ABSTRACT
Title of dissertation: ESSAYS ON
PREFERENTIAL TRADE AGREEMENTS
Edith Laget
Doctor of Philosophy, 2019
Dissertation directed by: Professor Nuno Lima?o
Department of Economics
Preferential Trade Agreements (PTAs) are a prominent feature of current
globalization. Over the last decades, country participation in PTAs has become
widespread, with each member of the World Trade Organization having signed an
average of 10 PTAs, up from 3 PTAs in 1990. Most importantly, the proliferation of
PTAs was accompanied by a significant deepening of their scope. Their content now
spans diverse behind-the-border disciplines, such as investment, technical barriers
to trade (TBT), sanitary and phytosanitary (SPS) measures , intellectual property
rights, visa and asylum, labor market laws and environmental regulations.
In order to quantify the impacts of the provisions included in PTAs on various
outcomes of interest, such as gross trade, foreign direct investment (FDI), global
value chains, quality, and so on, trade economists face several empirical obstacles to
model Non-Tariff Measures (NTMs). While it is straightforward to model provision,
such as an import quota or an export tax with an ad-valorem tariff equivalent, this
technique is not suitable for other provisions that do not purely deal with market
access. Many NTMs are implemented to address behind-the-border issues rather
than to discriminate against foreign businesses. Therefore, it would be misleading
to restrict the effect of certain trade policies to their market access dimension only.
The objective of this dissertation is two-fold: it is, first, to understand the impacts
of the overall content of PTAs on economic outcomes and, second, to shed light on
the relationships of specific disciplines included in PTAs with those outcomes.
In the first chapter, I review the recent evolution of trade and investment
integration and how the content of PTAs has been reshaped over the years. The
second and third chapters are dedicated to the analysis of two of the most frequent
provisions in PTAs ? TBT and SPS provisions. In the theoretical part, I augment
the structure of the traditional Melitz model to assess the impact on quality of such
provisions. I model TBT/SPS measures as domestic regulations ensuring minimum
quality of goods. By integrating these regulations, PTAs change the economic struc-
ture of the model (with respect to minimum quality enforcement) from segmented
to joint markets. I highlight two potential channels to explain the change in quality
of exported goods following the enforcement of a PTA with TBT/SPS provisions.
The first channel for quality improvement is driven by the increase in market size.
With larger markets to export to, firms have the incentive to differentiate vertically
their products in order to capture bigger shares. The second channel depends on
the type of provisions implemented. I consider two cases, mutual recognition ver-
sus harmonization of TBT/SPS measures. The ultimate impact on quality of this
regulatory channel depends on the new reference for minimum quality once a PTA
is signed. Then, I empirically test the importance of these two channels using new
data on the content of PTAs and estimates of the quality of imported goods. I find
that mutual recognition positively impacts quality relatively more than harmoniza-
tion. This result is driven by PTAs between developed and developing countries.
The effect is heterogenous across sectors, with bigger impact of deep PTAs on goods
that have a wider scope for quality differentiation. Finally, I study the impact of
PTAs on FDI and find that deep PTAs promote foreign investments. The impact
is bigger for projects related to service activities, as well as North-South investment
relationships.
ESSAYS ON
PREFERENTIAL TRADE AGREEMENTS
by
Edith Laget
Dissertation submitted to the Faculty of the Graduate School of the
University of Maryland, College Park in partial fulfillment
of the requirements for the degree of
Doctor of Philosophy
2019
Advisory Committee:
Professor Nuno Lima?o, Chair
Professor Eunhee Lee
Professor Felipe Saffie
Professor John Shea
Professor Lars Olson
?c Copyright by
Edith Laget
2019

Dedication
A He?le?ne
ii
Acknowledgments
I am indebted to my advisor, Professor Nuno Lima?o, for his constant guidance
and support, and for his enormous patience throughout this long process. I am
thankful to Professor John Shea, Professor Eunhee Lee, Professor Felipe Saffie, and
Professor Lars Olson for giving me valuable feedbacks and kindly agreeing to serve
in my dissertation committee.
I would like to thank all my friends from the current and past trade groups
Lerong Li, Alejandro Graziano, Karam Jo, Heehyun Lim, Marisol Rodriguez Cha-
truc, Yang Xu and Jingting Fan, who were always keen to discuss ideas and helped
me solve issues with my research. I also thank Professor Basile Grassi for his friend-
ship and support, from our first days at the ENSAE until the final review of this
dissertation.
I owe a great deal of gratitude to my colleagues at the World Bank Nadia
Rocha, Michele Ruta and Gonzalo Varela for their trust, practical advices and in-
credible mentoring.
None of this would have been possible without the constant emotional support
of my entire family, in particular my parents Catherine and Michel Laget and my
brother Etienne. I have no word to describe how deeply grateful I am to my husband
Antoine for accompanying and supporting me in the most perfect way throughout
this difficult journey.
Most importantly, thank you my little He?le?ne for giving me the last piece of
motivation to finish my thesis.
iii
Table of Contents
Dedication ii
Acknowledgements iii
Table of Contents iv
List of Tables vii
List of Figures ix
List of Abbreviations x
1 Introduction to modern Preferential Trade Agreements 1
1.1 Evolution of trade integration . . . . . . . . . . . . . . . . . . . . . . 1
1.2 The rise of non-tariff measures . . . . . . . . . . . . . . . . . . . . . . 7
1.3 Content of modern PTAs . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.4 Thesis overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.5 Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.6 Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2 A model of TBT/SPS measures in PTAs 23
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.2 Related literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.3 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.3.1 TBT and SPS measures . . . . . . . . . . . . . . . . . . . . . 33
2.3.2 Consumer?s problem . . . . . . . . . . . . . . . . . . . . . . . 37
2.3.3 The firm?s profit maximization problem . . . . . . . . . . . . . 38
2.3.4 Productivity thresholds . . . . . . . . . . . . . . . . . . . . . . 43
2.4 Aggregated quality . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
2.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.6 Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
2.7 Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3 Impact of deep PTAs on quality 52
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.2 Literature on quality estimation . . . . . . . . . . . . . . . . . . . . . 54
3.3 Estimation of quality . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
iv
3.3.1 Demand for imports . . . . . . . . . . . . . . . . . . . . . . . 58
3.3.2 Dealing with price endogeneity . . . . . . . . . . . . . . . . . 59
3.3.3 Impact of TBT/SPS provisions on quality . . . . . . . . . . . 62
3.4 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.4.1 Data for the quality estimation . . . . . . . . . . . . . . . . . 69
3.4.2 Data on TBT/SPS chapter . . . . . . . . . . . . . . . . . . . . 70
3.4.3 ISO survey . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
3.4.4 Specific trade concerns . . . . . . . . . . . . . . . . . . . . . . 76
3.5 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
3.5.1 Quality estimates . . . . . . . . . . . . . . . . . . . . . . . . . 77
3.5.2 Baseline effect of TBT/SPS measures on quality . . . . . . . . 81
3.5.3 Interaction with level of development . . . . . . . . . . . . . . 83
3.5.4 Interaction with ISO 9001 . . . . . . . . . . . . . . . . . . . . 84
3.5.5 Interaction with quality ladder . . . . . . . . . . . . . . . . . . 85
3.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
3.7 Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
3.8 Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4 Deep PTAs and FDI 104
4.1 Disclaimer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
4.2 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
4.3 Related literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
4.4 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
4.4.1 Preferential trade agreements . . . . . . . . . . . . . . . . . . 112
4.4.2 Investments . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
4.4.3 Control variables . . . . . . . . . . . . . . . . . . . . . . . . . 118
4.5 Effect of PTAs on FDI . . . . . . . . . . . . . . . . . . . . . . . . . . 118
4.5.1 Empirical strategy . . . . . . . . . . . . . . . . . . . . . . . . 119
4.5.2 Baseline results . . . . . . . . . . . . . . . . . . . . . . . . . . 120
4.5.3 The impact of single disciplines . . . . . . . . . . . . . . . . . 121
4.5.4 Level of development . . . . . . . . . . . . . . . . . . . . . . . 124
4.6 Robustness checks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
4.6.1 Composition concerns . . . . . . . . . . . . . . . . . . . . . . 126
4.6.2 Simultaneity concerns . . . . . . . . . . . . . . . . . . . . . . 126
4.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
4.8 Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
4.9 Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
5 Conclusion 144
A Additional tables and figures 146
A.1 Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
v
B Derivation of the optimal TBT/SPS-related costs and productivity thresholds161
B.1 Maximization problem under no recognition . . . . . . . . . . . . . . 161
B.1.1 Finding optimal qualities under NR . . . . . . . . . . . . . . . 162
B.1.2 Finding productivity threshold under NR . . . . . . . . . . . . 162
B.2 Maximization problem under full integration . . . . . . . . . . . . . . 165
B.2.1 Finding optimal qualities under MR or HR . . . . . . . . . . . 166
B.2.2 Finding productivity threshold under MR or HR . . . . . . . . 166
Bibliography 170
vi
List of Tables
1.1 Non-tariff measure classification from UNCTAD . . . . . . . . . . . . 21
1.2 Frequency of different types of provisions in PTAs . . . . . . . . . . . 22
2.1 Examples of most important trade NTMs perceived by businesses in
the pharmaceutical industry . . . . . . . . . . . . . . . . . . . . . . . 50
2.2 Number of TBT concerns by type and objectives . . . . . . . . . . . 51
3.1 Elasticities of substitution from Broda and Weinstein (2006) . . . . . 93
3.2 Summary statistics for the estimation of the import demand equation 93
3.3 Quality and factors of endowment . . . . . . . . . . . . . . . . . . . . 94
3.4 Relationship between unit prices and quality . . . . . . . . . . . . . . 94
3.5 Distribution of quality estimates by broad sectors . . . . . . . . . . . 95
3.6 Baseline estimation of the impact of TBT and SPS provisions on
average quality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
3.7 Baseline estimation of the impact of TBT provisions on average quality 97
3.8 Baseline estimation of the impact of SPS provisions on average quality 98
3.9 Impact of TBT measures on quality across development levels . . . . 99
3.10 Impact of SPS measures on quality based on the level of development
of partners . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
3.11 TBT/SPS measures interaction with ISO 9001 certification gap . . . 101
3.12 Interaction between TBT measures and ladder length . . . . . . . . . 102
3.13 Interaction between SPS measures and ladder length . . . . . . . . . 103
4.1 Business activities and industries included in the classification of fDi
Markets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
4.2 Impact of total depth on investment . . . . . . . . . . . . . . . . . . . 134
4.3 FDI in extractive activities are not driven by PTAs . . . . . . . . . . 135
4.4 Disciplines level estimations . . . . . . . . . . . . . . . . . . . . . . . 136
4.5 Disciplines level estimations interacted with services . . . . . . . . . . 136
4.6 Disciplines effects by level of development . . . . . . . . . . . . . . . 137
4.7 Interaction with China in North-South agreements . . . . . . . . . . . 137
4.8 Robustness check: removing top 1 percent of projects announcements 138
4.9 Lobbying contributions by issues . . . . . . . . . . . . . . . . . . . . 138
4.10 robustness check: removing lobbying intensive sectors . . . . . . . . . 139
4.11 Industries and business activities partitioning . . . . . . . . . . . . . 140
A.1 List of WTO-plus disciplines included in PTAs . . . . . . . . . . . . . 146
vii
A.2 List of WTO-X disciplines included in PTAs . . . . . . . . . . . . . . 148
A.3 Template for TBT chapters? mapping . . . . . . . . . . . . . . . . . . 150
A.4 Template for SPS chapters? mapping . . . . . . . . . . . . . . . . . . 152
A.5 Baseline estimation of the impact of TBT provisions on average qual-
ity without sector fixed effects . . . . . . . . . . . . . . . . . . . . . . 155
A.6 Baseline estimation of the impact of TBT provisions on average qual-
ity without sector fixed effects . . . . . . . . . . . . . . . . . . . . . . 156
A.7 TBT lagged interaction with level of development . . . . . . . . . . . 157
A.8 SPS lagged interaction with level of development . . . . . . . . . . . 158
A.9 TBT lagged interaction with ladder length . . . . . . . . . . . . . . . 159
A.10 SPS lagged interaction with ladder length . . . . . . . . . . . . . . . 160
viii
List of Figures
1.1 Evolution of the number and number of provisions of PTAs . . . . . . 16
1.2 Number of partners sharing a PTA as of 1990 . . . . . . . . . . . . . 17
1.3 Number of partners sharing a PTA as of 2015 . . . . . . . . . . . . . 17
1.4 Average number of policy areas included in PTAs as of 1990 . . . . . 18
1.5 Average number of policy areas included in PTAs as of 2015 . . . . . 18
1.6 Overall incidence of NTMs . . . . . . . . . . . . . . . . . . . . . . . . 19
1.7 Number of TBT and SPS-specific trade concerns raised to the WTO . 20
2.1 Incidence of TBT and SPS chapters in PTAs . . . . . . . . . . . . . . 48
2.2 Minimum quality requirements . . . . . . . . . . . . . . . . . . . . . . 49
2.3 Domestic and foreign minimum quality standards . . . . . . . . . . . 49
3.1 Number of concerns raised for SPS and TBT measures . . . . . . . . 89
3.2 Average quality ladders by broad industries . . . . . . . . . . . . . . 90
3.3 Average correspondence between Rauch?s and HS classificaiton . . . . 91
3.4 Total ISO 9001 certifications by broad HS industries . . . . . . . . . . 92
4.1 BITs and investment related disciplines in PTAs . . . . . . . . . . . . 130
4.2 Frequency of inclusion of investment related disciplines in PTAs . . . 131
4.3 Correlation between fDi markets and UNCTAD bilateral investment
data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
ix
List of Abbreviations
BACI Base d?Analyse du Commerce International
CEPII Centre de?tudes prospectives et d?informations internationales
CES Constant Elasticity of Substitution
CFI cost, insurance and freight
EMA European Medicine Agency
EU European Union
FDA Food Drug Administration
FDI Foreign Direct Investment
FOB free on board
GATT General Agreement on Tariffs and Trade
GDP Gross Domestic Product
GVC Global Value Chain
IPR Intellectual Property Rights
ISO International Organization for Standardization
MFN Most Favored Nation
MRA Mutual Recognition Agreement
NAFTA North American Free Trade Agreement
NTB Non-Tariff Barrier
NTM Non-Tariff Measure
PTA Preferential Trade Agreement
SPS Sanitary and Phytosanitary
STC Specific Trade Concerns
TBT Technical Barriers to Trade
TPP Trans-Pacific Partnership
TRAINS Trade Analysis Information System
TRIPS Trade-Related Aspects of Intellectual Property Rights
TTIP Transatlantic Trade and Investment Partnership
UNCTAD United Nations Conference on Trade and Development
US United States
USD US dollar
USTR United States Trade Representative
WIOD World Input-Output Database
WTO World Trade Organization
x
Chapter 1: Introduction to modern Preferential Trade Agreements
1.1 Evolution of trade integration
Trade has arguably been the most important driver of economic growth for
both advanced and developing economies in the second half of the 20th century.
Economists agree on the long term benefits of trade, based on empirical evidence
pointing out that economies more open to trade tend to show stronger growth and
development performance. After a long period of import substitution industrializa-
tion, developing economies that eventually shifted from protectionist to export-led
growth strategies recovered faster from the early 1980s global economic recession.
The early 1990s represented a tipping point for globalization, as an increasing num-
ber of countries embraced market-economy systems as well as global trade and
investment integration.
The positive relationship between trade and economic growth has long-been
established in the trade literature. Trade openness generates economic gains by
leveraging comparative advantage and enabling a more efficient exploitation of pro-
duction factors. Because it enhances competition and economic diversification, trade
can also a channel for increased productivity, positive knowledge spillover, and bet-
ter resilience to external shocks. For all these reasons, the quest for the gains from
1
trade has pushed governments seeking deeper international integration to reduce
the overall level of trade restrictions, which they have accomplished at both the
multilateral and regional levels.
Since the creation of the General Agreement on Tariffs and Trade (GATT) in
1947, member countries have engaged in several rounds of multilateral negotiations
to increase market access. Over the years, the negotiations have welcomed more
countries and supported more ambitious reforms. The desire to adjust international
trade rules to a far more complex policy environment and to deepen multilateral
commitments ultimately led to the foundation of the World Trade Organization
(WTO) in 1995. As part of the Uruguay Round agenda and subsequent ministerial
conferences, WTO member countries continuously extended market access and ne-
gotiated policies in areas such as trade in services, rules of origin, dispute settlement
or temporary trade remedies (safeguards, countervailing duties, and temporary ex-
port restraints), and also introduced non-trade related policies dealing with beyond
market access issues such as intellectual property and domestic regulations on non-
tariff measures (NTMs). Broadly defined, NTMs are ?policies other than custom
tariffs that can potentially have an adverse effect on trade in goods and services,
changing quantities traded, or prices or both.?1 Table (1.1) shows the UNCTAD
classification of NTMs in terms of scope and design. NTMs vary by their mode
of application, and consist of technical, non-technical or export-specific measures.
1As noted by the International Trade Center (ITC) ?NTM is a neutral concept and does
not necessarily imply a particular direction of impact. It is not synonymous with the fre-
quently used term, non-tariff barrier (NTB), which implies a negative impact on trade. NTBs
may be view as a subset of NTMs that have a protectionist or discriminatory intent.? (see:
https://ntmsurvey.intracen.org/support-materials/understanding-ntms)
2
NTMs are by far more complicated to interpret than traditional trade policies such
as customs tariffs. Their classification requires in depth knowledge of the legal ter-
minology used in the regulatory texts establishing them. The ways countries have
addressed the surge in such measures to mitigate their impact on trade is at the
center of my dissertation and I will discuss this issue more in the subsequent sections
of the present introductory chapter.
Despite the common and persistent interest in reducing trade barriers in their
multiple forms (traditional customs tariffs and NTMs with adverse effects on trade),
progress on trade integration within the WTO?s multilateral framework has stalled
since the last round of negotiations that began in Doha in 2001. The reform of agri-
cultural subsidies, among other proposed rules, is one of the most contentious reasons
for this round?s failure. 2 The conflicting positions associated with the WTO?s con-
sensus based decision-making process has prevented its current 164 members from
reaching an agreement so far. Arguably, this lack of progress at the global level has
led many member states to look for faster regional or bilateral integration solutions,
resulting in a proliferation of Preferential Trade Agreements (PTAs).
From early on, regional integration has coexisted with multilateral integration
through first the GATT and then the WTO. Under the GATT?s Article XXIV,
signatories are allowed to grant special treatment to one another by establishing
regional agreements, provided that they substantiality eliminate duties and other
barriers on all trade in the region (disapproving partial agreements on a limited
2In short, developed member states led by the European Union (EU) and the United States
(US), have refused to lower their domestic subsidies on agriculture, in exchange of increased market
access granted by their developing counterparts.
3
number of goods), and that outside countries do not face higher barriers than before
the establishment of the regional agreements. The Treaty of Rome, establishing the
European Economic Community, which was signed in 1957 and entered into force in
1958, was the first regional trade agreement notified to the GATT under its Article
XXIV. The 1994 North American Free Trade Agreement (NAFTA) was the second
major such regional trade agreements to be notified to the GATT.
From the 1950s to the 1990s, the number of PTAs increased continuously to
almost 70. It was around the time of NAFTA?s creation that regional integration
really boomed in terms of membership and scope of commitments. PTA creation
rapidly accelerated to reach close to 300 agreements currently in force. New data on
the content of PTAs compiled by the World Bank3 shows that more than 40 percent
of agreements include ?deeper? provisions tackling issues, such as anti-dumping and
countervailing measures, rules on competition, investment, movement of capital,
intellectual property rights, and TBT and SPS measures. Other areas, such as
labor market and environmental regulations, visa and asylum rule are also included
in at least 20 percent of active agreements.4 The data also shows that agreements
signed prior to the WTO include on average 9 policy areas, whereas the most recent
ones cover more than 15 different policy areas. Figures (1.1) to (1.5) display this
evolution of PTAs in terms of scope and depth.
The prevalence of PTAs and the deepening of their commitments has spurred a
3World Bank?s Content of PTAs database available at:
https://datacatalog.worldbank.org/dataset/content-deep-trade-agreements. More on this
data in the next sections.
4See tables A.1 and A.2 for the full lists of provisions included in trade agreements, and further
comments on the data in the subsequent section of this introductory chapter.
4
lot of interest in trade policy research. In his chapter of the Handbook on Commer-
cial Policy dedicated to PTAs [Lima?o, 2016], Lima?o organizes this research around
two basic questions: ?What explains the formation and proliferation of so many
PTAs and what are their basic trade and welfare effects on members?? He demon-
strates that conventional tools are inadequate to address these two questions when it
comes to modern PTAs, and therefore calls for careful revisions of traditional mod-
els of PTAs. Specifically, the estimated effects of PTAs obtained from the standard
gravity approach raise two puzzles. The first one is the PTA formation puzzle ? if
PTAs are expected to grant market access to the signatories, then why are so many
PTAs signed despite their heterogenous and sometimes small trade effects that are
commonly estimated in the literature? The second puzzle concerns PTA?s trade
elasticity. This puzzle emerges after correcting the statistical issues that led to the
first puzzle ? long-run (corrected) estimates of PTA effects on trade volumes are
actually too large to be explained by the observed reduction of trade barriers given
reasonable estimates of trade elasticities. In light of these two puzzles, reducing
modern PTAs to their effect on preferential tariffs would be a mistake, as it would
overlook their impact on behind-the-border barriers. Lima?o therefore recommends
that scholars carefully ?augment the economic and policy structure of traditional
models of PTAs.?
From the negotiations of the defunct Transatlantic Trade and Investment Part-
nership (TTIP) between the US and the EU, to the newly signed EU-Mercosur trade
agreement, modern PTAs tend to animate the public debate for the mere reason
that it is complicated to predict their economic impact. Proponents claim that in-
5
creased business opportunities would lower prices and create jobs, while opponents
fear that the agreement would lower the capacity of local governments to regulate
industries and give too much power to transnational corporations. Moreover, negoti-
ations of such so-called ?mega trade deals? are not happening in the most favorable
environment. The world economy is slowly recovering from the Great Recession but
with lingering effects on trade and FDI flows and voters? confidence. Within the
last decade, a backlash against globalization has gained momentum and has been
exploited by populist political movements, which ultimately found their way into
several newly-elected governments. In this context, trade economists? role to pro-
vide unbiased evidence on the effects of trade and investment integration is more
important than ever.
This thesis aims to contribute to restoring faith in the potential of rule-
based international economic integration. It follows the recommendations stated
in [Lima?o, 2016] by improving traditional models of PTAs and incorporating speci-
ficities of modern agreements in order to grasp their full impact on various outcomes
of interest. Throughout the text, I will use interchangeably the terms ?modern? or
?deep? PTAs to qualify those agreements that address behind-the-border measures.
In the first part of my dissertation, I study one of the most common policies included
in PTAs, namely TBT and SPS provisions.
In the following two chapters, I explain how I modify the traditional [Melitz,
2003] heterogenous firms model to account for the effects of TBT and SPS measures
on firms? decisions to produce and export. Specifically, I incorporate a quality choice
mechanism that enables me to quantify the impact of TBT and SPS provisions
6
in PTAs on the average quality of exported goods. I show how such provisions
change the market structure of standard models of PTAs by modifying the way firms
optimally choose quality, from separable to joint decisions. Models of modern PTA
can help explore their impact on various key outcomes, such as trade, global value
chains (GVC) integration, and foreign direct investment (FDI). The last chapter of
this dissertation focuses on the impact of deep agreements on FDI.
1.2 The rise of non-tariff measures
Before going into the details of the analysis of modern PTAs, it is worth
giving an overview of the closely related subject of NTMs. As custom tariffs were
progressively reduced under the WTO and PTAs, policy makers and private sector
stakeholders have become increasingly concerned by the rise of these multiple and
often less transparent obstacles to trade. As mentioned above, NTM is a broad
term referring to any measure other than tariffs, which alters (purposely or not)
the competitiveness of foreign goods. These measures comprise a large array of
instruments, such as quantitative restrictions, TBT and SPS measures, pre-shipment
inspections, etc. The WTO dedicated a whole report to this topic a few years
ago [WTO, 2012] shedding light on how NTMs are becoming a major source of
trade costs.
Because of their complex nature, there is not a unified approach for identifying
the magnitude of NTMs and assessing their effects on international trade. One way
to do this is to compute their ad valorem equivalents. [Kee et al., 2008] and [Kee
7
et al., 2009] develop indicators to compare the relative effects of tariffs and NTMs.
They estimate ad valorem equivalents of NTMs, and use them to compute an Overall
Trade Restrictiveness Index. Building on the original work of [Anderson and Neary,
1994], they show that NTM?s effects are on average equivalent to a 12 percent
applied tariff, which increases the restrictiveness already imposed by traditional
trade barriers by 87 percent.
A more basic measure relies on a frequency index that accounts only for the
presence or absence of an NTM in a country. Figure (1.6) displays the aggregated
number of NTMs recorded in the UNCTAD TRAINS database. This NTM dataset
covers more than 55.000 measures implemented in 109 countries. Since countries
are not required to declare their use of NTMs, data collection is a resource-intensive
task conducted by legal experts, who review legal documents, regulations, direc-
tives, rules and the like in the most comprehensive and comparable possible fashion.
The figure shows that SPS and TBT are by far the most commonly used behind-
the-border measures. For the sake of health and safety, governments are allowed to
implement non-discriminatory quality standards, technical regulations, risk assess-
ments and the like, on any goods available in their market. This is stated in the
WTO?s TBT and SPS Agreements that entered into force with the WTO?s estab-
lishment in 1995. These agreements recognize the right of each member country ?to
take measures necessary to pursue national security, to prevent deceptive practices
or to protect human health or safety, animal or plant life, or health or the environ-
ment.? They also require that any such TBT/SPS measure be as non-disruptive as
possible for trade.
8
Yet, the frontier between legitimate policy objectives and protectionist motives
is often blurry. TBT/SPS measures can impose disproportionate costs on foreign
producers or create an unfair disadvantage for foreign competitors. Although these
measures should not be protectionist in nature, and should not apply differently to
foreign and domestic producers, they often result in increasing costs for exporters
which must comply with additional regulations. This is particularly burdensome for
developing countries? exporters, who might lack the capacity to comply with such
measures because of inadequate production and/or certification processes. On the
other hand, there is evidence that developing countries have increased their use of
TBT/SPS measures at least as much as developed countries. Figure (1.7) shows
the evolution of TBT and SPS Specific Trade Concerns raised to the WTO against
developed, developing and least developed countries.
TBT/SPS-related NTMs can take endless forms. The annual National Trade
Estimate Report on Foreign Trade Barriers prepared by the Office of the United
States Trade Representative (USTR) provides concrete examples of TBT/SPS prac-
tices and estimates of their trade costs whenever possible. The report is a compre-
hensive survey discussing the state of major trade barriers (of any type) currently
faced by the US (and others) stakeholders in the largest export markets (61 coun-
tries, plus the European Union, Taiwan, Hong Kong, and the Arab League). For
example the 2019 National Trade Estimate Report5 indicates that India imposes
security and safety requirements for telecommunication equipment. This TBT mea-
sure is described as follows: ?The Compulsory Registration Order prescribes safety
5https://ustr.gov/sites/default/files/2019 National Trade Estimate Report.pdf
9
standards and in-country testing requirements for electronic and information and
communications technology goods. The policy, which entered into force in January
2014, mandates that manufacturers register their products and have them certi-
fied by laboratories accredited by the Bureau of Indian Standards (BIS), even if
the products have already been certified by accredited international laboratories.
[...] U.S. stakeholders have raised concerns regarding delays in product registration
due to the lack of government testing capacity, a cumbersome registration process,
canceled registrations for administrative reasons unrelated to safety, and additional
compliance costs that can exceed tens of millions of dollars, including costs associ-
ated with factory-level and component-level testing.? An example of an SPS barrier
is discussed regarding South Korea?s rules for beef and beef products: ?Prior to
2008, Korea restricted the importation of U.S. beef and beef products, citing bovine
spongiform encephalopathy (BSE)-related concerns. In 2008, the United States and
Korea reached a bilateral agreement to fully reopen Korea?s market to U.S. beef
and beef products. However, as a transitional measure, U.S. beef and beef products
imported into Korea must be derived from animals less than 30 months of age.?
A growing strand of the literature analyzes the impacts on trade of the surge of
NTMs. Some studies have investigated the use of NTMs in place of low import tariffs
for protectionist purposes and showed that the shares of trade covered by NTMs
have dramatically increased since the financial crisis. [Bown and Crowley, 2013a,
Bown and Crowley, 2013b, Bown and Crowley, 2014] use data on temporary trade
barriers to show that their increasing use is correlated with countries? commitment
to multilateral tariff discipline.
10
Statistical and quantitative evidence associated with surveys like the USTR?s
shows that SPS and TBT measures have a high potential for trade diversion for those
exporters that might not be able to comply with the safety standards. Through
the WTO and PTAs, policy makers have worked to find a compromise between
maintaining the legitimate objectives of SPS and TBT measures and reducing the
trade costs they generate. As argued, by opponents of the TTIP, increased market
access should not come at the expense of product quality, consumers, animal or
environment safety.
1.3 Content of modern PTAs
As mentioned above, most of the 294 PTAs currently in force and notified to
the WTO have been signed within the last three decades, a period during which the
number of agreements has multiplied five-fold. The most dramatic reforms in terms
of trade policy emerged in the mid-90s when countries started to include provisions
lying outside the WTO?s mandate with the objective to tackle the rapid increase in
NTMs.
In this thesis, I use a new dataset on the content of PTAs developed by the
World Bank. Its first version6, released in 2016, maps the content of all 279 PTAs
that are in force and notified to the WTO as of 2015. The mapping methodology
is based on the work of [Horn et al., 2010], which was also used in the World Trade
Report 2011 [WTO, 2011]. Specifically [Horn et al., 2010] identified a comprehen-
sive set of 52 ?disciplines? (or policy areas) in 14 EU and 14 US PTAs notified to
6denoted 1.0 throughout the text
11
the WTO as of 2008.7 As noted by [Horn et al., 2010], ?the classification is largely
based in the articles headings in the case of the EU agreements, and on the chapters
headings in the case of the US agreements?. The World Bank extended the PTA
coverage to all agreements in force and notified to the WTO, but keeping the original
mapping across 52 disciplines.8 Each discipline is classified based on its connection
to the WTO mandate. The first category, denoted WTO-plus, corresponds to the
14 disciplines that fall under the current mandate of the WTO and are already sub-
ject to some form of commitment in its agreements, such as tariffs, anti-dumping,
TBT and SPS. The second category, denoted WTO-extra, groups the other 38 dis-
ciplines that are outside the current mandate of the WTO, such as investment and
competition policy. Table (A.1) lists the 14 WTO-plus disciplines and table (A.2)
the 38 WTO-X. On top of the area coverage, the data provide information on the
legal enforcement of each discipline within an agreement. The enforcement coding
is based on the analysis of the legal language of the treaty text and the possibility
of recourse to dispute settlement. All my empirical applications and content-related
measures are based on the identification of legally enforceable disciplines. A disci-
pline is considered legally enforceable ?if the language used is sufficiently precise and
committing (use of shall instead of should) and if it has not been excluded from dis-
pute settlement procedures under the PTA.? This first version of the dataset refers
7The 14 EU agreements are: EEA, EC-Turkey, EC-Tunisia, EC-Israel, EC-Morocco, EC-
Jordan, EC-South Africa, EC-Mexico, EC-FYRoM, EC-Egypt, EC-Croatia, EC-Chile, EC-
Albania, CARIFORUM, and the 14 US agreements are: US-Israel, NAFTA, US-Jordan, US-
Singapore, US-Chile, US-Australia, US-Morocco, CAFTA-DR, US-Bahrain, US-Oman, US-Peru,
US-Colombia, US-Panama, US-South Korea.
8 [Hofmann et al., 2017] provide a thorough description of the database and the way the mapping
of PTAs was implemented.
12
to the ?horizontal? mapping of PTAs. By giving the breadth of a PTA, the data
corresponds to the first dimension of an ambitious analytical research agenda aiming
at screening the whole content of PTAs. Apart from their legal enforceability, disci-
plines are not differentiated with one another across agreements. Nevertheless, the
horizontal breadth provides a rich source of information that enables me to model
the scope of a PTA more precisely than the usual dummy variables that have been
used in the literature so far.9
In a second round of efforts aiming at complementing the current knowledge
of PTAs, the Content database has recently been extended to its vertical dimension.
As part of my work in the World Bank?s Global Trade Unit, I got access to this
most recent and comprehensive mapping of PTAs to date before its release. The
vertical mapping in the Content database 2.0 provides further information on a sub-
set of 18 policy areas based on their economic importance and their frequency of
occurrence in PTAs. On average 40 questions, or provisions to use the terminology
employed by the World Bank, were identified by legal experts to characterize each
policy areas. A guideline was provided to the experts in order to maintain a har-
monized vertical mapping across policy areas. Whenever applicable, the provisions
coded were sorted under pre-defined categories, such as scope and definition, sec-
toral application, transparency, substantial commitments, and exceptions. Experts
were also required to assess the level of enforceability of the provisions, based on the
legal language used in the chapters or the existence of a dispute settlement mecha-
9The Design of Trade Agreements Database (DESTA) is another source of information on the
content of PTAs. It covers only 7 provisions, but extends the mapping to 620 agreements (including
partial free trade agreements). This greater number of agreements does not match the usual PTA
dummies used in the literature, which are often based on the list of PTAs notified to the WTO.
13
nism. As a result of this second mapping, the increased granularity of the Content
of PTAs? database expanded the number of variables from 52 to more than 1000.
1.4 Thesis overview
This introductory chapter has given an overview of the topics that have mo-
tivated my research. While emphasizing the complementarity between the multi-
lateral trading system and PTAs, I showed the stakes and challenges to properly
deal with NTMs in empirical analysis. The rest of my dissertation is organized as
follows:
In the next chapter, I lay out an augmented version of the [Melitz, 2003]
model with endogenous fixed effects and quality choice. TBT and SPS measures
are modeled as minimum quality requirement mechanisms that bring information to
the consumers. This model enables me to draw predictions regarding the impact on
quality of the presence of TBT and SPS provisions in trade agreements. I highlight
two channels for quality improvement in the context of trade liberalization. On the
one hand, firms might be inclined to vertically differentiate, meaning to increase
the quality of their exports, to capture bigger shares following market enlargement.
On the other hand, changes in TBT/SPS-related regulations may also have a direct
impact on quality by changing minimum quality requirements.
The third chapter tests the importance of the two channels with an empirical
exercise relying on an estimated measure of imported good quality. I follow closely
the literature [Khandelwal, 2010] to obtain quality estimates for bilateral trade
14
flows. I then use the second version of the Content of PTA dataset described in the
previous section to test which TBT/SPS provisions have the most significant impact
on the quality of traded goods. I study potential sources of heterogeneity in those
impacts, by integration approach mutual recognition versus harmonization, by level
of development of the signatories, and by sector. I find that mutual recognition
positively impacts quality relatively more than harmonization. This result is driven
by PTAs between developed and developing countries. Deep PTAs have bigger
impact on goods that have a wider scope for quality differentiation.
The fourth chapter presents research that I have conducted at the World Bank
in collaboration with Nadia Rocha and Gonzalo Varela.10 It is part of a wider re-
search agenda aiming at understanding the impact of the content of PTAs on various
economic outcomes.11 Our work focuses on a set of investment-related provisions
and shows how their contributions to FDI vary across sectors and level of devel-
opment. We find that deep PTAs promote foreign investments. The impact is
bigger for projects related to service activities, as well as North-South investment
relationships.
Finally, chapter 5 concludes and offers suggestions for future work.
10The usual disclaimer applies for this collaborative work: this chapter has not undergone the
review accorded to official World Bank publications. The findings, interpretations, and conclusions
expressed herein are those of the authors and do not necessarily reflect the views of the World
Bank Group and its affiliated organizations, or those of the Executive Directors of the World Bank
or the governments they represent.
11Others have shown how the content dataset can be used to reveal differences in effects between
shallow and deep PTAs. [Mattoo et al., 2017] show that deeper agreement (with more provisions)
promote trade in aggregate terms with limited diversion. The authors argue that deeper agree-
ments can actually have a positive spillover effect on trade with third countries if their design
and implementation are non-discriminatory. Deep trade agreements are also expected to facilitate
vertical FDI [Osnago et al., 2015], and to ease operations along GVCs [Laget et al., 2018].
15
1.5 Figures
Source: Author?s calculation based on the World Bank?s Content of PTAs
database version 1.0.
Figure 1.1: Evolution of the number and number of provisions of PTAs
16
Figure 1.2: Number of partners sharing a PTA as of 1990
Source: Author?s calculation based on the World Bank?s Content of PTAs
database version 1.0.
Figure 1.3: Number of partners sharing a PTA as of 2015
17
?-
Figure 1.4: Average number of policy areas included in PTAs as of 1990
Source: Author?s calculation based on the World Bank?s Content of PTAs
database version 1.0.
Figure 1.5: Average number of policy areas included in PTAs as of 2015
18
?.. 
Note: Shares of NTMs recorded in the UNCTAD TRAINS database.
Figure 1.6: Overall incidence of NTMs
19
Developed economies Developing & emerging economies
Least developed economies
Sources: WTO?s TBT and SPS committee.
Figure 1.7: Number of TBT and SPS-specific trade concerns raised to the WTO
20
0 50 100 150 200
1995 SPSTBT
1996 SPSTBT
1997 SPSTBT
1998 SPSTBT
1999 SPSTBT
2000 SPSTBT
2001 SPSTBT
2002 SPSTBT
2003 SPSTBT
2004 SPSTBT
2005 SPSTBT
2006 SPSTBT
2007 SPSTBT
2008 SPSTBT
2009 SPSTBT
2010 SPSTBT
2011 SPSTBT
2012 SPSTBT
2013 SPSTBT
2014 SPSTBT
2015 SPSTBT
2016 SPSTBT
2017 SPSTBT
2018 SPSTBT
1.6 Tables
A Sanitary and Phytosanitary measures
Technical Measures B Technical Barriers to trade
C Pre-shipment inspection and other formalities
D Contingent trade protective measures
E Non-automatic licensing and quantity control measures
F Price control measures, additional taxes and charges
G Finance measures
H Measures affecting competition
I Trade-related investment measures
Non Technical Measures
J Distribution restrictions
K Restriction on post-sales services
L Subsidies
M Government procurement restrictions
N Intellectual property
O Rules of origin
Export measures P Export related measures
Table 1.1: Non-tariff measure classification from UNCTAD
21
All countries ? 279 PTAs North - North ? 49 PTAs South - South ? 125 PTAs North - South ? 105 PTAs
FTA Industrial 0.98 FTA Industrial 1.00 FTA Industrial 0.97 FTA Industrial 0.99
FTA Industrial 0.98 FTA Industrial 1.00 FTA Industrial 0.97 FTA Industrial 0.99
FTA Agriculture 0.98 FTA Agriculture 1.00 FTA Agriculture 0.97 FTA Agriculture 0.98
Customs 0.81 Customs 0.90 Export Taxes 0.74 Customs 0.91
Export Taxes 0.76 Export Taxes 0.82 Customs 0.70 Export Taxes 0.77
AD 0.56 AD 0.71 Competition Policy 0.49 TRIPs 0.71
TRIPs 0.54 TRIPs 0.71 TBT 0.49 AD 0.65
State Aid 0.53 Movement of Capital 0.71 SPS 0.45 Movement of Capital 0.65
Movement of Capital 0.50 State Aid 0.69 AD 0.42 GATS 0.61
GATS 0.50 GATS 0.69 State Aid 0.41 State Aid 0.59
TBT 0.49 Public Procurement 0.63 GATS 0.33 CVM 0.55
CVM 0.46 CVM 0.61 CVM 0.33 IPR 0.55
STE 0.43 TBT 0.57 TRIPs 0.33 STE 0.51
Competition Policy 0.41 STE 0.55 STE 0.31 Public Procurement 0.48
SPS 0.40 SPS 0.51 Movement of Capital 0.30 Investment 0.48
Public Procurement 0.38 IPR 0.51 Investment 0.27 TBT 0.46
Investment 0.38 TRIMs 0.47 Public Procurement 0.21 TRIMs 0.37
IPR 0.37 Investment 0.45 TRIMs 0.19 Competition Policy 0.30
TRIMs 0.31 Competition Policy 0.41 Statistics 0.16 SPS 0.30
Labour Market Regulation 0.11 Visa and Asylum 0.31 IPR 0.15 Environmental Laws 0.12
Visa and Asylum 0.11 Labour Market Regulation 0.31 Energy 0.06 Visa and Asylum 0.12
Environmental Laws 0.10 Environmental Laws 0.29 Agriculture 0.05 Labour Market Regulation 0.10
Statistics 0.10 Social Matters 0.27 Labour Market Regulation 0.04 Anti-Corruption 0.09
Energy 0.08 Energy 0.24 Mining 0.04 Agriculture 0.07
Agriculture 0.08 Research and Technology 0.22 Industrial Cooperation 0.03 Data Protection 0.07
Social Matters 0.07 Education and Training 0.22 Illicit Drugs 0.03 Social Matters 0.06
Approximation of Legislation 0.05 Taxation 0.22 Visa and Asylum 0.02 Illegal Immigration 0.04
Industrial Cooperation 0.05 Economic Policy Dialogue 0.20 Public Administration 0.02 Information Society 0.03
Research and Technology 0.05 Agriculture 0.18 Environmental Laws 0.02 Energy 0.03
Anti-Corruption 0.05 Industrial Cooperation 0.18 Approximation of Legislation 0.02 Consumer Protection 0.03
Illegal Immigration 0.05 Cultural Cooperation 0.18 Consumer Protection 0.02 Public Administration 0.03
Taxation 0.05 Financial Assistance 0.18 Taxation 0.02 Financial Assistance 0.03
Education and Training 0.04 Approximation of Legislation 0.18 Anti-Corruption 0.02 Approximation of Legislation 0.03
Financial Assistance 0.04 Illegal Immigration 0.18 Audio Visual 0.02 Regional Cooperation 0.02
Consumer Protection 0.04 Statistics 0.14 Terrorism 0.02 Research and Technology 0.02
Cultural Cooperation 0.04 Consumer Protection 0.12 Money Laundering 0.02 SME 0.02
Data Protection 0.04 Nuclear Safety 0.12 Information Society 0.01 Health 0.02
Economic Policy Dialogue 0.04 Regional Cooperation 0.06 Regional Cooperation 0.01 Cultural Cooperation 0.01
Public Administration 0.03 Public Administration 0.06 Research and Technology 0.01 Industrial Cooperation 0.01
Mining 0.03 Health 0.04 Education and Training 0.01 Innovation Policies 0.01
Nuclear Safety 0.02 Audio Visual 0.04 SME 0.01 Mining 0.01
Regional Cooperation 0.02 Anti-Corruption 0.04 Data Protection 0.01 Political Dialogue -
SME 0.02 Data Protection 0.04 Economic Policy Dialogue - Education and Training -
Audio Visual 0.01 SME 0.04 Innovation Policies - Economic Policy Dialogue -
Health 0.01 Mining 0.04 Political Dialogue - Human Rights -
Illicit Drugs 0.01 Terrorism 0.04 Cultural Cooperation - Taxation -
Information Society 0.01 Human Rights 0.02 Social Matters - Statistics -
Terrorism 0.01 Civil Protection 0.02 Health - Illicit Drugs -
Money Laundering 0.01 Information Society - Financial Assistance - Money Laundering -
Civil Protection 0.004 Political Dialogue - Human Rights - Audio Visual -
Human Rights 0.004 Money Laundering - Nuclear Safety - Terrorism -
Innovation Policies 0.004 Illicit Drugs - Civil Protection - Nuclear Safety -
Source: World Bank?s Content of PTAs database version 1.0
Note: Frequency of legally enforceable provisions on PTAs as of 2015. In blue
WTO-plus provisions, in black WTO-X provisions, in bold-black WTO-X pro-
visions related to investments
Table 1.2: Frequency of different types of provisions in PTAs
22
Chapter 2: A model of TBT/SPS measures in PTAs
2.1 Introduction
PTAs have become a prominent feature of the world trading system over the
past 20 years. As they have increased in number, they have also broadened their
scope by including provisions beyond merely market access. In order to quantify
the impacts of the provisions included in PTAs, the trade policy literature faces
several challenges to model the economic and policy structure of Non-Tariff Measures
(NTMs). While it is straightforward to model provisions such as import quotas or
export taxes with ad-valorem tariff equivalents, this technique is not suitable for
provisions that do not purely deal with market access. Specifically, many NTMs are
implemented to address important behind-the-border issues and not to discriminate
against foreign businesses. Therefore, it would be misleading to restrict the effect
of certain trade policies to their market access dimension. The goal of this chapter
is to model the impact of two of the most frequent provisions in PTAs ? Technical
Barriers to Trade (TBT) and Sanitary and Phytosanitary (SPS) provisions ? by
incorporating their impact on quality.
Based on the first version of the World Bank?s Content of PTAs dataset,
figure (2.1) shows that the number of PTAs that include TBT or SPS provisions
23
has been consistently increasing in recent years. In 2015, 74 percent of PTAs in
force included TBT or SPS provisions, compared with 38 percent in 1995 (the year
at which the WTO?s Agreements on Technical Barriers to Trade and Sanitary and
Phytosanitary Measures entered force). Moreover, all agreements that entered into
force since 2009 have included TBT or SPS provisions. This supports the view that
the perceived importance of standard-related NTMs as an obstacle to trade has
increased over time, thus providing an incentive for countries to include TBT and
SPS provisions in the negotiations of PTAs.
TBT/SPS measures encompass concurrent procedures that consist of stan-
dards, technical regulations, risk assessments, inspections, and conformity assess-
ment procedures. Standards and technical regulations specify the required technical
characteristics of a product (such as the level of electromagnetic radiation distur-
bance for some medical devices). Risk assessment refers to the likelihood of adverse
effects of imports on humans, animals or the environment (which may depend on the
origin of the shipment and the customs declaration). Inspections may be carried
out in production facilities to verify good manufacturing practices (the US Food
and Drug Administration carries out such inspections within and outside the US).
Conformity assessment procedures define the testing procedures necessary to assess
the conformity of products to established norms (testing is undertaken by third
party conformity assessment bodies that must themselves be accredited/certified by
governments). All these procedures are implemented to ensure minimum quality of
products.
In general, TBT/SPS measures are not explicitly introduced for the purpose
24
of trade protection. They serve the laudable goals of protecting consumers, animals
and environment health and safety. For example, in Europe, all products sold in the
market must comply with relevant directives and regulations set by the European
Commission. Upon completion of appropriate testing, these products are affixed
with a CE marking visible to consumers and attesting to their conformity with EU
regulations. Similarly, in the US, the Consumer Product Safety Commission is the
federal regulatory agency in charge of ?protecting the public against unreasonable
risks of injury or death from consumer products?, notably by examining imported
shipments of consumer products. Overall, these measures bring information by sig-
naling quality to consumers who do not have the capacity to evaluate the quality
of a product by themselves. Voluntary adoption of TBT standards can also help
markets operate more efficiently, by enabling firms to produce compatible goods
and benefit from economies of scale. Nevertheless, neither the externality for the
efficiency motives are not explicitly modeled in what follows. The next model sim-
ply focuses on the consequences of using TBT/SPS measures, that is establishing
minimum quality requirements.
Yet, without cooperation on TBT/SPS procedures, differences can easily in-
hibit trade. Although the purpose of this chapter is not to study how countries
design their TBT/SPS regulations, it is worth noting that different economic funda-
mentals can lead to different regulations. TBT/SPS measures are designed through
the collaborative work of public and private experts. Because economic funda-
mentals, such as industry bargaining power, consumer preferences, institutions and
technology levels, are the underlying factors guiding the choice of TBT/SPS mea-
25
sures, different countries with different fundamentals or preferences are likely to set
different regulations. When TBT/SPS measures are implemented by two regions
sharing equivalent economic fundamentals, like the US and the EU in some sectors
for example, regulatory dissimilarity can generate unnecessary costs to exporters.
Different sources of information reveal the prevalence of TBT and SPS mea-
sures relative to other NTMs. As already shown in the Introductory chapter, the
UNCTAD NTM database reveals that TBT and SPS are by far the most commonly
implemented NTMs (see figure (1.6)). The UNCTAD NTM data is collected with-
out taking into account the potential adverse effects of NTMS on trade, but other
sources, such as the WTO?s Specific Trade Concerns (STC) database, address this
by recording the complaints of member states that face TBT/SPS measures with
alleged discriminatory impacts on their exports. Figure (1.7), from chapter 1, shows
that the incidence of complaints about TBT and SPS measures has been growing
in recent years. Surveys carried out among stakeholders also reveal the burden
imposed by differences in TBT/SPS measures between markets. Table (2.1) lists
examples of regulatory divergences between the American and European pharma-
ceutical industries, that are perceived by businesses as the most important sources
of barriers to trade. For instance, differences in the processes to approve new drugs
enforced by the FDA for the US and the European Medicine Agency (EMA) for
the EU significantly slow down the entry of new drugs in new markets. The total
trade costs generated by these regulatory divergences (and not just the differences
between FDA and EMA approval procedures) have been estimated using ad-valorem
26
equivalents and amount to USD 7.3 billion for the pharmaceutical sector.1
It is not the existence of TBT/SPS measures per se that creates barriers to
trade, but rather their variety. Indeed, when regulations differ from one market to
the other, exporting firms must incur additional costs to prove that their products
comply with each export market?s regulations. These costs might even prevent firms
from exporting. In the previous chapter, I showed that TBT/SPS-related provisions
were systematically included in the most recent PTAs. These were included in trade
agreements even before the creation of the WTO in order to minimize the trade bar-
riers caused by differences between regulations across trading partners.2 TBT/SPS-
related provisions in PTAs may take the form of mutual recognition or harmonization
of regulations.3 Practically, TBT/SPS-related provisions in PTAs allow exporters
to prove their compliance with quality requirements only once, without needing to
repeat the process of inspection, sampling, testing, certification, and the like in each
destination market. TBT/SPS-related provisions arguably reduce barriers to trade,
but we do not know their ultimate impact on the quality of imported goods. Because
the stringency of TBT/SPS measures differs across countries, opponents to modern
PTAs often argue that such types of integration will lower standards and have a
negative impact on quality of imports. Conversely, market enlargement enabled by
the reduction of fixed costs to export may foster vertical differentiation, or in other
1See [Ecorys, 2009] for more details on the methodology and costs for other sectors.
2Under the SPS and TBT Agreements, which came into force in 1995 with the creation of
the WTO, members are strongly encourage to adopt international standards as a means to facili-
tate trade. Nevertheless, the WTO recognizes its members? rights to implement non-harmonized
measures, as long as it is to achieve legitimate policy objectives.
3Chapter 3 provides more details on the different forms of TBT/SPS-related provisions existing
in PTAs.
27
words differentiation through quality improvement. These two hypothesis need to
be studied in a comprehensive theoretical framework that can accommodate market
enlargement and regulatory mechanisms.
In this chapter, I examine the economic effects of the presence of TBT/SPS-
related provisions in PTAs. I provide a theoretical framework, in which changes in
the competitive and regulatory environments have an impact on both prices and
quality. As previously mentioned, treating TBT/SPS measures as pure market ac-
cess regulations is misleading and might overlook their direct impact on consumers?
health and safety. Therefore, when studying such behind-the-border instruments
that pertain to domestic policies, standard trade models require some adjustments
to integrate their dual effect on consumers? protection (by ensuring a minimum
quality of goods) and trade barriers (by increasing fixed costs to export).
In what follows, I propose a variant of the [Melitz, 2003] model with quality
choice. Concretely, I model TBT/SPS measures as minimum quality requirements
for the firms. Quality is costly to produce and directly linked to fixed investments.
Firms optimally choose fixed investments to improve the quality of their produc-
tion, while fulfilling minimum quality requirements. The trade policy of interest is
the existence of a PTA that includes TBT/SPS regulations. The model is flexible
enough to accommodate mutual recognition and harmonization of TBT/SPS mea-
sures. Without TBT/SPS integration, exporting firms face additional fixed costs
to enter foreign markets because, on top of the domestic TBT/SPS compliances
costs, they must show that their goods also comply with the destination market?s
regulations. Mutual recognition policies enable signatory countries to acknowledge
28
their trading partners? regulations as being equivalent to their own and therefore
eliminate redundant TBT/SPS compliance costs. Domestic regulations are not ex-
pected to change in this case. Whereas harmonization policies support the adoption
of common regulations, implying changes in regulations from at least one of the sig-
natories. I show analytically that the implementation of both mutual recognition or
harmonization policies results in an endogenous reduction of the fixed investments
related to quality.
2.2 Related literature
This chapter is firstly related to the literature studying the economic impact
of TBT/SPS measures. The model developed in the next sections provides the
theoretical foundations to study the effects of mutual recognition or harmoniza-
tion of TBT/SPS measures. Researchers have empirically studied the effects of
such measures on trade and highlighted differences across sectors, integration ap-
proaches or trading partners. Using a measure of shared standards built from the
Perinorm database4, [Moenius, 2004] finds that country-specific standards hinder
trade in sectors producing simple goods (e.g. food, beverages and agricultural prod-
ucts), but promote trade for more complex goods (e.g. machinery and electronics).
The reason is that even though national standards may impose adaptation costs
on exporters, they also provide exporters with valuable information to integrate
a foreign market. Moenius argues that the negative effects due to adaption costs
4Perinorm is a bibliographic database (subscription only) that references national, European
and international standards from more than 200 standards publishing organizations in 23 countries,
with a total of more than 1,400,000 records.
29
are outweighed by the positive effects stemming from the reduction of transaction
costs. [Portugal-Perez et al., 2010] find that EU standards for electronic products
that are harmonized to international standards have a significant and positive net
effect on EU imports. [Chen and Mattoo, 2008] compare the effects of the harmo-
nization and mutual recognition of TBT measures in PTAs on signatories and third
countries. Instead of building a measure of shared standards, they rely on offi-
cial documents to identify the provisions associated with harmonization and mutual
recognition. Their analysis confirms that both types of agreement increase trade
between participating countries, but find different implications for third countries.
A harmonization of standards tends to increase exports from developed third parties
but reduces them from developing ones. One reason might be that harmonized stan-
dards are likely to be stricter in the participating countries, which would increase
the costs of adaptation of the excluded developing countries more than the benefit
of the economies of scale engendered by the harmonized region. By contrast, the
authors find that mutual recognition has more symmetric benefits for participating
and excluded countries. Unless the agreement contains restrictive rules of origin,
imports from both developed and developing excluded countries increase. Using a
similar strategy based on the screening of the treaties? texts, [Disdier et al., 2015]
analyze the impact on trade of technical regulations incorporated in North-South
trade agreements. To map the legal texts, the authors follow the same template
as [Budetta and Piermartini, 2009] and search for phrases associated with harmo-
nization or mutual recognition. They find that developing countries expand their
trade with their high income partners following deep integration agreements on stan-
30
dards. But this happened at the expense of their trade with non-bloc developing
partners.
This chapter is also related to the literature on quality choice, in particular to
the works of [Antoniades, 2015], [Johnson, 2012] and [Gervais, 2013]. These studies
focus on predicting trade patterns based on heterogeneous firm characteristics and
measuring the gains from trade openness. I instead focus on the changes in quality
from the implementation of a behind-the-border policy. As explained in the intro-
ductory chapter, commitments related to TBT and SPS measures are expected to
generate wider effects than the traditional PTAs? commitments on market access.
The implementation of such NTMs has implications for both barriers to trade and
product quality. By changing the way TBT and SPS measures are implemented, I
argue that PTAs trigger supply-side responses affecting endogenous variables that
eventually modify the average level of product quality. Using a multi-country, multi-
sector, heterogeneous firms model that incorporates a quality choice mechanism, I
show how changes in trade barriers due to TBT and SPS commitments eventually
impact the overall level of product quality.
Finally, while I do not derive a full general equilibrium model, I show what
would be the challenges of doing so. The traditional literature on PTAs focuses
on market access mechanisms. Leading examples are [Caliendo and Parro, 2014],
who develop a structural model with sectoral input-output linkages to quantify the
impact of NAFTA, and [Anderson et al., 2016], who calculate the general equilib-
rium effects of all PTAs using iterated gravity estimations. On the empirical side
it also has a long history of using Computable General Equilibrium (CGE) tech-
31
niques. It largely resorts on the use of ad-valorem equivalents to inform policy
makers on the potential impact of ?mega-trade deals?, such as the defunct TTIP
( [Francois et al., 2013] and [Fontagne? et al., 2013]) or most recently the renegotia-
tion of NAFTA ( [Dadkhah et al., 2019]). While the use of ad-valorem equivalents
or restrictiveness indices provides interesting insights on the global consequences
of tackling behind-the-border barriers, the CGE methodology must rely on restric-
tive modeling assumptions and is not flexible enough to address structural changes
following TBT/SPS provisions for instance. Concretely, in a scenario in which coun-
tries neither mutually recognize nor harmonize their TBT/SPS measures, markets
are segmented and firms make separate optimal quality decisions based on their
destination-specific profits. Following the policy change of entering into a trade
agreement with TBT/SPS provisions, markets are not segmented anymore (at least
from the quality point of view) and firms optimally choose quality based on their
overall profits. Therefore, the structure of my model moves from separated markets
to joint markets, which ultimately limits my ability to solve the model analytically.
2.3 Model
In this section, I develop a static model in which consumers value the quality
of the goods they consume. Firms operate in a monopolistically competitive en-
vironment as in [Melitz, 2003] and compete through both price and quality. The
quality introduced in this model has both horizontal and vertical attributes. The
former is due to the consumer?s love for quality and the latter to the firm?s abil-
32
ity to differentiate through quality. I assume that countries implement TBT and
SPS measures solely to ensure minimum quality.5 In practice, firms must prove
that their production satisfies a series of minimum quality requirements (guided by
the adoption of standards). In the current framework, the TBT/SPS measures, or
equivalently the minimum quality requirements, are exogenous. However, the choice
of quality is endogenous to the firms. The government?s decision to harmonize with
or mutually recognize its trading partner?s own TBT/SPS measures represents the
trade policy of interest.
Quality is costly to produce and I assume it is related to fixed costs. Hence, by
increasing fixed costs, minimum quality requirements influence the firms? decision
to enter a market. In terms of notation, there are N countries indexed by i for
exporters and j for importers, and S products indexed by s. Each country j is
populated with Lj consumers supplying one unit of work and whose wage rate is
wj. They demand xijs units of product s produced in country i, of quality qijs.
2.3.1 TBT and SPS measures
One way to model the fact that a product is safe for consumption is to estab-
lish market-specific minimum quality requirements. Figure 2.2 shows the quality
adjusted consumers? demand q? x in terms of product quality q. The zero demand
5As in [Costinot, 2008], quality standards are implemented to protect consumers against harm-
ful products. This type of standards help reduce the information cost for consumers who can
appraise and compare different product qualities satisfying different standards. Another charac-
teristic of standards that I do not consider in this chapter is their network effect of making products
compatible with each other. [Gandal and Shy, 2001] study effect of mutual recognition when the
network effects of standards are at play, i.e. when the value of a product increases with the number
of consumers that use compatible products.
33
to the left of q illustrates the fact that a product can only be sold in a market if it
satisfies the minimum quality requirements imposed by the authorities, otherwise it
is not authorized for sale in the market. Consumers? love for quality explains the
increasing relationship between quality and quality-adjusted demand.
On the supply side, the monopolistic firms optimally choose the quality of
their production accounting for consumers? quality-adjusted demand. The central
modeling feature is that firms can improve the quality of their production by invest-
ing into fixed costs. The quality of product s manufactured and sold in the same
domestic market i is described by the functional form:
q = f?siis iis (2.1)
The parameter ?s is the technology of quality production and indicates how much
quality improves after investing in one more unit of fixed costs. It is also tied to
the product-level perception of quality. For instance, quality is not perceived and
valued the same way for wines versus cars.
For each product s, there is a minimum quality requirement that operates as
a selection mechanism. Firms? optimal quality must satisfy the domestic minimum
quality requirement qiis > q or equivalently the corresponding minimum fixed costis
threshold fiis > f .iis
Quality requirements are market specific, as countries may not enforce the
same regulations. Figure 2.3 shows a scenario where country i has adopted more
stringent regulations than country j. A firm located in country i perceives q as a
i
34
domestic regulation and q as a foreign regulation. Without agreement on TBT/SPS
j
measures, imported goods may be subject to duplicated minimum quality require-
ments, adding to the fixed costs required to comply with each destination market?s
regulations. In the example of figure 2.3, even though quality requirements in mar-
ket i are higher than that of market j, a firm located in i is still required to spend
1
at least q ?sj in additional fixed costs to be able to sell in market j.
The quality equation for imported products is:
q = [(f )?ijsijs iis (f
1??ijs ?s
ijs) ] , where 0 ? ?ijs ? 1 (2.2)
In which the policy parameter ?ijs represents the degree of integration between
country j?s and country i?s regulations for product s. If ?ijs = 0, countries i and
j do not recognize/harmonize their regulations, whereas if ?ijs = 1 they fully rec-
ognize/harmonize their regulations. Notice here that it is not necessary to have
different minimum quality requirements across countries to create distortion. The
fact that countries might not recognize compliance with foreign regulations will
force firms to further spend on fijs to establish the quality of their exported pro-
duction. For sake of simplicity, I derive equilibrium solutions for three cases, the no
recognition/harmonization case, superscripted with NR, and two closely related in-
tegration cases: full mutual recognition MR and harmonization HR which differ by
the minimum quality constraint q ultimately enforced by the PTA. Under mutual
s
recognition, one can argue that countries would stick to their domestic regulations
in order to save costs from adaption. Under harmonization, the minimum quality
35
is expected to change to the most stringent direction denoted qHR, at least more
s
stringent for the countries with the lowest minimum standards.
To summarize, the corresponding functional forms under no recognition and
full integration is:
??????(f
?s
ijs) if qijs ? q
NR ? jsno recognition qijs = ???0 else????(f ?siis) if qijs ? qis
mutual recognition qMRijs = ???? ?i, j (2.3)? 0 else???? ?s?(fiis) if qijs ? q
HR
s
harmonization qHRijs = ???0 else
Markets are segmented under no recognition (NR), which implies that firms make
independent decisions regarding quality for each of the markets they can export to.
They will pick an optimal domestic quality investment fiis and as many bilateral
fixed cost investments fijs as there are exporting markets. However, under full in-
tegration (MR or HR) markets are not separable anymore. Consequently, firms
make one single optimal joint decision regarding the quality of their product. They
only pick fiis, which will be recognized by any country j sharing TBT/SPS provi-
sions with country i. The distinction between separable markets and joint markets
is the central mechanism of behind-the-border mechanism entailed by TBT/SPS
integration.
36
2.3.2 Consumer?s problem
The representative consumer?s problem is a quality-augmented version of the
consumer?s problem in the Melitz model. Preferences are represented by a Dixit-
Stiglitz utility function, in which the demand for a given variety xijs(?) is shifted by
a quality multiplier qijs(?). The consumer?s income Ej is the sum of labor income
wjLj and firms? profit ?j. The representative consumer?s maximization problem
6 in
country j is:
?(?? ) ?sS N ? ??s 1 js?s?1
max Uj = (qijs(?)x ?ijs(?)) s d? (2.4)
{xijs(?)} ?s=1 i=1 ???isS ?N ? ?S
s.t. pijs(?)xijs(?)d? = ?jsEj (2.5)
s=1 i=1 ???is s=1
Where xijs(?) is consumer j?s demand for the variety ? of a product s that is pro-
duced in country i, which is associated with price pijs(?) and quality qijs(?), and ?is
the set of varieties. Finally, ?s > 1 is the elasticity of substitution between varieties
consumed and ?js is the share of income Ej that country j spends on product s.
Solving the consumer?s maximization problem gives the usual CES demand function
for differentiated goods adjusted for quality. At identical prices, consumers demand
6The present model does not incorporate negative externality from consuming very low quality
goods, but actually focuses on the government?s response to solve it. Practically, a negative
consumption externality (modeled as an extra term in the consumer?s utility function equal to -?
if the aggregate consumption is of quality below some minimum standard and 0 otherwise) would
encourage the benevolent government to restrict quality to a certain minimum threshold. The
government would still choose to restrict quality as a second best for any other levels of negative
externality.
37
more products of higher quality.
xijs(?) = p (?)
??sq (?)?s?1P? ?s?1ijs ijs js ?jsEj (2.6)
For which I define P?js, the quality-adjusted price index in country j for product s:
[? ] 1N ? ( )?(?s?1) 1??spijs(?)
P?js = d? (2.7)
??? qi=1 is ijs(?)
2.3.3 The firm?s profit maximization problem
As usual in a Melitz-like heterogeneous firms model, firms differ in productivity
and each of them monopolistically supplies a single variety ? of a differentiated
good. Ex-ante there is a large pool of identical firms. In addition to the fixed costs
related to the quality regulations introduced in the previous section, firms producing
product s must pay fixed entry cost f es to enter their domestic market.
Upon e(ntry), each firm draws its productivity ? from a Pareto distribution?s
Gis(?) = 1? bis , where bis is the minimum productivity and ?s is the productiv-?
ity dispersion. The corresponding production function is therefore y(?) = ?(?)`(?),
where `(?) is the labor input. Firms may decide to export, in which case they have
to pay variable iceberg trade costs ?ijs ? 1. A firm must ship ?ijsxijs units of goods
in order to sell xijs units of that good to country j. I set ?iis = 1 so that variable
international trade costs are expressed relative to domestic trade costs and payable
in domestic labor units. Importing country j may levy tariff tijs on imports from
country i. In addition to the iceberg costs, tariffs create a wedge between factory
38
and consumer prices pijs = tijs?ijspis. Moreover, exporting firms face other fixed
operating costs Fijs related to doing business in country j, such as market research,
distribution and marketing costs. Lastly, TBT/SPS domestic regulations require
firms to fulfill minimum quality requirements. In this extended Melitz model, firms
not only decide whether or not to export and at which price, but also choose the
quality of their product to attract more demand. They improve quality by mak-
ing fixed cost investments as described by equation (2.2). All fixed costs ? entry
costs f eis, regulatory costs fiis and fijs, and distribution costs Fijs ? are payable in
domestic labor unit wi.
Firms optimally choose their export destinations, corresponding prices and
investments in quality to maximize their total profits from all domestic and foreign
sales. Total profits of a firm from country i producing variety ? and selling to a
certain number of destinations k are as follows:
?is(?) = ?pis(?)xiis(?)? wi`ii?s?(?)? wifiis ? wiFii?s
? ?domesticis (?)piks(?)xiks(?)
+ ? ? wi?`?iks(?)? wifiks ? wiFikstiksk=6 i ?
?foreignis (?)
xiis(?) ?iksxiks(?)
s.t. `iis(?) = and `iks(?) =
?(?) ?(?)
Without loss of generality and because of CES preferences, I can separate the price
optimization from the destination and quality optimization. Under monopolistic
39
competition, a firm of productivity ? sets the usual price with constant mark-up:
?s tijs?ijswi
pijs(?) = (2.8)
?s ? 1 ?(?)
Given this price, the firm chooses the optimal qualities and destination markets
for its goods sold domestically or intended for export. The timing for export decision
and optimal quality choice is as follows: first, the firm selects potential destination
markets; and second, given the optimal quality associated to those destinations it
verifies whether its total revenues can cover the fixed investment in exported quality
f and the usual fixed exporting costs F .
Moreover, the way firms pick their optimal quality and destination markets
depends on the state of recognition of TBT/SPS measures. When countries do not
recognize each other?s regulations (NR), markets are segmented and firms pick a
set of market-specific qualities solely based on the sales in each market. Market
selection operates as in Melitz through the condition of positive profits, and fixed
exporting costs F for one market do not influence the choice to export to other
markets.
However, when countries fully integrate their regulations (MR or HR), firms
make a unique joint decision on their optimal quality, based on their total sales
from all markets. Therefore, it is important to think about quality optimization in
terms of total profits defined by equation (2.8). The decision to export to several
markets ultimately impacts the optimal quality, but also the total amount of fixed
costs (from investment in quality and exports). Depending on the size of the other
40
fixed costs to export F , a firm may decide to drop one destination market and
choose a lower quality. The sequence of decisions for the joint market scenario can
be implemented backward: i) the firm evaluates what would be its optimal quality if
it were to export to k markets, ii) if the total revenues from exports cannot cover the
total fixed costs to exports and fixed investment in quality, then the firm can drop
a market and choose a lower quality optimized on k ? 1 markets. For k markets,
there are k! possible combinations of destination choices. This cannot be solved
with algebra because revenues depend on moving endogenous variables and cannot
easily be compared to fixed export costs. Nevertheless, in what follows I describe
the optimal choice of quality for k markets and argue in the next section that the
exporting thresholds can be determined by the most constrained firms (- i.e. for
which the minimum quality constraint binds).
Appendix B details the methodology to solve the following firm?s profit max-
imization problem under non-recognition (NR):
? ( )k 1??? E ? t ? ?kj j ij ijwi ?(??1)
max?(fi1, fi2, ..., fik) = fij ? wi (fij + F? ij
)
? ? 1 ?P?
j=1 j j=1
(2.9)
subject to
?j ? (1, ..., k), fij > f (2.10)j
41
The resulting optimal interior solutions for qualities are given by:
?? ( ) ?? ?s1??s(?s?1)1??s
opt,NR ??? (tjis?ijswi) ?jsEj ?s?1?qijs = ?s ? 1??? ?? ? ?? (2.11)witijs P? sjs
A
( )1??s
where ?s =
?s(?s?1) ?s and corner solutions qopt,NR? ijs = q . Not surprisingly,?s ?s 1 js
quality increases with firm?s productivity and the perception of quality ?s within
a product category. The term A represents the role played by market size in the
optimal quality decision. Larger markets are associated with higher incentive for
quality improvement, as firms will be able to reap the costs from exporting to new
markets by increasing their supplied quality and attracting more demand.
On the other hand, if countries fully recognize each other?s TBT/SPS measures
(MR or HR), firms only have to pick their domestic quality, which will be recognized
everywhere there is a mutual recognition agreement. Consequently, this TBT/SPS
agreement provides incentives to invest into higher quality production, since firms
expect to recover their fixed costs from higher sales abroad. Appendix B shows that
if firms choose to export to the whole PTA market, the resulting optimal interior
solution for quality under full integration (MR or HR) is given by:
? ? ( ) ? ?s1??s(?s?1)
opt,MR/HR ? ?jsEjq = ? ??s?1?is s ? (2.12)1 ?s
{j?PTA } P?jsijs
and corner solution qopt,MRis = q . Here we observe the impact of market size en-s
largement on quality, which is the first of the two effects implied by TBT/SPS
42
provisions. Integrated markets offer firms a broader range of export destinations.
If other export fixed costs are not too large to prevent firms from reaching new
destinations, (2.12) shows how increased market access affects quality. This ex-
pression shows that mutual recognition agreements provide incentives to invest into
higher quality production since firms can expect to recover their fixed export costs
from higher sales abroad. For all other countries outside the PTA with TBT/SPS
provisions, the firms will pick qopt,NRijs as defined by expression (2.11).
Finally, the expression of optimal quality for non-exporters is similar under no
integration or full integration:
[ ] ?s
opt ?isE 1??s(?i s?1)qiis = ?
?s?1
s ? ? (2.13)w?sP? 1 ?si is
TBT/SPS provisions also affect non-exporters through general equilibrium changes
in price index and wages. Without solving the general equilibrium problem, it is
not clear how change in competition will affect non-exporters.
2.3.4 Productivity thresholds
The constrained maximization problem used to derive the above solution for
optimal quality ensures that minimum quality requirements are satisfied for both
interior and corner solutions. The productivity threshold to export is obtained from
finding the minimum ? for which profits are positive.
? Under no recognition (NR), each destination market has a separate produc-
tivity threshold:
43
( ) 1
1??s(?s?1)
?,NR ? F
?s ?s(?s?1)
?s(?s 1) ijs
?1
? ?s?1ijs = qjs 1 + [?ijs] (2.14)q
js
? E
where I denote ? js jijs = ?s w?s P? 1??si is
? Under mutual recognition or harmonization, as explained in appendix B, firms
can always opt out of an export market and adjust their optimal quality accord-
ingly, thus lowering the fixed costs to export (from both TBT/SPS measures
and Melitz-like costs). Ultimately the minimum threshold is driven down to
the corner solution threshold that cannot be lowered further.
( ? ) 1 ? ? ?1? ? ?1s ?s(?s?1) s
?,MR/HR 1??s(?s?1) { ? ?j PTA F
? = q ijs
} ijs
?s(?s?1)
is 1 + ? ? ?s ijsq
s ? {j?PTAij?s}? ?
A
(2.15)
The difference between MR and HR lies in the minimum quality regulation
ultimately enforced by the PTA, q for MR and qHR for HR. Notice that
is s
market size also has an impact of the export threshold through the term A.
2.4 Aggregated quality
In this section, I derive the expression of the aggregated quality?of imported
goods. Let us define the aggregated quality of imported goods by q? = qoptijs ?is??is ijs (?)d?.
The integral can be decomposed between the mass of firms for which the minimum
quality constraint binds (qopt = q) and those that are not constrained and choose
an interior solution defined by equations (2.11) or (2.12).
The integral over the constrained firms with probability P(? = ??) is zero,
44
because the Pareto distribution is continuous at ??. The aggregated quality is
therefore integrated over the interior solutions and becomes:
? ? ? ? ( )?? ?sopt ? opt ijsq?ijs = Mijs qijs (?)gis(?|? > ?ijs)d? = Mijs qijs (?) d?
?? ?? ?
?s ?s ? ?s(?s?1)= M [? ] 1??s(?s?1) (? ) 1??s(?s?1)? ?s(?s?1) ijs ijs ijs1? ? ?? (? ???1) ss s ?
?s
where Mijs represent the number of exporters from i to j, which is also equivalent to
? E
the number of varieties, and ? js jijs = ?s w?s P? 1??s . Next, I substitute in the equationsi is
of the productivity thresholds (2.14) and (2.15) to obtain the simple expressions:
( ) 1
F?s ?s(?s?1)ijs
q?NRijs = ?sb
?s
isMijsq 1 +js( ?qjs ) 1?s ?s(?s?1)
MR/HR ?s {j?PTA }
Fijs
q?ijs = ?sbisM
ijs
ijsq 1 +s q
s
As expected, average quality is positively correlated with minimum quality require-
ments q, but also depends on the level of competition through the term Mijs corre-
sponding to the number of exporters. Therefore, there is both a regulatory channel
and a market size channel behind the average aggregate level of quality. In the next
chapter, I empirically investigate which of the two channels prevails in the overall
impact of a PTA including TBT/SPS-related provisions.
I rewrite the above equations in terms of variables that will be easier to control
for in the econometric specification. Let us recall that the number of exporting
45
firms from i to j, Mijs, can be expressed probabilistically as a share of M
e
is, the
firms originally entering market i: Mijs = prob(? > ?
?
ijs)M
e
is = [1?Gis(??ijs)]M eis =
( bis )?s? M eis. Plugging in this expression in the above quality expressions gives:?ijs
( ) 1
F?s ?s(?s?1)?
NR s
1??s(?s?1) ijs
q?ijs = ?sb
?sM e [? ? ? 1??1 ? (? ?1)is is ijs] s (q ) s s 1 + (2.16)js q
js
[? ] ?s ( ? ) 1?s?1 F?s ?s(?s?1)
MR/HR ? 1??s(?s?1)?s e 1 ? {j?PTA } ijsq? = ? b M ? (q ) ?s(?s ijs s?1)ijs is is 1 +
ijs
s q
j s
( ) (2.17)1??
where ? = ?s(?s?1)
s
?s tijs?ijswi ?jsEj
ijs ? . Some intuitive remarks can be made?s ?s 1 P?js witijs
by comparing (2.16) and (2.17). Since ?s > 1, TBT/S?PS agreements increase qual-
ity through access to a larger market (from Ej to Ej). Tariffs t and iceberg
trade costs ? depress the quality of imports. The overall productivity level in the
exporting country bis is positively related to quality. Increases in the regulatory
parameter q can actually increase or decrease the average level of quality. Indeed,
to this point I have set the ex-post minimum quality requirement after mutual
recognition of TBT/SPS measures to the one implemented in the domestic country.
However, one could think about a situation in which exporting firms can choose
where to obtain their certifications and will eventually get certified in the country
where it is the cheapest. Also, countries might agree to harmonize to more stringent
TBT/SPS measures, committing for example to international standards. The reg-
ulatory channel will depend on whether the agreement calls for mutual recognition
46
versus harmonization.
2.5 Conclusion
The model developed in the present chapter can serve different empirical and
quantitative purposes. The framework reveals how the economic structure of the
model is changed by PTAs including TBT/SPS provisions. Fully solving the model
cannot be done analytically, but could be approached with quantitative methods in
further work. In the next chapter, I show how empirical analysis of the impact of
TBT/SPS provisions on quality can shed light on the market size and regulatory
channels described in the present theoretical chapter.
47
2.6 Figures
1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015
year
New PTA signed w/ sps or tbt provision New PTA signed w/o sps or tbt provision
stock of PTAs stock of PTAs w/ sps or tbt provisions
Source: World Bank?s Content of PTAs database version 1.0
Figure 2.1: Incidence of TBT and SPS chapters in PTAs
48
New PTAs signed
0 5 10 15 20
0 100 200 300
Stock of PTAs
Figure 2.2: Minimum quality requirements
Figure 2.3: Domestic and foreign minimum quality standards
49
2.7 Tables
(a) Most important perceived NTMs from the EU to the US
1 Restrictions or bans on use of specific chemicals
2 Classification and labeling requirements for chemical products
3 Threat of 100% container scanning
4 FDA New Drug Approval Process
5 Drug precursor legislation
6 Double certification need caused by the European Union Authorized Economic Op-
erator program and the US Customs CTPAT program
7 US state level safety certifications
8 Prior authorization for sensitive product categories
9 Pre-shipment inspections
10 Non recognition of ?Made in EU?
(b) Most important perceived NTMs from the US to the EU
1 EU pricing policy member state differences
2 Health technology assessment differences
3 Different rules in various Member States concerning authorization of pharmaceuti-
cals
4 International reference pricing
5 Therapeutic reference pricing
6 Differences in the enforcement of the unified customs system across EU member
states
7 Parallel trade allowance
8 Restrictions concerning information distribution to patients
9 New EU Member States compliance with undisclosed data protection
10 Prohibitions (e.g. security, sensitive products, political reasons, etc.)
Source: [Ecorys, 2009]
Table 2.1: Examples of most important trade NTMs perceived by businesses in the
pharmaceutical industry
50
Types of concerns raised Stated objectives for the mea-
sures raised as STCs
further information, clarification 387 Protection of Human health or 277
safety
unnecessary barrier to trade 347 Protection of the environment 137
transparency 319 Other 108
other issues raised (free text) 263 Prevention of deceptive practices 108
and consumer pro
rationale, legitimacy 235 Consumer information, Labelling 101
international standards 228 Not specified 82
discrimination 177 Quality requirements 47
time to adapt, ?reasonable inter- 142 Harmonization 26
val?
non-product related processes 43 National security requirements 19
and procedural methods
special and differential treatment 23 Protection of animal or plant life 16
or health
technical assistance 8 Reducing trade barriers and facil- 0
itating trade
Cost saving and productivity en- 0
hancement
Table 2.2: Number of TBT concerns by type and objectives
51
Chapter 3: Impact of deep PTAs on quality
3.1 Introduction
The role of quality as a determinant of trade patterns has been studied in the
theoretical literature since [Linder, 1961], who argued that rich countries produce
and consume higher quality goods than developing countries. His work led to the
extensively tested prediction that countries of similar income per capita tend to
trade more with each other, also known as the Linder hypothesis. Subsequent
theoretical and empirical work has demonstrated the existence of both supply-side
and demand-side relationships between countries? development level and the quality
of traded goods.
The role of quality operating on the demand side is supported by the empirical
work of [Hallak, 2006] who finds that rich countries tend to import more from
counterparts producing goods of higher unit values. [Schott, 2004] and [Hummels
and Klenow, 2005] provide evidence for the role of quality on the supply side, showing
that quality (also measured with unit values) increases with exporters? income.
Beyond its role in determining trade patterns, quality upgrading has been
linked to exporters? performance, adoption of new technologies, and to the rise
52
of income inequality in developing countries.1 From a consumer or environmental
welfare point of view, quality upgrading is beneficial for health and safety. The
fundamental questions addressed by these studies include why product quality varies
across countries and over time, and how it can be influenced by trade policy.
In the previous chapter, I highlighted the existence of two possible channels for
quality adjustment following a TBT/SPS integration, the market size and regulatory
channels. In this chapter, I rely on the theoretical model of chapter 2 to empirically
estimate the impact of deep trade agreements on the quality of traded goods.
After deriving a measure of imported good quality at the industry level, I
investigate the change in quality associated with PTAs that include TBT/SPS-
related issues. Using the second version of the Content of PTAs dataset, I shed light
on the role of TBT/SPS chapters in changing the quality of imported goods. I study
the impact of different provisions that can be included in PTAs: mutual recognition
and harmonization of standards, technical regulations, and conformity assessments
of TBT measures; and mutual recognition and harmonization of standards, and
inspection provisions of SPS measures. In each econometric specification, I control
for the rest of PTAs content with a measure of total depth. Moreover, the literature
shows that rich countries tend to trade products of higher quality. But rich countries
are also most likely to sign deeper PTAs that include TBT/SPS provisions, which
raises concerns on the endogeneity of such policies. I address this issue by controlling
1 [Verhoogen, Eric A., 2008] finds that Mexican firms, which employed higher skilled workers
and produced higher quality goods, performed better during the 1994 Peso crisis. [Bustos, 2011]
sheds light on firms? incentive to innovate (and hence improve quality) following trade liberalization.
[Goldberg and Pavcnik, 2007] shows that because of the changes in relative demand for skilled labor,
quality improvement prompted by the demand from richer countries has been a determinant in
the increase of income inequality in developing countries.
53
for country-pair fixed effects.
The empirical analysis shows that including TBT provisions in a PTA increases
the quality of imports, with effects twice as large as when standards are mutually
recognized rather than harmonized. Inclusion of SPS provisions also increases the
quality of imports, but the distinction between mutual recognition and harmoniza-
tion is less clear. The country-pair effects account for a large share of these impacts,
which hints at the presence of policy endogeneity. Results are driven by PTAs be-
tween high- and low-income countries, supporting the importance for developing
countries to tackle behind-the-border issues in their trade agreements. I also find
that the estimated positive impacts of TBT/SPS provisions on quality are driven by
products with a wider scope for quality improvement, such as differentiated goods.
3.2 Literature on quality estimation
Unless one is interested in studying a specific product for which quality at-
tributes may be observable, comparable product qualities for a wide range of indus-
tries are usually difficult to measure directly. Example of quality-related attributes
that can be observed are the efficiency of solar panels, i.e. how effectively they can
convert solar radiation into electricity, or the sharpness of cameras? lenses, as mea-
sured by the index of Perceptual Megapixel. [Crozet et al., 2011] use wine terroir
scores available for each producer in France as a proxy for wine quality and derive
stylized facts regarding French exporters? performance in the wine sector. Unfor-
tunately, finding consistent attributes for a large range of sectors and/or countries
54
is unfortunately not feasible. Nevertheless, a good measure of quality is necessary
to feed the analysis of the present chapter. This section provides an overview of
the techniques employed in the literature to measure quality for larger samples of
products and countries and highlights their data requirements.
The common ground of all existing techniques is that quality is expensive to
produce and higher costs are eventually passed through sale prices. Therefore, a
natural attempt to measure quality is to resort to unit prices by assuming that
higher sale prices reflect higher quality. At both the country and firm level, this
solution is convenient because it only requires trade values and quantities. [Schott,
2004] and [Baldwin and Harrigan, 2011] compute unit values from product-level US
import data and [Hummels and Klenow, 2005] use cross-section bilateral trade at
the 6-digit level. These studies support the intuition that unit value increases with
the GDP per capita of exporters and importers, and shed light on the role of quality
in determining aggregate trade patterns. [Johnson, 2012] studies the link between
export prices and price thresholds to access different markets. He uses cross-section
bilateral trade data from the CEPII?s BACI database and finds that higher trade
unit values are associated with destination markets with higher entry thresholds.
Unit values computed with firm-level data allow researchers to test predictions
from micro-founded theories. Using Chinese firm-level data, [Manova and Zhang,
2012] show that firms charging higher unit values export to more destinations. With
Colombian data, [Kugler and Verhoogen, 2011] find that the best performing firms
tend to use more expensive inputs and export higher unit value outputs. Also with
Colombian data, [Brooks, 2006] highlights the role of product quality in explain-
55
ing why Colombian firms are found to under-export manufactured goods to the
US. [Hallak and Sivadasan, 2013] confirm the positive relationship between export
performance and output quality by using data from India, US, Chile and Colombia.
The major drawback of using export prices to infer quality is that unit val-
ues are also likely to reflect the cost of factors of production, which weakens the
identification of quality from the variation of unit prices. To address this issue,
researchers have developed ways to indirectly estimate quality, defined as a vertical
characteristic that shifts consumers? utilities.2 Many studies build on the straightfor-
ward intuition that after controlling for prices and horizontal characteristics, higher
quality products should drive higher demand.
The estimation of import demand functions to derive quality has become stan-
dard in the most recent literature. In practice, import demand equations are es-
timated while accounting for the endogeneity of prices. Then the resulting price
elasticities are used to compute quality from the relevant fixed effects and/or the
demand residuals. The main challenge is to obtain consistent estimates of price
elasticities. Researchers have turned towards instrumental variables strategies, us-
ing exchange rate related measures to instrument for prices. The variation used
to identify demand parameters ultimately depends on the detail level of the data
employed for the estimation. With product-level US data, [Khandelwal, 2010] es-
timates import shares controlling for unit prices and sectoral varieties to derive
quality estimates based on the demand residuals. [Hallak and Schott, 2011] obtain
2The taste for quality is not to be confounded with the horizontal love of variety, which justifies
why low quality products are still traded along higher quality products of same price.
56
quality estimates from the trade balances also conditional on export prices. On the
firm-level side, [Khandelwal et al., 2013] use Chinese customs data, [Gervais, 2013]
uses data from the US Census of Manufactures, and [Piveteau and Smagghue, 2019]
use French customs data.
Overall, the scope and detail of the data, as well as the selected estimation
strategy, ultimately depend on the research questions to be addressed. In this
chapter, I am interested in evaluating the heterogenous impact of the content of
TBT/SPS measures included in PTAs. Since I am studying the whole realm of
PTAs enforced and notified to the WTO over a long time span, I need to derive
time-varying quality measures for the largest set of countries possible. The level of
product disaggregation will consequently be pinned down by the data availability.
I choose to follow closely the methodology developed by [Khandelwal, 2010] after
adjusting his demand equation with the one obtained from the preceding chapter.
3.3 Estimation of quality
In this section, I describe the methodology implemented to estimate the qual-
ity of imported goods at the origin-destination-product-year level. I follow the lit-
erature that estimates quality using the demand-side variation of trade, based on
the following straightforward assumption: conditional on prices, higher demand for
imports should reflect higher quality.
57
3.3.1 Demand for imports
Drawing on the model developed in chapter 2, I derive the aggregate demand
for imports from the CES consumers? deman?d function (2.6) over all firms by using?
the Pareto distribution. I denote Z? = Mijs ? Zd? the aggregated function of any?
variable Z that depends on the productivity level?. Fo(r a)ny constant H that does? ? ? ??
not depends on ?, the aggregation falls to M H ? d? = M H ijsijst ?? .? ijst ?s?1
The aggregated log import demand of country j, for product s, from origin i,
at time t is written as follows:
ln?xijst = ???sl?n??pijs?t + ln ?M?i?js? ?P?? ?s?1t jst ??? ?ijs ?jsEj?t ? ?? ?+ ?(?s ? 1??)ln qijs?t (3.1)?s ? 1
(1) (2) (3) (5)
(4)
Where (1) is the unit price, whose elasticity is ?s. The unit price is proxied by trade
unit value obtained by dividing trade value over quantity at the HS 6-digit level,
which is the most disaggregated level available for a large sample of countries; (2)
represents the number of varieties within a product class s. This variety variable
reflects the horizontal taste of consumers for a large choice of products. For sake
of simplicity, I follow the literature and approximate this variable by either the
population or GDP of the exporting country; (3) is composed of the quality-adjusted
price index and aggregated income in the destination country. It does not depend on
the origin country, so it can be accounted for by the use of product-destination-time
(or variety-time) fixed effects; (4) will be accounted for by destination-sector-time
and country-pair-sector fixed effects; lastly (5), conditional on prices and varieties,
58
the remaining term represents the quality of imported goods and can be retrieved
from the residual demand denoted ijst given the estimates of ??s. The instrumental
variables would allow us to estimate the elasticities of substitution; alternatively
these parameters can be conveniently taken from [Broda and Weinstein, 2006] at
the HS 6-digit level.3 I estimate the demand curve (3.1) for each HS 2-digit industry,
and retrieve quality from the estimated residuals at the HS 6-digit level. Finally,
I compute a trade weighted measure of quality at the 2-digit industry level using
import shares as weights.
3.3.2 Dealing with price endogeneity
As explained in the literature review section, the main challenge entailed by
the estimation of the import demand equation is the endogeneity of prices that
might come from both simultaneity and measurement error problems. Indeed, since
quality is costly to produce, the additional production costs dedicated to quality
improvement are expected to be passed through to prices. Thus, unit values are
likely correlated with the residual term, raising simultaneity concerns. Also, given
that unit value is constructed as the ratio of trade flows over quantity, it is highly
dependent on the level of aggregation and product classification, which makes it
prone to measurement errors. Both sources of endogeneity are expected to bias
downward the estimates of ?s.
I do not identify ?s in my estimation of import demand, instead I use the price
3Doing so includes additional noises to the estimates and should be addressed in further revi-
sions.
59
elasticities that have been consistently identified by [Broda and Weinstein, 2006].
However, I still include an instrumental variable in my estimation of equation (3.1)
to test for over-identification.4 This instrument is constructed from the interaction
of the bilateral nominal exchange rate (varying at the country-pair-time level) and
a proxy for transportation costs (varying at the industry level):
eLUCit/$
XrateTransportIVijst = ? CIFs/FOBs (3.2)
eLUCjt/$
Where eLUC are the nominal exchange rates of country k?s local currency againstkt/$
the USD, CIF is the cost, insurance and freight price, and FOB is the free on board
price averaged at the HS 2-digit industry level.
The use of an instrument derived from exchange rates is frequent in the lit-
erature that estimates quality (see [Khandelwal, 2010], [Hallak and Schott, 2011],
and [Piveteau and Smagghue, 2019]). As pointed out by [Piveteau and Smagghue,
2019], this choice may be questionable if one believes that goods imported from
countries with stronger currencies have higher quality. There would be a positive
relationship between quality and the exchange rate at the cross-section level. How-
ever, at least in the short/medium run, firms are not expected to quickly react to
changes in exchange rates by upgrading the quality of their exports following an
importer devaluation of an importers? currency. Therefore, despite existing cross-
section variations in the exchange rate pass-through into import prices across sectors,
product characteristics, or currencies of invoicing, time varying exchange rates seem
4If ?s from [Broda and Weinstein, 2006] are correctly identified, one should find a coefficient of
-1 for the variable unit price multiplied by sigmas.
60
to be a good candidate to instrument for prices in the import demand equation.
To add industry-level variation to the instrument, I then interact the exchange
rate with a proxy for transportation costs. Databases with large coverage of bilateral
trade flows usually do not provide widespread measures of bilateral transportation
costs.5 The usual way to circumvent this limitation is to use the mirror nature of
trade data. In a perfect world, each trade flow should be reported twice, once by
the importing economy and once by the exporting one. On the one hand, importers
declare CIF flows, and on the other hand exporters report FOB trade flows. There-
fore, the ratio CIF/FOB can give us a proxy for unit transportation/insurance costs
at the industry level. Transportation costs satisfy the relevance condition with no
objection, but there might be concerns about the exclusion restriction. Quality
might be correlated with distance (hence higher transportation costs) as supported
by [Hummels and Skiba, 2004] who investigate the ?Washington Apples? effect in
which higher quality goods may be exported to more distant countries in order to
lower the unit transportation costs.
To address this concern, [Khandelwal, 2010] argues that ?as long as trans-
portation costs do not affect deviations from average quality the instrument remains
valid? because the Washington Apples effect is identified using cross-section vari-
ations rather than variation in transportation costs over time. To summarize, the
estimation of quality relies on the following specification, estimated with OLS for
5Unlike the US official trade data which provides transportation cost at the most disaggregated
level, a measure used by [Khandelwal, 2010]
61
each of 94 HS 2-digit industries:
ln?x =? ln?ijst 1 pijst + ?2 ln popit + ?jst + ?ijs + ijst
ln pijst =?XrateTransportIVijst + ajst + aijs + ?ijst (3.3)
? ?? ijstln qijst =
?s ? 1
Results of the quality estimations are reported in the subsequent sections.
3.3.3 Impact of TBT/SPS provisions on quality
I use the quality estimates from (3.3) to assess the impact of TBT/SPS provi-
sions included in PTAs on imported good quality. Below, I recall the equations for
aggregated quality derived in the second chapter (2.16) and (2.17):
( ) 1
F?s ?s(?s?1)?s 1??s(?s?1) ijs
q?NR ?s e ? 1??? 1 s ? (? ?1)ijs =?sb s s sis ?Mis [??i?js] ? (?q ) ?? ? ? 1 +js qjsA[ NR B ?? ?NR? ] CNR?s ( ? ) 1?s?1 F?s ?s(?s?1)1??
? s
(?s?1)
MR s e ? {j?PTAq? =? b M ? (q )1???s? ? ? 1 + ijs} ijs?s(?s 1)ijs s is ? is ijs is qj ?? ? ? sBMR/HR ?? ?
A CMR/HR MR/HR
( )1??s
Where ? ?s(?s?1) ?s
tijs?ijswi ?jsEj
ijs = ? . The change in the functional form?s ?s 1 P?js witijs
of quality moving from q?NRijs to q?
MR
ijs sheds light on the two main motives for quality
improvement. I refer to the change from A to AMR/HR as the market size channel.
It shows that the reduction of trade barriers enabled by the enforcement of a PTA
implies an increase in market access (sum of ??s), which triggers competitively mo-
62
tivated quality improvements to capture bigger market shares abroad. The number
of domestic firms M eis will eventually change due to the general equilibrium effects,
which increases competition at home too. Together these two terms only depend on
the origin country j and can be taken care of with destination-time fixed effects.
On the other hand, the change from B to BMR/HR corresponds to the regula-
tory channel. In an integration associated with mutual recognition, the change in
the ex-post minimum quality requirements (induced by the option to seek the least
expensive TBT/SPS measures) explains how TBT/SPS-related chapters in PTAs
can impact quality. Basically, mutual recognition allows firms to minimize the costs
related to TBT/SPS measures either by saving on redundant foreign costs, or by
getting access to cheaper requirements. The new generation of PTAs has often faced
strong backlash from public opinion, which fears that integration to larger markets
might level down standards and undermine countries? sovereignty in setting their
domestic legislation. In response to voters? pressure (and probably to lobbies too),
we increasingly observe provisions granting exceptions to sensitive sectors in the
content of PTAs. Nevertheless, when integration occurs between parties at different
levels of development, the least developed partners are virtually always required to
raise their standards, sometimes in the context of a pre-accession agreement (as is
the case for each EU enlargement).
The above partial equilibrium equations can also help predict the effect of
harmonization on quality. Policy makers claim that the harmonization approach to
integrating TBT/SPS measures is the most conservative one, as regulations should
be raised to the most stringent level possible. Under that condition, the regulation
63
channel should have a positive direct impact on quality by raising q, at least for the
economy that must adjust its regulation to be in line with the higher standards.
Finally the change from C to CMR/HR combines the two channels above but will be
captured by country-pair and sector fixed effects.
To test these predictions, I first estimate the following baseline equation with
OLS:
ln Q?ijst = ?1Provision
TBTSPS
ijt + ?Depthijt + ait + ajt + aij + ast + ?ijst (3.4)
The way I code the content of PTAs is explained in detail in the next section. In
a nutshell, the independent variables included in (3.4) describe the extent to which
a PTA addresses various policy areas, and in particular TBT/SPS-related policies.
The main variable of interest is ProvisionTBTSPSijt , which is a dummy that indicates
whether and how TBT or SPS measures are addressed when a PTA is signed between
countries i and j. The list of provisions is indicated in Tables (A.3) and (A.4). If
the provision is addressed in a PTA, the variable turns on and stays at one from
the date of enforcement. The variable Depthijt is defined as the sum of 50 dummy
variables that indicate the presence of the main policy areas in PTAs other than
TBT or SPS-related areas (see Tables (A.1) and (A.2) for the other possible areas
included in PTAs). This allows me to control for the horizontal scope of trade
agreements in areas other than reduction of tariff barriers and TBT/SPS issues.
The directional fixed effects ait and ajt control for country-level determinants of
quality, such as market size. The pair fixed effects aij capture traditional gravity-
64
like variables (such as common language, distance and common borders) that may
affect the level of quality traded between a pair of countries. Finally, ast captures
the evolution of innovation within an industry s.
Next, I investigate heterogeneity of the results by interacting the TBT/SPS
provisions with first country-level and then sector-level variables. The model pre-
dicts that quality should change more if the prior difference between parties? regu-
lations is larger.
To test this hypothesis, I first use dummy variables indicating the difference
in development level between the parties. I create three variables identifying trade
relationships between Northern and Southern countries. The variables are denoted
NN when trade occurs between two high-income countries, NS for trade between
a high-income country and a low- or middle-income country, and SS for trade be-
tween two developing countries. The level of income is based on the World Bank?s
classification of development. The corresponding econometric specification is:
ln Q?ijst =?NNProvision
TBTSPS
ijt ?NNij
+? ProvisionTBTSPSNS ijt ?NSij
(3.5)
+?SSProvision
TBTSPS
ijt ?NSij
+?2Depthijt + ait + ajt + aij + ast + ?ijst
I use the intensity of ISO 9001 certification as a proxy for the overall level
of existing quality at the country-sector level. To account for the potential bias
induced by large differences in the number of firms across countries, I divide the
65
number of ISO certifications by each country?s total export.6 Finally, I compute
the intensity gap (or difference in intensity gap) between parties and denote it as
ISO9001ij . The resulting specification is:
ln Q?ijst =?1Provision
TBTSPS
ijt + ?2Provision
TBTSPS
ijt ? ISO9001ijs
(3.6)
+ ?3Depthijt + ait + ajt + aij + ast + ?ijst
Finally, I interact the TBT/SPS provision with a variety-level measure of prod-
uct differentiation (where a variety is defined as an origin-product pair). Products
differ in their level of complexity, which implies that the potential range of quality
variation might not be the same for homogenous versus highly differentiated prod-
ucts, or for countries in which the technology level is high or low. I motivate my
investigation of this type of heterogeneity with the work of [Khandelwal, 2010], who
defines a measure of scope for quality differentiation or ?quality ladders? at the
product level. In his paper, the ladders are defined as the range of quality within
HS 10-digit products and across all origin countries. I follow his method by testing
three measures of ladders: the full range, the inter-decile range, and the interquar-
tile range, whose the last two measures should be more robust to outliers than the
first measure. Since I am ultimately working with bilateral measures of quality, I
construct the ladders at the HS 6-digit level, my finest level of observation.
Ladder = pctiletop(qua?iht lityiht)? pctilebottom(qua?lityiht) (3.7)
6Dividing the ISO certifications by the number of firms in a country, drastically reduces the
number of observations.
66
where pctiletop is either max, 90th or 75th percentile, and pctilebottom is min, 10th or
25th percentiles. Then, I take the trade weighted average of Ladders to obtain an
industry-level aggregated measure of the quality range at the 2-digit level.
?
IndLadderist = wihtLadderiht (3.8)
h?s
Finally, I include the interaction term between TBT/SPS provisions and the indus-
try level ladders to obtain the following specification:
ln Q?ijst =?1Provision
TBTSPS
ijt + ?2Provision
TBTSPS
ijt ? IndLadderist
(3.9)
+ ?3Ladderist + ?4Depthijt + ait + ajt + aij + ast + ?ijst
I expect the coefficient ?2 to be positive, revealing that sectors with more scope for
quality improvement will experience a higher increase in quality following mutual
recognition or harmonization of TBT/SPS measures.
To understand the determinants of the quality ladders, I use the definition
of optimal quality (3.3) and derive an expression for the ladder?s lower bound. If
one assumes that firms from developed countries have access to better technology
than their developing country competitors (bNs > bSs from the Pareto minimum
productivity parameters), then I can compute log(qoptNjst)? log(q
opt
Sjst), where optimal
quality for the developed country firms is evaluated at the minimum productivity
bNs. It is reasonable to think that over the whole set of developed countries at
least one will be associated with a minimum productivity parameter that enables
firms to easily pass the tests for minimum quality certification. Conversely, over the
67
whole set of developing countries at least one will be associated with a minimum
productivity parameter that does not allow a firm to pass these tests. Therefore to
get the ladder?s lower bound, log(qoptSjst) is evaluated at the productivity threshold
?? that allows firms to fulfill minimum quality certifications.
[ ( )
log(qopt opt
?s tNjstwN
Njst)? log(qSjst) = ??? ? ( s
log
1 ?s(?s 1) ) tSjstwS ( )] (3.10)
? ?Njst+ (1 ?s) log + (?s ?
bNs
1) log
? ?Sjst ?Sjst
Equation (3.10) shows that the ladder intuitively lengthens with the level of technol-
ogy in the developed countries. All else equal (same wages, same tariffs and iceberg
trade costs) longer ladders are associated with higher ?s, or in other words with
more differentiated goods. Moreover, similar to [Khandelwal, 2010] the ladders can
be indexed by the valuation parameter ?s, so that industries with higher valuation
for quality certification are associated with longer ladders.
3.4 Data
In this section, I first describe the data employed to estimate quality and its
derived ladders. Then, I present in detail the novel data used to identify the change
in trade policies related to TBT and SPS measures. I complete the data section
with a presentation of the survey data on ISO certification and a brief description
of the Specific Trade Concerns data.
68
3.4.1 Data for the quality estimation
For the estimation of product quality, I use bilateral import data from the
Base pour l?Analyse du Commerce International (BACI) database developed by
CEPII, which provides import values and quantities at the 6-digit product level
between 1995 and 2015. The CEPII applies the procedure described in [Gaulier
and Zignago, 2010] to the underlying UN-Comtrade bilateral trade data in order
to reconcile the declarations to the United Nation Statistics Division of almost 200
exporting and importing countries. Normally, each flow should be reported twice to
the United Nations Statistics Division, once from the exporter and once from the
importer. The first exercise aims to verify the consistency of this mirror reporting.
Then the procedure draws on this same principle of mirror declarations of trade
flows to fill in information that is not reported (when only one of the two partners
reports a trade flow).
In a first step, the CEPII procedure transforms the reported CIF import values
as FOB, in order to match import with FOB export declarations. The transforma-
tion removes the transport costs. However, unlike the US customs data for example,
product level transportation costs are not reported. The transformation from CIF
to FOB needs to be estimated using a gravity-like model including the usual CEPII
gravity variables (share of border, landlocked country, same language, colonial ties),
year fixed effects and product level median unit-values. Then, the import-export
FOB mirror values are averaged at the product country-pair level. For obvious
capacity and technical reasons, reports from low-income countries such as Malawi
69
are not as reliable as those from the US. To account for the potential difference in
reporting accuracy, the CEPII applies a ?reliability? weight to compute the average
import-export FOB values. Those weights are based on the geographical and sec-
toral specialization of the reporting countries. Overall, this mirror data procedure
has two advantages, it allows CEPII to extend the coverage of the available data
and increases the reliability of the data reported by low income countries.
As explained in the previous section, the import demand equation is endoge-
nous in prices. To compute the instrumental variable for the unit values, I use the
CIF/FOB ratios to proxy for transportation costs. These ratios are also taken from
CEPII and benefit from the same methodology relying on mirror data to fill in miss-
ing prices. Finally, nominal exchange rates, GDP and population variables are taken
from the World Development Indicators dataset. The elasticities of substitution are
taken from [Broda and Weinstein, 2006].
3.4.2 Data on TBT/SPS chapter
I use the new dataset on the vertical Content of PTAs developed by the World
Bank to identify what kind of TBT/SPS measures are included in PTAs.7
As mentioned in the introductory chapter, the dataset is the result of a second
round of research on the content of PTAs. In a first round, the contents of 260 PTAs
notified to the WTO (as of 2015) were mapped across 52 policy areas. Legal experts
in charge of the mapping classified the policy areas between two categories with
respect to the WTO?s mandate. The nature of the first mapping could be described
7See [Espitia et al., 2019] for a complete description on the TBT chapter
70
as horizontal, in the sense that it is limited to the general question ?does the trade
agreement include a chapter on ?xyz? area??. It provides information on the breadth
of an agreement and the legal enforceability of area-specific commitments.
The second mapping was undertaken to go deeper into the details of specific
policy areas, extending the original 52 degrees of freedom into more than 1000 pro-
visions. In this chapter, I use this ?vertical? mapping of the TBT and SPS chapters
in order to identify their mode of integration. The TBT mapping covers the 255
PTAs that entered into force between 1960 and 2017 and included TBT provisions.
The template of the TBT mapping reported in Table (A.3) was first developed
by [Budetta and Piermartini, 2009], who originally focused on PTAs signed by the
EU, the US and Mexico. The first set of provisions indicates whether an agreement
refers to the WTO?s TBT Agreement and whether it goes beyond the commitments
taken under the WTO in terms of new areas covered or sector-specific engagements.8
The next three sets of provisions indicate the type of integration approach ? mu-
tual recognition versus three modes of harmonization ? adopted or promoted for
standards, technical regulations and conformity assessment procedures. The last set
of provisions describes the modalities of implementation, notably the way disputes
8The TBT Agreement calls for WTO members to use existing international standards as a
basis for their technical regulations. It encourages countries to accept as equivalent the technical
regulations of other Members if these regulations adequately fulfill the objectives of their own
domestic regulations (TBT Agreement, Article 2.7). Its Article 6 supports mutual recognition
agreements that recognize the tests and certifications of another country independently. It re-
quires transparency of new technical regulation, by expecting members to publish and notify the
WTO Secretariat (Article 2.9) and to make available inquiry point (Article 10). Under Article
11, members are expected to provide technical assistance to other WTO members for compliance
with TBT measures. Article 14 refers to the Dispute Settlement Body for consultations on TBT
matters, and finally the TBT Agreement establishes the Committee on TBT (Article 13) to deal
with the administration of the Agreement, which will later on be mandated to analyze Specific
Trade Concerns raised by members on matters related to trade protection and TBT measures.
71
shall be handled, the transparency of measures, the cooperation and technical assis-
tance offered to/by the signatories. I focus on the provisions describing the type of
integration, making the distinction between mutual recognition and harmonization.
I look at the impact of different provisions related to standards, technical regulations
and conformity assessment one at the time, but ultimately I combine those three
measures to create two composite indices, one for mutual recognition and one for
harmonization.
The SPS mapping covers 283 PTAs signed through 2016 and builds on the
work of [Jackson and Vitikata, 2016] as well as [Budetta and Piermartini, 2009].
It is organized in a similar way as the TBT mapping, with a first section asking
whether the PTA refers to the WTO mandate.9 Then there are three sections on
standards, risk assessment, and inspection, distinguishing between mutual recog-
nition and harmonization (to specific, regional or international rules). Finally, the
mapping is completed with provisions related to transparency, institutions and other
areas of cooperation. I also look at the individual impacts of SPS provisions one at
the time, before eventually moving to the creation of two indices that distinguish
between mutual recognition and harmonization. Table (A.4) shows the complete
list of questions coded for the SPS chapters.
The mapping of all TBT and SPS questions also records the degree of enforce-
ability of the provisions. The lowest degree of enforceability is assigned when treaties
simply address a provision by using non-binding legal language (with words such as
9Similar to its TBT counterpart, the SPS Agreement states the rights and obligations of WTO
members (Article 2), the recommendations relative to Harmonization (Article 3), Equivalence
(Article 4), Risk assessment (Article 5), Transparency (Article 7), technical assistance (Article 9),
Dispute Settlement (Article 11) and the creation of the Administration SPS committee.
72
?may?, ?should?, etc). If parties expressed the desire to commit or encourage the
implementation of a provision without resorting to a dispute settlement mechanism,
the level of enforceability is considered average. And finally, the highest level of en-
forceability refers to cases in which a provision is introduced with legal obligation or
clear ways of undertaking actions, notably with the creation of a dispute settlement
mechanism. For my analysis, I follow previous works in this domain and focus on
provisions that are at the minimum binding, arguing that weak enforceability does
not reflect a genuine commitment from the signatories.
The TBT and SPS chapters have been mapped to 34 and 59 provisions re-
spectively. Therefore for the sake of clarity, I first group the individual provisions
by their mode of integration, meaning either mutual recognition or harmonization.
I create TBTMutualRecoall and TBT
Harmo
all to indicate whether the agreement includes
any provisions related to the mutual recognition or harmonization of TBT mea-
sures, so these variables are 0/1 dummies. The same is done for the SPS measures.
Specifically, i) TBTMutualRecoall = 1 if any of the questions ?Is mutual recognition in
force?? for Standards, Technical Regulations, or Conformity Assessment are true.
ii) TBTHarmoall = 1 if any of the questions ?Are there specified existing standards
to which countries shall harmonize??, ? Is the use or creation of regional standards
promoted??, ?Is the use of international standards promoted??, or ?Do parties par-
ticipate in international or regional accreditation agencies?? also for Standards,
Technical Regulations, or Conformity Assessment are true. iii) SPSMutualRecoall = 1
if ?Standards ? Is equivalence recognized??, ?Standards ? Is mutual recognition
recognized?? or ?Audits/Control inspection ? Is mutual recognition in force?? is
73
true. iv) Finally, SPSHarmoall = 1 if any of ?Standards: - Do parties reference inter-
national standards??, ?Standards: - Do parties recognize the adaption to regional
conditions??, ?Standards & Inspection ? Are there specified existing standards to
which countries shall harmonize??, ?Standards& Inspection ? Is the creation of con-
certed/regional standards referenced??, or ?Inspection ? Is the use of international
standards promoted?? is true.
Next, I look at more disaggregated indices, but still focus on the mode of
liberalization (i.e. mutual recognition versus harmonization). I distinguish Stan-
dards, Conformity Assessment and Technical Regulation related provisions of TBT
measures, and Standards or Audit/Inspection related provisions of SPS measures.
When several provisions relate to one of the above categories, I create an index
reflecting the presence of ?any? relevant provisions. For example, the TBTHarmoStandard
index is a 0/1 dummy if any of the following provisions is true: ?Standards ? Are
there specified existing standards to which countries shall harmonize??, ?Standards
? Is the use or creation of regional standards promoted??, or ? Standards ? Is the
use of international standards promoted??.
3.4.3 ISO survey
In specification (3.6) I interact the TBT/SPS provisions with the ISO 9001
certifications gap between trading partners (relative to each partner?s total exports).
The purpose of this variable is to measure how distant countries are in terms of
quality-related best practice. Despite being disconnected from the ultimate goals
74
of TBT/SPS measures, the ISO 9001 standards family is a useful tool to gauge the
level of a company or organization?s commitment towards ?quality management?.
The ISO 9001 standard requires organizations to set out their objectives re-
garding the improvement of their quality management system. It helps businesses
and organizations to be more efficient and improve customer satisfaction. As an
example: the Se?ne?galaise Des Eaux (SDE), which supplies drinking water to five
million citizens, has been using ISO 9001 since 2002. The standard helps the orga-
nization to anticipate and meet the needs of its customers, as well as ensuring the
quality of the water and services. An example of customers? need is an improved
facility of payment. To satisfy this need, the SDE is diversifying its payment meth-
ods by offering mobile phone service. The advantage of using information on the
ISO 9001 standard family is that it can be adopted by any entity from the private
or public sector, regardless of its activity. It is a widespread standard adopted by
more than one million companies across the world and has the advantage of being
certifiable. Certification allows to track the incidence of ISO 9001 standards across
countries and industries.
Every year ISO carries out a global survey to collect the number of newly
granted ISO 9001 certificates. Since the ISO does not perform certification itself,
it relies on the responses of accredited certification bodies reporting the number of
ISO standards granted. The ISO acknowledges that the fluctuations in the num-
ber of certificates from year to year should be interpreted cautiously, because of
uncontrolled variations in the number of participating certification bodies. For this
reason, I collapse the data at the country-industry level and use the average number
75
of certifications across the surveyed period 2009-2014. The ISO survey is reported
at the SITC 2-digit level, which I map to the HS 2-digit classification. Certifications
have been reported by entities from 183 countries. Figure (3.4) shows large hetero-
geneity in certifications across sectors. More homogenous sectors, such as animal
and animal products, do not seek the quality management certification as much as
more complex sectors, such as transportation. This does not mean that the sectors
that do not seek ISO certifications are not concerned with quality, but simply reveals
that the chosen ISO measure has limitations and another quality index should be
used for those less differentiated sectors.
3.4.4 Specific trade concerns
Finally, I use data on the TBT and SPS Specific Trade Concerns (STCs)
notified to the WTO to identify the sectors for which TBT and SPS measures have
the largest incidence.10 The STCs database publishes the information reported
to the two special SPS and TBT committees established within the WTO. The
purpose of these committees is to provide member countries a ?forum for exchange
of information on all aspects related to the implementation of the SPS and TBT
Agreements?. If one country estimates that a trading partner is imposing unfair
SPS/TBT measures that have a significant effect on trade, it can raise its concern to
one of the two committees and report which country is imposing the new SPS/TBT
measure, the affected products, and how it affects trade. The WTO publishes
10 [Fontagne? et al., 2015] use the STC database to shed light on the effect of non-tariff barriers
on trade, [Orefice, 2017] looks at the positive relationship between tariff reduction and TBT/SPS
measures increase.
76
all this information in two SPS and TBT STCs databases, which report: 1) the
country or countries raising the concern and the country imposing the measure,
2) the HS 4-digit product of concern, 3) the year the concern was raised (data
available between 1995 and 2011), 4) whether the concern has been resolved, and
5) the reason for the concern (quality requirements, protection of the environment,
harmonization, consumer information, etc.) The STC data helps me establish that
all sectors are exposed to the TBT measures, but SPS measures only apply to the
Animal, Vegetable, Food and related products sectors.
3.5 Results
In this section, I first present the results from the estimation of quality and
quality ladders and show that the estimates verify the general findings on quality
from the literature. Then I move to the estimation of the effects of TBT and SPS
measures on the estimated qualities and test for potential sources of heterogeneity
in their effects.
3.5.1 Quality estimates
Quality is estimated after running 94 separate econometric specifications (3.3)
corresponding to each HS 2-digit industry. The identification of the price coefficient
comes from the origin-destination-6-digit-product variation within a 2-digit industry.
The resulting quality estimates are available at the HS 6-digit product level, but
are eventually aggregated to the 2-digit industry level using import share weights.
77
Ultimately, I decided to work with an industry-level measure of quality in order to
identify potential heterogeneity in the impact of TBT/SPS measures, which are only
observable at this broader level of classification. The summary statistics from the
econometric specifications are reported in table (3.2). The comparison of the price
coefficients between the OLS and the IV estimations shows how accounting for the
problem of endogeneity improve the sign and magnitude of the unit price coefficient.
The F ? stats of the first stage are large, confirming the relevance of instruments
for the price variable. the R ? squared and F ? stat of the quality equation also
confirm the validity of the estimations.
The inferred qualities are regressed on different proxies for countries? develop-
ment level: GDP per capita, the capital-labor ratio, and the share of population
with high education. The results, reported in table (3.3), show that developed coun-
tries (with higher GDP per capita, higher capita-labor ratios and higher education
attainment) export higher quality products relative to developing countries, which
is in line with the findings in the literature. Results are robust to the variables
(population versus GDP ) used in the import demand equation to control for the
number of varieties produced in the exporting countries. In the rest of this chapter,
I present the results using the measure of quality that was obtained from the spec-
ifications using population as a control variable for the number of varieties, but the
main findings remain the same with the alternative GDP control.
In a second step, I construct the HS 6-digit quality ladders based on the
formula (3.7) applied to log qualities. For a given product, the ladder is a proxy
of the scope for quality improvement available to firms. I remove outliers that
78
could distort the ladders by computing the interdecile and interquartile ranges. I
then aggregate the product ladders at the HS 2-digit industry level using import
shares as in (3.8). Figure (3.2) shows that the lengths of the industry ladders are
more or less constant over time for a given industry, at least within the relatively
short time span available, during which we would not observe frequent and major
technological progresses. Statistics on the dispersion, maximum, minimum and
median also confirm the stability of industry ladders over time.
Moreover, the figure displays the distribution of scopes for quality improvement
across industries. Intuitively, the prepared foods, mineral products and vegetable
products industries have the shortest ladders, while transportation and raw hides
have the longest. This ranking also emerged from the elasticities of substitution,
with long ladders corresponding to more differentiated goods with small elasticities
of substitution, and short ladders corresponding to less differentiated goods that are
easier to substitute and present smaller margins for quality differentiation.11 An-
other comparison of the long versus short ladder can be done with respect to [Rauch,
1999] classification of differentiated products. Figure (3.3) show the correspondence
between the Rauch?s and HS classifications. The underlying correspondence is es-
tablished at the HS 6-digit, but in the chart I display the proportion of homoge-
neous/differentiated in each broad HS categories. Long ladder categories overlap
with the differentiated Rauch?s category quite well, comforting the interpretation of
sectors with longer ladder to be more differentiated.
As a last test on the estimates for quality, I run a similar regression of unit
11But would have expected machinery and electrical goods to be associated with a longer ladder.
79
values on quality as in [Khandelwal, 2010] to assess the relationship between prices
and qualities. If prices were the best proxy for product quality, one would observe
a positive correlation between them. Khandelwal argues that this would be the
case if products only possess vertical (quality) attributes. However, in the presence
of horizontal (taste/variety) attributes, the relationship may reverse and may vary
with the length of the ladders. I ran the following specification:
log(pijst) = ?1 log(q?ijst) +?2 log(q?ijst)? log(IndustryLadderijht) +?ht + ijst (3.11)
where pijst is the unit value at the product level, q?ijst is the estimated quality
from the previous stage, IndustryLadderijht is the corresponding scope for quality
improvement, and ?ht are the product-time fixed effects. The results, reported in
table (3.4), indicate a very small but positive and significant coefficient ?2. This
means that the unit prices of products associated with longer ladders are more
correlated with quality than for products with short ladders. In other words, the
quality of more differentiated goods may be approximated by unit prices (even
though the relationship is weak), but the quality of homogeneous goods is poorly
identified by unit prices. The last remark is on the negative sign of the coefficient
associated with quality. In the present study, quality is inferred from prices but
conditional on the demand for imports. As import demand and prices are negatively
correlated, so are estimated quality and prices.
80
3.5.2 Baseline effect of TBT/SPS measures on quality
Using the inferred measure of quality, I can estimate the change in quality
following the entry into force of a PTA that includes TBT and SPS measures.
Quality is aggregated at the country-pair by year by 2-digit industry level using trade
shares as weights. The first set of results are based on the baseline specification (3.4).
The results for the baseline specification raise the importance of including
country-pair fixed effects to account for bilateral characteristics that might deter-
mine the quality of goods traded between countries, such as historical or cultural
ties, distance12, or common language. Table (3.6) shows that, out of the four ag-
gregated indices of interest, only the effect of mutual recognition of SPS measures
?survives? the inclusion of country-pair fixed effects. The other variables are sig-
nificant and positive before controlling for pair fixed effects, but are not significant
afterwards. This sheds light on the presence of policy endogeneity explained by the
fact that rich countries (which import and export high quality products) are likely
to spread their own TBT/SPS regulations through PTAs negotiated with their main
trading partners. The control variable TotalDepth is significant and positive. Since
its definition does not depend on the presence of TBT or SPS chapters in PTA, this
coefficient reflects the remaining effect of the content of a PTA going beyond the
chapters related to TBT and SPS measures. While the improvement of the overall
institutional environment seems to positively affect quality, provisions other than
TBT/SPS are not directly targeting issues on health and safety, and therefore shall
12cf. The Washington Apple effect, countries ship higher quality goods to long distance destina-
tion in order to reap for the cost of shipping
81
raise less endogeneity concerns for quality.
I then look at the disaggregated indices. Results for the TBT-related pro-
visions, presented in table (3.7), suggest that the choice of integration approach
is important to the impact on quality. Choosing mutual recognition (for any of
the standards, technical regulation, or conformity assessment categories) increases
quality by twice as much as the harmonization approach. One explanation might
be that harmonization actually entails adaptation costs to new regulations, and re-
sources involved to comply with the new regulations cannot be invested into quality
improvement (at least in the aftermath of the regulatory change).
The coefficient on provisions captures the combined effect of increased market
access and change in minimum quality regulation. 13 The reported coefficients tell
us that a mutual recognition of TBT standards increases quality by 22 percent. The
distributions of quality estimates reported for each sector in table (3.5) reveal that
a 20 percent impact will not change dramatically the level of quality.14
Moving to the SPS results, table (3.8) indicates that the positive and signif-
icant relationship between quality and mutual recognition is driven by the provi-
sion on inspections. Surprisingly, the comparison between mutual recognition and
harmonization is reversed, as harmonization seems to increase quality more than
13Results reported in appendix tables (A.5) and (A.6) show that between 35 and 40 percent of
the overall impact of TBT/SPS provisions can be account for by sectoral composition.
14Because quality is an abstract measure is it also pertinent to relate this result to the comparison
of export quality across countries. In 2003, Australia and Singapore signed a trade agreement and
committed to mutually recognize their standards and technical regulations on TBTs. On average,
the agreement is expected to have increased the quality of goods exported from Singapore to
Australia by about 20 percent. As an illustration, in the inferred data this corresponds to an
increase in quality of electrical machines or paper and paper products to the level of Israel, or an
increase in the quality of headgear to the level of New Zealand.
82
mutual recognition on average. Sectors concerned with SPS measures tend to be
more sensitive and protected by governments than those concerned with TBT mea-
sures, and they also tend to be more homogenous. Therefore, the scope for vertical
differentiation is more limited.
3.5.3 Interaction with level of development
In the next three sections, I investigate the heterogeneity of the relationships
between TBT or SPS provisions and quality. First, I look at differences across coun-
tries? level of development. I define as ?northern? all countries classified as high
income in the World Bank?s WDI database. The ?southern? category comprises
all other income levels from upper-middle to low-income. I create three dummies
(North-North, North-South and South-South) to identify the trading relationships
between developed and developing countries. Results from the econometric specifi-
cation 3.5 are reported in tables (3.9) and (3.10).
Mutual recognition of TBT measures tends to improve the quality of traded
goods involving a developing country. Mutual recognition is easier to implement
than harmonization and it is therefore not surprising that low income countries
would benefit from it. The negative sign of the coefficient on the interaction between
developed countries is more difficult to interpret. One would need to investigate
further whether the increased competition from low productivity firms is driving
this result. The third and fourth columns of table (3.9) show that the positive
relationship between harmonization of TBT measures and quality is driven by North-
83
South trade. This is where the institutional gap is the biggest and is actually
where most of the PTAs including a harmonization approach are signed. Minimum
standards are expected to strengthen in developing countries, leading to a significant
and positive regulatory channel (confirmed by the results reported in column (4)).
As expected, harmonization does not have an impact on quality of imports between
developed countries because ?standards? are already high for all partners.
The results for the interaction between SPS measures and the level of devel-
opment are less striking. Only the mutual recognition of SPS measures between
developed and developing countries appears positive and significant in table (3.10),
while all remaining coefficients are insignificant. The decomposition of SPS indices
does not improve the results either. Perhaps the relevant heterogeneity does not fully
reside in the difference of development levels, but rather in the sectoral composition.
This is investigated in the next section.
3.5.4 Interaction with ISO 9001
I present the second exercise to investigate the heterogeneity of the effects of
TBT and SPS measures on quality. I interact the integration approach indices with
the gap of ISO 9001 certifications between trading partners as indicated in equa-
tion 3.6. The ISO gap is computed as the difference between the number of firms?
certifications for this well-known quality management standard over time. Before
taking the difference, the number of certifications is first scaled to the country?s
total exports, to account for the fact that larger markets may have more firms and
84
therefore more chances to appear in the ISO 9001 certifications survey. The result-
ing variable is then aggregated at the country-pair-industry level. Thanks to its
large country coverage and multi-sector availability, this variable provides a useful
proxy for how quality is valued by firms residing in each market.
Table (3.11) shows the results from including these interactions with TBT
and SPS aggregated indices. Large gaps in quality management certification do
not seem to matter for TBT provisions (whichever integration approach is imple-
mented). SPS provisions have a negative impact on quality for sectors with the
larger gap of quality management certification. These results give a hint on un-
derlying sectoral differences. Focusing on the sectors concerned by SPS provisions
(i.e. HS ? 24 ) and based on figure (3.4), larger ISO gaps are expected to occur in
differentiated industries, such as prepared food (HS 16-24), which account for rel-
ative more certifications than homogeneous industries such as animal or vegetable
products.
3.5.5 Interaction with quality ladder
Finally I look at the sectoral heterogeneity as described by equation 3.9, which
interacts the TBT and SPS provisions with the quality ladder measure of the scope
for quality improvement. The ladder measures the range between the smallest and
the largest inferred quality for a given variety. As shown in figure (3.2), the rank-
ing of average industry ladders remains more or less the same over time and is
consistent with the degree of differentiability of goods. Therefore, I decide to use
85
the trade-weighted average over the sample period to end up with a measure at
the country-pair-industry level. The variable LadderFullRange is evaluated by
taking into account the whole range of quality estimates and gives the difference
between maximum and minimum qualities, whereas LadderInterquartile computes
the range between the 25th and 75th percentiles, allowing us to abstract from the
potential effects of outliers. Both variables are continuous.
Results, reported in tables (3.12) and (3.13), show that industries with larger
scope for quality improvement (i.e. longer ladders) are more impacted by both mu-
tual recognition and harmonization of TBT and SPS provisions. The results are
similar for the two measures of quality ladder used, full range and inter-quartile,
showing that even after trimming outliers and hence reducing the ladder range,
TBT and SPS provisions still have different sectoral effects. As competition in-
creases following a bilateral or regional integration, firms will be pushed to improve
quality of their goods in order to vertically differentiate themselves and appeal to
consumers? taste for quality. The wider the potential quality range (maybe for
technical reasons), the easier it is for firms to differentiate and compete.
The results reported in the third columns of both the TBT and SPS tables
show that a proper theoretical framework is required to understand the effect of har-
monization. When I account for the scope of quality improvement, which varies with
country pair and industry, harmonization provisions appear to be negatively corre-
lated with quality. The interpretation of this result tells us that the quality of fully
homogenous goods is negatively impacted by harmonization provisions (whereas
mutual recognition provisions do not have any effect on homogenous goods). This
86
is related to the impact on trade of harmonization on homogenous versus differenti-
ated goods found in the empirical literature. [Portugal-Perez et al., 2010] find that
international harmonization of EU standards has a positive and significant impact
on imports of electronics. Meanwhile, [Czubala et al., 2009] find that the same type
of harmonization has a negative effect on imports of textiles and apparel, which are
more homogenous than electronics. Standards have a dual impact on trade costs,
as they bring adaptation costs but also reduce information costs. The opposite ef-
fects of standards harmonization on trade of homogenous and differentiated goods
might reside in the relative weights of these information versus adaptation costs;
assuming that differentiated goods benefit more from the information conveyed by
standards. Harmonization of standards might divert trade of homogenous goods
because it entails relatively high adaptation costs. Since I estimate quality from the
import demand residuals, if imports decrease because of standards harmonization
and if prices are sluggish to adjust to the new standards, then estimated quality
should decrease.
3.6 Conclusion
This empirical chapter contributes to the large literature that estimates the
impact of PTAs. The novel dataset on the vertical Content of PTAs enables us to
disentangle the various modes of integration. Results show that provisions matter
differently for quality improvement.
I find that on average mutual recognition of TBT measures increases quality
87
by twice as much as harmonization, whereas quality benefits more from the har-
monization of SPS measures than from mutual recognition. However, the baseline
results hide a lot of heterogeneity across markets and sectors. Further exploration
reveals that TBT and SPS provisions are more impactful whenever a developing
country is involved in the PTA (this is in line with findings on the impact of deep
PTAs on GVC trade [Laget et al., 2018] or FDI as shown in the next chapter). I
also find big sectoral differences. Differentiated goods, which tend to display bigger
scope for quality improvement, are positively impacted by TBT and SPS provisions
whatever the mode of integration.
A large share of those effects is driven by country-pair effects revealing the
presence of policy endogeneity. While the total depth of PTAs still has a positive
and significant impact on quality after controlling for the pair fixed effects, the ef-
fects of the aggregate measures of mutual recognition or harmonization are either
reduced or muted. This can be explained by the fact that, by nature, TBT/SPS pro-
visions directly target issues on quality, whereas other provisions focus on different
institutional challenges unrelated to health and safety issues.
88
3.7 Figures
Mineral products (HS 25-27) 17
Wood and paper (HS 44-49) 29
Miscellaneous (HS 91-99) 42
Raw hides, skins, leather and furs (HS 41-43) 46
Plastics and rubbers (HS 39-40) 52
Metals (HS 72-84) 91
Stone and glass (HS 68-71) 100
Machinery and electrical (HS 85-86) 108
Footwear and headgear (HS 64-67) 138
Transportation (HS 87-90) 150
Animal and animal products (HS 1-5) 286
Vegetable products (HS 6-15) 455
Textiles (HS 50-63) 1022
Foodstuffs (HS 16-24) 1753
Chemnical and allied industries (HS 28-38) 2082
0 500 1,000 1,500 2,000
Cumulative number of TBT concerns
Transportation (HS 87-90) 1
Textiles (HS 50-63) 2
Chemnical and allied industries(HS 28-38) 3
Wood and paper (HS 44-49) 5
Raw hides, skins, leather and furs (HS 41-43) 7
Foodstuffs(HS 16-24) 45
Animal and animal products(HS 1-5) 85
Vegetable products(HS 6-15) 86
0 20 40 60 80
Cumulative number of SPS concerns
Source: WTO?s TBT and SPS committee.
Figure 3.1: Number of concerns raised for SPS and TBT measures
89
Log quality ladders
1995 2000 2005 2010 2015
year
Animal and animal products (HS 1-5) Vegetable products (HS 6-15)
Prepared food, beverages and tobacco (HS 16-24) Mineral products (HS 25-27)
Chemical and allied industries (HS 28-38) Plastics and rubbers (HS 39-40)
Raw hides, skins, leather and furs (HS 41-43) Wood and paper (HS 44-49)
Textiles (HS 50-63) Footwear and headgear (HS 64-67)
Stone, ceramic and glass (HS 68-71) Base metals (HS 72-84)
Machinery and electrical (HS 85-86) Transportation (HS 87-90)
Miscellaneous (HS 91-99)
Source: Author?s calculation based on estimated qualities.
Figure 3.2: Average quality ladders by broad industries
90
-2 0 2 4 6
Source: Author?s calculation based on STIC rev.2 and HS 1996 correspondence tables
Figure 3.3: Average correspondence between Rauch?s and HS classificaiton
91
Source: ISO?s annual survey on ISO 9001 standards certifications.
Note: Total number of ISO 9001 certifications granted between 2009 and 2014. Original
SITC classification manually mapped to HS classification.
Figure 3.4: Total ISO 9001 certifications by broad HS industries
92
worldwide ISO 9001 certifications
0 50,000 100000
Animal and animal products (HS 1-5)
Raw hides, skins, leather and furs (HS 41-43)
Mineral products (HS 25-27)
Vegetable products (HS 6-15)
Miscellaneous (HS 91-99)
Textiles (HS 50-63)
Wood and paper (HS 44-49)
Stone, ceramic and glass (HS 68-71)
Prepared food, beverages and tobacco (HS 16-24)
Chemical and allied industries (HS 28-38)
Plastics and rubbers (HS 39-40)
Machinery and electrical (HS 85-86)
Transportation (HS 87-90)
Base metals (HS 72-84)
3.8 Tables
HS sections BW median sigma
Animal and animal products (HS 1-5) 5.075
Prepared food, beverages and tobacco (HS 16-24) 4.25
Textiles (HS 50-63) 3.65
Vegetable products (HS 6-15) 3.35
Base metals (HS 72-84) 3.275
Wood and paper (HS 44-49) 3.2
Plastics and rubbers (HS 39-40) 3.125
Mineral products (HS 25-27) 3.05
Footwear and headgear (HS 64-67) 3
Chemical and allied industries (HS 28-38) 2.8
Transportation (HS 87-90) 2.55
Stone, ceramic and glass (HS 68-71) 2.4
Raw hides, skins, leather and furs (HS 41-43) 2.35
Machinery and electrical (HS 85-86) 2.2
Miscellaneous (HS 91-99) 2.2
Table 3.1: Elasticities of substitution from Broda and Weinstein (2006)
Mean Median 25th 75th
OLS price coefficient 0.009 0.011 -0.049 0.076
IV price coefficient -0.969 -1.136 -2.815 0.353
F-stat of quality estimates 62 38 7 93
F-stat first stage 2925 2574 1605 3820
R-squared 0.700 0.701 0.682 0.730
Table 3.2: Summary statistics for the estimation of the import demand equation
93
ln(quality) ln(quality) ln(quality) ln(quality) ln(quality) ln(quality)
VARIABLES population population population GDP GDP GDP
control control control control control control
(1) (2) (3) (4) (5) (6)
ln(PCGDPiht) 0.632*** 0.572***
(0.0232) (0.0230)
ln(Kiht/Liht) 0.661*** 0.596***
(0.0225) (0.0223)
ln(HighEduciht) 1.584*** 1.518***
(0.292) (0.302)
Observations 3,181,373 3,098,059 1,356,962 3,180,353 3,097,508 1,356,908
R-squared 0.417 0.422 0.518 0.341 0.345 0.421
Destination-time FE Yes Yes Yes Yes Yes Yes
Product-time FE Yes Yes Yes Yes Yes Yes
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Table 3.3: Quality and factors of endowment
log(UnitV alueijht)
VARIABLES (1) (2) (3) (4)
log(qualitypopcontrolijht ) -0.0131*** -0.00947***
(0.000198) (0.000164)
log(qualitypopcontrolijht )? log(ladderijht) 6.59e-05***
(0.00000281)
log(qualitypopcontrolijht )? log(ladder
interquartile
ijht ) 1.30e-05***
(0.00000289)
log(qualitygdpcontrolijht ) -0.00330*** -0.00295***
(0.000147) (0.000128)
log(qualitygdpcontrolijht )? log(ladderijht) 6.92e-05***
(2.49e-06)
log(qualitygdpcontrolijht )? log(ladder
interquartile
ijht ) 1.65e-05***
(1.73e-05)
Observations 3,096,827 3,096,827 3,084,433 3,084,433
R-squared 0.595 0.595 0.595 0.595
Product-year FE Yes Yes Yes Yes
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Table 3.4: Relationship between unit prices and quality
94
p1 p5 p10 p25 p50 p75 p90 p95 p99
Animal and animal products (HS 1-5) -7.52 -3.31 -1.85 -0.39 0.11 1.79 7.91 13.58 31.32
Vegetable products (HS 6-15) -6.19 -2.13 -1.05 -0.19 0.18 0.89 2.81 5.70 19.28
Prepared food, beverages and tobacco (HS 16-24) -2.14 -0.64 -0.33 -0.06 0.10 0.31 0.65 1.04 3.00
Mineral products (HS 25-27) -5.14 -1.71 -0.80 -0.12 0.20 0.87 2.70 5.37 17.03
Chemical and allied industries (HS 28-38) -3.83 -1.42 -0.70 -0.06 0.29 0.99 4.51 29.08 72.46
Plastics and rubbers (HS 39-40) -6.74 -1.49 -0.55 0.11 0.82 4.65 14.91 23.67 47.53
Raw hides, skins, leather and furs (HS 41-43) -7.47 -2.42 -1.16 -0.09 0.81 4.06 11.33 18.73 40.18
Wood and paper (HS 44-49) -4.56 -1.40 -0.62 -0.03 0.34 0.95 2.88 6.30 18.34
Textiles (HS 50-63) -3.10 -1.06 -0.48 0.02 0.42 1.17 3.03 5.62 17.22
Footwear and headgear (HS 64-67) -9.12 -3.05 -1.46 -0.17 0.84 3.09 8.41 14.07 30.51
Stone, ceramic and glass (HS 68-71) -6.08 -2.26 -1.15 -0.19 0.40 1.57 3.96 6.60 18.47
Base metals (HS 72-84) -5.23 -1.09 -0.34 0.08 0.86 7.55 31.66 51.73 74.82
Machinery and electrical (HS 85-86) -12.34 -2.39 -0.44 0.93 2.46 5.40 10.86 15.23 25.93
Transportation (HS 87-90) -10.61 -2.31 -0.91 -0.04 0.41 9.27 31.95 46.37 69.78
Miscellaneous (HS 91-99) -5.86 -2.41 -1.31 -0.22 0.65 2.50 6.86 12.00 29.25
Table 3.5: Distribution of quality estimates by broad sectors
95
ln(quality)
VARIABLES (1) (2) (3) (4) (5) (6) (7) (8)
TBTMutualRecoall 0.526*** 0.0308
(0.103) (0.0659)
TBTHarmoall 1.100*** 0.100
(0.0914) (0.0638)
SPSMutualRecoall 0.593*** 0.236**
(0.0800) (0.117)
SPSHarmoall
0.343*** -0.0331
(0.0446) (0.0407)
TotalDepth 0.0438*** 0.00429*** 0.0294*** 0.00514*** 0.0183*** 0.00398*** 0.0132*** 0.00469***
(0.00255) (0.00136) (0.00249) (0.00145) (0.00125) (0.00123) (0.00141) (0.00136)
Observations 3,196,889 3,196,886 3,196,889 3,196,886 667,363 667,361 667,363 667,361
R-squared 0.435 0.465 0.436 0.465 0.231 0.253 0.231 0.253
Origin-time FE Yes Yes Yes Yes Yes Yes Yes Yes
Destination-time FE Yes Yes Yes Yes Yes Yes Yes Yes
HS2 FE Yes Yes Yes Yes Yes Yes Yes Yes
Pair FE No Yes No Yes No Yes No Yes
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Table 3.6: Baseline estimation of the impact of TBT and SPS provisions on average
quality
96
ln(quality)
VARIABLES (1) (2) (3) (4) (5) (6)
TBTMutualRecoStandard 0.224***
(0.0741)
TBTMutualRecoTechReg 0.224***
(0.0741)
TBTMutualRecoConformity 0.232***
(0.0659)
TBTHarmoStandard 0.111*
(0.0632)
TBTHarmoTechReg 0.150**
(0.0621)
TBTHarmoConformity 0.137**
(0.0622)
TotalDepth 0.0124*** 0.0124*** 0.0126*** 0.0114*** 0.0118*** 0.0117***
(0.00140) (0.00140) (0.00138) (0.00136) (0.00137) (0.00136)
Observations 3,196,883 3,196,883 3,196,883 3,196,883 3,196,883 3,196,883
R-squared 0.456 0.456 0.456 0.456 0.456 0.456
Origin-time FE Yes Yes Yes Yes Yes Yes
Destination-time FE Yes Yes Yes Yes Yes Yes
HS2 FE Yes Yes Yes Yes Yes Yes
Pair FE Yes Yes Yes Yes Yes Yes
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Table 3.7: Baseline estimation of the impact of TBT provisions on average quality
97
ln(quality)
VARIABLES (1) (2) (3) (4)
SPSMutualRecoStandard 0.0258
(0.125)
SPSMutualRecoInspection 0.194*
(0.115)
SPSHarmoStandard 0.343*
(0.184)
SPSHarmoInspection -0.0758
(0.0488)
TotalDepth 0.00350*** 0.00338*** 0.00347*** 0.00424***
(0.00120) (0.00120) (0.00120) (0.00127)
Observations 667,361 667,361 667,361 667,361
R-squared 0.216 0.216 0.216 0.216
Origin-time FE Yes Yes Yes Yes
Destination-time FE Yes Yes Yes Yes
HS2 FE Yes Yes Yes Yes
Pair FE Yes Yes Yes Yes
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Table 3.8: Baseline estimation of the impact of SPS provisions on average quality
98
VARIABLES ln(quality)
(1) (2)
TBTMutualRecoall ?NorthNorth -0.143**
(0.0717)
TBTMutualRecoall ?NorthSouth 0.324***
(0.110)
TBTMutualRecoall ? SouthSouth 0.923***
(0.226)
TBTHarmoall ?NorthNorth -0.0922
(0.0795)
TBTHarmoall ?NorthSouth 0.226***
(0.0734)
TBTHarmoall ? SouthSouth 0.272
(0.216)
TotalDepth 0.00407*** 0.00479***
(0.00131) (0.00140)
Observations 3,196,883 3,196,883
R-squared 0.456 0.456
Origin-time FE Yes Yes
Destination-time FE Yes Yes
HS2 FE Yes Yes
Pair FE Yes Yes
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Table 3.9: Impact of TBT measures on quality across development levels
99
VARIABLES ln(quality)
(1) (2)
SPSMutualRecoall ?NorthNorth -0.242
(0.364)
SPSMutualRecoall ?NorthSouth 0.336*
(0.172)
SPSMutualRecoall ? SouthSouth 0.0969
(0.152)
SPSHarmoall ?NorthNorth 0.0405
(0.0588)
SPSHarmoall ?NorthSouth -0.0772
(0.0546)
SPSHarmoall ? SouthSouth -0.0534
(0.0774)
TotalDepth 0.00333*** 0.00423***
(0.00120) (0.00134)
Observations 667,361 667,361
R-squared 0.216 0.216
Origin-time FE Yes Yes
Destination-time FE Yes Yes
HS2 FE Yes Yes
Pair FE Yes Yes
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Table 3.10: Impact of SPS measures on quality based on the level of development
of partners
100
VARIABLES ln(quality)
(1) (2) (3) (4)
TBTMutualRecoall 0.0730***
(0.0255)
TBTMutualRecoall ? isogap9001 -355.3
(376.1)
TBTHarmoall 0.0417*
(0.0226)
TBTHarmoall ? isogap9001 0.608
(8.142)
SPSMutualRecoall 0.183
(0.122)
SPSMutualRecoall ? isogap9001 -13.62***
(2.996)
SPSHarmoall -0.00953
(0.0274)
SPSHarmoall ? isogap9001 -11.33**
(5.057)
isogap9001 -7.815 -7.735 -14.93 -14.94
(36.80) (36.77) (17.79) (17.78)
TotalDepth 0.0122*** 0.0115*** 0.00317** 0.00343**
(0.00141) (0.00138) (0.00126) (0.00134)
Observations 3,099,089 3,099,089 650,400 650,400
R-squared 0.458 0.458 0.215 0.215
Origin-time FE Yes Yes Yes Yes
Destination-time FE Yes Yes Yes Yes
HS2 FE Yes Yes Yes Yes
Pair FE Yes Yes Yes Yes
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Table 3.11: TBT/SPS measures interaction with ISO 9001 certification gap
101
VARIABLES ln(quality)
(1) (2) (3) (4)
TBTMutualRecoall 0.0560 0.0275
(0.0653) (0.0523)
TBTMutualRecoall ? LadderFullRange 0.0103*
(0.00623)
TBTMutualRecoall ? LadderInterquartile 0.0325***
(0.0122)
TBTHarmoall -0.163*** 0.109**
(0.0629) (0.0497)
TBTHarmoall ? LadderFullRange 0.0311***
(0.00568)
TBTHarmoall ? LadderInterquartile -0.0131
(0.00993)
LadderFullRange 0.363*** 0.361***
(0.00188) (0.00190)
LadderInterquartile 0.648*** 0.670***
(0.00298) (0.00323)
TotalDepth 0.00132 0.00144 0.00139 0.00160
(0.00108) (0.00107) (0.00115) (0.00116)
Observations 3,096,694 3,096,694 3,096,694 3,096,696
R-squared 0.599 0.611 0.599 0.620
Origin-time FE Yes Yes Yes Yes
Destination-time FE Yes Yes Yes Yes
HS2 FE Yes Yes Yes Yes
Pair FE Yes Yes Yes Yes
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Table 3.12: Interaction between TBT measures and ladder length
102
VARIABLES ln(quality)
(1) (2) (3) (4)
SPSMutualRecoall -0.173 0.145
(0.124) (0.106)
SPSMutualRecoall ? LadderFullRange 0.157***
(0.0335)
SPSMutualRecoall ? LadderInterquartile 0.0832**
(0.0390)
SPSHarmoall -0.148*** -0.00345
(0.0323) (0.0270)
SPSHarmoall ? LadderFullRange 0.0483***
(0.00772)
SPSHarmoall ? LadderInterquartile 0.0403***
(0.0109)
LadderFullRange 0.279*** 0.269***
(0.00480) (0.00489)
LadderInterquartile 0.461*** 0.471***
(0.00798) (0.00871)
TotalDepth 0.00333*** 0.00204* 0.00409*** 0.00251**
(0.00111) (0.00108) (0.00111) (0.00125)
Observations 645,996 645,996 645,996 645,996
R-squared 0.388 0.359 0.390 0.394
Origin-time FE Yes Yes Yes Yes
Destination-time FE Yes Yes Yes Yes
HS2 FE Yes Yes Yes Yes
Pair FE Yes Yes Yes Yes
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Table 3.13: Interaction between SPS measures and ladder length
103
Chapter 4: Deep PTAs and FDI
4.1 Disclaimer
This chapter is adapted from joint work with Nadia Rocha and Gonzalo Varela.
It has not undergone the review accorded to official World Bank publications. The
findings, interpretations, and conclusions expressed herein are those of the authors
and do not necessarily reflect the views of the World Bank Group and its affili-
ated organizations, or those of the Executive Directors of the World Bank or the
governments they represent
4.2 Introduction
Because investment is complementary to trade, PTAs provide relevant insti-
tutional frameworks to partner countries that wish to regulate their mutual foreign
investments. In this chapter, we study the impact of deep trade agreements on
foreign direct investment (FDI) decisions, and examine three sub-questions. First,
is the impact of trade agreements on FDI heterogeneous across types of business
activities? Second, is this impact heterogeneous across disciplines covered in the
agreements? Third, does the level of development of home and host countries mat-
104
ter for this impact? We exploit the World Bank?s dataset on the content of PTAs
and data on bilateral greenfield investment decisions at the activity level. We find
that deep trade agreements matter for investment decisions: every additional disci-
pline into a PTA increases FDI by 1.4 percent, on average. Deep agreements do not
impact FDI in natural resources and extractive activities, and have heterogeneous
effects across production and services related activities. Results also reveal that dis-
ciplines that go beyond the mandate of the WTO matter more for FDI. Disciplines
related to investment liberalization and protection, IPR or migration stand out to
increase FDI, whereas disciplines on labor market regulations reduce investment.
Finally, we find that our results are mostly driven by investment between developed
and developing countries.
As PTAs are growing in number and deepening in content, the growing frag-
mentation of production has shifted the composition of trade towards more flows
of differentiated intermediate products and less of homogenous goods. This has re-
sulted into a rise in global value chains (GVC), reinforcing the need to concurrently
regulate the exchange of goods and services with FDI under a common frame-
work [Antra?s and Staiger, 2012]. Moreover, Bilateral Investment Treaties (BITs)
that have traditionally regulated foreign investments are losing ground as more in-
vestment related disciplines are incorporated in PTAs (see Figure 4.8). By removing
economic uncertainty regarding market access rules and beyond the border business
conditions, PTAs are increasingly relevant in the context of cross-border trade and
investment. Their prevalence has revived interest in the literature on regionalism,
as researchers examine the role of this new generation of deep trade agreements in
105
shaping the exchange of goods and services as well as investments. Understanding
the economic effects of PTAs is essential to design and implement them efficiently.
This chapter contributes to the large debate on the economic impact of deep
trade agreements by empirically investigating the key disciplines that promote FDI.
We use a new dataset on the content of PTAs developed by the World Bank and
bilateral cross-border data on greenfield investments run by the Financial Times.
The latter provides information on the sector and business activity associated with
an FDI operation. An activity defines the actual function of a project and can be de-
scribed as a task along the supply process to end-users: from upstream research and
development, passing by manufacturing (i.e. production), to downstream sale and
retail. We argue that these activities provide a better understanding of which parts
of the supply chain are most impacted by PTAs than sectors defined by standard
industrial classification.1
After quantifying the average relationship of deep trade agreements on FDI,
we explore three sources of heterogeneity in this effect, to reveal which factors drive
our results: i) by sector - we first distinguish FDI by the type of business activity
that is financed (differentiating between extractives, production and services related
activities); ii) by type of discipline - we isolate the effect of specific disciplines
included in PTAs; and iii) by level of development of the signatory countries. The
first exercise relates to the impact of PTAs on resource seeking FDI, the possibility to
isolate investment in extractive activities inform us on resource driven FDI. Studying
1Indeed, because a sector characterizes the end-use of goods or services provided by an FDI
recipient, it generally entails a succession of activities and does not allow to accurately locate which
part of the supply chain is involved on the FDI operation.
106
the impact of individual disciplines give us some hint on the efficiency seeking motive
of FDI. By lowering the cost and uncertainty to do business across borders, deep
PTAs facilitate operations along the value chains. Finally the third exercise which
focuses on income differences between partners inform us on market seeking FDI.
The literature has shown that this category of horizontal FDI tends to occur more
between developed and developing countries.
Our key finding is that deep PTAs matter for FDI between member countries.
Their positive impact persists even after accounting for the presence of BITs or shal-
low PTAs, which confirms that the disciplines covered in deep PTAs go beyond the
commitments of traditional agreements. Adding a discipline to a PTA increases FDI
by 1.4 percent. This result is driven by production and services related activities,
and we also find that deep PTAs do not affect investments in resource extraction
activities. We then estimate separately the impact of individual disciplines2 related
to investments. Among the core disciplines of PTAs - i.e. those disciplines that
have a clear economic content, as opposed to other disciplines that do not (e.g.
cultural cooperation, anti-terrorism), those that matter for FDI are: investment,
intellectual property rights (IPR), visa and asylum, environmental laws, labor mar-
ket regulations, movement of capital and competition policy. We find that all have
a large and positive impact and increases FDI in service-related activities between
32 and 50 percent. However, labor and environmental disciplines decrease FDI in
2There are 18 core disciplines, which include tariff liberalization for industrial and agriculture
goods, technical barriers to trade (TBT) and sanitary and phythosanitary (SPS) measures, export
taxes and anti-dumping and countervailing measures, trade related intellectual property (TRIPs)
and trade related investment measures (TRIMs), movement of capital, state owned enterprises,
state aid, competition policies, intellectual property rights (IPR), investment, public procurement
and services.
107
production-related activities by 48 and 110 percent respectively, while the remain-
ing disciplines have no significant impact. Production activities tend to employ
relatively more low-skill workers and are more polluting than services activities, two
characteristics that make stringent labor market regulations and environmental laws
more binding.3 Finally, we find that our discipline level results are driven by busi-
nesses between developing and developed economies for which the prior regulatory
gap is the widest.
The results on services and development level are consistent with what the
GVC literature has found.4 One of the reasons for these mirroring responses to
deep PTAs is the catalyst role that FDI plays on GVC integration by providing
foreign capital and technical know-how. Research has demonstrated that this high
degree of complementarity between GVCs and FDI is enhanced by policies, such
as those included in deep PTAs, providing support to the entry and upgrading in
GVCs.5
4.3 Related literature
The economic effects of PTAs have been thoroughly studied in the literature.
Researchers have looked at their impact on trade flows, investment, growth and
welfare. Our paper relates to the strand that studies the role of trade agreements in
3On the contrary, service activities tend not to be adversely affected by labor and environmental
disciplines because of the high-skill intensity and intangibility of such activities.
4Using sectoral data on value added trade, [Laget et al., 2018] find that deep PTAs matter more
for GVC trade in services and have a larger impact for trade in intermediates between developed
and developing countries.
5See the studies on FDI spillovers in the context of GVCs [Amendolagine et al., 2017] and [Farole
and Winkler, 2014].
108
increasing cross-border investments through different channels. One potential chan-
nel is driven by the complementarity between trade and investment that is based on
the production structure of multinational enterprises.6 [Markusen and Maskus, 2002]
provides a comprehensive framework to analyze and understand these mechanisms.
A core argument is that, by reducing trade barriers, PTAs facilitate the exchange of
inputs (tangible or not) and therefore stimulate investments along the value chain.
An important implication is that the trade creation and trade diversion effects of
PTAs may translate into FDI relocation, [Baltagi et al., 2008] and [Tintelnot, 2017]
provide empirical and counterfactual evidencefor this mechanism.
Another channel linking PTAs with FDI comes from the liberalization of ser-
vice, investment and other behind-the-border disciplines. [Dee and Gali, 2005] show
that PTAs in general have a significant impact on investment flows through their
non-trade disciplines. [Bu?the and Milner, 2008] offer a political economy perspec-
tive and argue that PTAs serve as commitment mechanisms to foreign investors
regarding the treatment of their assets that is more credible than domestic reg-
ulations. [Osnago et al., 2015] find that deep trade agreements facilitate vertical
FDI because they reduce the contractual uncertainty associated with the difference
between PTA members? institutions.
In terms of methodology, recent works have grasped the importance of ac-
counting for the variation in content to distinguish the effects of deep versus shallow
PTAs. Advances in data availability regarding the content of PTAs have enabled re-
6The relationship between trade and FDI is complex and both ?tariff jumping? substitutability
and ?value chain? complementarity motives are supported by the theoretical analysis. [Fontagne?,
1999] bring empirical evidences for the dominance of the complementarity relationship.
109
searchers to depart from the use of restrictive dummy variables and incorporate more
sophisticated measures of PTA depth. Equipped with these new empirical tools,
some papers have focused on the impact of PTA depth on trade flows, while others
have addressed the same question for FDI. [Mattoo et al., 2017] find that deep trade
agreements increase gross trade and can have a positive spillover effect on trade with
third countries if their design and implementation are non-discriminatory. [Orefice
and Rocha, 2014] study trade in parts and components, and [Laget et al., 2018] study
trade in value added, both finding that deep PTAs increase countries? participation
in GVCs by enhancing regulatory frameworks and easing cross-border operations.
On the FDI side, [Lesher and Miroudot, 2006] construct an index of investment
disciplines in PTAs but limit their coverage to 24 North-South agreements. Several
papers have attempted to build similar type of indices based on careful analysis of
the information contained in agreements? treaties. However, conducted this exercise
on the universe of PTAs as well as their overall content, until the World Bank?s
Content of PTA database became available. [Osnago et al., 2015] are the first to
use this dataset to construct a comprehensive measure of PTA depth and use it to
explore the impact of deep integration on vertical investments.
Several papers analyze the impact of specific policy areas on foreign invest-
ments (whether applied domestically or multilaterally). Examples are: [Helpman,
1992], [Ferrantino, 1993], and [Mansfield and Lee, 1996] for the impact of Intellec-
tual Property Rights; [Javorcik and Spatareanu, 2005] for labor market regulations;
and [Hanna, 2010] for environmental laws. However, by approaching each of these
disciplines separately, this body of research omits the important role provided by a
110
broad PTA framework in shaping the exchange of goods, services, and investments.
We fill this gap by exploring the role of relevant disciplines for FDI within the frame-
work of deep PTAs. Finally, our paper also relates to the strand of literature that
studies the determinants of FDI involving developing countries. [Go?mez-Mera et al.,
2014] show how PTAs channel market seeking FDI from developing countries. Sim-
ilarly to our results, [Medvedev, 2012] finds that the positive relationship between
FDI and PTA is driven by North-South investment.
Our quantitative analysis is performed using a gravity model of investment to
test first the impact of overall integration, and then the role of specific disciplines, in
promoting cross-border investment flows. Following cutting-edge research on gravity
models of trade, we use a Pseudo Poisson Maximum Likelihood (PPML) strategy
to address the methodological challenge of the large number of zeros in bilateral
investment data. To the best of our knowledge, this paper is the first to analyze the
role of specific disciplines on foreign investment within the context of a PTA. Our
large time and country coverage, as well as activity-level information, enable us to
shed light on important heterogeneity in our results.
The rest of the paper is organized as follows; Section 2 introduces the data used
to identify the content of PTAs and the data used to measure bilateral investment;
Section 3 describes the methodology used for our empirical analysis; Section 4 dis-
cusses our findings, Section 5 presents robustness checks and addresses endogeneity
concerns; and Section 6 concludes.
111
4.4 Data
In this section, we discuss the data used to measure the depth of a trade
agreement and describe in detail our source for data on bilateral investments at the
sector and activity level.
4.4.1 Preferential trade agreements
We first estimate the overall effects of deep trade agreements by using a new
measure of depth. We then estimate the effects of specific disciplines included in
PTAs. Our variables of depth and disciplines come from the first version of the
World Bank database on the Content of PTAs.
For our baseline measure of depth, we simply follow previous works that use
this content dataset or other similar sources, by counting the number of disciplines
included in a PTA. We focus on the number of legally enforceable disciplines as
they are expected to be more impactful than disciplines that are mentioned in
treaties without any sign of substantial co?mmitment from the parties. Specifically,
we define the variable TotalDepth = 52 kijt k=1Discijt - i.e. the simple count of
dummy variables for legally enforceable disciplines Disckijt included in the agreement
between country i and j at time t. Once an agreement is ratified, its content is not
expected to change unless there is an enlargement in the future. The variable takes a
value of zero before an agreement is signed and then turns on to the above sum after
the signature, and remains the same until the end of the sample period. Therefore,
the identification of the impact of depth variable comes from time variation within
112
each pair of countries
Another way to measure depth is to separate the disciplines between those
that fall under the current mandate of the WTO (such as tariffs, customs and anti-
dumping) and those that go beyond this mandate and are not subject to any kind
of WTO agreements (examples are agreements on investment, competition policy,
labour market laws, or environmental regulations). Following [Horn et al., 2010]
we call the former measure?of depth WTO-plus and the latter WTO-X and define
them as: WTOplus = 14
?
ijt p=1 Disc
p
ijt and WTOextraijt =
38
x=1 Disc
x
ijt, where
discpijt are 14 WTO+ disciplines and Disc
x
ijt are 38 WTO-X disciplines included in
an agreement between countries i and j in year t (see tables (A.1) and (A.2) for
the list of disciplines).
In the second stage of our analysis, we use dummy variables to identify the
commitment towards individual disciplines in PTAs. While including these dummies
sequentially, we also make sure that we control for the rest of the PTA content by
including a measure of the ?remaining? depth as the sum of disciplines except the
one singled out in the specification. We sequentially test all the disciplines mapped
in the database, eventually restricting ourselves only to those having a significant
impact on FDI. As shown in figure (4.8), the selected disciplines are also the most
frequent WTO-X areas included in PTAs.
113
4.4.2 Investments
In this section we describe the data source for our dependent variable measur-
ing cross-border investments. Reliable investment data is not as readily available
as trade data. When available, the data usually come from surveys conducted by
central banks or statistical offices and are used to comply with reporting obliga-
tions, such as balance of payments or international investment position statistics.
Such data reporting often lacks uniformity across countries, which undermines its
quality.7 Also, investment data are rarely available at the bilateral sectoral level.
To circumvent these limitations, we use investments from the fDi Markets database
collected by fDi Intelligence, a division of Financial Times. The data is available
at the firm-project level and corresponds to announcements of ?Greenfield FDI in
a new physical project or expansion of an existing investment, which creates new
jobs, and capital investment?. The announcements are collected from publicly avail-
able sources, such as media sources, industry and investment promotion agencies, or
market search and publication companies. Each project identified is cross-referenced
against multiple sources.
One concern is that recorded announcements differ in their advancement. The
dataset includes (but does not clearly distinguish) FDI projects that have been
announced or opened by a company, which either means that firms have made their
final decision and projects are moving towards implementation in the former case, or
7Despites the international reporting practices set by the IMF, countries tend to i) deviate from
the 10 percent ownership threshold used to define FDI, ii) not use uniform industry classification,
or iii) not report all types of investment properly (short-term intra-company loans, re-invested
earnings are often missing).
114
are already fully operational in the latter.8 Yet, to the extent possible, project status
is updated if a company makes further announcements. Large announced projects
above $1bn are researched on a quarterly basis to verify their progress. If the
project is cancelled, it is removed from the database. Working with announcements
represents an important advantage for our identification strategy in that it reduces
the scope for endogeneity caused by simultaneity.9 fDi Intelligence mentions that
it takes on average two years for a project to materialize after an announcement is
made in the media. Hence, regressing announcements on PTA variables at time t
is equivalent to regressing investment disbursement on lagged PTA variables, which
is one of the common techniques used when there is concern about endogeneity.
Another advantage of using the fDi Markets database is the availability of company-
level investment data. This allows us to address endogeneity concerns related to
the fact that large investment decisions may determine the adherence to certain
disciplines in PTAs by host countries, which would imply that it is large investments
that determine PTAs rather than the opposite. This is why, in the final part of our
analysis, we remove the largest investments from the sample, as well as the sectors
that spend the most resources in lobbying for specific content in PTAs, to check
robustness of results to that potential source of endogeneity.
Overall, the data covers more than 60,000 companies from 177 source countries
investing in 162 destination countries from 2003 to 2015. Each company?s project is
8The data does not cover merger and acquisition or other equity and non-equity investments.
Typically, projects are captured at the announcement stage for capital-intensive projects and at
the opening stage of services operations with limited capital investment required. Capital-intensive
projects take on average 2 years to become operational.
9Recent academic research using fDi Markets data includes: [Crescenzi et al., 2013]; [Paniagua
and Sapena, 2014]; [Castellani and Pieri, 2016]; [Amoroso et al., 2015]; [Antonietti et al., 2015]
115
classified according to 39 industries and 18 business activities based on end-users.
Table (4.9) shows how the different sectors and activities overlap in the data. Busi-
ness activities are defined as the actual function of a project, whereas sectors are
based on the receiving company?s core business area. Examples of project?s an-
nouncements are: Wiseway Group (Australia ? a transportation services provider)
is investing in China in the Transportation industry in a project related to Sales,
Marketing and Support activity; ProLogis (United States ? a leasing and prop-
erty management company) is investing in Sweden in the Real Estate industry in a
construction project. This means that an industry traditionally classified as man-
ufacturing may receive investments to develop a service-related activity and vice
versa. In the data for example, 65 percent of the investment received by the ?con-
sumer electronics? sector is dedicated to production activities but 16 percent falls
in retail, 5 percent in research and development, 4 percent in marketing, 2 percent
in logistics, distribution and transportation, and so on (all service-like activities).
We think that these activities best describe operations along the supply chain.
In figure (4.3), we compare our source for bilateral FDI announcements data
with bilateral FDI data from UNCTAD. Even though the correlation fluctuates
over time, on average our chosen measure of FDI tracks official data on cross-border
investments relatively well. We explain the discrepancy by the fact that our data
is not based on declarations of firms? balance sheets and therefore does not record
the full set of balance of payment flows, which implies that our constructed measure
of inward (or outward) FDI cannot perfectly match official measures. First, inward
and outward notions require a reporting country, which is not the case for the fDi
116
markets data as its compilation is performed by experts from the Financial Times.
Second, outward (inward) flows are computed by netting out any transactions that
decrease the stake of resident (foreign) investors in foreign (resident) enterprises
from transactions that increase it.
A crucial stage of the empirical work resides in dealing with the absences of
announcements. As acknowledged in the trade literature, zero trade (or investment)
flows do not occur randomly, which means that samples restricted on positive values
may yield biased estimates. It is particularly important to account for zero flows
when studying the effects of deep integration. The nature of our investment data
necessitates assumptions concerning the presence of zero investment flows. Contrary
to officially reported trade or investment data, we cannot apply the mirror method
to complete missing observations, and lack of announcements in the news can be
left to interpretation. Nevertheless, we assume that if a significant investment ever
materializes, it will be covered in the news or announced in some other ways that
fDi markets will be able to identify. We generate zero investment flows by doing
the following: starting from the sample of countries included in the fDi markets
database, we generate all possible source/destination pairs across the sample period
and replace missing observations by zeros. We then keep only the pairs (for the
whole period) that ever invested at some point during the period.10
Finally, we do not make assumption regarding the motives of investments.
The data and its coarse industry/activity classification do not allow us to distin-
10This way, at the aggregate activity level our estimation sample is composed of 35 percent
(38,300) non-zero against 65 percent (71,033) zero observations for FDI announcements.
117
guish between vertical and horizontal FDI. Even though the theoretical literature
has long distinguished market seeking (i.e. horizontal) FDI and efficiency seeking
(i.e.vertical) FDI ( [Markusen, 1984], [Helpman, 1984], recent works have shed light
on the complex strategies of multinationals. In this paper we study the impact of
deep integration on the intention to invest in Greenfield projects in general.
4.4.3 Control variables
We control for existence of expired trade agreements by including a corre-
sponding dummy variable. Past agreements are taken from Mario Larch?s Regional
Trade Agreements Database. We also control for bilateral investment treaties to
isolate the effect of investment disciplines included in PTAs only. Finally, we use
fixed effects to minimize the scope for endogeneity. We initially include country-
pair, source-time and destination-time fixed effects. When we move to the business
activity level, we control additionally for pair-industry, source-industry-time and
destination-industry-time fixed effects.
4.5 Effect of PTAs on FDI
In this section we introduce the gravity model used to estimate the effect of
the depth of a PTA on FDIs.
118
4.5.1 Empirical strategy
The gravity model is the workhorse estimation technique for applied interna-
tional trade analysis. This model has both empirical and theoretical advantages;
it predicts trade flows well and can be derived from a large class of structural
general equilibrium trade models. Theoretical studies of multinationals and trade
( [Markusen and Venables, 1998, Markusen and Venables, 2000], [Egger and Pfaf-
fermayr, 2000]) have found that both types of activity are determined by the same
exogenous factors (distance, market size, trade/investment barriers). Despite the
lack of micro-founded theory to fit FDI patterns from gravity determinants (as [An-
derson and van Wincoop, 2003] does for trade flows), the gravity model is a successful
tool for predicting FDI flows and has frequently been used in the literature.
In our estimation strategy, we follow the best practices for estimating a struc-
tural gravity model. We use directional country-time fixed effects to account for the
multilateral resistance terms. These fixed effects absorb any country-specific charac-
teristics that may vary with time, such as national policies, institutions and exchange
rates. Because we use panel data, we can incorporate country-pair fixed effects to
account for time invariant differences between origin and destination countries, such
as factor endowments, that can predict FDI flows. By including this extensive set
of fixed effects, we alleviate the risk of endogeneity caused by omitted variables.
Finally, we rely on the Pseudo Poisson Maximum Likelihood (PPML) estimator,
which has been commonly adopted to estimate gravity equations. [Santos Silva and
Tenreyro, 2006] shows that this non-linear estimator produces consistent estimates
119
in the presence of heteroskedasticity while accounting for the large number of zero
flows.
Initially, we estimate our baseline specification using both OLS and PPML
methods to show that both estimators lead to similar results.
OLS : log(FDIijt) = ?1Depthijt +Bcontrolsijt + ?ij + ?it + ?jt + ijt (4.1)
PPML : FDIijt = exp (?1Depthijt +Bcontrolsijt + ?ij + ?it + ?jt) + ijt (4.2)
where FDIijt measures greenfield investments between country i and j at time t,
Depthijt is a measure of PTA depth, controlsijt are dummy variables accounting
for present and past PTA agreements (PTAijt and pastPTAijt respectively), and
bilateral investment treaties (BITijt); and ?ij, ?it, and ?jt represent respectively
country-pair industry, reporter-industry-time and partner industry time fixed ef-
fects.
4.5.2 Baseline results
Table (4.2) reports the coefficients of total depth, WTO-plus and WTO-extra
variables for both OLS and PPML estimators. The first three columns columns are
estimated with OLS and show that the total depth of PTAs matters for investment,
whereas a shallow measure of PTAs using only a dummy variable does not have a
significant effect. The bilateral investment treaty dummy is not significant, which
is also the case in the literature when the effects of BITs and PTAs are estimated
concurrently. Moving to the best practices for the gravity estimation, which rec-
120
ommend using PPML, we find that the total depth variable is also significant and
positive when using the PPML estimator. One additional discipline increases FDI
announcements by 1.4 percent. Splitting the depth of the PTAs between WTO-plus
and WTO-extra disciplines as done in columns (4) and (6) of table (4.2), we find
that the disciplines that go beyond the WTO?s mandate drive the positive effects.
In table (4.3), we test whether deep PTAs matter for FDI in extractive activities.
The interaction with a dummy for extraction shows that deeper agreements do not
promote FDI in extraction as opposed to services and production activities. This
confirms that deep PTAs really apply to more complex activities that are more
rules-intensive. Consequently, for the rest of our analysis we focus on production
and service activities only. These baseline results are in line with results from similar
specifications regressing GVC trade on the total depth of PTAs (see [Laget et al.,
2018]). They illustrate the fact that investment and value-added trade are two man-
ifestations of the same GVC activities. In the next section we further investigate
the effects of specific disciplines.
4.5.3 The impact of single disciplines
The data on the content of PTAs allows us to isolate the effect of single disci-
plines on investment while controlling for the overall depth. The following specifi-
cation estimates the effect of investment related disciplines while accounting for the
rest of a PTA depth:
( )
FDIijt = exp ?1Depth
?/d
ijt + ?dDisc
d
ijt +Bcontrolsijt + ?ij + ?it + ?jt + ijt (4.3)
121
where Discdijt is a dummy variable that indicates whether the discipline d is included
in an agreement. The variable Depth?/dijt controls for the rest of the agreement content
summing across all other disciplines except d. Table (4.4) reports the results of
the PPML estimation at the discipline level. Commitments relative to intellectual
property rights (IPR) increase FDI announcements by 36 percent. When included in
PTAs, IPR protect investors? main assets (brand or innovation efforts) and guarantee
secured returns on those assets. Visa and asylum increases FDI announcements
by 43 percent shedding light on the importance of facilitating the movement of
persons to promote business activities. On the other hand, labor market regulations
are negatively correlated with FDI and decreased announcements by 80 percent.
The main objective of labor disciplines is not to increase market access but rather
to improve social welfare by fostering workers? bargaining power, which does not
liberalize the business environment.
At this point, it is worth commenting on the magnitude of the coefficients of
the disciplines in table (4.4) relative to the coefficient of the total depth variable,
which is computed as the sum of discipline dummies. Some of these disciplines
have a significant and positive impact, while others have a negative impact or no
impact at all on FDI. For this reason, summing all the individual effects is netting
out positive and negative impacts and results in a much smaller absolute coefficient
for total depth. Moreover, accounting for the difference in range and variance of
the aggregate depth vs. disciplines, we find that a one standard deviation increase
in total depth is associated with a 0.04 standard deviation increase in FDI and a
one standard deviation increase in IPR is associated with a 0.03 standard deviation
122
increase in FDI.
The next estimation studies the potential heterogeneity in the effect of disci-
plines across business activities. We create a dummy variable that identifies services
activities. Services refers to any activity that is neither production nor extractive
that we interact with the discipline dummy. Since we have excluded extractive ac-
tivities from the rest of the analysis, the control group represents the production
activities. Table (4.9) summarizes the distribution of the FDI announcements across
business activities and shows that our sample is biased towards services (with 53 per-
cent of the outstanding amount of announcements for services projects, 37 percent
for manufacturing, and 9 percent for extractive). The new specification is written:
?
FDI = exp(? Depth?/d + ? Discd + ? Discdijt 1 ijt d ijt d ijt ? Services
(4.4)
+Bcontrolsijt + ?ij + ?it + ?jt) + ijt
Table (4.5) reports the results of the interactions with service-related activities.
Our eight disciplines of interest turn out to be positively correlated with announce-
ments of projects related to service activities. Service activities, that are skill- and
knowledge-intensive, benefit the most from the reduction of uncertainty conferred
by PTAs. As for environmental laws, when production and service activities are
combined, this discipline is not significant, but separating the effects on each activ-
ity reveals that such discipline is negatively correlated with the intention to invest in
production activities. This might be because environment policy areas cover com-
mitments on Good Manufacturing Practices that ?constraint? mostly production
123
processes. We would need measurement on the vertical depth to test this hypothesis.
The opposite effects of labor market regulations found for services and production
can be explained by the difference in workers? skill employed in those two types of
activities. On top of ILO labor standards (prohibition of child labor, respect of
human right, and so on.), labor disciplines in PTAs cover (and are not limited to)
right to collective bargaining, freedom of association, minimum wages, unemploy-
ment benefit, cost of firing. Such regulations tend to bind for low-skilled workers
and hence refrain investors from investing in production-like activities, which tend
use more unskilled workers than services.
As mentioned in the introduction, business activities can be viewed as ?tasks?
performed along the supply chain, and are different from standards industrial clas-
sifications. Therefore, significant effects on services activities but none on manu-
facturing activities should not be interpreted as an absence of impact on goods.
Indeed, products from the textile sector have attracted 18 percent of investment on
manufacturing activities, but also 77 percent on retail services activities.
4.5.4 Level of development
The content of trade agreements varies greatly with the level of development
of signatories and disciplines may matter differently for developed or developing
countries. We first investigate the interaction between the overall PTA depth and the
level of development of partner countries. We then use the same type of interaction
to understand how the effects of disciplines varies with the development level. The
124
regression specification is as follows:
?
FDIijt = exp(? Depth
?/d
1 ijt +Depth
?/d
ijt ?Devij + ?dDiscdijt + ?dDiscdijt ?Devij
+Bcontrolsijt + ?ij + ?it + ?jt) + ijt
(4.5)
Where Devij is a vector of three dummies of the possible North-North, North-
South and South-South country pairs. North is defined as the group of high-income
countries while South comprises low- and middle-income and LDC countries.
Results reported in table (4.6)show that North-South agreements are driving
our results on overall depth. Without interaction we found in the baseline spec-
ification that an additional discipline increase FDI by 0.14 percent regardless of
the development level of the member countries. This effect rises to 0.24 percent for
FDI between developed and developing countries. Moreover, investment, IPR, labor
market regulations and movement of capital have a bigger positive and significant
impact on FDI between North-South than pairs of similar income. This sheds light
on the role that such disciplines can play to enhance the investment climate and
provide stability, two factors that are critical when prior institutional gap is the
widest as in between developed and developing countries.
The absence of significant effect of South-South agreements (at both overall
and discipline levels) is due to the fact that such PTAs rarely included WTO?X dis-
ciplines and hence do not significantly commit on investment related issues. On the
contrary, North-North agreements often include investment-related disciplines, but
125
institution levels being already similar across high-income countries, commitment
established in PTAs do not influence the pattern of FDI between these countries.
4.6 Robustness checks
4.6.1 Composition concerns
There is evidence supporting the existence of templates for the content of
trade agreements, which are championed by the US and the EU, the two major
economies leading active regional integration agendas (see [Horn et al., 2010]). This
raises concerns about the validity of our results and questions the existence of a
composition effect.
To test whether our findings are driven by the presence of such templates,
we run our specifications by sequentially removing countries from the estimation
samples. We find that both baseline and discipline level estimations remain the
same, ruling out any role played by a particular economy in our results.
Similar concern might be raised with respect to China and its weight in North-
South relationships. Table (4.7) shows that the results we obtained by interacting
disciplines with development levels remain the same after we single out China from
the North-South interacted terms.
4.6.2 Simultaneity concerns
We now address the concerns of endogenous trade agreements. As already
mentioned in the data description section, the use of announcements (instead of
126
actual disbursements) reduces the scope for endogeneity. The delay between an-
nouncements and actual disbursements (that amounts to two years on average)
implies that regressing announcements on PTAs at time t is equivalent to regressing
actual FDI on PTAs at time t? 2 on average, which is a common modification un-
dertaken in the empirical literature. The lag between dependent and independent
variables reduces the scope for investors to influence the content of PTAs.
Nevertheless, the two-year lag is an average and is not directly reported by
fDi markets, leaving room for some endogeneity concerns due to simultaneity. The
main concern is that influential investors might still lobby for specific content in a
particular PTA. We can test whether the presence of ?big players? drives our main
results. There are two possibilities to consider, i) firms might not be organized
and only the biggest will engage in lobbying for trade policies, and ii) firms are
organized and are represented by lobbyists at the industry-level. We can test these
two hypotheses thanks to the project-level information available for our investment
data.
For the first scenario, we proxy big firms by those having the largest aggregated
announcements across the period at the country pair level. We argue that removing
the top one percent should leave the effect of PTAs exogenous to the remaining
smaller firms. Table (4.8) confirms that our baseline and discipline level results are
robust to removing the top 1 percent of announcements.
We then move to the industry level lobbying hypothesis. To determine which
industries lobby most for trade-related issues, we rely on the US Lobbying Disclo-
sure Act of 1995, which requires any lobbying activities engaged within the US to
127
be disclosed.11 The Center for Responsive Politics publishes the full reports of the
disclosed lobbying activities indicating the client name (including foreign entities),
lobbyist name, contribution level and year, and more interestingly industry classifi-
cation and the general issue that is lobbied for. Among the possible lobbied issues,
we focus on the ones that might relate to the content of PTAs, including: tariffs,
trade, labor antitrust and workplace, environment, copyright patent and trademark.
Table (4.9) gives the level of contributions of each lobbied issue by sector and rel-
ative to the sector-level total contribution in 2003. The sectors that are relatively
more engaged in PTA-related lobbying turn out to be Health, Construction and
Defense. Table (4.10) shows that our results are robust to removing these three
lobbying intensive sectors from our estimation sample.
4.7 Conclusion
This paper contributes to the existing literature on the relationship between
PTAs and FDI. Our baseline results show that on average adding a discipline to a
deep PTA increases FDI by 1.4 percent. In general, WTO-extra disciplines matter
more for FDI than WTO-plus. All our disciplines of interest (Investment, IPR, Visa
and Asylum, Movement of Capital, Competition policy, Labor market regulations
and Environmental laws) promote FDI in service-related activities. Concerning
production-related activities, only visa and asylum has a positive and significant
impact on FDI, while the inclusion of Labor market regulations and Environmental
11The US Lobbying Disclosure Act of 1995 contributes to the ?public awareness of the efforts of
paid lobbyist to influence the public decision making process in both the legislative and executive
branches of the Federal Government?
128
laws reduces FDI. We also find that discipline level results are mostly driven by
investments between developed and developing countries, where the institutional
gap is the widest. Finally, we address causality concerns by ruling out the potential
influence of lobbying activities at both the firm and industry levels.
129
4.8 Figures
Sources: Author?s calculation based on content of trade agreement
database
Figure 4.1: BITs and investment related disciplines in PTAs
130
Movement	of	Capital	
Competition	Policy	
Investment	
IPR	
Visa	and	Asylum	
Labor	Market	Regulation	
Environmental	Laws	
Statistics	
Energy	
Agriculture	
Social	Matters	
Research	and	Technology	
Industrial	Cooperation	
Approximation	of	Legislation	
Taxation	
Illegal	Immigration	
Anti-Corruption	
Financial	Assistance	
Education	and	Training	
Consumer	Protection	
Economic	Policy	Dialogue	
Cultural	Cooperation	
Data	Protection	
Public	Administration	
Mining	
Regional	Cooperation	
Nuclear	Safety	
SME	
Terrorism	
Information	Society	
Illicit	Drugs	
Health	
Audio	Visual	
Money	Laundering	
Human	Rights	
Innovation	Policies	
Civil	Protection	
Political	Dialogue	
0.00	 0.10	 0.20	 0.30	 0.40	 0.50	
Figure 4.2: Frequency of inclusion of investment related disciplines in PTAs
131
2003 2005 2007 2009 2011 2013 2015
Year
Investment vs. FDI inward Investment(t-1) vs. FDI inward
Investment(t-2) vs. FDI inward
Figure 4.3: Correlation between fDi markets and UNCTAD bilateral investment
data
132
.4 .5 .6 .7 .8
4.9 Tables
18 Business Activities 39 Industries
Business Services Aerospace Hotels & Tourism
Construction Alternative/Renewable Industrial Machinery,
energy Equipment & Tools
Customer Contact Centre Automotive Components Leisure & Entertainment
Design, Development & Automotive OEM Medical Devices
Testing
Education & Training Beverages Metals
Electricity Biotechnology Minerals
Extraction Building & Construction Non-Automotive Transport
Materials OEM
Headquarters Business Machines & Paper, Printing & Packag-
Equipment ing
ICT & Internet Infrastruc- Business Services Pharmaceuticals
ture
Logistics, Distribution & Ceramics & Glass Plastics
Transportation
Maintenance & Servicing Chemicals Real Estate
Productiona Coal, Oil and Natural Rubber
Gas
Recycling Communications Semiconductors
Research & Development Consumer Electronics Software & IT services
Retail Consumer Products Space & Defense
Sales, Marketing & Support Electronic Components Textiles
Shared Services Centre Engines & Turbines Transportation
Technical Support Centre Financial Services Warehousing & Storage
Food & Tobacco Wood Products
Healthcare
Table 4.1: Business activities and industries included in the classification of fDi
Markets
afDi markets uses the term ?manufacturing?, in order to avoid any confusion with the traditional
meaning of manufacturing in the trade and investment literature, we decided to replace it by
?production? which describes the activity in terms of the tasks performed rather than its industrial
sector.
133
OLS PPML
VARIABLES log(FDI announcements) FDI announcements
(1) (2) (3) (4) (5) (6)
PTA 0.0345 0.0664 -0.0823 0.173 -0.205 -0.174
(0.0699) (0.0877) (0.126) (0.264) (0.130) (0.243)
expired PTA 0.100 0.146 0.192 0.174 0.179
(0.125) (0.128) (0.137) (0.175) (0.181)
BIT -0.00306 0.00985 0.00132 -0.109 -0.110
(0.0661) (0.0668) (0.0670) (0.0684) (0.0689)
Total Depth 0.00868* 0.0143**
(0.00524) (0.00578)
WTO-X 0.0163* 0.0155*
(0.00857) (0.00865)
WTO-plus -0.0190 0.0107
(0.0254) (0.0233)
Observations 33,622 33,622 33,622 33,622 109,383 109,383
R-squared 0.619 0.619 0.619 0.619 0.800 0.800
Country-pair FE Yes Yes Yes Yes Yes Yes
Country-time FE Yes Yes Yes Yes Yes Yes
Business activities All All All All All All
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Table 4.2: Impact of total depth on investment
134
PPML
VARIABLES FDI announcements
(1) (2)
Total depth 0.0142**
(0.00577)
Total depth ? extraction 0.00873
(0.0128)
WTO-X 0.0169*
(0.00878)
WTO-plus 0.00765
(0.0235)
WTO-X ? extraction -0.0145
(0.0568)
WTO-plus ? extraction 0.0229
(0.0397)
BIT -0.109 -0.111
(0.0684) (0.0689)
PTA -0.217* -0.167
(0.132) (0.243)
expired PTA 0.167 0.174
(0.175) (0.180)
Observations 109,383 109,383
R-squared 0.800 0.800
Country-pair FE Yes Yes
Country-time FE Yes Yes
Business activities All All
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Table 4.3: FDI in extractive activities are not driven by PTAs
135
VARIABLES Dependent variable: FDI announcements, PPML estimation
Visa Environment Labor Movement Competition
xyz disciplines: Investment IPR and laws market of policy
asylum regulations capital
(1) (2) (3) (4) (5) (6) (7)
Total depth (xyz exculded) 0.0116* 0.00690 0.00111 0.0272*** 0.0366*** 0.0136* 0.0192***
(0.00691) (0.00648) (0.00743) (0.00827) (0.00653) (0.00727) (0.00708)
xyz discipline 0.224 0.361*** 0.430*** -0.302 -0.818*** 0.0665 -0.142
(0.210) (0.130) (0.143) (0.188) (0.145) (0.181) (0.221)
Observations 103,953 103,953 103,953 103,953 103,953 103,953 103,953
R-squared 0.799 0.800 0.800 0.799 0.801 0.799 0.799
Country-pair FE Yes Yes Yes Yes Yes Yes Yes
Country-time FE Yes Yes Yes Yes Yes Yes Yes
Business activities P&S P&S P&S P&S P&S P&S P&S
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Table 4.4: Disciplines level estimations
VARIABLES Dependent variable: FDI announcements, PPML estimation
Visa Environment Labor Movement Competition
xyz disciplines: Investment IPR and laws market of policy
asylum regulations capital
(1) (2) (3) (4) (5) (6) (7)
Total depth (xyz discipline excluded) 0.0115* 0.00723 0.000498 0.0270*** 0.0362*** 0.0139* 0.0192***
(0.00690) (0.00646) (0.00748) (0.00821) (0.00654) (0.00731) (0.00718)
xyz discipline 0.0402 0.178 0.260 -0.485** -1.100*** -0.119 -0.307
(0.226) (0.152) (0.167) (0.195) (0.194) (0.196) (0.242)
xyz discipline ? service 0.340** 0.345** 0.321* 0.372** 0.511** 0.367** 0.334**
(0.158) (0.171) (0.177) (0.185) (0.256) (0.174) (0.162)
Observations 103,953 103,953 103,953 103,953 103,953 103,953 103,953
R-squared 0.800 0.801 0.800 0.800 0.801 0.800 0.800
Country-pair FE Yes Yes Yes Yes Yes Yes Yes
Country-time FE Yes Yes Yes Yes Yes Yes Yes
Business activities P&S P&S P&S P&S P&S P&S P&S
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Table 4.5: Disciplines level estimations interacted with services
136
VARIABLES Dependent variable: FDI announcements, PPML estimation
Total Visa Environment Labor Movement Competition
xyz disciplines: depth Investment IPR and laws market of policy
asylum regulations capital
(1) (2) (3) (4) (5) (6) (7) (8)
Total depth (without xyz) 0.00534 0.00256 0.0248*** 0.0371*** 0.0125* 0.0145** 0.0235***
(0.00680) (0.00797) (0.00840) (0.00659) (0.00729) (0.00583) (0.00744)
discipline*South-South -0.00837 -0.108 -0.229 0.683 -0.784 -0.353 0.178 0.333
(0.02190) (0.38000) (0.31100) (0.52500) (0.53900) (0.45100) (0.27200) (0.28400)
discipline*North-North -0.00184 0.0947 0.0914 -0.561*** -0.793*** -0.415** -0.0515 -0.699***
(0.00693) (0.13200) (0.29900) (0.21400) (0.16600) (0.20500) (0.10100) (0.25300)
discipline*South/North 0.0244*** 0.786*** 0.489*** -0.0235 -0.884*** 0.388** 0.705*** 0.0433
(0.00663) (0.17900) (0.14300) (0.19400) (0.17200) (0.18600) (0.15200) (0.22700)
Observations 103,953 103,953 103,953 103,953 103,953 103,953 103,953 103,953
R-squared 0.8 0.801 0.8 0.8 0.801 0.801 0.801 0.8
Country-pair FE Yes Yes Yes Yes Yes Yes Yes Yes
Country-time FE Yes Yes Yes Yes Yes Yes Yes Yes
Business activities P&S P&S P&S P&S P&S P&S P&S P&S
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Table 4.6: Disciplines effects by level of development
VARIABLES Dependent variable: FDI announcements, PPML estimation
Total Visa Environment Labor Movement Competition
xyz disciplines: depth Investment IPR and laws market of policy
asylum regulations capital
(1) (2) (3) (4) (5) (6) (7) (8)
Total depth without discipline 0.00521 0.00194 0.0250*** 0.0374*** 0.0277*** 0.0125* 0.0234***
-0.00685 -0.00801 -0.00854 -0.00662 -0.0088 -0.00729 -0.00746
Discipline*HH -0.00154 0.0992 0.112 -0.563*** -0.794*** -0.838*** -0.391 -0.700***
-0.00693 -0.133 -0.303 -0.215 -0.166 -0.203 -0.467 -0.253
Discipline*LL(with China) 0.0514 0.586 0.252 - - 0.0445 -0.413** -
-0.0431 -0.514 -1.247 - - -0.504 -0.205
Discipline*LL(without China) -0.0153 -0.271 -0.264 0.676 -0.809 0.585 -0.148 0.334
-0.0239 -0.483 -0.305 -0.528 -0.538 -1.346 -0.41 -0.284
Discipline*LH(with China) -0.00254 0.553*** 0.0434 0.0609 0.341 0.391** -0.195 -0.0567
-0.0146 -0.187 -0.183 -0.189 -0.412 -0.186 -0.269 -0.243
Discipline*LH(without China) 0.0416* 1.068** 0.851*** -0.707 -0.109 -3.002*** 0.919** -0.286
-0.0224 -0.486 -0.272 -0.495 -0.526 -0.574 -0.421 -0.271
Observations 103,953 103,953 103,953 103,953 103,953 103,953 103,953 103,953
R-squared 0.801 0.801 0.8 0.8 0.801 0.801 0.802 0.8
Country-pair FE Yes Yes Yes Yes Yes Yes Yes Yes
Country-time FE Yes Yes Yes Yes Yes Yes Yes Yes
Business activities P&S P&S P&S P&S P&S P&S P&S P&S
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Table 4.7: Interaction with China in North-South agreements
137
VARIABLES Dependent variable: FDI announcements, PPML estimation
Total Visa Environment Labor Movement Competition
xyz disciplines: depth Investment IPR and laws market of policy
asylum regulations capital
(1) (2) (3) (4) (5) (6) (7) (8)
Total depth (d excluded) 0.0079 0.00289 -0.00438 0.0221*** 0.0256*** 0.00734 0.0120*
-0.00609 -0.00665 -0.00751 -0.0085 -0.00667 -0.00746 -0.00717
Discipline(d) 0.000947 -0.0527 0.0652 0.206 -0.605*** -0.966*** -0.124 -0.328
-0.00708 -0.13 -0.147 -0.162 -0.19 -0.207 -0.211 -0.225
Discipline(d)* services 0.0135** 0.283** 0.370** 0.333** 0.502*** 0.549** 0.370** 0.416***
-0.00605 -0.136 -0.165 -0.156 -0.172 -0.237 -0.161 -0.149
Observations 103,941 103,941 103,941 103,941 103,941 103,941 103,941 103,941
R-squared 0.805 0.805 0.806 0.805 0.806 0.806 0.806 0.806
Country-pair FE Yes Yes Yes Yes Yes Yes Yes Yes
Country-time FE Yes Yes Yes Yes Yes Yes Yes Yes
Business activities P&S P&S P&S P&S P&S P&S P&S P&S
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Table 4.8: Robustness check: removing top 1 percent of projects announcements
General Sector
Sector Contribution
issue percentage
Copyright, Patent & Trademark Health $39,800,000 16%
Copyright, Patent & Trademark Transportation $8,730,704 6%
Copyright, Patent & Trademark Construction $1,835,212 5%
Environment & Superfund Defense $9,905,470 4%
Environment & Superfund Energy/Nat Resource $7,355,459 4%
Environment & Superfund Construction $1,139,960 3%
Labor, Antitrust & Workplace Defense $6,871,012 3%
Labor, Antitrust & Workplace Construction $1,202,697 3%
Labor, Antitrust & Workplace Labor $1,068,717 2%
Tariffs Defense $6,130,000 3%
Tariffs Health $5,488,536 2%
Tariffs Transportation $2,516,683 2%
Trade Health $28,900,000 12%
Trade Defense $17,400,000 8%
Trade Transportation $7,459,139 5%
Table 4.9: Lobbying contributions by issues
138
VARIABLES Dependent variable: FDI announcements, PPML estimation
Total Visa Environment Labor Movement Competition
xyz disciplines: depth Investment IPR and laws market of policy
asylum regulations capital
(1) (2) (3) (4) (5) (6) (7) (8)
Total depth (without xyz) 0.0138** 0.00698 -0.000181 0.0257*** 0.0360*** 0.0134* 0.0193***
-0.00603 -0.00657 -0.0077 -0.00841 -0.00667 -0.00741 -0.00729
xyz discipline 0.00777 -0.0254 0.18 0.278 -0.454** -1.112*** -0.106 -0.317
-0.00764 -0.14 -0.159 -0.174 -0.206 -0.213 -0.201 -0.248
xyz discipline ? services 0.0128* 0.332** 0.358** 0.312* 0.371* 0.529* 0.372** 0.344**
-0.00675 -0.154 -0.178 -0.183 -0.195 -0.272 -0.182 -0.169
Observations 101,526 101,526 101,526 101,526 101,526 101,526 101,526 101,526
R-squared 0.799 0.799 0.8 0.799 0.799 0.801 0.799 0.799
Country-pair FE Yes Yes Yes Yes Yes Yes Yes Yes
Country-time FE Yes Yes Yes Yes Yes Yes Yes Yes
Business activities P&S P&S P&S P&S P&S P&S P&S P&S
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Table 4.10: robustness check: removing lobbying intensive sectors
139
140
Table 4.11: Industries and business activities partitioning
Activities
Sector:
outstanding
% row total
% column total
- 46,677 157 - 58 9,034 2,313 - 4,265 - 1,820 10,097 2 567 - 2,866 22 1
Aerospace - 59.9 0.2 - 0.1 11.6 3.0 - 5.5 - 2.3 13.0 0.0 0.7 - 3.7 0.0 0.0
- 1.2 0.0 - 0.3 4.2 8.4 - 2.6 - 0.3 37.9 0.0 0.5 - 0.6 0.1 0.0
Alternative - 52,281 48 - 12 478 48 513,332 1,239 - 508 107 204 1,059 519 56,056 - 2
& Renewable - 8.4 0.0 - 0.0 0.1 0.0 82.0 0.2 - 0.1 0.0 0.0 0.2 0.1 9.0 - 0.0
energy - 1.3 0.0 - 0.1 0.2 0.2 53.4 0.8 - 0.1 0.4 0.6 1.0 0.1 12.3 - 0.0
Automotive - 232,371 - - 43 8,329 459 - 3,934 - 5,119 739 50 3,054 771 1,236 110 107
Components - 90.7 - - 0.0 3.2 0.2 - 1.5 - 2.0 0.3 0.0 1.2 0.3 0.5 0.0 0.0
- 5.8 - - 0.3 3.9 1.7 - 2.4 - 0.8 2.8 0.2 2.9 0.1 0.3 0.7 0.7
Automotive OEM - 504,999 103 - 60 18,667 2,120 - 6,413 435 2,429 1,882 157 7,166 8,289 4,375 73 118
(Original Equipment - 90.6 0.0 - 0.0 3.3 0.4 - 1.2 0.1 0.4 0.3 0.0 1.3 1.5 0.8 0.0 0.0
Manufacturer) - 12.6 0.0 - 0.4 8.7 7.7 - 3.9 0.1 0.4 7.1 0.5 6.9 1.3 1.0 0.5 0.8
137 54,786 110 - 13 176 16 - 1,186 - 2,443 - 4 349 1,270 4,577 91 -
Beverages 0.2 84.1 0.2 - 0.0 0.3 0.0 - 1.8 - 3.7 - 0.0 0.5 1.9 7.0 0.1 -
0.0 1.4 0.0 - 0.1 0.1 0.1 - 0.7 - 0.4 - 0.0 0.3 0.2 1.0 0.6 -
- 19,054 76 - 18 801 69 - 1,385 - 205 10 - 9,345 - 1,112 4 12
Biotechnology - 59.4 0.2 - 0.1 2.5 0.2 - 4.3 - 0.6 0.0 - 29.1 - 3.5 0.0 0.0
- 0.5 0.0 - 0.1 0.4 0.3 - 0.8 - 0.0 0.0 - 9.0 - 0.2 0.0 0.1
Building - 128,699 - - - 264 25 - 533 63 998 - 8 43 973 68 6 -
& Construction - 97.7 - - - 0.2 0.0 - 0.4 0.0 0.8 - 0.0 0.0 0.7 0.1 0.0 -
Materials - 3.2 - - - 0.1 0.1 - 0.3 0.0 0.2 - 0.0 0.0 0.2 0.0 0.0 -
Business - 40,457 254 - 511 3,947 331 - 4,391 446 1,105 879 784 1,583 529 1,492 2,013 1,781
Machines - 66.9 0.4 - 0.8 6.5 0.5 - 7.3 0.7 1.8 1.5 1.3 2.6 0.9 2.5 3.3 2.9
& Equipment - 1.0 0.0 - 3.0 1.8 1.2 - 2.7 0.1 0.2 3.3 2.4 1.5 0.1 0.3 12.7 11.5
Business - 275 108,157 41,434 5,300 4,098 11,939 - 7,665 2,306 443 142 23,981 1,657 5,097 42 3,627 1,544
Services - 0.1 49.7 19.0 2.4 1.9 5.5 - 3.5 1.1 0.2 0.1 11.0 0.8 2.3 0.0 1.7 0.7
- 0.0 18.5 2.9 31.4 1.9 43.3 - 4.7 0.5 0.1 0.5 72.1 1.6 0.8 0.0 23.0 9.9
Ceramics - 45,892 - - 47 72 2 - 477 - 542 93 34 25 2,555 139 21 -
& Glass - 92.0 - - 0.1 0.1 0.0 - 1.0 - 1.1 0.2 0.1 0.1 5.1 0.3 0.0 -
- 1.1 - - 0.3 0.0 0.0 - 0.3 - 0.1 0.3 0.1 0.0 0.4 0.0 0.1 -
2,142 480,193 63 - 24 4,959 646 - 4,179 - 6,470 272 430 4,904 1,058 5,063 269 193
Chemicals Continued on next page
E x t r a c t i o n
P r o d u c t i o n
B u s i n e s s
S e r v i c e s
C o n s t r u c t i o n
C u s t o m e r
C o n t a c t C e n t e r
D e s i g n , D e v e l o p m e n t
& T e s t i n g
E d u c a t i o n &
T r a i n i n g
E l e c t r i c i t y
H e a d q u a r t e r s
I C T & I n t e r n e t
I n f r a s t r u c t u r e
L o g i s t i c s ,
D i s t r i b u t i o n &
T r a n s p o r t a t i o n
M a i n t e n a n c e &
S e r v i c i n g
R e c y c l i n g
R e s e a r c h &
D e v e l o p m e n t
R e t a i l
S a l e s , M a r k e t i n g &
S u p p o r t
S h a r e d
S e r v i c e s C e n t e r
T e c h n i c a l S u p p o r t
C e n t e r
141
Table 4.11 ? Continued from previous page
Activities
Sector:
outstanding
% row total
% column total
0.4 94.0 0.0 - 0.0 1.0 0.1 - 0.8 - 1.3 0.1 0.1 1.0 0.2 1.0 0.1 0.0
0.2 12.0 0.0 - 0.1 2.3 2.3 - 2.5 - 1.1 1.0 1.3 4.7 0.2 1.1 1.7 1.2
Coal, Oil 661,085 596,779 1,452 - 207 1,541 487 448,845 2,655 672 51,942 384 343 3,418 10,929 16,224 1,211 1
and Natural Gas 36.8 33.2 0.1 - 0.0 0.1 0.0 25.0 0.1 0.0 2.9 0.0 0.0 0.2 0.6 0.9 0.1 0.0
66.5 14.9 0.2 - 1.2 0.7 1.8 46.6 1.6 0.2 8.6 1.4 1.0 3.3 1.7 3.6 7.7 0.0
- 35,025 922 - 3,130 46,282 1,092 - 15,928 363,192 4,090 833 154 8,624 4,153 20,848 1,275 1,859
Communications - 6.9 0.2 - 0.6 9.1 0.2 - 3.1 71.6 0.8 0.2 0.0 1.7 0.8 4.1 0.3 0.4
- 0.9 0.2 - 18.6 21.6 4.0 - 9.7 83.2 0.7 3.1 0.5 8.3 0.6 4.6 8.1 12.0
Consumer - 45,436 18 - 161 1,905 562 - 2,920 - 1,571 390 541 3,070 10,966 2,467 68 142
Electronics - 64.7 0.0 - 0.2 2.7 0.8 - 4.2 - 2.2 0.6 0.8 4.4 15.6 3.5 0.1 0.2
- 1.1 0.0 - 1.0 0.9 2.0 - 1.8 - 0.3 1.5 1.6 3.0 1.7 0.5 0.4 0.9
Consumer - 32,816 28 - 454 1,454 129 - 4,422 - 23,298 111 18 1,088 211,751 5,021 151 48
Products - 11.7 0.0 - 0.2 0.5 0.0 - 1.6 - 8.3 0.0 0.0 0.4 75.4 1.8 0.1 0.0
- 0.8 0.0 - 2.7 0.7 0.5 - 2.7 - 3.8 0.4 0.1 1.1 32.7 1.1 1.0 0.3
Electronic - 199,482 57 - 38 7,379 440 - 12,939 120 1,975 486 496 2,442 883 4,121 86 209
Components - 86.3 0.0 - 0.0 3.2 0.2 - 5.6 0.1 0.9 0.2 0.2 1.1 0.4 1.8 0.0 0.1
- 5.0 0.0 - 0.2 3.4 1.6 - 7.9 0.0 0.3 1.8 1.5 2.4 0.1 0.9 0.5 1.3
Engines - 33,516 21 - 18 2,307 80 - 838 - 437 422 - 664 - 1,275 125 12
& Turbines - 84.4 0.1 - 0.0 5.8 0.2 - 2.1 - 1.1 1.1 - 1.7 - 3.2 0.3 0.0
- 0.8 0.0 - 0.1 1.1 0.3 - 0.5 - 0.1 1.6 - 0.6 - 0.3 0.8 0.1
Financial - - 415,908 - 2,239 1,170 712 - 19,204 4,043 713 - - 141 8 69,453 1,895 728
Services - - 80.6 - 0.4 0.2 0.1 - 3.7 0.8 0.1 - - 0.0 0.0 13.5 0.4 0.1
- - 71.1 - 13.3 0.5 2.6 - 11.7 0.9 0.1 - - 0.1 0.0 15.3 12.0 4.7
Food - 169,263 238 - 132 1,450 222 - 5,389 - 27,106 38 17 2,011 171,624 6,284 336 21
& Tobacco - 44.1 0.1 - 0.0 0.4 0.1 - 1.4 - 7.1 0.0 0.0 0.5 44.7 1.6 0.1 0.0
- 4.2 0.0 - 0.8 0.7 0.8 - 3.3 - 4.5 0.1 0.1 1.9 26.5 1.4 2.1 0.1
- 195 1,378 16,501 14 434 130 - 276 - - - - 939 17 686 12 -
Healthcare - 0.9 6.7 80.2 0.1 2.1 0.6 - 1.3 - - - - 4.6 0.1 3.3 0.1 -
- 0.0 0.2 1.1 0.1 0.2 0.5 - 0.2 - - - - 0.9 0.0 0.2 0.1 -
Hotels - - 4 346,737 202 109 143 - 2,031 - - - - - 603 1,215 84 -
& Tourism - - 0.0 98.7 0.1 0.0 0.0 - 0.6 - - - - - 0.2 0.3 0.0 -
- - 0.0 24.0 1.2 0.1 0.5 - 1.2 - - - - - 0.1 0.3 0.5 -
Continued on next page
E x t r a c t i o n
P r o d u c t i o n
B u s i n e s s
S e r v i c e s
C o n s t r u c t i o n
C u s t o m e r
C o n t a c t C e n t e r
D e s i g n , D e v e l o p m e n t
& T e s t i n g
E d u c a t i o n &
T r a i n i n g
E l e c t r i c i t y
H e a d q u a r t e r s
I C T & I n t e r n e t
I n f r a s t r u c t u r e
L o g i s t i c s ,
D i s t r i b u t i o n &
T r a n s p o r t a t i o n
M a i n t e n a n c e &
S e r v i c i n g
R e c y c l i n g
R e s e a r c h &
D e v e l o p m e n t
R e t a i l
S a l e s , M a r k e t i n g &
S u p p o r t
S h a r e d
S e r v i c e s C e n t e r
T e c h n i c a l S u p p o r t
C e n t e r
142
Table 4.11 ? Continued from previous page
Activities
Sector:
outstanding
% row total
% column total
Industrial Machinery, - 111,920 232 - 101 8,880 1,605 6 10,647 204 5,318 2,827 316 2,152 305 9,024 597 383
Equipment - 72.4 0.2 - 0.1 5.7 1.0 0.0 6.9 0.1 3.4 1.8 0.2 1.4 0.2 5.8 0.4 0.2
& Tools - 2.8 0.0 - 0.6 4.2 5.8 0.0 6.5 0.0 0.9 10.6 0.9 2.1 0.0 2.0 3.8 2.5
Leisure - 41 141 52,939 3 66 198 - 417 320 - - - - 30,799 1,272 - -
& Entertainment - 0.0 0.2 61.4 0.0 0.1 0.2 - 0.5 0.4 - - - - 35.7 1.5 - -
- 0.0 0.0 3.7 0.0 0.0 0.7 - 0.3 0.1 - - - - 4.8 0.3 - -
Medical - 19,987 314 - 21 2,507 762 - 2,613 - 1,680 170 - 3,975 168 1,641 192 45
Devices - 58.7 0.9 - 0.1 7.4 2.2 - 7.7 - 4.9 0.5 - 11.7 0.5 4.8 0.6 0.1
- 0.5 0.1 - 0.1 1.2 2.8 - 1.6 - 0.3 0.6 - 3.8 0.0 0.4 1.2 0.3
307,344 489,236 94 - 7 744 124 - 2,051 73 7,082 1,258 3,043 1,154 536 2,456 2 47
Metals 37.7 60.0 0.0 - 0.0 0.1 0.0 - 0.3 0.0 0.9 0.2 0.4 0.1 0.1 0.3 0.0 0.0
30.9 12.2 0.0 - 0.0 0.3 0.5 - 1.2 0.0 1.2 4.7 9.1 1.1 0.1 0.5 0.0 0.3
22,762 7,365 17 - - 814 314 - 244 - 138 - 40 20 313 194 - -
Minerals 70.6 22.9 0.1 - - 2.5 1.0 - 0.8 - 0.4 - 0.1 0.1 1.0 0.6 - -
2.3 0.2 0.0 - - 0.4 1.1 - 0.1 - 0.0 - 0.1 0.0 0.0 0.0 - -
Non-Automotive - 35,218 167 - - 504 204 - 217 - 2,400 3,732 209 284 1,607 858 18 -
Transport OEM - 77.5 0.4 - - 1.1 0.4 - 0.5 - 5.3 8.2 0.5 0.6 3.5 1.9 0.0 -
- 0.9 0.0 - - 0.2 0.7 - 0.1 - 0.4 14.0 0.6 0.3 0.2 0.2 0.1 -
Paper, - 110,310 53 - 18 570 26 - 187 - 2,004 100 1,196 126 29 476 46 -
Printing - 95.8 0.0 - 0.0 0.5 0.0 - 0.2 - 1.7 0.1 1.0 0.1 0.0 0.4 0.0 -
& Packaging - 2.7 0.0 - 0.1 0.3 0.1 - 0.1 - 0.3 0.4 3.6 0.1 0.0 0.1 0.3 -
- 76,207 425 - 1 5,109 195 - 6,750 383 2,497 - - 22,613 898 5,312 876 19
Pharmaceuticals - 62.8 0.4 - 0.0 4.2 0.2 - 5.6 0.3 2.1 - - 18.6 0.7 4.4 0.7 0.0
- 1.9 0.1 - 0.0 2.4 0.7 - 4.1 0.1 0.4 - - 21.8 0.1 1.2 5.5 0.1
- 95,523 21 - 3 2,653 132 - 2,369 - 787 91 721 550 201 744 14 30
Plastics - 92.0 0.0 - 0.0 2.6 0.1 - 2.3 - 0.8 0.1 0.7 0.5 0.2 0.7 0.0 0.0
- 2.4 0.0 - 0.0 1.2 0.5 - 1.4 - 0.1 0.3 2.2 0.5 0.0 0.2 0.1 0.2
- 719 32,129 987,122 - 160 11 - 621 - 3,642 1 - - 238 33,917 70 -
Real Estate - 0.1 3.0 93.2 - 0.0 0.0 - 0.1 - 0.3 0.0 - - 0.0 3.2 0.0 -
- 0.0 5.5 68.3 - 0.1 0.0 - 0.4 - 0.6 0.0 - - 0.0 7.5 0.4 -
- 87,490 - - 8 1,387 2 - 712 - 1,167 353 471 424 1,123 514 194 -
Rubber - 93.2 - - 0.0 1.5 0.0 - 0.8 - 1.2 0.4 0.5 0.5 1.2 0.5 0.2 -
Continued on next page
E x t r a c t i o n
P r o d u c t i o n
B u s i n e s s
S e r v i c e s
C o n s t r u c t i o n
C u s t o m e r
C o n t a c t C e n t e r
D e s i g n , D e v e l o p m e n t
& T e s t i n g
E d u c a t i o n &
T r a i n i n g
E l e c t r i c i t y
H e a d q u a r t e r s
I C T & I n t e r n e t
I n f r a s t r u c t u r e
L o g i s t i c s ,
D i s t r i b u t i o n &
T r a n s p o r t a t i o n
M a i n t e n a n c e &
S e r v i c i n g
R e c y c l i n g
R e s e a r c h &
D e v e l o p m e n t
R e t a i l
S a l e s , M a r k e t i n g &
S u p p o r t
S h a r e d
S e r v i c e s C e n t e r
T e c h n i c a l S u p p o r t
C e n t e r
143
Table 4.11 ? Continued from previous page
Activities
Sector:
outstanding
% row total
% column total
- 2.2 - - 0.0 0.6 0.0 - 0.4 - 0.2 1.3 1.4 0.4 0.2 0.1 1.2 -
- 174,354 41 - 14 16,238 127 - 6,365 142 773 112 - 9,424 27 2,048 46 142
Semiconductors - 83.1 0.0 - 0.0 7.7 0.1 - 3.0 0.1 0.4 0.1 - 4.5 0.0 1.0 0.0 0.1
- 4.3 0.0 - 0.1 7.6 0.5 - 3.9 0.0 0.1 0.4 - 9.1 0.0 0.5 0.3 0.9
Software - 1,964 21,920 - 3,697 57,763 1,049 - 21,773 63,523 1,575 290 - 10,218 291 81,499 2,076 8,086
& IT services - 0.7 7.9 - 1.3 20.9 0.4 - 7.9 23.0 0.6 0.1 - 3.7 0.1 29.6 0.8 2.9
- 0.0 3.7 - 21.9 27.0 3.8 - 13.2 14.6 0.3 1.1 - 9.9 0.0 17.9 13.1 52.0
- 2,237 37 - - 1,041 83 - 975 - 53 134 - 129 - 727 22 -
Space - 41.1 0.7 - - 19.1 1.5 - 17.9 - 1.0 2.5 - 2.4 - 13.4 0.4 -
& Defence - 0.1 0.0 - - 0.5 0.3 - 0.6 - 0.0 0.5 - 0.1 - 0.2 0.1 -
- 40,342 2 120 53 220 70 - 2,096 - 7,779 305 - 241 175,844 930 21 21
Textiles - 17.7 0.0 0.1 0.0 0.1 0.0 - 0.9 - 3.4 0.1 - 0.1 77.1 0.4 0.0 0.0
- 1.0 0.0 0.0 0.3 0.1 0.3 - 1.3 - 1.3 1.1 - 0.2 27.2 0.2 0.1 0.1
- 169 584 150 259 419 649 - 4,043 637 257,228 385 - 46 891 107,181 141 -
Transportation - 0.0 0.2 0.0 0.1 0.1 0.2 - 1.1 0.2 69.0 0.1 - 0.0 0.2 28.8 0.0 -
- 0.0 0.1 0.0 1.5 0.2 2.4 - 2.5 0.1 42.5 1.4 - 0.0 0.1 23.6 0.9 -
Warehousing - 9,204 - 270 - - 5 - 235 - 177,685 - - - 29 556 1 -
& Storage - 4.9 - 0.1 - - 0.0 - 0.1 - 94.5 - - - 0.0 0.3 0.0 -
- 0.2 - 0.0 - - 0.0 - 0.1 - 29.3 - - - 0.0 0.1 0.0 -
Wood - 32,895 - - - 44 31 - 67 - 623 6 52 20 1,539 747 - -
Products - 91.3 - - - 0.1 0.1 - 0.2 - 1.7 0.0 0.1 0.1 4.3 2.1 - -
- 0.8 - - - 0.0 0.1 - 0.0 - 0.1 0.0 0.2 0.0 0.2 0.2 - -
E x t r a c t i o n
P r o d u c t i o n
B u s i n e s s
S e r v i c e s
C o n s t r u c t i o n
C u s t o m e r
C o n t a c t C e n t e r
D e s i g n , D e v e l o p m e n t
& T e s t i n g
E d u c a t i o n &
T r a i n i n g
E l e c t r i c i t y
H e a d q u a r t e r s
I C T & I n t e r n e t
I n f r a s t r u c t u r e
L o g i s t i c s ,
D i s t r i b u t i o n &
T r a n s p o r t a t i o n
M a i n t e n a n c e &
S e r v i c i n g
R e c y c l i n g
R e s e a r c h &
D e v e l o p m e n t
R e t a i l
S a l e s , M a r k e t i n g &
S u p p o r t
S h a r e d
S e r v i c e s C e n t e r
T e c h n i c a l S u p p o r t
C e n t e r
Chapter 5: Conclusion
In light of the rise of offshoring and the expansion of GVCs, [Antra?s and
Staiger, 2012] make a case for a structural reshaping of PTAs arguing that ?effec-
tive trade agreements and the institutions that support them will have to evolve,
from a market access focus toward a focus on deep integration, and from a reliance
on simple and broadly-applied rules that guide the member-governments in their
negotiations and shape their agreements, toward a collection of more-individualized
agreements that can better reflect member-specific idiosyncratic needs?. This struc-
tural reshaping calls for an adaptation of the traditional tools used to evaluate the
economic impact of PTAs.
This dissertation contributes to this agenda by using theoretical and empirical
techniques to study the impact of specific behind-the-border provisions included in
PTAs. The different exercises conducted in the three analytical chapters show that
modern PTAs have differentiated effects on their members and across sectors. It
focuses on the investment and quality impacts of deep PTAs and shows that, at the
policy level, PTAs can have positive impacts. To go further, one could study in more
details the effects of deep PTAs on welfare with more elaborated quantitative tools.
The complexity of the model derived in the second chapter hints that computational
144
tools would be necessary to solve the general equilibrium problem.
To conclude this thesis, I want to emphasize the importance of leveraging the
assets of both multilateral and preferential trading systems. Global trade has been
under strain from the year-long trade war between the US and China. Despite the
difficulties encountered since its last major round of negotiations, the WTO has a
determinant role to play in solving the current trade tensions. Because no other
institution has brought more stability to international trade, it is in the interest of
all to preserve the multilateral institution. Countries should also continue to seek
integration through PTAs, as they have proven to be good alternatives to address
behind-the-border issues.
145
Appendix A: Additional tables and figures
A.1 Tables
Table A.1: List of WTO-plus disciplines included in
PTAs
WTO-plus policy areas
AD Retention of antidumping rights and obligations
under the WTO Agreement (Art VI GATT). Un-
fair trade practices.
Customs Provision of information; publication on the inter-
net of new laws and regulations; training. Incl.
provisions on trade facilitation.
CVM Retention of countervailing measures rights and
obligations under the WTO Agreement (Art VI
GATT).
Export Taxes Elimination of export taxes. Examples: Elimina-
tion of customs duties on exports, elimination of
duties, taxes or other charges on exports.
FTA Agriculture Tariff liberalization with regard to agriculture
goods; elimination of non-tariff measures.
FTA Industrial or Customs Tariff liberalization with regard to industrial
goods; elimination of non-tariff measures.
GATS Liberalization of trade in services.
Public Procurement Progressive liberalization; national treatment
and/or non-discrimination principle; publication
of laws and regulations on the internet; specifica-
tion on public procurement regime.
SPS Affirmation of rights and obligations under the
WTO Agreement on SPS; harmonization of SPS
measures.
State Aid Assessment of anticompetitive behavior; annual
reporting on the value and distribution of state
aid given; provision of information. Incl. export
subsidies on products.
Continued on next page
146
Table A.1 ? Continued from previous page
WTO-plus policy areas
STE GATT Art. XVII. Establishment or maintenance
of a state enterprise in accordance with and affirm-
ing provisions of GATT. Non-discrimination re-
garding production and marketing condition; pro-
vision of information. Incl. provisions on public
undertakings.
TBT Affirmation of rights and obligations under WTO
Agreement on TBT; provision of information;
harmonization of regulations; mutual recognition
agreements.
TRIMs Provisions concerning requirements for local con-
tent and export performance on FDI. Applies only
to measures that affect trade in goods.
TRIPs Harmonization of standards; enforcement; na-
tional treatment; most-favored nation treatment
and any other policy covered by TRIPs. Interna-
tional treaties referenced in TRIPS: Paris Conven-
tion, Berne Convention, Rome Convention, IPIC
Treaty.
147
Table A.2: List of WTO-X disciplines included in PTAs
WTO-X policy areas
Anti-Corruption Regulations concerning criminal offense measures
in matters affecting international trade and invest-
ment
Competition Policy Maintenance of measures to proscribe anticompet-
itive business conduct; harmonization of compe-
tition laws; establishment or maintenance of an
independent competition authority
Environmental Laws Development of environmental standards; enforce-
ment of national environmental laws; establish-
ment of sanctions for violation of environmental
laws; publications of laws and regulation
IPR Accession to international treaties not referenced
in the TRIPs Agreement
Investment Information exchange; Development of legal
frameworks; Harmonization and simplification of
procedures; National treatment; establishment of
mechanism for the settlement of disputes
Labor Market Regulation Regulation of the national labour market; affirma-
tion of International Labour Organization (ILO)
commitments; enforcement
Movement of Capital Liberalization of capital movement; prohibition of
new restrictions
Consumer Protection Harmonization of consumer protection laws; ex-
change of information and experts; training
Data Protection Exchange of information and experts; joint
projects
Agriculture Technical assistance to conduct modernization
projects; exchange of information
Approximation of Legislation Application of EC legislation in national legisla-
tion
Audio Visual Promotion of the industry; encouragement of co-
production
Civil Protection Implementation of harmonized rules
Innovation Policies Participation in framework programs; promotion
of technology transfers
Cultural Cooperation Promotion of joint initiatives and local culture
Economic Policy Dialogue Exchange of ideas and opinions; joint studies
Education and Training Measures to improve the general level of education
Energy Exchange of information; technology transfer;
joint studies
Continued on next page
148
Table A.2 ? Continued from previous page
WTO-X policy areas
Financial Assistance Set of rules guiding the granting and administra-
tion of financial assistance
Health Monitoring of diseases; development of health in-
formation systems; exchange of information
Human Rights Respect for human rights
Illegal Immigration Conclusion of re-admission agreements; prevention
and control of illegal immigration
Illicit Drugs Treatment and rehabilitation of drug addicts; joint
projects on prevention of consumption; reduction
of drug supply; information exchange
Industrial Cooperation Assistance in conducting modernization projects;
facilitation and access to credit to finance
Information Society Exchange of information; dissemination of new
technologies; training
Mining Exchange of information and experience; develop-
ment of joint initiatives
Money Laundering Harmonization of standards; technical and admin-
istrative assistance
Nuclear Safety Development of laws and regulations; supervision
of the transportation of radioactive materials
Political Dialogue Convergence of the parties? positions on interna-
tional issues
Public Administration Technical assistance; exchange of information;
joint projects; Training
Regional Cooperation Promotion of regional cooperation; technical assis-
tance programmes
Research and Technology Joint research projects; exchange of researchers;
development of public-private partnership
SME Technical assistance for small and medium enter-
prises; facilitation of the access to finance
Social Matters Coordination of social security systems; non-
discrimination regarding working conditions
Statistics Harmonization and/or development of statistical
methods; training
Taxation Assistance in conducting fiscal system reforms
Terrorism Exchange of information and experience; joint re-
search and studies
Visa and Asylum Exchange of information; drafting legislation;
training
149
Table A.3: Template for TBT chapters? mapping
Provisions in TBT chapter
Reference to WTO-TBT agreement:
TBT?1 Does the agreement refer to the WTO TBT Agreement?
TBT?2 Does the agreement use the same definitions as the TBT Agreements?
TBT?3 Does the agreement go beyond the TBT Agreement?
Integration approach:
TBT?4 Standards: ? Is mutual recognition in force?
TBT?5 Standards: ? Is there a time schedule for achieving mutual recognition?
TBT?6 Standards: ? Is the burden of justifying non-equivalence on the import-
ing country?
TBT?7 Standards: ? Is the burden of justifying non-equivalence on the import-
ing country?
TBT?8 Standards: ? Is the use or creation of regional standards promoted?
TBT?9 Standards: ? Is the use of international standards promoted?
TBT?10 Technical Regulations: ? Is mutual recognition in force?
TBT?11 Technical Regulations: ? Is there a time schedule for achieving mutual
recognition?
TBT?12 Technical Regulations: ? Is the burden of justifying non-equivalence on
the importing country?
TBT?13 Technical Regulations: ? Are there specified existing standards to which
countries shall harmonize?
TBT?14 Technical Regulations: ? Is the use or creation of regional standards
promoted?
TBT?15 Technical Regulations: ? Is the use of international standards pro-
moted?
TBT?16 Conformity Assessment: ? Is mutual recognition in force?
TBT?17 Conformity Assessment: ? Is there a time schedule for achieving mutual
recognition?
TBT?18 Conformity Assessment: ? Do parties participate in international or
regional accreditation agencies?
TBT?19 Conformity Assessment: ? Is the burden of justifying non-equivalence
on the importing country?
TBT?20 Conformity Assessment: ? Are there specified existing standards to
which shall harmonize?
TBT?21 Conformity Assessment: ? Is the use or creation of regional standards
promoted?
TBT?22 Conformity Assessment: ? Is the use of international standards pro-
moted?
Transparency requirements:
TBT?23 Is the time period allowed for comments specified?
TBT?24 Is the time period allowed for comments longer than 60 days?
TBT?25 Contact points/consultations for exchange of information
Continued on next page
150
Table A.3 ? Continued from previous page
Provisions in TBT chapter
Institutions:
TBT?26 Is a regional body established?
TBT?27 Is there a regional dispute settlement body?
TBT?28 Are there regional consultations foreseen to resolve disputes?
TBT?29 Is there a mechanism to issue recommendations?
TBT?30 Are recommendations mandatory?
TBT?31 Is the recourse to the DS for technical regulations disallowed?
Further cooperation among members:
TBT?32 Common policy/standardization program (beyond trade-related objec-
tives)
TBT?33 Technical Assistance
TBT?34 Metrology
151
Table A.4: Template for SPS chapters? mapping
Provisions in SPS chapter
Reference to WTO-SPS agreement:
SPS?1 Does the Agreement refer to the WTO SPS Agreement?
SPS?2 Does the Agreement use the same definitions as the SPS Agreement?
SPS?3 Does the Agreement use the same rules as the SPS Agreement?
SPS?4 Are any specific Annexes of the SPS Agreement adopted?
Integration approach:
SPS?5 Standards: ? Do parties recognize the adaption to regional conditions
(including regionalization, zoning and/or compartmentalization)
SPS?6 Standards: ? Do parties reference international standards?
SPS?7 Standards: ? Is equivalence recognized?
SPS?8 Standards: ? Are Parties encouraged to take into account other parties?
standards when elaborating new standards?
SPS?9 Standards: ? Is the burden of justifying non-equivalence on the import-
ing country?
SPS?10 Standards: ? Is mutual recognition recognized?
SPS?11 Standards: ? Is there a time schedule for achieving mutual recognition?
SPS?12 Standards: ? Are there specified existing standards to which countries
shall harmonize?
SPS?13 Standards: ? Is the creation of concerted/regional standards referenced?
SPS?14 Risk assessment: ? Do the parties recognize that SPS measures are
based on documented and scientific (if not available, objective) evidence?
SPS?15 Risk assessment: ? if yes, length of period (days)
SPS?16 Risk assessment: ? Is the burden of evaluating risk on the exporting
country?
SPS?17 Risk assessment: ? Is there reference to international stan-
dards/procedures?
SPS?18 Audits/Control inspection: ? Is there a provision on control and inspec-
tion?
SPS?19 Audits/Control inspection: ? Are there provisions for pre-certification
processes for exporting firms?
SPS?20 Audits/Control inspection: ? Are there provisions for advance rulings?
SPS?21 Audits/Control inspection: ? Is mutual recognition in force?
SPS?22 Audits/Control inspection: ? Does the importing party have the right
to audit the exporting party?s competent authorities, inspection systems,
or production procedure?
SPS?23 Audits/Control inspection: ? Is the burden of justifying non-equivalence
on the importing country?
SPS?24 Audits/Control inspection: ? Is the participation of interested parties
referenced?
SPS?25 Audits/Control inspection: ? Are there specified existing standards to
which countries shall harmonize?
Continued on next page
152
Table A.4 ? Continued from previous page
Provisions in SPS chapter
SPS?26 Audits/Control inspection: ? Is the use or creation of regional standards
promoted?
SPS?27 Audits/Control inspection: ? Is the use of international standards pro-
moted?
Transparency:
SPS?28 Is there a transparency provision?
SPS?29 Is there a provision on exchange of information?
SPS?30 Is there a provision on electronic publication?
SPS?31 Is there a duty to translate the document in the language of the other
party(ies)?
SPS?32 Is there a limitation to the obligation to notify, for reasons of law en-
forcement, public interest, or commercial interest?
SPS?33 Do parties have to notify each other prior to the entry into force of a new
standard or regulation?
SPS?34 Is there a specified minimum time period for comments?
SPS?35 Is there a derogation clause on notification period for emergency?
SPS?37 Does the Agreement allow the participation of interested parties of the
other party in the development of standards?
SPS?38 Does the agreement specifically reference the participation of regulatory
authorities of the other party in the development of standards?
Institutions:
SPS?39 Do parties establish SPS contact/enquiry points?
SPS?40 Do parties establish a SPS committee?
SPS?41 Is there a fixed periodical meeting for the committee?
SPS?42 Is the SPS Committee the designated first place for dispute resolution?
SPS?43 Is the SPS Committee open?
SPS?44 Do parties establish a working group?
SPS?45 Is there a mechanism to issue recommendations?
SPS?46 Is there a mechanism mandated to issue administrative decisions?
SPS?47 Is a body for administering the Agreement established?
SPS?48 Is the recourse to the DS for SPS chapter disallowed?
International Regulatory cooperation areas:
SPS?49 Is there a general IRC clause/Common policy/standardization program
(beyond trade-related objectives)?
SPS?50 Is there a provision on technical assistance?
SPS?51 Is there a provision for technical consultations/cooperation?
Other areas of cooperation:
SPS?52 Is there an MRA in force?
SPS?53 Is there a provision on special/preferential treatment?
SPS?54 Is there a provision on certifications?
SPS?55 Is there a provision on labelling, marking, and packaging?
SPS?56 Is there a provision on traceability?
Continued on next page
153
Table A.4 ? Continued from previous page
Provisions in SPS chapter
SPS?57 Is coordination for participating in international or regional accreditation
agencies referenced?
SPS?58 Is testing data to be made available?
154
ln(quality)
VARIABLES (1) (2) (3) (4) (5) (6)
TBTMutualRecoStandard 0.361***
(0.0685)
TBTMutualRecoTechReg 0.361***
(0.0685)
TBTMutualRecoConformity 0.354***
(0.0610)
TBTHarmoStandard 0.191***
(0.0572)
TBTHarmoTechReg 0.227***
(0.0566)
TBTHarmoConformity 0.198***
(0.0557)
TotalDepth 0.00802*** 0.00802*** 0.00813*** 0.00655*** 0.00695*** 0.00669***
(0.00128) (0.00128) (0.00126) (0.00124) (0.00124) (0.00124)
Observations 3,196,883 3,196,883 3,196,883 3,196,883 3,196,883 3,196,883
R-squared 0.030 0.030 0.030 0.030 0.030 0.030
Origin-time FE Yes Yes Yes Yes Yes Yes
Destination-time FE Yes Yes Yes Yes Yes Yes
HS2 FE No No No No No No
Pair FE Yes Yes Yes Yes Yes Yes
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Table A.5: Baseline estimation of the impact of TBT provisions on average quality
without sector fixed effects
155
ln(quality)
VARIABLES (1) (2) (3) (4)
SPSMutualRecoStandard 0.0629
(0.131)
SPSMutualRecoInspection 0.171
(0.115)
SPSHarmoStandard 0.324*
(0.179)
SPSHarmoInspection -0.115**
(0.0499)
TotalDepth 0.00354*** 0.00349*** 0.00357*** 0.00470***
(0.00120) (0.00120) (0.00120) (0.00126)
Observations 667,361 667,361 667,361 667,361
R-squared 0.041 0.041 0.041 0.041
Origin-time FE Yes Yes Yes Yes
Destination-time FE Yes Yes Yes Yes
HS2 FE Yes Yes Yes Yes
Pair FE Yes Yes Yes Yes
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Table A.6: Baseline estimation of the impact of TBT provisions on average quality
without sector fixed effects
156
VARIABLES ln(quality)
(1) (2)
TBTMutualRecoall (t? 2)?NorthNorth -0.175***
(0.0678)
TBTMutualRecoall (t? 2)?NorthSouth 0.220*
(0.122)
TBTMutualRecoall (t? 2)? SouthSouth 0.725***
(0.234)
TBTHarmoall (t? 2)?NorthNorth -0.106
(0.0745)
TBTHarmoall (t? 2)?NorthSouth 0.152**
(0.0717)
TBTHarmoall (t? 2)? SouthSouth 0.0729
(0.220)
TotalDepth 0.00274** 0.00309**
(0.00136) (0.00137)
Observations 2,543,354 2,543,354
R-squared 0.492 0.492
Origin-time FE Yes Yes
Destination-time FE Yes Yes
HS2 FE Yes Yes
pair FE yes yes
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Table A.7: TBT lagged interaction with level of development
157
VARIABLES ln(quality)
(1) (2)
SPSMutualRecoall (t? 2)?NorthNorth 0.0947
(0.263)
SPSMutualRecoall (t? 2)?NorthSouth 0.352*
(0.203)
SPSMutualRecoall (t? 2)? SouthSouth -0.246*
(0.149)
SPSHarmoall (t? 2)?NorthNorth 0.0682
(0.0684)
SPSHarmoall (t? 2)?NorthSouth -0.103*
(0.0609)
SPSHarmoall (t? 2)? SouthSouth -0.156*
(0.0851)
TotalDepth 0.00347*** 0.00427***
(0.00132) (0.00138)
Observations 526,034 526,034
R-squared 0.246 0.246
Origin-time FE Yes Yes
Destination-time FE Yes Yes
HS2 FE Yes Yes
pair FE yes yes
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Table A.8: SPS lagged interaction with level of development
158
VARIABLES ln(quality)
(1) (2) (3) (4)
TBTMutualRecoall (t? 2) -0.153** -0.0669
(0.0678) (0.0548)
TBTMutualRecoall (t? 2)? LadderFullRange 0.0190***
(0.00659)
TBTMutualRecoall (t? 2)? LadderInterquartile 0.0302**
(0.0133)
TBTHarmoall (t? 2) -0.299*** 0.0872*
(0.0667) (0.0472)
TBTHarmoall (t? 2)? LadderFullRange 0.0377***
(0.00613)
TBTHarmoall (t? 2)? LadderInterquartile -0.0225**
(0.00983)
LadderFullRange 0.355*** 0.353***
(0.00205) (0.00206)
LadderInterquartile 0.641*** 0.645***
(0.00314) (0.00329)
TotalDepth 0.000361 0.000272 0.000561 0.000502
(0.00114) (0.00113) (0.00115) (0.00114)
Observations 2,489,024 2,489,024 2,489,024 2,489,024
R-squared 0.615 0.629 0.615 0.629
Origin-time FE Yes Yes Yes Yes
Destination-time FE Yes Yes Yes Yes
HS2 FE Yes Yes Yes Yes
Pair FE Yes Yes Yes Yes
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Table A.9: TBT lagged interaction with ladder length
159
VARIABLES ln(quality)
(1) (2) (3) (4)
SPSMutualRecoall (t? 2) -0.252* 0.00679
(0.130) (0.118)
SPSMutualRecoall (t? 2)? LadderFullRange 0.138***
(0.0354)
SPSMutualRecoall (t? 2)? LadderInterquartile 0.0256
(0.0468)
SPSHarmoall (t? 2) -0.227*** -0.0220
(0.0482) (0.0397)
SPSHarmoall (t? 2)? LadderFullRange 0.0738***
(0.0128)
SPSHarmoall (t? 2)? LadderInterquartile 0.0507***
(0.0193)
LadderFullRange 0.278*** 0.269***
(0.00530) (0.00540)
LadderInterquartile 0.492*** 0.483***
(0.00891) (0.00950)
TotalDepth 0.00324*** 0.00188 0.00345*** 0.00219*
(0.00121) (0.00120) (0.00126) (0.00124)
Observations 513,656 513,656 513,656 513,656
R-squared 0.415 0.425 0.417 0.425
Origin-time FE Yes Yes Yes Yes
Destination-time FE Yes Yes Yes Yes
HS2 FE Yes Yes Yes Yes
Pair FE Yes Yes Yes Yes
Robust standard errors in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Table A.10: SPS lagged interaction with ladder length
160
Appendix B: Derivation of the optimal TBT/SPS-related costs and
productivity thresholds
B.1 Maximization problem under no recognition
In the no recognition scenario (NR), optimal qualities are chosen for each des-
tination markets separately and if firms cannot eventually sustain the usual Melitz
fixed costs to export, then dropping a market does not influence their quality choices
for the remaining destination markets. The firm?s profit maximization problem if
export to each markets (1, 2, ...k) and domestic market i is written:
? ( )k 1?? k?jE ?j ? tij?ijwi ?(??1)
max?(fi1, fi2, ..., fik) = f ? w (f + F )
? ? ? ij i ij ij1 ?P?
j=1 j j=1
(B.1)
subject to
?j ? (1, ..., k), fij > f (B.2)j
?
The Lagrangian is defined as L(f, ?) = ?(f) ? kj=1 ?j(f ? f ), where f =j ij
(fi1, fi2, ..., fik) and ? = (?1, ..., ?k).
161
B.1.1 Finding optimal qualities under NR
Maximum under constraints must satisfy the following equations:
? Interior solution ?L(f,?) = ??(f) + ?j = 0, ?j ? (1, ..., k)?fij ?fij [ ]
? complementary slackness ??L(f,?) = ? f ? f = 0
?? ij ij
? original constraint ?L(f,?) = f ? fij ? 0?? j
? ?ij ? 0?j ? (1, ..., k)
The interior solutions verify:
( )1??
?L(f, ?) ?(? ? 1) ? tij?ijwi ?(??1)?1
= ?jEj fij ? wi + ?j = 0 (B.3)?fij ? ? ? 1 ?P?j
Which requires ?(??1)?1 < 0 to have a maximum (i.e. the second order conditions
are negative).
? If the constraint for market j does not bind: ?j = 0 and therefore f < fj ij
?? ( ) ? 1??1??(??1)?1opt wi ? ? tij?ijwifij = (B.4)?jEj ?(? ? 1) ? ? 1 ?P?j
? If the constraint binds, then ?j > 0 and f = fj ij
B.1.2 Finding productivity threshold under NR
The productivity thresholds are such that ? optj(fij ) > 0,?j
162
? If f optij = f j
( )1??
? ( )jEj ? tij?ijwi
? (f ) = f?(??1)j ( ) ? wi f + Fij ij > 0? ? ? 1 ?P? j jj 1??
??1?jEj ? t
( )
ij?ijwi
? f?(??1) ? wi f + Fij > 0
? ? ? 1( P? ) j j (B.5)j
( wi f +)Fj ij???1 > 1??
?jEj ? tij?ijwi f?(??1)
? ??1 P?j j
Finally using q = f?
( ) 1
1??(??1)
?,NR ? F
? ?(??1)
ij ?1
?ij = q
?(? 1)
j 1 + [? ]??1ij (B.6)q
j
( )
? 1???(? 1) ? t ?Where ? = ij ij
wi ?jEj
ij ? ??1 P?j witij
163
? if f optij as in (B.11)
( ) ?1?? ? ( ) ?
?(??1)
??1 ?(??1)?1
? ?jEj ? tij?ijwi wi ? ? tij?ijwi ?
? ???? 1 ?P?j ?jEj ?(? ? 1) ?? ? 1? ( )
?P?j ?
1
??1 ?(??1)?1
wi ? ? tij?ijwi ?
> w ?? ?i + Fij?
?jEj ?(? ? 1) ? ? 1 ?P?j
( ) ? ( ) ? ?(??1)1?? ? ??1??(??1)?1? wi ?(? ? 1) ?jEj ? tij?ijwi wi ? ? tij?ijwi
wi ?(????1) ? ? ? 1 ?P?j ?jEj??(? ? 1) ? ??1( ) ?P?j???
1
??1 ?(??1)?1
wi ? ? tij?ijwi
> w ?i + F? ? ij???jEj ?(? 1) ? 1 ?P?j
( ) ? ( ) ? ?(??1)1?? ??1 ?(??1)?1
? wi ?(? ? 1) ?jEj ? tij?ijwi ? wi ? ? tij?ijwi ?
wi ?(????1) ? ? ? 1 ?P?j ?jEj??(? ? 1) ? ?? ?
1 ?P?j
1
?? ( )??1??(??1)?1wi ? ? tij?ijwi ?> wi + F? ? ij?? E ?(? 1) ? 1
?j j ?P?j? ( ) ? 1??1??(??1)?1? wi wi ? ? tij?ijwi
?(? ??1?) ?jEj ?(? ? 1) ? ? 1 ?P?j? ??? ( )
? 1
??1??(??1)?1wi ? ? tij?ijwi ?> wi + Fij?
?jEj ?(? ? 1) ? ? 1? ?P?j? ( ) ?? 1??1 ?(??1)?1? 1? ?(? ? 1) wi ? ? tij?ijwiwi > w F
?(? ? ? ? i ij1) ?jEj ?(? 1) ? 1 ?P?j
1??
? F??(?? ij1)?1 > [ ( ) ] 1
? ? ??1 ?(??1)?11 ?(? 1) wi ? ? tij?ijwi
?(??1) ?jEj ?(??1) ??1 P?j
(B.7)
164
Remark that the above threshold is bounded by the threshold obtained from
the corner solution f optij = f , therefore the overall productivity thresholdij
corresponds to the minimum of those two.
B.2 Maximization problem under full integration
Under full integration (NR or HR) the solutions for the optimal fixed costs are
not separable anymore. Firms choose a unique f opti that will apply domestically and
to any markets part of the PTA. The trade policy change corresponds to the move
from k inequality constraints to a unique inequality. While the decision to export
to different markets were separable in the non-recognition case, in this context the
choice of the unique f opti is entangled with the decisions to export to specific set
of markets. The problem could be solved computationally backwards. First fixing
the number of exports market and then solving for optimal f , then discussing what
happen for the marginal firm that cannot export to all k markets but would be
better off exporting to only k ? 1 markets.
The profit maximization problem can be rewritten as:
? ( )k 1?? ?k
?(??1) ?jEj ? tij?ijwi
max?(fi,11, ...,1k) = fi 1j ? w? i
fi ? wi 1jFij
? ? 1 ?P?
j=1 j j=1
(B.8)
subject to
fi > f (B.9)
Where f could be the maximum of all f or some other international standards.
j
165
And 1j are indicator variables (0, 1) representing the decision to export to market j
The new Lagrangian is defined as L(fi, ?) = ?(fi)? ?(f ? fi),
B.2.1 Finding optimal qualities under MR or HR
? Interior solution ?L(fi,?) = ??(fi) + ? = 0
?fi ?fi
?L(f ,?) [ ]? complementary slackness ? i = ? f ? f
?? i
= 0
? original constraint ?L(fi,?) = f ? fi ? 0??
? ? ? 0
The interior solution if the firm enters k markets verifies:
? ( )k 1???L(f, ?) ?(??1)?1 ?(? ? 1) ? tij?ijwi
= fi ?jEj ? wi + ? = 0 (B.10)?fij ? ? ? 1 ?P?j=1 j
? If the constraint does not bind: ? = 0 and therefore f < fi
?? ( ) ? 1k ??1 ?(??1)?1wi ? ? tij?ijwi
f opt ? ?ij = (B.11)?jEj ?(? ? 1) ? ? 1 ?P?j=1 j
? If the constraint binds, then ? > 0 and f = fi
B.2.2 Finding productivity threshold under MR or HR
The productivity thresholds are such that ?(f opt) > 0
166
? f opti = f
? ( )k 1??? E ? t ? w ?kj j ij ij i
?(f ) = f?(??1) ( ) ? wif ? wi Fij ? ij > 0? ? 1??P?j=1 j j=1k
?wi f +? (j=1 Fij?? 1 > )1??? E t ? w
f?(??1) k j j ? ij ij ij=1 ? ??1 P?j
(B.12)
Even if we consider that countries that sign a PTA have similar domestic
TBT/SPS regulations (i.e. f ? f), the expression of exporting threshold has
j
an ambiguous relationship with the number of destination markets. Both nu-
merator and denominator increases with the number of destinations, making
the overall change impossible to evaluate for regional PTAs (more than two
countries) without computational method. For bilateral PTAs, the denomina-
tor still increases with full integration, but the numerator remains the same
as the non recognition scenario (assuming f ? f). Therefore, the exporting
j
threshold defined by (B.12) shows that mutual recognition or harmonization
of TBT/SPS measures increases the number of exporters at least for bilateral
PTAs.
167
? if f optij as in (B.11)
?? ( ) ? ?(??1)? k ??1??(??1)?1 ? ( )k 1??? wi ? ? tij?ijwi ?jEj ? tij?ijwi
??j?Ej ?(? ? 1) ? ? 1 ?P?j ? ? ? ? 1( ) ?P?j=1 j=1? ?
j
???
1
k ??1
wi ? ? tij?ijwi ??(??1)?1 ?k ?> wi + Fij?
? E ?(? ? 1) ? ? 1
? ?P?j=1 j j j?? ( ) ?
j=1
1
k ??1 ?(??1)?1
? wi wi ? ? tij?ijwi ?
?(? ??1?) ?jEj ?(? ? 1) ? ? 1 ?P?j=1 j? ???? ( )
? 1
k ??1??(??1)?1 kwi ? ? tij?ijw ?i ?> wi + Fij?
?jE ?(? ? 1) ? ? 1j=1 ?j ?P?j( ) ? j=11??1 ?(??1)?1
? ? ??k ?k? 1 ?(? 1) wi ? ? tij?ijwiw ?i > w F? i ij?(? 1) ?jEj ?(? ? 1) ? ? 1 ?P?j=1 j j=1
(B.13)
?k
? 1?? F? ij(? )?(??1)? j=11 = [? ( ) ] 1 (B.14)? ? ??1 ?(??1)?11 ?(? 1) k wi ? ? tij?ijwi
?(??1) j=1 ?jEj ?(??1) ??1 P?j
The resulting expression for the productivity threshold in the case of mutual
recognition or harmonization shows that if a firm productivity falls just below
this threshold it can always drop a market and select a smaller optimal quality.
This up to the point where the optimal quality hits the minimum quality
requirement applicable in the integrated market. Therefore, we can use (B.12)
168
as the minimum export threshold in the mutual recognition case.
169
Bibliography
[WTO, 2011] (2011). World Trade Report 2011: The WTO and preferential trade
agreements: From co-existence to coherence. World Trade Organization, Geneva.
[WTO, 2012] (2012). World Trade Report 2012: Trade and Public Policies: a Closer
Look at Non-Tariff Measures in the 21st Century. World Trade Organization,
Geneva.
[Amendolagine et al., 2017] Amendolagine, V., Presbitero, A. F., Rabellotti, R.,
Sanfilippo, M., and Seric, A. (2017). Fdi, global value chains, and local sourcing
in developing countries. Working paper, IMF.
[Amoroso et al., 2015] Amoroso, S., Dosso, M., and Moncada-Paterno?-Castello, P.
(2015). The impact of skill endowments and collective bargaining on knowledge-
intensive greenfield fdi. Technical report, SSRN.
[Anderson et al., 2016] Anderson, J. E., Larch, M., and Yotov, Y. V. (2016). Esti-
mating General Equilibrium Trade Policy Effects: GE PPML. Technical report,
LeBow College of Business, Drexel University.
[Anderson and Neary, 1994] Anderson, J. E. and Neary, J. P. (1994). Measuring the
restrictiveness of trade policy. The World Bank Economic Review, 8(2):151?169.
[Anderson and van Wincoop, 2003] Anderson, J. E. and van Wincoop, E. (2003).
Gravity with gravitas: A solution to the border puzzle. American Economic
Review.
[Antoniades, 2015] Antoniades, A. (2015). Heterogeneous Firms, Quality, and
Trade. Journal of International Economics.
[Antonietti et al., 2015] Antonietti, R., Bronzini, R., and Cainelli, G. (2015). In-
ward greenfield fdi and innovation. Economia e Politica Industriale.
[Antra?s and Staiger, 2012] Antra?s, P. and Staiger, R. W. (2012). Offshoring and
the role of trade agreements. American Economic Review.
[Baldwin and Harrigan, 2011] Baldwin, R. and Harrigan, J. (2011). Zeros, qual-
ity, and space: Trade theory and trade evidence. American Economic Journal:
Microeconomics.
170
[Baltagi et al., 2008] Baltagi, B. H., Egger, P., and Pfaffermayr, M. (2008). Esti-
mating regional trade agreement effects on fdi in an interdependent world. Journal
of Econometrics.
[Bown and Crowley, 2013a] Bown, C. P. and Crowley, M. A. (2013a). Import pro-
tection, business cycles, and exchange rates: Evidence from the Great Recession.
Journal of International Economics, 90(1):50?64.
[Bown and Crowley, 2013b] Bown, C. P. and Crowley, M. A. (2013b). Self-enforcing
trade agreements: Evidence from time-varying trade policy. American Economic
Review, 103(2):1071?90.
[Bown and Crowley, 2014] Bown, C. P. and Crowley, M. A. (2014). Emerging
Economies, Trade policy, and Macroeconomic Shocks. Journal of Development
Economics, 111(C):261?273.
[Broda and Weinstein, 2006] Broda, C. and Weinstein, D. E. (2006). Globalization
and the Gains From Variety. The Quarterly Journal of Economics, 121(2):541?
585.
[Brooks, 2006] Brooks, E. L. (2006). Why don?t firms export more? product quality
and colombian plants. Journal of Development Economics.
[Budetta and Piermartini, 2009] Budetta, M. and Piermartini, R. (2009). A Map-
ping of Regional Rules on Technical Barriers to Trade, . In Estevadeordal, A.,
Suominen, K., and Teh, R., editors, Regional Rules in the Global Trading System,
pages 250?315. Cambridge University Press.
[Bustos, 2011] Bustos, P. (2011). Trade liberalization, exports, and technology up-
grading: Evidence on the impact of mercosur on argentinian firms. American
Economic Review.
[Bu?the and Milner, 2008] Bu?the, T. and Milner, H. V. (2008). The politics of foreign
direct investment into developing countries: Increasing fdi through international
trade agreements? American Journal of Political Science.
[Caliendo and Parro, 2014] Caliendo, L. and Parro, F. (2014). Estimates of the
Trade and Welfare Effects of NAFTA. The Review of Economic Studies.
[Castellani and Pieri, 2016] Castellani, D. and Pieri, F. (2016). Outward invest-
ments and productivity: Evidence from european regions. Regional Studies.
[Chen and Mattoo, 2008] Chen, M. X. and Mattoo, A. (2008). Regionalism in Stan-
dards: Good or Bad for Trade? Canadian Journal of Economics, 41(3):838?863.
[Costinot, 2008] Costinot, A. (2008). A comparative institutional analysis of agree-
ments on product standards. Journal of International Economics, 75(1):197?213.
171
[Crescenzi et al., 2013] Crescenzi, R., Pietrobelli, C., and Rabellotti, R. (2013). In-
novation drivers, value chains and the geography of multinational corporations in
europe. Journal of Economic Geography.
[Crozet et al., 2011] Crozet, M., Head, K., and Mayer, T. (2011). Quality Sort-
ing and Trade: Firm-level Evidence for French Wine. The Review of Economic
Studies.
[Czubala et al., 2009] Czubala, W., Shepherd, B., and Wilson, J. S. (2009). Help
or hindrance? the impact of harmonised standards on african exports?. Journal
of African Economies.
[Dadkhah et al., 2019] Dadkhah, A., Ciuriak, D., and Xiao, C. (2019). Evaluat-
ing the impact of the usmca. Technical report, Global Trade Analysis Project
(GTAP).
[Dee and Gali, 2005] Dee, P. and Gali, J. (2005). The trade and investment effects
of preferential trading arrangements. In Ito, T. and Rose, A. K., editors, Interna-
tional Trade in East Asia, NBER-East Asia Seminar on Economics. University
of Chicago Press.
[Disdier et al., 2015] Disdier, A.-C., Fontagne?, L., and Cadot, O. (2015). North-
South Standards Harmonization and International Trade. The World Bank Eco-
nomic Review.
[Ecorys, 2009] Ecorys (2009). Non-Tariff Measures in EU-US Trade and Investment
- An Economic Analysis. Report prepared for the Europen Commission, Joseph
Francois (project leader).
[Egger and Pfaffermayr, 2000] Egger, P. and Pfaffermayr, M. (2000). Trade, multi-
national sales, and fdi in a three-factors model. Technical report, Johannes Kepler
University of Linz, Department of Economics.
[Espitia et al., 2019] Espitia, A., Batalla, S. P., Piermartini, R., and Rocha, N.
(2019). Mapping technical barriers to trade in preferential trade agreements:
Concepts, approach and stylized facts. Working paper, The World Bank.
[Farole and Winkler, 2014] Farole, T. and Winkler, D. (2014). Making Foreign Di-
rect Investment Work for Sub-Saharan Africa : Local Spillovers and Competitive-
ness in Global Value Chains. World Bank.
[Ferrantino, 1993] Ferrantino, M. J. (1993). The effect of intellectual property rights
on international trade and investment. Weltwirtschaftliches Archiv.
[Fontagne?, 1999] Fontagne?, L. (1999). Foreign direct investment and international
trade. Technical report, OECD Science, Technology and Industry Working Pa-
pers.
172
[Fontagne? et al., 2013] Fontagne?, L., Gourdon, J., and Jean, S. (2013). Transat-
lantic Trade: Whither Partnership, Which Economic Consequences? Technical
report, CEPII Policy Brief.
[Fontagne? et al., 2015] Fontagne?, L., Orefice, G., Piermartini, R., and Rocha, N.
(2015). Product Standards and Margins of Trade: Firm Level Evidence. Journal
of International Economics.
[Francois et al., 2013] Francois, J., Manchin, M., Norberg, H., Pindyuk, O., and
Tomberger, P. (2013). Reducing Transatlantic Barriers to Trade and Investment:
An Economic Assessment. CEPR Discussion Papers.
[Gandal and Shy, 2001] Gandal, N. and Shy, O. (2001). Standardization policy and
international trade. Journal of International Economics, 53(2):363?383.
[Gaulier and Zignago, 2010] Gaulier, G. and Zignago, S. (2010). Baci: International
trade database at the product-level. Cepii working paper, CEPII.
[Gervais, 2013] Gervais, A. (2013). Product Quality and Firm Heterogeneity in
International Trade. US Census Bureau Center for Economic Studies Paper No.
CES-WP-13-08.
[Goldberg and Pavcnik, 2007] Goldberg, P. K. and Pavcnik, N. (2007). Distribu-
tional Effects of Globalization in Developing Countries. Journal of Economic
Literature.
[Go?mez-Mera et al., 2014] Go?mez-Mera, L., Kenyon, T., Margalit, Y., Reis, J. G.,
and Varela, G. (2014). New Voices in Investment: A Survey of Investors from
Emerging Countries. The World Bank.
[Hallak, 2006] Hallak, J. C. (2006). Product quality and the direction of trade.
Journal of International Economics.
[Hallak and Schott, 2011] Hallak, J. C. and Schott, P. K. (2011). Estimating cross-
country differences in product quality. The Quarterly Journal of Economics.
[Hallak and Sivadasan, 2013] Hallak, J. C. and Sivadasan, J. (2013). Product and
process productivity: Implications for quality choice and conditional exporter
premia. Journal of International Economics.
[Hanna, 2010] Hanna, R. (2010). Us environmental regulation and fdi: Evidence
from a panel of us-based multinational firms. American Economic Journal: Ap-
plied Economics.
[Helpman, 1984] Helpman, E. (1984). A simple theory of international trade with
multinational corporations. Journal of Political Economy.
[Helpman, 1992] Helpman, E. (1992). Innovation, imitation, and intellectual prop-
erty rights. Working paper, National Bureau of Economic Research.
173
[Hofmann et al., 2017] Hofmann, C., Osnago, A., and Ruta, M. (2017). Horizontal
depth : a new database on the content of preferential trade agreements. Policy
Research Working Paper. The Worl Bank.
[Horn et al., 2010] Horn, H., Mavroidis, P. C., and Sapir, A. (2010). Beyond the
WTO? An Anatomy of EU and US Preferential Trade Agreements. The World
Economy, 33(11):1565?1588.
[Hummels and Klenow, 2005] Hummels, D. and Klenow, P. J. (2005). The variety
and quality of a nation?s exports. American Economic Review.
[Hummels and Skiba, 2004] Hummels, D. and Skiba, A. (2004). Shipping the good
apples out? an empirical confirmation of the alchian-allen conjecture. Journal of
Political Economy.
[Jackson and Vitikata, 2016] Jackson, L. A. and Vitikata, H. (2016). Cross-cutting
issues in regional trade agreements: sanitary and phytosanitary measures. In
Regional Trade Agreements and the Multilateral Trading System. World Trade
Organization.
[Javorcik and Spatareanu, 2005] Javorcik, B. S. and Spatareanu, M. (2005). Do
foreign investors care about labor market regulations? Download PDF Review of
World Economics.
[Johnson, 2012] Johnson, R. C. (2012). Trade and Prices with Heterogeneous Firms.
Journal of International Economics.
[Kee et al., 2008] Kee, H. L., Nicita, A., and Olarreaga, M. (2008). Import de-
mand elasticities and trade distortions. The Review of Economics and Statistics,
90(4):666?682.
[Kee et al., 2009] Kee, H. L., Nicita, A., and Olarreaga, M. (2009). Estimating
Trade Restrictiveness Indices. Economic Journal, 119(534):172?199.
[Khandelwal, 2010] Khandelwal, A. (2010). The Long and Short (of) Quality Lad-
ders. The Review of Economic Studies.
[Khandelwal et al., 2013] Khandelwal, A. K., Schott, P. K., and Wei, S.-J. (2013).
Trade Liberalization and Embedded Institutional Reform: Evidence from Chinese
Exporters. American Economic Review.
[Kugler and Verhoogen, 2011] Kugler, M. and Verhoogen, E. (2011). Prices, plant
size, and product quality. The Review of Economic Studies.
[Laget et al., 2018] Laget, E., Osnago, A., Rocha, N., and Ruta, M. (2018). Deep
Trade Agreements and Global Value Chains. Policy Research Working Papers.
The World Bank.
174
[Lesher and Miroudot, 2006] Lesher, M. and Miroudot, S. (2006). Analysis of the
economic impact of investment provisions in regional trade agreements. Technical
report, OECD Trade Policy Papers.
[Lima?o, 2016] Lima?o, N. (2016). Chapter 6 - preferential trade agreements. In
Bagwell, K. and Staiger, R. W., editors, Handbook of Commercial Policy. North-
Holland.
[Linder, 1961] Linder, S. B. (1961). An essay on trade and transformation. Almqvist
& Wiksell.
[Manova and Zhang, 2012] Manova, K. and Zhang, Z. (2012). Export prices across
firms and destinations. The Quarterly Journal of Economics.
[Mansfield and Lee, 1996] Mansfield, E. and Lee, J.-Y. (1996). The modern univer-
sity: contributor to industrial innovation and recipient of industrial r&d support.
Research Policy.
[Markusen, 1984] Markusen, J. R. (1984). Multinationals, multi-plant economies,
and the gains from trade. Journal of International Economics.
[Markusen and Maskus, 2002] Markusen, J. R. and Maskus, K. E. (2002). A uni-
fied approach to intra-industry trade and foreign direct investment. Frontiers of
Research in Intra-Industry Trade.
[Markusen and Venables, 1998] Markusen, J. R. and Venables, A. J. (1998). Multi-
national firms and the new trade theory. Journal of International Economics.
[Markusen and Venables, 2000] Markusen, J. R. and Venables, A. J. (2000). The
theory of endowment, intra-industry and multi-national trade. Journal of Inter-
national Economics.
[Mattoo et al., 2017] Mattoo, A., Mulabdic, A., and Ruta, M. (2017). Trade Cre-
ation and Trade Diversion in Deep Agreements. Policy Research Working Papers.
The World Bank.
[Medvedev, 2012] Medvedev, D. (2012). Beyond trade: The impact of preferential
trade agreements on fdi inflows. World Development.
[Melitz, 2003] Melitz, M. J. (2003). The Impact of Trade on Intra-Industry Reallo-
cations and Aggregate Industry Productivity. Econometrica, 71(6):1695?1725.
[Moenius, 2004] Moenius, J. (2004). Information Versus Product Adaptation: The
Role of Standards in Trade. International Business and Markets Research Center
Working Paper, Northwestern University.
[Orefice, 2017] Orefice, G. (2017). Non-tariff measures, specific trade concerns and
tariff reduction. The World Economy.
175
[Orefice and Rocha, 2014] Orefice, G. and Rocha, N. (2014). Deep integration and
production networks: An empirical analysis. The World Economy.
[Osnago et al., 2015] Osnago, A., Rocha, N., and Ruta, M. (2015). Deep Trade
Agreements and Vertical FDI: The Devil is in the Details. Policy Research Work-
ing Papers. The World Bank.
[Paniagua and Sapena, 2014] Paniagua, J. and Sapena, J. (2014). Is fdi doing good?
a golden rule for fdi ethics. Journal of Business Research.
[Piveteau and Smagghue, 2019] Piveteau, P. and Smagghue, G. (2019). Estimating
firm product quality using trade data. Journal of International Economics.
[Portugal-Perez et al., 2010] Portugal-Perez, A., Reyes, J.-D., and Wilson, J. S.
(2010). Beyond the information technology agreement: Harmonisation of stan-
dards and trade in electronics. The World Economy.
[Rauch, 1999] Rauch, J. E. (1999). Networks versus markets in international trade.
Journal of International Economics.
[Santos Silva and Tenreyro, 2006] Santos Silva, J. M. C. and Tenreyro, S. (2006).
The log of gravity. The Review of Economics and Statistics.
[Schott, 2004] Schott, P. K. (2004). Across-Product Versus Within-Product Spe-
cialization in International Trade. The Quarterly Journal of Economics.
[Tintelnot, 2017] Tintelnot, F. (2017). Global production with export platforms.
The Quarterly Journal of Economics.
[Verhoogen, Eric A., 2008] Verhoogen, Eric A. (2008). Trade, Quality Upgrading,
and Wage Inequality in the Mexican Manufacturing Sector. The Quarterly Journal
of Economics.
176