Cities, ethnic wars and water : three essays in

international economics
Tchapo Gbandi

To cite this version:

Tchapo Gbandi. Cities, ethnic wars and water : three essays in international economics. Economics
and Finance. Université de Pau et des Pays de l’Adour, 2020. English. �NNT : 2020PAUU2073�.
�tel-03012102�

HAL Id: tel-03012102

https://theses.hal.science/tel-03012102v1
Submitted on 18 Nov 2020

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est

archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents
entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non,
lished or not. The documents may come from émanant des établissements d’enseignement et de
teaching and research institutions in France or recherche français ou étrangers, des laboratoires
abroad, or from public or private research centers. publics ou privés.
 THÈSE
UNIVERSITE DE PAU ET DES PAYS DE L’ADOUR
École doctorale Sciences Sociales et Humanités (481)

Présentée et soutenue le 09 juillet 2020

par Tchapo GBANDI

pour obtenir le grade de docteur

de l’Université de Pau et des Pays de l’Adour (UPPA)
Spécialité : Sciences Economiques

VILLES, CONFLITS ETHNIQUES ET ACCORDS SUR

L’EAU: TROIS ESSAIS EN ECONOMIE
INTERNATIONALE

MEMBRES DU JURY
RAPPORTEURS

• Julie LOCHARD Professeur à l’Université Paris-Est Créteil (UPEC)

• Mathieu COUTTENIER Professeur à l’Université de Lyon, ENS Lyon

EXAMINATEURS

• Marion MERCIER Chercheur CNRS, LEDa-DIAL, Université Paris Dauphine-PSL

• Serge REY Professeur à l’Université de Pau et des Pays de l’Adour (UPPA)

• Théophile BOUGNA Chercheur, Banque Mondiale

DIRECTEUR

• Fabien CANDAU Professeur à l’Université de Pau et des Pays de l’Adour (UPPA)

CITIES, ETHNIC WARS AND WATER: THREE ESSAYS IN INTERNATIONAL ECONOMICS
Tchapo GBANDI- July, 9th 2020

Thèse réalisée au Centre d’Analyse Théorique et de Traitement des données économiques

(CATT)
Center for the Analysis of Trade and Economic Transition (CATT)

Université de Pau et des Pays de l’Adour (UPPA)

Collège Sciences Sociales et Humanités (SSH)
Avenue du Doyen Poplawski
BP 1633 64016 PAU Cedex
Tél : +33 (0)5 59 40 80 61 / 80 62
E-mail : [email protected]
Web : https://catt.univ-pau.fr

Directeur : Fabien CANDAU

[email protected]

-2-
 Ph.D. THESIS
UNIVERSITY OF PAU (UPPA)
Doctoral School of Social Sciences and Humanities (ED SSH-481)

Ph.D. dissertation in Economics by

Tchapo GBANDI
Publicly defended on
July, 9th 2019

CITIES, ETHNIC WARS AND WATER:

THREE ESSAYS IN INTERNATIONAL
ECONOMICS
Research center:
Center for the Analysis of Trade and Economic Transition (CATT)

COMMITTEE:
EVALUATORS
• Julie LOCHARD Professor, Université Paris-Est Créteil (UPEC)

• Mathieu COUTTENIER Professor, Université de Lyon, ENS Lyon

EXAMINERS
• Marion MERCIER Researcher, CNRS, LEDa-DIAL, Université Paris Dauphine-PSL

• Serge REY Professor, Université de Pau et des Pays de l’Adour (UPPA)

• Théophile BOUGNA Researcher in Economics, World Bank Group

ADVISOR
• Fabien CANDAU Professor, Université de Pau et des Pays de l’Adour (UPPA)
 Remerciements

Avant tout propos, je tiens à traduire toute ma profonde gratitude à l’endroit de tous
ceux et de toutes celles, qui, d’une manière ou d’une autre, ont permis l’aboutissement
heureux de cette thèse. Mes remerciements s’adressent en premier lieu à Monsieur
Fabien Candau, mon Directeur de thèse. Ses remarques, suggestions et critiques, sa
rigueur couplée à sa bonne humeur ont été d’un apport incommensurable. Son amour à
la recherche et sa combinaison parfaite entre travail et sympathie m'ont largement
inspiré. Trois années de collaboration, de travail acharné sous sa direction qui me
semblent déjà être plutôt trois années de plaisir vu l'environnement incitatif qu'il a
instauré. Je reste convaincu que la vie c'est avant tout les bonnes rencontres ; pour faire
court, merci pour tout.
En second lieu, ma reconnaissance va à l'endroit de mon grand cousin, oncle et tuteur
Ephrem, pour son parcours inspirant, ses incessants encouragements et son soutien
sans relâche jusqu'à aujourd'hui ; je lui dis tout simplement “MERCI”.
Je remercie particulièrement Geofroy (co-auteur) et Mamadou, que je peux appeler
désormais mes frères pour leur accueil et disponibilité depuis les premiers jours de nos
rencontres ; je crois qu'ils ont compris qu'ils pourront compter sur moi en toute
circonstance. Qu’ils trouvent ici l’expression de ma reconnaissance. La sagesse de
Carlito (Charles) et le dynamisme de Julie : merci de m'avoir permis de m'intégrer dans
votre équipe, sans oublier Estelle et Marie. Je tiens à remercier tous les membres du
corps professoral et tout le personnel administratif du Centre d'Analyse Théorique et
de Traitement des Données Economiques (CATT) : Mme C. Haritchabalet puis Mme
Cuyalaa, Mme Guenebeaud, Mme Joubert, Mme Perrin, et M. Cassagnard. J’ai
naturellement une pensée spéciale à mes amis et très sympathiques collègues
doctorants pour l’ambiance combien conviviale qu’ils ont su instaurer durant ces trois
années d’étude. Mention spéciale Kevin E., Hervey de Paris, Lemmi, Djawé, Andrad
Varella et Jean-Claude, pour leurs précieux soutiens et leurs aides, sans oublier Tiam,
Leonnel, Isaine, Mamoudou, Denis, Ralph, George et Kaba.
Mes remerciements vont également à l’endroit Monsieur M. Couchoro pour son
indéfectible soutien moral, ses conseils pratiques et ses expériences partagées à tous les
niveaux. Un clin d'œil à Monsieur G. Afawubo, M. Agbodji, Mme Zinsou, à qui j'adresse
mes remerciements les plus sincères pour leurs encouragements et conseils.
Comment pourrais-je omettre Nanou, Fatou et principalement Tanti Saly ; merci de
m'avoir adopté.
Pour finir, j’aimerais rendre un hommage particulier à mes parents, ma grand-mère et
mes sœurs et neveux, qui n’ont jamais douté de moi et qui m’ont toujours soutenu ; vos
prières me vont droit au cœur. A Jeanne et à Diamond, je n’ai pas de mots pour signifier
ma gratitude, merci, je n’ai jamais été loin.

i
Cities, Ethnic Wars and Water:
Three essays in International
Economics

ii
 Abstract

International trade is one of the key factors that have deeply reorganized the world economy.
This thesis addresses three different topics in the field of international economics. It firstly
brings new insights on the contribution of international trade to the geographical distribution
of populations in countries and secondly, determines trade effect on ethnic conflicts.
Afterwards, this document examines the reaction of world trade to international water
agreements that increasingly emerged in recent decades as a tool for sustainable and socially
equitable development. The contributions of this document stem not only from the originality
of the topics discussed and the methodologies adopted, but it also proposes and builds more
appropriate indicators that are better suited to capturing the issues under analysis.
Thus, the first chapter assesses whether international trade is relevant in explaining the
development of large cities in past European colonies. We argue that trade restrictions can
provide an advantage to one city, which may become the platform for exports and
catastrophically attracts people. By contrast, trade liberalization, by providing market access
to other cities, fosters the dispersion of economic activities and consequently a dispersion of
the population. The empirical investigations have led to the conclusion that international
trade does not drive the size of large cities in the former colonized countries. Only institutions
drive the size of primate cities: democracy goes hand in hand with agglomeration.
The second chapter analyses the relationship between trade and insecurity in African
countries. Focusing on ethnic conflicts, this part of the thesis argues that the type of trading
partners determines the opportunity cost of an ethnic conflict. The international trade of
ethnic groups, the regional trade and countries' internal trade may have heterogeneous
effects on peace. The data analyses results support this prediction by pointing out that
international ethnic trade and countries' internal trade are peace-promoting tools unlike
trade between countries that share at least one ethnic group, which appears as a factor of
ethnic conflicts. However, only international ethnic trade seems to reshape the national
identity of countries, which in turn reduces the likelihood of ethnic conflicts.
In the last chapter of the thesis, we discuss the growing concern about the international
management of shared water resources. As water constitutes an important input in the
production of goods, it seems necessary to assess the effectiveness of international water
agreements with regards to trade after several decades of implementation. One can argue that
these agreements are just "cheap talk", or rather pragmatic attempts to meet the sustainable
development objectives to take actions for better water management. Conversely, these
agreements, by allowing a better allocation of water resources between countries, can also
stimulate production (especially agricultural production) and ultimately trade. Thus, this
chapter firstly sheds light on the climate change indicators that are the main drivers of the
increasing implementation of shared water resources' agreements. Secondly, it examines
whether these agreements affect international trade before proposing their overall effects on
countries. In general, the study reveals that water agreements, motivated in particular by high
temperature signals, represent a cost to free trade in Europe, Asia and the Americas, except in
Africa. Therefore, it seems that these water agreements have substantial environmental
protection contents that restrict trade.

iii
 Résumé

Le commerce international est l'un des éléments majeurs qui a restructuré profondément la
vie des nations et de l'économie mondiale. Cette thèse aborde trois sujets différents dans le
domaine du commerce international. Elle apporte en premier lieu, de nouveaux
développements sur la contribution du commerce international à la répartition géographique
des populations au sein des nations et détermine en second lieu son influence sur les conflits
ethniques. Ensuite, elle analyse la réaction des échanges mondiaux aux accords
internationaux sur l'eau qui s'érigent ces dernières années comme un outil important du
développement durable et socialement équitable. Les contributions de ce document, tiennent
non seulement de l'originalité des questions traitées et des méthodes utilisées, tout en
proposant et en exploitant des mesures mieux indiquées pour capter les faits analysés.
Dans le premier chapitre, nous analysons si le commerce international a eu un rôle dans le
développement des grandes villes des anciens pays colonisés. D'un point de vue théorique, un
accès difficile au marché mondial de biens pourrait amener la principale ville à devenir la
plateforme des exportations, attirant ainsi les populations. Dans le cas contraire, il y aura une
meilleure répartition de l'activité économique dans le pays, et par conséquent moins
d'incitation des populations à s'agglomérer dans la capitale. Notre analyse montre que le
commerce international semble ne pas affecter la taille des grandes villes dans les anciens
pays colonisés. Les institutions restent par contre déterminantes pour expliquer la taille des
villes : la démocratie est source de concentration des populations dans les grands centres
urbains des pays en développement.
Le second chapitre examine la relation entre le commerce et l'insécurité dans les pays
africains. En nous concentrant sur les conflits ethniques, cette partie met en avant le fait que
le type de partenaires à l'échange déterminerait le coût d'opportunité à un conflit ethnique.
Les résultats obtenus valident cette intuition en démontrant que les échanges internationaux
des ethnies, le commerce régional et le commerce interne d'un pays ont des effets
hétérogènes sur la promotion de la paix. Ainsi, il ressort que le commerce international des
ethnies et le commerce interne des pays ont des effets pacificateurs contrairement au
commerce entre pays qui partagent au moins un groupe ethnique. Cependant, seul le
commerciale international des ethnies semble remodeler l'identité nationale qui à son tour
réduit la probabilité de conflits ethniques.
Le dernier chapitre se concentre sur la fulgurante augmentation des coopérations
internationales concernant la gestion des bassins d'eau communs. L'eau étant un élément
essentiel des processus de production des biens, il conviendrait de savoir, après plusieurs
décennies d'implémentation, l'impact de ces accords sur le commerce. En effet, trois
hypothèses concurrentes peuvent être émises. Il se pourrait que ces accords ne soient que des
discours politiques sans conséquences réelles n'impactant pas le commerce. Il est aussi
possible que ces accords se matérialisent par des tentatives pragmatiques visant à respecter
les objectifs de développement durable et dans ce cas, il est probable que ces accords ont un
effet négatif sur le commerce. A l'inverse, ces accords permettant une meilleure allocation des
ressources en eau entre pays, peut aussi stimuler la production (notamment agricole) et in
fine les échanges commerciaux. En général, les résultats révèlent que les accords sur l'eau,
motivés en particulier par les hausses de température, représentent un coût au libre-échange
en Europe, en Asie et en Amérique (mais pas en Afrique), il semble donc que ces accords sur
l'eau ont un contenu de protection environnementale important qui limitent les échanges
commerciaux.
iv
 Contents

Remerciements .................................................................................................................................. i
Abstract .............................................................................................................................................. iii
Résumé ................................................................................................................................................iv
Contents ............................................................................................................................................... v
List of figures .................................................................................................................................. viii
List of tables....................................................................................................................................... ix
General introduction ....................................................................................................................... 1
Trade and Institutions in Explaining Urban Giants: summary .............................................. 10
Beyond the Income Effect of Trade Integration on Ethnic Wars: summary ..................... 13
Climate Change, Water Treaties and International Trade: summary ................................. 16
I. Chapter 1 - Trade and Institutions: Explaining Urban Giants ............................... 21
Introduction ............................................................................................................................................... 21
I.1 Hypothesis, data and preliminary results ......................................................................... 25
I.1.1 Empirical Strategies .......................................................................................................... 25
I.1.2 Data .......................................................................................................................................... 29
I.1.3 First results ........................................................................................................................... 34
I.2 IV Strategy ..................................................................................................................................... 36
I.2.1 Two steps SLS identification strategy with a zero-stage .................................... 37
I.2.2 Results of the Zero-Stage ................................................................................................. 39
I.2.3 Results of the first stage................................................................................................... 40
I.2.4 Results, second stage ........................................................................................................ 41
I.2.5 Non parametric estimation: matching estimator approach .............................. 43
I.2.6 Discussions ........................................................................................................................... 45
I.3 Conclusion ..................................................................................................................................... 46
I.4 Appendices .................................................................................................................................... 47
I.4.1 Appendix A: Light nights and urban primacy.......................................................... 47
I.4.2 Appendix B: Outdegree Indicator of Network ........................................................ 48
I.4.3 Appendix C: Theoretical foundation of the market access ................................. 50

v
 I.4.4 Appendix D: Alternative measures of democracy in the zero stage
regressions ............................................................................................................................................ 51
I.4.5 Appendix E: Deepening the analysis ........................................................................... 53
II. Chapter 2 - Beyond the Income Effect of Trade Integration on Ethnic Wars ... 56
Introduction ............................................................................................................................................... 56
II.1 1 Empirical strategy ................................................................................................................... 60
II.1.1 A brief review of the literature ..................................................................................... 60
II.1.2 International, regional and internal trade at the ethnic group level.............. 62
II.1.3 Estimation ............................................................................................................................. 67
II.2 1 Results about the effects of trade on conflicts ............................................................. 69
II.2.1 Trade and conflicts without income effect ............................................................... 69
II.3 Inference from a two-step strategy ..................................................................................... 73
II.3.1 Empirical strategy .............................................................................................................. 73
II.3.2 Results .................................................................................................................................... 77
Conclusion .................................................................................................................................................. 81
II.4 Appendices .................................................................................................................................... 82
II.4.1 Appendix A: First resuts .................................................................................................. 82
II.4.2 Appendix B: First-stage IV estimates, explaining ethnic groups'
international trade ............................................................................................................................. 84
II.4.3 Appendix C: First-stage IV estimates, explaining regional trade ..................... 85
II.4.4 Appendix D: Robustness check analyses of the impact of Regional trade
with imports ......................................................................................................................................... 86
II.4.5 Appendix E: First-stage IV estimates, explaining countries’ local trade ....... 88
II.4.6 Appendix F: First-stage IV estimates .......................................................................... 89
III. Chapter 3 - Climate Change, Water Treaties and International Trade ........... 91
III.1 Introduction .................................................................................................................................. 91
III.1 About water treaties .................................................................................................................. 95
III.2 Estimating Elasticity ................................................................................................................101
III.3 The Structural Gravity Equation .........................................................................................101
III.4 Data and estimators .................................................................................................................103
III.5 Baseline results ..........................................................................................................................105
III.6 Heterogeneity analysis ...........................................................................................................106
III.7 Water treaties versus Regional Trade Agreements on Water .................................109
III.8 General Equilibrium Analysis...............................................................................................111
vi
 III.8.1 Methodology and data ....................................................................................................112
III.8.2 Results ..................................................................................................................................115
III.9 Conclusion ...................................................................................................................................118
III.10 Appendices ..............................................................................................................................120
III.10.1 Appendix A: Climate reasons of water agreements, unilateral estimation
approach 120
III.10.2 Appendix B: Reverse causality test .......................................................................122
III.10.3 Appendix C: Sensitivity of results to changes in trade elasticities............123
General conclusion......................................................................................................................125
References ......................................................................................................................................129

vii
 List of figures

Figure 1 : World exports trend ................................................................................................................... 1

Figure 2 : Cumulative function of water agreements ..................................................................... 10
Figure 3: Urbanization Primacy in 2010 and Density in 1500 ................................................... 23
Figure 4: World Trade Network in 1900 ............................................................................................. 48
Figure 5: Crops production imagery: cotton and cassava ............................................................ 64
Figure 6: Mineral resources location in Africa at the Ethnic/country level .......................... 66
Figure 7: Water Treaties Contents......................................................................................................... 97
Figure 8: The Impact of No Water Treaties on Trade (GETI) ....................................................117

viii
 List of tables

Table I-1: Urban Reversal.......................................................................................................................... 36

Table I-2: Democracy (ANRR index), zero stage .............................................................................. 39
Table I-3: Market Access, zero stage ..................................................................................................... 40
Table I-4: Democracy and Trade, First stage ..................................................................................... 41
Table I-5: Trade, Institutions and Urban Primacy ........................................................................... 43
Table I-6: Average effect of democracy on log urban primacy ................................................... 44
Table I-7: Effect of the constructed light night gap on urban primacy .................................... 47
Table I-8: Democracy (Polity 2 score), zero stage ............................................................................... 51
Table I-9: Democracy (PR, CL), zero stage ........................................................................................... 52
Table I-10: First and Second Nature of Urban Giants ........................................................................ 54
Table II-1: International Trade effect on ethnic wars onset ........................................................ 70
Table II-2: International Trade effect on ethnic wars onset ........................................................ 73
Table II-3: First stage regression............................................................................................................ 78
Table II-4: Second stage regression: Country fixed effects regression .................................... 80
Table II-5: Results from naïve regressions ......................................................................................... 83
Table II-6: IV first stage regressions: explaining ethnic groups' international trade ........ 84
Table II-7: IV first stage regressions: explaining regional trade ................................................ 85
Table II-8: Regional Trade effect on ethnic wars onset ................................................................. 87
Table II-9: IV first stage regressions: explaining local trade ....................................................... 88
Table II-10: IV first stage regressions explaining country total Export and local trade in
the Two-stage methodology ..................................................................................................................... 89
Table III-1: Climate reasons of water agreements: gravity estimation approach .............100
Table III-2: Baseline results ....................................................................................................................106
Table III-3: Heterogeneity analysis .....................................................................................................108
Table III-4: Heterogeneity analysis .....................................................................................................111
Table III-5: Direct, PIE, GETI effects of water agreements .........................................................116
Table III-6: Climate reasons of water agreements ........................................................................121
Table III-7: Strict exogeneity test of water treaties ......................................................................122
Table III-8: Doubling and halving trade elasticities ......................................................................123

ix
 General introduction

General introduction

Let begin this thesis with a popular quote from Adam Smith that goes: “Every man lives
by exchanging”. Indeed, this saying is not only true at the individual level, it also holds at
the aggregate scale: international trade is a fundamental factor of economics growth,
explaining a large part of the “Great Convergence” of emerging countries (Baldwin,
2017). Any country can't produce all the goods and services that people needs in its
homeland. The sharp decrease in trade costs (Jacks, Meissner, and Novy, 2008), the
growth of income (Baier and Bergstrang, 2001) and the political consensus about trade
liberalization (sometimes called the “Washington consensus”, see Rodrik, 2006) have
stimulated bilateral trade flows1 between countries at an impressive rate these latter
decades.

Figure 1 : World exports trend

Source: Author, replication from data of Federico and Tena-Junguito (2016). The
value of global exports, Time of value of world exports (million $) at constant
prices, relative to 1913.

1 Data in the figure are from https://ourworldindata.org/trade-and-globalization

1
 General introduction

However, it is obvious that this increasing level of international interactions is not

immune of effects on countries domestic’s characteristics. At the same time, the level of
trade should be affected by decisions countries make either domestically or
internationally, knowingly or unintentionally. Then, international trade, by connecting
nations all around the world has structurally affected countries economically,
institutionally, socially and spatially. This issue becomes so central to world politics as
the technological progress made in the last century has deeply connected people in the
world as ever.

As a result, international trade remains a topical issue. From development policies

perspective, trade has been so important that development institutions such as the
General Agreement on Tariffs and Trade (GATT) and the World Trade Organization
(WTO) were created, to smoothen bilateral transactions of goods and services. The
GATT was specifically created to reduce trade restrictions in order to promote
exchanges, reduce prices, enhance an efficient use of factors, and increase employment.
In development countries, trade is seen as a powerful means that can help lift people out
of poverty (despite the ambiguous results obtained by researchers, see Winter et al.
2004). In this vein, the goal of the WTO is not only to be a platform that settles world
market issues but, it acts also as an international think tank that responds to developing
countries needs in terms of trade. Besides, a lot of Regional Trade Agreements (RTA),
which are reciprocal treaties, has been promoted to foster exchanges in the world.
Besides, some single country use to propose to particular developing countries, different
strategies that can help them alleviate extreme poverty, by setting up initiatives like
General Scheme of Preferences (GSP). By doing so, the opportunity is given to vulnerable
countries to benefit from a scheme of tariff preference that removes imports duties
when exporting their products to the granter’s homeland.

It clearly appears that the political, social, and economic importance of trade is not to be
neglected. After all, it seems obvious that international trade affects development in
many aspects. This thesis tries to analyze two of them by answering the following
questions respectively in the first and second chapter: (i) Does trade affect the spatial
distribution of population within countries? (ii) To what extent does trade influences
ethnic conflicts? In the third chapter, instead of analyzing a link that goes from trade to
2
 General introduction

development, (iii) we try to comprehend the way trade reacts to an important input in
goods production: the so called “water”. More precisely, the question we raise in this
part of the thesis is whether trade reacts to international water treaties.

International trade may be a factor of agglomeration. As a matter of fact, a sharp growth

of cities' populations and particularly the increasing number of metropolitan areas, have
made urban planning a more serious matter in public policy agenda. The channels that
drive this growth of big cities are mainly the industrialization led by rural-urban
migration. According to the World Urbanization Prospects, the urban population in the
world has grown rapidly to about 4.2 billion people in 2018, more than four times the
urban population in 19502. Thus, urbanization appears to be a topical issue since it
concerns more than a half of the world's population (55% according to World
Urbanization Prospects in 2018). About seven out of ten people (68%) of the world
population is projected to be living in urban areas by 2050, following the projection of
the United Nation3. Besides, the UN World cities' report of 2016 reveals that more than
40% of the urban population is concentrated in a single city of more than one million
inhabitants in 28 countries4.

Urbanization has advantages but it also has drawbacks. In fact, urbanization, through
density and competitivity, paves the way for industrialization, growth and employment
thereby sustaining the agglomeration. As a consequence, in case of poor urban planning,
congestions issues, high level of unemployment rate, slums, environmental issues (CO2
emission) and inappropriate living conditions might emerge. Thus, before thinking of
policies that can help manage urban areas, it seems necessary to determine the causes of
its evolution. Why do large cities in terms of population still attract people? Have these
urban giants been affected by the distant past distribution of economic activities?

Based, among other sources, on the theoretical works from Krugman and Livas (1991)
and Krugman (19965), one of the seminal empirical works that investigates the link
between trade and urbanization goes back to Ades and Glaeser (1995). They analyzed
two reasons behind urban population growth: international trade and institutions. The

2 751 million of people in 1950.

3 https://www.un.org/development/desa/publications/2018-revision-of-world-urbanization-
prospects.html
4https://www.un.org/en/development/desa/population/publications/pdf/urbanization/the_worlds_citi

es_in_2016_data_booklet.pdf
5 The working paper of this research cited in the article of Glaeser (1995) is from 1992

3
 General introduction

theoretical line of research from Krugman and Livas (1996) leads to the conclusion that
closed domestic markets are a key factor that drives the emergence of urban giants.
They argue that producers are more likely to choose production sites with good access
to consumers to save transport costs. This will attract job seekers which will in turn fuel
the demand, promoting a self-reinforcing of the agglomeration process. Accordingly, this
will ensure a better access to intermediate inputs because of the proximity of the other
firms. Urbanization is then driven by the concentration of the demand and the inputs in
the same city. As a matter of fact, exporting from this agglomeration area is a key added
value for firms, since they can benefit not only from economy of scale as they can
directly rely on their existing infrastructure, but they can also benefit from a competitive
work force and the proximity of intermediate inputs. On the contrary, in case of low
transaction costs, goods can be shipped easily not only from the main city but also from
the interland. That is part of the point made in the model of Krugman and Livas (1996).
Besides, institutions could be part of the causes behind the spatial distribution of the
populations (Ades and Glaeser, 1995). Actually, governments when managing policies,
can purposely or unconsciously, influence the spatial distribution of the economic
activities in their countries and therefore that of the population.

For the sake of clarity on the main cause(s) of urban giants, we revisit in this first
chapter the work of Ades and Glaeser (1995) by handling the analysis on a panel data
instead of using cross section observations. Thus, the study focuses on primate cities,
define as the largest city in a country, which is by far, larger than any other city in the
urban hierarchy.

This work departs from the existing literature because it handles the explanation of big
cities growth by carrying out the analysis on only past colonial countries. Actually, the
historical events due to colonization make these countries very interesting in terms of
old and current institutional patterns. To conclude, we address the three (3) following
questions in the first chapter:

 Does trade matter in explaining primate cities populations?

 Apart from trade, do past institutions affect the population's location choice?
 Finally, we question if urban giants react regarding current institutions?

4
 General introduction

International trade seems also to be peace-related. One of the greatest and famous
policies that place international trade for the first time at the core of its strategy goes
back to the world wars period in Europe. Actually, the ignominy of these wars and the
fear to see it happen again inspired policy makers as well as peace and development
practitioners to rely on trade as a channel of promoting peace. Different theoretical
frameworks have been proposed to highlight the background of these insanities. They
argue that, in case countries depend on each other through trade flows, the opportunity
cost of war will increase, encouraging nations to engage some efforts for peace
(Montesquieu, 1958; Kant, 1995) when a potential conflict is coming into play in their
relationships. Thus, the spread of international trade and the gain from the exchanges
brought new paradigms that should prevent from disputes between countries because
they depend on each other. When this analysis holds at bilateral level, it may be noticed
on the one hand that, gains from international trade can thereby cause frustrations.

The causes of ethnic wars can be social, economic or political. In an economic point of
view, new paradigms can exacerbate the local probability of ethnic conflict or reduce it
through the way they redefine the ties between people. There is an increasing challenge
to peace and stability in Africa these last decades than ever. Countries such as Rwanda,
Sierra Leonne, Congo, Nigeria, South Africa have been affected a lot by this issue. At
least, one country over two has experience civil war in this continent. Apart from having
the highest ethnic diversity, Africa comparatively to the other continents, is also
characterized by a higher incidence of war (Osinubi and Osinubi, 2006 6). Though,
Elbadawi and Sambanis (2000) argue that these civil wars were not the consequence of
the ethnic fragmentation. Following Michalopoulos and Papaioannou (2015), we
consider in our context as ethnic wars, the conflicts characterized by political violence
which involve a use of force by a group with a political purpose or motivation. This
violence does not only record conflicts that take place within the context of a civil war; it
also accounts for violence against civilians, militia interactions, communal conflicts and
rioting.

How can trade affect conflicts in general? Massimiliano (2015), while proposing how
trade can increase resilience in fragile states, exposes three main mechanisms that can
lead to conflicts: the decrease in the opportunity cost of peace in case of a decline in real

6 https://www.tandfonline.com/doi/abs/10.1080/09718923.2006.11978376

5
 General introduction

income, the rapacity behavior that leads to violent competition for point-source
commodities, and the government (or rebel) taxation that can provide the means to
suppress (or enhance) violence.

In Africa, Rwanda is for instance the country that hosted the most recent and insane
ethnic war that shocked the world. According to the United Nations (UN), this genocide
caused more than 800 thousand deaths, mostly Tutsi in Rwanda only in four months in
the year 1994. Historically, even if the difference between Hutus and Tutsi had been
culturally accepted by Rwanda people, some argue that these differences had been
deepened by colonialism that brought them gradually to the war. On the one hand, one
can argue that if internal trade were developed in Rwanda, ethnic groups (Hutu and
Tutsi) would be much more tied together, that could increase the opportunity cost of
ethnic war7.

On the second hand, the intensity and duration of an ethnic war can be fueled by the
support of peers at the other side of the country’s border, by supplying them, goods,
arms, and refuge. In this vein, it is possible that, countries which have partitioned
ethnics groups will be more likely to experience ethnic group violence and/or to see it
aggravate, in case they develop trade relationship over their brotherhood link. Indeed,
this issue is important since during the formation of African states, 97.8% of all borders
split ethnic homelands in multiple parts8, meaning that almost all African countries are
concerned with split ethnic groups issues. Particularly, about 39.5% of the 827 ethnic
groups in Africa are partitioned (Michalopoulos et al., 2016). Then, trade between same
ethnic groups across a border can exacerbate the social distance with homeland ethnic
groups with any particular relationship. This trade is not to be neglected. Aker et al.
(2014) find that common ethnicity lowers the transaction costs particularly for
agricultural trade9. Finally, following "the Land of a Thousand Hills" history, the Hutus
who are much more farmers, have been excluded from the central government, and
international trade (particularly that of coffee) has been part of the issue (Kamola, 2007)
that brought them to conflict. On the third hand, the domestic integration (geography)
that may facilitate international trade integration of units that comprise these countries
can affect the gains of the different parts. Especially, ethnic groups that benefit from the

7 http://eprints.mdx.ac.uk/3809/
8 http://www.mathiasiwanowsky.com/TradeandEthnicity.pdf
9 https://www.sciencedirect.com/science/article/abs/pii/S0304387813001478

6
 General introduction

exports either because they can better ship their products, or because they have special
endowment factors, will have any incentive to engage in a dispute. As a consequence, the
incentive of an ethnic group conflict depends on the kind of trade partners. Therefore,
the main question of Chapter 2 is:

 How do ethnic wars in Africa react to ethnic groups’ international trade, regional
and internal trade of countries?

What about the trade and water agreements? A large part of the population in the world
is not aware of the quantity of water used during the production process of the goods
they buy or consume. For instance, in average 15400𝑚3 of water is used and polluted
when producing one ton of meat from beef cattle, the equivalent of six Olympic
swimming pool10 water capacity; either around 15400 liters of water for a kilogram of
this meat11. 3265 Liters of water are involved and polluted in the production of one
kilogram of eggs, in other words 206 liters of water per egg12. We must also be reminded
that, on average, 63213 liters of water are consumed and polluted for the production of a
bottle of 0.75 liter of wine (Bonamente et al., 2015). In the extractive industry, gold has
the highest level of water footprint, estimated on average at 716.000 liters per kilogram
of this metal (Hoekstra, 2015). As for crops, the water footprint statistics estimate at 214
liters, the quantity of water involved on average in the production of a kilogram of
tomatoes. Thus, when shipping goods from one place to another, one must bear in mind
not only the fact that a lot of water has been involved in their production, but also that
some goods transported can have water content.

As it is known, water is one of the most important substances on earth, necessary for the
survival of animals and plants. If there was no water, there would be no life on earth.
Therefore, water resources constitute a key resource to achieve the sustainable
development goals. It is essential to produce energy, to defend public health care,
sustainable livelihood, ecosystem and sustainable cities (OECD). As noticed by World

10 According to FINA (Fédération Internationale de Natation: http://www.fina.org/), an Olympic

swimming pool contains on average 2500 m3 of water
(https://www.fina.org/sites/default/files/finafacilities_rules.pdf)
11 https://waterfootprint.org/en/water-footprint/product-water-footprint/
12 If we conjecture that an egg weights 63 grams
13 i.e. more than 800 bottle of 0.75 liter capacity.

7
 General introduction

Bank14, water is at the core of every aspect of development and therefore constitutes an
imperative tool for the Sustainable Development Goals (SDG). It drives economic
growth, supports healthy ecosystems, and is essential and fundamental for life itself.
Nevertheless, regarding the increasing population growth, water becomes a scarce
resource. China is for instance one of the countries that is to water shortages and water
pollution.

Water issues become more serious because of the sur-exploitation of water resources
and particularly regarding environment challenges. A lot of water is used, displaced or
consumed in everyday life: for cooking, for sanitary needs, in industries, in agriculture,
etc. Let consider the mining sector activities, particularly the extraction of gold. It is
common that manual techniques that consist of panning are usually used to separate
gold from the sand, stones and other sediments. Hence, a large quantity of water is used
in the process. Then, apart from the fact that some rivers are emptied of their resources
and/or deviated for the sake of gold extraction, chemical and other pollution from the
panning areas can highly cause also damages to the ecosystem, and even affect the
health of the populations that depend on these resources, that is to say the riparian15.
Since it is found in small quantities, gold mining operations tend to cover wide areas
(cite this16). As a matter of fact, the gold mining boom in the amazon rainforest has been
the cause of the Amazon river chemical pollution. This presents a real issue since the
Amazon river is shared by at least four (4) countries.

We can also observe this fact in agriculture. Farmers located upstream of rivers that
cross several countries, by irrigating their farms, could pollute these latter making them
unsuitable for all those located downstream. The same picture is observed during the
construction of dams that obstruct the flow of water and reduce it for countries
downstream of rivers. For instance, normally a dam helps population by bringing
electricity, but it can betray the provision of countries downstream to the rivers or lakes.
As a matter of fact, the dam of Selingue (in Mali, 140 km from Bamako) has raised a lot of
voices. Apart from the riparian in Mali, and any other associations of users that were
afraid for the availability of water henceforth, the countries that also share this lake also
reacted to the creation of the dam. Actually, for the beginning of the power station, a

14 https://www.worldbank.org/en/topic/water/overview
15 https://www.brilliantearth.com/gold-mining-environment/
16 file:///C:/Users/t/Downloads/15_Abdul-Wahab.pdf

8
 General introduction

basin of 2.2 billion cube meters of water had been created for the need of the central,
and this operation had needed 6 months. Then riparian countries (and water
consumption association in Mali) requested for international collaborations/meetings
on the water in order to fix the rules of its usage.

As a consequence, the overexploitation of water resources can deeply affect water

provision and trigger tensions since there are some populations that could be negatively
impacted by the miss-exploitation of the resource. There is therefore a special need to
promote international cooperation or initiate joint management collaborations to
equitably manage and/or preserve water resources. In this vein, water treaties come
like crucial policy initiatives that pave the way for better distribution and/or efficient
use of water resources. Water management becomes a topical issue for policies in the
main interest of the populations, the health and for sustainable development. In recent
decades, a lot of water agreements have been signed between countries (Figure 2). The
targets of these treaties are not always the same; when some of the agreements main
objective is to preserve water quality, others ask for environment preservation, or join
management purpose. From that point of view, these water agreements should act as
restrictive policies against water sur-exploitation preventing from its miss-usage.

9
 General introduction

Figure 2 : Cumulative function of water agreements

Sources Authors

This chapter assesses the added value of these water-related collaborations at the
international level. It investigates particularly the following question:

 How (and where) does water agreement affect international trade?

To the best of our knowledge, no study has already highlighted this relation before.

Trade and Institutions in Explaining Urban Giants: summary

The urban empirical literature provides a rule which states that the size distribution of
cities within countries follows a powerful law17. This empirical observation known as
Zipf's law states that: in countries, the largest city is about twice the size of the second
largest city, three times the size of the third largest city, etc. But, in some countries the
cities size distribution does not follow this law since the population size in the largest
city is by far more than twice that of the second largest city in the urban hierarchy. This

17 https://academic.oup.com/qje/article-abstract/114/3/739/1848099?redirectedFrom=fulltext

10
 General introduction

chapter aims to highlight the causes behind this deviation from the law: what are the
main factors of the urban primate size?

To do so, we follow Ades and Glaeser (1995) who considered two main factors when
explaining urban giants’ growth; institutions and trade. It is possible that, population
would prefer to live in the city, attracted by industries, firms and businesses. In fact,
these latter would choose to be close to the administrative center in order to increase
their probability to get permissions, authorizations and grab opportunities. Besides,
governments can opt for fiscal centralization in order to take over all the cash channels.
In other case, a government may also prefer decentralization in case it feels weak in
managing heavy urbanization.

Political institutions are not the only drivers of urban populations. In case of bad access
to international market, the main city is likely to become the platform of exports and will
therefore attract people. Nevertheless, when the other cities can easily get access to the
international market, the migration to the main city will certainly decrease, thus
relieving the big city.

To do this, we concentrate our study on past European colonies because they show up
some institutional particularities that rooted during colonization period. We observe
that, countries that were densely populated hundred years ago have now small central
cities while we find no relationship for the non-colonized countries. This picture shown
for the first time in the field of urban economics is analogous to that of Acemoglu,
Johnson and Robinson (2002). By following their argument, we argue that colonizers
had settled extractive institutions in rich places18 pushing people out of the cities. This
provides a supplementary demonstration of the fact that the density in 1500 is an
indicator of past institutions. Then, past colonized countries present interesting patterns
in terms of institutions since colonization had deeply affected their long-term
institutional trend.

When highlighting the causes behind the urban primates, we depart from existing
literature in two main points. Firstly, we propose a much more extensive measure of
market access, instead of using direct measure that only captures part of trade frictions.
Secondly, since Head and Mayer (2002) weighted distance accounts for cities

18 Populated places were certainly rich places.

11
 General introduction

population, we propose our own measure of bilateral distance to avoid reverse causality
bias, as our dependent variable is also a measure of cities population. Finally, on the
contrary to Ades and Glaeser (1995) we rely on instrumental variables methods to
tackle the endogeneity not only between urban giants' growth and trade, but also
regarding institutions. As for institutions, the historical mortality rates of soldiers,
laborers, and bishops from Albouy (2012) during colonial times appears to be a good
candidate as an excluded instrument19. However, the invariability of this instrument has
brought us to propose a year by year zero stage regressions strategy that allows us to
get time-varying instrument. We interpret this as the exogenous consequence of distant
past shocks in nowadays institutional patterns. A more contemporaneous indicator by
relying on the regional waves of democratization and transitions to non-democracy
from Acemuglu (2017) is also used as instrument. As for the endogeneity of trade, by
following the previous techniques, we rely on past genetic distance as a proxy for
cultural proximity, which may impact the trust between partners and then on trade
exchanges. We also introduce for the first time in the empirical a new control that was
not available until now. This allows us to capture the rural-urban income gap, using
intensity of night lights from outer space.

The chapter is separated in two main parts. In the first part, we investigate whether past
patterns shape contemporaneous trend of primate cities. We found that, past density is
an indicator of bad governance since results reveal that, on average a country that was
twice more densely populated in 1500 is now a country where the share of its biggest
city in the total population is smaller than in other countries. In fact, countries densely
populated in 1500 were richer, increasing the incentive to settle extractive institutions
that act as repulsive forces for the native population. We also found a negative influence
of the past network of international trade on contemporaneous size of primate cities.
The first globalization has favored land tenants (in Latin America for instance Findlay
and O’Rourke, 2009) and thus urbanization was not profitable for these elite.

In the second part of this chapter, we provide instrumental variables estimates based on
instruments presented above. On the contrary to Ades and Glaeser (1995), all our
estimates show that the indicator of democracy, instrumented by past mortality rates

19 For more discussion on this instrument, see also De Sousa and Lochard (2009).

12
 General introduction

and by waves of democratization, favors agglomeration in the colonial countries' biggest

city. On the contrary, market access seems to not drive primate populations.

After all, the main conclusion from this chapter of the PhD thesis is that, only first cause
(i.e. distant past history) and institutions matter in explaining the urban giants’ size, not
trade. When spreading good institutions in terms of democracy and governance,
governments consciously or unconsciously favored the concentration of populations in
the biggest city. Then, it is worth noting that the channels that drive this relation must
be explore in order to provide public policies settlers some recommendations that can
encourage populations to stay and/or go in the other areas while steel enhancing their
institutions.

Beyond the Income Effect of Trade Integration on Ethnic

Wars: summary

This chapter tries to shed light on three different type of trade: international trade,
regional trade and internal trade. Its contribution is to propose different measures and
methodologies to precisely analyze the geographical impact of trade on conflicts. We
begin the debate here by exposing the three main mechanisms provided in the literature
that are used to justify the relation from trade to war. Firstly, when trade shocks
decrease revenue, the cost to participate to violent actions follows the same path. When
trade increases tax revenues, it in turn improves the redistributive capacity of the
government, reducing in fine people’s incentive to get involve in conflicts. Finally, the
literature also argues that resource that is highly valuable on international market
generates envy and expropriation that can lead to an escalation into conflicts. In
summary, the common point that emerges from these three mechanisms is that trade
affects the likelihood of conflicts through the income channel. As a consequence, if we
perfectly control for income, the trade effect must be insignificant. What if the trade
effect remains significant? Is there another mechanism that can fuel the escalation into
ethnic war apart from the pecuniary channel?

Actually, markets known as places of socialization may foster trust between partners.
But, in our context, in case of ethnic favoritism, business' owners that do not belong to

13
 General introduction

the right ethnic group could be stacked, paving the way to potential conflicts. Trade can
also affect the culture. On the one hand, globalization can cause cultural destruction
which in turn will increase the vulnerability of the communities (Maystre et al., 2014).
On the other hand, globalization by promoting new national identities/culture could
redefine the relation between ethnic groups. Trade can foster the destruction of the
environments through deforestation or can expose some localities to pollution that can
potentially lead to a conflict. This leads to the conclusion that trade can affect the
probability of war via non-pecuniary channel.

As for international trade that develops relation with people far from their country
homeland, it may also weaken institutional and cultural link with neighbors in the same
country. On the contrary, the international trade of ethnic groups might be a peace
promoting instrument. Secondly, we approximate the strength of the ties between ethnic
groups with internal trade and investigate how it affects the probability of ethnic war.
We assume that the opportunity cost of war will increase since the parties within a
country have a lot to lose. Local trade by reinforcing ties between communities inside
the country will have a peaceful effect since it increases the opportunity cost of war of all
the groups engaged. In some case, such as Nigeria, potential conflict resolution
structures are set up to prevent latent or potential conflicts20. Regional trade, for his
part, represents a diversion of socialization inside a country, but as argued by
Iwanowsky (2018) ethnic networks provide social collateral to enforce contracts and
can thus compensate for unfavorable institutions between countries. In fact, the
intensification of trade with peers across borders could decrease the dependence with
the others ethnic groups from the same country. We expect that, countries that are
partitioned are more likely to experience ethnic war as long as they depend more on
their peers at the other side of the border in term of trade: the ethnic network by being
empowered by trade, may deter the relations with the other ethnic groups from the
same country.

In order to approximate ethnic groups’ income, we follow Henderson et al. (2012) by

using night light pictures provided by satellites from outer space. These data are useful
to approximate income particularly in developing countries because these latter often
lack of reliable data at sub and/or supranational level.

20 https://ageconsearch.umn.edu/record/11837

14
 General introduction

We then evaluate in which sense these three kinds of trade affect ethnic war. Firstly,
international trade is defined as the value of goods addressed by an ethnic group out of
its country homeland. We got this variable by building for the first time in the economic
literature an indicator on ethnic exports by relying on a weighting scheme based on the
kind of goods produced by the ethnic group. Firstly, we identify the spatial locations of
over 45 agriculture products and 48 mining products across Africa that allowed us to
compute a weighting scheme that indicates the level of participation of every ethnic
group to its country exports. With this weighting scheme at hand, we use BACI trade by
products database to compute ethnic groups exports. The regional trade is built by
considering the total export of a country to its' partners with which it shared at least one
ethnic group. Finally, the internal trade designs the quality of relationship between
groups within the same country.

As it appears in the data description, our study relies on two types of data: the ethnic
level time-varying variable (the international trade of ethnic groups), and country level
time-varying variables (the regional trade and the local trade). Regarding this, this
chapter comprises two main sections. In the first section, we use a commonly a widely
known strategy to identify the effect of trade on war. But at least three caveats can be
formulated. Firstly, the measurement of ethnic trade potentially suffers from errors that
may bias the analysis, particularly with respect to the country level variables such as
regional trade and local trade. Secondly, by merging aggregate data to subnational level
observations, the OLS standard errors can be biased downward leading the aggregated
variables to be artificially significant (Moulton, 1990). Thirdly, the methodology in the
first section does not allow us to gauge the reaction of the ethnic groups' conflicts to
their national context. Hence, we borrow from the urban and labor economics literature
a two-step empirical analysis which allows us to overcome the weaknesses in the first
section. So, the first stage regression allows us to estimate the country-year effects that
depict the changes in national identities that influence conflicts after controlling for
income and unobserved ethnic fixed characteristics.

In general, the results reveal that, the likelihood of peace increases when ethnic groups
get involved in international trade and internal trade and decreases with regional trade.
On the contrary, only international trade appears to affect countries path in explaining
ethnic conflicts. As a consequence, we can therefore consider that only the international

15
 General introduction

trade integration has the potential to shape national identities and to reduce the risk of
ethnic wars: the performance on the world market fosters mutual confidence,
cooperation and good institutions inside nations that reduce the likelihood of ethnic
conflicts.

Climate Change, Water Treaties and International Trade:

summary

The literature on international trade has long studied the role of international
collaboration on trade. In this vein, researchers try for instance to comprehend the
effectiveness on regional trade agreements (RTA), and general agreement for trade for
trade facilitation such as GATT, WTO, etc. Then, the whole literature literally ignores
how international water cooperation can, for its part drive international trade. Actually,
the increasing number of environment issues, coupled with the sharply growth of the
population, have seriously affected water resources. The predictions for the upcoming
centuries are even more frightening since climate change will be coupled with dramatic
and significant reduction in precipitation and major changes in respect to the seasons.
Water being an important input, its scarcity may affect goods production and
thenceforth trade of agricultural and industrial goods characterized by a high-water
footprint. These climate changes are at the bone for sustainable development initiatives,
which comprise also institutional arrangements between nations aiming to manage
water resources. This is of particular importance since more than 600 freshwater-
related treaties have been signed to manage transboundary basins, according to the
International Freshwater Treaties Database.

The purpose of this article is threefold. In the first section, this research analyzes how
climate change drives the enforcement of Water Treaties (WTs). To date, we have no
clue of what kind of climate change motivates countries to cooperate. Our variable is
then a dummy indicator of the existence of a water agreement between two countries in
a given year: it takes one when the two countries have signed a new agreement in a
given year and zero otherwise. In the second section, this chapter aims at analyzing in to
what extent these treaties have impacted international trade. We argue on the one hand
that water treaties may have little or no effect because they engender too small
16
 General introduction

additional costs on production to influence international trade, or because they are just
“cheap talks” and are not a real commitment to implement significant protectionist
policies on water. In the second hand, we think that these treaties contribute to resolve
and/or prevent conflict by fostering cooperation between riparian nations and as a
result promote institutions and economic conditions thereby increasing trade. Indeed,
an agreement on the usage of water upstream a river, may make available more water to
the riparian located downstream the same river, fostering therefore their trade. Also,
agreements on dams may stimulate the production of goods and thereby international
trade. On the third hand, the possible negative effect of water agreements on trade
stems from the fact that, they are implemented in order to save freshwater ecosystems
and/or to foster a sustainable development. In the last section, we rely on the estimates
in section two to provide a general equilibrium model that allows us to assess not only
the global effect of the water treaties but also counterfactual analyses consisting of
suppressing all the WTs implemented to date.

Water agreements have been receiving increasing attention in recent decades and many
international institutions put interest in regulating this sector. For instance, the Institute
of International Law (IIL) in 1911 provides some rules among which was noticed the
prohibition of unilateral and detrimental exploitation of international basins and the
requirement to develop transboundary cooperation. In 1966, while the International
Law Association advocated an equitable utilization of shared watercourses, the United
Nations General Assembly adopted in 1977 the United Nations Convention on the Law of
the Non Navigational Uses of International Watercourses. All these efforts highlight the
growing interest in managing the water resources. But it is essential to notice that the
agreements on water resource have a very heterogeneous content. While some of them
focused on water quality, other try to deal with irrigation, hydropower, infrastructure
and under-groundwater issues. Some are motived by technical/financial cooperation
and/or monitoring provisions. In terms of their objective, there are regulatory treaties
that set rules to prescribe certain actions, and programmatic treaties in respect of a
common program more general treaties that set out principles and norms for
cooperation in transboundary basins.

In this study, the water treaties variable is built using information given by the
International Freshwater Treaty Database provided by the Program in Water Conflict

17
 General introduction

Management and Transformation (College of Earth, Ocean, and Atmospheric Sciences).

This database, which spans the years 1820 to 2007, gathered a full number of
international, freshwater-related agreements between pair or groups of countries. The
data on temperature and precipitations come from Dell et al. (2014) and those on
“Floods, Storms & Droughts" come from the Centre for Research on the Epidemiology of
Disasters. The virtual water associated with exports of agriculture products is built by
combining statistics from Water footprint (hereafter WFP) databases and the bilateral
exports of goods classified as agriculture products in BACI. The Regional Trade
Agreement (RTA) variable comes from Bergstrand's homepage and Data on trade are
from TRADHIST (CEPII) which provides a complete representation of bilateral trade
flows between countries all around the world. Our database finally covers the period
1955-2007 and concerns 154 countries.

After all, we firstly show that while a lot of attention has been focused on water quantity,
a lot of considerations are now accorded to water quality and environment purposes
particularly since the year 1990s. Secondly, it clearly appears that, among climate
change signals, particularly temperature drives the willingness to be engaged in a new
water treaty. In fact, the first section analyses led to the conclusion that countries with
higher temperature are more likely to sign water agreements. Thirdly, using a structural
gravity equation with importer-year, exporter-year effects and bilateral fixed effects, the
study generally reveals that agreements on water are generally a cost to free trade. This
confirms the protectionist role of WTs because we find that these treaties cause a
significant reduction in international trade between countries.

It also appears in the results that water protection are more effective when enforced in
the framework of a water agreement than when included in a regional trade agreement;
in fact, RTAs with water content seems to have no effect on bilateral trade. When we
focus on agricultural exports, findings still reveal that treaties on water resources
reduce agriculture goods exchanges and, without surprise, the strongest impact.

A heterogeneity analysis points out that water treaties reduce trade in Europe, Asia and
America. In contrast, it clearly appears that they have boosted international trade in
Africa. This positive effect in Africa may be explained by the fact that water agreements
have been particularly effective in this continent to foster cooperation between
countries. Regarding the results in the last section, the general equilibrium model makes
18
 General introduction

conclude have a significant effect of water treaties on trade, in particular in Asia. We find
that the enforcement of the treaties has a strong negative impact in Asia. At country
level, countries along the Nile and the Niger are the main winners of freshwater treaties
in Africa.

19
 Chapter 1 - Trade and Institutions: Explaining Urban Giants

Chapter 1 -
Trade and Institutions:
Explaining Urban Giants

20
 Chapter 1 - Trade and Institutions: Explaining Urban Giants

I. Chapter 1 - Trade and Institutions:

Explaining Urban Giants

Introduction

From the 18th century to date, urbanization has been an important process all around
the world and despite the apparent chaotic evolution of cities, much regularities in the
spatial economy have been explained by considering economic factors only. This is
typically the case of the Zipf’s law21, which is usually explained by random productivity
process of local economies and migrations between cities (Eeckhout, 2004). To describe
deviation from this law, it seems however necessary to introduce additional factors. For
instance how to explain that Mexico city is much bigger than just twice the size of the
second largest city in this country; is it the historical centralization of the political
powers in that city or its unique market access to the world? This is the question
addressed in this paper that aims to better understand what explains urban primacy22.
We find that democratic institutions are the main factor behind the concentration of a
nation's urban population in the main city; trade liberalization and good market access
are not significant.
Our research area is related to a vast literature. The agglomeration of the population in
the nation's largest city has been explained by many mechanisms, often related to
political institutions (Lipton, 1977; Bates, 1981). In countries where urbanization goes
hand in hand with industrialization, capital accumulation in the industrial sector can
lead rent seeker' governments to favor fiscal centralization and as a consequence, the
agglomeration of the population in the largest city (Davis and Henderson, 2003). In

21 The Zipf’s law states that the size of the largest cities is inversely proportional to their rank in the size
distribution of cities.
22 Our main question concerns the urban concentration in the biggest city because such a phenomenon is

less understood than other forms of urban change such as urbanization, density or urban growth.
Furthermore as it has been pointed out by Crenshaw and Anthony (2014), primate cities are (more than
other cities) global political theaters and consequently the identification of the causal impact of
globalization and institutions in these cities seems to be the most urgent/logical investigation. Obviously,
additional researches on the impact of these variables on different aspects of the spatial economy, using
different measures of urban concentration (see Anthony, 2014), would bring interesting additional
results.

21
 Chapter 1 - Trade and Institutions: Explaining Urban Giants

states with extensive regulation, where licenses and permits are necessary for a myriad
of activities (to start a business, to hire or to fire workers, to register property and so
on), the decision to be near the administrative center is beneficial since it increases the
probability of getting these authorizations. Such a probability can be a powerful
determinant of location choice for multinational firms which in turn attract domestic
firms and migrants. Furthermore, since bribes, red tape or more generally corruption
enable the obtaining of these rights, the centralization of power in the largest city is a
“cash machine” for governments. Urban favoritism for the largest city may be a strategy
to monopolize (or to polarize) the rent extracted by limiting the competition of
bureaucrats and politicians located in other regions. Fiscal centralization not only pulls
activities into the urban giants, it also pushes activities outside secondary cities. Indeed,
when local governments have little power, this centralization limits the urban planning
de facto and also leads to a kind of resistance against urbanization23. Without the ability
to integrate new migrants into the formal sector, there is a political fear of
agglomerating people. The incidence of crime or revolt is higher in cities, and this may
represent an incentive for weak local governments to discourage urbanization while on
the contrary, strong central government can favor its urban giant24. Ades and Glaeser
(1995) support this thesis and find that dictatorial governments have urban giants that
are 50 percent larger than democratic governments25. This result however, raises many
questions, the most important being: which kind of institutions is the most prone to
urban bias?26 Here we challenge the conclusion of Ades and Glaeser (1995) by working
on countries colonized by European powers during the past 500 years. We focus on
these countries because colonization has deeply modified economies and institutions in

23 For instance, Collier and Venables (2013) report that, in Liberia, governments have adopted a deliberate
policy of reducing infrastructure investment, in order to induce a return of urban residents to the
countryside. Feler and Henderson (2011) also document exclusionary policies in the development of
Brazilian cities, in particular during the period of dictatorship, where localities withheld public services to
deter in-migration. In contrast, the Chinese fiscal decentralization, started in the 1980s, has been a success
into spreading growth by fostering competition between local authorities (Bai, Hsieh and Song, 2019).
24 Beyond the spatial structural advantage in organizing demonstrations in area with a dense population

(in factories thanks to unions or in districts thanks to neighborhood effects and organizations), urban
residents located in the capital city also have an information advantage on public spending due to their
education, access to the media or to the focus of media on their city (Majumdar et al. 2004).
25 A related result has been found by Henderson and Wang (2007) showing that democratization favors

small cities.
26 For instance Acemoglu, Robinson and Verdier (2004) have analyzed kleptocratic governments that

succeeded in staying in power for long periods despite weak support from the population by relying on
divide-and-rule tactics. Do these weak dictatorships also have an urban bias?

22
 Chapter 1 - Trade and Institutions: Explaining Urban Giants

these countries, and while many consequences of this colonization has been analyzed in
depth, the analysis in terms of urban economics is lagging behind.

As illustrated by Figure 3, former European colonies that were relatively densely

populated in the 1500s now have smaller central cities27. By considering, along with
Acemoglu, Johnson and Robinson (2002), hereafter AJR, that density in 1500 was an
indicator of prosperity fostering predatory behaviors and an indicator of the difficulty of
settling, leading colonizers to set up extractive institutions, we come to the logical
conclusion that these institutions have reduced concentration in the central metropolis.

Figure 3: Urbanization Primacy in 2010 and Density in 1500

The main finding of AJR is that economic growth is determined by political institutions
and by nothing else and in particular not by the geography and culture of nations. Are
institutions also the unique determinant of urban primacy? While a vast literature on
that subject explains the different size of cities by different kinds of political institutions,
Krugman and Livas (1996) proved that agglomeration can emerge without political
favoritism28: a bad market access can be a sufficient determinant. Trade restrictions,
that provide advantage to one city which becomes the platform for exports and

27 The second plot presented in Figure 3 for non-European former colonies shows that this reversal is not
widespread; indeed for countries not colonized the relationship is not significant and the historical spatial
economy does not explain the current one.
28 Candau (2008) extends this model to consider how bad governance affects location choices.

23
 Chapter 1 - Trade and Institutions: Explaining Urban Giants

catastrophically attracts all the population29. By contrast, trade liberalization by

providing market access to other cities fosters the dispersion of activities. This
conclusion has been disputed a lot; the link between the spatial economy of countries
and trade integration is not monotonic, depending on the structure of the economy
considered (weak congestion costs, strong share of income spent on industrial goods,
etc), a good market access can also foster agglomeration in the largest city (Krugman,
1991; Behrens and Robert-Nicoud, 2014). Buenos Aires, London, Lagos, New-York, Seoul
are clear examples where good market access to the world has boosted the size of cities.
Faced with this theoretical debate, Ades and Glaeser (1995) found that a one-standard-
deviation increase in the share of trade in GDP reduces the size of the main city by about
13 percent. Their analysis using alternative measures of trade integration (import duties
and transportation spending) confirms this result; however, these variables are plagued
with potential endogeneity bias, and their results using instrumental variables are not
significant which leads them to question reverse causality. Nitsch (2006) using
geographic characteristics as an instrument for trade, is also skeptical about the role of
economic geography on urban concentration30. In contrast using a difference-in-
difference strategy, Karayalcin and Yilmazkuday (2015) show that the Uruguay round of
liberalization has reduced the relative size of mega-cities.
Our work differs from that literature on different aspects. First, since the direct measure
of trade costs (e.g. tariffs used by Ades and Glaeser, 1995) only represents a part of trade
frictions, we use an indirect measure of economic integration based on trade flows.
Regressing bilateral trade flows on distance between partners31 and aggregating these
results using economic geography theory, we compute indicators of market access of
these countries to the rest of the world. To our knowledge, this methodology first
introduced by Redding and Venables (2004) and used in various studies has never been
employed to study urbanization.

29 See also Helpman (1998) and Candau (2011) concerning the determinants (and conditions) of this
massive agglomeration in one city/region. Candau and Dienesch (2015) integrate in this framework,
heterogeneous workers and show that only high-skilled workers increasingly choose the core region
during the process of integration.
30 Furthermore Davis and Henderson (2003) in their detailed study of urbanization also confess:

“We experimented with a time varying measure, trade openness. This was statistically
insignificant in the IV estimation. Even the sign was sensitive to which instruments
were included. Thus we did not include the variable in the results presented in the
tables”
31 In contrast with the literature in international economics that use geographical distance weighted by

population (or simply geographical distance) in their trade gravity equation, we compute a bilateral
distance between partners that takes into account internal distance between cities in each country.

24
 Chapter 1 - Trade and Institutions: Explaining Urban Giants

Secondly, since simple indicators of openness (e.g. exports as a percentage of GDP as

used by Barrios et al.) introduce endogenous bias, we propose different instruments for
market access such as the genetic distance between partners. The main idea of this
instrumentation is that the genetic distance is an exogenous indicator of past
international migration and of international trade roads that can explain the current
trading system. We also use many controls not available until recently, such as the
differential between the intensity of night lights in the city and the intensity of lights in
other cities obtained from satellite pictures within a country that allows us to capture
spatial income differences.
Third, our results differ from the literature. We show that the rise and fall of mega-cities
are mainly explained by institutions. The basic explanation follows the AJR thesis
applied to urban economics. European colonialism led to the development of inclusive
institutions in previously poor areas which has favored agglomeration economies in
urban giants. In contrast, introducing extractive institutions in previously prosperous
areas have led to the dispersion of the population.
Section 2 presents our baseline empirical strategy, data and the first results. Section 3
presents an IV strategy and our results; Section 5 describes the conclusions of the study.

I.1 Hypothesis, data and preliminary results

I.1.1 Empirical Strategies

The aim of this paper is to study the causal impact of political institutions and
international trade on the concentration of people in the biggest cities. However, before
to enter into the details of these two determinants, let's first analyze the impact of
institutions in general, considering all the formal and informal constraints and rules (the
different codes of conduct, customs, laws, governance) that define political institutions
and also economics interactions necessary to trade toward long distance) that causes
the concentration of a nation's urban population in its first city. The most simple

25
 Chapter 1 - Trade and Institutions: Explaining Urban Giants

equation allowing to describe key identification issues of this question may be the
following32:

𝑢𝑖𝑡 = 𝛼𝐼𝑖𝑡 + 𝑓𝑖 + 𝑓𝑡 + 𝜀𝑖𝑡 (1)

Where 𝑢𝑖𝑡 is the population in the biggest metropolis on the total population33, 𝑓𝑡 are
time fixed effects to control for shocks affecting all individuals at the same time, 𝑓𝑖 are
country fixed effects aimed at capturing characteristics of countries and 𝐼𝑖𝑡 the variable
of institutions of the country 𝑖 at the period 𝑡.
The main parameter of interest is α, the elasticity of agglomeration with respect to
institutions. An unbiased OLS estimate for this parameter is obtained under the
following assumption:

𝐶𝑜𝑣(𝐼𝑖,𝑡 , 𝑓𝑖 + 𝑓𝑡 + 𝜀𝑖𝑡 ) = 𝐶𝑜𝑣(𝐼𝑖,𝑡 , 𝑓𝑖 ) + 𝐶𝑜𝑣(𝐼𝑖,𝑡 , 𝑓𝑡 ) + 𝐶𝑜𝑣(𝐼𝑖,𝑡 , 𝜀𝑖𝑡 ) = 0 (2)

As it is well known when the covariance in the left hand side of this equation differs
from zero, any of the three covariances on the right hand side is a source of bias. We
discuss these three terms in turn in order to build our empirical strategy.

Concerning the first term, 𝐶𝑜𝑣(𝐼𝑖,𝑡 , 𝑓𝑖 ), there are some reasons to consider that time
varying institutions depend on invariant country characteristics. For instance natural
resources or the level of development may have an impact on institutions. A related
concern is the fact that the degree of urban primacy may influence institutions.
Agglomeration economies may be higher in countries with urban giants and incomes
generated from spillover and/or pecuniary externalities can influence institutional
change. This relationship is even more obvious concerning international trade, higher
level of productivity in a particular sector (e.g. in finance) due to a concentration of
activities in one place may influence institutions, at least the formal institutions, of a
country. To analyze this bias, consider the following reversed causality between
institutions and concentration:

32We follow here a presentation proposed by Combes, Duranton and Gobillon (2011).
33This share has the advantage to better take into account the size and change of the biggest cities in
countries with rural economies (where the central city is not fueled solely by the flow of urban to urban
migration but also by the rural one) than classical measures such as the size of the biggest city on the
urban population which is moreover dependent on administrative definition of the urban population.

26
 Chapter 1 - Trade and Institutions: Explaining Urban Giants

𝐼𝑖𝑡 = 𝛽𝑢𝑖𝑡 + 𝜂𝑖 (3)

From classic sociological theory to political modernization theory (Weber, 1909; Lerner,
1964; Lipset, 1959), there is a vast literature that defends this kind of relationship
considering that the urban structure affects political mobilization and political change
(see Anthony (2014) for a survey).
Using Equation (1) without time effects to simplify, yields:

𝛽𝑓𝑖 + 𝜂𝑖 (4)
𝐼𝑖𝑡 =
1 − 𝛽𝛼

which allow to rewrite the previous covariance as follows:

(1 − 𝛽𝛼)𝐶𝑜𝑣(𝐼𝑖,𝑡 , 𝑓𝑖 ) = 𝛽𝑉𝑎𝑟(𝑓𝑖 ) + 𝐶𝑜𝑣(𝜂𝑖 , 𝑓𝑖 ) (5)

Hence the first origin of the bias comes from the variance of individual specific
characteristics 𝑉𝑎𝑟(𝑓𝑖 ) magnified by the elasticity of institutions with respect to the
concentration of people 𝛽. The second channel of an upward bias is the correlation
between individual characteristics that explain institutions and urban primacy, one can
think respectively to culture and productivity, which lead to conclude that such a bias is
likely34.

The other origins of bias in (2) is simple to understand, the covariance between
institutions and temporal chocks (e.g. a global financial crisis), 𝐶𝑜𝑣(𝐼𝑖,𝑡 , 𝑓𝑖 ) may differs
from zero. Finally a covariance between institutions and the error term is the usual
definition of endogeneity.

To reduce each of these problems, we first look for a lagged measure of institutions or
more precisely a measure of past institutional factors that explain the sustainability of
urban agglomeration before the industrial revolution. We choose the density in 1500 to
approximate political institutions around that period35. Density in 1500 is a good proxy
of the current institutions according to AJR, who show that countries that were
relatively rich in 1500 (density is considered as a proxy of economic prosperity) have

34 See for instance Mokyr (2016) who describe how “cultural entrepreneurs”, such as Newton, Bacon and
many others in England and France, have influenced innovations and growth after 1600.
35 Available on request, we also use urbanization in 1500 which gives the same result.

27
 Chapter 1 - Trade and Institutions: Explaining Urban Giants

fostered predatory behaviors and extractive institutions by colonizers. We denote this

variable 𝐼𝑖,1500 . We also use a proxy for the current state of political institutions by using
the binary view of Acemoglu, Naidu, Restrepo and Robinson (2017) who classify a
country as democratic when different databases of political institutions (Freedom
House, Polity IV) conclude as such, and dictatorial in the reverse case. This variable is
denoted 𝐼𝑖𝑡 .

To take into account institutions that plays a major role in explaining international
trade, we regress bilateral trade flows on distance, GDPs and fixed effects and we use the
estimates of fixed effects (which typically capture unobservable institutions that explain
trade) and distance to build a indicator of market access (also called market potential),
here after denoted Ω𝑖𝑡 . To reduce the bias of omitted variable, several variables of
control are also introduced among which a proxy of the income gap between the biggest
city and the rest of the country. While it seems trivial to say that people move from one
place to another according to income difference, this variable has never been used to
analyze urban primacy until now mainly due to the lack of data at the regional level in
developing countries. To approximate this spatial income gap in each country we use the
differential between the intensity of night lights in the biggest city and the intensity of
lights in the rest of the country from satellite pictures36. We also add classical controls
(mainly climate data) for characteristics that are specific to country and time. All these
controls are represented in the following equations by a vector 𝐶.

We start our analysis with a simple cross-sectional analysis (for the year 2010) running
ordinary least squares on the following equation:

𝑢𝑖 = 𝛼𝐼𝑖,𝑡−𝑥 + 𝛿Ω𝒊 + 𝛾𝐶𝑖 + 𝜀𝑖 (6)

where 𝐼𝑖,𝑡−𝑥 represent past institutions when the indicator of density in 1500 is used
(𝑡 − 𝑥 = 1500) or the current level when we use binary variable of democracy
(𝑡 − 𝑥 = 0) Many papers on the impact of institutions on urban primacy are based on
this type of cross-sectional analysis. This naive regression presents all the
aforementioned bias (and many other), the problem of endogeneity due omitting

36 Henderson, Storeygard and Weil (2012) have shown that data from outer space are a good proxy of
income.

28
 Chapter 1 - Trade and Institutions: Explaining Urban Giants

variables is particularly serious. However, the advantage of this estimation is that it

allows to include the invariant historical measure of institutions, 𝐼𝑖,1500 . This variable
which is also a measure of the urban primacy in 1500, helps to understand the evolution
of the spatial economy of nations over a long period of time. The coefficient 𝛼, if
negative, indicates a dispersion of the population from the biggest city, i.e. a reversal in
the attractiveness of the biggest city.

We then rely on the panel structure of the data to tackle endogeneity issues (due to
omitting variables) by estimating:

𝑢𝑖𝑡 = 𝛼𝐼𝑖,𝑡−𝑥 + 𝛿Ω𝒊𝒕 + 𝛾𝐶𝑖𝑡 + 𝑓𝑖 + 𝑓𝑡 + 𝜀𝑖𝑡 (7)

This equation is analyzed through the lens of the Fixed-Effect (FE) estimator over the
period 1992-2010. This fixed effects estimator is useful to reduce the endogeneity bias
of omitted variables. However, the FE procedure is useless to obtain consistent
estimates of the time-invariant proxy of institution, 𝐼𝑖,1500 , since there is a perfect
colinearity between fixed effects and institutions in 1500. As a result, the FE model
constraint us to use exclusively time varying institutions, 𝐼𝑖𝑡 , which potentially introduce
endogeneity from reversed causality.

To carry out inference on historical density, we thus use the Fixed Effects Filtered (FEF)
estimator of Pesaran and Zhou (2016). This estimation is in two steps, first a FE
estimation is used to obtain coefficients of the time-varying variables and these
estimates are used to filter out the time-varying effects. Then, the residuals from the first
step are averaged over time and computed as a dependent variable in a cross-section
OLS estimation that includes historical density and trade network.
To obtain unbiased estimate of the elasticity of agglomeration with respect to the
current institutions due to reverse causality, we propose two different Instrumental
Variables (IV) estimations.

I.1.2 Data

 Cities

29
 Chapter 1 - Trade and Institutions: Explaining Urban Giants

Urban concentration over the period 1962-2010 is calculated by dividing the population
in the biggest city by the total population using data from the United Nations, World
Urbanization Prospects. Data on density in 1500, 𝑢𝑖,1500, comes from the Clio-infra
database on urban settlement sizes37. Data on urbanization in 1500 comes from AJR.

To approximate the spatial urban-income gap, ∆𝑖𝑡 , we use the differential between the
intensity of night lights in the biggest city and the intensity of lights in other cities
obtained from satellite pictures within a country. More precisely, we draw the spatial
income gap between the biggest city and the remaining area of every country by using
geomatics’ tools on Night Lights Data provided by the National Oceanic and Atmospheric
Administration’s (NOAA) from 1992 to 2010. NOAA processes the raw data collected by
the United States Air Force Defense Meteorological Satellite Program (DMSP). The
approach used here aims to compute for each country the mean of night lights intensity
per square kilometer within areas where the largest city is located. For this purpose, we
use Global Administrative Areas (GADM) shapefiles which provide an informative set of
countries’ administrative boundaries at national and sub-national level. We then cross
them with the Natural Earth populated places database which allows keeping the most
populated town location in every country. Night Lights’ rasters provided by NOAA
contain pixels to which are attributed digital numbers ranging from 0 to 63, reflecting
the brightness of the light. We used an average value for each pixel for year in which two
satellites collected the data. We compute, by country and year, the sum of pixels’ values
of the biggest city and the sum of pixels for the rest of the country (to approximate light
intensity outside the mega-city), we divide these numbers by the geographical surface.
The income gap is approximated by the difference between these two values. A lag of ten
year is used to avoid reverse causality38. These data limit the period analyzed to 1992-
2010.

 Trade
The indicator of market access is directly built from the economic geography theory, we
modify the market access of Redding and Venables (2004) to take into account the
specificity of our research. The first step is to estimate the following trade gravity

37 http://www.cgeh.nl/urbanisation-hub-clio-infra-database-urban-settlement-sizes-1500-2000
38 Robustness tests with different lags are presented in Appendix I.4.1.

30
 Chapter 1 - Trade and Institutions: Explaining Urban Giants

equation on bilateral exports, coming from COMTRADE, with individual fixed effects
(𝑓𝑖𝑥 , 𝑓𝑖𝑚 ) and distance using a pseudo-maximum likelihood (PPML) estimator39:

𝑋𝑖𝑗 = −𝜀ln𝜏𝑖𝑗 + 𝑎𝑖 𝑓𝑖𝑥 + 𝑎𝑗 𝑓𝑗𝑚 + 𝑣𝑖𝑗 (8)

where bilateral trade costs 𝜏𝑖𝑗 take the following common specification:

𝜏𝑖𝑗 = 𝑑𝑖𝑠𝑡𝑖𝑗 exp(𝛿1 𝑏𝑖𝑗 + 𝛿2 𝑐𝑖𝑗 + 𝛿3 𝑙𝑖𝑗 )

With 𝑏𝑖𝑗 , 𝑐𝑖𝑗 , 𝑙𝑖𝑗 are dummies characterizing bilateral trade barriers, respectively a
border, past colony links and a common language, 𝑑𝑖𝑠𝑡𝑖𝑗 is a weighted measure of
distance between cities. The literature usually uses two kind of distance, the simple
distances between capitals and a distances between capitals weighted by the population
of each countries. Since simple distances between capitals poorly represent the
geography of trade costs (in particular for large countries) and introduce bias in the
estimation (e.g. over-estimation of the border effect), many articles routinely use the
weighted measure proposed by Head and Mayer (2002) and freely available from the
GeoDist database of the CEPII. This distance between country 𝑖 and 𝑗 is calculated as
follows:

𝑑𝑖𝑠𝑡𝑖𝑗 = ∑ (𝑝𝑜𝑝𝑘 /𝑝𝑜𝑝𝑖 ) ∑ (𝑝𝑜𝑝𝑙 /𝑝𝑜𝑝𝑗 ) 𝑑𝑘𝑙

𝑘 𝑙

where 𝑝𝑜𝑝𝑘 and 𝑝𝑜𝑝𝑙 are the population of the largest cities 𝑘 and 𝑙. In brief, distances
are weighted by the share of the biggest city in the total population which is also our
dependent variable. Thus it seems obvious that we cannot use such a measure to explain
the concentration of the population in the nation's largest city without introducing
endogeneity bias. We thus deviate from the literature by building our own measure
using a weighting scheme based on internal distance between the capital and its cities:

𝑑𝑖𝑗
𝑑𝑖𝑠𝑡𝑖𝑗 = 𝑤𝑖𝑡ℎ 𝑑𝑖 = ∑ 𝑑𝑖𝑠𝑡𝑖𝑘 , 𝑑𝑗 = ∑ 𝑑𝑖𝑠𝑡𝑗𝑙
𝑑𝑖 𝑑𝑗 𝑘 𝑙

39Following the work of Silva and Tenreyro (2006), the use of pseudo-maximum likelihood estimators
(Poisson and derived econometric models) is justified for treating heteroskedasticity and dealing with the
presence of zero trade values.

31
 Chapter 1 - Trade and Institutions: Explaining Urban Giants

Where 𝑑𝑖𝑗 is the distance in km between the biggest cities 𝑖 and 𝑗 of each partners, and
𝑑𝑖 is the internal distance i.e. the sum of all distance between the biggest city 𝑖 and all
other cities 𝑘 belonging to the same country.

Lastly an important part of the gravity equation making the link between theory and
empiric are fixed effects 𝑓𝑖𝑥 and 𝑓𝑗𝑥 (and their symmetric expressions) that allow to
obtain the predicted value of supply and market capacity40. Using these predicted
values, with our weighted measure of bilateral distance, and estimates of the trade
elasticity 𝜀 give the market access:

Ω𝑖 = (exp(𝑓𝑖𝑚 ))𝛼̂𝑖 + ∑
̂𝑗
𝛼 𝛿̂ 𝛿̂ 𝛿̂
[(exp(𝑓𝑗𝑚 )) 𝑑𝑖𝑠𝑡𝑖𝑗 −𝜀̂ 𝑏𝑖𝑗1 𝑐𝑖𝑗2 𝑙𝑖𝑗3 ] (9)
𝑗≠𝑖

The gravity equation is estimated in cross-sections to obtain a market access that varies
over time, Ω𝑖𝑡 .
To get rid of multicollinearity problems in estimating (7) and also of the fact that the
market potential can be endogenous to the size of mega-cities, we also use an historical
measure of the market access not linked to trade flow but to trade relationship: we
calculate an indicator of networks, the so-called out-degree of trade, measuring the
number of arcs pointing to partners on the total number of nodes in the bilateral trade
network of countries in 190041.

 Institutions

Concerning institutions, one of the most recent dichotomous measure of democracy

proposed by Acemoglu, Naidu, Restrepo and Robinson (2017), hereafter ANRR is
chosen. This index, running from 1960 to 2010, combines information from several
sources including Freedom House and Polity IV databases. ANRR classify a country as
democratic when several databases conclude as such. We also use the instrument of
ANRR, who exploit regional waves of democratization and transitions to non-democracy
as a source of exogenous variation in democracy. More precisely ANRR divide the world

40 Individual fixed effects also captures internal trade costs that are specific to each country. Thus at the
difference of Redding and Venables (2004) we decide to not include a measure of internal distance beside
these fixed effects to avoid double counting (Redding and Venables (2004) makes the critical assumption
that every country is a circle which lead them to consider the following formula for internal distance:
̂ ⁄2 𝑎𝑟𝑒𝑎
𝜏 −𝜀 = 𝑑𝑖𝑠𝑡 −𝜀 with 𝑑𝑖𝑠𝑡 = 0.66( ); where 𝑎𝑟𝑒𝑎 is the surface of the country in km²).
𝜋
41 See Appendix I.4.2 which describes in details how this indicator has been built.

32
 Chapter 1 - Trade and Institutions: Explaining Urban Giants

in seven regions, 𝑅𝑖 (where 𝑖 is a country in Africa, East Asia and the Pacific, Eastern
Europe and Central Asia, Western Europe and other developed countries, Latin America
and the Caribbean, the Middle East and the North of Africa, and South Asia) and consider
a dummy 𝐷𝑖𝑡0 taking 1 when the country 𝑖 is a democracy at the start of the period.
Finally democracy in country 𝑖 is influenced by democracy in the set of countries
𝐽𝑖 = {𝑖 ′ : 𝑖 ′ ≠ 𝑖, 𝑅𝑖 ′ = 𝑅𝑖 , 𝐷𝑖 ′ 𝑡0 = 𝐷𝑖𝑡0 } which gives the following instrument of political
governance:

1 (10)
𝐺𝑖𝑡 = ∑ 𝐺𝑖 ′ 𝑡
|𝐽𝑖 | ′
𝑖 ∈𝐽𝑖

In the words of ANRR, 𝐺𝑖𝑡 is the jack-knifed average of democracy in a region × initial
regime cell, i.e. a distance-weighted average of democracy among “neighbors”. In order
to avoid confusion, it is noteworthy that the “geographical components” of this
instrument are radically different from the ones included in the market access, the
geographical distance is not taken into account and the neighborhood has a wide
definition (e.g. every country in Africa).
For robustness checks regarding the variable of democracy, we also use the polity2
variable, from the Polity-IV project and the Freedom House indices of political rights
(PR) and civil liberty (CL).
The mortality rates of soldiers, laborers, and bishops during colonial times, used to
instrument institutions, come from Albouy (2012).

 Other

Climate data, such as annual precipitation and temperature comes from the CERDI's
database. Genetic distance data are used to instrument the market access. These data,
based on parts of the DNA that varies through random mutation and drift, comes from
Spolaore and Wacziarg (2018) who use information on human micro-satellite variation
(Pemberton et al., 2013) at the population (not country) level which are then match to
countries using ethnic composition data. More precisely, we use the ancestral genetic
distance of Spolaore and Wacziarg (2018) who match to populations as they were in
1500 AD.

33
 Chapter 1 - Trade and Institutions: Explaining Urban Giants

I.1.3 First results

Table I-1, Column (1) and Column (2) examines the cross-sectional patterns running
ordinary least squares specification as described in Equation (6) to illustrate the raw
data patterns by regressing the log share of urban primacy in 2010 on two different
indicators of democracy. In Column 1, based on AJR, this indicator is the log of density in
1500. The bias of reversed causality is reduced by using this historical variable, but this
advantage comes at a cost: the effects of the current political institutions cannot be
analyzed. In Column 2, we use the indicator of democracy of ANRR. These two columns
make our results comparable with the literature on urban primacy that often use cross-
sectional data, but as discussed at length in Section 1, this specification may provide
biased estimates of the variables of interest. Then, in Column (3), (4), (5) and (6), we
exploit the panel structure of the data to tackle endogeneity issues by estimating
Equation (7) using the Fixed Effects (FE) and the Fixed Effects Filtered (FEF) models.

In Column 1, the elasticity of urban primacy to past density is significant, high and
negative, revealing a reversion in the size of the urban concentration relatively to its
total population. On average a country that was twice more densely populated in 1500 is
now a country where the share of its biggest city in the total population is 23% smaller
than in other countries. We argue that the level of density in 1500 being a proxy of bad
governance in the following centuries (countries densely populated in 1500 were richer,
which was an incentive to set extractive institutions42) represents a repulsive force for
the native population. This mechanism is consistent with the finding of Nunn and Puga
(2012) that rugged terrains were chosen in Africa because these places afforded
protection to those being raided during the slave trade. The other side of this result is
that countries with a weak concentration of the population in the nation's largest city in
1500 have benefited of inclusive institutions, such as the development of institutions of
private property, that made the growth of urban giants profitable by developing
agglomeration economies. As in the cross-section analysis, the impact of past density is

42 Incentive to set extractive institutions may have also been magnified by the fact that high density
countries were also countries were it was harder to settle for colonizers in reason of opposition and/or in
reason of disease, as a result the colonial rulers may have concentrated their capital in one unique
location, a port for instance, in order to extract rent at the minimal cost.

34
 Chapter 1 - Trade and Institutions: Explaining Urban Giants

still negative and significant (Column, 4-5) using the Fixed Effects Filtered (FEF)
estimator of Pesaran and Zhou (2017). Finally in Column 2, 3 and 6, the indicator of
democracy is always significant and influences positively the population growth in the
biggest city. All these results go against the finding of Ades and Glaeser (1995) who find
that bad governance fosters urban concentration at the top of the urban hierarchy.

In Column (1), the past network of trade is also significant and negative. This negative
impact of the past network of international trade may be related to the particular
specialization of some countries, the first globalization has favored land tenets in Latin
America for instance (Findlay and O'Rourke, 2009) and thus urbanization was possibly
not profitable for this elite. As a result this group may have put in place policies against
cities and in particular against large cities that were a fertile land for industrialization.
The introduction of fixed effects (Column 5) confirms this result, since the coefficient of
the past network, albeit smaller, is still significant. However the introduction of an
indicator of the current democracy raises some doubt about the validity of this variable
of past network which is no longer significant (Column 6) and thus no longer used in the
rest of the paper. Indeed this indicator of past network, reflect more past political
institutions and networks between kings than between traders.

Columns 3 and 4 present the positive impact of the market access. This result supports
the thesis that a decrease in trade costs, at least in the first step of development,
exacerbates regional disparities (Krugman, 1991). Activities and people tend to
concentrate where there is a large market and the market is large where activities are
concentrated; trade openness fosters this circular causality of agglomeration in the
biggest city.
Regarding controls, the income gap approximated by the difference in night lights
between the biggest cities and other cities has the expected positive impact.
The negative impact of temperature in the cross-section analysis (Columns 1-2) may be
driven in some countries by the indirect effect of new technologies such as air-
conditioning or large investment infrastructures (e.g. dam) that have improved
conditions of life in many cities that were initially inhospitable. Once individual fixed
effects are introduced, capturing the heterogeneity of countries on these aspects,
temperatures are no longer significant (Columns 3-6).

35
 Chapter 1 - Trade and Institutions: Explaining Urban Giants

Table I-1: Urban Reversal

OLS OLS FE FEF FEF FEF

Past Institution (Density in 1500) -0.234 -0.226 -0.252
(0.060)a (0.007)a (0.007)a
Democracy 0.705 0.036 0.076
(0.319)b (0.025)b (0.028)a
Past Trade Network (Outdegree in 1900) -0.616 -0.613 -0.418 -0.208
(0.134)a (0.236)b (0.110)a (0.185)
Market Access 0.067 0.026
(0.028)b (0.010)b
Income (Night Lights Data) -0.003 -0.019 0.065 0.062 0.058 0.094
(0.051) (0.082) (0.001)a (0.028)b (0.029)a (0.031)a
Landlock 0.057 0.044 0.223 0.206 0.567
0.310 0.446 0.186 0.123 (0.146)a
Temperature -0.818 -1.132 0.026 0.002 -0.055 -0.055
(0.380)a (0.449)a (0.821) (0.134) (0.135) (0.135)
Humidity -0.175 -0.182 -0.006 -0.005 -0.001 -0.001
(0.128) (0.118) (0.593) (0.011) (0.011) (0.011)
Constant 0.216 0.662 -2.441
(1.223) (1.174) (0.374)
R-square 0.549 0.447 0.999
Observations 30 30 315 322 277 270
Note: OLS estimator in Column 1-2, Fixed effects (FE) in Column 3, Fixed Effects Filtered (FEF) in Column 4-6.
All variables are in Log. (a), (b), (c) denote significance at the 1, 5 and 10 percent level respectively. Data on
urban primacy are from the World Development Indicators. Data on density in 1500 from the Clio-infra
database on urban settlement sizes, data on urbanization in 1500 from Acemoglu Johnson and Robinson
(2002). The binary variable of democracy comes from Acemoglu, Naidu, Restrepo and Robinson (2017). The
income gap between the biggest city and other cities is approximated by the differential between the intensity
of night lights in the biggest city and the intensity of lights in other cities obtained from satellite pictures. The
market access is computed from the estimation of a gravity equation using bilateral exportation from
COMTRADE, distance and geographical variables from the CEPII. Fixed Effect estimations include a full set of
year and country fixed effects.

I.2 IV Strategy

The main drawback of the previous analysis concerns the reverse causality of trade and
political institutions, indeed, as discussed in Section 2, political institutions may be
influenced by the spatial economy and furthermore international trade is often
dependent of the performance of the biggest city. To take into account these links, we
propose a two-stage least squares (2SLS) estimation strategy, with a “zero-stage” to find
suitable instruments for these institutions in the first stage. This 2SLS identification
strategy with a zero-stage is inspired by Docquier et al. (2016).

36
 Chapter 1 - Trade and Institutions: Explaining Urban Giants

I.2.1 Two steps SLS identification strategy with a zero-stage

The zero-stage is based on the finding that the mortality rate faced by settlers is a good
instrument of the current institution (AJR). Since this proposition has stirred up
questions (Albouy, 2012)43, we also take a more contemporaneous variable by using the
regional waves of democratization and transitions to non-democracy as an instrument
(this is the instrument used by ANRR, already described in the previous section).

Denoting 𝐻𝑖 the historical mortality rates of soldiers, laborers, and bishops during
colonial times (source: Albouy, 2012), 𝐼𝑖 the indicator of democracy (source: ANRR), and
𝐺𝑖 the measure of political governance based on waves of democratization (defined by
Equation (10)), we estimate the following equation:

𝐼𝑖 = 𝛽𝐻𝑖 + 𝛾𝐺𝑖 + 𝑣𝑖 (11)

This equation is simply an extension of AJR using a different dependent variable

(democracy instead of expropriation risks), and an additional variable 𝐺𝑖 . Finally while
AJR estimates this equation for one year only, we regress it on each year of our sample
̂𝑔 that varies over time44.
which gives an historical-geographical instrument, denoted 𝐻𝑖𝑡

Democracy effect identification will come from this time-varying predicted instrument
̂𝑔 that captures the discretionary variations in institutions. It reflects the exogenous
𝐻𝑖𝑡

component of democracy variable as a consequence of distant past shocks in the

contemporaneous institutions trend.

Concerning the endogeneity of trade, we follow the same strategy by using the ancestral
genetic distance (Spolaore and Wacziard, 2018). The relevance condition asserting that
this variable is a good instrument to predict the endogenous dependent variable, i.e.
trade, is well established. For instance, Guiso, Sapienza and Zingales (2009) find that
genetic distance is a good proxy for cultural proximity, which impacts on the trust
between two partners and then on trade exchanges. Spolaore and Wacziard (2018)
present clear evidences that genetic distance acts as a temporary barrier to the diffusion
of innovations and development. Giuliano, Spilimbergo and Tonon (2014) find that

43See however the response of AJR (2012).

44 ̂𝑔 ≡ 𝛽̂ 𝐻 + 𝛾̂𝐺 for each year
In other terms, from the estimation of (11) we compute a different 𝐻𝑖 𝑖 𝑖
between 1962 and 2010.

37
 Chapter 1 - Trade and Institutions: Explaining Urban Giants

genetic distance is a good proxy for transportation costs between countries in reason of
international migrations. The variations in genetic distance are partially explained by
international migrations which partially cause international trade via business and
social networks. See for instance Rauch and Trindade (2002) on the impact of the ethnic
networks on trade. The market access is then regressed on ancestral genetic distance,
𝑔
denoted 𝐷𝑖 , for each year of our sample, which enables to get an instrument of the
̂𝑔 .
market access based on genetic distance that varies over time, hereafter denoted 𝐷𝑖𝑡

The first step explains the variation of institutions by using the exogenous variable of
institutions defined in the previous step:

̂𝑔 + 𝑓 + 𝑓 + 𝐶 + 𝜒
𝐼𝑖𝑡 = 𝜍𝐻 𝑖 𝑡 𝑖𝑡 𝒊𝒕
(12)
𝑖𝑡

and in a similar way the variation over time of the market access:

̂𝑔 + 𝑓 + 𝑓 + 𝐶 + 𝑣
Ω𝑖𝑡 = 𝜚𝐷 𝑖 𝑡 𝑖𝑡 𝒊𝒕
(13)
𝑖𝑡

̂𝑔 which, by
The interest of this first step is that 𝜍 is estimated from within changes in 𝐻𝑖𝑡

depending on the distance-weighted average of democracy among neighbors, is an

exogenous variable of the current urban primacy. In a similar way 𝜚 is not estimated by
cross-country difference but by within changes of a genetic distance between partners
which is certainly not a function of trade. Individual fixed effects in the two equations
are particularly important to control for differences in the level of development. The
term 𝐶𝑖𝑡 represents three variables of control that are used in the two stages: the
precipitation, the temperature and the gap in the intensity of night lights.

Finally the second step uses the predicted values of the “historical-geographical”
institutions and of the predicted value of “genetic” market access as follows:

̂ + 𝜚̂𝐷
𝑢𝑖𝑡 = 𝜍̂𝐻
𝑔 ̂ +𝑓 +𝑓 +𝐶 +𝜖
𝑖
𝑔
𝑡 𝑖𝑡 𝑖𝑡
(14)
𝑖𝑡 𝑖𝑡

38
 Chapter 1 - Trade and Institutions: Explaining Urban Giants

I.2.2 Results of the Zero-Stage

Table I-2 confirms the positive impact of waves of democratization on institutions.

Results are less clear concerning the effect of the mortality rate of colonizers.
Coefficients associated to the colonizers mortality rate are significant only between
1962 and 1972 which may suggest an erosion of the bad influence of past colonial
institutions. However, additional estimations provided in the online appendix shows
that the initial result of AJR is verified over the whole period when we use a different
indicator of democracy, i.e. the mortality rate of colonizers explains institutions as
measured by the polity 2 index of democracy, the Freedom House indices of political
rights (PR) and civil liberty (CL). It is also noteworthy that this indicator of mortality has
a small impact on institution in comparison with waves of democratization45.

Table I-2: Democracy (ANRR index), zero stage

1962 1965 1972 1982 1992 2002 2010
Waves of democratization 0.957 0.961 0.696 0.810 0.894 0.929 0.979
𝐺𝑖 (0.076)a (0.076)a (0.216)a (0.161)a (0.153)a (0.198)a (0.149)a
Mortality rate of colonizers -0.030 -0.042 -0.165 -0.063 -0.076 0.028 0.010
𝐻𝑖 (0.017)c (0.017)b (0.096)c (0.080) (0.087) (0.066) (0.067)
Constant 0.147 0.206 0.783 0.338 0.383 -0.038 -0.016
(0.127) (0.123) (0.470) (0.362) (0.425) (0.328) (0.296)
R-squared 0.773 0.776 0.515 0.356 0.427 0.558 0.538
Observations 35 35 36 36 36 36 36
Note: OLS estimator. Mortality rate variable is in Log. (a), (b), (c) denote significance at the 1, 5 and 10 percent
level respectively. The mortality rate of colonizers (in log) comes from Albouy (2012). Data on democracy are
from Acemoglu, Naidu, Restrepo, Robinson (2017), denoted ANRR in the title of this table, and instrumented
with their instrument based on waves of democratization.

In Table I-3 we present some of the zero-stage of the market access. Throughout the
period we find a negative impact of the genetic distance on the market access.

45Not reported here we also compute our instrument by using only the predicted value of the indicator of
governance based on the waves of democratization (i.e. by excluding the mortality rate such as 𝐻̂𝑔 ≡ 𝛾̂𝐺 )
𝑖 𝑖
and we find similar results in the following stages. The small estimates of 𝐻𝑖 (see Table I-2) may explain
why our results hold with and without this variable.

39
 Chapter 1 - Trade and Institutions: Explaining Urban Giants

Table I-3: Market Access, zero stage

1962 1972 1982 1992 2002 2010
Genetic Distance -0.421 -0.360 -0.550 -0.393 -0.345 -0.364
(0.108)a (0.159)b (0.187)b (0.192)b (0.188)b (0.105)b
Constant 0.820 -0.115 0.328 0.813 1.602 1.422
(0.321)b (0.481) (0.565) (0.587) (0.585)a (0.313)a
R-square 0.185 0.066 0.095 0.047 0.037 0.151
Observations 37 38 38 38 38 37
Note: OLS estimator. All variables are in Log. (a), (b), (c) denote significance at the 1, 5 and
10 percent level respectively. Data on genetic distance in 1500 are from Spolaore and
Wacziarg (2018). The market access is computed from the estimation of a gravity equation
using bilateral exportation from COMTRADE, distance and geographical variables from the
CEPII.

I.2.3 Results of the first stage

Table I-4 (Column 1) presents the first stage described by Equation (12). As expected
our historical and geographical instrument based on mortality rates and on waves of
globalization significantly explains the level of democracy. This first stage confirms the
different results obtained by AJR and ANRR. The coefficients on the instrument variables
are highly significant. We present the F-test statistics that reject the joint null effect
hypothesis of these IVs and they are above 10, the threshold proposed by Staiger and
Sotck (1997). As a robustness check we also run an estimation by excluding the
mortality rate in the zero-stage, and thus by using an instrument in this first step only
based on the waves of democratization. We do not report this estimation since results
are almost identical (we find a coefficient of 0.871 with a RSE of 0.052).
Column (3) presents the first stage regressing the market access on its instrument with
various controls (see Equation (13)). As required, the instrument of trade openness
based on genetic distance significantly explains the market access. This result confirms,
in a different way, the main finding of Giuliano, Spilimbergo and Tonon (2014) that
genetic distance captures transportation costs between countries.
Column (2) and (4) propose a simple test to verify that there is no other problems of
endogeneity in this strategy by analyzing whether the market access explains the
current level of democracy. Indeed the causal link of globalization on democracy has be
discussed a lot by economists and political scientists (e.g. Milner and Mukherjee, 2009)
and not directly addressed in this paper until now. Our result confirms the finding of this

40
 Chapter 1 - Trade and Institutions: Explaining Urban Giants

literature that trade openness does not explain the process of democratization (when
adequate IV are used).
Column (3) shows that the instrument of the market access has no impact on
institutions, while Column (4) shows that the instrument of institutions does not impact
on the market access.

Table I-4: Democracy and Trade, First stage

Democracy ANRR Polity2
Market Acces Eq Market Acces Eq
Democracy Eq (12) Democracy Eq (12)
(13) (13)
Democracy (𝐻̂𝑖𝑡𝑔 ) 0.944 0.918 0.021 0.864 0.843 0.001
Instru: Mortality+Waves (0.038)a (0.042)a (0.017) (0.037)a (0.040)a (0.001)
Market Access (𝐷̂𝑖𝑡𝑔 ) -0.352 -1.005 0.993 -7.461 -1.005 -1.003
Instru: Genetic dist (0.328) (0.093)a (0.085)a (4.623) (0.093)a (0.085)a
Temperature -0.646 -0.847 0.622 0.457 -7.081 -7.148 0.622 0.440
(0.318)b (0.341)b (0.196) (0.191)b (4.474) (4.806) (0.196) (0.190)b
Precipitation -0.006 -0.002 -0.006 -0.009 -0.202 -0.086 -0.006 -0.009
(0.042) (0.045) (0.017) (0.012) (0.571) (0.614) (0.017) (0.013)
Constant 1.920 2.878 -1.834 -1.353 20.785 28.925 -1.834 -1.311
(0.928)b (1.063)a (0.574)a (0.555)b (13.060) (14.744)a (0.574)a (0.556)b
R-square 0.74 0.73 0.98 0.98 0.76 0.75 0.98 0.98
Observations 1705 1535 1623 1535 1705 1535 1623 1535
F-test 604.99 236.90 116.38 76.10 556.77 225.97 116.38 73.78
Note: 2SLS procedure. (a), (b), (c) denote significance at the 1, 5 and 10 percent level respectively. The market access
is computed from the estimation of a gravity equation using bilateral exportation from COMTRADE, distance and
geographical variables from the CEPII. This market access is instrumented using genetic distance in 1500 from
Spolaore and Wacziarg (2018). Data on democracy are from Acemoglu, Naidu, Restrepo, Robinson (2017) and
instrumented with their instrument, denoted ANRR, based on waves of democratization and with the mortality rate
of colonizers provided by Albouy (2012). All estimations include time effects and individual fixed effects.

Column (5), (6), (7) and (8) present exactly the same specification with a different
measure of democracy. We use the variable Polity2 instead of the variable of democracy
of ANRR. Results are not affected by this change.

I.2.4 Results, second stage

Table I-5 (Column 1) shows the results obtained by estimating the second stage
described by Equation (14). Column 2 presents a “half-naive” estimation, where the
democracy is instrumented but the market access is not. Column 3 presents the
symmetrical estimation, where the market access is instrumented while the democracy
is not.

41
 Chapter 1 - Trade and Institutions: Explaining Urban Giants

In all estimations reported in Table I-5, whatever the specification used, we find that the
indicator of democracy, instrumented by past mortality rates and by waves of
democratization favors agglomeration in the biggest city46. These results indicate that
the negative impact of democracy found by Ades and Glaeser (1995) and Henderson and
Wang (2007) may be due to a lack of control and/or to endogeneity bias. Another
possibility, which is more interesting, is that there is something specific in our sample of
countries. Ades and Glaeser (1995) for instance consider a sample of seventy countries
without considering the history of these States. Our analysis is slightly different since we
focus our research exclusively on countries colonized by Europeans 500 years ago
following the arguments of AJR and of many historians that the long run effects of
colonization are still unclear. As a result, it is possible that countries not colonized by
Europeans have developed institutions that shape differently their economic
geographies.

Concerning the second hypothesis, the market access is never significant, which reveals
how the previous results done without controls on time-varying institutions and
without an exogenous measure of trade openness were misleading. This is in accordance
with the skeptical view of Ades and Glaeser (1995) and Henderson and Wang (2007)
about the impact of international trade on the urban pattern. The same estimations are
done in Column (4), (5) and (6) but with the variable polity2 and based on the first step
of Columns (5) and (6) of the previous Table I-4. The estimation of (14) in Column (4),
which is the most careful estimation correcting for the endogeneity of the market access
and of institutions, provides the same result than Column 147.

46 When we compute the instrument of democracy by only considering waves of democratization, we get a
very similar result not reported here but available on request (the coefficient is equal to 0.115 and highly
significant with a RSE of 0.036).
47 On the contrary the naive estimation in Column (5), where the market access is not instrumented,

shows that this variable is significant (however only at 10%), revealing the interest of the IV strategy
concerning the impact of trade integration on urban primacy.

42
 Chapter 1 - Trade and Institutions: Explaining Urban Giants

Table I-5: Trade, Institutions and Urban Primacy

dep: Democracy ANRR Polity2
Eq (14) Naive reg Eq (14) Naive reg
Market Access (Ω𝑖𝑡 ) 0.048 0.052
(0.031) (0.031)c
Market Access (𝜚̂𝐷̂𝑖𝑡𝑔 ) -0.158 -0.146 -0.250 -0.148
Instru: Genetic (0.145) (0.173) (0.154) (0.166)
Democracy (𝐼𝑖𝑡 ) 0.038 0.007
(0.016 )a (0.004 )c
Democracy (𝜍̂𝐻̂𝑖𝑡𝑔 ) 0.104 0.092 0.009 0.006
Instru: Mortality+Waves (0.034)a (0.034)a (0.004)b (0.004)c
Light 0.085 0.074 0.075 0.083 0.071 0.081
(0.021)a (0.021)a (0.020)a (0.021)a (0.021)a (0.020)a
Temperature 0.218 0.051 0.171 0.225 0.022 0.159
(0.136) (0.121) (0.135) (0.136)c (0.120) (0.133)
Precipitation -0.008 -0.007 -0.006 -0.007 -0.006 -0.007
(0.012) (0.012) (0.011) (0.012) (0.012) (0.012)
Constant -1.753 -1.761 -1.551 -1.492 -1.636 -1.508
(0.373)a (0.346)a (0.369)a (0.352)a (0.334)a (0.386)a
R-square 0.99 0.99 0.99 0.99 0.99 0.99
Observations 306 306 306 306 306 306
Note: 2SLS procedure. ( ), ( ), ( ) denote significance at the 1, 5 and 10 percent level respectively. The
a b c

market access is computed from the estimation of a gravity equation using bilateral exportation from
COMTRADE, distance and geographical variables from the CEPII. This market access is instrumented using
genetic distance in 1500 from Spolaore and Wacziarg (2018). Data on democracy are from Acemoglu,
Naidu, Restrepo, Robinson (2017) and instrumented with their instrument, denoted ANRR, based on waves
of democratization and with the mortality rate of colonizers provided by Albouy (2012). All estimations
include time effects and individual fixed effects.

I.2.5 Non parametric estimation: matching estimator approach

The previous results are based on a linear functional relationship and on strong
assumptions regarding the IV strategy. The combination of non-random selection into
the process of democratization and the omission of non-linear relationships, can biased
estimates. By finding for each observation in the treatment group, a statistical “twins” in
the control group with the same characteristics, and by using these observations to
compute a counterfactual outcome without treatment for the observations at hand,
matching estimators potentially address these problems without specifying any
parametric assumption, and are thus used here as an alternative strategy.
For any observation, we observe the treatment 𝐼𝑖𝑡 (democratic regime or not) and the
outcome for this treatment: the relative size of the largest city under a democracy,
denoted 𝑢𝑖𝑡 (1), or under another regime, denoted 𝑢𝑖𝑡 (0). We use the Average Treatment

43
 Chapter 1 - Trade and Institutions: Explaining Urban Giants

Effect on the entire sample (ATE) and the Average Treatment effect for the Treated
(ATT) by performing the nearest neighbor matching (Abadie and Imbens, 2006). At least
two assumptions are crucial, the first one is the “conditional mean independence”
leading to assume that the political regime 𝐼𝑖𝑡 is independent of urban primacy 𝑢𝑖𝑡 (1)
and 𝑢𝑖𝑡 (0), conditional on a set of covariates. The second is the stable-unit-treatment-
value assumption (SUTVA) or the “non-interference” assumption, leading in our case to
consider on the one hand that the treatment 𝐼𝑖𝑡 is identical for each treated observation
and on the second hand that being under a democratic regime in a given country does
not influence untreated urban primacy.
We retain covariates used in the previous section apart from temperature and
precipitation, to which we add country-year GDP per capita and the rural population
growth. Using these covariates for the five nearest neighbors, Table I-6 (Row 1 and 3)
presents the ATEs and ATTs. Since the nearest-neighbor matching estimators may not
be consistent when matching on more than one variable48, Table I-6 (Row 2 and 4) also
presents bias-corrected estimator results. Reported ATEs and ATTs are positive,
statistically significant and then confirm results obtained in the previous section.
Democracy causes an increase in the relative size of the biggest city by an average of
1.39 (i.e. 𝑒 0.332 ) point. The sensitivity analysis based on different nearest neighbors (1 to
4) confirms this conclusion.

Table I-6: Average effect of democracy on log urban primacy

Outcome: log_Prim_pop
# of observation (# of Democracies) 703 (532)
0.514
(1) A-I ATEs
( 0.053 )a
0.285
(2) A-I ATEs (Bias adjusted)
( 0.059 )a
0.645
(3) A-I ATTs
( 0.060 )a
0.332
(4) A-I ATTs (Bias adjusted)
( 0.066 )a
0.25
(5) ATT (Propensity score matching)
( 0.121 )b
Note: Matching procedure, all variables are in Log. (a), (b), (c) denote significance at the 1, 5 and 10
percent level respectively. The market access is computed from the estimation of a gravity equation
using bilateral exportation from COMTRADE, distance and geographical variables from the CEPII. Data
on democracy, the treatment, is from Acemoglu, Naidu, Restrepo, Robinson (2016).

48 see Abadie and Imbens (2006)

44
 Chapter 1 - Trade and Institutions: Explaining Urban Giants

As a robustness check, in particular regarding the selection bias, we also use the
propensity-score estimator. In contrast with the previous matching procedure,
propensity-score matching relies on only one characteristic which is the probability of
being in democracy conditionally on covariates; therefore, the matching procedure can
be less precise. We adopt a logit model to estimate these propensity scores. The
covariates are the same ones used for matching earlier. In practice, we take into account
the fact that the propensity scores are estimated rather than known when calculating
standard errors49. Table I-6 presents this PSM-ATT which is computed by calculating the
average of the difference between the observed and potential outcome for each
observation in the treated group. This last analysis confirms the results obtained so far,
the urban primacy has been promoted by democratic regimes.

I.2.6 Discussions

The link between urban planning and institutions studied here is binary (dictatorship
versus democracy) which makes our paper more easily comparable to the literature.
However, political regimes are richer than this simple binary difference and many
interesting results are certainly hidden by this presentation. This is well summarized by
Duranton (2008) who wrote that “this type of explanation [i.e. Ades and Glaeser (1995)]
implicitly assumes fairly strong state institutions able to tax their countryside and
redistribute the proceeds to the primate city. On the contrary, it may be argued that
undemocratic and unstable regimes are weak and favor primate cities by default” (see
Appendix I.4.5). We find that in democratic states, the political regime has no role
whatever the urban pattern. In contrast, in countries with weak dictatorship and with an
urban giant, more democracies foster the development of the largest city, maybe by
“default” as argued by Duranton (2008).

49 The newest command in Stata allows this.

45
 Chapter 1 - Trade and Institutions: Explaining Urban Giants

I.3 Conclusion

Institutions and first cause matter, not trade, in explaining the relative fall of the biggest
cities in former European colonies. Our analysis documents the negative effects of bad
governance on the current level of urbanization. We find that the reversal in urban
primacy is due to past institutions related to colonization. Bad governance and
extractive institutions inherited from colonization have reduced the population growth
of the biggest city in many countries.
Regarding the literature, we have improved the analysis on different aspects by working
with panel data (e.g. the analysis of Ades and Glaeser (1995) is in cross-section only), by
using new controls (e.g. satellite data lights, market access measures) and new
instruments (settlers mortality, waves of democratization) to explain urban primacy. All
our different strategies (using OLS, IV, matching techniques) lead to the same
conclusion: the democracy of the former European colonies has favored cities at the top
of the urban hierarchy. As a result, we do not share the conclusion of Ades and Glaeser
(1995) who wrote that what “was true in Rome, 50 B.C.E., is still true in many countries
today”, on the contrary, we document that dictatorship is no longer a force of
agglomeration in the biggest cities of our sample. This result has concrete consequences,
it means that in many countries where democratic institutions are improved, we can
forecast a more significant growth of the biggest city than what has been predicted until
now with major implications for infrastructures, public health, sanitation, productivity
and inequality.

46
 Chapter 1 - Trade and Institutions: Explaining Urban Giants

I.4 Appendices

I.4.1 Appendix A: Light nights and urban primacy

Table I-7 presents our baseline estimation using the FE estimator with different
retarded measure of income gap approximated by the differential of night lights
between the biggest city and the rest of the country. The current differential seems to
have no effect (Column 1), while the differential of light intensity in the night observed
five (Column 2) or ten years ago (Column 3) explains the current attraction of the
biggest city.

Table I-7: Effect of the constructed light night gap on urban primacy

(1) (2) (3)

FE FE FE
Light gap (level) 0.029
(0.021)
Light gap (Lag = 5) 0.074
(0.032)b
Light gap (lag = 10) 0.085
(0.024)a
Temperature 0.193 0.553 0.848
(0.186) (0.256)b (0.277)a
Humidity -0.022 -0.027 -0.020
(0.019) (0.028) (0.017)
Constant 1.241 0.425 -0.067
(0.541)b (0.733) (0.802)
R-squared 0.983 0.982 0.989
Observations 686 505 322
Note: FE estimator. All variables are in Log. ( ), ( ), ( ) denote significance
a b c

at the 1, 5 and 10 percent level respectively. The income gap (Light gap)
between the biggest city and other cities is approximated by the
differential between the intensity of night lights in the biggest city and the
intensity of lights in other cities obtained from satellite pictures (1992-
2010). All estimations include a full set of year and country fixed effects.

47
 Chapter 1 - Trade and Institutions: Explaining Urban Giants

I.4.2 Appendix B: Outdegree Indicator of Network

Historical data on trade comes from Fouquin and Hugot (2016). Working with historical
data obviously lead to deal with entities that does not exist in 1900 (e.g. colonial
empires). To compute a network indicator at the level of countries (as they are known
today), we assume that a country trading to a group of countries is connected to all of
these countries. For instance when Italy traded with “Italian East Africa” we consider
that Italy was connected to Eritrea, Somalia and Ethiopia. In the reverse case where
“Italian East Africa” traded with Italy, we assume that Eritrea, Somalia and Ethiopia
were connected with Italy. The nature of this database explains why we have not go
beyond an indicator of network based on “links”, indeed building the network of trade
flows would lead to make critical assumptions on the distribution of trade between
countries. Figure below presents this network.

Figure 4: World Trade Network in 1900

Source: Authors

The size of vertices is proportional to the number of outcoming arcs: a large circle
indicates that the corresponding country exports to a large number of exporting

48
 Chapter 1 - Trade and Institutions: Explaining Urban Giants

countries. Considering a dummy variable taking, 𝐿𝑖𝑗 taking one when country 𝑖 and 𝑗 are
trade connected, and zero otherwise, the outdegree measure is computed as follows:

∑𝑗≠𝑖 𝐿𝑖𝑗 (15)

𝑂𝑖𝑗 =
𝑁−1

where 𝑖 is the exporter and 𝑁 represents the number of node/country in this network.

49
 Chapter 1 - Trade and Institutions: Explaining Urban Giants

I.4.3 Appendix C: Theoretical foundation of the market access

Here we briefly present the theory behind the Redding and Venables (2004)
methodology to build the market access variable.
In standard urban economics model of the NEG (e.g. Krugman and Livas, 1998),
individuals consume a composite basket of different varieties produced under
increasing returns and sold under monopolistic competition. These goods are exchanged
between cities and with the rest of the world under iceberg trade costs denoted 𝜏.
Agents lives in monocentric cities and have to commute to work in the central business
district where jobs are located. Market clearing on the goods market gives the following
export equation (denoted 𝑋𝑖𝑗 , expressed in value) from city 𝑖 to 𝑗:

−𝜀
𝑌𝑗 (16)
𝑋𝑖𝑗 = 𝜏𝑖𝑗 𝑆
𝑃𝑗 𝑖

where 𝜀 is the relative elasticity of export with respect to bilateral trade costs i.e.
𝜀 = 𝜕ln(𝑋𝑖𝑗 ⁄𝑋𝑗𝑗 )⁄𝜕ln(𝜏𝑖𝑗 ), 𝑌𝑗 is the nominal income of each importing partner 𝑗 that
depends on wage, commuting costs and land rent. The term 𝑠𝑖 takes into account the
exporting capacity of firms located in 𝑖 (mainly the number of firms and the factory price
that depends on wage and markup). Lastly, 𝑃𝑗 is a reversed measure of the price index in
𝑗 that depends on trade costs and prices, called supplier access of importer 𝑗 (since it is a
weighted sum of the supplier market capacities):

−𝜀
𝑃𝑗 = 𝑠𝑗 𝜏𝑗𝑗 −𝜀
∑ 𝑠𝑖 𝜏𝑖𝑗 (17)
𝑖≠𝑗

−𝜀
Now considering the gravity Equation (16), one can remark that the term 𝜏𝑖𝑗 𝑌𝑗 ⁄𝑃𝑗 is a
measure of the real income of consumers/importers located in 𝑗 net of transportation
costs, in other words it represents the market access of firms that exports from 𝑖 to 𝑗.
The aggregate market access obtained from the location 𝑖 is thus given by:

𝑌𝑖 𝑌
−𝜀 𝑗 (18)
𝛺𝑗 = 𝜏𝑖𝑖−𝜀 + ∑ 𝜏𝑖𝑗
𝑃𝑖 𝑃𝑗
𝑖≠𝑗

50
 Chapter 1 - Trade and Institutions: Explaining Urban Giants

I.4.4 Appendix D: Alternative measures of democracy in the

zero stage regressions

To verify that the non-significant effect of colonizers mortality rate presented above is a
statistical issue depending on the characteristics of the dependent variables which is
dichotomous, we provided robustness check for these first stage regressions concerning
institutions using other measures of democracy, usually exploit in the literature and
mainly used in the construction of ANNR index of democracy. That is to say, the polity 2
score of democracy, and the Freedom House indices of political rights (PR) and civil
liberty (CL). When polity-2 score ranges from -10 to 10 with a higher score indicating
more democracy, the PR and CL indices range from 1 to 7 with a higher score indicating
less freedom. The estimations confirm a negative and significant effect of colonizers
mortality rate on these different measures of democracy and a positive and still highly
significant impact of waves of democratization.
Table (Tab3-1) presents this first stage for some years and confirms the negative sign of
colonizers mortality rate on democracy and the positive impact of waves of
democratization on the polity2 score of democracy.

Table I-8: Democracy (Polity 2 score), zero stage

1962 1972 1982 1992 2002 2010

Waves of democratization 10.228 12.025 11.301 14.209 12.446 10.616
Zi (1.605)a (2.757)a (2.586)a (1.347)a (2.284) a (2.215)a
Mortality rate of colonizers -2.560 -2.777 -2.559 -1.756 -1.013 -1.297
Mi (1.203)b (1.406)c (1.199)b (0.843)b (0.705) (0.760)c
Constant 7.693 6.939 5.660 2.288 0.346 3.074
(5.613) (6.467) (5.648) (3.872) (3.320) (3.185)
R-square 0.659 0.632 0.424 0.669 0.642 0.515
Observations 35 36 36 36 36 36
Note: OLS estimator. (a), (b), (c) denote significance at the 1, 5 and 10 percent level respectively. The mortality
rate of colonizers (in log) comes from Albouy (2012). Data on democracy are from Acemoglu, Naidu, Restrepo,
Robinson (2016) and instrumented with their instrument based on waves of democratization.

Countries where colonizers face high mortality rate have now a lower level of
democracy than other countries. This result is significant over a large part of the period
(excepted four years around the end). Notice also the erosion of this historical heritage,
the coefficient of the mortality rate declines from -2.5 in 1962 to -1.3 in 2010.

51
 Chapter 1 - Trade and Institutions: Explaining Urban Giants

The results are similar when using the Freedom House indices of political rights (PR)
and civil liberty (CL).

Table I-9: Democracy (PR, CL), zero stage

CL
1995 2000 2002 2004 2006 2008 2010
Waves of democratization -3.181 -2.684 -2.918 -2.670 -2.522 -2.597 -2.259
Zi (0.460)a (0.519)a (0.500)a (0.473)a (0.436)a (0.476)a (0.491)a
Mortality rate of colonizers 0.931 0.675 0.660 0.722 0.601 0.617 0.658
Mi (0.161)a (0.237)a (0.228)a (0.200)a (0.194)a (0.198)a (0.222)a
Constant 2.206 2.593 2.676 2.073 2.465 2.537 2.112
(0.857)b (1.129) (1.100) (0.932)b (0.830)a (0.868)a (1.005)b
R-square 0.719 0.574 0.588 0.568 0.548 0.520 0.427
Observations 36 36 36 36 36 36 36
PR
1995 2000 2002 2004 2006 2008 2010
Waves of democratization -3.663 -4.098 -3.848 -3.778 -3.815 -4.047 -3.754
Zi (0.456)a (0.627)a (0.619)a (0.632)a (0.609)a (0.650)a (0.684)a
Mortality rate of colonizers 0.982 0.449 0.499 0.602 0.441 0.561 0.518
Mi (0.155)a (0.246)c (0.262)c (0.274)b (0.248)c (0.240)b (0.253)b
Constant 2.039 4.414 3.911 3.373 4.080 3.836 3.778
(0.731)a (1.222)a (1.148)a (1.190)a (0.909)a (0.791)a (0.966)a
R-square 0.695 0.573 0.604 0.577 0.616 0.640 0.530
Observations 36 36 36 36 36 36 36
Note: OLS estimator. ( ), ( ), ( ) denote significance at the 1, 5 and 10 percent level respectively. The mortality rate of
a b c

colonizers (in log) comes from Albouy (2012). Data on democracy, the political rights (PR) and civil liberties (CL)
indices, are from Freedom House dataset. Waves of democratization variable, is from ANRR (2017).

52
 Chapter 1 - Trade and Institutions: Explaining Urban Giants

I.4.5 Appendix E: Deepening the analysis

In this appendix we analyze whether our results can be generalized inside our
categorization of democracy versus dictatorship and whether the existence of urban
giant can shape our result. We conjecture that for a democratic regime, the existing
urban pattern has no role. To win an election, politicians need to win in different regions
and cities and thus the degree of democratization in these countries may not be a
determinant of urban bias and of urban primacy. Considering now an unstable
dictatorship, such a regime may have a too weak power to reverse the spatial pattern of
a country where activities are dispersed; but the situation can be different in these
regimes if there is already an urban giant. In a weak dictatorship, it can be efficient to
concentrate the governmental policy in the biggest city.

How to define a “weak dictatorship” or the “existence” of an urban primacy? We

consider that dictatorship (as defined by ANRR) with weak military power are weak
dictatorship. We calculate the average “military personal on labor force ratio” over the
period 1992-2010 and each country is labeled as “high military power country” ("low
military power country") when its associated group mean ratio is equal or above
(below) the 75 percentile of its density. While this computation is certainly not
exhaustive, it undoubtedly covers cases of weak dictatorships. Concerning democracy,
based on ANRR, we create a dummy variable taking one when the country has never
experienced a period of dictatorship during all the period and zero when it had
experienced at least one year of dictatorship.

Finally based on the Zipf’s law, establishing that within a country, the size of the largest
cities is inversely proportional to their rank, a dichotomous measure of the current
existence of urban primacy is built. A country is considered as exhibiting urban primacy
if its most populous city has more than twice the population of the second city. On the
contrary, this variable takes zero when the second city in the urban hierarchy has more
than half of the population of the biggest city.

In Table I-10 (column 1 & 2) we study how the share of the population in the biggest city
evolves in countries with democratic rules and where the level of urban primacy is low.
In that case, there is no evidence that institutions influence the relative growth of the
biggest city. Similar results are obtained in democratic countries with urban giants. The
53
 Chapter 1 - Trade and Institutions: Explaining Urban Giants

hypothesis that an urban bias can be beneficial to democratic regime (for instance to
win election) in countries where an urban giant dominates the landscape is not obvious;
there is no apparent relationship between political institutions and urban primacy in
countries already democratized whatever the urban pattern. On the opposite in
countries with weak dictatorship and urban primacy, more democracies foster the
development of the largest city (see Table I-10, col 3). The same result is obtained
concerning countries that already host an urban giant, i.e. in countries where the biggest
city is higher that the size predicted by the Zipf's law, democracies favors the
agglomeration of the population there.

Table I-10: First and Second Nature of Urban Giants

(1) (2) (3) (4)

Density 1500 -0.184 -0.174 0.081 -0.262
(0.039)a (0.037)a (0.065) (0.027)a
Market Access 1.018 0.559 -1.054 -0.411
Instru: Genetic (0.660) (0.620) (0.428)c (0.321)
Democracy -0.065 -0.008 0.166 0.136
Instru: Mort+ANRR (0.153) (0.125) (0.039)a (0.027)a
Light gap -0.013 -0.020 0.272 0.127
(0.055) (0.054) (0.067)a (0.056)b
Temperature 0.256 0.083 -0.204 -0.005
(0.159) (0.095) (0.463) (0.156)
Precipitation 0.015 0.022 -0.063 -0.010
(0.022) (0.018) (0.010)a (0.017)a
Constant -5.021 -0.631 1.445 -1.340
(1.854)a (0.434)a (1.712) (1.057)
R-square 0.994 0.995 0.999 0.998
Observations 81 99 54 180
Note: Column (1): Democracy without urban primacy, Column (2): Democracy with
urban primacy, Column (3): Weak dictatorship with urban primacy, Column (4): All
with urban primacy. Estimation: 2SLS procedure, all variables are in Log. (a), (b), (c)
denote significance at the 1, 5 and 10 percent level respectively. The market access is
computed from the estimation of a gravity equation using bilateral exportation from
COMTRADE, distance and geographical variables from the CEPII. This market access is
instrumented using genetic distance in 1500 from Spolaore and Wacziarg (2017) and
the mortality rate of colonizers from Albouy (2012). Data on democracy are from
Acemoglu, Naidu, Restrepo, Robinson (2016) and instrumented with their instrument,
denoted ANRR, based on waves of democratization and with the mortality rate of
colonizers provided by Albouy (2012). All estimations include a full set of year and
country fixed effects.

54
 Chapter 2 – Beyond the Income Effect of Trade Integration on Ethnic Wars

Chapter 2 -
Beyond the Income Effect
of Trade Integration on
Ethnic Wars

55
 Chapter 2 – Beyond the Income Effect of Trade Integration on Ethnic Wars

II. Chapter 2 - Beyond the Income Effect

of Trade Integration on Ethnic Wars

Introduction

The objective of this paper is to deepen our understanding of the impact of trade on
ethnic conflicts inside each nation (civil war) in Africa. Many conflicts are rooted in the
long term of a national history, but transnational links and interactions with other states
can also influence civil wars. In particular, international trade is considered to affect
conflicts by three mechanisms. The first one is the opportunity cost effect, when trade
shocks decrease revenue, the cost to participate to violent actions follows the same
path50. The second one, is the resource effect, formal trade fosters tax revenues and then
can improve the capacity of the government to buy peace by financing public goods and
by redistributing funds to rebels51.
The third one, is the rapacity effect, a resource that is highly valuable on international
market generates envy and expropriation that can escalate into conflicts52.
All these mechanisms have a common ground, trade affects conflicts by its impact on
real incomes53. This means that by properly controlling for income, trade should have an
insignificant effect on conflicts. To test this, we use night light pictures provided by
satellites from outer space (Henderson et al., 2012) to approximate income at the ethnic
level, and find that trade still has a significant impact on conflicts. Why? Our thesis is that
beyond its monetary effect, trade has many other consequences on institutions, culture,
and on the environment that can play on conflicts. Markets are places of socialization
and then trade by fostering exchanges may have a role in building trust between
communities. Obviously, such a peaceful effect can be reversed if the situation is
considered as unfair. From a theoretical point of view, the exposure to trade implies that

50 Montesquieu (1758), Mayer, Martin and Thoenig (2008.a).

51 See Angrist and Kugler (2008) who analyze how the increase in the price of coca in Columbia, in part
driven by the American demand, has raised revenues for the guerrilla.
52 The rapacity effect is often analyzed by studying trade in mineral products. According to Cali (2015) a
10% increase in the price of oil, raises the risk of conflict by 2.2 percent on average across countries.
53 See also Calvo and Mercier (2019).

56
 Chapter 2 – Beyond the Income Effect of Trade Integration on Ethnic Wars

only the more productive firms enter the export market while the least one are forced to
exit (Melitz, 2003), however this “fair game” can be distorted by bad governance and
deficient institutions. For instance in countries characterized by ethnic favoritism it is
possible that firms with a sufficient productivity to export are stuck on the domestic
market simply because they don't belong to the right group, which may generate
grievance against the system54.
The same resentment may occur in the case of forced trade, defined by Fleurbaey (2015)
as a situation where “the disadvantaged agent would no longer accept the trade if his
disadvantage was removed”55.
In this last case, the income effect of trade can be offset by an ethical one. Indeed forced
trade can be a Pareto improvement bringing pecuniary benefits to all parties however
the unethical nature remains because the benefit of the exploited ethnic group is
conditional to its disadvantage.
Concerning culture, on the one hand, globalization is at the source of cultural destruction
which may render communities more vulnerable (e.g. Maystre, Olivier, Thoenig and
Verdier, 2014). On the other hand, the process of cultural homogenization can also help
to relativize differences between ethnic groups and/or can be an opportunity to build a
new culture by hybridization56.
In that case, globalization by promoting a new national identity can have a peaceful
effect. Finally, trade can affect the environment of different groups (via deforestation or
the rise of local pollution haven) that can be at the source of conflicts.
These non-pecuniary effects of trade, can be different depending on the market
geographical scale at which transactions are done. International trade can be selective,
fostering trade between a community and a far distant partner, and can then weaken
institutional and cultural links inside a country. In other term international trade can be
a substitute to local/internal trade and can thus separate communities. Local trade can
have stronger peaceful effect at least if it is based on voluntary trade and not on unequal
resources (i.e. forced trade) which can pave the way for ethnic conflicts. Regional trade

54 Chaney (2016) presents a model where firms with the potential to export, are prevented from doing so
because they lack sufficient liquidity. One can extrapolate this model by considering that if governments
provide these liquidities arbitrary (crony capitalism) then the outcome becomes economically unfair.
55 Forced trade occurs when the source of trade is related to the economical position of agents (e.g.

inequality). Sexual tourism is an emblematic example but one can think of many other exchanges
apparently less scandalous that can be classify in this category.
56 See Rauch and Trindade (2009), Cowen (2002) and Caplan and Cowen (2004).

57
 Chapter 2 – Beyond the Income Effect of Trade Integration on Ethnic Wars

with neighbor countries can have ambiguous effects. Like international trade it
represents a diversion of socialization inside the country, but as argued by Iwanowsky
(2018) ethnic networks provide social collateral to enforce contracts and can thus
compensate for unfavorable institutions between countries.

Finally, the interest to analyze these three different types of trade is also based on the
fact that different types of goods are traded at different geographical scales and involved
different economical, political and social effects. In Africa, mineral resources and specific
agricultural products (e.g. coffee, cotton etc) represent a significant part of export, while
more basic goods are exchanged on internal or regional markets and concern very
different people. This means that the conflict effect of trade can be different depending
on the geographical trade scale analyzed.

Consequently, this article contributes to the literature of trade by studying three

different type of trade. Using the geographical distribution of 175 crops on a 10 km by
10 km cell grid (Monfreda et al., 2008) and the spatial distribution of mineral resources
in Africa, we build a proxy of the international trade of ethnic groups concerning
agricultural and mineral products. In addition, considering the total value of trade (and
not only agricultural and mineral products), we also compute a indicator of regional
trade between countries that share at least one ethnic group. Finally, we compute an
indicator of internal/domestic trade to analyze its effect on conflicts. We find that at least
two different types of trade have a very similar effect: international and national trade
integration reduce the likelihood of war. This finding holds whatever the specification or
the empirical strategy adopted (i.e. with different controls, estimators, IV strategies).
Results concerning the regional trade integration (between countries that share at least
one ethnic group) are less robust nonetheless a positive effect is obtained once the
endogeneity bias due to reverse causality is taken into account.

All these results are obtained by controlling for income, which is an argument in favor of
our thesis that trade, has effects beyond revenues affecting for instance culture,
institutions and the environment of ethnies. Putting differently trade affect conflicts, not
solely via its income effect, but by an explanatory beam that is related to the “national
identity” of countries. The national identity is a collective product that implied repeated
exchanges between different ethnic groups to build people's sense of belonging to the

58
 Chapter 2 – Beyond the Income Effect of Trade Integration on Ethnic Wars

same nation. Trade, depending on its nature, can foster or deter this national identity
(via the institutional/cultural/environmental channels already described).

To verify how trade affects national identities and finally conflicts, we propose a two-
step approach. The first stage assesses the importance of the variation of income at the
ethnic level against other variations occurring at the national level as a source of conflict.
The source of identification is based on the fact that some ethnic group have been
partitioned by arbitrary borders while other ethnies have not57, and thus the “national
identity” of these two different ethnic groups are certainly different. Ethnic groups that
have been splitting between two different countries have followed different social,
cultural, institutional, and even economical path in comparison with non-partitioned
ethnic groups. Formally, we regress the likelihood of civil conflicts on national time-
varying characteristics (captured with country-year fixed effects), incomes at the ethnic
level (that vary over time), ethnic and time fixed effects. The second stage uses the
country time varying effects, interpreted as a measure of the “changing national
identity”, estimated in the first stage and regress its predicted value on the total exports
of countries (and a set of time dummies and controls). The idea is that time varying
effect obtained in the first step captures all the effects that play on conflicts that are not
related to income, such as cultural, environmental or institutional effects. Instrumental
variables are used to deal with estimation bias. The main finding is that export
significantly reduces conflicts via other channel of diffusion than the income gain.
We are not aware of any work using proxies of income and trade data at ethnic groups’
level, or that develops a similar two-stage approach to study civil wars58. However this
article has been influenced by Gleditsch (2007) and Martin et al. (2008.b). Gleditsch
(2007) analyzes trade between neighborhood countries (approximated by a ratio of
import on GDP of countries that share a border). He finds that the likelihood of peace is
much more higher for countries which are highly trade-integrated with their
neighboring countries than for countries with no trade (the risk of conflict is divided by
2). We find the opposite result by going beyond the the pooled model of Gleditsch
(2007) by using fixed effects in order to treat the endogeneous bias due to omitted time-

57 Indeed, in many African countries a significant fraction of the population belongs to ethnic groups that
are partitioned among different states, see Michalopoulos and Papaioannou. (2016) for a deep analysis of
the long-run consequences of ethnic partitioning in Africa.
58 This two-stage regression is now a standard method in urban economics to analyze agglomeration

economies (Combes, Duranton and Gobillon, 2008) and in labor economics (Abowd et al., 1999).

59
 Chapter 2 – Beyond the Income Effect of Trade Integration on Ethnic Wars

invariant characteristics, and an IV strategy to account for the fact that regional trade is
also a function of conflicts. There are some anecdotal evidences of our result. For
instance the Movement of Democratic Forces of Casamance, a violent separatist
movement in Senegal, was composed of a partitioned ethny, the Diola, and was
supported by the neighboring Guinea-Bissau and Gambia precisely because of the
influence of the Diola in these countries (Michalopoulos and Papaioannou, 2016).
Martin et al. (2008b) consider a different dimension of trade, instead of analyzing
regional trade, they take into account the total amount of trade (more precisely a ratio of
multilateral trade flows on income). They use fixed effects and develop an IV strategy
and find a non-linear effect of trade integration on conflicts depending on the intensity
of conflicts. For low-intensity civil wars, trade openness increases the probability of a
conflict while the reverse holds for high-intensity wars. In comparison with Martin et al.
(2008b), we also analyze domestic and regional trade and we propose a two-stage
analysis enabling to infer the impact of trade on all specific characteristics of countries
that vary over time and that impact on ethnic wars conditioned on incomes.

There is also research that partly overlaps with what we do, yet with very different
objectives and with different methodologies. We briefly present these studies in the next
section. The rest of the paper is structured as follow. Section 2 and 3 present
respectively the empirical strategy and the results concerning the direct impact of the
three different types of trade integration on conflicts. Section 4 discusses our two-step
approach and presents results concerning the impact of trade on time-varying
characteristics of countries (e.g. cultural changes) that affects wars.

II.1 1 Empirical strategy

II.1.1 A brief review of the literature

More than half of the world's nations have been affected by civil wars in the past fifty
years59. As a result, economists, social and political scientists have made great efforts to
better understand the causes of these conflicts. Economic factors such as the ability to

59According to Blattman and Miguel (2010) who consider civil war as a conflict that leads to at least
twenty five battle deaths per year.

60
 Chapter 2 – Beyond the Income Effect of Trade Integration on Ethnic Wars

finance a rebellion, political instability and a rich endowment of resources significantly

explain civil wars (Fearon and Laitin, 2003; Collier and Hoeffler, 2004; Hegre et al.,
2001). Other determinants such as ethnic fragmentation, political grievances, and
income inequality are also common causes of these conflicts (Hegre and Sambanis 2006;
Laitin 2007; Michalopoulos and Papaioannou, 2016). Finally the effect of international
trade, first on income and then on conflict, has been analysed by many authors. A good
example is Bazzi and Blattman (2014) who analyze how income affects the risk of
political instability using data on export price shocks. Their objective is to study how
these shocks, by catastrophically impacting on income, influence conflicts. However, in
their own words, “this paper uses trade shocks as an example”, and quite logically the
authors do not dig into the detail of who trade with who or about the different effects of
trade as we do here. In comparison, the current study is not interested by income
shocks, quite the opposite, our analysis aims to purge the distributional effect of trade
on income to analyze whether there is a residual effect precisely due to other factors.
There are finally many case studies analyzing the link between international trade and
conflicts. Studying the 1994 genocide in Rwanda that occurred during an economic crisis
brought on by the collapse of international coffee prices, Kamola (2007) describes how
the production of coffee, implemented by colonizers, driven by an external demand has
shaped the economy but also the distribution of power and finally conflicts between
Hutu and Tutsi. Still considering the impact of coffee trade, but this time in Columbia,
Dube and Vargas (2013) find that the sharp fall in international prices of this good
during the 1990s has caused more violence in coffee-producing municipalities than in
other municipalities. A reversed relationship has been found for mineral products: a 10
percent increase in the price of oil raises the number of conflict events by 2 percent in
Nigeria according to Cali (2015). Similar results have been found for the Democratic
Republic of Congo (Maystadt et al. 2014), and for Sub-Saharan Africa (Berman and
Couttenier. 2015). Our work differs from these studies according to the data used, the
theory tested and the empirical strategy proposed.

61
 Chapter 2 – Beyond the Income Effect of Trade Integration on Ethnic Wars

II.1.2 International, regional and internal trade at the ethnic

group level

In this section we analyze three types of trade, internal trade inside a nation, the
regional trade of countries where ethnic groups have been partitioned between nations
and the international trade of ethnic groups. As explained in introduction these different
type of trade may have a different impact on conflicts due to the nature of goods
exchanged, and to the different individuals and ethnies that are involved. As a result it is
possible that these exchanges due to their differences affect conflicts differently. There is
obviously no data on trade at the ethnic level, which involves important assumptions to
approximate these flows. Trade variables come from BACI, a database developed by the
CEPII, and are combined with other sources, hereafter described, to measure these three
different type of variables.

 The international trade of ethnic groups, °𝑟𝑖𝑡 .

International trade data at the subnational level in Africa, and even more, data at the
ethnic level are not available; we thus use a weighting strategy based on trade of
particular goods: agricultural goods and natural resources. Formally the sum of exports
in the agricultural sector, denoted 𝑎, and the mining sector, 𝑚 , are taken into account as
follows:

45 48 (19)
°𝑟𝑖𝑡 = ∑(𝑃𝑟𝑖𝑎 × 𝑋𝑖𝑡𝑎 ) + ∑ (𝑃𝑟𝑖𝑚 × 𝑋𝑖𝑡𝑚 )
𝑎=1 𝑚=1

𝑋𝑖𝑡𝑘 represents country 𝑖 exportation at year 𝑡 for a good in sector 𝑘 = 𝑎, 𝑚; indice 𝑎

designs the agricultural goods and indice 𝑚 the mining goods. The weight 𝑃𝑟𝑖𝑎 is the
ratio of the ethnic group production of a good in the agriculture sector on it national
𝐸𝑟𝑖𝑎
output, such as 𝑃𝑟𝑖𝑎 = where 𝐸𝑟𝑖𝑎 is the sum of the mean production of crop a by the
𝐸𝑖𝑎

ethnic group 𝑟 (in country 𝑖) over the period 1997-2003, and 𝐸𝑖𝑎 is the sum of the mean
production of crop 𝑎 in country 𝑖 over the same period.

As for agricultural products, after identifying natural resources location toward Africa,
𝑆𝑟𝑖𝑚
we construct ethnic weight of every resource product as 𝑃𝑟𝑖𝑚 = ∑ , where 𝑆𝑟𝑖𝑚 is the
𝑔 𝑆𝑔𝑖𝑚

62
 Chapter 2 – Beyond the Income Effect of Trade Integration on Ethnic Wars

area (surface) of the ethnic group where the mine 𝑚 is localized and 𝑆𝑔𝑖𝑚 represents the
surface of each of the ethnic groups 𝑔 in the country 𝑖 that exploits this resources 𝑚.

It is noteworthy to note that, we focus on exportation of agricultural goods and mineral

products because these two sectors represent an important share of exportation in
Africa.

Since the weighting scheme is based on the production of agricultural and mineral goods
at the ethnic group's level, we have to geolocate these productions and then to match
them with the geographical boundary of ethnic groups.

Concerning the agricultural sector, we rely on geographical global data set (raster) of
croplands on a 5 minute by 5 minute (10 km by 10 km) latitude/longitude grid extracted
from EarthStat60 that provides information on land use concerning “Harvested Area and
Yield” (see Monfreda et al., 2008). This database contains raster on 17561 agricultural
products with four raters per crop comprising pixels of 10 times 10 square kilometers
over the period 1997-2003. As a result, we have information on (i) the average
proportion harvested for a given agricultural product; (ii) the average number of
hectares harvested per pixel; (iii) the total production of the crops in tons over each
pixel; and (iv) the quality of the data according to their sources. We mainly use raster
about the production of crops (in tons) in order to account for that some agricultural
products are harvested more than once a year. Figure 5 shows the coverage of cotton
and cassava in Africa. The more a pixel is bright, the higher the production within the
area covered.

60http://www.earthstat.org/
61It should be noticed that upstream work has been done to match the products name from EarthStat with
those from Baci. At the end of this exercise, 47 raster out of the 175 in EarthStat were retained.

63
 Chapter 2 – Beyond the Income Effect of Trade Integration on Ethnic Wars

Figure 5: Crops production imagery: cotton and cassava

Source: Authors with Rasters from EARTHSTAT

64
 Chapter 2 – Beyond the Income Effect of Trade Integration on Ethnic Wars

These data are then merged with the spatial coverage of ethnic group reported by
Murdock (1959) and digitized by Nathan Nunn62. Finally we calculate the production
within each ethnic groups area for the different crops identified (48 different crops),
which gives 𝐸𝑟𝑖𝑎 . Once divided by the national production 𝐸𝑖𝑎 , we get our weight of the
national export 𝑃𝑟𝑖𝑎 .

Concerning mining resources, we use the U.S. Geological Survey63 enabling to locate
mining resources toward Africa. This survey provides data on the spatial location of a
large panel of mineral resources all around the world (Mineral Resources Data System /
MRDS) with the name of the deposit, the deposit description, the geological features, the
data source and references64. Then, we merge the Murdock (1959) map of ethnographic
regions shape file, with countries boundaries65 and finally with the resources locations
shapefile in order to identify in which ethnic group and country the natural resources
are located in order to get 𝑆𝑔𝑖𝑚 , 𝑃𝑟𝑖𝑚 and °𝑟𝑖𝑡𝑚 . Figure 5 illustrates the result of this
computation.

62 https://worldmap.harvard.edu/data/geonode:murdock_ea_2010_3
63 https://mrdata.usgs.gov/mrds/
64 https://mrdata.usgs.gov/mrds/package.php
65 ESRI Shape file format: http://www.maplibrary.org/library/stacks/Africa/index.htm

65
 Chapter 2 – Beyond the Income Effect of Trade Integration on Ethnic Wars

Figure 6: Mineral resources location in Africa at the Ethnic/country level

Source: Author's map based on Murdock (1959) and the Mineral Resources Data System.

 The regional trade of countries where ethnic groups have been

partitioned between nations, 𝑖𝑡
𝑆𝑝𝑙𝑖𝑡
.

We compute the sum of bilateral exports of countries that share at least one ethnic
group:

𝑖𝑡
𝑆𝑝𝑙𝑖𝑡
= ∑ 𝑋𝑖𝑗𝑡 (20)
𝑗

where 𝑋𝑖𝑗𝑡 represents time varying trade between two countries 𝑖 and 𝑗 that shares at
least one ethnic group. We consider neighboring trade because the likelihood of civil
wars for countries where ethnic groups have been partitioned seems strong
(Michapoulos and Papaioannou, 2016). We argue that, on top of the ancestral ties
between people, the intensification of trade with peers across borders may deter the

66
 Chapter 2 – Beyond the Income Effect of Trade Integration on Ethnic Wars

relations with the other ethnic groups from the same country. In other words, the ethnic
network or the strength of ethnic ties may affect the dependence with other ethnic
groups inside their country. To identify countries that share the same ethnic groups, we
use the database of Michapoulos and Papaioannou (2016) where an ethnic group is
considered as partitioned when at least 5 percent of its ancestral area belongs to more
than one country.

 The local trade of ethnic groups, 𝑖𝑡 .

Internal flows come from EORA Database Input-Output Tables66. This database contains
the Input-Output tables for 195 countries and allows computing an indicator of internal
trade (the sum of the final demand and trade between sectors). This indicator is often
used in international economics to measure the internal trade of nations.

II.1.3 Estimation

For each type of trade we estimate the following equation:

(21)
𝑖𝑡
Pr(𝑊𝑎𝑟𝑟𝑖𝑡 ) = 𝛾0 + 𝛾1 {𝑖𝑡
𝑆𝑝𝑙𝑖𝑡
+ 𝛾2 𝑌𝑟𝑖𝑡 + 𝛾2 𝐸𝑟𝑖𝑡 + 𝑍𝑖𝑡 + 𝑓𝑟 + 𝑓𝑖 + 𝑓𝑡 + 𝜀𝑟𝑖𝑡
°𝑟𝑖𝑡

where the probability of war, (𝑊𝑎𝑟𝑟𝑖𝑡 ), is approximated by the onset of a new ethnic civil
war and comes from the Ethnic Power Relation (EPR) database compiled by Wimmer et
al. (2009). It is a binary variable taking one when an ethnic group is engaged in a new
ethnic war in a given year and zero otherwise. These data disentangle pure civil war
from ethnic war67 based on PRIO/UCDP conflicts database.

66Data can be download at https://worldmrio.com/countrywise/

67More precisely, Wimmer, Cederman and Min (2009) define ethnic wars as follows: “we distinguish
between ethnic and nonethnic conflicts using the aims of the armed organization and their recruitment
and alliance structures (...)We identify as “ethnic” the aims of achieving ethnonational self-determination,
a more favorable ethnic balance-of-power in government, ethnoregional autonomy, the end of ethnic and
racial discrimination, language and other cultural rights, and so forth. In ethnic wars, armed organizations

67
 Chapter 2 – Beyond the Income Effect of Trade Integration on Ethnic Wars

We measure the income of an ethnic homeland 𝑟 in country i at time, 𝑌𝑟𝑖𝑡 , by exploiting

night light pictures provided by satellites. These data are useful to approximate income
particularly in developing countries because these latter often lack of reliable data at sub
and/or supranational level (Henderson et al., 2012). In order to get night lights intensity
at ethnic group level, we use ESRI African countries shapefile that we merge with
Murdock (1959) ethnic group's boundaries shapefile. Night lights raster come from the
National Oceanic and Atmospheric Administration (NOAA) and are available from 1992
to 2013. A raster contain pixels of approximately one kilometer square. Every pixel is
associated to a number that goes from 0 to 63 indicating the brightness of the light.
These raster are matched with the shapefile (Murdock, 1959) obtained earlier and the
mean night lights density is computed within each ethnic group-country boundaries for
every year. Since satellite pictures do not distinguish gaze flaring activities from real
lights, we exclude the corresponding spatial areas.

The ethnic-country-year controls 𝐸𝑟𝑖𝑡 characterizes ethnic groups that are excluded
from the central government68. This variable is equal to 1 when the ethnic group is
excluded from the central government and 0 otherwise.

𝑍𝑖𝑡 is a vector of country level variables that account for country-year specific effects, we
use two variables: GDP growth (source: World Bank WDI ), an indicator of democracy
(source: Policy-IV project, variable: Polity2).

Finally 𝑓𝑟 denotes time invariant ethnic characteristics or ethnic fixed effects, 𝑓𝑡 allowed
to get rid of time specific shocks that can drive the relation between ethnic wars and the
interest variables, 𝑓𝑖 control for countries characteristic that are invariant in time.
Instead of relying on contemporaneous variables at the right-hand side of our
specification, we use their lags by one year to alleviate endogeneity issues. However, this
concern still persists since we should not probably account for unobservable and/or
observable time-varying indicators that may also drive the relationship at issue. For the
sake of this respondent, we rely on instrumental variables approach by providing and
discussing instruments for each of the endogenous variables, in the previous sections.

also recruit fighters predominantly among their leaders’ ethnic group and forge alliances on the basis of
ethnic similarity”.
68 See EPR (https://icr.ethz.ch/data/epr/) for more details.

68
 Chapter 2 – Beyond the Income Effect of Trade Integration on Ethnic Wars

Concerning the estimator, the use of a logit model augmented with ethnic fixed-effects
leads to a loss of a significant number of observations (Martin et al. 2008a, Martin et al.
2008b). To avoid this loss of data, we can compute a time-demean transformation of
each variable within each unique ethnic group-country combination69, and apply an
ordinary least square (OLS) estimator on these transformations. This procedure
corresponds exactly to a within Fixed-Effects estimation with a Frich-Waugh
transformation method. In our context, this equates to a linear probability model by
using the within regression estimator. However, we rely on this logit model to assess the
effect of ethnic groups’ time invariant characteristics on the probability of ethnic
conflicts (Table II-5 in Appendix A)

II.2 1 Results about the effects of trade on conflicts

II.2.1 Trade and conflicts without income effect

We successively estimate Equation (21) by using our indicator of international, ethnic-

regional and internal trade. Whatever the indicator considered, the inclusion of time-
varying variables (Polity and GDP growth) does not affect results. We find that
international trade as well as internal trade fosters peace between ethnic groups. It is
noteworthy that we control here for income at the ethnic level, thus this result leaves
room to consider that trade have peaceful effects that may be related to other channels.

69 A time-demean transformation of a variable 𝑋𝑟𝑖𝑡 , denoted ∆𝑋𝑟𝑖𝑡 , is given by ∆𝑋𝑟𝑖𝑡 = 𝑋𝑟𝑖𝑡 − 𝑋𝑟𝑖 where 𝑋𝑟𝑖
is the mean of the variable 𝑋𝑟𝑖𝑡 over time of each ethnic group in every country. This transformation
allows absorbing the fixed-effects.

69
 Chapter 2 – Beyond the Income Effect of Trade Integration on Ethnic Wars

Table II-1: International Trade effect on ethnic wars onset

(1) (2) (3) (4) (5) (6)

International trade Regional trade Internal trade
FE FE FE FE FE FE
Trade (𝑖𝑡 , 𝑖𝑡
𝑆𝑝𝑙𝑖𝑡
, °𝑟𝑖𝑡 ) -0.0013 -0.0012 -0.0000 0.0004 -0.0049 -0.0058
(0.0005)b (0.0005)b (0.0007) (0.0009) (0.0021)b (0.0022)a
Rest of Trade 0.0022 0.0019
(0.0017) (0.0017)
Income (𝑌𝑟𝑖𝑡 ) -0.0042 -0.0040 -0.0044 -0.0040 -0.0036 -0.0031
(0.0021)b (0.0021)c (0.0021)b (0.0021)c (0.0020)c (0.0020)
Excluded ethnic groups 0.0079 0.0095 0.0065 0.0092 0.0044 0.0082
(0.0032)b (0.0036)a (0.0033)c (0.0039)b (0.0026)c (0.0032)b
Polity2 0.0004 0.0006 0.0008
(0.0003) (0.0004) (0.0003)a
GDPgrowth 0.0002 0.0004 0.0005
(0.0001)a (0.0002)b (0.0002)a
Constant -0.0093 -0.0082 -0.0263 -0.0236 0.1006 0.1233
(0.0080) (0.0079) (0.0127)b (0.0127)c (0.0487)b (0.0507)b
Observations 5,57 5,57 5,554 5,554 5,586 5,586
R-squared 0.018 0.018 0.015 0.017 0.013 0.016
Year FE Yes Yes Yes Yes Yes Yes
Ethnic Group-Country FE Yes Yes Yes Yes Yes Yes
Notes: The dependent variable is a binary variable taking 1 when an ethnic group is engaged in a new ethnic
war in a given year. All explanatory variables are lagged by one year. Trade designs respectively International
trade, Internal trade and Regional trade. Robust standard errors clustered at “Ethnic group and country” level
reported in parentheses. (a) (b) (c) denote significance at the 1, 5 and 10 percent level respectively.

A conflict may be costly in various ways for ethnic groups since insecurity and the
destruction of trade facilities may compromise the shipping of products and/or the
functioning of the exchange platforms (such as markets places, exports platforms, etc.).
Even if we conjecture that the use of one year lag of the interest variables is enough to
tackle the reverse causality issue, the endogenous bias due to uncontrolled confounding
variables issue may still keep doubt on the reliability of the results in Table II-1. We
therefore rely on instrumental variables (IV) strategy to overcome this issue.

 International trade

To correct the endogeneity of international trade at the ethnic group, we rely on

weather conditions (temperature) as a source of discretionary variation of ethnic trade.

70
 Chapter 2 – Beyond the Income Effect of Trade Integration on Ethnic Wars

Good weather conditions are reliable instruments in our case because a large part of the
African population depends on agriculture activities and since 95% of cultivated land in
Africa is not irrigated70, weather matters. We compute an indicator of temperature at
ethnic level, with the CHELSA71 dataset that provides high spatial resolution of mean,
maximum and minimum temperatures and mean precipitation for the Earth's land
surface, for the years 1979 to 2013. A good weather in terms of temperature may
provide favorable conditions for agricultural exploitation. As shown in Table II-6,
moderate temperature fosters trade, while higher level temperature negatively affects
the international trade of ethnic groups. We thus use these instruments to run the IV
regressions. As a result, the coefficient on international trade is still significant and
negative72 , more foreign trade reduces the risk of escalating into war (Table II-2,
Column 1 and 2).

The exclusion restriction condition behind our IV estimation is that weather conditions,
particularly temperature in our case, can affect the probability of ethnic conflicts only
through trade. This argument is supported by Burke et al. (2015) who prove that climate
conditions do not drive conflicts alone73. Nevertheless, researches in the field of
psychology state that hot temperature can stimulate violence behavior by directly
affecting feelings of hostility that can lead to aggressive thoughts (see Anderson, 2001).
Accounting for this argument should lead to a violation of the exclusion restriction
condition: hot temperature can drive violence and in turn may be positively correlated
to ethnic groups likelihood to conflicts. In such a case, we can infer that our estimated
coefficient is biased upward, and since our estimated coefficient is negative, the real
causal effect should bound to be negative also.

 Regional Trade

70 https://www.worldbank.org/en/programs/cooperation-in-international-waters-in-africa
71 Climatologies at High resolution for the Earth's Land Surface Areas
72 We provide robustness check to this finding by also capturing trade with the total exports at country

level. Actually, we estimate the same relationship by using the country total exports to gauge if results
from our constructed ethnic exports data still hold. We keep using the same instrument (ethnic level
temperature) for trade to make results comparable. As we expected, the results confirm that international
trade is a peace promote. Actually, the IV regression leads to a coefficient of -0.0342 significant at 10%.
73 https://www.annualreviews.org/doi/pdf/10.1146/annurev-economics-080614-115430

71
 Chapter 2 – Beyond the Income Effect of Trade Integration on Ethnic Wars

Regarding regional trade, we deal with the endogeneity issue by using a variable of
remoteness (𝑅𝑒𝑚𝑖𝑡 ) widely used in the literature of trade (Baier et Bergstrand, 2004,
Martin et al., 2008a,b, Candau and Dienesch, 2015) . We restrict here the panel of
partners 𝑗 to countries that share at least one ethnic group with country 𝑖. Then, the
remoteness instrument for the regional trade in this case is:

𝑅 (22)
𝐺𝐷𝑃𝑗𝑡
𝑅𝑒𝑚𝑖𝑡 = −𝑙𝑛 (∑ )
𝑑𝑖𝑗
𝑗≠𝑖

where 𝑅 is a set of partners with which country 𝑖 share at least an ethnic group.

This indicator is a good candidate to explain trade because it contains two important
factors of a gravity equation, distance with partners and theirs GDPs, that explain the
lion share of the variation in bilateral trade. A country with low remoteness has more
alternative sources of exports and is therefore more open to trade. In the regression, we
also control for trade with the rest of countries that do not share an ethnic group with
country 𝑖 (is different from countries in 𝑅), called in the different tables “Rest of Trade”.

We find in Table (Table II-2, Column 3 and 4) a new result; ethnic/regional trade fosters
wars. In Appendix D we lead a robustness check by modifying our indicator of regional
trade by using imports (instead of export) from partners with which country 𝑖 shares an
ethnic group. We also follow Gleditsch (2007) by building an indicator of regional trade
using imports from partners with which country 𝑖 shares a border. Similar results are
obtained whatever the indicator used (see Appendix D). These results contrast with the
seminal result of Gleditsch (2007) who find that the risk of conflict is divided by two
when countries are highly trade-integrated with their neighboring partners.

 Internal Trade

The results with internal trade are also not immune to endogeneity problems. Actually,
ethnic war can disrupts trade by affecting trade facilities and by corroding trust between
communities. As for international trade, we rely on the weather conditions to predict
country production on which our indicator of internal trade is built. The precipitation
data comes from the precipitation raster of CHELSA. In addition to this instrument, we

72
 Chapter 2 – Beyond the Income Effect of Trade Integration on Ethnic Wars

also consider the three years lag of internal flows as a source of it contemporaneous
variation in Column 6. The results from the first stage regressions are presented in Table
II-9. The findings from the second-stage IV regressions reveal that internal trade
unambiguously promotes peace; however, this integration has less impact on ethnic
conflicts than the international ethnic trade integration effect (see Table II-2, column 5 &
6).

Table II-2: International Trade effect on ethnic wars onset

(1) (2) (3) (4) (5) (6)

International trade Regional trade Internal trade
IV IV IV IV IV IV
Trade (𝑖𝑡 , 𝑖𝑡
𝑆𝑝𝑙𝑖𝑡
, °𝑟𝑖𝑡 ) -0.1170 -0.1185 0.0165 0.0266 -0.0065 -0.0082
(0.0610)c (0.0576)b (0.0060)a (0.0096)a (0.0031)b (0.0035)b
Rest of Trade -0.0031 -0.0063
(0.0025) (0.0034)c
Income (𝑌𝑟𝑖𝑡 ) -0.0044 -0.0044 -0.0048 -0.0029 -0.0015 -0.0007
(0.0021)b (0.0021)b (0.0022)b (0.0021) (0.0022) (0.0020)
Excluded ethnic groups 0.0087 0.0092 -0.0054 0.0053 -0.0023 0.0028
(0.0032)a (0.0036)b (0.0045) (0.0034) (0.0023) (0.0028)
Polity2 0.0001 0.0043 0.0013
(0.0003) (0.0016)a (0.0005)a
GDP growth 0.0001 0.0003 0.0003
(0.0001) (0.0002)c (0.0001)b
Constant 0.0622 0.0626 -0.0315 -0.0211 0.1532 0.1900
(0.0394) (0.0373)c (0.0132)b (0.0127)c (0.0743)b (0.0810)b
Observations 5,57 5,57 5,528 5,528 4,202 4,202
R-squared 0.017 0.018 0.018 0.020 0.012 0.016
Year FE Yes Yes Yes Yes Yes Yes
Ethnic Group-Country FE Yes Yes Yes Yes Yes Yes
Notes: The dependent variable is a binary variable taking 1 when an ethnic group is engaged in a new ethnic
war in a given year. All explanatory variables are lagged by one year. Trade designs respectively International
trade, Internal trade and Regional trade. Robust standard errors clustered at “Ethnic group and country” level
reported in parentheses. (a) (b) (c) denote significance at the 1, 5 and 10 percent level respectively.

II.3 Inference from a two-step strategy

II.3.1 Empirical strategy

The previous analysis aims at using commonly and widely known strategy to identify
the effect of trade on war. But at least three caveats can be formulated.

73
 Chapter 2 – Beyond the Income Effect of Trade Integration on Ethnic Wars

Firstly, the measurement of ethnic trade potentially suffers of errors that may bias the
analysis. Secondly, by merging aggregate data (country level variables) to micro
observations (ethnic groups) we make a strong assumption of independent disturbances
which may be not verified for data with grouped structure (Moulton, 1990). This is
particularly the case of estimates concerning regional and internal trade, which are
country-level variables. In that case, OLS standard errors can be biased downward
leading the aggregated variables to be artificially significant. Furthermore single-stage
estimation is problematic because we mislead country shocks affecting conflicts with
idiosyncratic risk, e.g. a particular type of ethnic groups has an endemic risk of conflict
not totally controlled due to missing variables.
Thirdly, our specification suffers of various problems. To make them clear, let's rewrite
the Equation (23) in its simplest form by dropping the time dimension, all the variables
of control and by focusing on international trade as follows:

𝑃𝑟(𝑤𝑎𝑟𝑟𝑖 ) = 𝛾0 + 𝛾1 °𝑟𝑖𝑡 + 𝑓𝑟 + 𝑓𝑖 + 𝜀𝑟𝑖 (23)

An unbiased estimate for 𝛾1 is obtained under the following assumption:

𝐶𝑜𝑣(°𝑟𝑖 , 𝑓𝑟 + 𝑓𝑖 + 𝜀𝑟𝑖 ) = 𝐶𝑜𝑣(°𝑟𝑖𝑡 , 𝑓𝑖 ) + 𝐶𝑜𝑣(°𝑟𝑖𝑡 , 𝑓𝑟 ) + 𝐶𝑜𝑣(°𝑟𝑖𝑡 , 𝜀𝑟𝑖 )=0 (24)

When the covariance in the left hand side of this equation differs from zero, any of the
three covariances on the right hand side is a source of bias. Since this equation is the
workhorse model in economics and in political sciences when analyzing conflicts, we
propose to discuss these covariances successively.

Concerning the first term 𝐶𝑜𝑣(°𝑟𝑖 , 𝑓𝑖 ), there are many reasons to consider that the time-
varying trade integration of ethnic groups depend on invariant country characteristics.
For instance the level of development (industrialized or not) and the technology used
often characterizes a country, and are crucial determinants of trade. For instance to
capture comparative advantage, Costinot (2009) used country (and product) fixed
effects. Furthermore many political and economic institutions are country-specific and
influence the trade integration of nations. For instance, the concept of forced trade that
we define in the introduction clearly evolves in an institutional framework that depends
on the long run history of nations, it is thus quite naïve to consider that the covariance
between °𝑟𝑖𝑡 and 𝑓𝑖 is null.

74
 Chapter 2 – Beyond the Income Effect of Trade Integration on Ethnic Wars

The second term 𝐶𝑜𝑣(°𝑟𝑖 , 𝑓𝑟 ) is interesting to discuss using the theory of factor
endowments (Heckscher, Ohlin and Samuelson) and of the resource curse theory
(Corden and Neary, 1982) from the point of view of regional economics. When a
resource, valuable on the international market, is discovered in one region, often in one
ethnies, the wage and price increase in this sector are spreading throughout the whole
economy, including in sectors with no productivity gains which generate a loss of
competitiveness74. Consequently the industrial and agricultural sectors that are owned
by other ethnic groups are in difficulties (ask for fiscal redistribution, which in reaction
fosters the desire of independence of the exporting resource ethnic group, and this
situation sometimes degenerates into conflicts) which affects their export. To sum up,
the factor endowment of ethnic group, 𝑓𝑟 , being a natural resources or not, is highly
correlated with the trade exchange of this ethnic group °𝑟𝑖 .

The third term 𝐶𝑜𝑣(°𝑟𝑖𝑡 , 𝜀𝑟𝑖 ) is the classical origin of an endogeneous bias. While in the
previous section, we have devoted significant effort to address this last issue, the first
two points remain problematic. We thus propose a two stages estimation approach75
that resolves partially these problems. The first stage is an estimation of the probability
of wars on income (𝑌𝑟𝑖𝑡 ), on country-year effects (𝑓𝑖 × 𝑓𝑡 ), ethnic group's fixed effects (𝑓𝑟 )
such as:

𝑃𝑟(𝑤𝑎𝑟𝑟𝑖𝑡 ) = 𝛾1 𝑌𝑟𝑖𝑡 + 𝑓𝑖 × 𝑓𝑡 + 𝑓𝑟 + 𝜀𝑟𝑖𝑡 (25)

The estimation of the country-year fixed effects (𝑓𝑖̂

× 𝑓𝑡 ) is interpreted as the change in
national identities that influences conflicts after controlling for observed (𝑌𝑟𝑖𝑡 )76 and
unobserved (𝑓𝑟 ) ethnic characteristics.

74 See Corden and Neary (1982). We are particularly grateful to Peter Neary to have discussed with us the
implication of “booming sector and de-industrialization” at the regional scale.
75 As a robustness check of the analysis in the first section, we could have followed Berman and Couttenier

(2015) by using a specification at country level as in Bazzi and Blattman (2014); where the dependent
variable is a dummy that equals 1 if at least one ethnic conflict event was recorded during year 𝑡 in
country 𝑖. Nevertheless, this will not allow us to account for ethnic group level income; in fact studying the
effect of trade on conflicts beyond the ethnic groups’ income effect is one of the essential points of this
study.
76 Income can also be endogenously determined by conflicts. Here, we still present results with a lag of one

year to reduce this problem; similar results with different lag are also obtained (not reported but available
on request).

75
 Chapter 2 – Beyond the Income Effect of Trade Integration on Ethnic Wars

Thus this method allows exploiting the fact that there are two kinds of ethnic groups,
some that belong exclusively to one country and those that are shared by at least two
countries77. The former dimension helps to identify changes over time within countries,
while the latter provides the identification that accounts for differences between
countries over time. The identification of differences in term of wars between countries
comes from the ethnic groups that have been partitioned. Indeed as shown by
Michalopoulos and Papaioannou (2016), these ethnies share the same characteristics
that we control for with ethnic fixed effects f r , then everything that varies for these
partitioned ethnic groups inevitably comes from the national context and then measured
by the country-year fixed effects (𝑓𝑖̂
× 𝑓𝑡 ).

Finally this approach by differentiating ethnic effects from country trends, resolves the
problem of 𝐶𝑜𝑣(°𝑟𝑖 , 𝑓𝑖 ) ≠ 0 (as the next stage will make clear).

In the second step, we analyze how trade, now measured at the country 𝑖 and time 𝑡
level, can explain change in national identities (𝑓𝑖̂
× 𝑓𝑡 ) that affects conflicts:

(26)
𝑖𝑡
× 𝑓𝑡 = 𝛾1 {𝑖𝑡
𝑓𝑖̂ 𝑆𝑝𝑙𝑖𝑡
+ 𝑍𝑖𝑡 + 𝑓𝑡 + 𝜀𝑖𝑡
°𝑟𝑖𝑡

These equations allow capturing how trade, by affecting change in the national
identity 𝑓𝑖 × 𝑓𝑡 , explains civil conflicts. In other words, we assess here the extent to
which, the three kind of trade drive the country-year endemic propensity to ethnic war.
Since these changes are not related to the income changes of ethnic groups, our second
stage enables to assess the effect of trade on non-monetary variables. As briefly
discussed in the introduction, 𝑓𝑖 × 𝑓𝑡 may be interpreted as the “national identity” of a
country on which globalization has various effects. The process of cultural

77 This strategy is used in urban economics to identify the source of agglomeration economies at the urban
level (Combes et al 2008). Wages are explained by city-time effects (here nation-time) and individual fixed
effect (here ethnic groups fixed effects) to control for characteristics such as the human capital of
individuals. Then in a second stage, the predicted value of city-time effects are regressed on the city
density (here national exports).

76
 Chapter 2 – Beyond the Income Effect of Trade Integration on Ethnic Wars

homogenization can help to relativize differences between ethnic groups and/or can be
an opportunity to build a new culture by hybridization78.

In that case, globalization has peaceful effects79. Obviously, trade can have effect on
𝑓𝑖 × 𝑓𝑡 that are unrelated to culture. The resource curse is a well-known example.
Extraction of highly profitable resources sold on the world market can foster bad
governance. In turn, institutional change with its share of injustice is a fertile ground for
conflicts. Maybe less known, the relocation of polluting firms in developing countries are
often the cause of a reduction/delay in the restrictiveness of environmental law at the
national level (see Candau and Dienesch, 2017), and this may lead to a destruction of the
local environment of some ethnic groups. Since these pollution haven are clearly
motivated by international trade, the coefficient 𝛾1 also capture these kinds of effect.

II.3.2 Results

Table II-3 presents the first stage results from the estimation of Equation (26). In
Column (1), we consider only ethnic group's level of income as explanatory variables.
Column (2) performs the same regression as Column (1) by accounting for some of the
country level time-varying and observable characteristics. In Column (3) we consider
the full set of country level-time varying particularities by introducing the country-year
fixed-effects.

The main finding in this table is that the income growth of ethnic groups reduces the risk
of ethnic conflict even if we control for the full set of country year covariates. This result
confirms at the local level what has been analyzed until now at the national one in the
literature. It is noteworthy to observe the significant change in the R-squared from the
first to the third column. Indeed, even if the first column shows that there is a
correlation between wars and income, this latter alone seems to be a poor predictor of

78Rauch and Trindade (2009) provide a theoretical foundation to analyze the creation of cultural goods by
hybridization. They find that globalization is detrimental because the destruction of local cultures reduces
the possibility of hybridization and generates a cultural blowback. On the contrary to Cowen (2002) view
of globalization as a positive process of creative destruction. Caplan and Cowen (2004) defends that no
great culture has risen in isolation. The fear that globalization builds up a culture of the least common
denominator, has been modeled by Francois and van Ypersele (2002) who analyzes the pro and cons of
the dumbing down argument.
79 After all in Europe, the reunification of Germany has been a success in part thanks to its export

performance at the world level.

77
 Chapter 2 – Beyond the Income Effect of Trade Integration on Ethnic Wars

wars. The introduction of the full set of country-year covariates in Table II-3, Column
(3), improves the goodness-of-fit measure of our regression. This confirms that country
level time varying characteristics are important and need to be account for when
explaining ethnic wars.

Table II-3: First stage regression

Dep var: Ethnic war ( 𝒘𝒂𝒓𝒓𝒊𝒕 ) (1) (2) (3)

Log Income -0.0043 -0.0040 -0.0037
(0.0019)b (0.0019)b (0.0022)c
Polity2 0.0004
(0.0002)c
GDP growth 0.0005
(0.0002)a
Constant -0.0133 -0.0115 -0.0037
(0.0070)c (0.0069)c (0.0022)c
Observations 6,348 6,348 6,348
R-squared 0.011 0.013 0.260
Ethnic Group-Country FE Yes Yes Yes
Year FE No No No
Country × Year FE ( 𝑓𝑖 × 𝑓𝑡 ) No No Yes
Notes: The dependent variable is a binary variable taking 1 when an ethnic group is
engaged in a new ethnic war in a given year. All explanatory variables are lagged by
one year. Column 1 and 2 are estimated without taking into account country-year
characteristics. Column 3 accounts for both year FE and country-year effects. Results
look exactly the same when we do not introduce year FE in column 3. The predicted
country-year FE are extracted from this latter regression. Robust standard errors
clustered at “Ethnic group and country” level reported in parentheses. (a), (b), (c)
denote significance at the 1, 5 and 10 percent level respectively.

Table II-4 presents the second stage of our analysis, by showing the results of the effect
of our three kinds of trade on countries time-varying determinant of the risk of ethnic
conflicts. The results in Columns (1), (2), (4), (5), (7) and (8) are first analyses and do
not account for the endogeneity issues.. On the one hand, these results show that, both
international ethnic trade and local trade are negatively associated to the probability of
a country to experience an ethnic war. On the second hand, the coefficients on regional
trade are not significant, while that of the internal trade, albeit negative, is not
significantly robust (only weakly significant in Column 8).

Since reverse causality is likely, we need to find valid instruments to address this
endogeneity issue of our three variables of interest.

78
 Chapter 2 – Beyond the Income Effect of Trade Integration on Ethnic Wars

Concerning international trade, we resolve the endogeneity issue by considering the

remoteness variable as described previously (with some minor changes: since we now
consider total trade, the sum in Equation (22) is done on all the trade partners of
country 𝑖 all around the world). In a similar way, the endogeneity issue concerning
regional trade is solved by relying on the regional remoteness instrument proposed in
Equation (22). Finally, to deal with the endogeneity bias on internal trade we use the
weather conditions (precipitations, also drawn from raster of CHELSA) at the country
level, and the two years lag of the internal flows. Here, we account for the level of
interaction between people within countries, by using their population density as a
control variable.

The first stage estimations for the IV results in column (3), (6) and (9) of Table II-4 are
presented respectively in column (1), (2) and (3) of Table Table II-10 (Appendix F). The
results on total trade are close to that obtained in the first section (see Table II-2,
column 1 and 2). On the contrary, the effects of regional and internal trade on the
probability of countries risk of ethnic conflicts are not significant.

79
 Chapter 2 – Beyond the Income Effect of Trade Integration on Ethnic Wars

Table II-4: Second stage regression: Country fixed effects regression

Dep var : 𝑓𝑖 × 𝑓𝑡 (1) (2) (3) (4) (5) (6) (7) (8) (9)
International trade Regional trade Local trade
Trade (𝑖𝑡 , 𝑖𝑡
𝑆𝑝𝑙𝑖𝑡
, °𝑟𝑖𝑡 ) -0.0052 -0.0053 -0.0087 -0.0014 -0.0013 0.0010 -0.0035 -0.0032 -0.0023
(0.0021)b (0.0021)b (0.0046)c (0.0016) (0.0015) (0.0011) (0.0021) (0.0019)c (0.0014)
Rest of Trade -0.0030 -0.0034 -0.0047
(0.0018)c (0.0022 (0.0029)
Polity2 -0.0008 -0.0008 -0.0003 -0.0003 -0.0005 -0.0006
(0.0005) (0.0006) (0.0006) (0.0006) (0.0005) (0.0005)
GDP growth 0.0005 0.0008 0.0004 0.0004 0.0005 0.0009
(0.0004) (0.0004)c (0.0004) (0.0004) (0.0004) (0.0004)b
Density -0.0004 0.0014
(0.0030) (0.0027)
Constant 0.0242 0.0239 0.0402 0.0171 0.0181 0.0202 0.0781 0.0733 0.0458
(0.0107)b (0.0096)b (0.0220)c (0.0112) (0.0123) (0.0141) (0.0473) (0.0437) (0.0291)
Estimator OLS OLS IV OLS OLS IV OLS OLS IV
Observations 569 569 480 421 421 418 536 536 445
R-squared 0.070 0.080 0.051 0.069 0.078 0.070 0.057 0.064 0.072
Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes
Notes: This Table presents the results from the second step (Equation (26)). All explanatory variables are lagged by one year. Trade designs respectively
International trade, Internal trade and Regional trade. Predicted values of time-varying country effects are used as dependent variables and come from Column 3
of Table II-3. Robust standard errors clustered at “Country” level reported in parentheses. (a), (b), (c) denote significance at the 1, 5 and 10 percent level
respectively.

80
 Chapter 2 – Beyond the Income Effect of Trade Integration on Ethnic Wars

This two-step analysis helps us to better distinguish the effect of trade on income to
other effects. While the one-step approach presented in the first section supports the
thesis that both internal and international trade have a peaceful effect, the current
analysis relativize the role of internal and regional trade. Only international trade
integration has a robust significant negative effect on ethnic conflicts. Consequently at
the end of this careful analysis, we consider that only the effect of international trade
integration effect has the potential to shape national identities and to reduce the risk of
ethnic wars.

Conclusion

The beneficial impact of international trade on peace is an old idea that has been at the
heart and soul of many trade agreements in Europe, in Latin America and in Africa.
Integration in the world economy has been conceived as a way that promotes peace. The
contribution of this paper is to propose different measures and methodologies to
precisely analyze the geographical impact of trade on conflicts. More precisely we build
indicators of international, regional and internal trade and assess their effects on
conflicts at the level of ethnic groups. We propose original empirical strategies, among
which a two stages analysis enabling to separate the income effect to other effects of
trade.

We find that international trade clearly favors peace while internal and regional trades
seem less significant. To the famous quotation of Arrow (1972), that “virtually every
commercial transaction has within itself an element of trust”, we reverse the causality
and guess that the performance on the world market fosters mutual confidence,
cooperation and good institutions inside nations that reduce the likelihood of ethnic
conflicts.

81
 Chapter 2 – Beyond the Income Effect of Trade Integration on Ethnic Wars

II.4 Appendices

II.4.1 Appendix A: First resuts

We also test the effect of some ethnic group's controls that are time invariant on the
probability of ethnic war when ethnic trade and ethnic income are controlled for (See
Table II-5). We then consider: dummy variable taking 1 in case of the presence of a
major lake/river (source: Michapoulos and Papaioannou, 2016), a variable indicating
the distance from the sea, a dummy variable taking 1 when an on-shore oil field and gas
deposit is in the historical homeland of an ethny (called petroleum, source: the
Petroleum Dataset v.1.1); a geographical variable such as the Log of ethnic groups'
homeland surface(lnkm2split, source: Global Mapping International/Colorado Springs),
a binary variable taking 1 when a city with a population larger than 20 000 in 1400 was
in the historical homeland (called city1400, Chandler, 1987); a variable indicating the
ethnic group population in the 1960s.

82
 Chapter 2 – Beyond the Income Effect of Trade Integration on Ethnic Wars

Table II-5: Results from naïve regressions

(1) (2) (3)

International Internal Regional
trade trade trade
Logit Logit Logit
Trade -0.0635 -0.0878 -0.3391
(0.0617) (0.0748) (0.1385)b
Income -0.4210 -0.4869 -0.4938
(0.1290)a (0.1223)a (0.1238)a
Excluded ethnic group 1.4729 0.9372 1.2706
(0.5725)b (0.4531)b (0.5080)b
River -0.3031 -0.5979 -1.3582
(0.5254) (0.5182) (0.5673)b
Distance from sea 1.3573 1.6835 0.7867
(0.8606) (0.8439)b (0.8036)
Petroleum -0.1082 -0.1282 -0.1960
(0.4913) (0.4824) (0.7837)
Land surface 0.3462 0.2283 0.1439
(0.2007)c (0.1865) (0.1852)
City1400 0.4730 0.5509 0.6393
(0.6334) (0.5981) (0.6321)
Population1960 0.2195 0.0447 0.0160
(0.1915) (0.1149) (0.1342)
Constant -12.7101 -10.4668 -1.2566
(2.7066)a (1.9478)a (3.7211)
Observations 5,566 5,496 5,582
Pseudo R-squared 0.2753 0.2424 0.2707
Notes: The dependent variable is a binary variable taking 1 when an
ethnic group is engaged in a new ethnic war in a given year and 0
otherwise. All the time varying explanatory variables are lagged by one
year. Trade designs respectively International trade, Internal trade and
Regional trade. Robust standard errors clustered at the “Country pair”
level reported in parentheses. (a) (b) (c) denote significance at the 1, 5 and
10 percent level respectively.

83
 Chapter 2 – Beyond the Income Effect of Trade Integration on Ethnic Wars

II.4.2 Appendix B: First-stage IV estimates, explaining ethnic

groups' international trade

The following tables present the first stage regressions used in Table II-2, column (1)
and (2). It shows a strong and positive relation between temperature and ethnic groups'
international trade. We can also use the mean temperature (( 𝑀𝑎𝑥_𝑇𝑒𝑚𝑝𝑒𝑟𝑎𝑡𝑢𝑟𝑒 +
𝑀𝑖𝑛_𝑇𝑒𝑚𝑝𝑒𝑟𝑎𝑡𝑢𝑟𝑒 ) /2 ) at the ethnic groups' level as instrument, but this measure
seems less relevant. For instance, a very high level of temperature recorded during a
relatively temperate period can bias the average temperature during that period and
conversely. In the meantime, a very high level of temperature during a given period is
sufficient to dammage (or negatively affect) the growth of agricultural crops,
particularly in Africa where agricultural technologies (ex: greenhouse cultivation,
irrigation practices) are less available. As a consequence, we think that it is more
meaningful to using both the minimum and maximum temperature variables that we
have constructed from the rasters proposed by CHELSA.

Table II-6: IV first stage regressions: explaining ethnic groups' international trade

International trade
(1) (2)
Min Temperature 3.3024 3.4202
(0.6394)a (0.6509)a
Max Temperature -1.1613 -1.3089
(0.4469)a (0.4566)a
Ethnic group Income -0.0019 -0.0034
(0.0051) (0.0051)
Excluded ethnic group 0.0094 -0.0020
(0.0193) (0.0202)
Polity2 -0.0029
(0.0008)a
GDP growth -0.0007
(0.0004)c
Constant -16.3772 -16.1294
(4.7937)a (4.6871)a
Observations 5,57 5,57
R-squared 0.165 0.168
F-Test 26.68 13.87
Year FE Yes Yes
Ethnic Group--Country FE Yes Yes

Notes: This table presents the first stage regressions results concerning
the IV estimations in column 1 and 2 of Table II-2. Robust standard errors
clustered at “Ethnic group and country” level reported in parentheses. (a)
(b) (c) denote significance at the 1, 5 and 10 percent level respectively. We
present the F-tests statistics on instruments; they are always above the
critical value of 10, recommended by Staiger and Stock (1997).

84
 Chapter 2 – Beyond the Income Effect of Trade Integration on Ethnic Wars

II.4.3 Appendix C: First-stage IV estimates, explaining regional

trade

The following tables present the first stage regressions used in Table II-2, Column (3)
and (4). The remoteness instrument is restrained to the panel of partners 𝑗 with which
countries 𝑖 shares at least one ethnic group: we name this instrument regional
remoteness. The high significance of the coefficients on the instrument and the F test of
weak instruments in these first-stage regressions sufficiently reveal that the regional
remoteness is reasonably a strong instrument for regional trade.

Table II-7: IV first stage regressions: explaining regional trade

Regional trade
(1) (2)
Regional remoteness -0.8364 -0.5761
(0.1862)a (0.1635)a
Rest of Trade 0.3199 0.3136
(0.0926)a (0.0879)a
Ethnic group Income 0.0377 -0.0288
(0.0547) (0.0495)
Excluded ethnic group 0.8261 0.2653
(0.1094)a (0.1178)b
Polity2 -0.1332
(0.0110)a
GDP growth 0.0047
(0.0028)c
Constant -13.9745 -9.9041
(3.6814)a (3.2686)a
Observations 5,528 5,528
R-squared 0.340 0.399
F-Test 39.16 23.54
Year FE Yes Yes
Ethnic Group--Country FE Yes Yes
Notes: This table presents the first stage regressions results
concerning the IV estimations in column 5 and 6 of for estimations
in Table II-2. Robust standard errors clustered at “Ethnic group
and country” level reported in parentheses. (a) (b) (c) denote
significance at the 1, 5 and 10 percent level respectively. We
present the F-tests statistics on instruments; they are always
above the critical value of 10, recommended by Staiger and Stock
(1997).

85
 Chapter 2 – Beyond the Income Effect of Trade Integration on Ethnic Wars

II.4.4 Appendix D: Robustness check analyses of the impact of

Regional trade with imports

To make our analysis comparable with Gleditsch (2007), in Table II-8, we build our
indicator of regional trade on importations (instead of export as in the text). In Column 1
and 2 we compute an measure of the regional integration by summing all the import
from countries that share at least one ethnic groups with the country i studied. In
Column 3 and 4 we use the variable of Gleditsch that sums all the import from
neighboring countries. The same positive effect than the one presented in the text is
obtained. The difference between our result and the negative sign found by the seminal
paper of Gleditsch (2007) may come from the additional controls introduced here.

86
 Chapter 2 – Beyond the Income Effect of Trade Integration on Ethnic Wars

Table II-8: Regional Trade effect on ethnic wars onset

(1) (2) (3) (4)

Regional trade Regional trade
Importations Importation: Gleditsch
IV IV IV IV
Regional Trade 0.0169 0.0255 0.0332 0.0792
(0.0060)a (0.0090)a (0.0121)a (0.0287)a
Rest of Trade -0.0133 -0.0190 -0.0177 -0.0384
(0.0047)a (0.0062)a (0.0070)b (0.0142)a
Income -0.0090 -0.0094 -0.0130 -0.0194
(0.0030)a (0.0031)a (0.0041)a (0.0062)a
Excluded ethnic groups -0.0042 0.0058 -0.0110 0.0101
(0.0042) (0.0034)c (0.0066)c (0.0041)b
Polity2 0.0036 0.0104
(0.0013)a (0.0038)a
GDP growth 0.0009 0.0011
(0.0003)a (0.0004)a
Constant 0.0360 0.0035
(0.0239) (0.0190)
Observations 5,469 5,469 5,386 5,386
R-squared 0.017 0.020 0.022 0.023
Year FE Yes Yes Yes Yes
Ethnic Group-Country FE Yes Yes Yes Yes
Notes: The dependent variable is a binary variable taking 1 when an ethnic group is engaged in a new
ethnic war in a given year. All explanatory variables are lagged by one year. Regional trade in column
(1) and (2) designs imports from all the countries with which country 𝑖 shares an ethnic group. The
endogeneity on this variable is tackled using regional remoteness on imports partners with which
country i shares an ethnic group. Rest of Trade in column (1) and (2) captures imports from all the
other countries with which country 𝑖 does not share an ethnic group. Regional trade in column (3) and
(4) designs imports from all the countries with which country 𝑖 shares a border. The endogeneity on
this variable is tacled using regional remoteness on imports partners with which country 𝑖 shares a
border. Rest of Trade in column (3) and (4) captures imports from all the other countries with which
country 𝑖 does not share a border. Robust standard errors clustered at “Ethnic group and country” level
reported in parentheses. (a) (b) (c) denote significance at the 1, 5 and 10 percent level respectively.

87
 Chapter 2 – Beyond the Income Effect of Trade Integration on Ethnic Wars

II.4.5 Appendix E: First-stage IV estimates, explaining countries’

local trade

The following table presents the first stage regressions used in Table II-2, column (5)
and (6). It provides sufficient evidence that precipitations drives internal trade in
accordance with our expectations.

Table II-9: IV first stage regressions: explaining local trade

Local trade
(1) (2)
Country Precipitation 0.0077 0.0075
(0.0023)a (0.0023)a
Local trade( lag3 ) 0.5875 0.5864
(0.0236)a (0.0240)a
Ethnic group Income 0.0299 0.0309
(0.0111)a (0.0111)a
Excluded ethnic group 0.1000 0.1061
(0.0420)b (0.0412)b
Polity2 0.0018
(0.0022)
GDP growth -0.0000
(0.0009)
Constant 9.4759 9.5047
(0.5453)a (0.5543)a
Observations 4,734 4,734
R-squared 0.802 0.802
F-Test 392.47 370.06
Year FE Yes Yes
Ethnic Group--Country FE Yes Yes
Notes: This table presents the first stage regressions results
concerning the IV estimations in column 3 and 4 of Table II-2. Robust
standard errors clustered at “Ethnic group and country” level
reported in parentheses. (a), (b), (c) denote significance at the 1, 5 and
10 percent level respectively. We present the F-tests statistics on
instruments; they are always above the critical value of 10,
recommended by Staiger and Stock (1997).

88
 Chapter 2 – Beyond the Income Effect of Trade Integration on Ethnic Wars

II.4.6 Appendix F: First-stage IV estimates

The following table presents the first stage regressions used in Table II-4. The
remoteness indicator, the precipitations and the regional remoteness of countries
appear to be strong instruments respectively in their corresponding first stage IV
regression. It is worth noting that, the instruments used in this two-stage methodology
are the same as in the first section, particularly for countries internal and regional trade,
because they are same variables throughout the paper.

Table II-10: IV first stage regressions explaining country total Export and local trade in
the Two-stage methodology

(1) (2) (3)

International Internal Regional
trade trade trade
Remoteness -2.5770
(0.5973)a
Regional Remoteness -1.2461
(0.3558)a
Country level Precipitations 0.1293 0.0055
(0.0516)b (0.0026)b
Country level Precipitations square -0.0001
(0.0007)
Local trade( lag3 ) 0.9975
(0.0111)a
Rest of Trade 0.7169
(0.1962)a
Polity2 0.0015 0.0007 0.0888
(0.0534) (0.0020) (0.0536)
GDP growth 0.0064 0.0021 -0.0212
(0.0130) (0.0027) (0.0182)
Density -0.0028
(0.0088)
Constant -51.5302 0.0364 -22.8002
(12.9587)a (0.2768) (6.4555)a
Observations 580 505 489
R-squared 0.277 0.977 0.435
F-Test 13.67 2848.43 12.27
Year FE Yes Yes Yes
Notes: This table presents the first stage regressions results concerning the IV
estimations in column 3, 6 and 9 respectively in Table II-4. Robust standard errors
clustered at “country” level reported in parentheses. (a), (b), (c) denote significance at the
1, 5 and 10 percent level respectively. We present the F-tests statistics on instruments;
they are always above the critical value of 10, recommended by Staiger and Stock
(1997).

89
 Chapter 3 – Climate Change, Water Treaties and International Trade

Chapter 3 -
Climate Change, Water
Treaties and International
Trade

90
 Chapter 3 – Climate Change, Water Treaties and International Trade

III. Chapter 3 - Climate Change, Water

Treaties and International Trade

III.1 Introduction

More than 40% of the world population lives at close proximity to a fresh water
resource. The number of rivers that cross and/or separate countries are numerous:
more than 151 countries share transboundary river basins which are often vital for
agricultural production. During the past century, pollution, population and economic
growth have affected water resources80 and in the next century climate change will have
dramatic consequences with a significant reduction in precipitation and major changes
in respect of wet/dry seasons. This will affect the production and trade of agricultural
and industrial goods characterized by a high water footprint (e.g. a single cotton T-shirt
requires 2,500 liters of freshwater81 ). Consequently, climate change can be a catalyst for
institutional arrangements between nations aiming to manage water resources. In
particular, international watercourses offer a unique opportunity to fathom the effect of
climate change on WTs over time since more than 600 freshwater-related treaties have
been signed to manage transboundary basins82.

Surprisingly few studies have been undertaken to try to explain the rationale behind the
implementation of WTs. To the best of our knowledge, only Tir and Ackerman (2009)
explain how political institutions (preponderant power distribution, democratic
governance), development and water scarcity increase the likelihood of international
river cooperation between contiguous riparian states. In contrast to this study, we focus
on the effects of climate change by considering exogenous variables such as
temperature, drought, precipitation, floods and storms. Taking into account these
different variables is important in order to identify which kind of climate change fosters

80 See for instance Duarte, Pinilla and Serrano (2014) who analyzes the water footprint of the Spanish
agricultural sector over a long period of time (1860-2010).
81 See Debaere (2014)
82 According to the International Freshwater Treaties Database

91
 Chapter 3 – Climate Change, Water Treaties and International Trade

the development of WTs. We find that only temperature change significantly explains
the development of WTs. Precipitation, floods and storms have no effect.

Beyond the link between climate change and WTs, we aim to understand the impact of
these bilateral treaties on international trade. While it is a common knowledge among
economists that international trade can be a way to cope with the impact of climate
change by enabling water-poor countries to import water-intensive goods in order to
alleviate the pressure of scarcity (see Candau, Regnacq and Schlick, 2019), it is not clear
how current policies such as WTs affect international exchanges. Moreover, in the
absence of any studies, opposing arguments about the effects of such treaties are equally
credible.

First, these treaties may have little or no effect because they engender too small
additional costs on production to influence international trade. This argument has been
used to question the validity of the pollution haven hypothesis. For instance,
Smarzynska Javorcik and Wei (2003) argue that environmental norms’ costs are too
small in comparison to other costs to influence the relocation of polluting firms to
countries with laxer standards83. Since many WTs include environmental standards, the
same kind of argument about the insignificant impact of these policies can be applied.
The absence of impact can also come from the fact that these treaties are “cheap talk”
and are not a real commitment to implement significant protectionist policies on water.

It is also possible that these treaties contribute to resolve/prevent political conflict by

fostering cooperation between riparian nations84 and as a result promote institutions
and economic conditions thereby increasing trade. This second hypothesis, that WTs
foster international trade may be defended by also considering the multilateral effects of
these treaties. Indeed, a treaty on water protection between two countries can enable
downstream countries to use more water (or water of better quality), fostering
agricultural production and export85. Finally, some treaties in respect of large dams, by

83 Candau and Dienesch (2017) debunk this latter argument.

84 Hensel and Brochman (2008) for instance show that the presence of a WT decreases the potential for
militarization and Zawahri and Gerlak (2009) discuss how securing access to freshwater has fostered
cooperation in Africa. However as shown by Dinar (2009) even such a relationship is not sure, this author
indeed develops a theoretical model (also tested on seventy-four country dyadic observations) and
suggests an inverted U-shaped curve for the relationship between water scarcity and cooperation.
85 Agreements on water are often bilateral, according to the Atlas of International Freshwater Agreements,

two-thirds of basins have three or more riparian states, but only 20 percent of the agreements signed are
multilateral.

92
 Chapter 3 – Climate Change, Water Treaties and International Trade

producing hydroelectricity, may stimulate the production of goods and thereby

international trade.

Third, WTs may have a negative trade effect if they are implemented in order to save
freshwater ecosystems and/or to foster a sustainable development86. For instance the
United Nations Convention (on Non-Navigational Uses of International Watercourses) is
concerned with “equitable and reasonable use” and “the obligation not to cause
significant harm” to other States in the same basin. The ASEAN Agreement on the
Conservation of Nature and Natural Resources recommends “to promote
environmentally sound agricultural practice by, inter alia, controlling the application of
pesticides, fertilizers and other chemical” and “to promote pollution control and the
development of environmentally sound industrial processes and products”. The
Convention on Environmental Impact Assessment in a Transboundary Context
(Appendix I) provides a list of activities that are forbidden or regulated. These three
examples are just illustrative of the increasing number of WTs that may negatively
influence the export of agricultural and industrial goods. Of course, this may be mere
rhetoric i.e. de jure concerns about water issues, but if WTs are de facto applied with a
stringent implementation they have the potential to imply a loss of competitiveness.

Our methodology is based on a long tradition in the extant literature that analyzes the
effects of regional trade agreements (RTAs) on trade. Using gravity equations, it has
been found that RTAs have been successful in promoting trade growth in Europe
(Mayer, Vicard, Zignago, 2019), North America (Caliendo and Parro, 2015) and Africa
(Candau, Guepie and Schlick, 2018). Our analysis is linked to this literature since recent
trade agreements go beyond tariffs and include provisions on capital mobility,
competition, regulation of the labor market and environmental policies that are often
related to water issues. The effect of environmental provisions on international trade
has not yet been studied87 but many works have analyzed these so-called “deep RTAs” in
general and have found ambiguous results. Mattoo et al. (2017) find that deep
agreements lead to more trade creation and less trade diversion than shallow
86 In case of emission leakages and transboundary spillovers, WTs may be a tool to coordinate and to
legitimize green lobbying on environmental policy outcomes. See Conconi (2003) and Mason, Umanskaya
and Barbier (2018) who analyze the conditions under which lobby influence the environmental policy
including water regulation.
87 Baghdadi, Zarzoso and Zitouna (2013) found that environmental provisions of RTAs reduce polluting

emissions. This is an important result but regarding our concern it is a result on which we can only
speculate about the related effect on international trade.

93
 Chapter 3 – Climate Change, Water Treaties and International Trade

agreements88. Kohl, Brakman and Garretsen (2016) find instead that agreements
including domestic regulatory policies reduce trade.

Since RTAs are very heterogeneous in their contents, a consensus is hard to reach. By
focusing on WTs, our aim is to reduce this heterogeneity in order to provide a clearer
analysis. Indeed the definition of WTs is very concrete. First, all of the international
freshwater treaties that are included in the database must address freshwater as a
consumable resources in order to be included. Second, WTs address non-navigational
issues such as water management, flood control, hydropower projects, or allocations for
consumptive or nonconsumptive uses in international basins. Third, these treaties
should state in some way how these items are addressed by the signatories to be
included in the database. Documents dealing primarily with navigation, fishing, or
border issues are excluded and in a similar way document that are too flaw about the
water services are also excluded.

Our results confirm the protectionist role of WTs, we find that these treaties cause a
significant reduction in international trade between countries in almost all continents.
We compare this result to the effect of RTAs with water content and find that these
provisions have no effect on bilateral trade. This might be an interesting results for
environmental economics and management: water protection appears to be more
effective when enforced in the framework of a WT than when included in a RTAs.
Putting differently while is often defended that in front to climate change and to achieve
development goals, countries need to retool trade agreement in order to include social
and environmental issues (e.g. Shaffer and Fleurbaey, 2018), the current thesis based on
water, defends instead the enforcement of treaties on specific topics.

Pursuing our analysis at the continental scale, we find that WTs reduce trade in Asia and
in America and foster international trade in Africa. The content of the WTs explains this
outcome as African treaties on freshwater are less related to environmental issues and
often include joint management and technical cooperation provisions, which are helpful
in stimulating international trade. Cooperation on watercourses stimulates further
cooperation and the growth of the agricultural sector which is the main African export
engine.

88 See Bagwell, Bown and Staiger (2016) for a review.

94
 Chapter 3 – Climate Change, Water Treaties and International Trade

Using these results, we undertake a counterfactual analysis consisting of suppressing all

the WTs implemented to date. We find that the enforcement of a WT has a strong
negative impact on international trade in all countries except African countries. The
effects are the strongest in Asia due to the high elasticity of trade there. At country level,
countries along the Nile and the Niger are the main winners of freshwater treaties in
Africa.

The remainder of the paper is organized as follows. In Section II, WTs are briefly
outlined. In Section III, the data and the empirical strategy deployed are discussed as
well as the main results of the gravity structural estimation. Section IV presents the
counterfactual analysis and the final section outlines the study's conclusion.

III.1 About water treaties

Negotiations and treaties in respect of international rivers have a long history. Focusing
on the 20th century, some basic rules were established by the Institute of International
Law (IIL) in 1911, among which was the rule prohibiting unilateral and detrimental
exploitation of international basins and the requirement to develop transboundary
cooperation. In 1966, the International Law Association advocated an “equitable
utilization” of shared watercourses and proposed best practices in respect of water
management (public participation, flexible allocation methods, conflict resolution). In
1997, the United Nations General Assembly adopted the United Nations Convention on
the Law of the Non Navigational Uses of International Watercourses. This convention
followed the UNECE Water Convention89 adopted in Helsinki in 1992 and amended in
2003 in order to allow accession by countries outside the UNECE region. These two
global conventions promote the development of institutions that foster sustainable
development.

However, these conventions have only been ratified by a limited number of countries.
Europe is the most advanced continent since the majority of basin countries have
ratified the UNECE Water Convention with the aim of facilitating cooperation in order to

89More precisely called the Convention on the Protection and Use of Transboundary Watercourses and
International Lakes.

95
 Chapter 3 – Climate Change, Water Treaties and International Trade

ensure the sustainable use of transboundary water resources. Countries in North

America have also signed a significant number of treaties with a notable legal principles
component in the governance architecture of basins. In Southern Africa, an important
Water Protocol has been signed by members of the Southern Africa Development
Community, and finally many African countries have signed the African convention on
the conservation of nature and natural resources that includes provision on water. In
Asia, treaties are less global and more regional/bilateral, such as the agreement between
Kazakhstan and China, or the agreement between China and Russia on management and
protection of Transboundary Rivers.

In brief, international or regional WTs have a very heterogeneous content. They cover,
inter alia, water quality, irrigation, hydropower, infrastructure (barrages, canals, dykes)
and under-groundwater90. Some treaties contains technical/financial cooperation
and/or monitoring provisions (e.g. mechanisms to monitor treaty compliance), and
define the compensation schemes applicable in exchange for water rights or access. The
objectives of WTs also vary from one treaty to another. There are a) regulatory treaties
that set rules to prescribe certain actions (e.g. water allocation) or to proscribe others
(e.g. pollution emissions), b) programmatic treaties in respect of a common program
(e.g. dam construction) or,c) more general treaties that set out principles and norms for
cooperation in transboundary basins.

Past analysis in respect of the Transboundary Freshwater Dispute Database (TFDD)

reveals that most treaties address hydroelectric and water supplies, rather than water
quality issues (Hamner and Wolf, 1998; Beach et al., 2000)91. On recent data, we find a
different picture. Figure 7 classifies the treaties according to the four different treaty
subject matters: the environment, water quality, water quantity and hydro-
power/electricity generation. We find that during the 1990s, an increasing number of
treaties relating to the environment and to water quality were signed. The number of
treaties relating to dams, however, in that period had clearly reached a plateau.

90 Only three transboundary aquifers have an operational arrangement for water cooperation: the
Genovese aquifer, the Nubian Sandstone aquifer, the Northwestern Sahara aquifer system.
91 These analyses also show that bilateral treaties outnumber multilateral treaties which may explain the

difficulty to manage the resource without a systemic agreement.

96
 Chapter 3 – Climate Change, Water Treaties and International Trade

Figure 7: Water Treaties Contents

Source: Authors.

Depending on the level of country development, the incentive to sign a WT, as well as
the nature of the treaty, can vary. These treaties are certainly influenced by the level of
specialization in the industrial sector (e.g. polluting firms) and in the agricultural sector.
Indeed 70% of the fresh-water used in this last sector is an intermediate input of the
agricultural sector. Consequently, depending of the climate and of the agricultural
technology available, the need to enforce WTs is going to be more or less pressing.

We test this hypothesis which has previously only been partially addressed by Tir and
Ackerman (2009). These authors presented a Cox regression of river treaties taking into
account institutional variables (democracy, power distribution) and the availability of
water but with strong assumptions that cast doubts on their results in the light of
climate change. For instance, due to data limitations, they consider that the total amount
of water available per country is invariant over time. Data on temperature, drought,
precipitation, floods and storms are not taken into account. To date, we have no clue of
what kind of climate change motivates countries to cooperate and it seems thus
important to consider these variables. To do so, consider that a water agreement is

97
 Chapter 3 – Climate Change, Water Treaties and International Trade

signed when the utility gain, 𝛤odt , is higher that the negotiation cost, 𝐶odt for a country-
pair (𝑜, 𝑑) at year 𝑡 such as:

𝛤odt > 𝐶odt (27)

by assuming with Martin, Mayer and Thoenig (2012) that 𝐶odt is the unobserved
component of the decision process submitted to stochastic shocks, then the inequality
(27) is transformed into a probability of water agreement formation. We now have to
choose the distribution of 𝐶odt that determines the functional form taken by this
probability. Assuming a Type I extreme value distribution concerning 𝐶odt (see Train,
2003), we get the functional form which is a logit probability:

𝑃(𝑊𝑇odt = 1) = exp(𝛤
exp(𝛤odt ) (28)
odt )+1

where the gains depends here on climatic conditions:

𝛤odt = 𝛼𝐶𝑙𝑖𝑚𝑜𝑡 + 𝛽𝐶𝑙𝑖𝑚𝑑𝑡 + 𝐶𝑜𝑑 + 𝑓𝑜 + 𝑓𝑑 + 𝑓𝑡 (29)

where 𝑊𝑇odt is a dichotomous indicator of the existence of a water agreement between

country o and country d at time t ; it takes one when the two countries have signed a
new agreement in a given year and zero otherwise92. This dummy is built from the
International Freshwater Treaty Database93 provided by the “Program in Water Conflict
Management and Transformation” (College of Earth, Ocean, and Atmospheric Sciences94
). This database gathered a full number of international, freshwater-related agreements
between pair or groups of countries. These treaties concern “water as a scarce or
consumable resource, a quantity to be managed, or an ecosystem to be improved or
maintained [...] water rights, water allocations, water pollution, principles for equitably
addressing water needs, hydropower/reservoir/flood control development, and
environmental issues and the rights of riverine ecological systems”. Agreements
concerning fishing rights, navigation rights and tariffs, or delineation of rivers as
borders are mentioned in some treaties.

92 A full explanation on the construction of the variable W𝑊𝑇odt is provided in the “Data and estimators”
subsection.
93 https://transboundarywaters.science.oregonstate.edu/content/international-freshwater-treaties-
database.
94 Oregon State University.

98
 Chapter 3 – Climate Change, Water Treaties and International Trade

𝐶𝑙𝑖𝑚𝑜𝑡 and 𝐶𝑙𝑖𝑚𝑑𝑡 are a set of climatic variables consisting in temperature, precipitation,
floods, storms and droughts. Data on temperature and precipitations come from Dell et
al. (2014) and data on “Floods, Storms & Droughts” come from the Centre for Research
on the Epidemiology of Disasters95. 𝑓𝑜 (𝑓𝑑 ) and 𝑓𝑡 design respectively countries fixed
effects and year specific effects. These fixed effects takes into account the technology
used in the agricultural and/or the availability of water in each country. 𝐶𝑜𝑑 is a set of
control variables that allow to account for time-invariant ties within each pair of
countries such as colonial relations, same language, same border.

95 Available on request at https://www.emdat.be/

99
 Chapter 3 – Climate Change, Water Treaties and International Trade

Table III-1: Climate reasons of water agreements: gravity estimation approach

(1) (2) (3) (4)
Logit Probit LPM LPM
Origin Temperature 6.869 2.551 0.357 0.365
(2.394)a (0.763)a (0.043)a (0.043)a
Origin Precipitation -0.183 -0.027 -0.011 -0.014
(0.475) (0.171) (0.012) (0.012)
Origin Drought 0.334 0.148 0.008 0.006
(0.179)c (0.070)b (0.011) (0.011)
Origin Flood -0.015 0.003 -0.014 -0.015
(0.174) (0.066) (0.007)c (0.008)b
Origin Storm -0.769 -0.271 -0.004 -0.005
(0.305)b (0.108)b (0.008) (0.008)
Destination Temperature 8.211 2.873 0.368 0.377
(2.584)a (0.803)a (0.043)a (0.043)a
Destination Precipitation -0.149 -0.017 -0.011 -0.014
(0.481) (0.171) (0.012) (0.012)
Destination Drought 0.327 0.148 0.006 0.004
(0.181)c (0.070)b (0.011) (0.011)
Destination Flood -0.004 0.009 -0.011 -0.012
(0.173) (0.066) (0.008) (0.008)
Destination Storm -0.770 -0.263 -0.005 -0.005
(0.305)b (0.108)b (0.008) (0.008)
log Distance -1.809 -0.780 -0.084
(0.095)a (0.042)a (0.009)a
Contiguity 0.348 0.293 1.005
(0.149)b (0.057)a (0.092)a
Comcol 1.309 0.604 0.017
(0.188)a (0.073)a (0.021)
Smctry -0.476 -0.175 0.716
(0.200)b (0.085)b (0.129)a
Constant -43.010 15,597 -1.159 -2.084
(7.710)a (2.651)a (0.191)a (0.219)
Observations 190,739 190,739 832,333 832,333
Pseudo R-squared 0.319 0.319 0.014 0.003
Origin-Destination FE No No No Yes
Origin FE Yes Yes Yes Yes
Destination FE Yes Yes Yes Yes
Year FE Yes Yes Yes Yes
Robust standard errors clustered at the “country pair” level reported in parentheses. (a), (b),
(c) denote significance at the 1, 5 and 10 percent level respectively.
In column (3) & (4), all coefficients are multiplied by 10 for readability and presentation
purposes.

100
 Chapter 3 – Climate Change, Water Treaties and International Trade

Table III-1 presents results from the estimation of Equation (29)). We also estimate this
relationship using a Probit in Column 2 and a Linear Probability Model (LPM) in Column
3 and 4.

The first interesting result of Column 1-3 is the difference between climatic conditions
that are rarely significant and traditional bilateral variables such as distance that
strongly explain the enforcement of water treaties. However, while precipitation,
drought and flood do not motivate a political action toward WTs, temperatures are
significant both at the origin and at destination. Temperature is the most and rapidly
perceived indicator of climate change raising awareness on global warming and
explaining the enforcement of water treaties. It is interesting to observe that this result
holds despite the fact that we control for country heterogeneity and proximity96.

Since the Logit (and Probit) model does not support well the high number of fixed
effects, we use in Column 4 the Linear Probability Model (LPM) with origin-destination
fixed effects (instead of traditional bilateral variables). This last estimation confirms that
countries with higher temperatures are more likely to sign a WT.

III.2 Estimating Elasticity

III.3 The Structural Gravity Equation

To analyze the effect of WT on trade we use a simple structural gravity equation. Instead
of presenting the theoretical model in detail (postponed to Section 4), we start by this
gravity equation that explains bilateral flows between country 𝑜 and country 𝑑, denoted
𝑋𝑜𝑑 , such as:

𝑌𝑜 𝑌𝑑 (30)
𝑋odt = 𝑜𝑑
1−𝜎
𝑜 𝑃𝑑1−𝜎

where σ is the elasticity of substitution between varieties ( 𝜎 > 1 ), 𝑜𝑑 a reversed

measure of trade costs 𝜏𝑜𝑑 (𝑜𝑑 = 𝜏𝑜1−𝜎 i.e an indicator of trade openness) between 𝑜
and 𝑑 , 𝑌𝑑 and 𝑌𝑜 the aggregated expenditures/incomes at the destination of exports 𝑑

96This finding is verified in the Online Appendix A where a robustness check is done by considering the
choice to sign an agreement country by country (and not on a bilateral basis).

101
 Chapter 3 – Climate Change, Water Treaties and International Trade

and at origin 𝑜. 1−𝜎

𝑜 represents the market potential in 𝑜. This term is sometimes
considered as an indicator of the market access from 𝑜 and/or called outward
multilateral resistance because it represents a GDP share weighted measure of trade
cost resistance that exporters in 𝑜 face when shipping their goods to consumers on their
own and outward markets. The term P𝑑1−𝜎 in this gravity equation (30) is the
accessibility-weighted sum of exporters 𝑜 capabilities also called inward multilateral
resistance since it is a reversed measure of the openness of a nation to import from the
world. Anderson and Yotov (2010) also consider this term as the buyer's incidence
because it represents the weighted sum of trade costs paid by buyers.

This gravity equation is estimated using the pseudo-maximum likelihood (PML)

estimator as follows:

𝑋odt = exp(𝛼 + 𝑓𝑜𝑡 + 𝑓𝑑𝑡 + 𝑜𝑑𝑡 + 𝜖𝑜𝑑𝑡 ) (31)

where 𝑓𝑜𝑡 and 𝑓𝑑𝑡 are time-varying countries-specific effects approximating exporting
1−𝜎𝑠 1−𝜎𝑠
and importing capacity, 𝑌𝑜,𝑠 /𝑜,𝑠 and 𝐸𝑜,𝑠 /𝑃𝑜,𝑠 in Equation (30) at time 𝑡, 𝛼 is a
constant.

To control for other bilateral relationships, binary bilateral variables are used i.e.
dummies for colonial links, borders, common language and physical distance. Since this
strategy raises doubts regarding the possibility of omitted variables, we compare results
with estimations including bilateral fixed effects 𝑓𝑜𝑑 to control for all unobserved time-
unvarying bilateral determinants of exports (Baier and Bergstrand, 2007; Magee, 2008).

Consequently trade costs in (31) take the following form:

odt = 𝜓𝑅𝑇𝐴𝑜𝑑𝑡 + 𝑊𝑇𝑜𝑑𝑡 + 𝑓𝑜𝑑 (32)

where 𝑅𝑇𝐴𝑜𝑑𝑡 takes 1 at the year 𝑡 when a regional agreement between these countries
enters into force and zero otherwise. In a similar way, 𝑊𝑇𝑜𝑑𝑡 (for Water Treaty) takes 1
after the signature of a water agreement. As the data source provides only the signing
date of the treaties documents, we assume that, once adopted, the agreement is valid till
the end of the period considered (i.e. 2007). This is obviously a strong assumption; we
thus also lead estimation by considering that 𝑊𝑇𝑜𝑑𝑡 takes 1 only during the first five
years after the signature. We also consider a dummy 𝑊𝑇𝑜𝑑𝑡 for a period of ten years.
102
 Chapter 3 – Climate Change, Water Treaties and International Trade

III.4 Data and estimators

The Water Agreement variable (𝑊𝑇𝑜𝑑𝑡 ) is a dummy indicator built from the
International Freshwater Treaty Database97 provided by the “Program in Water Conflict
Management and Transformation” (College of Earth, Ocean, and Atmospheric
Sciences98). This database, which spans the years 1820 to 2007, gathered a full number
of international, freshwater-related agreements between pair or groups of countries. To
build our dummy of WTs, we extract from the International Freshwater Treaty
Database, all the international water treaties since 1920 and countries that are involved
in each of them. The groups that signed the treaties are comprised of two countries in
case of bilateral agreements and more than two in case of multilateral agreements.
Afterwards, we generate, for each treaty, a set of the bilateral combinations of countries
that comprised the signing group. Consequently, the number of bilateral combinations
per treaty per year is given by 𝑛! = 𝑛 × ( 𝑛 − 1 ) × ( 𝑛 − 2 ) × . . .× 1, with 𝑛 ≥ 2; 𝑛
designing the number of countries involved in the treaty. The constructed variable takes
one in years when the agreement is signed and zero otherwise. It still takes one, even in
years a country had signed more than one agreements that have the same issue. We
firstly generate the water agreement's onsets per issue before gathering them within a
general water agreement whatever the issue. Since, we only have the signing date of the
agreements, we also provide on top of the general water agreement onset dummy, two
other indicators based on the assumptions that a treaty is effective respectively during 5
years and 10 years.

The virtual water associated with exports of agriculture products is built by combining
statistics from Water footprint (hereafter WFP) databases99 and the bilateral exports of
goods classified as agriculture products in BACI. The virtual water designs the quantity
of water involves in the production process of a traded goods. These statistics are
proposed in cubic meters of water per ton of production. The Water footprint network
uses to produce information about the amount of water used in the production process

97 https://transboundarywaters.science.oregonstate.edu/content/international-freshwater-treaties-
database
98 Oregon State University
99 https://waterfootprint.org/en/resources/waterstat/product-water-footprint-statistics/

103
 Chapter 3 – Climate Change, Water Treaties and International Trade

of goods and services. It tells not only about the volume of water consumed in the
production of goods such as wheat, rice, maize, cotton, but it informs also about more
than one process products. The WFP distinguishes three main components of water
used in the process of production of products, depending on their sources. (i) The green
component of water footprint consists of water from precipitation that is stored in the
root zone of the soil and evaporated, transpired or incorporated by plants. (ii) The blue
part of water footprint designs that from surface or groundwater resources and is either
evaporated, incorporated into a product or taken from one body of water and returned to
another, or returned at a different time. (iii) Finally, the Grey portion of water footprint
indicates the volume of fresh water required to assimilate pollutants to meet specific water
quality standards. With these informations at hand, we firstly make a correspondence
between agriculture products (comprising crops and animal) from WFP database and
those of BACI. We then compute, for each agriculture goods the green, blue and grey
component of their virtual water. Finally, we sum up these three values for every
agriculture product, and we aggregate all the agriculture good's virtual water to get the
agriculture exports virtual water variable.

The Regional Trade Agreement (𝑅𝑇𝐴𝑜𝑑𝑡 ) variable comes from Bergstrand's

homepage100.

Data on trade are from TRADHIST (CEPII101) which provides a complete representation
of bilateral trade flows between countries all around the world. Specific bilateral
characteristics come from GEODIST (CEPII) and are coded one (zero otherwise) when
country-pairs are contiguous (𝑐𝑜𝑛𝑡𝑖𝑔𝑖𝑗 ), when one country of the pair was the colonizer
of the other (𝑐𝑜𝑙𝑖𝑗 ), whether the two countries were formerly parts of the same country
(𝑠𝑚𝑐𝑡𝑟𝑦𝑖𝑗 ), and when at least 9% of the population in both countries speak the same
language (𝑙𝑎𝑛𝑔𝑖𝑗 ). The Gross domestic Product (GDP) data is also extracted from CEPII.

Our database covers the period 1955-2007 and concerns 154 countries.

Equation (31) is estimated with the Poisson Pseudo-Maximum-Likelihood (PPML)

estimator routinely used in the trade literature in order to handle zero trade flows and

100 https://www3.nd.edu/~jbergstr/
101 Centre d’études prospectives et d’informations internationales.

104
 Chapter 3 – Climate Change, Water Treaties and International Trade

heteroscedasticity102. To manage the large number of FE (exporter/importer-year fixed

effects alongside with country pair fixed effects), we rely on a method developed by
Correia et al. (2019)103.

III.5 Baseline results

Table III-2 presents the findings obtained from Equation (31). In Column 1, where an
exhaustive set of country-year varying characteristics and country-pair FE are
introduced, we find that water agreements unambiguously cause a reduction of bilateral
trade104.

In Column 2, we modify our dummy of WT that now takes one during the first five years
after which the agreement has been signed, and in Column 3, the dummy takes one
during ten years. In each regression we confirm the previous negative impact of these
treaties.

To be sure that the results obtained so far are not the result of an aggregation bias and
to test whether the sign of the WTs elasticity changes with the sector considered and/or
with the water content of international trade we lead two additional investigations. In
Column 4 we consider only trade in agricultural goods and in Column 5 the dependent
variable is the “virtual water” of agricultural trade. As already emphasized, the
agricultural sector is by far the highest consumer of water (around 70% of the total
consumption in the world according to the FAO) and it is thus interesting to take a
particular attention to this sector.

In each case, we still find that WTs reduce trade exchanges and without surprise the
strongest impact is observed in the agricultural sector.

102 See Silva Santos and Tenreyro (2006) for the seminal article and Silva Santos and Tenreyro (2011) who
show that the Poisson pseudo maximum likelihood estimator performed better than other estimators
even when the conditional variance is far from being proportional to the conditional mean and even when
the dependent variable has a very large proportion of zeros. Finally Head and Mayer (2015) compare
several estimators and favor the PPML estimator. Interestingly Fally (2015, Proposition 1) demonstrates
that the estimated fixed effects with PPML are perfectly consistent with the multilateral resistances of the
theoretical model.
103 https://arxiv.org/abs/1903.01690
104 The same regression has been done with a temporal trend and is not reported here because exactly the

same results have been obtained.

105
 Chapter 3 – Climate Change, Water Treaties and International Trade

In the Online Appendix B we analyze the potential endogeneity from the simultaneity
between trade and water agreements by using the test proposed by Wooldridge (2002)
and used by Baier and Bergstrand (2007) concerning RTAs. The rationale behind this
test is that there is no anticipation when the future level of WTs (𝑊𝑇𝑖𝑗,𝑡+1 ) is
uncorrelated with the current trade flows. We find a strict exogeneity of water treaties
(i.e. a non-significant effect of 𝑊𝑇𝑖𝑗,𝑡+1 on trade, see Table III-7, highlighting that there is
no obvious reverse causality.

Table III-2: Baseline results

(1) (2) (3) (4) (5)
Agri Virtual
dependent var => Total trade Agri
Water
Water Agreements -0.132 -0.069 -0.087 -0.250 -0.177
(0.048)a (0.034)b (0.028)a (0.056)a (0.086)b
RTAs 0.191 0.190 0.186 -0.014 0.560
(0.030)a (0.030)a (0.030)a (0.038) (0.167)a
Constant 22.143 22.144 22.147 12.749
(0.013)a (0.013)a (0.013)a (0.019)a
Observations 829,491 829,491 829,491 165,512 84,51
R-squared 0.990 0.990 0.990 0.979 0.994
Country pair FE Yes Yes Yes Yes Yes
Origin*Year
Yes Yes Yes Yes Yes
+ Dest *Year FE
Robust standard errors clustered at the “country pair” level reported in parentheses. (a), (b), (c)
denote significance at the 1, 5 and 10 percent level respectively.
In Col 1, the dummy WT takes one for the year when a water treaty is signed, in Col 2 WT takes
one during the first five years after the signature and in Col 3 during the ten years that follow.

III.6 Heterogeneity analysis

We aim to test whether the effect of water agreements varies across space. Depending
on the level of development (and of institutions) it is indeed possible that some WT are
de jure and not de facto, i.e. not applied in certain regions of the world, while in other
regions these agreements may be more effective impacting on trade and welfare. Due to
the fact that some rivers flow across continents, we estimate Equation (31) on a
continental basis. Such a choice is motivated by the localized characteristic of the
resource which cross countries at the continental level. We follow the Water Conflict
Management and Transformation program at the Oregon State University, which

106
 Chapter 3 – Climate Change, Water Treaties and International Trade

considers four continents: Africa, America, Asia, Europe. We also lead robustness check
by separating North America and the Central and South America, the results obtained
are similar.

In all the Columns of Table III-3, we use the set of fixed and time varying effects
presented previously as well as RTAs in order to be sure that the coefficient of WTs does
not take into account other confounding factors105.

Coefficients are negative and significant in Europe, in America106 and in Asia. The
coefficient seems particularly strong is Asia (-0,509) and such a strong result may come
from the lack of control on the dimension 𝑖𝑗𝑡, indeed at the exception of RTAs, we have
no other variable that varies over space and time. We thus introduce a variable counting
the number of General Agreement on Tariffs and Trade/World Trade Organization
(GATT/WTO) members in the country pair and a dummy taking one in the case of a
bilateral conflict and zero otherwise107108. With these additional controls the effect of
WTs are almost unchanged in all continents at the exception of Asia where we get a
more realistic effect of −0.302.

105 RTAs are introduced systematically but not reported in tables to save space (results are conventional
and similar of what has been obtained in the previous section).
106 To test the heterogeneity inside the continent, we also create a dummy considering North America

(Canada, USA, Mexico) and the rest of the America. We find the same negative impact of WTs. The
coefficient is equal to -0.164 (RSE: 0.08) in North America and -0.933 (RSE:0.164) in Central and South
America.
107 The GATT/WTO membership comes from the CEPII database. Bilateral conflicts are militarized

Interstate Disputes (MID) from the Correlates of War (COW) project

(https://www.correlatesofwar.org/data-sets). This dummy takes the value one when MID in the original
database takes the values 3, 4 or 5, which correspond respectively to the (i) “display of force” (e.g. a
decision of mobilization or a border violation), (ii) “the use of force” (e.g. an attack or an occupation of
territory) and (iii) an “interstate war” (more than 1000 military deaths).
108 To save space the coefficient of these controls are not presented, we find that the GATT/WTO

membership significantly impact trade (0.406, RSE: 0.179) while bilateral conflicts are not significant.

107
 Chapter 3 – Climate Change, Water Treaties and International Trade

Table III-3: Heterogeneity analysis

(1) (2) (3) (4)

WT in Africa 0.118 0.127 0.204 0.319

(0.111) (0.108) (0.076)a (0.084)a
WT in Europe -0.099 -0.091 -0.018 -0.034
(0.056)c (0.061) (0.038) (0.029)
WT in Asia -0.509 -0.302 -0.477 -0.329
(0.080)a (0.152)a (0.098)a (0.077)a
WT in America -0.315 -0.312 -0.289 -0.297
(0.084)a (0.085)a (0.072)a (0.070)a
Constant 22.1 22.1 22.1 22.1
(0.015)a (0.015)a (0.015)a (0.015)a
Observations 829491 829491 829491 829491
R-squared 0.990 0.990 0.990 0.990
Country pair FE Yes Yes Yes Yes
Origin*Year
Yes Yes Yes Yes
+ Desti*Year FE
Note: Robust standard errors clustered at the “country pair” level reported in parentheses. (a), (b),
(c) denote significance at the 1, 5 and 10 percent level respectively. In all column RTAs in Africa,
Asia, America and Europe have been introduced. In column 2, GATT/WTO membership and a
dummy for bilateral conflicts have been added.

In Column 3 and 4, we do not use these 𝑖𝑗𝑡 controls but we introduce the dummy taking
1 respectively during the five and ten years that a WT has been signed. All the previous
results are verified at the exception of WT in Africa that are now significant and positive.
This positive effect may be explained by the fact that water agreements have been
particularly effective in Africa to foster cooperation between countries. For instance
Zawahri and Gerlak (2009) describe the growing creation of commissions between
riparian states to promote shared water management in the areas of flood control,
technical assistance, and infrastructure development in the Congo, Zambezi, and
Orange/Senqu River basins. These negotiations as well as the institutionalization of
relationship may have been a vector to develop regional trade.

In the Online Appendix C, we treat another source of heterogeneity that concerns the
content of WTs. Indeed as already discussed in the first section of this paper, some
treaties are signed to promote water quality while others have a very different aims
such as the production of hydroelectricity. We find that treaties with a content related to
water quality reduces trade in Asia, while restriction on water quantity reduces trade in

108
 Chapter 3 – Climate Change, Water Treaties and International Trade

America. Treaties on economic and sustainable development also affect negatively

bilateral export. More surprisingly treaties on hydropower have also a negative impact
on bilateral trade109. By destabilizing the society, these large projects may have a
detrimental effect on international trade and on the capacity to produce and to export
(at least on the short run that we study). It is also possible, that such a desegregation of
the data without enough observations for each WTs contents provides a biased picture
that should be taken with caution.

III.7 Water treaties versus Regional Trade Agreements on Water

Beside these conventions on water, the other kind of international treaties not included
in our database are RTAs with environmental provisions related to water quality and/or
water resources. Protection of the water resources with regulation against water
pollution are indeed often clearly mentioned110, in particular concerning the
management of transboundary river basins. For instance, the Bulgaria-EC, art. 81(2)
establishes that “Cooperation shall concern: - combating local, regional and
transboundary air and water pollution- water quality, particularly of transboundary
waterways (including the Danube and of the Black Sea)”. In a similar way the COMESA
aims to “take measures to control trans-boundary, air and water pollution arising from
mining, fishing and agricultural activities.” The East African Community, art. 111(2)
stipulates that “action by the Community relating to the environment shall have the
following objectives: to ensure sustainable utilization of natural resources like lakes.”
These quotations come from the codebook of the database TREND (TRade &
ENvironment Database) that identifies 300 different types of environmental provisions
in 730 trade agreements (Morin, Dür and Lechner, 2019) 111 . We use this dataset to

109 This result was not expected since a strong correlation between energy production and economic
growth reported in the literature logically invites us to consider hydropower as a positive determinant of
bilateral trade. Such a negative result might be explained by taking into account the fact that some
hydropower projects are not inclusive, in particular in developing countries where there have sometimes
been financed without taking into account the lives and production of local communities.
110 Sometimes (albeit rarely) it is also stipulated that water is a “sensitive subject” that cannot be

liberalized by the RTAs. For instance in the CETA, art. X.08(1) it is written that “water in its natural state,
such as water in lakes, rivers, reservoirs, aquifers and water basins, is not a good or a product and
therefore, except for Chapter XX – Trade and Environment and Chapter XX – Sustainable Development, is
not subject to the terms of this Agreement.”
111 http://www.chaire-epi.ulaval.ca/en/trend

109
 Chapter 3 – Climate Change, Water Treaties and International Trade

create a dummy of all water agreements that deal with water issues. First, we retain
RTAs that include statements on water pollution concerns and/or rivers, basins and
lakes management. Secondly, since the TREND database only provides the names of the
trade agreements, we match the list of members (countries) that are involved in these
RTAs. Finally, we follow the same method presented earlier for our Water Agreements
variable (see section 3.2 Data and estimators) to propose a new indicator of RTAs which
deals specifically with water management (called “RTAs Water” in the different table of
results).

In reason of the potential collinearities between RTAs, RTAs with water provision and
WTs we introduce these variables successively in the estimation of the gravity equation
(31). In Column 1 of Table III-4 only our new dummy of RTAs with a content concerning
water resource is introduced, the coefficient is significantly positive, albeit relatively
small. In Column 2, the dummy of WTs is added without affecting the previous results,
by fostering trade RTAs with water provision have the opposite effect than WTs. This
result is counter-intuitive and not in line with discussions on the effect of “green” RTAs.
Indeed several analysts consider that these environmental clauses are often ambitious
and conclude that these kind of RTAs are laboratories where negotiators experiment
and promote new provisions that are often effectively implemented (Baldwin and Low
2009; Morin, Brandi and Berger, 2019). Conducing a survey of government officials,
trade negotiators and other experts, Clive and Yamaguchi (2018) report that according
to respondents “these RTAs were a significant factor in the improvement of wastewater
treatment systems [...] and have encouraged exporters to develop environmental
certification to comply with standards and hence avoid risk to reputation and potential
liability damages”. This survey is however based on a very limited number of
respondents, but it indicates that a negative sign is expected and not the positive one
reported here. It is likely that the positive effect obtained in our analysis comes from the
lack of control, representing more the positive effect of RTAs in general than the positive
effect of RTAs with water provisions. In Column 3 we thus control for RTAs and find that
RTAs with water provision are finally unsignificant. This confirms the skeptical analysis
of some researchers that view environmental provisions as “fig leafs” enabling to make
trade agreements more suitable for voters and their representatives (Berger, Brandi and
Bruhn, 2017). In Column 4, the coefficient of WTs is almost unaffected by the
introduction of RTAs dummies.
110
 Chapter 3 – Climate Change, Water Treaties and International Trade

Table III-4: Heterogeneity analysis

(1) (2) (3) (4)

Export flows
RTAs Water 0.032 0.032 0.020 0.020
(0.016)a (0.016)a (0.016) (0.016)
RTAs 0.190 0.190
(0.030)a (0.030)a
Water Treaties -0.148 -0.133
(0.050)a (0.048)a
Constant 22.2 22.2 22.1 22.1
(0.002)a (0.002)a (0.013)a (0.013)a
Observations 829491 829491 829491 829491
R-squared 0.990 0.990 0.990 0.990
Country pair FE Yes Yes Yes Yes
Origin*Year
Yes Yes Yes Yes
+ Desti*Year FE
Robust standard errors clustered at the “country pair” level reported in parentheses. (a), (b),
(c) denote significance at the 1, 5 and 10 percent level respectively.

III.8 General Equilibrium Analysis

The previous analysis provides the trade elasticity of WT but by controlling for
multilateral resistance and other bilateral or individual characteristics this estimation
gives only the partial effect of WT. For instance an agreement between two countries
can have a negative impact of their trade, but can also have positive (or negative) effects
on trade with other partners, which is typically captured by individual fixed effects. In a
similar way, WT can have income effects by affecting revenues in sectors where water is
an important input. To analyze the general equilibrium effects of WT, we need to present
in detail the model from which the structural gravity Equation (30) comes from. To
study how WT have globally affected international trade, we lead a counterfactual
analysis consisting in suppressing all WTs. Our analysis is thus an ex-post exercise in the
same vein of Caliendo and Parro (2015) and Mayer, Vicard and Zignago (2018) that
consist to quantify what would be the costs of un-doing what have been achieved. Such
an exercise seems interesting in a period where trade integration is put into question in
many part of the world and when at the same time climate change calls for action in
particular regarding water. We use the theoretical model of Arkolakis, Costinot and

111
 Chapter 3 – Climate Change, Water Treaties and International Trade

Rodriguez-Clare (2012) in its most simple form for reasons of transparency: a) the
estimation of trade friction obtained in the previous section can be used directly without
additional data such as tariffs and input-output data that are often lacking (in particular
for African countries) and/or based on many assumptions, b) the model based on
perfect competition, imposes the minimal assumptions and provides results that are
easy to interpret.

III.8.1 Methodology and data

To assess the general equilibrium effect of water treaties, we use the theoretical model
of Arkolakis, Costinot and Rodriguez-Clare (2012) where the real market potential of
exporters in the structural gravity Equation (30) is defined by:

𝑛
(33)
1−𝜎
𝑜 = ∑(𝜏𝑜𝑑 ⁄𝑃𝑑 )1−𝜎 𝑌𝑑
𝑑=1

while the price index of the consumption basket in the destination country is:

𝑛
(34)
𝑃𝑑1−𝜎 = ∑(𝜏𝑜𝑑 ⁄𝑜 )1−𝜎 𝑌𝑜
𝑜=1

Considering a Log-differentiation of the gravity Equation (30) we present, hereafter and

step by step, the impact of a change in trade costs due to WT.

By focusing our analysis on a change of 𝑜𝑑 in the numerator of (30), from 𝑜𝑑 to 𝑐𝑜𝑑 we
obtain the direct effect of trade costs. The upper-script 𝑐 is used to characterize the
counterfactual experiment. This direct effect takes the following simple form:

𝑐𝑜𝑑 (35)
𝐷𝑖𝑟𝑒𝑐𝑡𝑜𝑑 = 𝑜𝑑 = = exp[𝜓(𝑊𝑇(1)𝑜𝑑 − 𝑊𝑇(0)𝑜𝑑 )]
𝑜𝑑

where WT(0) means no treaty on water and WT(1) the enforcement of a treaty. The
“dot” is used in this paper to represent the proportional change in a variable between its

112
 Chapter 3 – Climate Change, Water Treaties and International Trade

initial value and the counterfactual scenario112. As shown in this equation (35), the
direct effect does not take into account price indices.

Now adding in this analysis how multilateral resistances vary after regional trade
liberalization gives the Price Index Effect of WTs113:

𝑜 𝑃𝑑 (36)
𝑃𝐼𝐸𝑜𝑑 = exp[𝜓(𝑊𝑇(1)𝑜𝑑 − 𝑊𝑇(0)𝑜𝑑 )]
𝑐𝑜 𝑃𝑑𝑐

To compute this, we first use the estimate of 𝜓 obtained from Equation (32). Then we
use the result of the gravity regression to measure trade costs 𝑜𝑑 , i.e. 𝑜𝑑 ≡

𝑒𝑥𝑝[𝜓̂𝑊𝑇𝑜𝑑𝑡 ]. Using this measure of 𝑜𝑑 with expenditures 𝑌𝑜 and 𝑌𝑑 in Equation (33)
and (34) with the contraction mapping of Head and Mayer (2015) gives the multilateral
resistances 𝛱𝑜 and 𝑃𝑑 . Then using the previous measure of 𝑜𝑑 , we get from Equation

(35) the counterfactual trade costs 𝑐𝑜𝑑 , i.e. 𝑐𝑜𝑑 ≡ 𝑜𝑑 exp[𝜓̂(𝑊𝑇(1)𝑜𝑑 − 𝑊𝑇(0)𝑜𝑑 )].
Using again the contraction mapping with 𝑐𝑜𝑑 and with the same expenditures 𝑌𝑜 and 𝑌𝑑
provides the counterfactual multilateral resistances 𝑐𝑜 and 𝑃𝑑𝑐 . All these findings give
the PIE of water treaties presented in Equation (36). This equation takes into account
multilateral resistance, i.e. any change between a pair of countries gives counterfactual
changes in trade flows for all country pairs that are taken into account in prices indices.
Then trade diversion, which is a reduction of trade flows from outside the block after an
agreement is taken into account.

The computation of the PIE is based on the assumption that WTs do not affect incomes.
However it is obviously possible that WTs affect incomes and wages in sectors that use
water in an intensive way. Taking into account this change, the General Equilibrium
Trade Effect (GETI), is defined as follows:

𝑌𝑜′ 𝑋𝑑′ 𝑜 𝑃𝑑 (37)

𝑃𝐼𝐸𝑜𝑑 = exp[𝜓(𝑊𝑇(1)𝑜𝑑 − 𝑊𝑇(0)𝑜𝑑 )]
𝑌𝑜 𝑋𝑑𝑐 𝑐𝑜 𝑃𝑑𝑐

112 The literature usually work with a “hat”, a notation here preserved to notify the predicted value of
coefficients.
113 Head and Mayer (2015) call this effect the Modular Trade Impact in reference to Anderson (2011). We

prefer the term Price Index Effect which may be more telling.

113
 Chapter 3 – Climate Change, Water Treaties and International Trade

Where 𝑌𝑜′ and 𝑋𝑑′ denote respectively the production at the origin and the expenditure at
destination after a change due to WTs.

The GETI allows quantifying the fact that a treaty affects both signatory and non-
signatory members by diverting trade from partners outside the block (Viner, 1950).
This diversionary effect turns out to be very concrete in the case of water basins/rivers
that are shared by many countries. An agreement or a disagreement between two
upstream countries can have either a beneficial or deleterious effect on third countries
depending on how the resource is managed upstream. The potential for trade diversion
is substantial since the number of shared basins/rivers is large114.

Considering the production side with labor as the sole factor of production in each
country 𝑖 = (𝑜, 𝑑), 𝑌𝑖 = 𝑤𝑖 𝐿𝑖 , and by assuming that the change in the labor force is
constant, gives 𝑤̇ = 𝑌̇. Since trade deficit are constant, change in expenditures equals
change in incomes (indeed with 𝑋𝑑 = 𝑤𝑑 𝐿𝑑 (1 + 𝑑𝑑 )) where 𝑑𝑑 is the deficit of country
𝑑, gives 𝑋̇𝑑 = 𝑤̇𝑑 = 𝑌𝑑̇ ). To determine the equilibrium change in income we use the
share of expenditure of consumers in 𝑜 spent on goods produced in 𝑑, 𝜋𝑜𝑑 = 𝑋𝑜𝑑 ⁄𝑋𝑜 .
Finally, the change in expenditure due to a trade shock is given by:

̇ 𝑜𝑑 𝑌𝑜̇ 1−𝜎 (38)

𝜋̇ 𝑜𝑑 =
∑𝑙 𝜋𝑙𝑑 ̇ 𝑙𝑑 𝑌𝑙̇ 1−𝜎

Inserting this expression in the market clearing enables to solve the system and to get
the income change due to the removal of a WT:

1
𝑛
𝜋 ̇ 𝑌̇1−𝜎 (39)
̇𝑌𝑑 = ∑ 𝑜𝑑 𝑜𝑑 𝑜 𝑌𝑜̇ 𝑋𝑜
𝑌𝑑 ∑ 𝜋  𝑌
̇ ̇ 1−𝜎
𝑜=1 𝑙 𝑙𝑑 𝑙𝑑 𝑙

Using the direct effects calculated earlier, with 𝑌𝑜 approximated by GDPs, and the trade
share 𝜋𝑜𝑑 of each country 𝑜, gives from (39) a system of equations defining 𝑌𝑜̇ , which

114 Nine different countries share the Congo, Niger, Nile, Rhine, and Zambezi, and at least five different
nations are concerned by the Amazon, Aral Sea, Ganges Brahmaputra-Meghna, Jordan, Kura-Araks, La
Plata, Lake Chad, Mekong, Neman, Tarim, Tigris Euphrates-Shatt al Arab, Vistula, and Volga basins (Wolf et
al., 1999).

114
 Chapter 3 – Climate Change, Water Treaties and International Trade

once inserting in the trade share expenditure (38)115, gives the General Trade
Equilibrium Impact (GETI) of WT: 𝜋̇ 𝑜𝑑 𝑌𝑑̇ .

The general equilibrium analysis is computed on a database that comprises 154

countries. Since we need a square dataset for the analysis, the countries that constitute
the sample are those which allow to get not only a balanced panel database but also
have information on trade with self (internal trade flows). These internal flows that
depict the expenditure made to get goods produce domestically, allow calculating
multilateral resistances after a change in trade costs. To approximate these internal
flows that are generally defined by country production minus it's exports, we rely on the
Input-Output Tables proposed by EORA Database. We consider the year 2000 to assess
the impact of a change in water agreements and we take Finland as a reference. Our
results are independent of this choice (however to avoid any doubt about that, we have
checked that changing reference country does not modify our results; these results are
available on request).

III.8.2 Results

The experiment in Table III-5 is to turn off dummies of WTs in order to calculate the
counterfactual trade flows for all pairs. Column 1 reports the trade elasticity estimated
previously, Column 2 the direct effect of WT. We lead the experiment separately for each
continent, considering that WTs agreements are suppressed in one place only. For
instance the first line in Table III-5 gives the amount of trade created of WTs in Africa in a
world where Asia, America and Europe already have their current WTs when the direct
effect (Column 2), the PIE (Column 3), and the GETI (Column 4) are taken into account.
Line 2 considers the trade creation due to WTs in Europe ceteris paribus in particular
when WTs in Africa, Asia, and Europe are not modified. The PIE and GETI are the median
values of the real/counterfactual trade ratio for countries relevant in the experiment.
These calculations are computed for members and non-members for each variable in the
following separated columns of Table III-5.

115To resolve the system we need an estimate of the constant elasticity of substitution between variety,
we use 𝜎 = 4.03 which is the number obtained in the meta-analysis of Head and Mayer (2014).

115
 Chapter 3 – Climate Change, Water Treaties and International Trade

Table III-5: Direct, PIE, GETI effects of water agreements

Coeff PTI MTI GETI

Members Yes Yes Yes No Yes No
Water agreement * Africa 0.118 1.125 1.009 1 1.025 0.997
Water agreement * Europe -0.099 0.906 0.995 1.003 0.992 1.008
Water agreement * Asia -0.509 0.601 0.962 1.017 1.037 1.163
Water agreement * Americas -0.315 0.73 0.986 1.015 0.919 0.963
Notes: The MTI, GETI are the median values of the real/counterfactual trade ratio for countries relevant
in the experiment.

The direct impact of WTs is an increase in the delivered price of goods exported in
Europe, in America and in Asia which lead to reduction of trade there. The effect is the
strongest in Asia due to the high trade elasticity of WT. However while the direct effect
leads to a strong reduction, the fact to take into account the price index effect gives a less
radical result (trade change by 4% (=1-0.962)). Such a result comes from the fact that
with WTs, members suffers a loss of competitiveness and then the market access for
partners outside the bloc is easier, which counteract partially the direct negative effect.

The opposite results are obtained in Africa, where the delivered price of exported goods
decreases without WTs. The 'third country effect' however tends to minimize this result.
While water agreements have a high direct positive effect (+12%), they also divert trade
from partners outside the bloc, the PIE leads to an increase of trade around 1%.

Considering now the following column, it is first noteworthy that there is a small
difference between the PIE and the GETI, which illustrates that the shock implemented
is a small change in trade costs with weak effect on incomes. This is particularly true in
Europe where the amount of trade is almost unaffected by the GETI. However in Africa,
the improvement in income due to WTs leads to an increase of international trade from
to 1% to 2.5%. In Appendix C we lead some sensitivity analysis by doubling (and
dividing by two) the current trade elasticities in order to analyze how results are
affected. We find that while significant changes occur at the level of the PTI, the final
picture (once multilateral effects are taken into account) is very similar to the one taken
here.

116
 Chapter 3 – Climate Change, Water Treaties and International Trade

Figure 8: The Impact of No Water Treaties on Trade (GETI)

Source: Authors

117
 Chapter 3 – Climate Change, Water Treaties and International Trade

We then analyze the past shocks at the national level and report the result of WTs on
trade with GETI on Map in Figure 8. Countries sharing Nile and Niger are countries that
lose the most from a removal of WTs. In contrast the multilateral effect of WTs in Asia is
strong, hence a withdrawal of these agreement leads to the strongest increase in trade
and income in China, Pakistan, Thailand, Vietnam and Cambodia. We lack data to lead
this counterfactual analysis for small countries like Bhutan and Laos, but this would
certainly reinforce the result obtained here. For instance in Bhutan, hydropower
accounts for 32% of the country's total exports and 8% of its gross domestic product in
2018. The detrimental effect of no WTs is smaller in America for the simple reason that
basins are smaller and less shared than in Asia.

III.9 Conclusion

The history of international water treaties dates as far back as 2500 BC and it is even
stated that the first treaty in the history of mankind was a water treaty116. But
surprisingly these WTs have not been analyzed in details by economists, or more
precisely only the link to water conflicts has been studied. While this analysis is crucial,
it is now time to deepen our understanding of these treaties. First, we aim to know what
are the exogenous factors (i.e. beyond conflicts), that motivate the enforcement of these
treaties. We find that many climatic variables have a weak influence at the exception of
temperatures which strongly (and in a very robust way) explain WTs. In the context of
climate change and of a hotter earth, such a result indicates that many more WTs are
going to be enforced in the future.

Second, if the literature suspect that WTs can resolve conflicts, it is surprising to not
analyze how such a strong effect can influence other variables. To the best of our
knowledge, the current analysis is the first assessment of the consequences of
transboundary freshwater treaties on trade. There is various reason to consider that
WTs influence international trade going from a positive effect along a pacification of
economical and political relationship and a better sharing of a common resource, to a
negative effect linked to the fact that these treaties may impose new environmental
regulation on producers and then of the cost of production limiting the capacity to

116 see https://www.un.org/waterforlifedecade/transboundary_waters.shtml

118
 Chapter 3 – Climate Change, Water Treaties and International Trade

export. To analyze this issue, we use the most advanced tools in international trade
allowing to control for omitted variables, such as comparative advantages and specific
characteristics of countries that vary over time (importers and exporters’ time effects)
or bilateral relationships such as the presence of shared basin/rivers, common language
or political ties (by bilateral fixed effects and specific dummies such as RTAs).

We find that WTs signed in Asia and America have a significant negative effect on the
international trade of members of these treaties. The content of treaties explains this
outcome, since many WTs have been implemented to foster the sustainable
development of resources. By contrast in Africa, WTs have boosted international trade,
perhaps by being part of conflicts’ resolution processes and/or by fostering agricultural
development. Understanding what distinguishes African countries and the treaties
implemented there from what happens in the rest of the world in greater detail is
beyond the scope of this paper but offers an intriguing area for further research. We also
find that WTs have a significant effect on trade while RTAs that include provisions on
water are not significant. This is an important result in term of international
management; it is indeed often argued that trade agreements should be modified to take
into account environmental and social issues. From the point of view of the current
analysis on water, it would be more efficient to directly negotiate treaties on these topics
instead of adding provisions to existing RTAs that are rarely enforced.

Our contribution is mainly positive by analyzing the impact of WTs on welfare in a very
simple trade model. This work can be pursued by using a more sophisticated model (i.e.
a multi-sector models, monopolistic competition, variable markups) and by
investigating the welfare effect of these water treaties. This last exercise is a promising
challenge, requiring to analyze how water services enter in the utility function and to
estimate (or to calibrate) the preference parameter in order to assess whether WTs can
be simultaneously trade reducing and welfare improving.

119
 Chapter 3 – Climate Change, Water Treaties and International Trade

III.10 Appendices

III.10.1 Appendix A: Climate reasons of water agreements,

unilateral estimation approach

In this appendix we analyze how results regarding the formation of water agreements
are affected by deviating from the bilateral analysis and by analyzing the incentive to
sign an agreement in 𝑜 regarding climatic conditions in this country:

𝑊𝑇𝑜𝑡 = 𝛼𝐶𝑙𝑖𝑚𝑜𝑡 + 𝑓𝑜 + 𝑜𝑑𝑡 + 𝜖𝑜𝑡 (40)

where 𝑊𝑇𝑜𝑡 is the Water Agreement variable and 𝐶𝑙𝑖𝑚𝑜𝑡 the vector of climatic
conditions described in the text.

Table III-6 provides the results of this estimation using first a Logit model. For
robustness check, we also build a dependent variable that tells about the total number of
water agreements signed by a country during a given year. Since this latter appears to be
a count variable, we rely on appropriate estimation techniques such as Poisson
estimator, Negative Binomial regression and Zero Inflated Poisson regressor, that help
account for this kind of data. In Column 1 we introduce temperature alone in order to
avoid problems of multicollinearity. We find that hotter temperatures explain the
enforcement of water treaties. Not reported here the same estimations has been done by
inserting in a similar way precipitation, floods, storms and droughts separately and in
each cases these variables were not significant. This is verified in Column 2-5 where
temperature is the sole variable that significantly explains water treaties.

120
 Chapter 3 – Climate Change, Water Treaties and International Trade

Table III-6: Climate reasons of water agreements

(1) (2) (3) (4) (5)

Water agreements
Temperature 0.639 0.633 1.319 1.011 0.854
(0.176)a (0.178)a (0.203)a (0.217)a (0.167)a
Precipitation 0.022 -0.095 -0.039 0.042 c
(0.054) (0.072) (0.065) (0.022)c
Drought 0.266 0.107 0.372 0.189
(0.306) (0.386) (0.361) (0.356)
Flood -0.114 -0.050 -0.228 -0.073
(0.242) (0.286) (0.286) (0.325)
Storm -0.201 -0.486 -0.193 -0.611
(0.323) (0.314) (0.366) (0.412)
Constant -17.07 -17.128 -36.053 -31.144 -26.362
(3.597)a (3.630)a (2.575)a (0.101)a (2.089)a
Observations 3024 3024 7828 7828 7828
R-squared 0.135 0.136 0.274
Country FE Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes
Country fixed effects and year fixed effects are account for in all columns. The first and second columns
results are based on logit estimations. The third, the fourth and fifth columns results come respectively
from “Poisson estimator”, “Negative binomial regression” and a “Zero inflated Poisson regressor”.
a: p<0.01, b: p<0.05, c: p<0.1

We thus find here the same result than what has been reported with the bilateral
equation. More precisely temperature is the critical variable to sign an agreement.

121
 Chapter 3 – Climate Change, Water Treaties and International Trade

III.10.2 Appendix B: Reverse causality test

In the table below, we analyze the reliability of our estimations with regard to reverse
causality issues. As proposed by Wooldridge (2002), we check the possibility that trade
and water agreements variables can be co-determinate by testing if there is an
anticipation behavior in trade when a future treaty is going to be enforced (𝑊𝑇𝑜𝑑,𝑡+1 ).
The strict exogeneity test consists in the absence of potential feedback effect. We then
augment our preferred results in Table III-2 with future level of water treaties
(𝑊𝑇𝑜𝑑,𝑡+1).

In column (1), we firstly rely on a parsimonious specification to test the significativity of

the coefficient on 𝑊𝑇𝑜𝑑,𝑡+1. In columns (2) and (3), we move to more complete
specifications that include all our main rights hand variables and the result on 𝑊𝑇𝑜𝑑,𝑡+1
always remains unchanged. More concretely, the results in Table III-7 show a non-
significant effect of 𝑊𝑇𝑜𝑑,𝑡+1 confirming that there is no feedback effect from trade to
water treaties.

Table III-7: Strict exogeneity test of water treaties

(1) (2) (3)

Export flows
Water Agreements -0.128 a
(0.048)
Water Agreements (t+1) -0.101 -0.078 -0.083
(0.063) (0.061) (0.062)
RTAs 0.213 a 0.213 a
(0.030) (0.030)
Constant 22.175 a 22.085 a 22.086 a
(0.000) (0.013) (0.013)
Observations 804,766 804,766 804,766
R-squared 0.990 0.990 0.990
Country pair FE Yes Yes Yes
Origin *Year + Destin*Year FE Yes Yes Yes
Robust standard errors clustered at the “country pair” level reported in
parentheses. (a), (b), (c) denote significance at the 1, 5 and 10 percent level
respectively.

122
 Chapter 3 – Climate Change, Water Treaties and International Trade

III.10.3 Appendix C: Sensitivity of results to changes in

trade elasticities

Many estimators have been used in the literature on the trade gravity equation before to
reach the current consensus concerning the PPML estimator. Consequently, these
analyses provide a neat picture on how our results could be affected with another
estimator. Using the Tobit/Heckman estimator provide higher coefficients of trade
agreement than the PPML estimator while the Zero Inflated Poisson (ZIP) ML estimator
gives a similar coefficient (Martin and Pham (2015)). In a similar way, comparing the
PPML estimator with the OLS estimator (truncated and not), the threshold Tobit of
Eaton and Tamura (1994), the poisson and gamma pseudo-maximum likelihood
estimator, Head and Mayer (2014) find that in each case the PPML yields smaller
estimates of the trade gains compared to these estimators. More precisely with PPML
the coefficient of RTAs is equal to 0.29 while the highest coefficient is obtained with the
LSDV (least squares with country dummies) and equals 0.63. To conclude, our results
based on PPML are conservative in the sense that other estimators would provide a
stronger effect of WTs on trade and then on welfare. Thus at the top of Table (Table
III-8) we test how results are affected by doubling all our elasticity by two. Results are
sensitive to this dramatic change, in particular the direct effect (PTI) is strong but once
general equilibrium effects are taken into account, the final picture in term of MTI and
GETI.

Table III-8: Doubling and halving trade elasticities

Coeff * 2 PTI MTI GETI

Yes Yes Yes No Yes No
Water agreement * Africa 100% 13% 0% 0% 3% 0%
Water agreement * Europa 100% -9% 0% 0% -1% 1%
Water agreement * Asia 100% -40% -4% 2% 5% 17%
Water agreement * Americas 100% -27% -1% 1% -9% -5%

Coeff / 2 PTI MTI GETI

Yes Yes Yes No Yes No
Water agreement * Africa -50% -6% 0% 0% -1% 0%
Water agreement * Europa -50% 5% 0% 0% 0% 0%
Water agreement * Asia -50% 29% 2% -1% -2% -8%
Water agreement * Americas -50% 17% 1% -1% 4% 2%
Source: Authors

123
 Chapter 3 – Climate Change, Water Treaties and International Trade

The elasticities obtained with PPML are smaller than those with other estimators but
they are however still high. WTs reduce trade by around 8% (𝑒 −0.087 − 1 ) in the general
case (coef from Table III-2, Col 3), but at the continental level the coefficient used in
simulations may be considered as too high to be realistic, in particular for Asia.
Robustness analysis, consisting to introduce new controls specific to pair of countries
and varying over time (level 𝑖𝑗𝑡) have indeed shown that these continental elasticities
can be divided by two. At the bottom of Table III-8 we thus divide all the previous
elasticities by two and analyse how the effects vary. The strongest effect is obtained in
Asia, where with elasticity twice smaller, the change in welfare is only 5% higher.

124
 General conclusion

General conclusion

There have been growing concerns about the links between the increasing importance
of trade and development issues. In this thesis, we provided new data and rely on
advance technical tools to provide some answers to three important interrogations in
the fields of international economics. The importance of the studies stems not only from
the fact that they address topical issues, but also because they lead to specific economic
developments that should help countries to re-think about some measures of public
interest. The findings and conclusions presented are mainly drawn from empirical
analyses of the underlying questions.

Foremost, we give new insight into the effect of trade on the spatial distribution of
population within countries. By analyzing the population growth at the top of the urban
hierarchy, we test two hypotheses explaining the rise of mega-cities: trade and political
institutions. We find that democratic institutions are the main factor behind the
concentration of a nation’s urban population in the main city. Contrary to the literature,
our results reveal that extractive institutions reduce the size of the biggest city. This
finding should alert the public administrations in these countries to not fail to develop
measures and initiatives that should limit urban agglomeration while institutions are
being improved.

In the second chapter, we question the implications of international, regional and

domestic trade, in explaining ethnic disputes by controlling for income effects using
satellite data on night lights. As a result, we find that domestic and international trade
integrations have peaceful effects while regional trade integration fosters civil conflicts.
For the sake of enhancing the existing methodology and to comprehend if country
patterns that drive the probability of ethnic war are influenced by trade, we also present
a two-step approach. In the first stage, country time-varying effects that explain
conflicts, purged of income effects, are estimated and are interpreted as a proxy of
changes in national identities. In a second step we analyze how trade integration affects
these national identities. The latter step reveals that only international trade integration
builds a new national identity that reduces the likelihood of ethnic wars. Since, African

125
 General conclusion

countries are still looking for solutions in undermining the insecurity on the continent;
we consider that this chapter can help to shed lights on a certain number of aspects.
Thus, by relying on some arguments and empirical evidences, our findings reveal that
trade integration in the world market can be also part of the solutions. Pan-African
organizations and African governments needs to consider how best to exploit this
channel to reduce particularly ethnic conflict around the continent.

Finally, we analyze the reaction of international trade to international water treaties.

Before reaching that analysis, this chapter aims at highlighting how climate change has
affected the enforcement of Water Treaties (WTs). Among different climate indicators,
only temperature has fostered the formation of international cooperation on water
resources. Besides, we conduct empirical analysis on the basis of the most recent
advancement in the structural gravity model in the literature on international trade and
find that freshwater-related treaties have caused a decrease in bilateral trade in Europe,
Asia and America but have fostered trade in Africa. In general, WTs that protect the
water resources reduce trade. Results differ in Africa where WTs are less related to
environmental issues and include joint management and technical cooperation
measures which stimulate international trade. We also go through a general equilibrium
model that allows us to provide the global effect of WTs and we find that they have a
significant effect on trade, in particular in Asia. The first major conclusion that can be
drawn from this section is that water agreements have a real effect even if this impact
differs from one continent to another. The second conclusion is that it is more effective
to implement water agreements especially instead of RTAs with provisions that concern
the management of transboundary basins if policies want to be really effective in
handling international water issues.

Even though this thesis seems to be an elaborated peace of work, it is not immune to
criticism. Besides, the realization of this work has generated new ideas that deserve
consideration for future works. In particular, as shown in the first chapter of this Ph.D.
thesis, countries past and contemporaneous institutional patterns are the main causes of
the urban giants' growth. In sum, the study concludes that democracy goes hand in hand
with agglomeration in past colonial countries. We can suggest five (5) extensions or
developments that can be made to this research. Firstly, in order to shed light on the
reasons behind our findings, we could extend the present framework by accounting for

126
 General conclusion

the following question: is it the attractiveness of these biggest cities or their

demography patterns that drive their population growth in case where countries are
improving their institutions? Secondly, a related extension would be to analyze the case
where our dependent variable is an economic growth of primate cities. Thirdly, from a
methodological point of view, the great challenge of the empirical analysis in this
chapter has been to find time-varying historical instruments. We overcome this problem
by proposing year-by-year regressions whose predicted values allowed us to generate
country-year instruments. One would like to go a step further to provide statistical
validity of this approach, even if we are confident of the reliability of our approach,
because the robustness check with a semi-parametric method confirms our findings. In
addition to this issue, it would also have been preferable to deal with the endogeneity of
the rural-urban income gap by providing exogenous instruments that may determine its
discretionary variation. Fourthly, the further step that would certainly be promising for
this literature will be to move to microeconomics analysis to capture the market access
variable. We have indeed gone one step further from the existing literature by relying on
a market access variable that encompasses several indicators of trade costs. However,
the latter still did not account for domestic factors that are also part of trade costs, such
as the existence of infrastructure, and reflects poorly their quality, which should either
allow goods to be exported directly, or easily convey to the main export platform. Thus,
the microeconomic analysis should be based on a survey in the potential secondary
cities, and should allow for instance to identify the availability of facilities that can help
in transporting goods either to the main city or directly to the international market.
Information on transaction costs117 such as the distance of paved road to the urban
giants, the distance of paved road to the nearest border, and dummy indicators that
capture whether there exists exports' facilities or whether the region uses to produce
cash crops, etc. should be collected. This variable should be regressed on a variable
depicting the difference between the urban giants' growth and that of the secondary
cities. Finally, one could distinguish the analysis by accounting for the nature of goods
exchange. In fact, the picture will be more contrasted when it comes to separating the
manufacturing sector exports from the agriculture sector exports (that need space and
land), since the latter can keep the work force from converging to the main city.

117As argued in the economic literature (see for instance Behrens et al., 2018) transaction costs may
determine the spatial distribution on economic activities.

127
 General conclusion

As for the second chapter, one of its' limitations is the lack of data that can help evaluate
more precisely not only the internal or domestic trade, but also the ethnic network
regarding trade. Further research should be done to improve and provide more reliable
constructed indicators of trade that can help alleviate this limitation. Also, beyond
channels that bring from trade to war, income and institutions' quality are always
mentioned in the literature, but they have never really been tested. Thus, further
analyses should address this issue in order to give a solid empirical background to these
theoretical arguments. This chapter could also benefit from several extensions. Our
constructed income variable might be a rough description of ethnic wealth, since the
repartition of night lights within the ethnic areas should be very heterogeneous or
dispersed, particularly in rural areas. In these areas, the amount of light is often
concentrated in single places (for instance around marketplaces or administrative
centers; etc.). One could imagine that this may just signal the presence of the
government through a public lighting programs rather than the level of income in the
area. Besides, our argument which supported that trade with peers across the border
should be conflict promoting do not account for the total ethnic network. Trade within
two countries that share at least one ethnic group can also be drawn by ethnic migration
outside their country's homeland. The same picture should be observed when analyzing
internal trade effect on ethnic peace since internal trade can be foster by internal ethnic
group population's migration.

In the third chapter, we find that addressing water issues in the world generally leads to
a compulsion of international trade of members particularly in Europe, Asia and
America. By contrast, in Africa these water agreements have boosted international trade,
perhaps by being part of a conflict resolution process and/or by fostering agricultural
development. In the last part of this chapter, we provide in addition to the partial effect
of water treaties on trade, general equilibrium effects that account successively for the
price index effects, the multilateral resistance and changes in countries' incomes and
wages. But it worth noting that the model that helps to draw the general equilibrium
effects account only for labor as a factor of production. Then, it would likely for future
works to rely on more sophisticated model that accounts not only capital in the
production function, but also that would distinguish different sectors in the economy. It
will also be useful to undertake a welfare effect analysis in order to gauge if the loss in
trade is associated to other gains that are beneficial for the populations.
128
 References

References

[1] Abowd, J. M., F. Kramarz, and D. N. Margolis (1999). High wage workers and high
wage firms. Econometrica 67(2), 251–334.
[2] Acemoglu, D., Naidu, Restrepo and Robinson (2017), 'Democracy Does Cause
Growth', forthcoming Journal of Political Economy
[3] Acemoglu, D., Robinson, J. A. and T. Verdier, (2004), 'Kleptocracy and Divide-and-
Rule: A Model of Personal Rule', Journal of the European Economic Association,
2(2-3):162-192.
[4] Acemoglu, D., S. Johnson, and J A. Robinson (2002), 'Reversal of Fortune:
Geography and Institutions in the Making of the Modern World Income
Distribution'. The Quarterly Journal of Economics, 117(4): 1231-1294.
[5] Acemoglu, D., S. Johnson, and J A. Robinson (2012), 'The Colonial Origins of
Comparative Development: An Empirical Investigation: Reply'. American
Economic Review, 102(6): 3077–311
[6] Ades, A. and E. Glaeser (1995), 'Trade and Circuses: Explaining Urban Giants',
Quarterly Journal of Economics 110: 195-227.
[7] Albouy, D. (2012), 'The Colonial Origins of Comparative Development: An
Empirical Investigation: Comment', American Economic Review, 102 (6): 3059-
76.
[8] Anderson, J. E., van Wincoop, E., 2003. Gravity with Gravitas: A Solution to the
Border Puzzle. The American Economic Review 93 (1): 170–192.
[9] Anderson, J., 2011. The Gravity Model. The Annual Review of Economics 3 (1),
133–160.
[10] Anderson, J., Yotov, Y., 2010. The Changing Incidence of Geography.
American Economic Review 100, 2157–2186
[11] Angrist, J,D. and D. Kugler (2008), Rural windfall or a new resource curse?
coca, income and civil conflict in colombia. The Review of Economics and
Statistics, 90(2): 191-215.
[12] Anthony, R. M., (2014), 'Bringing up the past: Political experience and the
distribution of urban populations', Cities, Volume 37, Pages 33-46

129
 References

[13] Arkokalis, C., Costinot A., Rodriguez-Clare A., 2012. New Trade Models,
Same Old Gains ? The American Economic Review, 102 (1): 94-130.
[14] Arkolakis C, A Costinot, D Donaldson, Rodríguez-Clare A ., 2019. The
Elusive Pro-Competitive Effects of Trade. The Review of Economic Studies,
Volume 86, Issue 1, Pages 46–80.
[15] Arrow, K., 1972, Gifts and Exchanges, Philosophy and Public Affairs 1(4),
343-362.
[16] Baghdadi, Leila ; Martínez Zarzoso, Inmaculada ; Zitouna, Habib, Are RTA
agreements with environmental provisions reducing emissions? Journal of
International Economics Volume 90, Issue 2, 90.2: 378-390.
[17] Bai, C-H., C-T. Hsieh and Z., Song (2019), 'Special Deals with Chinese
Characteristics', forthcoming in NBER Macroeconomics Annual.
[18] Baier S. L., Bergstrand J. H., 2001. The growth of world trade: tariffs,
transport costs, and income similarity. Journal of International Economics, 53(1):
1-27.
[19] Baier, S. and Bergstrand, J. (2004), “Economic Determinants of Free Trade
Agreements”, Journal of International Economics, 64, 29–63.
[20] Baier, S. L., Bergstrand, J.H., 2007. Do Free Trade Agreements Actually
Increase Members' International Trade?” Journal of International Economics, 71
(1) : 72-95.
[21] Baldwin, R. and P. Low. 2009. Multilateralising Regionalism: Challenges for
the Global Trading System. Cambridge: Cambridge University Press.
[22] Baldwin, R., (2016), The Great Convergence, Information Technology and
the New Globalization. The Belknap Press of Harvard University Press.
[23] Bates, R. H, (1981), Markets and States in Tropical Africa: The political
basis of agricultural policies. Berkeley: University of California Press.
[24] Bazzi S, Blattman C. 2014. Economic shocks and conflict: evidence from
commodity prices. Am. Econ. J. Macroeconomics. 6(4):1–38.
[25] Beach, H., Hamner J., Hewitt J., Kaufman E., Kurki A., Oppenheimer J., Wolf
A., 2000. Transboundary Freshwater Dispute Resolution. Tokyo: United Nations
University Press.

130
 References

[26] Behrens, K., Brown, W. M., & Bougna, T. (2018). The World Is Not Yet Flat:
Transport Costs Matter!, The Review of Economics and Statistics, 100(4): 712-
724.
[27] Behrens, K, Mion, G, Murata, Y, Südekum, J, 2018. Quantifying the Gap
between Equilibrium and Optimum Under Monopolistic Competition, HSE
Working papers WP BRP 185/EC
[28] Behrens, K. and F. Robert‐Nicoud (2014), 'Survival of the Fittest in Cities:
Urbanisation and Inequality', The Economic Journal, 124(581): 1371-1400.
[29] Berger A., C. Brandi and D. Bruhn, 2017. Environmental Provisions in
Trade Agreements: Promises at the Trade and Environment Interface. German
Development Institute, Briefing Paper 16.
[30] Berman, N., and M. Couttenier. 2015. “External Shocks, Internal Shots: The
Geography of Civil Conflicts.” The Review of Economics and Statistics, vol. 97,
issue 4, 758-776
[31] Blattman, C., and E. Miguel (2010): “Civil War,” Journal of Economic
Literature, 48(1), 3-57.
[32] Bosker M and J de Ree (2014), Ethnicity and the spread of civil war.
Journal of Development Economics, vol 108(C), 206-221.
[33] Calì, M. (2015) Trading Away from Conflict Using Trade to Increase
Resilience in Fragile States. World Bank Report.
[34] Caliendo, L., Parro F., 2015. Estimates of the Trade and Welfare Effects of
NAFTA, The Review of Economic Studies, 82(1): 1-44.
[35] Calvo, T. and Mercier, M., 2019. Analyse économique des causes des
guerres civiles: un état des lieux. DIAL Working Paper 2019 / 11.
[36] Candau F., Jean S., 2009. What Are EU Trade Preferences Worth for Sub-
Saharan Africa and Other Developing Countries? in Trade Preference Erosion:
Measurement and Policy Response ed B. Hoekman, W. J. Martin et C. A. Primo
Braga. Palgrave MacMillan Issues.
[37] Candau, F. (2008), 'Good Governance, Trade and Agglomeration', Papers in
Regional Science, 87(4): 483-504.
[38] Candau, F. (2011), 'Is Agglomeration Desirable?', Annals of Economics and
Statistics, 101/102: 203- 228.

131
 References

[39] Candau, F. and E. Dienesch (2015), 'Spatial distribution of skills and

regional trade integration'. The Annals of Regional Science, 54(2): 451-480.
[40] Candau, F., E., Dienesch, 2017. "Pollution Haven and Corruption Paradise"
Journal of Environmental Economics and Management, Volume 85, pp 171-192.
[41] Candau, F., Guepie, G., Schlick, J., 2019. Moving to autarky, trade creation
and home market effect: an exhaustive analysis of regional trade agreements in
Africa. Applied Economics, 51(30): 3293-3309.
[42] Candau, F., Regnacq C., Schlick, J., 2019. Climate Change and the Revealed
Comparative Advantage of Water. F.R.E.I.T Working Paper 1602.
[43] Caplan, B, and T Cowen. 2004. Do We Underestimate the Benefits of
Cultural Competition? American Economic Review, 94(2): 402407.
[44] Chandler, T. (1987): Four Thousand Years of Urban Growth: An Historical
Census. Edwin Mellon Press, Lewiston, NY.
[45] Chaney, T., 2016, Liquidity constrained exporters. Journal of Economic
Dynamics & Control, vol. 72 , 141-154.
[46] Clive George and Shunta Yamaguchi, 2018. Assessing Implementation of
Environmental Provisions in Regional Trade Agreements. OECD Trade and
Environment Working Papers 2018/01
[47] Collier, P. and A. J. Venables (2013), 'Housing and Urbanization in Africa:
unleashing a formal market process', Centre for the Study of African Economies,
Working Paper Series 2013-01, University of Oxford.
[48] Collier, Paul, and Anke Hoeffler. 2004. “Greed and Grievance in Civil War.”
Oxford Economic Papers, 56(4): 563–95.
[49] Combes, P.-P., G. Duranton, and L. Gobillon (2008). Spatial wage
disparities: Sorting matters! Journal of Urban Economics 63(2), 723–742.
[50] Combes, PP., G. Duranton and L. Gobillon, 2011. 'The identification of
agglomeration economies', Journal of Economic Geography, 11(2): 253–266.
[51] Conconi, P., 2003. Green Lobbies and Transboundary Pollution in Large
Open Economies. Journal of International Economics 59: 399-422.
[52] Correia S., Guimarães P., Zylkin T., 2019. PPMLHDFE: Fast Poisson
Estimation with High-Dimensional Fixed Effects. [econ.EM :
https://arxiv.org/abs/1903.01690].

132
 References

[53] Costinot, A., and Rodríguez-Clare, A., 2014. Trade Theory with Numbers:
Quantifying the Consequences of Globalization. In Gopinath, G, E. Helpman and K.
Rogoff (eds), Handbook of International Economics, Chapter 4, Volume 4.
[54] Cowen T. (2002) Creative destruction: How globalization is shaping the
worlds cultures. Princeton. NJ: Princeton University Press. Desmet
[55] Crenshaw E.M. and R. M Anthony (2014), City Size and Political
Contention, International Journal of Sociology, 44(4):7-33.
[56] Davis, J. and J.V. Henderson (2003), 'Evidence on the political economy of
the urbanization process', Journal of Urban Economics 53: 98–125.
[57] De Sousa, J. and J. Lochard, 2009. Histoire coloniale et commerce
international, Revue économique, 3(60): 635-645.
[58] Debaere, P., 2014. The Global Economics of Water: Is Water a Source of
Comparative Advantage? American Economic Journal: Applied Economics, 6(2):
32–48.
[59] Dell, M., B F Jones, B A Olken, 2014. What Do We Learn from the Weather?
The New Climate–Economy Literature. Journal of Economic Literature, Vol. 52,
No. 3, pp. 740-798.
[60] Dinar, S., 2009. Scarcity and cooperation along international rivers. Global
Environmental Politics, 9(1):109–135.
[61] Docquier, F., E. Lodigiani, H. Rapoport and M. Schiff, Emigration and
democracy, Journal of Development Economics 120 (2016) 209–223.
[62] Duarte, R., V Pinilla, A Serrano, 2014. The water footprint of the Spanish
agricultural sector: 1860–2010. Ecological Economics, Vol. 108, pp 200-207.
[63] Dube, O., and J. Vargas. 2013. “Commodity Price Shocks and Civil Conflict:
Evidence from Colombia.” Review of Economic Studies 80 (4): 1384–1421.
[64] Duranton, G., (2008), 'Viewpoint: From cities to productivity and growth in
developing countries', Canadian Journal of Economics, 41(3):689-736
[65] Eaton, J., Tamura, A., 1994. Bilateralism and Regionalism in Japanese and
U.S. Trade and Direct Foreign Investment Patterns. Journal of the Japanese and
International Economies 8 (4), 478 – 510.
[66] Eeckhout, J. (2004), Gibrat's Law for (All) Cities, American Economic
Review, 94(5): 1429-1451.

133
 References

[67] Fally, T., 2015. Structural Gravity and Fixed Effects, Journal of
International Economics, 97 (1): 7685.
[68] Fearon, James D. 2006. “Ethnic Mobilization and Ethnic Violence.” In The
Oxford Handbook of Political Economy, ed. Barry R. Weingast and Donald A.
Wittman, 852–68. Oxford and New York: Oxford University Press.
[69] Fearon, James D., and David D. Laitin. 2003. “Ethnicity, Insurgency, and
Civil War.” American Political Science Review, 97(1): 75–90.
[70] Feler, L. and J.V. Henderson (2011), 'Exclusionary policies in urban
development: Under-servicing migrant households in Brazilian cities', Journal of
Urban Economics, 69(3):253-272.
[71] Findlay, R. and K.H. O'Rourke (2009), Power and Plenty: Trade, War, and
the World Economy in the Second Millennium. Princeton U Press.
[72] Fleurbaey M, 2015. Forced trades in a free market, in C. Binder, G.
Codognato, M. Teschl, Y. Xu (eds.), Individual and Collective Choice and Social
Welfare. Essays in Honor of Nick Baigent, Springer.
[73] Francois, Patrick and van Ypersele, Tanguy. 2002. On the Protection of
Cultural Goods. Journal of International Economics 56(2): 359 369.
[74] Gasiorek, M., Peter Holmes, Jim Rollo, Zhenkun Wang, Javier Lopez
Gonzalez, Maximiliano Mendez Parra, Maryla Maliszewska, Wojtek Paczynski,
Xavier Cirera, Dirk Willenbockel, Francesca Foliano, Marcelo Olarreaga, Kamala
Dawar (2010): Mid-term Evaluation of the EU’s Generalised System of
Preferences, CARIS – Centre for the Analysis of Regional Integration at Sussex
[75] Giuliano, P., A. Spilimbergo and G. Tonon (2014), 'Genetic distance,
transportation costs, and trade', Journal Economic Geography, 14 (1): 179-198.
[76] Gleditsch K, 2007. Transnational Dimensions of Civil War. Journal of Peace
Research, vol. 44, no. 3, 2007, pp. 293–309.
[77] Gouel C, and D Laborde (2019), The Crucial Role of International Trade in
Adaptation to Climate Change. NBER Working Paper No. 25221.
[78] Guiso, L., P. Sapienza and L. Zingales (2009), 'Cultural Biases in Economic
Exchange?', The Quarterly Journal of Economics, 124 (3): 1095-1131.
[79] Hamner, J.,Wolf, A., 1998. Patterns in International Water Resource
Treaties: The Transboundary Freshwater Dispute Database. Colorado Journal of
International Environmental Law and Policy, Yearbook.

134
 References

[80] Harris, C. (1954), “The Market as a Factor in the Localization of Industry in

the United States”, Annals of the Association of American Geographers 64:315–
348.
[81] Head K., T Mayer. Market Potential and the Location of Japanese Firms in
the European Union. The Review of Economics and Statistics, 2004, 86 (4),
pp.959-972.
[82] Head K., T. Mayer. Regional wage and employment responses to market
potential in the EU. Regional Science and Urban Economics 36 (2006) 573–594
[83] Head, K. and T. Mayer (2002), 'Illusory Border Effects: Distance
Mismeasurement Inflates Estimates of Home Bias in Trade', CEPII Working Paper
2002-01.
[84] Head, K., Mayer, T., 2014. Gravity Equations : Workhorse, Toolkit, and
Cookbook. in Gopinath, G, E. Helpman and K. Rogoff (eds), Handbook of
International Economics, Chapter 3, Volume 4.
[85] Hegre, H, N Sambanis,. 2006. Sensitivity Analysis of Empirical Results on
Civil War Onset. The Journal of Conflict Resolution, 50(4): 508–535.
[86] Hegre, H, T Ellingsen, S Gates, and N. P. Gleditsch. 2001. “Toward a
Democratic Civil Peace? Democracy, Political Change, and Civil War, 1816–1992.”
American Political Science Review, 95(1): 33–48.
[87] Helpman, H., (1998), The Size of Region, In D. Pines, E. Sadka, and I. Zildcha
(eds.). Topic in public economics. Theorical and Applied Analysis. Cambridge:
Cambridge University Press: 33-54.
[88] Henderson J. V., Storeyguard A., Weil D. N., 2012, Measuring Economic
Growth from Outer Space, American Economic Review, 102(2):994-1028.
[89] Henderson, J. V. and H.G. Wang (2007), 'Urbanization and City Growth: the
Role of Institutions', Regional Science and Urban Economics, 37(1): 283-313.
[90] Henderson, J. V., A. Storeygard and D. N. Weil. (2012), 'Measuring
Economic Growth from Outer Space', American Economic Review, 102(2): 994–
1028
[91] Hensel, P., Brochmann, M., 2008. Armed Conflict over International Rivers.
Paper presented at the American Political Science Association Annual Meeting,
Boston, MA.

135
 References

[92] Iwanowsky M., 2018. The Role of Ethnic Networks in Africa: Evidence from
Cross-Country Trade. WP Institute for International Economic Studies, Stockholm
University.
[93] Jacks, David S., Christopher M. Meissner and Dennis Novy. 2008. "Trade
Costs, 1870-2000." American Economic Review, 98(2):529-34.
[94] Kamola I (2007) The global coffee economy and the production of
genocide in Rwanda, Third World Quarterly, 28:3, 571-592
[95] Karayalcin, C. and H. Yilmazkuday (2015), 'Trade and Cities', World Bank
Economic Review, 29(3): 523-549.
[96] Kaufmann, D., A. Kraay and M. Mastruzzi (2010), The worldwide
governance indicators : methodology and analytical issues, Policy Research
Working Paper Series 5430, The World Bank.
[97] Khol, T., Brakman, S., Garretsen, H., 2016. Do Trade Agreements Stimulate
International Trade Differently ? Evidence from 296 Trade Agreements. The
World Economy, 39 (1) : 97-131.
[98] Kiszewski A., Mellinger A., Spielman A., Malaney P., Sachs SE., and Sachs J.
2004. “A Global Index Representing The Stability of Malaria Transmission”. The
American Journal of Tropical Medecine and Hygiene, 70(5) 486-98
[99] Krugman, P. and E.R. Livas (1996), 'Trade policy and the Third World
metropolis', Journal of Development Economics, 49(1): 137-150.
[100] Krugman, P., (1991), 'Increasing Returns and Economic Geography', The
Journal of Political Economy, 99(3): 483-499
[101] Laitin, David D., 2007. Nations, States, and Violence. Oxford & New York:
Oxford University Press
[102] Lerner, D., (1958), The Passing of Traditional Society: Modernizing the
Middle East. New York: Free Press.
[103] Lipset, S. M. (1959), 'Some Social Requisites of Democracy: Economic
Development and Political Legitimacy', American Political Science Review, 53(1):
69–105.
[104] Lipton, M. (1977), Why Poor People Stay Poor: Urban Bias in World
Development, Cambridge, MA: Harvard University Press.
[105] Magee, C. S. P., 2008. New Measures of Trade Creation and Trade
Diversion, Journal of International Economics, 75 (2) : 349-362.

136
 References

[106] Majumdar, S., A. Mani and S. Mukan (2004), 'Politics, information and the
urban bias', Journal of Development Economics, 75(1): 137-165.
[107] Mariani F, M. Mercier and T. Verdier (2018), Diasporas and conflict.
Journal of Economic Geography, Volume 18, Issue 4, July 2018, Pages 761–793.
[108] Marshall, M G. and K Jaggers (2007), Polity IV Project: Dataset Users’
Manual, George Mason University and Center for Systemic Peace.
[109] Martin P., Mayer T., and M., Thoenig, 2008a. Make Trade Not War? Review
of Economic Studies, 75, 865-900.
[110] Martin P., Mayer T., and M., Thoenig, 2008b. Civil War and International
Trade. Journal of European Economic Association, 6(2-3):541-550.
[111] Mason, C.F., Umanskaya, V.I., Barbier, E. 2018. Trade, Transboundary
Pollution, and Foreign Lobbying. Environmental Resource Economics 70: 223.
[112] Mattoo A, A Mulabdic, M. Ruta, 2017. Trade Creation and Trade Diversion
in Deep Agreements, World Bank Policy Research Working Paper n. 8206.
[113] Mayer T, Vicard V, Zignago S., 2019. The Cost of Non-Europe, Revisited.
Economic Policy, Volume 34, Issue 98, Pages 145–199.
[114] Maystadt, J-F., G. De Luca, P. G. Sekeris, J. Ulimwengu, and R. Folledo. 2014.
“Mineral Resources and Conflicts in DRC: A Case of Ecological Fallacy?” Oxford
Economic Papers 66 (3): 721–49.
[115] Maystre N., Olivier J., Thoenig M., Verdier T., 2014. Product-Based Cultural
Change: Is the Village Global? Journal of International Economics Volume 92,
Issue 2, Pages 212-230
[116] Melitz M, J. S., 2003. The Impact of Trade on Intra-Industry Reallocations
and Aggregate Industry Productivity, Econometrica , 71(6), 1695-1725.
[117] Merrill, J., 2008. Water Management and Decision-Making in the Nile
Basin: A Case Study of the Nile Basin Initiative. University of South Florida,
https://scholarcommons.usf.edu/etd/402.
[118] Michalopoulos S., 2012. “The Origins of Ethnolinguistic Diversity”,
American Economic Review, 102(4), 1508-39.
[119] Michalopoulos S.,and Papaioannou E, 2016. The Long Run Effects of the
Scramble for Africa, American Economic Review, 106(7), 1802-1848.
[120] Milner V. H. and B. Mukherjee (2009), 'Democratization and Economic
Globalization', Annual Review of Political Sciences, 12(1):163–81

137
 References

[121] Mokyr, J., (2016), A Culture of Growth: The Origins of the Modern
Economy, Princeton University Press.
[122] Monfreda, C., N. Ramankutty, and J. A. Foley (2008), Farming the planet: 2.
Geographic distribution of crop areas, yields, physiological types, and net primary
production in the year 2000, Global Biogeochem. Cycles, 22, GB1022,
doi:10.1029/2007GB002947.
[123] Montesquieu, (1758), De l'esprit des lois
[124] Morin, JF, A. Dür and L. Lechner, 2018. Mapping the trade and
environment nexus: Insights from a new dataset, Global Environmental Politics,
vol. 18(1).
[125] Morin, JF, C Brandi and A Berger. 2019. The Multilateralization of PTA
Environmental Clauses - Scenarios for the Future? in Shifting Landscape of Global
Trade Governance (edited by M. Elsig, M. Hahn, and G. Spilker). Cambridge:
Cambridge University Press : 207-231.
[126] Moulton, B. R., 1990. An Illustration of a Pitfall in Estimating the Effects of
Aggregate Variables on Micro Units. The Review of Economics and Statistics, Vol.
72, No. 2, pp. 334-338
[127] Murdock, G. P. (1959): Africa: Its Peoples and their Culture History.
McGraw-Hill.
[128] Nel P.,and Righarts M, 2008. Natural Disasters and the Risk of Violent Civil
Conflict, International Studies Quarterly, 52(1), 159-185.
[129] Nitsch, V., (2006), 'Trade Openness and Urban Concentration: New
Evidence', Journal of Economic Integration 21(2), 340-362.
[130] Nunn, N. and D Puga (2012), 'Ruggedness: The Blessing of Bad Geography
in Africa', Review of Economics and Statistics, 94(1):20-36.
[131] Pemberton, T. J., M. DeGiorgio, and N. A. Rosenberg (2013), 'Population
Structure in a Comprehensive Genomic Data Set on Human Microsatellite
Variation', G3: Genes, Genomes, Genetics, 3(5): 903-919.
[132] Pesaran, H. and Q. Zhou (2016), Estimation of time-invariant effects in
static panel data models, Econometric Reviews, 37(10): 1137-1171.
[133] Rauch J.E. and R. Trindade, 2009. Neckties in the tropics: a model of
international trade and cultural diversity. Canadian Econ Journal, 42(3): 809-
1205.

138
 References

[134] Rauch, J E. and V. Trindade (2002), 'Ethnic Chinese Networks in

International Trade', Review of Economics and Statistics, 84(1): 116-130
[135] Redding, S., and A. Venables, 2004 “Economic Geography and International
Inequality,” Journal of International Economics 62:1 (2004), 53– 82.
[136] Rodrik, D., 2018. What Do Trade Agreements Really Do? Journal of
Economic Perspectives, 23(2) : 73-90.
[137] Rodrik, D., World Bank, (2006). Goodbye Washington Consensus, Hello
Washington Confusion? A Review of the World Bank's "Economic Growth in the
1990s: Learning from a Decade of Reform". Journal of Economic Literature, 44(4),
973-987.
[138] Santos Silva, J. and S. Tenreyro (2006), 'The log of gravity', The Review of
Economics and Statistics, 88: 641-658.
[139] Santos Silva, J., Tenreyro, S., 2011. Further Simulation Evidence on the
Performance of the Poisson-PML Estimator. Economics Letters 112 (2), 220–222.
[140] Santos Silva, S., Tenreyro, S., 2006. The Log of Gravity. Review of
Economics and Statistics, 88: 641-658.
[141] Shaffer G., M. Fleurbaey 2018, “Adapting trade policy to social,
environmental, and development goals”, T20 Policy Brief Draft.
[142] Smarzynska Javorcik, B. S., Wei, S-J., 2003. Pollution Havens and Foreign
Direct Investment: Dirty Secret or Popular Myth? The B.E. Journal of Economic
Analysis & Policy, 3(2): 1-34.
[143] Spolaore, E., and R. Wacziarg (2018), 'Ancestry and Development: New
Evidence', 33(5): 748-762.
[144] Staiger, D. and J. Stock (1997), 'Instrumental Variables Regression and
Weak Instruments', Econometrica, 65(3): 557-586.
[145] Tir J. and J.T. Ackerman, 2009. Politics of formalized river cooperation.
Journal of Peace Research, 46(5):623–640.
[146] Viner, J., 1950. The Customs Union Issue. Carnegie Endowment for
International Peace: New York.
[147] Weber, M. (1909), Urbanization and Social Structure in the Ancient World,
in Weber: Selections in Translation, edited by W. G. Runciman, 290–314.
Cambridge: Cambridge Univ. Press

139
 References

[148] Wimmer A., Cederman LE., and B., Min, 2009. Ethnic Politics and Armed
Conflict: A Configurational Analysis of a New Global Data Set. American
Sociological Review, 74(2):316-337.
[149] Winters, L., Alan, Neil McCulloch, and Andrew McKay. 2004. "Trade
Liberalization and Poverty: The Evidence So Far." Journal of Economic Literature,
42 (1): 72-115.
[150] Wolf A, Arius J, Danielson J, Ward B., Pender J., 1999. International River
Basins of the World. Water Resources Development, Vol. 15, No. 4, 387-427.
[151] Wooldridge, J. M., 2002. Econometric Analysis of Cross-Section and Panel
Data. MIT Press, Cambridge, MA.

140
 École doctorale Sciences Sociales et Humanités (481)
Laboratoire :
Centre d’Analyse Théorique et de Traitement des données économiques (CATT)
Université de Pau et des Pays de l’Adour (UPPA)
Contact:
[email protected]

Cities, Ethnic Wars and Water: Three essays in International Economics

Abstract

International trade is one of the key factors that have deeply reorganized the world economy.
This thesis addresses three different topics in the field of international economics. It firstly
brings new insights on the contribution of international trade to the geographical distribution
of populations in countries and secondly, determines trade effect on ethnic conflicts.
Afterwards, this document examines the reaction of world trade to international water
agreements that increasingly emerged in recent decades as a tool for sustainable and socially
equitable development. The contributions of this document stem not only from the originality
of the topics discussed and the methodologies adopted, but it also proposes and builds more
appropriate indicators that are better suited to capturing the issues under analysis.
Thus, the first chapter assesses whether international trade is relevant in explaining the
development of large cities in past European colonies. We argue that trade restrictions can
provide an advantage to one city, which may become the platform for exports and
catastrophically attracts people. By contrast, trade liberalization, by providing market access to
other cities, fosters the dispersion of economic activities and consequently a dispersion of the
population. The empirical investigations have led to the conclusion that international trade
does not drive the size of large cities in the former colonized countries. Only institutions drive
the size of primate cities: democracy goes hand in hand with agglomeration.
The second chapter analyses the relationship between trade and insecurity in African countries.
Focusing on ethnic conflicts, this part of the thesis argues that the type of trading partners
determines the opportunity cost of an ethnic conflict. The international trade of ethnic groups,
the regional trade and countries' internal trade may have heterogeneous effects on peace. The
data analyses results support this prediction by pointing out that international ethnic trade and
countries' internal trade are peace-promoting tools unlike trade between countries that share
at least one ethnic group, which appears as a factor of ethnic conflicts. However, only
international ethnic trade seems to reshape the national identity of countries, which in turn
reduces the likelihood of ethnic conflicts.
In the last chapter of the thesis, we discuss the growing concern about the international
management of shared water resources. As water constitutes an important input in the
production of goods, it seems necessary to assess the effectiveness of international water
agreements with regards to trade after several decades of implementation. One can argue that
these agreements are just "cheap talk", or rather pragmatic attempts to meet the sustainable
development objectives to take actions for better water management. Conversely, these
agreements, by allowing a better allocation of water resources between countries, can also
stimulate production (especially agricultural production) and ultimately trade. Thus, this
chapter firstly sheds light on the climate change indicators that are the main drivers of the
increasing implementation of shared water resources' agreements. Secondly, it examines
whether these agreements affect international trade before proposing their overall effects on
countries. In general, the study reveals that water agreements, motivated in particular by high
temperature signals, represent a cost to free trade in Europe, Asia and the Americas, except in
Africa. Therefore, it seems that these water agreements have substantial environmental
protection contents that restrict trade.
Résumé

Le commerce international est l'un des éléments majeurs qui a restructuré profondément la vie
des nations et de l'économie mondiale. Cette thèse aborde trois sujets différents dans le
domaine du commerce international. Elle apporte en premier lieu, de nouveaux
développements sur la contribution du commerce international à la répartition géographique
des populations au sein des nations et détermine en second lieu son influence sur les conflits
ethniques. Ensuite, elle analyse la réaction des échanges mondiaux aux accords internationaux
sur l'eau qui s'érigent ces dernières années comme un outil important du développement
durable et socialement équitable. Les contributions de ce document, tiennent non seulement de
l'originalité des questions traitées et des méthodes utilisées, tout en proposant et en exploitant
des mesures mieux indiquées pour capter les faits analysés.
Dans le premier chapitre, nous analysons si le commerce international a eu un rôle dans le
développement des grandes villes des anciens pays colonisés. D'un point de vue théorique, un
accès difficile au marché mondial de biens pourrait amener la principale ville à devenir la
plateforme des exportations, attirant ainsi les populations. Dans le cas contraire, il y aura une
meilleure répartition de l'activité économique dans le pays, et par conséquent moins
d'incitation des populations à s'agglomérer dans la capitale. Notre analyse montre que le
commerce international semble ne pas affecter la taille des grandes villes dans les anciens pays
colonisés. Les institutions restent par contre déterminantes pour expliquer la taille des villes :
la démocratie est source de concentration des populations dans les grands centres urbains des
pays en développement.
Le second chapitre examine la relation entre le commerce et l'insécurité dans les pays africains.
En nous concentrant sur les conflits ethniques, cette partie met en avant le fait que le type de
partenaires à l'échange déterminerait le coût d'opportunité à un conflit ethnique. Les résultats
obtenus valident cette intuition en démontrant que les échanges internationaux des ethnies, le
commerce régional et le commerce interne d'un pays ont des effets hétérogènes sur la
promotion de la paix. Ainsi, il ressort que le commerce international des ethnies et le commerce
interne des pays ont des effets pacificateurs contrairement au commerce entre pays qui
partagent au moins un groupe ethnique. Cependant, seul le commerciale international des
ethnies semble remodeler l'identité nationale qui à son tour réduit la probabilité de conflits
ethniques.
Le dernier chapitre se concentre sur la fulgurante augmentation des coopérations
internationales concernant la gestion des bassins d'eau communs. L'eau étant un élément
essentiel des processus de production des biens, il conviendrait de savoir, après plusieurs
décennies d'implémentation, l'impact de ces accords sur le commerce. En effet, trois hypothèses
concurrentes peuvent être émises. Il se pourrait que ces accords ne soient que des discours
politiques sans conséquences réelles n'impactant pas le commerce. Il est aussi possible que ces
accords se matérialisent par des tentatives pragmatiques visant à respecter les objectifs de
développement durable et dans ce cas, il est probable que ces accords ont un effet négatif sur le
commerce. A l'inverse, ces accords permettant une meilleure allocation des ressources en eau
entre pays, peut aussi stimuler la production (notamment agricole) et in fine les échanges
commerciaux. En général, les résultats révèlent que les accords sur l'eau, motivés en particulier
par les hausses de température, représentent un coût au libre-échange en Europe, en Asie et en
Amérique (mais pas en Afrique), il semble donc que ces accords sur l'eau ont un contenu de
protection environnementale important qui limitent les échanges commerciaux
 Ph.D. THESIS
UNIVERSITY OF PAU (UPPA)
Doctoral School of Social Sciences and Humanities (ED SSH-481)

Ph. D. dissertation in Economics

by Tchapo GBANDI

Publicly defended on July, 9th 2020

CITIES, ETHNIC WARS AND WATER:

THREE ESSAYS IN INTERNATIONAL
ECONOMICS
Research center:
Center for the Analysis of Trade and Economic Transition (CATT)

COMMITTEE
EVALUATORS
• Julie LOCHARD Professor, Université Paris-Est Créteil (UPEC)

• Mathieu COUTTENIER Professor, Université de Lyon, ENS Lyon

EXAMINERS
• Marion MERCIER Researcher, CNRS, LEDa-DIAL, Université Paris Dauphine-PSL

• Serge REY Professor, Université de Pau et des Pays de l’Adour (UPPA)

• Théophile BOUGNA Researcher in Economics, World Bank Group

ADVISOR
• Fabien CANDAU Professor, Université de Pau et des Pays de l’Adour (UPPA)