit-normal distribution. According to HMR (2008) this

selection equation was derived from a firm level decision, and it therefore does not contain

the unobserved and endogenous variable 𝑊𝑖𝑗 that is related to the fraction of exporting firms.

Moreover, the Probit can be used to derive consistent estimates of 𝑊𝑖𝑗𝑡.

62

As in HMR (2008) we divided equation (5.d) by the standard deviation 𝜎𝜂 before specifying the Probit

equation to avoid imposing conjunct normality (𝜎𝜂2 ≡ 𝜎𝑢

2 + 𝜎𝑣2 = 1) . However we omitted the star as a

superscript to indicate it. Empirical estimations, as suggested by WTO & UNCTAD (2012), usually ignores this

step. We run models with and without this procedure and we have found no meaningful differences in the model

coefficients.

106

Let ��𝑖𝑗𝑡 be the predicted probability of exports from 𝑗 to 𝑖, using the estimates

from the Probit equation, and let ��𝑖𝑗𝑡 = 𝛷−1(��𝑖𝑗𝑡) be the predicted value of the latent variable

𝑧𝑖𝑗𝑡. Then a consistent estimate for 𝑊𝑖𝑗 can be obtained from

𝑊𝑖𝑗𝑡 = max {(𝑍𝑖𝑗𝑡)𝛿𝑡

− 1, 0},

where 𝛿𝑡 ≡ 𝜎𝑛𝑡(𝑘 − 휀 + 1)/(휀 − 1).

(6.d)

According to our assumptions, consistent estimation of the log-linear model

requires control for both the endogenous number of exporters (via 𝑤𝑖𝑗) and the selection of

country pairs into trading partners (which generates a correlation between the unobserved 𝑢𝑖𝑗

and the dependent variables). Thus, estimates for 𝐸[𝑤𝑖𝑗|. , 𝑇𝑖𝑗 = 1] and 𝐸[𝑢𝑖𝑗𝑡|. , 𝑇𝑖𝑗𝑡 = 1] are

needed.

Additionally, 𝐸[𝑢𝑖𝑗𝑡|. , 𝑇𝑖𝑗𝑡 = 1] = 𝑐𝑜𝑟𝑟(𝑢𝑖𝑗𝑡, 𝜂𝑖𝑗𝑡) (𝜎𝑢𝑡

𝜎𝜂𝑡) ��𝑖𝑗𝑡 . As 𝜂𝑖𝑗𝑡 has a unit

normal distribution, the inverse Mills ratio, as in Heckman (1979) seminal paper, is thus a

consistent estimation of ��𝑖𝑗𝑡63 . Therefore, 𝑧𝑖𝑗𝑡 ≡ ��𝑖𝑗𝑡 + ��𝑖𝑗𝑡 is a consistent estimate for

𝐸[𝑧𝑖𝑗𝑡|. , 𝑇𝑖𝑗𝑡 = 1] and ��𝑖𝑗𝑡 ≡ ln{exp[𝛿(��𝑖𝑗𝑡 + ��𝑖𝑗𝑡)] − 1} is a consistent estimate for

[𝑤𝑖𝑗𝑡|. , 𝑇𝑖𝑗𝑡 = 1] . Thus the model can be estimate using the transformation

𝑚𝑖𝑗𝑡 = 𝛽0 + 𝜑𝑐𝑜𝑛𝑠𝑖𝑡 + 𝜍𝑝𝑖𝑡 + 𝜓𝑝𝑟𝑜𝑑𝑗𝑡 − 𝜚𝑐𝑗𝑡 − 𝛾𝑑𝑖𝑗𝑡 + ln{exp[𝛿(��𝑖𝑗𝑡 + ��𝑖𝑗𝑡)] − 1}

+ 𝛽𝑢𝜂 ��𝑖𝑗𝑡 + ℯ𝑖𝑗𝑡

(7.d)

where 𝛽𝑢𝜂 ≡ 𝑐𝑜𝑟𝑟(𝑢𝑖𝑗𝑡 , 𝜂𝑖𝑗𝑡) (𝜎𝑢𝑡

𝜎𝜂𝑡) and ℯ𝑖𝑗𝑡 is an i.i.d error term satisfying 𝐸[𝑒𝑖𝑗| . , 𝑇𝑖𝑗 =

1] = 0. Note that equation (7.d) is nonlinear in 𝛿𝑡. However, HMR (2008) tests indicate that

a linear model can be estimated. Following the paper specifications we dropped the Pareto

assumption 𝐺(. ) and revert it to the general specification for 𝑉𝑖𝑗𝑡64. Thus, 𝑣𝑖𝑗𝑡 ≡ 𝑣(𝑧𝑖𝑗𝑡) is

now an arbitrary and increasing function of 𝑧𝑖𝑗𝑡 . Now, it is possible to control for

𝐸[𝑉𝑖𝑗𝑡 |. , 𝑇𝑖𝑗𝑡 = 1] using 𝑣(𝑧𝑖𝑗𝑡), approximated with a polynomial in 𝑧𝑖𝑗𝑡 replacing ��𝑖𝑗𝑡 ≡

63

The inverse Mills ratio is written as ��𝑖𝑗 = 𝜙(��𝑖𝑗)/𝛷(��𝑖𝑗), that is, the probability density function (pdf) over

the cumulative density function (cdf).

64 𝑉𝑖𝑗 = {∫ 𝑎

𝑎𝑖𝑗

𝑎𝐿

1−for aij ≥ aL

0 otherwise is presented and discussed in the previous chapter.

107

ln{exp[𝛿𝑡(��𝑖𝑗𝑡 + ��𝑖𝑗𝑡)] − 1} in the equation at the second-stage. Accordingly, our final

estimate equation will be

𝑚𝑖𝑗𝑡 = 𝛽0 + 𝜑𝑐𝑜𝑛𝑠𝑗𝑡 + 𝜍𝑝𝑗𝑡 + 𝜓𝑝𝑟𝑜𝑑𝑖𝑡 − 𝜚𝑐𝑖𝑡 − 𝛾𝑑i𝑗𝑡 + 𝛽1𝑧𝑖𝑗𝑡

+𝛽𝑢𝜂 ��𝑖𝑗𝑡 + ℯ𝑖𝑗𝑡.

(8.d)

As we mentioned before, the estimation of the controls for sample selection and

firm heterogeneity are not straightforward in the case of panel data. There are a few empirical

studies that tackle sample selection models for panel data, but none of them are conclusive in

terms of what is an optimal estimation approach. Among them, an even smaller number of

papers deal with control for firm heterogeneity as in HMR (2008), and again, none of them is

conclusive about optimal estimation procedure.

Most of the papers apply the solution proposed by Wooldridge (1995, p.121-130)

which makes use of a Chamberlain-Mundlak approach (see works by Egger et al (2009);

Egger and Pfaffermayr (2011)). Of course, the approach is focused in treating sample bias

instead of dealing with the firm heterogeneity control proposed by HMR (2008). Additional

tests can be used to diagnosis sample selection bias prior to employing corrections, which are

in general more complicated (Semykina & Wooldridge, 2010; Wooldridge, 1995).

In this study we chose to estimate a Probit for each 𝑡. In this way, we also have a

𝑧𝑖𝑗 and ��𝑖𝑗 for each 𝑡. This approach was used by Martínez-zarzoso, Vidovic, & Voicu (2014)

returning results aligned with other studies, although authors are not stickily employing the

corrections procedures presented in Wooldridge (1995). Instead, they are applying the testing

procedures for sample bias to have the controls. Our estimates of different models with and

without the controls provide more robustness to coefficient values as we are not strictly

employing the strategy used by HMR (2008). But, as HMR (2008) explicitly assume the

existence of firm heterogeneity, the estimates could be considered inconsistent without these

controls.

Testing for sample bias comprises estimating a Probit model – from equation

(4.d) – for each period 𝑡 and, then estimate an OLS with pooled data with controls for the

selection biases (Wooldridge 1995, p.121-130). If the controls are significant in the pooled

model, thus estimates of fixed effect on panel data isn’t efficient without the correction for

sample selection bias. It is important to note that this test was developed to assess sample

selection bias, but we are including HMR (2008) firm heterogeneity control in the same way,

108

as the calculation of firm heterogeneity controls has the inverse Mills ratio as a component.

Test results can be seen in Table 12.

Table 12 – Test for sample selection and firm heterogeneity Biases

Dependent variable: Bilateral trade

Polled – Test

��𝑖𝑗𝑡 -0.394

(0.257)

��𝑖𝑗𝑡 -0.690

(1.279)

��𝑖𝑗𝑡2 0.081

(0.541)

��𝑖𝑗𝑡3 -0.010

(0.063)

Constant -19.618***

(2.617) Observations 6,634

R2 0.420

Adjusted R2 0.418

F Statistic 199.352*** (df = 24; 6609)

Note: *p<0.1; **p<0.05; ***p<0.01

Independent variables omitted

for clarity. Robust stand errors.

Note that neither ηijt nor zijt

coefficients are significant in the pooled data model,

meaning that estimate via FE model on panel data is consistent accordingly to this test. Our

FE model has as individuals the country-pairs and time dimension from 1996 to 2012. Taking

country-pairs as ids is very common in the literature (Baltagi, Egger, & Pfaffermayr, 2014;

Gómez-Herrera, 2012).

We carried out some tests to confirm that FE model is preferred over alternatives.

An F test for FE model and OLS, with the null hypothesis of non-significant effect of

heterogeneity across individuals, returned a p-value < 2.2e-16. In the presence of significant

effects for individuals heterogeneity FE model is preferred over a simple OLS with pooled

data.

FE is also a better choice when compared to Random effects, if we use a

Houseman test with the null hypothesis of non-correlation between errors and regressors, we

get a p-value =1.914e-13. Thus, the Random Effect model is inconsistent (see Green, 2008,

chapter 9).

A Lagrange Multiplier Test (Breusch-Pagan), with null hypothesis of no need for

fixed time effect, pointed to the need for including time fixed effect in our FE model (p-

value< 2.2e-16).

109

Regarding model fitness, we tested for serial correlation, heteroskedasticity and

multi-collinearity. Breusch-Godfrey/Woodridge test don’t allow us to reject the hypothesis

for serial correlation (p-value = 5.117e-08). Also, the Breusch-Pagan test for panel data

revealed the problem of heteroskedasticity (p-value < 2.2e-16). We treated both problems

with “arellano” correction in R cran. Thus we got a robust covariance matrix of parameters

for fixed effects panel data according to White method(Arellano, 1987; White, 1980).

Finally, testing for multi-collinearity (for polled data) returned no Variation

Inflation Factor (VIF) higher then 5, except for controls and powered variables.

All the models and tests were run in the Comprehensive R Archive Network (R

Cran) making use of plm, glm, sandwich, lmtest, sampleSelection, car and tseries. Box 1

summarizes the major estimates steps.

Box 1 – Summary of Major Estimates Steps

Step # Name Description

1 Database Org. Data gathering and organization.

2 1st Stage - Probit Estimation of equation 5.d to obtain controls for firm heterogeneity

and sample bias. We estimated one model per year – 18 in total.

3 HMR-Model Estimation of cross sectional data (OLS) with fixed effect per year

(equation 7.d)

4 Test for Sample Selection Estimation of pooled data, id= country pair, with controls for firm

heterogeneity and sample bias based on Wooldridge (1995), p.121-

130. Test returned no significant coefficient for both controls.

5 FE or OLS? F test, H0=non-significant heterogeneity across individuals( p-value <

2.2e-16).

6 FE or Random Effects? Houseman Test, H0= non-correlation between errors and regressors

(p-value =1.914e-13).

7 FE with fixed effect for

time?

Lagrange Multiplier Test (Breusch-Pagan), H0=no need for fixed

time effect (p-value< 2.2e-16).

8 FE + time fixed effect

(Table 14 - column 4)

Baseline model without the controls from step 2 and technological

variables.

9 FE (Table 14 - column 5) Baseline model + interest variables Tech. Gap and Demand Lag.

10 FE (Table 14 - column 6) Baseline model + variables Tech. Gap + Demand Lag + controls

estimated in step 2.

11 FE (6) serial correlation? Breusch-Godfrey/Woodridge test, H0= no serial correlation. p-value

= 5.117e-08

12 FE (6)

heteroskedasticity?

Breusch-Pagan test, H0= no heteroskedasticity. p-value < 2.2e-16.

13 Multi-collinearity

(pooled data)?

Test carried out on pooled data. No Variation Inflation Factor (VIF)

higher then 5, except for controls and powered variables.

14 FE (6) + arellano (Table

14 - column 7)

We treated problems in step 11 and 12 with “arellano” correction.

Thus we got a robust covariance matrix of parameters for fixed

effects panel data according to White method(Arellano, 1987; White,

1980).

Source: Prepared by the authors.

110

3.2 Data

Although trade data availability is increasing, to extend analysis out of OCDE

countries are still a considerable challenge since a lot of countries have not organized

information that easily integrates with trade-level data from some structured sources.

Data on soybean trade flows used in this study comes from BACI database

developed by CEPII at a high level of product disaggregation. This database is based on

original data reported by the United Nations Statistical Division (COMTRADE database).

BACI was built using an original approach that reconciles the declarations of the exporter

and importer – enabling to considerably extend the number of countries in the dataset (for

detailed information see Guillaume & Zignago, 2010). General gravity data – distance,

colonial ties, common language, contiguous or landlocked territories, among others, are from

GeoDist database also by CEPII (for detailed information see Mayer & Zignago, 2011).

Agricultural related data, such as production, average yield, arable-land and others

were collected from the FAOSTAT database. Data on exchange rates comes from the

International Monetary Fund (IMF), and prices data from World Bank Commodity Price Data

(The Pink Sheet).

Country-level data on biosafety used to build our technological variables comes

from different sources. Data on approval of genetically modified varieties comes mainly from

International Service for the Acquisition of Agri-biotech Applications (ISAAA) database.

However, when needed – because of missing or incomplete information – data was fulfilled

by information from Global Agricultural Information Network (GAINS) report by the

Foreign Agricultural Service of the USDA (FAS-USDA) and data from Biosafety-Clearing

House (BCH) databases.

The GAINS reports are country-specific reports prepared by country authorities

relating the general state of technology regulation, public view and adoption and other related

issues. The BCH databases includes notification of first transboundary movements of living

modified organisms (LMO), and countries’ profile containing number of experts, number of

laws and other regulations on the topic, and copy of regulatory documents for public

consultation.

Complementary information was required mainly because not all countries have

passed regulation to prior risk analysis before approving a new GM variety for any purpose.

Thus, a zero approval could mean unrestricted imports (approval of all varieties being

produced) or a general ban on importing GM food/feed. In general, GAINS reports for

111

several years were consulted to build the variable for mandatory labeling, and when required

(because of missing of information) we consulted legal documents issued by country

authorities to fulfill the blanks.

To estimate sample selection and firm-heterogeneity biases it was necessary to

add to database all possible combinations of trade partners. Reasonably countries that have

no commercial relation with any of the other countries for a given year were dropped from

the sample. These type of non-positive flows have no meaning information to our study.

The final database used for the estimation of the Probit model contained 39,751

observations for 16 years, 42 variables and 84 countries. From those, only 6,634 observations

had positive trade flows. The significant reduction of the database calls for a correct assess of

sample selection bias to assure that estimates are consistent. However, as we have seen in 3.1,

according to Wooldridge (1995) test, estimates of FE model isn’t inconsistent without these

controls in our case. Nonetheless, these controls enter the HMR (2008) model via the

specification of the theoretical model.

Final panel database, using the controls generated in first stage has 6,634

observations (id=country-pairs) and 44 variables, containing trade flows of soybean between

1995 and 2012 for 84 countries. In Box 2 we briefly describe the variables actually used in

the models.

Box 2– Model’s Variables Descriptions

Variable Name Description Source

Bilateral Trade Natural logarithm of annual exports of soybean (HS6-120100) from country 𝑗 to

country 𝑖, calculated from the original variable “v”.

BACI-CEPII

Production j

Natural logarithm of annual outcome of soybean of country 𝑗, calculated from the

original variable “Production Quantity” given in metric tons. It includes declared

and estimated data.

FAOSTAT

Consumption i

Natural logarithm of annual domestic supply (imports-exports+stocks) of soybeans

in country 𝑖, including uses as food, feed, seed, processing, waste and others given

in metric tons.

FAOSTAT

Distance

Natural logarithm of geodesic distances between most populated cities in 𝑗 and 𝑖. Distance (or dist in original database) was calculated following the great circle

formula.

GeoDist -

CEPII

Land Border Dummy variable assuming value 1 if 𝑖 and 𝑗 are contiguous, 0 otherwise. GeoDist -

CEPII

Language Dummy variable assuming value 1 if 𝑖 and 𝑗 share a common language spoken for

at least by 20% of the population, 0 otherwise.

GeoDist -

CEPII

Colony Dummy variable assuming value 1 if 𝑖 and 𝑗 had/have a colonial tie, 0 otherwise. GeoDist -

CEPII

Same Country

Dummy variable assuming value 1 if 𝑖 and 𝑗 is considered the same country. Value

1 is settled when countries were/are considered the same state or the same

administrative entity for a long period (25-50 years in the twentieth century, 75 year

in the ninetieth and 100 years before).

GeoDist -

CEPII

Continued…

112

Variable Name Description Source

Exch. Rate 𝑖 Natural logarithm of the inverse of country 𝑖′s official annual exchange rate

given in value of local currencies in terms of 1 US dollar (1/ex_rate).

FMI database

Exch. rate 𝑗 Natural logarithm of the inverse of country 𝑗′s official annual exchange rate

given in value of local currencies in terms of 1 US dollar (1/ex_rate).

FMI database

Landlocked j Dummy variables assuming value 1 if a country 𝑗 is landlocked. GeoDist -CEPII

Landlocked i Dummy variables assuming value 1 if a country 𝑖 is landlocked. GeoDist -CEPII

Yield j

Natural logarithm of country 𝑗’𝑠 average yield per unit of harvested area for

crop products. In most of the cases yield data are not recorded but obtained

by dividing the production data by the data on area harvested.

FAOSTAT

Land

Natural logarithm of differences between 𝑖 and 𝑗 in terms of land under

temporary agricultural crops (multiple-cropped areas are counted only

once), temporary meadows for mowing or pasture, land under market and

kitchen gardens and land temporarily fallow (less than five years). Variable

named as arable-land in the original database.

FAOSTAT

RTA

Dummy variable assuming value 1 if countries 𝑖 and 𝑗 make part of

Regional Trade Agreement (RTA), and 0 otherwise.

International

Economics Data and

Programs - (Sousa,

2012)

Other Goods 𝑖

Natural logarithm of an index built by aggregating annual imports of major

substitutes goods for soybean meal by 𝑖, such as meat meal (HS6-230110),

fishmeal (HS6-230120), cottonseed meal (HS6-230610), linseed meal

(HS6-230620) and groundnut meal (HS6-230500). Value given in tones.

BACI-CEPII

Mand. Label Dummy variable assuming value 1 if country 𝑖 has implemented mandatory

labeling rules and country 𝑗 hasn’t, and 0 otherwise.

GAINS Report -

USDA

Price

Natural logarithm of average annual prices of soybean in global markets

given in nominal USD.

World Bank

Commodity Price Data

( The Pink Sheet)

Remoteness 𝑖 Natural logarithm of country 𝑖’𝑠 𝑎𝑛𝑛𝑢𝑎𝑙 remoteness index as proposed by

Head (2003) – i.e. 𝑅𝑒𝑚𝑖𝑡 = ∑ 𝑑𝑖𝑗/(𝐺𝐷𝑃𝑗𝑡/𝐺𝐷𝑃𝑤𝑡)𝑖𝑡 . Annual nominal GDP

data comes from IMF databases and distance from GeoDist.

IMF Data

GeoDist -CEPII

Remoteness 𝑗

Natural logarithm of country 𝑗′𝑠 remoteness index as proposed by Head

(2003) – i.e. 𝑅𝑒𝑚𝑖𝑡 = ∑ 𝑑𝑖𝑗/(𝐺𝐷𝑃𝑖𝑡/𝐺𝐷𝑃𝑤𝑡)𝑗𝑡 . Annual nominal GDP data

comes from IMF databases and distance from GeoDist.

IMF Data

GeoDist -CEPII

Tech. Gap

Variable calculated as the difference between numbers of approved varieties

for production in country 𝑗and the total varieties of GM-soybean adopted by

other 𝑗’𝑠. Raw data includes approval of GM-variety in producing countries.

ISAAA approval

database

Demand Lag

Variable calculated as the difference between number of approved varieties

for production in country 𝑗 and the total varieties approved for consumption

in country 𝑖. Original data includes approval of GM-variety in producing

and importing countries.

ISAAA approval

database

Source: Prepared by the author.

3.2.1 Variables Transformation

In this subsection we shortly provide further information on transformation we

have made in some interest variables.

General approach to create the variable Demand lag consisted of three steps. First,

we gathered data on approvals for cultivation, food and feed for all the varieties of soybean

available, as reported in ISAAA database. Second, for each 𝑡 and variety we created a

dummy variable (𝑎𝑠𝑠𝑖𝑗) assuming value 1 when country 𝑖 has approved a variety for

113

cultivation not approved for consumption as food, feed or both in country 𝑗65. Third, we

aggregated the values of dummy variables per year to have the total of demand lag for each

pair of countries in year 𝑡.

Further steps were necessary to solve the problem of including countries with

“zero” approved events into the model. We know that many countries have not implemented

regulatory frameworks, or have no assessment capacity to carry out tests for identification of

imports of unauthorized events. On the other hand, we also know that some countries had

banned importation of any GMOs into their territories. The combination of these issues turns

out to lead to two equally factual interpretations for zero approval. Some of them should be

considered as a general approval for any event available, and others should be considered as

actual bans. A related problem emerges when a country implements risk assessments

measures after 1996 – the first year of commercial release of GMOs.

Therefore, we analyzed all zero approvals case-by-case to determine if the value

of total events approved for each year should be ascribed, or zero approval actually means a

ban. Underlying information to discriminate between these two types of zero came from

GAINS reports, but also from BCH archives and the Africa Centre for Biosafety report

(Moola & Munnik, 2007; BCH, 2015; FAS, 2015). As bans to GMOs were usually of public

concerns and well disseminated across specialist and media, we created a dummy for bans to

correct for false zero approvals (𝑏𝑎𝑛𝑗). If a ban were reported by one of the data sources for

a specific 𝑡, so the dummy variable was set to 1. Otherwise, if no ban was reported, the

dummy is set to 0. By multiplying both we get the adjusted asymmetry between 𝑖 and 𝑗 in

terms of approved varieties – i.e. (𝑎𝑠𝑠𝑖𝑗 ∗ 𝑏𝑎𝑛𝑗 = 𝑎𝑑𝑗𝑢𝑠𝑡𝑒𝑑_𝑎𝑠𝑠_𝑖𝑗).

In the case of Tech. Gap we took a more pragmatic approach. We simply found

the total number of different varieties available and approved for production in at least one

country 𝑗 for a given 𝑡 and subtracted the number of approved varieties in particular country 𝑗

to have a “distance” between country varieties and the state of the art of technology. That was

possible because major producing countries have taken very clear legal positions towards the

technology since the early-1990s. It’s worth noting that we haven’t used logarithm

transformation to these variables, as they are derived of dummy variables.

Lastly, it is important to relate some minors handling on independent variables,

beyond linearization, intended to increase the sample size. We have summed a constant

(0.01) to arable-land and Other Goods 𝑖 original entries to avoid a significant loss of

65

This database will be used to analyze impacts of individual approvals in trade in a future analysis.

114

observations. Although not formally proved as consistent, analysts broadly use censoring

techniques in econometric studies, especially in the case of independent variables. From the

economic perspective we can assume that most of zeroes are missing data for countries with

insignificant share of lands classified as arable-land or imports of meals derived of other

vegetable sources.

3.3 Results and Discussion

In this section we present and discuss our findings. First, it will be helpful to

retake our model and our main assumptions to go further on the analysis. As in equation (8.d)

our model predicts trade flows as a function of countries 𝑖 and 𝑗′𝑠 sizes, country j's costs of

production, country i's index price or remoteness, fixed and variables costs of trade and the

additional controls for firm heterogeneity (or fraction of exporting firms) and sample bias.

Our interest variables are defined as the Demand Lag between trade partners and Tech. Gap

between exporting countries in international markets.

As usual, country’s size is expected to increase bilateral trade, whereas production

and trade costs and remoteness are expected to decrease. Firm heterogeneity is expected to be

a significant control when variability of individual firms productivity are high, making with

some countries have lower fraction of firms productive enough to breakeven the fixed cost of

trade. Sample selection bias will be significant when biases generated by the unobserved

country-pair level shocks 𝑢𝑖𝑗 and 𝜂𝑖𝑗 are high. As our interest variables and the case study

itself point to the likelihood of strong negative effects of high level of “technology hatred”, it

is possible that some flows dropped to zero after adoption in some producing countries.

Finally, we expect that asymmetry of approval between trade partners decreases

trade since it increases the overall costs of trading. On the other hand, we expect that

opportunity costs of delaying adoption will also decreases overall bilateral trade as adoption

can cut down production costs and not all countries are averse to the technology. Noteworthy,

none of the studies we had access controlled for the opportunity costs of not adopting the

technology. Descriptive statistics can be consulted in Table 13. See variables’ description in

Box 2 for more details on variables transformations.

115

Table 13 – Descriptive Statistics

Statistic N Mean St. Dev. Min Max

t 6,634 2,003.9 5.0 1,995 2,012

Bilateral Trade 6,634 5.6 3.6 0.0 16.5

Production 𝑗 6,634 13.7 3.5 3.4 18.3

Distance 6,634 8.3 1.2 4.1 9.9

Land Border 6,634 0.2 0.4 0 1

Language 6,634 0.2 0.4 0 1

Colony 6,634 0.1 0.2 0 1

Same Country 6,634 0.05 0.2 0 1

Price 6,634 5.8 0.3 5.3 6.4

Land 6,634 1.3 2.3 -6.1 16.6

Yield 𝑗 6,634 10.0 0.3 8.1 10.6

Tech. Gap 6,634 7.4 3.9 0 15

Mand. Label 6,634 0.3 0.4 0 1

Demand Lag 6,634 1.5 2.9 0 15

RTA 6,634 0.3 0.5 0 1

Exch. Rate 𝑖 6,634 -1.8 2.5 -9.9 3.1

Exch. Rate 𝑗 6,634 -1.9 2.5 -9.9 3.1

Consumption 𝑖 6,634 5.6 2.9 -2.3 11.2

Other Goods 𝑖 6,634 10.8 2.0 -3.9 14.3

Landlocked 𝑗 6,634 0.2 0.4 0 1

Landlocked 𝑖 6,634 0.1 0.3 0 1

Remoteness 𝑗 6,634 18.0 1.0 16.5 22.3

Remoteness 𝑖 6,634 17.8 0.9 15.7 24.3

��∗𝑖𝑗𝑡

6,634 1.0 0.6 0.0 3.5


6,634 0.9 0.5 0.3 6.3


2 6,634 1.0 1.7 0.1 39.3


3 6,634 1.5 7.3 0.02 246.2

Source: prepared by the authors.

As seen in the method section, we estimated the model in the second stage using

two different estimators (Two-Stage HMR and FE model). For the FE model, we

progressively introduced our interest variables and controls to better assess the coefficients

stability. Results can be seen in Table 14 below.

116

Table 14 – Estimates Results

Dependent variable: Bilateral Trade

Probit OLS-HMR OLS-HMR I FE FE-I FE-II FE-III

(1) (2) (3) (4) (5) (6) (7)

Production j 0.027*** 0.408*** 0.302*** 0.395*** 0.382*** 0.381*** 0.381***

(0.001) (0.016) (0.025) (0.030) (0.032) (0.026) (0.032)

Remoteness 𝑗 0.003* -0.073* -0.120*** -0.045 -0.049 -0.043 -0.043

(0.002) (0.039) (0.039) (0.097) (0.096) (0.071) (0.096)

Exch. Rate 𝑗 0.008*** 0.122*** 0.082*** 0.041 0.043 0.045* 0.045

(0.001) (0.017) (0.018) (0.034) (0.034) (0.027) (0.034) Landlocked j -0.029*** 1.455*** 1.678*** 1.141*** 1.165*** 1.169*** 1.169***

(0.004) (0.116) (0.117) (0.244) (0.243) (0.183) (0.243)

Yield j 0.054*** 1.226*** 1.001*** 1.138*** 1.141*** 1.143*** 1.143*** (0.005) (0.127) (0.130) (0.270) (0.270) (0.196) (0.270)

Consumption i 0.009*** 0.317*** 0.267*** 0.302*** 0.300*** 0.301*** 0.301***

(0.001) (0.017) (0.018) (0.020) (0.020) (0.017) (0.020)

Remoteness 𝑖 -0.003** -0.110*** -0.086** -0.102** -0.101** -0.105*** -0.105***

(0.010) (0.038) (0.037) (0.040) (0.040) (0.038) (0.040)

Exch. Rate 𝑖 0.008*** -0.075*** -0.111*** -0.066*** -0.071*** -0.072*** -0.072***

(0.001) (0.014) (0.015) (0.014) (0.014) (0.014) (0.014)

Landlocked i 0.010* -0.623*** -0.573*** -0.589*** -0.604*** -0.607*** -0.607***

(0.006) (0.116) (0.116) (0.117) (0.117) (0.118) (0.117)

Other Goods 𝑖 0.028*** 0.162*** 0.060** 0.217*** 0.213*** 0.212*** 0.212***

(0.001) (0.023) (0.030) (0.027) (0.027) (0.024) (0.027)

Distance -0.063*** 0.058 0.370*** -0.303*** -0.287*** -0.289*** -0.289***

(0.002) (0.061) (0.075) (0.067) (0.067) (0.069) (0.067) Land Border 0.193*** 1.916*** 0.884*** 1.543*** 1.533*** 1.534*** 1.534***

(0.014) (0.128) (0.165) (0.141) (0.141) (0.132) (0.141)

Language 0.059*** 0.019 -0.344*** -0.014 -0.019 -0.015 -0.015 (0.006) (0.092) (0.099) (0.105) (0.105) (0.097) (0.105)

Colony 0.006 1.022*** 0.917*** 0.950*** 0.953*** 0.949*** 0.949***

(0.010) (0.160) (0.159) (0.143) (0.142) (0.164) (0.142) Same Country 0.032** -0.210 -0.360* -0.389* -0.381* -0.382* -0.382*

(0.015) (0.201) (0.201) (0.213) (0.213) (0.208) (0.213)

Land 0.008*** 0.211*** 0.179*** 0.182*** 0.186*** 0.187*** 0.187*** (0.001) (0.022) (0.023) (0.027) (0.027) (0.024) (0.027)

RTA 0.033*** -0.210* -0.327*** -0.401*** -0.379*** -0.390*** -0.390***

(0.005) (0.127) (0.126) (0.143) (0.142) (0.131) (0.142)

Price 5.223** 5.229** 5.246** 5.246**

(2.273) (2.313) (2.518) (2.317)

Mand. Label -0.029*** (0.023)

Tech. Gap -0.034*** -0.376*** -0.222*** -0.163** -0.164*** -0.164**

(0.003) (0.046) (0.050) (0.065) (0.060) (0.065) Demand Lag -0.021*** -0.316*** -0.249*** -0.159*** -0.160*** -0.160***

(0.003) (0.046) (0.049) (0.061) (0.057) (0.061)


1.164*** -0.022 -0.022

(0.261) (0.433) (0.438)


7.995*** -0.512 -0.512

(1.069) (1.737) (1.576)


2 -2.249*** 0.082 0.082

(0.439) (0.673) (0.599)


3 0.209*** -0.018 -0.018

(0.052) (0.073) (0.063)

Constant -4.689*** -13.107*** -16.403***

(0.384) (1.611) (1.857)

Observations 39,751 6,634 6,634 6,634 6,634 6,634 6,634

R2 0.425 0.438 0.253 0.254 0.255 0.255 Akaike Inf. Crit. 24,191.690

F Statistic

135.398***

(df = 36;

6597)

128.565***

(df = 40; 6593)

47.974***

(df = 39;

5522)

53.285***

(df = 35;

5526)

Note: *p<0.1; **p<0.05; ***p<0.01 With the exception of Model (5) all models have robust stand errors (arellano). Dummies for year fixed effects were used to all models and

are omitted.

117

As tests indicates the consistence of FE models we are going to set model (7) as

our benchmark – hereinafter FE (7). But in order to check coefficients stability, we are also

going to consider results from HMR-I (column 3) along with the Probit results (column 1) –

hereinafter HMR-I (3). HMR-I (3) is particularly interesting to breakdown the effects into

intensive and extensive margins, as well as the impacts of firm heterogeneity. Considering

models together do not lead to contradictory interpretations.

The Probit was estimated by pooling the data with fixed effects for years. Note

that it is not the procedure used to estimate ��𝒊𝒋𝒕 and ��𝒊𝒋𝒕

controls. The same procedure is used

in HMR (2008) when authors seek to demonstrate their results are not particular to a single

year.

Note that log-linear form allows us to interpret the effects as trade elasticity for

continuous variables, whereas the coefficients for technological variables should be

interpreted in a different way, since they are reported as integers. Also, the dummy controls

for trade costs, such as language, colony, same country, land and RTA, cannot have their

coefficients interpreted as trade elasticities.

Sample Selection and Firm Heterogeneity Biases

As expected, the controls for firm heterogeneity and selection bias play a role only

in the HMR-I (3). Firm heterogeneity is related to prohibitive fixed costs leading to the lack

of trade between country pairs. This latter effect is important to us since we are assuming that

adoption of GM technology by Argentina and the US along with non-adoption by Brazil from

1996-2005 may have caused changes in the extensive and intensive margin of soybean trade.

In other words, zero trade flows could occur because of strong rejection of GM technology

by some important markets – increased demand lags.

A comparison between models in columns (2) and (3) provides a picture about the

general impact the proposed controls have on the coefficients. The absence of controls will

cause an upward bias for some coefficients. Overall, variables controlling for sizes, country

𝑖′𝑠 remoteness and bilateral trade costs returned upward biased coefficients, whereas country

𝑗’s remoteness is downward. Interestingly, coefficients with higher importance for extensive

margin of trade not only return downward biases but also became significant after the

controls introduction (see coefficients for Language66

, Same Country, and RTA in columns

(1) to (3)).

66

Language will become statistically insignificant in the benchmark model (7).

118

Given the high Variation Inflation Factor (VIF), the value of the coefficients of

these controls cannot be interpreted as consistent in terms of trade elasticities impacts.

However, they are consistent with results from HMR (2008) showing that firm heterogeneity

overcomes the effects of selection bias. Also, the coefficients are positive, meaning that

countries with higher proportion of exporting firms and unobserved characteristic of

countries determining positive trade flows will also impact positively in the volume of trade.

Consequently, the measures of the effects of trade frictions in model (2) can be

considered upward biased, as they confound the true effect of trade frictions with their

indirect effect on the proportion of exporting firms (Helpman et al., 2008). Note that the most

impacted coefficients are those from variables representing trade frictions with predominant

effects in the margins of soybean trade (Land Border, Same Country, Distance, RTA,

Language), coherent with results reported by HMR (2008).

Note that none of the controls are significant in FE (7). Given the test for sample

selection bias proposed by Wooldridge (1995) we can still consider the estimates consistent.

Comparing model HMR-I (3) with FE (7) we can see that some coefficients will return

greater coefficients and others smaller ones. However, remoteness and exchange rate of

exporters and common language will be statistically insignificant in (7) but significant in (3).

Impacts of country-level variables

We have country-level variables for countries 𝑖 and 𝑗 in the model. Country 𝑗′𝑠

variables controlling for country specificities – technological variables excluded – are

Production 𝑗, Exch. Rate 𝑗, Landlocked 𝑗, Yield 𝑗 and Remoteness 𝑗. Country 𝑖′𝑠 variables are

Consumption 𝑖, Exch. Rate 𝑖, Landlocked 𝑖, Other Goods 𝑖 and Remoteness 𝑖.

From the supply side, based on theory and studies on aggregated data, we can

expect that soybean trade will increase along with higher outcomes of soybean and average

yields of exports. Conversely, trade volumes tend to be decreased by higher exchange rates,

lack of maritime exits and remoteness.

Production and yield returned the expected impact on trade flows as we can see in

columns (3) and (7). Given the production elasticity of trade reported in FE (3), for each

increase of say, 10%, in production, trade will increase in 3.8%. Trade elasticity of

production is not higher because of internal consumption of soybean in major producing

countries (e.g. China) and some of the larger exporters producing small quantities of soybean

(eg. international hubs). Note that production also impacts on the probability of exporting,

119

but the relevant impact will be at the intensive margin of trade. Exporter with higher average

yields, which can be seen as a result of technologies available for growers at the country

level, also increases the volume of trade significantly. The elasticity in this case, will make

with an increase in average productivity level of 10 % make exports increase by 11.4%. The

parameter in 𝑇𝑖 in Eaton and Kortum (2002) model, representing the general level of

technology in a country, can be a theoretical explanation for the high yield elasticity of

exports. If 𝑇𝑗 is high it will increase the chances of having more productive firms in every

industry.

Other variables for countries 𝑗 , however, need further considerations.

Contradictorily, landlocked exporters tend to export considerably higher volumes when

compared to countries with maritime access67

. However, this effect has a negative impact on

the extensive margin of trade, indicating that landlocked country has on average less trade

connections than others. A look at the data will make clear that this seeming odd result is

valid particularly for the soybean case, given intra-bloc trade in the European Union and

significant shares of world exports of landlocked countries like Paraguay, Switzerland,

Bolivia, Zambia, Austria and Hungary drive this coefficient up. Noteworthy, many of these

countries are large importers of soybeans from other producing countries.

Remoteness and variations of the exchange rate in exporting countries are not

statistically significant in FE(7), and have small coefficient values in HMR-I (3) models. This

result can be explained by the concentrated international supply of soybeans – primarily in

the Americas. In other words, as growers in Brazil, USA and Argentina produce almost the

totality of global traded soybeans, the remoteness of exporting countries tends to be

insignificant because of geographical issues – remember that remoteness index is build based

on distances and GDPs.

Variations in exchange rate in exporting countries are expected to impact trade in

two different ways in the case of soybeans. On the one hand it makes soybean export prices

in countries with devaluated currency cheaper. On the other hand, it makes the inputs used in

production more expensive. The net gains of devaluation will depend on this two-fold impact

on relative prices. Some studies show that currency devaluations in South America displace

exports of US in international markets (Andino & Koo, 2005). However, our results show

67

Landlocked countries tend to trade less (extensive and intensive margins) due to their remoteness leading to

increased trade costs – including transaction and transportation costs. According to World Bank they trade on

average 60% less when compared to countries with maritime boundaries.

120

that impacts on the likelihood of exports is possible very low and positive (0.08%), meaning

that currency valuation in country 𝑗 increases the likelihood of positive trade flows,

marginally. Regarding the intensive margin of trade, HMR-I (3) returned an elasticity of

0.082 for this variable. It can be a result of the predominance of increases of imported inputs

relative prices as well as an indication of the weakness of this variable to explain bilateral

trade of soybeans as the international prices are always given in USD dollars regardless

variation of local currencies.

From the demand side, theory predicts that bilateral trade tends to increase along

with higher demand levels, valuated exchange rates, absence or high imperfectness of

substitute goods, easy access to the sea and low levels of remoteness.

Results show that bilateral trade of soybean actually increases along with total

consumption (or size). One increase of 10% in consumption, will lead to an increase of 3% of

imports. Similar to suppliers’ size control, i.e. production, this number isn’t higher because of

larger consuming markets also producing certain amounts of soybean, such as China. The

effect of substitutes goods (Other Goods i) is interesting, since it shows that soybean has no a

close substitute for meal production in international markets – that is the major end product.

For each increase of 10% in imports of other meals feedstock, such as meat and bone meals,

fishmeal, among others, imports of soybean increase by 2.12% on average. HMR-I (3)

returns coefficients with the same sign and significance but with smaller values.

Importing countries with high degrees of remoteness, devaluated currencies and

landlocked will actually import less volumes of soybean. Landlocked countries import on

average 60% less when compared to countries with maritime access. Currency devaluations

and increases in the remoteness index have smaller impacts, but considering the magnitude of

small percentages applied on huge amounts of soybean traded, the economic significance can

be considerable to some cases. HMR-I (3) has similar impacts in terms of sign, magnitude

and significance of the coefficients. Noteworthy, landlocked country tends to have more

positive trade inflows when compared to others, accordingly to results of model (1).

In sum, we can say size and state of technology in producing countries are the key

drivers of exports volume. On the demand side, size, remoteness, valuated currencies and

access to maritime routes are the most important variables impacting trade inflows. In

addition, the lack of international close substitutes for soybean will also play a role in

determining demand levels.

121

Impact of trade-level variables

At the trade-level we have considered the usual Distance, Land Border, Language,

Colony, RTA and Same Country variables. In addition, seeking to control to factor

endowments and their impacts on input prices, we also considered the difference in land

endowment between trade partners through the variable Land.

First, it is important to note that common language is not significant according to

our benchmark model FE (7). On the other hand, this variable is highly significant in the

HMR-I (3) model, in which countries sharing a common language do trade 41.05% less than

others, but they have 5.9% more chances of trading among them. As in our benchmark model

this variable has no significance, we cannot say much about the effect of language on

international trade of soybean. Thinking of Brazil, Argentina and the United States, it is

difficult to think about an unambiguous effect, as countries speaking Portuguese are few and

with reduced economic importance. On the other hand, the number of countries speaking

English and Spanish will be higher and their economic importance too. This and other

qualifications peculiar to our case, such as the importance of China as a consumer market,

and absence of larger producers speaking Mandarin, will make the analysis of language effect

ambiguous.

A colonial tie, on the other hand, has a positive and meaningful impact on the

trade of soybean. Countries with colonial ties trade approx. 158% more when compared to

others. This can be seen as a shadow of former colonial economic relations, in which the

colonies provided primary products to the industrialized colonists. Interestingly, this effect is

seen exclusively at the intensive margin of trade.

On the other hand, countries sharing a same colonist for long periods will trade

reduced volumes of soybean (46.5% approx.). It can be a result of agricultural specialization

of colonies, making the trade of agricultural products smaller between two agricultural-based

countries, but intense between industrialized and agricultural countries. Impact on extensive

margins, however, will be positive (see column (1). As expected, RTAs will have the very

same type of impact on trade of soybeans. They do connect countries (extensive margin) but

they don’t increase volumes exported – remember that RTA does not mean a bilateral trade

agreement for soybean trade.

Distance, as a proxy for variable costs of trade based on the Samuelson’s “iceberg

costs” is expected to decrease trade volumes. According to our estimates, a 10% increase in

distance will lead to 28.9% decrease in the volume of trade (see column (7)). That is a

122

consistent result, as international freights tend to be a significant component of trade costs.

The access to the Pacific Ocean by the U.S. will make “iceberg costs” of exporting to China

smaller when compared to costs of Brazilian exports to the same market. Graphical analysis

shows that the absence of market rejection made of the U.S. the major supplier of soybean in

China. Additionally, results from HMR-I (3) suggest that distance has a negative impact on

the probability of trade, but a positive effect on volumes. This may be a simple result of gains

of scale. As distant markets will imply in higher costs of trade, higher volumes of trade can

represent some decreases on total costs.

Related to distance effects, the estimates show that countries sharing a land border

will trade significantly more than others. It is an additional indication that distance is a very

important variable to explain the soybean trade. Along with the fact handlers usually operates

closer to major producing or importing markets, we can say that this effects are merely

confirming that transportation costs have a large share in overall costs of soybean production

and distribution.

Finally, as colonial ties has also indicated, we can expect that endowment of land

will be important to partially explain the formation and direction of trade flows. This variable

is a proxy for differences in input costs 𝑐𝑗 and 𝑐𝑖 that will lead to certain degree of industry

specialization, ceteris paribus. It is a source of trade mostly thought from the HO models, but

also treatable in the framework developed by Eaton and Kortum (2002), if we assume that

relative endowments of land will impact the production costs of agricultural goods. The

difference in land endowment will impact mostly on the extensive margin of trade. If this

difference increases by 10%, trade will increase by approximately 1.87%. Note that the

coefficients are very similar across simulations.

Not related to bilateral trade costs but also important enough, the coefficient for

soybean prices in international markets (Price) is statically and economically significant. If

internal annual prices are increased by a very small amount the value of exports tend to

increase considerably (elasticity =5.2). It reinforces the thesis of lack of close substitutes for

soybeans in international markets, showing that imports are inelastic to certain ranges of

price variation.

In sum, the trade between countries located geographically closer, sharing a

border, with colonial ties and differences in factors endowment (land) is more frequent and

higher in volume when compared to others. Remarkably, countries under regional trade

agreements and considered as the same country trade more often but in lower volumes.

123

Impacts of technological variables

Finally, we can discuss the effects of the technological variables, which we are

assuming to have significant impacts on trade. We have three technology-related variables in

our model - Mand. Labeling, Tech. Gap, Demand Lag.

Precisely, differences in mandatory labeling regimes are more a proxy for

regulation asymmetry than for technological change itself, impacting mostly on the extensive

margin of trade. This assumption is based on several studies, which points out mandatory

labeling as one of the most detrimental factors impacting trade of GMOs(Carter & Gruère,

2006; Foster, 2010; Regimes & Giannakas, 2013). Indeed, labels enable consumers to tell

apart food originating from conventional and GM grains. In other words, without the label

consumers could not identify or decide about technologies of production.

There are evidences that countries with mandatory labeling imported soybeans

primary from Brazil, at least until the end of 2012 (P. R. S. Oliveira et al., 2013). Authors

estimated that when mandatory labeling regimes are increased by one year, imports from

Brazil increase by 4.5%.

The Demand Lag variable, in turn, was defined based on the general idea of

imitation lag developed by Posner (1961). Although many empirical works have been strictly

analyzing effects of divergences in regulation or acceptance of the technology in consuming

markets and impacts on producing ones, they rarely contribute to theoretical advances in this

field. In addition, the problem of the demand lag as a variable impacting on trade and

depending on the pace of technology adoption and acceptance has been ignored even among

evolutionary economists – perhaps because of the absence of a case in which this effect were

remarkable enough.

In this study we assume that the lag will enlarge market shares of countries

employing conventional (old) technology in those markets averse to the technology. The

duration and the proportion of the lag will determine the intensity of backward effects of

technology. As major players have passed regulations on risk assessment of new varieties

mainly based on the case-by-case approach, asymmetric and asynchronous approval is a good

proxy for the actual demand lag. That is especially true when we assume that policymakers

are voting new approvals accordingly to public opinion in their countries.

Moreover, approval asymmetry will be drastic as much as adverse countries have

good monitoring capacity to control entry of agricultural products into their territories. As we

have discussed in Chapter II, policymakers had/have a central role in keeping commercial

124

risks at a manageable level. Thus, if importing countries can keep effective bans on

unapproved varieties, an increase of uncovered approvals in exporting countries can

potentially affect or even cripple some trade flows.

The technological gap, on the other hand, has received considerable attention

among the evolutionary economists, but it is usually thought from the perspective of

technological differences between a country-pair. In this work we explicitly take a different

interpretation of the gap. We believe that a relative position of exporting countries in terms of

technology innovation and adoption when compared to general state of technology

worldwide is key to determine trade volumes.

We believe that international competition and innovation rate will create an

international technology frontier, which is constantly moving forward given the pace and

nature of innovation in the economic system. The distance of a country 𝑗, in terms of

technology employed in production, from this technological frontier is the technological gap.

As we consider the relative innovation and adoption rate around the world, this control is

somehow similar to the seminal concept of remoteness put by J. E. Anderson & Wincoop

(2003).

As technology will impact on the overall production costs and not all the countries

developed technology “hatred”, there will be gains of adoption that should be considered in

the models in order to think an overall technological impact. From that, trade should increase

as countries adopted newer and better technologies of production, in markets without the

drawback effects of technology rejection. Conversely, an increase of the technological gap

will decrease trade both in extensive and intensive margins, in those markets without high

levels of technology “hatred”. That is because in “normal” markets, “normal” technological

effects are expected to stand out – i.e. the positive impact of technology on overall production

costs will be a source of comparative advantage.

Considering the relative importance of technology gap and demand lag, we used

the costs related to mandatory labeling as the excludable variable required for estimating the

Probit at the first stage. The existence of a mandatory labeling regime in destination and

absence in sourcing countries decreases the probability of trade almost by 3%.

Beforehand, we can see that the coefficients of our other two technological

variables behavior steadily in all models. FE (7) points out that opportunity costs of not

adopting available technology are not only statistically but also economically significant. For

each variety that a producing country doesn’t grant approval for any reason trade will

decrease by 16.4% on average. Looking at the coefficients of HMR-I (3) we could say that

125

impacts could be even larger – 3.4% in the probability of trade and 22.2% in the volume of

trade.

Interestingly, the demand lag will have a very similar impact on trade volumes. A

difference in approved varieties between exporting and importing countries will decrease

trade on average by 16%. Again HMR-I (3) returns greater coefficients - 2.1% decreases in

the trade probability and 24.9% in trade volume.

The similarity of the impacts of both variables is an indication of the existence of

certain market rationality in approving new varieties, which balances opportunity costs of not

approving new and better varieties, and the commercial risks involved in adopting too many

varieties in the context of technology “hatred” in important markets. From this perspective,

the pace of adoption in Brazil and Argentina ended up balancing the opportunity costs of

non-adoption and the commercial risks of adoption in a way that the impacts on international

trade were not catastrophic.

The high coefficients for the technological variables are merely reflecting the

replacement effect we have seen in graphical analysis in Chapter II. As the EU is a giant

consuming market, the average impact of uncovered approvals is expected to have huge

impacts on trade pattern. These results indicate that, indeed, regulation and aversion in

developed countries can impact production and well fare in producing countries. Vigani et al

(2012) found an impact of 33% in trade flows to an increase of regulatory dissimilarity. This

can be an indication that not considering the effects of the opportunity costs or the

technological gap can bias upward the coefficient for the demand lag.

It is important to highlight that the first seeds developed in second half of the

1990s didn’t have great impacts on yield or overall cost reductions in tropical areas. In this

case, opportunity costs were kept relatively low in Brazil during this period. As the demand

lag was more remarkable during the first period, commercial risks were more relevant back

there. Thus, the effect of demand lag and small advantage of new technology were central to

hold Brazilian competitiveness during the national official ban on cultivation of GMOs.

Notwithstanding, the technological advance and development of new and better

varieties increased considerably the opportunity costs of non-adoption. From a more

evolutionary perspective technological advantage can reach up an irreversible degree for

later-movers. In other words, in the absence of non-technological advantages if a proved

better technology is available and adopted by other producing countries, the late-mover can

fail to catch-up, loosing significant shares of the market or even leaving the marketplace.

126

In sum, at the same time countries with more similar pace of approvals – smaller

demand-lags – tend to trade more (extensive and intensive margins) there are opportunity

costs of not approving better varieties that can decrease trade. In other others, we are saying

that countries faced a kind of technological trade-off in deciding about adoption under

technology rejection.

This results are paramount to police design, since it brings on scene

counterintuitive reasoning. First, adoption of new technologies will not be always desirable at

any cost. Second, if rejection is not even across different markets producing countries should

care not only about the commercial risks coming from demand lag but also those coming

from a technological gap, which can equally be detrimental to trade.

127

IV. CONCLUSIONS

The central aim of this dissertation was to study the relationship of technology

and trade under high levels of “hatred” toward a “new” product resulting from innovation.

The case of GM-soy is a good experiment, since both the adoption rate in producing

countries and the acceptance of the “new” product in international markets were asymmetric

and asynchronous across time and countries. Moreover, the production and consumption of

soybeans are concentrated at country level, mitigating the problem of the lack of trade data

disaggregated into conventional and GM grains.

To do so, we carried out a systematic observation of the soybean industry,

collected evidences of technological effects in trade of soybeans, put together theoretical

developments relating technology and trade, and finally, we have estimated a gravity

equation to assess our central hypotheses of a dual-technology effect on trade.

The case of GMOs starts with a true revolution in the seed industry, which will be

questioned by consumers – individual and industry – across different countries, in terms of

the new technology benefits and risks. This asymmetry of perceptions and interests on the

two edges of the supply chain will create tensions and risks to be managed by the middlemen.

As very often grains are produced in a country and processed in another, trade conflicts will

arise when producing countries adopt technologies that will be rejected by consumers in

importing countries.

In this scenario, policymakers will have a central role in keeping risks at a

manageable level to avoid collapse of trade flows, even if they are considering also other

production benefits related to the technology. Interesting, the vertical integration of the

supply chain, via ownership of enterprises operating at all logistic levels, the partnerships

between seed companies and grain handlers and high levels of industry concentration were

appropriate and necessary to avoid commercial collapses. That is because without

coordination the industry would not be able to keep full capacity in a system strongly

dependent on high economies of scale.

From the end-consumers’ perspective, the growths of per capita income

worldwide and the emergence of the certification industry will represent a structural change

in the consumption behavior. For the first time in the history of agriculture, it was possible to

see a technology intended to impact primary production process turns out to be of great

concern to society as whole. Consumers from high-income countries are, somehow,

developing a “taste for technology” that can impact significantly the agricultural production,

128

especially in the developing countries. In other words, they don’t demand only food, but they

demand food produced through one or another technology – e.g. organics, legal and fair

employment contracts, etc. Noteworthy, the industrial consumer, in turn, is also demanding

varieties that deliver higher values in terms of intrinsic characteristics – standard sizes, high

oil content, etc.

In the international arena, Brazil, Argentina and the United States at the supply

side, and China and the EU members at the demand side, are the protagonists of the history of

GM-soybean trade. Brazil adopted the technology ten years after growers in Argentina and

the U.S. planted genetically modified seeds for the first time. The EU members, the world

second largest destination of soybeans, developed very adverse regulatory frameworks for

approving and labeling GMOs. But, China has not raised significant bans against the

consumption of GMOs, being an important player to mitigate global commercial risks.

Graphical analysis and findings of other applied and theoretical studies can

illustrate the existence of significant technological effects in trade. In addition, they suggest

that impacts are based on different perception of benefits that consumers have in different

countries. However, many of these studies contribute very few to understanding and

investigating a general relation of “demand lag” and trade, although they usually focus on

estimating the effect of regulatory frictions on bilateral trade. The lack of theoretical interest

by the majority of the studies is a drawback to advance in many important unsolved questions

about the interaction of technology change and trade.

Seeking to shed some lights on the main stylized facts observed on applied studies

in the field, we have revisited three models of trade considering technological effects. The

natural starting point was the Ricardo model of trade as described by Eaton and Kortum

(2012). A second interesting contribution came from Helpman et al (2008). Lastly,

developments of evolutionary economists, departing from Posner (1961), contributed

considerably to our understanding of the case studied and specification of our empirical

model.

Considering our purposes, a clear inadequacy of the neoclassical models is the

lack of a treatment for different tastes across countries. In this respect, we can say that theory

need to advance in this matter to better approach cases in which technological or other

backward effects based on tastes matter.

Assuming that technology differences creates comparative advantages, the theory

of firm heterogeneity and its analytical developments, such as the Two-Stage estimation of

the gravity equation, and the central concepts of demand lag and technological gap, we

129

structured a model in which bilateral trade is a function of sizes, costs of trade and

technological variables.

The inclusion of a type of opportunity costs of no-adoption, namely the

technology-gap, is our main contribution in the matter of correct specification of empirical

models for the case of technology backward effects. Without this control, the analysis is

partial, since rejection is not verified in everywhere. Thus, in part of the marketplace, the

common effects of innovation leading to efficiency gains and, thus, comparative advantage

will be predominant.

We found that both the technological gap and the demand lag had significant

impacts on GM-soybean trade. Impacts can be verified in both the extensive and intensive

margin of trade. Interestingly, both effects are very similar showing that the trade-off

between the opportunity costs of non-approval and the increased commercial risks of

approving varieties not approved in destination markets were in someway balanced. For each

difference in approval considering the technological frontier – technological gap – countries

faced a volume of trade 16.4% lower. For each difference in approval considering varieties

approved in destination markets – demand lag – a country suffered a cut down of 16% in the

volume of trade.

We believe results can contribute to better designs of technological and trade

policies, since they provide a broader perspective of technological effects on trade. We also

believe that other similar cases marked by technological frictions can emerge at anytime, as

the “preference for technology” seems to be a structural change in consumption patterns.

The increased concern about the production means and their relation with ethical,

environmental, social and economic factors will likely increase trade conflicts if multilateral

bodies have reduced coordination power – as seen in the case of modern biotechnology and

trade.

Clearly, impacts on agriculture tend to be more noteworthy, given the especial

features faced by this sector, as intense competition in international markets and increased

awareness of consumers about food safety.

In addition, the lack of marked effects in other industries does not mean that

technological gap and demand lag is not operating in trade patterns in reduced proportions –

as Posner have already being pointing since the 1960s.

In conclusion, future research on the relationship of technological change and

trade is needed to advance in theoretical developments, especially to relax some of the strong

assumptions on consumers’ behavior. Empirical studies of other products and industries is

130

also required to investigate if this effect is specific to agricultural goods subjected to

commercialization approval, or the technological effect can be thought from a more general

perspective.

131

REFERENCES

Anderson, J. E., & Wincoop, E. Van. (2003). Gravity with Gravitas : A Solution to the

Border Puzzle. The American Economic Review, 93(1), 170–192.

Anderson, K. (2010). Economic impacts of policies affecting crop biotechnology and

trade. New Biotechnology, 27(5), 558–64. doi:10.1016/j.nbt.2010.05.012

Anderson, K., & Jackson, L. A. (2004). Standards , trade and protection : the case of GMOs

(No. Discussion Paper: 0411). Adelaide.

Andino, J., & Koo, W. W. (2005). The Impact of Brazil and Argentina ’ s Currency

Devaluation on U . S . Soybean Trade. Agribusiness & Applied Economics, (574), 18.

Arellano, M. (1987). Computing robust standard errors for within group estimators.

Oxford Bulletin of Economics and Statistics. doi:10.1111/j.1468-

0084.1987.mp49004006.x

Asian Food Information Center. (2008). Executive summary Food Biotechnology :

Consumer perceptions of food biotechnology in Asia.

Aw, B. Y., Roberts, M. J., & Xu, D. Y. (2008). R&D investments, exporting, and the

evolution of firm productivity. American Economic Review, 98(2), 451–456.

doi:10.1257/aer.98.2.451

Baltagi, B. H., Egger, P., & Pfaffermayr, M. (2014). Panel Data Gravity Models of

International Trade (No. 4616). Working Papers CESifo. Munich.

Bärwald Bohm, G. M., Rombaldi, C. V., Genovese, M. I., Castilhos, D., Rodrigues Alves, B. J.,

& Gouvêa Rumjanek, N. (2014). Glyphosate effects on yield, nitrogen fixation, and

seed quality in glyphosate-resistant soybean. Crop Science, 54(4), 1737–1743.

doi:10.2135/cropsci2013.07.0470

Bernard, A. B., Eaton, J., Jensen, J. B., & Kortum, S. (2003). Plants and Productivity Trade

132

in International. The American Economic Review, 93(4), 1268–1290.

doi:10.2307/3132288

BIO. (2015). Biotechnology Industry Organization: Agricultural Product Database.

Washington, D.C.

Brookes, G., & Barfoot, P. (2014). The global income and production effects 1996 – 2012

Economic impact of GM crops. GM Crops, 5698(March),

http://dx.doi.org/10.4161/gmcr.28098. doi:10.4161/gmcr.28098

Brossard, D., Shanahan, J., & Nesbitt, R. C. (Eds.). (2007). The Public. The Media &

Agricultural Biotechnology. Oxon: CABI.

Bullock, D. S., & Desquilbet, M. (2002). The economics of non-GMO segregation and

identity preservation. Food Policy, 27(1), 81–99. doi:10.1016/S0306-

9192(02)00004-0

Burnquist, H. L., Shutes, K., Rau, M.-L., Souza, M. N. F., & Faria, R. N. (2011).

Heterogeneity Index of Trade and Actual Heterogeneity Index – the case of

maximum residue levels ( MRLs ) for pesticides. In Agricultural & Applied

Economics Association’s 2011 AAEA & NAREA Joint Annual Meeting (p. 24).

Pittsburgh.

Capalbo, D. M. F., Arantes, O. M. N., Maia, A. G., Borges, I. C., & Silveira, J. M. F. J. da.

(2015). A Study of Stakeholder Views to Shape a Communication Strategy for GMO

in Brazil. Frontiers in Bioengineering and Biotechnology, 3(November), 1–10.

doi:10.3389/fbioe.2015.00179

Carter, C., & Gruère, G. (2006). International approval and labeling regulations of

genetically modified food in major trading countries. In R. E. ( U. of M. Just, D.

Alston, Juliian M. (University of California, & B. Zilberman, David (University of

California (Eds.), Regulating agricultural biotechnology: economics and policy (pp.

459–480). New York, NY: Springer New York.

133

Carvalho, F. (1995). O comportamento das exportações brasileiras ea dinâmica do

complexo agroindustrial. 1995. São Paulo University.

CERA. (2015). GM Crop Database. Center for Environmental Risk Assessment (CERA).

Washington D.C.: ILSI Research Foundation.

Charbonneau, K. B. (2014). Multiple Fixed Effects in Binary Response Panel Data Models.

Ottawa.

Chavas, J. P., Shi, G., & Lauer, J. (2014). The effects of GM Technology on maize yield.

Crop Science, 54(4), 1331–1335. doi:10.2135/cropsci2013.10.0709

Choi, E. K. (2010). International trade in genetically modified products. International

Review of Economics & Finance, 19(3), 383–391. doi:10.1016/j.iref.2009.04.002

Curtis, K. R., & Mccluskey, J. J. (2004). Consumer Acceptance of Genetically Modified

Food Products in the Developing World. AgBioForum, 7(2002), 70–75.

Dalgin, M., Mitra, D., & Trindade, V. (2004). Inequality, Non-Homothetic Preferences, and

Trade: A Gravity Approach (No. 10800). NBER WORKING PAPER SERIES. Cambridge.

doi:10.1017/CBO9781107415324.004

Deardorff, A. V. (1980). The General Validity of the Law of Comparative Advantage.

Journal of Political Economy, 88(5), 941–957. doi:10.1086/260915

Desquilbet, M., & Bullock, D. S. (2009a). Who Pays the Costs of Non-GMO Segregation

and Identity Preservation? American Journal of Agricultural Economics, 91(3), 656–

672. doi:10.1111/j.1467-8276.2009.01262.x

Desquilbet, M., & Bullock, D. S. (2009b). Who Pays the Costs of Non-GMO Segregation

and Identity Preservation? American Journal of Agricultural Economics, 91(3), 656–

672. doi:10.1111/j.1467-8276.2009.01262.x

Diamante, A., & Izquierdo, J. (2004). Manejo y gestión de la biotecnología agrícola

134

apropiada para pequeños productores: estudio de caso Argentina. Buenos Aires.

Retrieved from http://www.argenbio.org/adc/uploads/pdf/manejo_y_gestion.doc

Disdie, A.-C., & Fontagné, L. (2009). Trade Impact of European Measures on GMOs

Condemned by the WTO Panel. Review of World Eocnomics, 146(3), 495–514.

doi:101007/s1029001401883

Disdier, A.-C., & Fontagné, L. (2010). Trade impact of European measures on GMOs

condemned by the WTO panel. Review of World Economics, 146(3), 495–514.

doi:10.1007/s10290-010-0057-7

Dornbusch, R., Fischer, S., & Samuelson, P. a. (1977). Comparative Advantage, Trade, and

Payments in a Ricardian Model with a Continuum of Goods. The American Economic

Review, 67(5), 823–839. doi:10.1126/science.151.3712.867-a

Dosi, G. (1982). Technological paradigms and technological trajectories: a suggested

interpretation of the determinants and directions of technical change. Research

Policy.

Dosi, G., Grazzi, M., & Moschella, D. (2015). Technology and costs in international

competitiveness: From countries and sectors to firms. Research Policy, 44(10),

1795–1814. doi:10.1016/j.respol.2015.05.012

Dosi, G., & Soete, L. (1983). Technology Gaps And Cost‐Based Adjustment: Some

Explorations On The Determinants Of International Competitiveness.

Metroeconomica, xxxv(3), 197–222.

Dosi, G., & Soete, L. (1988). Technical change and international trade. In L. Soete (Ed.),

The Economics of Technical Change and International Trade (pp. 401–431).

Eaton, J., & Kortum, S. (2002). Technology, geography, and trade. Econometrica, 70(5),

1741–1779.

Eaton, J., & Kortum, S. (2002). Technology, Geography, and Trade. Econometrica, 70(5),

135

1741–1779. doi:10.1111/1468-0262.00352

FAO. (2015). FAOSTAT - FOOD AND AGRICULTURE ORGANIZATION OF THE UNITED

NATIONS. Rome: FAO-Statistics division. Retrieved from

http://faostat.fao.org/site/537/default.aspx

FAOSTAT. (2015). FOOD AND AGRICULTURE ORGANIZATION OF THE UNITED

NATIONS STATISTICS DIVISION.

Faria, R. N. de, & Wieck, C. (2015). Empirical evidence on the trade impact of

asynchronous regulatory approval of new GMO events. Food Policy, 53, 22–32.

doi:10.1016/j.foodpol.2015.03.005

FAS-USDA. (2015). EU-28 Agricultural Biotechnology Annual. Washington D.C.

Feinberg, S. (2000). Do world product mandates really matter? Journal of International

Business Studies, 31(1), 155–167.

Fieler, A. C. (2011). Nonhomotheticity and Bilateral Trade: Evidence and a Quantitative

Explanation. Econometrica, 79(4), 1069–1101. doi:10.3982/ECTA8346

Finger, R., & Hartmann, M. (2009). Adoption Patterns of Herbicide-Tolerant Soybeans in

Argentina. AgBioForum, 12, 404–411.

Fisher, E. O. N. (2011). Heckscher-Ohlin theory when countries have different

technologies. International Review of Economics and Finance, 20(2), 202–210.

doi:10.1016/j.iref.2010.11.009

Foster, M. (2010). Evidence of price premiums for non-GM grains in world markets.

Australian Agricultural and Resource Economics Society. ABARE, Australia.

Galvão, A. (Celeres). (2014). Informativo Biotecnologia. Uberlândia.

Gaskell, G., Stares, S., Allansdottir, A., & Allum, N. (2010). Europeans and Biotechnology in

2010 Winds of change? Luxembourg.

136

Gómez-Herrera, E. (2012). Comparing alternative methods to estimate gravity models

of bilateral trade. Empirical Economics, 44(3), 1087–1111. doi:10.1007/s00181-

012-0576-2

Greenpeace. (2004). Public’s perception of Genetically Engineered Food: Summary of

Report. Retrieved from htpp://www.greenpeace.org.hk

Griffith, D., Hu, M., & Ryans, J. (2000). Process standardization across intra-and inter-

cultural relationships. Journal of International Business, 31(2), 303–324.

Grimsrud, K. (2004). Consumer attitudes to genetically modified food in Norway.

American Agricultural Economics Association. Retrieved from

http://onlinelibrary.wiley.com/doi/10.1111/j.1477-9552.2004.tb00080.x/full

Gruere, G. P. (2011). Asynchronous approvals of GM products and the Codex annex:

what low level presence policy for Vietnam? Discussion Paper, (October), 48.

Guillaume, G., & Zignago, S. (2010). BACI: International Trade Database at the Product-

Level. The 1994-2007 Version (No. No. 2010-23). Paris. Retrieved from

http://www.cepii.fr/CEPII/fr/publications/wp/abstract.asp?NoDoc=2726

Hager, R. J., & Dubinyuk, Y. (2014). Ukraine: Agricultural Biotechnology Annual Report.

Washington D.C.

Hallak, J. C. (2006). Product quality and the direction of trade. Journal of International

Economics, 68(1), 238–265. doi:10.1016/j.jinteco.2005.04.001

Head, K., & Mayer, T. (2013). Gravity Equations: Workhorse, Toolkit, and Cookbook.

Handbook of International Economics, 63. doi:10.1016/B978-0-444-54314-

1.00003-3

Heckman, J. (1979). Sample selection bias as a specification error. Econometrica: Journal

of the Econometric Society, 47(1), 153–161. Retrieved from

http://www.jstor.org/stable/1912352

137

Helpman, E., Melitz, M., & Rubestein, Y. (2008). Estimating trade flows: Trading partners

and trading volumes. The Quarterly Journal of Economics, CXXIII(May), 441–487.

HighQuest, & Soytech. (2011). How the Global Oilseed and Grain Trade Works. U.S.

Soybean Export Council. Danvers MA.

Hopenhayn, H. A. (1992). Entry, Exit, and firm Dynamics in Long Run Equilibrium.

Econometrica, 60(5), pp. 1127–1150. doi:10.2307/2951541

Howard, P. H. (2009). Visualizing Consolidation in the Global Seed Industry: 1996–2008.

Sustainability, 1(4), 1266–1287. doi:10.3390/su1041266

Huang, J., Qiu, H., Bai, J., & Pray, C. (2006). Awareness, acceptance of and willingness to

buy genetically modified foods in Urban China. Appetite, 46(2), 144–51.

doi:10.1016/j.appet.2005.11.005

Huang, J., Yang, J., & Yang, W. (2012). Trade and Economic Implications of Low Level

Presence and Asynchronous Authorizations of Agricultural Biotechnology

Varieties: A Case Study in China. In 28th International Conference of Agricultural

Economists: The Global Bioeconomy (pp. 1–37). Foz do Iguaçu- Brazil.

Hunter, L. (1991). The contribution of nonhomothetic preferences to trade. Journal of

International Economics, 30(3-4), 345–358. doi:10.1016/0022-1996(91)90027-4

III, R. H., & Kashlak, R. (1999). National influences on multinational corporation control

system selection. MIR: Management International Review, 39(2), 167–189.

Isaac, G. E., Perdikis, N., & Kerr, W. a. (2004). Cracking export markets with genetically

modified crops: What is the entry mode strategy? International Marketing Review,

21(4/5), 536–548. doi:10.1108/02651330410547180

Josling, T., Orden, D., & Roberts, D. (2004). Food Regulation and Trade: Toward a Safe

and Open Global Food System. Peterson Institute.

138

Kalaitzandonakes, N., Kaufman, J., & Miller, D. (2014). Potential economic impacts of

zero thresholds for unapproved GMOs: The EU case. Food Policy, 45, 146–157.

doi:10.1016/j.foodpol.2013.06.013

Kopicki, A. (N. Y. T. (2013, July 27). Strong Support for Labeling Modified Foods. N.Y.

Times. NYC.

Krugman, P. R. (1979). A Model of Innovation , Technology Transfer , and the World

Distribution of Income. Journal of Political Economy, 87(2), 253–266.

Krugman, P. R. (1985). Increasing returns and the theory of international trade.

Kugler, M., & Verhoogen, E. (2008). The Quality-Complementarity Hypothesis: Theory and

Evidence from Colombia (No. 14418). Cambridge.

Lagos, E. J., & Jie, M. (2013). Agricultural Biotechnology Annual": China - Peoples Republic

of. Whashington, DC.

Lence, S. H., & Hayes, D. J. (2001). Response to an Asymmetric Demand for Attributes :

An Application to the Market for Genetically Modified Crops, (92).

Linders, G.-J. M., & De Groot, H. L. F. (2006). Estimation of the Gravity Equation in the

Presence of Zero Flows. Tinbergen Institute Discussion Paper. Amsterdam.

doi:10.2139/ssrn.924160

LOC. (2014). Restrictions on Genetically Modified Organisms (Vol. 6462). Washington D.C.

Madhok, A., & Osegowitsch, T. (2000). The international biotechnology industry: a

dynamic capabilities perspective. Journal of International Business, 31(2), 325–335.

Maggi, G. (1993). Technology gap and international trade: an evolutionary model.

Journal of Evolutionary Economics, 109–126.

Markley, B., & Yankelevich, A. (2012). Uruguay: Agricultural Biotechnology Annual.

Whashington, DC.

139

Martínez-zarzoso, I., Vidovic, M., & Voicu, A. M. (2014). EU-Accession Effects on Sectoral

Trade : A Helpman-Melitz-Rubinstein Approach with Panel Data (No. No. 4903).

Category 8: Trade Policy. Munich.

Mayer, T., & Zignago, S. (2011). Notes on CEPII’s distance measures: the GeoDist database.

CEPII Working Paper. Paris. Retrieved from http://mpra.ub.uni-

muenchen.de/36347/2/MPRA_paper_36347.pdf

Mccluskey, J. J., Grimsrud, K., & Wahl, T. I. (2006). Comparison of Consumer Responses

to Genetically Modified Foods in Asia, North America, and Europe. In R. Just, J.

Alston, & D. Zilberman (Eds.), Regulating agricultural biotechnology : economics and

policy (p. 732). New York, NY: Springer New York.

Melitz, M. J. (2003). The Impact of Trade on Intra-Industry Reallocations and Aggregate

Industry Productivity. Econometrica, 71(6), 1695–1725.

Meyer, K. (2001). Institutions, transaction costs, and entry mode choice in Eastern

Europe. Journal of International Business Studies, 2, 357–367.

Moola, S., & Munnik, V. (2007). GMOs in Africa : food and agriculture Status report 2007.

Africa. Johannesburg.

Moschini, G. (2004). Innovation and Trade with Market Failure: The Case of Genetically

Modified Products, 86(3), 634–648.

Noussair, C., Robin, S., & Ruffieux, B. (2002). Do consumers not care about biotech foods

or do they just not read the labels? Economics Letters, 75(1), 47–53.

doi:10.1016/S0165-1765(01)00594-8

Oliveira, A. L. R., Silveira, J. M. F. J., & Alvin, A. M. (2012). Cartagena protocol , biosafety

and grain segregation : The effects on the soybean logistics in Brazil. E3 Journal of

Agricultural Research and Development, 2(1), 17–30. Retrieved from

http://ssrn.com/abstract=2002261

140

Oliveira, P. R. S., Silveira, J. M. F. J., Magalhães, M. M., & Souza, R. F. (2013). Adoption Of

Technologies With Market Rejection And Return On Exports: Brazil, United States,

Argentina And The Case Of Gm-Soybeans. In 17th ICABR Conference: “INNOVATION

AND POLICY FOR THE BIOECONOMY.” Ravello - Italy.

Otero, A. (2015). Agricultural Biotechnology Annual: Mexico. Whashington, DC.

Parcell, J. L., & Kalaitzandonakes, N. G. (2004). Do agricultural commodity prices

respond to bans against bioengineered crops? Canadian Journal of Agricultural

EconomicsRevue Canadienne D Agroeconomie, 52(2), 201–209. Retrieved from

http://www3.interscience.wiley.com/journal/118748499/abstract

Pew Reserch Center. (2003, June 20). Broad Opposition to Genetically Modified Foods.

Global Attitudes & Trends, pp. 6–7. Whashington, DC. Retrieved from

http://www.pewglobal.org/2003/06/20/broad-opposition-to-genetically-

modified-foods/#about-this-surve

Posner, M. (1961). International trade and technical change. Oxford Economic Papers,

13(3), 323–341. Retrieved from

http://oep.oxfordjournals.org/content/13/3/323.short

Qaim, M., & Zilberman, D. (2003). Yield effects of genetically modified crops in

developing countries. Science (New York, N.Y.), 299(5608), 900–902.

doi:10.1126/science.1080609

Regimes, R., & Giannakas, K. (2013). Inserting Gm Products Into The Food Chain: The

Market And Welfare Effects Of Different Labeling And Regulatory. American Journal

of Agricultural Economics, 86(1), 42–60.

Rugman, A. (1976). Risk reduction by international diversification. Journal of

International Business Studies, 7(2), 75–80.

Samuelson, P. (1952). The transfer problem and transport costs: the terms of trade

when impediments are absent. The Economic Journal, 62(246), 278–304. Retrieved

141

from http://www.jstor.org/stable/2227005

Schumpeter, J. A. (2003). Capitalism , Socialism and Democracy. Taylor & Francis e-

Library.

Secretariat of the Convention on Biological Diversity. Cartagena Protocol on Biosafety to

the Convention on Biological Diversity (2000). doi:10.1016/S0006-

3207(03)00053-3

Semykina, A., & Wooldridge, J. M. (2010). Estimating panel data models in the presence

of endogeneity and selection. Journal of Econometrics, 157(2), 375–380.

doi:10.1016/j.jeconom.2010.03.039

Smart, R. D., Blum, M., & Wesseler, J. (2015). EU Member States’ Voting for Authorizing

Genetically Engineered Crops: a Regulatory Gridlock. German Journal of

Agricultural Economics, 64(4), 244–262. Retrieved from <Go to

ISI>://CCC:000368519000004

Smyth, S., Kerr, W. a., & Davey, K. a. (2006). Closing markets to biotechnology: does it

pose an economic risk if markets are globalised? In International Journal of

Technology and Globalisation (Vol. 2, p. 377). doi:10.1504/IJTG.2006.011923

Sousa, J. de. (2012). The currency union effect on trade is decreasing over time.

Economics Letters, 117(3), 917–920. doi:10.1016/j.econlet.2012.07.009

Stein, A. J., & Rodriguez-Cerezo, E. (2009). The global pipeline of new GM crops

Implications of asynchronous approval. Luxembourg. doi:102791/1287

Strossman, C. (2009). Agricultural Biotechnology Annual: Switzerland. USDA Global

Agricultural Information Network. Whashington, DC.

Sturges, W., Perkins, H., Boyd, J. D., Stalcup, M. C., Talley, L. D., Johnson, G. C., …

Hernandez, G. (2003). Insect-Resistant GM Rice in Farmers ’ Fields : Assessing

Productivity and Health Effects in China.

142

Teh, R., & Piermartini, R. (2009). Demystifying modelling methods for trade policy.

The Mellman Group. (2006). Memorandum On Food And Biotechnology: Review Of Public

Opinion Research.

Tinbergen, J. (1962). Shaping the world economy: suggestions for an international

economic policy. Twentieth Century Fund, New York. Retrieved from

http://onlinelibrary.wiley.com/doi/10.1002/tie.5060050113/abstract

Tomich, F. A., & Leite, C. A. M. (1999). Competitividade das exportações brasileiras de

frutas selecionadas. Universidade Federal de Viçosa.

Vigani, M., & Olper, A. (2013). GMO standards, endogenous policy and the market for

information. Food Policy, 43, 32–43. doi:10.1016/j.foodpol.2013.08.001

Vigani, M., Raimondi, V., & Olper, A. (2012). International trade and endogenous

standards: the case of GMO regulations. World Trade Review, 11(03), 415–437.

doi:10.1017/S1474745612000262

White, H. (1980). A Heteroskedasticity-Consistent Covariance Matrix Estimator and a

Direct Test for Heteroskedasticity. Econometrica, 48(4), 817–838.

doi:10.2307/1912934

Wilkinson, J., & Castelli, P. (2000). A transnacionalização da indústria de sementes no

Brasil: biotecnologias, patentes e biodiversidade. ActionAid Brasil.

Winchester, N., Rau, M. L., Goetz, C., Larue, B., Otsuki, T., Shutes, K., … Nunes de Faria, R.

(2012). The Impact of Regulatory Heterogeneity on Agri-food Trade. World

Economy, 35(8), 973–993. doi:10.1111/j.1467-9701.2012.01457.x

Winham, G. R. (2009). The GMO Panel: Applications of WTO Law to Trade in

Agricultural Biotech Products. Journal of European Integration, 31(3), 409–429.

doi:10.1080/07036330902782261

143

Wooldridge, J. M. (1995). Selection corrections for panel data models under conditional

mean independence assumptions. Journal of Econometrics, 68, 115–132.

doi:10.1016/0304-4076(94)01645-G

WTO, & UNCTAD. (2012). Analyzing bilateral trade using gravity equation. In A Pratical

Guide to Trade Policy Analysis (pp. 101–136).

Yankelevich, A. (2011). Paraguay: Agricultural Biotechnology Annual. Global Agricultural

Information Network - USDA. Washington D.C.

universidade estadual de campinas instituto de … · tese de doutorado paulo ricardo da silva...

Documents