NBER WORKING PAPER SERIES

GLOBALIZATION AND POLITICAL STRUCTURE

Gino Gancia
Giacomo A.M. Ponzetto
Jaume Ventura

Working Paper 22046

http://www.nber.org/papers/w22046

NATIONAL BUREAU OF ECONOMIC RESEARCH

1050 Massachusetts Avenue
Cambridge, MA 02138
February 2016, Revised June 2021

We thank Janko Heineken, Marta Santamaria and Jagdish Tripathy for excellent research
assistance. We are grateful for helpful comments by Giovanni Andreottola, Pablo Beramendi,
Graziella Bertocchi, Alessandra Bonfiglioli and participants at numerous conferences and
seminars. We acknowledge financial support from the European Research Council (ERC) under
the European Union's Horizon 2020 research and innovation program (Advanced Grant 693512 -
GEPPS and Starting Grant 714905 - CITIZINGLOBAL), the Spanish Ministry of Economy and
Competitiveness (RYC-2013-13838) and its Severo Ochoa Programme for Centres of Excellence
in R&D (SEV-2015-0563 and CEX2019-000915-S), the Generalitat de Catalunya through
CERCA and the SGR Programme (2014-SGR-830 and 2017-SGR-1393), the Institut d'Estudis
Autonòmics, and the Leverhulme Trust through a 2019 Research Fellowship. The views
expressed herein are those of the authors and do not necessarily reflect the views of the National
Bureau of Economic Research.

NBER working papers are circulated for discussion and comment purposes. They have not been peer-
reviewed or been subject to the review by the NBER Board of Directors that accompanies official
NBER publications.

© 2016 by Gino Gancia, Giacomo A.M. Ponzetto, and Jaume Ventura. All rights reserved. Short sections
of text, not to exceed two paragraphs, may be quoted without explicit permission provided that full
credit, including © notice, is given to the source.
Globalization and Political Structure
Gino Gancia, Giacomo A.M. Ponzetto, and Jaume Ventura
NBER Working Paper No. 22046
February 2016, Revised June 2021
JEL No. D71,F15,F55,H77,O57

ABSTRACT

This paper develops a theoretical framework to study the interaction between globalization and
political structure. We show that political structure adapts in a non-monotonic way to declining
transport costs. Borders hamper trade. At an earlier stage, the political response to expanding
trade opportunities consists of removing borders by increasing country size. At a later stage,
instead, it consists of removing the cost of borders by creating international unions. This leads to
a reduction in country size. Moreover, diplomacy replaces conquest as a tool to ensure market
access. These predictions are consistent with historical evidence on trade, territorial changes and
membership of international unions.

Gino Gancia Jaume Ventura

Queen Mary University of London CREI
Mile End Road Ramon Trias Fargas, 25-27
London E1 4NS 08005-Barcelona
United Kingdom SPAIN
[email protected] and Barcelona GSE
and also NBER
Giacomo A.M. Ponzetto [email protected]
CREI, Universitat Pompeu Fabra,
IPEG and Barcelona GSE
C/ Ramon Trias Fargas, 25-27
08005 Barcelona
Spain
[email protected]
 1 Introduction

Since the Industrial Revolution, the cost of distance has been falling dramatically thanks to
a stream of major technological innovations, like the railroad, the steamship, the telegraph,
the jet engine, containerization and most recently the internet. Such technological progress
has fundamentally transformed worldwide economic geography and fuelled a continuous ex-
pansion in the size of markets. In 1820, international trade was very modest: about 2% of
world output. Over the following century, international trade grew more than four-fold to
8%. This …rst wave of globalization was cut short in the interwar period, which saw inter-
national trade decline to about 5% of world output. After World War II, a second wave
of globalization started and still continues today. By 2010, international trade had reached
unprecedented levels, surpassing 20% of world output.
These two ages of globalization saw political geography evolve in opposite directions. In
the nineteenth century, economic and political integration proceeded together. Sovereign
states grew larger and fewer, from 125 in 1820 to merely 54 at the eve of the Great War.
Conversely, in the postwar era economic integration has been accompanied by political frag-
mentation. The number of countries has risen to a record high of more than 190. At the
same time, there has been a proliferation of international treaties and institutions aimed
especially at fostering economic integration, such as the World Trade Organization and the
European Union. These trends are illustrated graphically in Figure 1, which shows the his-
torical evolution of the number of sovereign states in the world and the number of members
of the GATT/WTO, along with average exports as a share of GDP.1
The sharp reversal in the link between economic and political integration presents an open
puzzle, which we address in this paper. We show that the observed evolution of political
structure during both waves of globalization can be understood as an e¢ cient response to the
falling cost of distance. Our starting point is that borders hamper trade, so expanding trade
opportunities make borders costlier. Political structure needs to adapt by removing borders
or by reducing their cost. We …nd that this adaptation entails a non-monotonic evolution of
e¢ cient country size. At an earlier stage, the e¢ cient political response to expanding trade
opportunities is to remove borders by creating larger countries. At a later stage, it is to
remove the cost of borders by creating international unions. This induces a reduction in the
size of countries.

1
The data on the number of states are from Butcher and Gri¢ ths (2013). The trade share is from
Maddison (2001). See Appendix A.1 for more details.

1
 200 20

180 18

160 16

140 14

120 12

100 10

80 8

60 6

40 4

20 2

0 0
1810 1830 1850 1870 1890 1910 1930 1950 1970 1990 2010
Country WTO Trade

Figure 1: Economic Integration and Political Integration. The …gure plots the trade share
(right axis), and the number of countries and of WTO members (left axis). See Appendix
A.1 for details on data.

Naturally, our theory does not imply that political geography responds to economic e¢ -
ciency alone. On the contrary, our model also sheds light on historical patterns of interstate
con‡ict. When creating international unions is too costly, we show it is appealing for the
Great Powers to wage war to conquer markets. Empire-building enables them to trade with
their colonies while imposing extractive institutions. As e¢ cient market size keeps growing,
however, we …nd that this strategy reaches a breaking point. Con‡ict and imperialism are
no longer advantageous for the Great Powers. They prefer gaining access to world markets
through peaceful diplomacy, by building supranational institutions that respect all mem-
bers’political autonomy. These theoretical predictions are borne out by empirical evidence.
During the …rst wave of globalization, increasing trade led to an expansion in country size
accompanied by con‡ict. Both links were broken during the second wave of globalization,
which saw instead the peaceful rise of supranational institutions.
To derive our results, in Section 2 we set up our model of a world with a continuum of
basic geographical units, which we refer to as localities. Each contains people that share
common preferences. Goods can be transported at a negligible cost within localities, but at

2
a positive cost across localities. Governments perform two tasks: (i) they enforce contracts,
protect property rights and enact economic regulations that help markets work; and (ii) they
provide public services such as education and welfare programs. We study two alternative
ways of organizing the government. The …rst is a single-level government that performs both
tasks. We refer to this single-level government as a country. The second possibility is a
two-level government. The lower level provides public services and we also refer to it as a
country; while the higher level regulates markets and we refer to it as an international union.
Our goal is to study how the shape of government changes as the cost of distance declines.
We make two standard assumptions about the e¤ects of government. The …rst is the
presence of border e¤ects. Localities that do not belong to the same country nor to the same
international union have di¤erent economic regulations. Therefore, they can trade only a
limited range of goods. The second assumption is policy uniformity. If two localities belong
to the same country, they receive the same basket of public services even though they would
prefer di¤erent baskets.2
Without costs of government, the optimal political structure would be a two-level gov-
ernment. The …rst level would be a continuum of country governments, one for each locality,
providing each of them with its preferred basket of public services. The second level would
be an international union of all localities that regulates markets and eliminates all border
e¤ects. Unfortunately, this political structure is too expensive.
We make two standard assumptions about the costs of government. The …rst is the
presence of economies of scale. There are some costs of setting up and running a country that
are …xed or independent of the number of localities that belong to it. The second assumption
is the presence of economies of scope. Coordinating di¤erent levels of government is costly
and, as a result, a two-level government is more expensive than a single-level government.
Economies of scale and scope a¤ect the optimal political structure. Economies of scale
make it desirable to have a discrete number of country governments rather than a continuum
of them. Localities are willing to accept public services less than ideally tailored to their
preferences in order to bene…t from economies of scale. Economies of scope, if large enough,
make it desirable to have a single- rather than a two-level government. Localities may also

2
We assume that local preferences di¤er with respect to public services but not economic regulation.
Market-enabling economic regulation aims primarily at increasing e¢ ciency, and this goal is widely shared by
people with di¤erent preferences (Coase 1960; Posner [1973] 2014). Accordingly, government functions such
as contract enforcement, monetary policy, or the policing of anti-competitive practices are often entrusted to
apolitical technocrats. Public services, however, are a focus of political tension because people have di¤erent
views on how children should be educated, on the proper size and scope of the welfare system, and so on.
Such preferences vary systematically across localities re‡ecting their distinctive history and culture.

3
be willing to accept higher trade costs to take advantage of economies of scope.
The equilibrium political structure balances four classic forces: border e¤ects, preference
heterogeneity, economies of scale and economies of scope. We study how a reduction in
the cost of distance a¤ects this balance. Globalization, i.e., economic integration, is an
endogenous outcome in our theory. An exogenous decline in the cost of distance fosters
trade directly for a given political structure. It also fosters trade indirectly by producing
endogenous changes in political structure aimed at facilitating trade.
In Section 3 we assume that localities bargain e¢ ciently in a world ruled by law and
diplomacy. When trade opportunities are limited, the bene…t of creating an international
union does not justify sacri…cing economies of scope. Thus, a single-level government is
optimal. As trade opportunities expand, the incentives to remove borders grow. The number
of countries declines and the mismatch between each locality’s ideal and actual provision of
public services grows. Eventually, this mismatch is large enough to justify the move to a
two-level government. The world political structure shifts from a few large countries to many
small countries within an international union. This two-level government is more expensive,
but it is nonetheless desirable because it facilitates trade and improves preference-matching
in the provision of public services.
In Section 4 we allow a set of core localities to wage war and build empires. War is
costly but it allows core localities to remove borders with their colonies while imposing their
preferences on them. Thus, we …nd that there is an intermediate stage of globalization in
which empires are formed. Eventually this stage ends, empires collapse, and international
unions are created to promote trade and peace. Core localities choose to avoid the cost of
war and to replace conquered colonies with free partners in an international union. The
cause of imperial collapse at a late stage of globalization is the same as the cause for the
rise of empires at an early stage: namely, the desire to reap gains from trade in the most
cost-e¢ cient manner.
In Section 5 we further relax our assumptions of symmetry and allow core localities to
have lower costs of trading with one another. We show that this gives rise to a richer and
more gradual pattern of international integration. World-straddling empires are not replaced
at once by a world union, but may rather give rise …rst to a regional union of core localities,
which only later grows to encompass the periphery too.
In Section 6 we show three signi…cant patterns in historical data since 1870. First, in-
creases in trade predict territorial expansion before international unions. Second, this corre-
lation is reversed after international unions are created. Third, membership of international

4
unions predicts territorial contraction. This evidence is consistent with the predictions of
our theory.

Related Literature The motivating facts presented in Figure 1 were …rst noted by Kahler
and Lake (2004) and Lake and O’Mahony (2004). They highlight the puzzling reversal in
the link between economic and political integration. They stress that it poses an obstacle
for e¢ ciency-based explanations of political structure, which have so far failed to account
for it and for the emergence of supranational institutions during the second but not the …rst
wave of globalization.
Several economic theories can help explain why economic integration has been accompa-
nied by political fragmentation since World War II (Bolton and Roland 1996, 1997; Alesina,
Spolaore and Wacziarg 2000; Casella 2001; Casella and Feinstein 2002). All of them, how-
ever, predict a monotonic e¤ect of globalization on political structure, and thus fail to explain
why the …rst wave of globalization was accompanied by a decline in the number of countries.
Nor can they explain the creation of international unions. Closest to our own work, Alesina,
Spolaore and Wacziarg (2000) add border e¤ects to Alesina and Spolaore’s (1997) seminal
theory of country formation based on the trade-o¤ between preference heterogeneity and
economies of scale.3 They explain the increase in the number of countries during the second
wave of globalization by interpreting globalization as an exogenous weakening of the border
e¤ect. As borders become less costly, e¢ cient political structure reacts by creating more
borders.4
Our model is the …rst to account for the non-monotonic relationship between globaliza-
tion and political structure, and to explain the appearance of multi-level governance during
the second wave of globalization. We obtain these results because of two innovations relative
to prior work. First, we recognize that economies of scope are limited.5 As a consequence,
a broader set of political structures can be e¢ cient. Hence, we can explain the shift from
a single-level to a two-level government. Second, we consider a more primitive technolog-
ical driver of globalization: expanding trade opportunities caused by a gradual decline of

3
Desmet et al. (2011) calibrate this trade-o¤ for European countries.
4
Bolton and Roland (1996, 1997) focus on income distribution and …nd that heterogeneous countries may
break up if their barriers to external trade decline. Casella (2001) and Casella and Feinstein (2002) study
how preferences for public goods can endogenously become more heterogeneous as market size expands and
enables greater specialization.
5
Alesina and Spolaore (2003) o¤er an insightful discussion of economies of scope. Chapter 2 analyzes
an arbitrary system of overlapping jurisdictions. Chapter 9 presents a system of overlapping jurisdictions
constrained to form a pyramidal hierarchy.

5
transport costs, which makes borders costlier. This creates incentives to remove borders
rather than to create them. In our theory, the weakening of the border e¤ect occurs only
endogenously as political structure adapts to new trade opportunities. Both our innovations
are crucial: without either, theories based on economic e¢ ciency would be unable to explain
why economic and political integration went together until World War II, and then took
di¤erent directions.6
Our work is also related to the literature on trade and war. The idea that trade promotes
peace was formalized by Polachek (1980). It is based on the premise that con‡ict harms
trade and hence trade openness raises the opportunity cost of war (Alesina and Spolaore
2003; Rohner, Thoenig and Zilibotti 2013). The opposite idea that trade generates military
con‡ict is instead expressed in neo-Marxist theories of imperialism. Trade can also make
countries dependent on others and therefore vulnerable (Bonfatti and O’Rourke 2018).7
Our paper suggests that these seemingly antithetical views capture two di¤erent stages of
the same model. We provide a uni…ed explanation for why territorial changes are more
associated with military con‡ict in the …rst wave of globalization than in the second, an
empirical pattern noted by Lake and O’Mahony (2006). Consistent with our result that
international unions remove the market-access incentive for waging war, Martin, Mayer and
Thoenig (2012) …nd evidence that regional trade agreements promote peaceful relations.
The e¤ect of war on country formation has been studied by Alesina and Spolaore (2003),
Gri¢ ths (2014) and Gennaioli and Voth (2015), among others. These papers show that
changes in military technology can explain country size, investment in state capacity and
the provision of public goods. However, they …nd a monotonic e¤ect of military technology
on political structure. Our theory also recognizes con‡ict as one of the determinants of
country formation and of the provision of public goods. In our model, waging war is one
reason why countries grow large. We show, however, that changes in military technology
alone are not su¢ cient to explain persuasively the observed evolution of political structure.
Instead, we …nd that expanding trade opportunities are key to explain endogenously the
creation of international unions along with the switch from a world of aggression in which

6
One way to interpret our …ndings is that, through our two innovations, we provide an explanation of
the decline in the border e¤ect that drives Alesina, Spolaore and Wacziarg’s (2000) results. The reaction to
this decline is the same in our model as in theirs: an increase in the number of countries. However, they
take the decline as exogenously given. Instead, we show it is an endogenous consequence of the creation of
international unions.
7
Martin, Mayer and Thoenig (2008) show that bilateral trade lowers the probability of con‡ict, but
multilateral trade openness decreases dependence on any given country and hence the cost of a bilateral
con‡ict.

6
countries grew large to one of diplomacy in which countries became smaller.
Finally, our work is related more broadly to the economic analysis of federalism and of
the geographic structure of government. Our model embeds the key trade-o¤ that lies at the
heart of the classic theory of …scal federalism (Oates 1972). Centralization reaps economies
of scale and bene…ts from policy coordination, but it imposes a uniform policy on localities
with di¤erent preferences. Models of political centralization and decentralization have been
applied most often to the architecture of government at the sub-national level (Lockwood
2006; Treisman 2007). However, the same insights apply to the study of international unions
(Hooghe and Marks 2003; Alesina, Angeloni and Etro 2005; Ruta 2005). Prior research in this
…eld has overwhelmingly focused on the optimal size and composition of an exogenously given
number of government tiers. Surprisingly, the literature has devoted much less attention to
the choice between a single-level and a multi-level governance structure, which our analysis
focuses on (an exception is Bo¤a, Piolatto and Ponzetto 2016).8

2 A Model of Political Structure

In this section we develop a stylized model of the world that contains the basic ingredients
of our theory: geography, markets and preferences. The model mixes these ingredients
imposing a high degree of symmetry. This allows us to derive our basic results about the
e¤ects of expanding trade opportunities on political structure quickly and intuitively.
The concept of locality is a key primitive in our theory. We model the world as a set
of places within which there are neither geographical nor cultural distances, and we label
them localities. Thus, localities consist of a group of people sharing common preferences
and inhabiting a particular territory. This approach, which is common in the literature,
simpli…es the study of how people with di¤erent preferences interact and organize themselves
into political entities. But it is silent about how these di¤erent preferences arose in the …rst
place and how they evolve over time. It also abstracts from domestic con‡ict.
The concept of trade opportunities is another important primitive in our theory. Geo-
graphical distances introduce trade costs across localities. In particular, we adopt the usual
assumption of iceberg transport costs across localities. We examine the implications of ex-

8
Federalism is a way of making public-good provision more e¢ cient by decentralizing it to a sub-national
level of government (Oates 1972). We view economic unions instead as a way of making market regulation
more e¢ cient by centralizing it to a supra-national level. Intuitively, the two structures may well coexist
because they solve di¤erent problems. We derive these results formally in Appendix A.4.4, where we present
an extension of our model that includes a third, sub-national, level of government.

7
panding trade opportunities created by an exogenous decline in these costs.

2.1 Basic Setup

We consider a world with a continuum of atomistic localities, l 2 [0; 1]. Each locality contains
a positive measure of identical individuals. Wl denotes the welfare of the representative
individual of locality l. For short, we refer to this individual as “locality l.” Then, the
welfare of locality l is:
Wl = WlM + WlG , (1)

where WlM is the utility derived from the consumption of market goods and WlG is the utility
derived from public services.
Governments provide public services and regulate markets, so government activity a¤ects
both welfare components. A political structure for this world consists of two partitions of
the set of localities [0; 1] into governments: a public-service partition P with typical element
Pn 2 P ; and an economic-regulation partition R with typical element Rn 2 R.
If P = R, we say that the world has a single-level governance structure, and we refer
to the common elements of P and R as country governments or countries. Each of these
countries provides both public services and market regulation to its constituent localities.
If P 6= R, we say that the world has a two-level governance structure. It will presently
become clear that governments have a pyramidal hierarchy: if the partitions P and R do
not coincide, the …ner partition P is always a re…nement of the coarser one R. Hence, we
refer to the (smaller) elements of P as country governments or countries, and the (larger)
elements of R as international unions or unions. Countries provide public services to their
constituent localities, while unions regulate the markets of their constituent countries.
We develop a model of the partitions P and R, that is, a model of how localities organize
themselves into countries and how countries organize themselves into unions. We start from
assumptions about preferences, technology and the costs of government, and we determine
how welfare Wl depends on political structure (P; R).

8
 2.1.1 Markets and Trade

There is a continuum of industries, i 2 [0; 1]. Let cl (i) be the consumption of goods of
industry i by locality l. The utility function takes the following form:
Z 1
WlM = ln cl (i) di. (2)
0

The production of …nal consumption goods requires industry-speci…c di¤erentiated input

varieties, m 2 [0; 1]. De…ne cl (m; i) as the amount of inputs of the variety m for industry i
used by locality l in the production of …nal goods. Then:
Z 1
cl (i) = exp ln cl (m; i) dm . (3)
0

These consumption preferences and production technology are symmetric across and within
industries, and we specify convenient Cobb-Douglas functional forms. As a result, each
locality spends an identical fraction of its income on each variety of each industry.
To introduce gains from specialization and trade, we adopt a simple symmetric version
of the Ricardian model. Each locality is endowed with one unit of labor in each industry.
This unit can produce one unit of the variety with the same index as the locality (m = l),
or e units of any other variety (m 6= l). Since > 0, each locality has a technological
advantage in its “own”variety. The parameter measures the extent to which technologies
di¤er across localities and, therefore, the potential gains from specialization and trade.
There are technological barriers to trade. We assume uniform iceberg transport costs
across localities so that only a fraction e < 1 of the goods shipped from l to m 6= l arrives
to destination. To focus on the most interesting case in which trade costs are not prohibitive
and to ensure positive gains from trade, we assume that > > 0. Our measure of trade
opportunities is the wedge > 0, which captures the potential gains from trade.
This wedge increases as improvements in transportation technology reduce physical trade
costs . Trade opportunities can thus range from = 0 when trade costs are prohibitive
( = ) to a maximum of = when trade costs are nil ( = 0).
Policy-induced barriers to trade (i.e., border e¤ects) arise when di¤erent governments
regulate markets. Speci…cally, we assume that exchanging goods in a fraction 2 (0; 1) of
industries requires legal enforcement of contracts. In these industries, varieties cannot be
traded between localities that have di¤erent governments regulating their markets, i.e., that
belong to di¤erent elements Rn and Rn0 . The reason is that foreigners correctly anticipate

9
that domestic courts will discriminate against them ex post. In the remaining set of indus-
tries, contracts are self-enforcing and thus varieties can be traded without restrictions. This
formulation captures a simple yet realistic microfoundation for the well-known …nding that
borders obstruct trade.9
A market equilibrium is a set of prices and quantities such that individuals maximize
utility and markets clear. Appendix A.3.1 shows that there exists a unique market equilib-
rium. Traded industries specialize in each locality’s own input variety, export essentially
all of their production and import all remaining input varieties. Thus, in the measure
of industries that are traded across borders, consumption of the full measure of varieties is
cl (m; i) = e . In the remaining measure 1 of industries, this is also the consumption
in locality l of varieties from other localities it shares economic regulation with. However,
locality l is forced to produce locally all input varieties that correspond to localities that do
not belong to the same Rn . Thus, consumption of these nontraded varieties is cl (m; i) = e .
Equilibrium utility from consuming market goods equals:
Z 1
WlM = + 1 + R
Il=m dm , (4)
0

R
where Il=m is an indicator variable which takes value 1 if localities l and m belong to the
same Rn , and zero otherwise. Equation (4) shows the impact of border e¤ects. A decline
in transport costs raises trade opportunities ( ) in every industry. However, border e¤ects
prevent a locality from reaping the gains from trade in a mass of industries that require
contract enforcement. As a consequence, the value of removing each border e¤ect is pro-
portional to , where trade opportunities measure the potential gains from trade in any
single industry and the mass of industries subject to border e¤ects.

2.1.2 Governments

Each locality has a …xed endowment of resources, normalized to one, that can be allocated to
produce public services. These public services consist of a basket of di¤erentiated varieties,
x 2 [0; 1]. The basket provided to locality l is characterized by a density function gl (x)
R1
de…ned over these varieties, with gl (x) 0 and 0 gl (x) dx = 1. The utility derived from

9
This microfoundation is analyzed by Broner and Ventura (2011). There are other microfoundations,
though. For instance, tari¤s and non-tari¤ barriers are also policies that discriminate against foreigners and
limit the range of goods that can be traded.

10
these public services is: Z 1
WlG = l (x) u (gl (x)) dx K, (5)
0

where l (x) 0, u (gl (x)) = 1=gl (x) and K is a cost function to be de…ned shortly. We
refer to the …rst and second terms of Equation (5) as the bene…ts and costs of public services
respectively.
We now introduce three assumptions about governments. The …rst assumption is about
preference heterogeneity. Each locality has a di¤erent ideal variety of public services. We
de…ne and order the basic varieties such that the ideal one for locality l is x = l. We assume
that l (x) = > 0 if x = l; and l (x) = 0 otherwise.
The second and third assumptions de…ne the cost function K. First, there are economies
of scale in the provision of public services. Building and maintaining a government reduces
the utility from public services by a …xed total amount > 0. This cost is equally shared
among the constituent localities. Second, there are economies of scope across government
functions. Membership of a union reduces the utility from public services by an amount
> 0. This captures the costs of oversight and coordination between a country and the
union.
These assumptions imply the following utility from public services:

WlG = R1 IlU , (6)

gl (l) I P dm
0 l=m

P
where Il=m is an indicator variable that takes value 1 if if localities l and m belong to the
same Pn , and zero otherwise; and IlU is an indicator variable that takes value 1 if locality l is
a member of a union (Rn 6= Pn ), and zero otherwise. The …rst term in Equation (6) means
that the value of public services for locality l depends on the amount of its ideal variety
that is provided. The second term means that each locality’s share of the …xed cost of
government declines with the size of the country. The parameter measures the magnitude
of these economies of scale. The third term means that being a member of a union is costly.
The parameter measures the magnitude of these economies of scope.
To complete the model we need to make assumptions on how localities interact. We
consider law and diplomacy in Section 3, and war and conquest in Section 4. In both cases,
the world’s political structure is determined by the interplay of the forces that follow from
our assumptions. Although these assumptions are standard in the literature, we provide
some additional discussion of them in Appendix A.2.

11
 3 Law and Diplomacy

E¢ cient bargaining among localities delivers Pareto e¢ cient outcomes and constitutes a
natural benchmark. In this case, the equilibrium political structure is obtained by solving
the following maximization problem:
Z 1
(P; R) = arg max ! l Wl dl, (7)
0

R1
where f! l gl2[0;1] is a set of Pareto weights such that 0
! l dl = 1. Given the symmetry of our
model, we focus on the case in which the bargaining process treats all localities in the same
way: ! l = 1 for all l 2 [0; 1]. This political structure can be intepreted as the utilitarian
welfare optimum because it maximizes average world welfare. We view it as the description
of a world in which all localities have the right to choose their own political structure. This
is a world ruled by law and diplomacy.
An implication of the maximization problem (7) is that each country Pn provides a uni-
form bundle that contains equal amounts of the ideal varieties of its constituent localities.10
That is, locality l receives the following bundle of public services:
8
> 1
< R1 P
if Il=x =1
P
gl (x) = I
0 l=m
dm (8)
>
: 0 P
if Il=x = 0.

Thus, we can re-write Equation (6) as follows:

Z 1
WlG = P
Il=m dm R1 IlU . (9)
0 I P dm
0 l=m

The …rst term means that the value of public services for locality l declines with the size
of its country. As more localities join the country, the public services provided are farther
away from the ideal of each member locality. The parameter measures the importance of
this preference mismatch.
Combining Equations (1), (4), and (9), we obtain:
Z 1 Z 1
R P
Wl = + 1 + Il=m dm Il=m dm R1 IlU . (10)
P
0 0 0
Il=m dm

10
This is welfare-maximizing since localities have convex preferences.

12
Equation (10) shows how political structure determines welfare. It reveals the key trade-o¤
that underlies our theory. A desirable political structure should facilitate trade, accommo-
date preference heterogeneity and take advantage of economies of scale and scope. But these
goals cannot be achieved simultaneously and something must give.

3.1 Equilibrium Political Structure

Two preliminary results simplify the analysis of the maximization problem in Equation (7).
The …rst is that P and R contain equal-sized elements. All localities prefer the same optimal
country size, so the equilibrium is symmetric. We denote by S and U the sizes of each element
Pn 2 P and Rn 2 R respectively.11 The second result is that, as anticipated, P is a re…nement
of R. If it is ever worth paying the costs of having a two-level governance structure, this
is because localities desire a lower-level government that provides public services adapted
to their speci…c preferences, and a higher-level government that reduces border e¤ects and
facilitates trade. Thus, we can express welfare as a function of S and U as follows:

Wl = W F (S; U ) = + (1 + U) S IU , (11)
S

where I U is an indicator variable that takes value 1 if S 6= U , and zero otherwise.

Equation (11) implies that the equilibrium political structure features either P = R or
P 6= R = f[0; 1]g.12 In the …rst case, the world is organized in a single-level governance
structure with a set of countries that provide public services and regulate markets. In
the second case, the world is organized in a two-level governance structure with countries
providing public services and a world union regulating markets.
We now …nd the equilibrium political structure in three steps. First, we compute the
welfare W F (S1 ; S1 ) generated by the single-level governance structure, where S1 is the op-
timal country size without a world union. This political structure takes full advantage of
economies of scope. Country size trades o¤ preference heterogeneity against both economies

11
Throughout, we disregard the constraint that the number of countries and unions, 1=S and 1=U , must
be a natural number. In Appendix A.3.2 we introduce this integer constraint and show that the equilibrium
political strcuture remains symmetric and qualitatively analogous to the tractable approximation we use in
the main text.
12
We know that, if P 6= R, there is only one world economic union because the marginal cost of adding
members is constant and the marginal bene…t is growing with the size of the union. Thus, having many
small unions is not optimal.

13
of scale and facilitating trade: s
S1 = . (12)

The size of countries in the absence of unions is increasing with economies of scale ( ) and
the importance of trade ( ). It is decreasing with preference heterogeneity ( ).13
Second, we compute the welfare W F (S2 ; 1) generated by a world with a union, where S2
is the optimal country size with a world union. This political structure gives up economies of
scope in order to remove border e¤ects and facilitate trade. Country size trades o¤ preference
heterogeneity and economies of scale:
r
S2 = . (13)

The size of countries with a world union is increasing with economies of scale ( ) and
decreasing with preference heterogeneity ( ). Country size is always smaller with a world
union than without it. The reason is that the union removes one of the incentives for
expanding country size, namely, facilitating trade.
The third step is to determine the equilibrium political structure. If W F (S1 ; S1 ) >
W F (S2 ; 1), the world is partitioned into countries of size S1 . If W F (S1 ; S1 ) < W F (S2 ; 1),
the world is partitioned into countries of size S2 that belong to a world union. Naturally,
in the knife-edge case in which W F (S1 ; S1 ) = W F (S2 ; 1), both solutions are equilibrium
political structures. A little algebra shows that the world union is preferred if and only if:
p p
+2 < +2 ( ). (14)

That is, the world union is preferred for high values of , and ; and low values of and .
A world union is more useful if the border e¤ect and trade opportunities are large and there
is substantial preference heterogeneity. A world union is less useful if economies of scale and
scope are sizable.

3.2 Globalization and Political Structure

With these results at hand, we can return to Figure 1 and ask again: Why did the …rst
wave of globalization reduce the number of countries but not generate unions? Why did the

13
Equation (12) assumes that > + , so there is enough preference heterogeneity to ensure that each
country is always smaller than the whole world.

14
 Equilibrium with Law and Diplomacy
2.5

Countries w/o Unions

1.5
κ

World Union

0.5

0
0 0.5 1 1.5 2 2.5 3 3.5 4
γ

Figure 2: Globalization and Political Structure. The …gure shows how equilibrium political
structure depends on economies of scope ( ) and trade opportunities ( ).

second wave of globalization increase the number of countries and lead to the creation of
unions? To answer these questions, we consider a gradual increase in trade opportunities
from 0 to , and we study how political structure changes as this process unfolds.
Figure 2 shows how equilibrium political structure depends on the two parameters that
measure economies of scope ( ) and trade opportunities ( ). For a given , the world chooses
a single-level political structure if is low and a two-level structure if is large. The dashed
arrow in Figure 2 displays a trajectory for the world economy, along a path of incrasing trade
opportunities with constant economies of scope. Equilibrium political structure is initially
single-level, but eventually shifts to two-level as trade opportunities cross a threshold value
U de…ned as follows:
p p
U +2 ( U) = +2 . (15)

If economies of scope are nil, the smallest trade opportunities leads to the formation of a
world union ( = 0 implies U = 0). If economies of scope are prohibitive, the world union
p p
is never an equilibrium ( > +2 ( ) 2 implies U > ). The comparative
statics of this threshold follow directly from our analysis of Equation (14). The larger the
border e¤ect ( ) and preference heterogeneity ( ), the smaller U. The larger economies of

15
 ln S
U

S1*

S 2*

γU γ

Figure 3: Globalization, Countries and Unions. The …gure shows how the world political
structure changes with trade opportunities ( ). The black line is the size of each country,
the green line is the world union.

scale ( ) and scope ( ), the larger U.

Figure 3 displays how political structure changes along the trajectory shown by the
dashed arrow in Figure 2, by plotting the equilibrium size of countries and unions as a
function of trade opportunities ( ). In the …rst wave of globalization ( < U ), it is too
expensive to create a world union, so increases in lead to an increase in country size. The
cost of reaping additional gains from trade is a growing preference mismatch. Eventually,
the preference mismatch has grown so large that it becomes cost-e¤ective to create a world
union. In the second wave of globalization ( > U ), the cost of reaping additional gains
from trade is the loss of economies of scope. The creation of a world union allows countries
to revert to a smaller size and reduce the preference mismatch. Further increases in have
no e¤ect on political structure.
The pattern we derive as a consequence of declining transport costs is borne out by
historical evidence on the two waves of globalization. In our theory, before the creation of
international unions country borders impose constant trade frictions (equal to the border

16
e¤ect ). Consistent with this prediction, O’Rourke and Williamson (1999) argue that the
…rst wave of globalization was driven overwhelmingly by declining trade costs and not by
the liberalization of cross-country trade. In our model, the creation of international unions
endogenously eliminates the trade frictions created by country borders between their mem-
bers. Consistent with this notion, O’Rourke and Williamson (1999) also suggest that trade
liberalization became important after the 1950s— when it was achieved through international
institutions such as the GATT/WTO.14
In deriving our results, we have taken transport costs as exogenous. This need not be
the case. For instance, the model could be extended by adding the possibility of investing
in trade-promoting technologies. Suppose governments can adopt a new technology with a
lower , hence a higher , after paying a …xed cost. Suppose also that the new technology,
once available, is adopted by all countries. This could be through coordination or simply
because new technologies, once discovered, can be copied by others. As we can see from
Equation (11), the bene…t from improving the trade technology is proportional to the size
of countries or unions. Once more, the reason is that the value of lowering trade costs is
proportional to the size of the market that is a¤ected. In turn, a better trade technology
increases the size of countries, but also makes unions relatively more attractive. This suggests
the existence of a complementarity between technological and political globalization. The
adoption of a better trade technology can trigger the formation of a union. But the union
can also make the adoption of such technology cost e¤ective. Moreover, the rise of large
countries may promote the adoption of trading technologies that could ultimately lead to
their collapse. This suggests that the secular increase in trade opportunities need not be a
gradual process, but rather can be marked by abrupt changes.

4 War and Conquest

We explore next how war and conquest a¤ect the relationship between globalization and
political structure. To do this, we assume that the world is divided into core and periphery.
The core contains a measure of localities with a superior military technology that can be

14
It is instructive to compare our results to Alesina, Spolaore and Wacziarg’s (2000) …nding that expanding
trade opportunities always reduce country size. Their analysis di¤ers from ours in two crucial ways. First,
they assume prohibitive economies of scope (in our model, ! 1), so the world always has a single-level
government. Second, they model expanding trade opportunities as an exogenous reduction in the border
e¤ect (in our model, a decline in with constant ). Such a reduction always reduces country size, so their
model can explain why the second wave of globalization led to smaller countries, but it cannot explain why
the …rst wave of globalization led to larger countries.

17
used to conquer other localities and form empires. The periphery contains the remaining
localities that do not have this military technology. We keep all assumptions regarding
preferences, technology and the costs of government. Thus, the model of the previous section
applies in the limit as ! 0.
Empires are an alternative form of government that provides public services and regulates
markets. Each empire contains a metropolis and its colonies. The metropolis consists of core
localities that unite to conquer periphery localities, which then become its colonies.
The downside of empire-building for the metropolis is that waging war and holding the
empire together reduces the utility that it derives from public services by an amount ! > 0.
This cost captures the diversion of government resources from providing public services in
the metropolis to waging colonial wars.
One upside of empire-building is that the metropolis can force its colonies to …nance an
imperial government that provides none of their ideal public services. Instead, an empire
provides the ideal public services of localities in its metropolis, at a lower cost to them thanks
to economies of scale: 8
> 1
< R1 M
if Il=m =1
M
gl (x) = I
0 l=m
dm (16)
>
: 0 M
if Il=m = 0,
M
where Il=m is an indicator variable that takes value 1 if localities l and m belong to the same
metropolis and zero otherwise.
Another gain from empire is that the metropolis expropriates market goods from it
colonies. We assume it loots a share 2 [0; 1] of the colonies’ output in the indus-
tries that do not rely on contract enforcement (a share 1 of the total).15 As a re-
sult, its consumption of the full measure of varieties in those industries rises to cl (m; i) =
R1 E R1 M E
e 1 + 0 Il=n dn= 0 Il=n dn where Il=n is an indicator variable that takes value 1 if
localities l and n belong to the same empire and zero otherwise.
The last bene…t of building an empire is that it enables trade in the industries that
require legal enforcement of contracts (a share of the total). The metropolis enjoys the
gains from trade creation that result from common market regulation in its empire, while it
incurs reduced costs of preference mismatch. Overall, the welfare of a member of an imperial

15
This assumption simpli…es the algebra, but our results remain qualitatively unchanged if we assume that
looting can take place in all industries, as we show in Appendix A.4.1.

18
metropolis is:
" R1 !# Z
E 1
Il=m dm E
Wl = + (1 ) + ln 1 + R01 M + Il=m dm
I dm
0 l=m 0
Z 1
M
Il=m dm R1 E
!. (17)
0 0
Il=m dm

From the perspective of the conquered colonies, instead, the gains from trade creation
are dwarfed by the costs of looting and especially of subjugation to an imperial government
that generates an unbounded preference mismatch. This division of the surplus captures in
our model the extractive nature of imperialism: the metropolis enjoys the gains from trade,
while the conquered localities su¤er under an exploitative colonial administration.
Periphery localities are incapable of resisting the core’s superior military technology.
Thus, a periphery locality remains free if and only if no core locality wishes to colonize it.
This may be the case in equilibrium, because each empire has a limited reach. We assume
that a metropolis of size M can subjugate a set of colonies of size M . This assumption
captures two considerations. The …rst is technological. Country size is important for military
success, so one of the reasons countries grow large is to prepare for war. The second is
ideological. As the disproportion between the size of the metropolis and its colonial targets
grows, the aggressive and undemocratic nature of imperialism becomes more obvious, and
eventually unpalatable on moral grounds for the citizens of the metropolis. Either force
alone su¢ ces for our results.
The equilibrium size of an empire does not depend only on the metropolis’s ability and
willingness to use violent repression in its colonies. It is also limited by competition between
rival empires, which may clash in the attempt to conquer the same territories. If the sum
of desired colonies is larger than the periphery, empires cannot avoid clashing. We assume
they then carve up the entire periphery into empires whose size is proportional to the size of
the respective metropolises. If instead the sum of desired colonies is smaller than the entire
periphery, there is no need for con‡ict between empires. We assume that rival metropolises
are then capable of avoiding unnecessary clashes.16

16
Essentially, this is the outcome of two assumptions. First, each metropolis can target which periphery
localities it tries to colonize. Second, if a periphery locality is targeted by multiple empires, each of them
has an equal probability of colonizing it.

19
4.1 Equilibrium Political Structure

With war and conquest, equilibrium political structure need no longer be globally e¢ cient.
Formally, equilibrium political structure is now determined in two stages.

1. Core localities choose whether to wage war and build empires. Localities in the pe-
riphery may become their colonies or remain free.

2. Localities that do not belong to an empire choose their political structure through
e¢ cient bargaining, as in our baseline model.

The world’s equilibrium political structure now consists of a set of empires that have a
combined size 1 F ; plus two partitions (P; R) of the free world which itself has a combined
size F . We solve for this equilibrium political structure in two steps. First, we determine
the political structure of the free world (P; R) for a given size F . Second, we determine the
number and size of empires and therefore the size of the free world F .17

4.1.1 The Free World

The analysis of the free world is essentially the same as in the previous section. The only
di¤erence is that now the combined size of the free world is F rather than 1. E¢ cient
bargaining ensures that free localities choose the optimal political structure. Equation (11)
still applies and, as a result, there are two cases to consider: S = U and S < U = F . The
optimal country sizes in these cases are still given by Equations (12) and (13), respectively.18
The union of the free world is preferred if and only if:
p p
+2 < F +2 ( ). (18)

Condition (18) generalizes Condition (14) for the case of a free world of size F . The main
di¤erence is that empires reduce the size of the free world. This reduction lowers the welfare
associated with a union of free localities. This union still costs to each member. But it is
now less e¢ cient at removing border e¤ects ( F instead of ).

17
For simplicity, we do not consider the possibility that empires could be members of international unions.
Appendix A.4.2 shows that our results are robust to relaxing this assumption.
18
We now assume that > F 2 + to ensure that countries are always smaller than the free world.

20
 4.1.2 Empires

Core localities must …rst decide whether to wage war and build an empire, or to forego war
and enter the free world. We start our analysis with two observations. First, a metropolis
always conquers all the colonies it can. Adding extra colonies is desirable from its perspec-
tive because it lowers the cost of government, enables looting and facilitates trade without
creating any preference mismatch (for the metropolis) in the provision of public services.
Second, the equilibrium political structure generically features either no empire, or empires
whose metropolises span the entire core.
These two observations imply that there are two cases. If 1= (1 + ), each metropolis
conquers colonies of size M , but empires do not need to clash because there remains a
free world F = 1 (1 + ) 0 that no empire is capable of subjugating. If instead >
1= (1 + ), there is con‡ict between empires and the entire periphery is colonized (F = 0).
Each metropolis carves out a share of it proportional to its share of the core, so its colonies
have size (1 ) M= . Summarizing, this implies that the size of each metropolis relative
to its empire is:
M 1
= max ; < 1. (19)
E 1+
The welfare of a core locality that builds an empire is:

1
Wl = W E (E) = + (1 ) + ln 1 + + E E !. (20)
E

The size of an empire trades o¤ preference heterogeneity against both economies of scale and
facilitating trade: s
E = , (21)

where E is the optimal empire size (for the the core localities).19
Comparing Equation (21) to Equations (12) and (13), we see immediately that empires
are larger than peaceful countries. The reason is that the metropolis does not internalize the
cost of the preference mismatch it imposes on its colonies: hence, appears instead of in
the denominator. An intuitive implication is that an empire is larger when its metropolis can
be smaller relative to its colonies ( ). Hence empires are smaller if the core is so large ( )
that empires clash in equilibrium and box each other in; or if they are su¢ ciently democratic

19 2
Equation (21) assumes that > ( = ) + , so there is enough preference heterogeneity to ensure
that the metropolis is smaller than the whole core and a fortiori the empire is smaller than the whole world.

21
they cannot repress a large share of their population ( ).
The equilibrium size of empires is also decreasing with preference heterogeneity ( ) and
increasing with the importance of trade ( ) and with economies of scale ( ). These com-
parative statics are analogous to those for peaceful countries. An empire is distinguished by
its extractive institutions, but these determine the equilibrium size of empires only through
economies of scale for the imperial government.
When are empires formed? If core localities wage war and build empires, their welfare
is W E (E ). If core localities instead refrain from waging war and choose to form countries
and unions by e¢ cient bargaining, their welfare is max W F (S1 ; S1 ) ; W F (S2 ; 1) .
If W E (E ) > max W F (S1 ; S1 ) ; W F (S2 ; 1) , empire-building is a dominant strategy
for core localities. Thus, in equilibrium they form = ( E ) empires of size E , and the size
of the free world is reduced to F = 1 = . Some algebra shows that empire-building is the
unique equilibrium if and only if:

1 p
! (1 ) ln 1 + +2 ( )
n p p o
< min 2 ( ); + 2 . (22)

Unsurprisingly, core localities wish to build empires if colonial wars are e¤ective enough: i.e.,
if their military cost (!) is low and their looting bene…ts ( ) are high, while the size of the
metropolis ( ) is small relative to the empire, because the core ( ) is small enough empires
do not need to clash and their government is su¢ ciently autocratic ( ) to subjugate a large
share of the empire’s population.
If instead W E (E ) max W F (S1 ; S1 ) ; W F (S2 ; 1) , there is always an equilibrium in
which no empires are formed, diplomacy rules and the size of the free world is F = 1. This
peaceful equilibrium need not be unique, though. A Pareto-ine¢ cient second equilibrium
also exists if W F (S2 ; 1) W E (E ) max W F (S1 ; S1 ) ; W F (S2 ; 1 = ) . In this case,
Pareto e¢ ciency requires all core localities to forego empire-building and create a peaceful
world union. Yet, if core localities expect other core localities to build empires, their best
response is to build an empire themselves. Empire-building is then a coordination failure
that lowers the welfare of every locality in the world; but it is also an equilibrium because
once it happens there is no incentive for a single core locality to abandon its empire and join
the free world.

22
 Equilibrium with War and Conquest
3

2.5

Empires

Peaceful Countries
2

1.5
κ

1 World Union

0.5

0
0 0.5 1 1.5 2 2.5 3 3.5 4
γ

Figure 4: Globalization, Con‡ict and Political Structure. The …gure shows how equilibrium
political structure depends on economies of scope ( ) and trade opportunities ( ).

4.2 Globalization and Political Structure

Returning to the relationship between globalization and political structure, Figure 4 shows
again how equilibrium political structure depends on the two parameters that measure
economies of scope and trade opportunities (i.e., and ). For simplicity, we disregard
Pareto-dominated equilibria and focus throughout on the unique Pareto-dominant equilib-
rium.Then there is an age of empires if and only if the following condition holds:

1 p p p
! < (1 ) ln 1 + +2 U U . (23)

Otherwise, empires are never built and the world is always ruled by law and diplomacy.
The dashed arrow in Figure 4 displays again a trajectory for the world economy along a
path of incrasing trade opportunities, with constant economies of scope. Initially the whole
world is free and there is no union. As trade opportunities reach a …rst threshold value
L > 0, empires are formed. Eventually, as trade opportunities reach a a second threshold
value H < , empires are abandoned and a world union is formed. This evolution is not

23
generic, though. A necessary and su¢ cient condition for L > 0 is that:

1 p p
! > (1 ) ln 1 + +2 (1 ). (24)

If this condition holds, core localities do not conquer colonies for the extractive bene…ts alone.
Empires become cost-e¤ective for the metropolis only when they have su¢ cient additional
bene…ts from trade creation: hence, after trade opportunities have grown enough.
Likewise, a necessary and su¢ cient condition for H < is that:

1 p p p
! > (1 ) ln 1 + +2 + . (25)

If this condition holds, expanding trade opportunities make empires grow so large that the
preference mismatch within the metropolis alone becomes enough to justify a move to a
two-level government. Perhaps surprisingly, the cause of imperial collapse at a late stage of
globalization is the same as the cause for the rise of empires at an early stage: namely, the
desire to remove border e¤ects and reap the gains from trade. It is just that, at some point,
it becomes more cost-e¢ cient to replace conquered colonies with free partners in a union.
Figure 5 displays how political structure changes along the trajectory shown by the
dashed arrow in Figure 4. At low levels of globalization ( < L ), the world contains only
free countries. There are no empires nor unions. As trade opportunities expand, the size
of countries grows. When trade opportunities cross the …rst threshold ( L < < H ), core
localities prefer to build empires. Empires are larger than countries and keep growing as
globalization proceeds. Eventually, trade opportunities cross the second threshold ( > H ).

Empires collapse and countries revert to a smaller size. A world union is created. After this,
there are no further changes in political structure.20

4.3 Other Drivers of Political Structure

Our analysis has focused on growing opportunities for trade as the explanation for the non-
monotonic evolution of political structure. In our model as in the real world, however,
several additional factors shape the number of countries and the decision to form colonial

20
Figure 5 depicts the case in which, during the age of empires, the free world always adopts a single-
level governance structure and the shift to the two-level governance structure coincides with the collapse of
empires. This need not be the case. If the size of the free world is large enough, an economic union of free
countries co-exists with empires.

24
 ln S
U

E*

S1*
S 2*

γL γH γ

Figure 5: Countries, Empires and Unions. The …gure shows how the world political structure
changes with expanding trade opportunities ( ). The black line is the size of peaceful
countries, the red line is the size of empires, the green line is the world union.

empires or international unions. A natural question is whether these other factors alone could
be driving the patterns we observe in historical data. In particular, could the three-stage
evolution of political structure observed in history and depicted in Figure 5 be explained
purely by changes in the costs and bene…ts of war and conquest?
To answer this question, Figure 6 shows how equilibrium political structure depends on
the cost of war (!) as well as trade opportunities ( ). The rise and subsequent fall of empires
could be explained by a decline and subsequent increase in ! alone. Such a non-monotonic
evolution of the cost of war plausibly …ts the historical consequences of advancing military
technology. In the nineteenth century, it enabled the great powers to make easy colonial
conquests. Conversely, in the twentieth century it led them to incur horri…c costs in two
World Wars, and eventually to develop nuclear weapons and the doctrine of mutually assured
destruction.
Yet, changes in the cost of war alone— i.e., movements along a vertical trajectory in
Figure 6— cannot explain why what follows empires (international unions) is di¤erent from

25
 Equilibrium with War and Conquest
0.6

0.55

0.5 Peaceful Countries

World Union
0.45

0.4
ω

0.35

0.3

0.25

Empires
0.2

0.15

0.1
0 0.5 1 1.5 2 2.5 3 3.5 4
γ

Figure 6: Trade, War and Political Structure. The …gure shows how equilibrium political
structure depends on the cost of war (!) and trade opportunities ( ).

what preceded them (peaceful countries). Expanding trade opportunities are necessary to
account for this pattern. They are also su¢ cient, as Figure 6 con…rms by plotting once again
as a dashed arrow the straight trajectory along a path of increasing trade opportunities, with
a constant cost of war. Nonetheless, joint changes in and ! along a U-shaped trajectory—
displayed as a dash-dotted arrow in Figure 6— paint a more realistic picture. The two
drivers of political structure are complementary: changes in the cost of war accelerate the
consequences of changes in trade opportunities. In the nineteenth century, empires arise
earlier as conquering colonies becomes cheaper. In the twentieth century, international unions
arise earlier as war becomes costlier.
This complementarity sheds light on the relationship between peace, trade and interna-
tional organizations. In our model, a goal of international unions is to reduce the appeal
of costly wars. Empirically, this was a key real-world motivation for creating supranational
institutions like the United Nations and the European Union. Figure 6 helps understand why
they were created only after World War II. By that point, trade opportunities had grown
large enough that a positive shock to the cost of war su¢ ced to trigger a switch from the
age of empires to that of peaceful unions.21

21
Our theory also points to a non-monotonic complementarity between military and trade technologies.

26
 The rise and fall of empires may re‡ect not only the gradual improvement of military
technology, but also its gradual spread. Our model captures the traditional view that con‡ict
among rival empires was caused by the spread of industrialization and the attendant rise of
new great powers on the world scene. As more localities acquire the technology to conquer
empires, the size of the core ( ) expands. So long as the core remains small enough ( <
1= (1 + )), rival powers can secure their desired colonies without clashing. Accordingly,
Britain and France could carve out two world-straddling empires without …ghting each other,
the Fashoda incident notwithstanding. In this case, the spread of great-power technology
increases the number of empires without a¤ecting their size. However, once technology
di¤usion reaches a critical threshold ( > 1= (1 + )) rival powers cannot avoid costly clashes
as they try to secure scarce colonies for themselves and deny them to each other. Accordingly,
the rise of imperial powers eventually came to be marked by wars like the Spanish-American
war and the Russo-Japanese war— and it was arguably a cause of both World Wars.
In our model, such clashes unambiguously reduce the size of the empire each core lo-
cality can conquer (@E =@ < 0). At the same time, they force a metropolis to remain
relatively large, so as to remain competitive with rivals in the military domain (@M =@ >
@E =@ ).22 Intuitively, costly great-power rivalry thus makes imperialism less appealing
(@W E (E ) =@ < 0). Just as an increase in the cost of warfare, a greater prevalence of
clashes between imperial powers (such as the World Wars) hastens the demise of empires
(@ H =@ < 0). Once again, though, the simultaneous expansion of trade opportunities is
not only a complementary driver of decolonization, but also the factor required to explain
the creation of international unions.
Finally, recall that the parameter can be interpreted as the limit to imperialism arising
from moral concerns among citizens of the metropolis. A decline in this limit can represent
the spread of democratic values. In our model, a decline in makes empires weakly smaller
(@E =@ 0) and less desirable (@W E (E ) =@ 0). This e¤ect is always present if

In an early stage of globalization ( < U ), trade-enabling technology (higher ) complements o¤ensive

war technology (lower !). An improvement in either tempts core localities to forego diplomacy and peace
and embark instead on imperial conquest, thereby raising the value of an increase in the other. In a later
stage of globalization ( < U ) the pattern switches to complementarity between trade-enabling technology
and defensive war technology (higher !), both of which hasten the demise of empires and the advent of
peaceful unions. While a dynamic theory of endogenous technological progress remains beyond the scope
of our model, these considerations suggest that a realistic U-shaped trajectory in Figure 5 could derive
from directed technical change, with causal feedback loops between trading technology and di¤erent types
of military technologies.
22
Great-power rivalry can prove so demanding that
n an imperial metropolis
o needs to expand while its
p
empire shrinks: @M =@ > 0 > @E =@ if > max 1= (1 + ) ; 2 = .

27
the binding constraint on imperialism arises from moral concerns in the metropolis ( <
1= (1 + )). If instead the binding constraint is competition between the great powers, the
e¤ect only arises if the increase in democratic values is large enough. Since competition
between empires grew over time, our theory suggests that democratization may have played
a greater role in slowing the rise of empires in the nineteenth century than in hastening their
demise in the twentieth.

5 Regional Unions

Perhaps the most unrealistic aspect of our baseline model is its lack of gradualism. If core
localities join a union, it is a world union. If a world union is created, country size is reduced
at once to its autarky level. However, a simple generalization of our theory shows that this
need not be the case. Sometimes, regional unions precede world unions.23
To introduce regional unions, we now consider a slightly richer treatment of geography.
Assume that, in addition to having a superior military technology, core localities are near
each other. Transportation costs for core-core trade are (with > 0), so the gains
from core-core trade are now + . Transportation costs for core-periphery and periphery-
periphery trade remain , so the gains from these types of trade are still . The rest of our
assumptions remain those of Section 4.
Appendix A.3.3 provides a full analysis of this model. Here we just state the new results
it generates. Technically, the key implication of this new assumption is that Equation (4),
indicating the utility from consuming market goods, must be replaced by:
Z 1 Z
WlM = + 1 + R
Il=m dm + Il2C (1 ) + R
Il=m dm , (26)
0 0

where Il2C is an indicator variable that equals 1 if l 2 C and zero otherwise. Border e¤ects
are larger for core-core trade ( ( + )) than for either core-periphery or periphery-periphery
trade ( ). Other things equal, core localities prefer sharing economic regulation with other
core localities. Periphery localities, instead, are indi¤erent about which localities they share
regulation with and care only about their total number.
With this additional assumption, a mixed political structure with a two-level government
in the core and a single-level government in the periphery becomes possible. Figure 7, which

23
An alternative form of gradualism is the creation of unions with increasing depth of international inte-
gration. Appendix A.4.3 provides a formalization of this idea.

28
 Equilibrium with Regional Unions
1.6

1.4

Peaceful Countries Empires

1.2

0.8
κ

0.6 Core Union

World Union

0.4

0.2

0
0 0.5 1 1.5
γ

Figure 7: Gradualism and Regionalism. The …gure shows how equilibrium political structure
depends on economies of scope ( ) and trade opportunities ( ).

assumes a large enough value for , shows a scenario in which a four-fold evolution takes
place along the trajectory of expanding trade opportunities depicted as usual by the dashed
arrow. The …rst stage is peaceful single-level government everywhere. The second is the
creation of colonial empires through which the core conquers and rules distant localities in
the periphery. As trade opportunities expand further, empires become overstretched and
collapse, and a peaceful union replaces them. The novelty of this third stage is that, unlike
in Figure 4, this is now a core union. Indeed, for the most part, in the real world core
countries replaced empires with freer trade not with their former colonies, but with other
core countries. Only as trade opportunities expand further a fourth and …nal stage is reached
in which the whole world forms a union.
The model could be further extended in fruitful ways. One extension is a world with
two, three or N peripheries that are located progressively farther away. In this world, there
is a union that starts at the core and grows outwardly with expanding trade opportunities.
When the …rst periphery joins the union, the size of its countries declines. When the second
periphery joins the union, the size of its countries also declines, and so on. The union
gradually advances outward and it keeps breaking up countries. The end point is the same
as in our baseline model, but the world approaches it gradually.

29
 A second extension is a world with two or more core-periphery structures, which we
can think of as continents or regions. In this world, within each region there is one union
that advances outward, breaking up countries. Across regions, however, there is no union
initially. Eventually, trade opportunities may expand so much that a world union becomes
cost-e¤ective, and the regional unions merge. The world approaches the same end point
again, but it now approaches it both gradually and regionally.
The bottom line is that asymmetric geography can explain the gradual appearance of
international unions with a limited geographic scope. As distance becomes less and less im-
portant, the world gradually converges to the single international union we analyzed earlier.

6 Historical Evidence on Trade and Country Size

Our theory provides a new perspective on the connection between trade, country size and
the emergence of international unions. Our motivation for developing this theory was to
improve our understanding of global trends, as presented in Figure 1. But the mechanisms
that connect trade and country size over time should also be at work when we compare the
trajectories of di¤erent countries. The extension with regional unions in Section 5 considers
trade costs that vary across countries. Moreover, localities could di¤er in their costs of
government, of waging war and of forming unions. This would introduce heterogeneity in
country size and also imply that di¤erent countries may join unions at di¤erent times.
We exploit such heterogeneity by examining cross-country historical data.24 Our empir-
ical evidence show that (i) increases in the volume of trade predict increases in country size
in the absence of international unions, but (ii) the creation of international unions weak-
ens or eliminates this pattern, and instead (iii) membership of international unions predicts
decreases in country size. These …ndings, though far from conclusive, suggest that the mech-
anisms our theory highlights are empirically relevant.

6.1 Data

We draw our data from the Cross-National Time-Series (CNTS) Data Archive. This dataset
provides information about land area and the volume of trade, measured as the sum of
imports and export per capita, for an unbalanced panel of countries with observations from

24
In a more speculative vein, in Appendix A.5 we provide a brief qualitative narrative showing how our
theory helps interpret historical events.

30
 Table 1: Descriptive Statistics.
Period Countries Land Area Expansion Contraction WTO ∆ Trade
Number Mean Mean Mean Mean Mean Std. Dev.
1870-1880 43 738,564 0.233 0.023 0.000 0.280 0.421
1880-1890 45 687,672 0.244 0.044 0.000 0.271 0.430
1890-1900 46 674,440 0.217 0.109 0.000 0.084 0.303
1900-1910 46 670,962 0.087 0.130 0.000 0.403 0.346
1910-1920 47 681,729 0.234 0.085 0.000 1.410 1.262
1920-1930 55 604,987 0.164 0.073 0.000 0.466 1.041
1950-1960 73 536,990 0.041 0.068 0.425 0.880 0.971
1960-1970 102 450,449 0.010 0.049 0.564 1.043 1.268
1970-1980 127 377,772 0.024 0.055 0.591 4.613 4.504
1980-1990 149 274,399 0.020 0.020 0.624 0.276 0.668
1990-2000 158 257,461 0.006 0.019 0.800 1.851 14.732
2000-2010 169 294,461 0.006 0.012 0.845 1.698 1.327
Notes: Countries and Land Area are measured at the beginning of each decade. Land Area is expressed in
thousand square miles. Expansion and Contraction are dummies taking value 1 if the land area of an existing
country expandend or contracted, respectively, over the decade. WTO is a dummy taking value 1 if a country
is a member of GATT/WTO at the beginning of each decade.

1870 to 2010.25
Given that land area changes slowly and discontinuously, we focus on decades and build a
dummy variable that takes value 1 if a country or empire has grown in size over the previous
ten years. We interpret averages over this variable as the “probability” of a territorial
expansion. Similarly, we build dummy variables for territorial contraction that take value
1 if the land area of a country or empire has fallen over the previous decade. We interpret
averages over this variable as the “probability”of a territorial contraction.
Table 1 contains descriptive statistics for the main variables of interest. For each decade,
it reports the number of countries with non-missing observations, their average land area, the
share of countries that expanded and that contracted their territories, the share of countries
that are members of the WTO (GATT before 1995), the average change in the volume of
trade and its standard deviation. Given that all years corresponding to the world wars have
no observations in the dataset, the decades around 1940 are missing.
A quick look at Table 1 con…rms the basic trends already discussed. Territorial expansions
are much more frequent than contractions in the …rst part of the sample, but the opposite is

25
Before 1870, trade data is missing for the majority of countries.

31
 800 2500

700
2000

600
1500

500

1000
400

500
300

200 0
1830
1835
1840
1845
1850
1855
1860
1865
1870
1875
1880
1885
1890
1895
1900
1905
1910
1920
1925
1930
1935
1946
1951
1956
1961
1966
1971
1976
1981
1986
1991
1996
2001
2006
country empire

Figure 8: The Size of Countries and Empires. The …gure plots the average size of countries
(left axis) and of empires (right axis) in thousand squared miles. See Appendix A.1 for
details on data.

true after 1950. The table also reports the average growth in the volume of trade per capita.
It shows that trade grows throughout the entire period, but at di¤erent speeds both over
the decades and across countries.26
Figure 8 shows the average land area of “internationally recognized” countries and of
thirteen major empires since 1830. The size of empires increases until World War II and falls
thereafter. Empires, besides being larger than countries, start reducing their size somewhat
later than countries. This picture is consistent with the data on the number of countries
shown in Figure 1. Despite the di¤erent data sources, both …gures tell a remarkably similar
story.27 During the nineteenth century there was a phase of political concentration in which
countries and empires expanded their territories. But this trend reversed during the twentieth

26
Caution should be taken when comparing some of these variables across decades since the composition
of the sample is changing.
27
Figure 8 is based on the “International System”(Singer and Small 1966), which includes countries with
international recognition and reliable data. See Appendix A.1 for more details. This classi…cation, however,
underestimates the number of independent political entities in the nineteenth-century developing world.
Figure 1 is instead based on Butcher and Gri¢ ths (2013), who provide a more comprehensive record.

32
century, and especially after World War II.
Finally, recall that our theory also implies that territorial changes should be associated
with con‡ict during the age of empires, but be more peaceful in era of international unions.
Data from the Correlates of War project shows that indeed before 1950 more than one third
of all territorial disputes involved military con‡ict, while after that date border changes were
peaceful in almost 90% of cases.

6.2 Regression Analysis

We start by studying how the probability of observing a territorial expansion depends on

changes in the volume of trade, by running a series of binomial regressions. To alleviate
simultaneity, we compute the change in the volume of trade over the previous decade. Fur-
thermore, to check if the correlation between trade and territorial expansion changes after
World War II, we include an interaction term between changes in the volume of trade and a
post-1945 dummy.
Table 2 reports the main results of estimating a logit model. Standard errors are corrected
for clustering by country, to accommodate autocorrelated shocks at the country level. In
column 1, we start by including as regressors the lagged change in the volume of trade and
its interaction with the post-1945 dummy. As predicted by our theory, the coe¢ cient for the
lagged change in trade is positive, the one for the interaction term is negative, and both are
highly signi…cant. In other words, trade predicts territorial expansion before World War II,
but not after it. In column 2, we also include the post-1945 dummy. The inclusion of the
constant and the post-war dummy mean that the identifying variation is deviations from the
global trends visible in Figures 1 and 8. Nevertheless, the two coe¢ cients of interest remain
signi…cant. In column 3, we add year …xed e¤ects and we keep them in the rest of the table.
While these time dummies control for shocks a¤ecting all countries equally, including some
common e¤ects of globalization, they do not a¤ect our main results. In fact, the coe¢ cients
for the trade variables become even more signi…cant.
We then add some control variables inspired by our model. In column 4, we add the log
of population and the urbanization rate at the beginning of each decade.28 The …rst variable
controls for the e¤ect of size while the second is a proxy for economic development, and both
are likely to be correlated with the military strength of a country.29 Both coe¢ cients are

28
These variables are also sourced from the CNTS Data Archive. The urbanization rate is the fraction of
population living in urban centers of at least 100,000 inhabitants.
29
More direct measures, such as GDP, are not available for the entire period of analysis.

33
 Table 2: Trade and Territorial Expansion.
Dependent variable: Expansion dummy
All All All All All All Pre1945 Post1945
(1) (2) (3) (4) (5) (6) (7) (8)

∆ Trade 0.818*** 0.285* 0.463* 0.607** 0.545** 0.650* 0.577** -0.179

[0.189] [0.176] [0.257] [0.239] [0.269] [0.335] [0.290] [0.142]
∆ Trade × Post1945 -1.294*** -0.287* -0.580* -0.896*** -0.776** -1.636**
[0.314] [0.177] [0.300] [0.314] [0.318] [0.819]
Post1945 -2.724*** -5.262*** -5.626*** -4.850*** -2.859
[0.472] [1.242] [1.566] [1.353] [2.405]
Log Population 0.595*** 0.599*** -2.053 0.640*** 0.460*
[0.141] [0.141] [1.702] [0.161] [0.271]
Urbanization Rate 0.002 0.002* 0.003 0.003 0.001
[0.001] [0.001] [0.004] [0.002] [0.002]
∆ Democracy -0.166** -0.159 -0.267** 0.113
[0.076] [0.132] [0.111] [0.103]

Country FE No No No No No Yes No No
Time FE No No Yes Yes Yes Yes Yes Yes
Observations 822 822 822 799 651 227 212 439
R² 0.090 0.218 0.260 0.362 0.326 0.386 0.214 0.102
Notes: All observations refer to 10-year periods. The dependent variable is a dummy taking value 1 if the country's land area
expanded over the decade and 0 otherwise. ∆ Trade and ∆ Democracy are changes over the previous decade. Post1945 is a
dummy for decades after 1945. All other variables are measured at the beginning of each decade. Constant always included
and Pseudo-R ² reported. Standard errors, clustered by country, are in brackets. *, ** and *** denote significance at 10%, 5%
and 1% respectively.

positive and signi…cant, but their inclusion does not a¤ect the trade variables. In column
5 we add the change in the level of democracy over the previous decade.30 Consistent with
our model and other existing theories, we …nd that democratization lowers the probability
of territorial expansion. Yet, its inclusion has almost no e¤ect on the coe¢ cient for trade. In
column 6 we add country …xed e¤ects. This speci…cation is quite demanding: all countries
with no changes in size are dropped and the coe¢ cients are identi…ed only from within-
country deviations from country-speci…c trends. As a result, sample size falls markedly.
Nonetheless, trade and its interaction with the post-1945 dummy are the only two coe¢ cients

30
We use the polity2 index sourced from the Polity IV database. Unfortunately, this index is missing for
some of the country-year cells in our dataset.

34
that remain signi…cant.
Finally, we allow for heterogeneity in all coe¢ cients before and after World War II. To this
end, in columns 7 and 8 we split the sample and re-estimate the speci…cations without country
…xed e¤ects separately for the decades before and after 1945. The results con…rm that trade
predicts territorial expansion in the …rst part of the sample, but not in the postwar era. As to
the remaining coe¢ cients, population has a larger and more signi…cant coe¢ cients in the pre-
1945 period, while democracy becomes insigni…cant after 1945. A possible interpretation of
these results is that variables capturing military strength or tolerance for aggression become
less relevant when territorial changes are more peaceful. However, the lack of signi…cance
may also indicate low statistical power due to the very few territorial expansions observed
in the post World War II period. To study how trade correlates with border changes after
1945, we need to turn to territorial contractions, which become relatively more frequent.
In Table 3, we run similar binomial regressions to study how the probability of territorial
contraction depends on changes in the volume of trade and on being part of an international
economic union, measured by WTO membership. As in the previous table, the change in
the volume of trade is computed over the previous ten years, the WTO dummy is measured
at the beginning of each decade, and standard errors are corrected for clustering by country.
We restrict the analysis to the period after 1945, when the WTO dummy starts to have
positive values, but results are similar if earlier decades are added.31 In column 1, we
include the lagged change in the volume of trade and the WTO dummy. Only the coe¢ cient
for the latter is positive and precisely estimated, indicating that countries joining the WTO
have a higher probability of a subsequent territorial contraction. In column 2, we add the
interaction between changes in trade and the WTO dummy. The coe¢ cient for trade turns
positive and become statistically signi…cant. When we add year …xed e¤ects, in column 3,
the negative coe¢ cient of the interaction term also become signi…cant. Growth in trade
is followed by a higher probability of territorial contraction, but not after a country joins
the WTO. This …nding suggests that the negative e¤ect of trade on country size may be
mediated by membership of international unions like the WTO, as our theory predicts.
In column 4 we add the log of population and the log of GDP per capita at the beginning of
each decade. These proxies for size and economic development, which where associated with
territorial expansion before World War II, are now correlated with territorial contraction.

31
Before 1950, territorial contractions are infrequent and occur in periphery countries, for which data are
scarce and less reliable. If we perform the analysis over the entire sample, we obtain similar results, but
these are driven by the observations after 1950. Hence, adding more decades is not very informative.

35
 Table 3: Trade, Unions and Territorial Contraction.
Dependent variable: Contraction dummy
Post1945 Post1945 Post1945 Post1945 Post1945 Post1945
(1) (2) (3) (4) (5) (6)

∆ Trade -0.003 0.097* 0.146*** 0.105** 0.104** 0.080

[0.011] [0.055] [0.051] [0.052] [0.052] [0.058]
WTO 1.555** 1.594** 2.120*** 1.744** 1.785** 2.423**
[0.724] [0.760] [0.754] [0.884] [0.886] [1.139]
∆ Trade × WTO -0.113 -0.146** -0.234 -0.251 -0.468
[0.072] [0.057] [0.284] [0.287] [0.355]
Log Population 0.503*** 0.490*** 0.558***
[0.143] [0.150] [0.210]
Log GDP per capita 0.549*** 0.557*** 0.140
[0.188] [0.205] [0.412]
∆ Democracy -0.106 -0.166
[0.139] [0.220]

Region FE No No No No No Yes
Time FE No No Yes Yes Yes Yes
Observations 588 532 532 530 486 355
R² 0.038 0.032 0.155 0.248 0.239 0.255
Notes: All observations refer to 10-year periods. The dependent variable is a dummy equal to 1 if the country's
land area contracted over the decade and 0 otherwise. WTO is a dummy for WTO/GATT membership. ∆ Trade
and ∆ Democracy are changes over the previous decade. All other variables are measured at the beginning of
each decade. Constant always included and Pseudo-R² reported. Standard errors, clustered by country, in
brackets. *, ** and *** denote significance at 10%, 5% and 1% respectively.

36
This suggest that the countries or empires that where growing before 1940 may be precisely
those that start to break up after it. In column 5 we also add the change in the democracy
index. While democratization is followed by a lower probability of territorial expansion in
the era of empires, it has no statistically signi…cant correlation with territorial contraction
after 1950. Adding these controls barely a¤ects the coe¢ cient for the WTO dummy, while
its interaction with trade becomes statistically insigni…cant. Country …xed e¤ects are not
feasible because they would shrink our sample size to a mere 63 observations. Instead, in
column 6 we add region …xed e¤ects. While the estimated coe¢ cient for the change in
the volume of trade loses precision, the WTO dummy remains positive and statistically
signi…cant. In sum, throughout all speci…cations, the most robust result is the positive
correlation between WTO membership and subsequent territorial contractions.
We have shown three signi…cant patterns in historical data since 1870. First, increases
in trade predict territorial expansion before international unions. Second, this correlation is
reversed after international unions are created. Third, membership of international unions
predicts territorial contraction. To the best of our knowledge, these empirical …ndings are
novel in the literature. While not conclusive, they are remarkably consistent with our the-
ory. The predictions of our model are also consistent with other empirical …ndings already
documented elsewhere. Besides the well-documented trade-creating e¤ect of international
unions, of particular interest is Martin, Mayer and Thoenig’s (2012) …nding that regional
trade agreements promote peaceful relations between member states.

7 Conclusions

In this paper we have studied the interaction between globalization and political structure.
We have shown that the expansion of trade opportunities can help explain two salient aspects
of the evolution of political structure over the last couple of centuries: (i) the rise and
subsequent fall in the size of countries observed during the nineteenth and twentieth century,
and (ii) the seemingly contradictory trends towards more political fragmentation at the
national level and more political integration at the supranational level in the second half
of the twentieth century. We have also provided some historical evidence in support of this
claim. Yet, we have deliberately left aside several important factors. We now brie‡y mention
three that seem particularly promising for future research.
First, we have modeled international unions as mere economic agreements to facilitate
trade. Although this simpli…cation provides a useful and realistic starting point, it does

37
not do justice to the full political scope of unions, which in reality also provide public
goods and coordinate policy choices with cross-border consequences.32 Hence, supranational
policymaking involves distributive con‡ict that we have abstracted from, and which can help
explain gradual changes in union membership.
Second, our concept of locality abstracts from internal heterogeneity in both preferences
and economic attributes. Yet, historical experience suggests that internal con‡ict has played
a role in the process of country formation and break-up (Bolton and Roland 1996, 1997).
It would be interesting to explore how expanding trade opportunities also a¤ect political
structure through their e¤ect on domestic heterogeneity and con‡ict. We know, for instance,
that international unions may exacerbate domestic con‡ict and shape political preferences
(Gancia, Ponzetto and Ventura 2020).
Finally, we have focused on the economic aspects of globalization: an expansion of trade
opportunities and how political structure adjusts to take advantage of them. Yet, there are
also cultural aspects of globalization, which may entail for instance a reduction in preference
heterogeneity. Exploring the simultaneous e¤ects of both economic and cultural drivers of
globalization on political structure seems a fruitful avenue for further research.

References

[1] Alesina, Alberto, Ignazio Angeloni and Federico Etro (2005). “International Unions,”
American Economic Review 95(3): 602–15.

[2] Alesina, Alberto, and Enrico Spolaore (1997). “On the Number and Size of Nations,”
Quarterly Journal of Economics 112(4): 1027–1056.

[3] — — (2003). The Size of Nations. Cambridge, MA: MIT Press.

[4] Alesina, Alberto, Enrico Spolaore and Romain Wacziarg (2000). “Economic Integration
and Political Disintegration,”American Economic Review 90(5): 1276–1296.

[5] Bo¤a, Federico, Amedeo Piolatto, and Giacomo A. M. Ponzetto (2016). “Political Cen-
tralization and Government Accountability,” Quarterly Journal of Economics 131(1):
381–422.

32
Globalization is likely to make the coordination role of unions more important because it exacerbates
international policy externalities (Epifani and Gancia 2009; Broner and Ventura 2011).

38
 [6] Bolton, Patrick, and Gerard Roland (1996). “Distributional Con‡icts, Factor Mobility
and Political Integration,”American Economic Review 86(2): 99–104.

[7] — — (1997). “The Breakup of Nations: A Political Economy Analysis,”Quarterly Jour-

nal of Economics 112(4): 1057–1090.

[8] Bonfatti, Roberto and Kevin H. O’Rourke (2018). “Growth, Import Dependence and
War,”Economic Journal, 128(614): 2222–2257.

[9] Broner, Fernando and Jaume Ventura (2011). “Globalization and Risk Sharing,”Review
of Economic Studies 78(1): 49-82.

[10] Butcher, Charles, and Ryan Gri¢ ths (2013). “Introducing the International System(s)
Dataset (ISD), 1816–2011,”International Interactions 39(5): 748–768.

[11] Casella, Alessandra (2001). “The Role of Market Size in the Formation of Jurisdictions,”
Review of Economic Studies 68(1): 83–108.

[12] Casella, Alessandra, and Feinstein, Jonathan S. (2002). “Public Goods in Trade: On
the Formation of Markets and Jurisdictions,” International Economic Review 43(2):
437–462.

[13] Coase, Ronald H. (1960). “The Problem of Social Cost,”Journal of Law and Economics
3: 1–44.

[14] Desmet, Klaus, Michel Le Breton, Ignacio Ortuño-Ortín and Shlomo Weber (2011).
“The Stability and Breakup of Nations: A Quantitative Analysis,”Journal of Economic
Growth 16(3): 183–213.

[15] Epifani Paolo and Gino Gancia (2009).“Openness, Government Size and the Terms of
Trade,”Review of Economic Studies 76: 629–668.

[16] Gancia, Gino, Giacomo Ponzetto and Jaume Ventura (2020). “A Theory of Economic
Unions,”Journal of Monetary Economics 109: 107–127.

[17] Gennaioli, Nicola and Joachim Voth (2015). “State Capacity and Military Con‡ict,”
Review of Economic Studies 82(4): 1409–1448.

[18] Gri¢ ths, Ryan (2014). “Secession and the Invisible Hand of the International System,”
Review of International Studies 40(3): 559–581.

39
[19] Hooghe, Liesbet, and Gary Marks (2003). “Unraveling the Central State, but How?
Types of Multi-Level Governance,”American Political Science Review 97(2): 233–243.

[20] Kahler, Miles, and David A. Lake (2004). “Governance in a Global Economy: Political
Authority in Transition,”PS: Political Science & Politics 37(3): 409–414.

[21] Lake, David A., and Angela O’Mahony (2004). “The Incredible Shrinking State: Ex-
plaining Change in the Territorial Size of Countries,” Journal of Con‡ict Resolution,
48(5): 699–722.

[22] Lake, David A., and Angela O’Mahony (2006). “Territory and War: State Size and
Patterns of Interstate Con‡ict,” in Territoriality and Con‡ict in an Era of Globaliza-
tion, edited by Miles Kahler and Barbara F. Walter, 133–55. Cambridge: Cambridge
University Press.

[23] Lockwood, Benjamin (2006). “The Political Economy of Decentralization.” In Hand-

book of Fiscal Federalism, edited by Ehtisham Ahmad and Giorgio Brosio, 33–60. Chel-
tenham: Edward Elgar.

[24] Maddison, Angus (2001). The World Economy: A Millennial Perspective. Paris: OECD.

[25] Martin, Philippe, Thierry Mayer and Mathias Thoenig (2008). “Make Trade Not War?”
Review of Economic Studies 75(3): 865–900.

[26] — — (2012). “The Geography of Con‡icts and Regional Trade Agreements,”American

Economic Journal: Macroeconomics 4(4): 1–35.

[27] Oates, Wallace E. (1972). Fiscal Federalism. New York, NY: Harcourt Brace Jovanovich.

[28] O’Rourke, Kevin H., and Je¤rey G. Williamson. (1999). Globalization and History: The
Evolution of a Nineteenth-Century Atlantic Economy. Cambridge, MA: MIT Press.

[29] Polachek, Solomon W. (1980). “Con‡ict and Trade,” Journal of Con‡ict Resolution
24(1): 55–78.

[30] Posner, Richard A. (2014). Economic Analysis of Law, 9th edition [1st edition 1973].
New York, NY: Wolters Kluwer Law & Business.

[31] Rohner, Dominic, Mathias Thoenig and Fabrizio Zilibotti (2013). “War Signals: A
Theory of Trade, Trust, and Con‡ict,”Review of Economic Studies, 80(3): 1114–1147.

40
[32] Ruta, Michele (2005). “Economic Theories of Political (Dis)integration,” Journal of
Economic Surveys 19(1): 1–21.

[33] Singer, David J., and Melvin Small (1966). “The Composition and Status Ordering of
the International System: 1815–1940,”World Politics 18(2): 236–282.

[34] Treisman, Daniel (2007). The Architecture of Government: Rethinking Political Decen-
tralization. Cambridge: Cambridge University Press.

41
 A Appendix –For Online Publication

A.1 Data Sources

The trade share reported in Figure 1 is merchandise exports as percent of GDP in 1990
prices, from Maddison (2001). Maddison provides trade data for selected countries in the
years 1820, 1870, 1913, 1929, 1950, 1973 and 1998. To avoid compositional e¤ects, we report
the value of merchandise export as a share of GDP for the set of countries with data for all
those years (Austria, Belgium, France, Italy, Spain, Switzerland, the United Kingdom and
the United States). The trade share computed using the data for all available countries in
every year is very similar to the one displayed in Figure 1.
The number of countries is reported for the same years. Data on the number of countries
in the twentieth century is not very controversial. For the nineteenth century, however,
some leading conventions grossly underestimate the number of countries. For example, the
“International System,”developed by Singer and Small (1966) and adopted in the Correlates
of War project or in the Cross-National Time-Series Data Archive, only includes countries
with international recognition. In particular, prior to 1920, the criteria to be recognized
as an independent country were to have population greater than 500,000 and to have had
diplomatic missions at or above the rank of chargé d’a¤aires with Britain and France. Clearly,
this de…nition is too strict for our purposes, which require the identi…cation of even relatively
small political units living in economic and political autarky. We follow Butcher and Gri¢ ths
(2013), who recognize the problem and o¤er alternative criteria to identify the number of
countries between 1816 and 2011.
The number of WTO members is from the WTO website.
The size of countries and empires displayed in Figure 8 is from the Cross-National Time-
Series (CNTS) Data Archive. It provides data on contiguous territorial area in thousand
square miles for all countries existing in a given year according to the International Sys-
tem. In a few instances, missing data have been imputed by cross-checking major territorial
changes from other sources (China and Persia before 1860). Area of empire is provided for
a consistent sample of 13 countries: Austria (formerly Austria-Hungary), Belgium, France,
Germany (formerly Prussia), Italy (formerly Sardinia), Japan, the Netherlands, Portugal,
Russia, Spain, Turkey (formerly the Ottoman Empire), the United Kingdom, and the United
States. For these countries, empire area includes “overseas”territories (i.e., colonies). Data
for the two World War periods, 1914–1918 and 1940–1945 (1938–1954 for empires) are miss-
ing.

1
 For the analysis in Section 6, the volume of trade is computed as the value of imports
plus exports per capita. Population is in millions. Urbanization is aggregate population in
cities of 100,000 and over divided by total population. For empires, trade, population and
urbanization refer to the metropolis only. Land area includes overseas territories.

A.2 Discussion of Assumptions

The starting point of our analysis is the idea that improvements in transport technology are
a major driver of globalization, which we model as a fall in trade costs between localities.
This view is uncontroversial: there is clear evidence that the secular rise in trade volumes
was made possible by better transport technologies. For instance, some of the major drivers
of the …rst wave of globalization are the adoption of the steamship (Pascali 2017), the
telegraph (Steinwender 2018) and the spread of railroads (Donaldson 2018). Likewise, the
main drivers of the second wave of globalization include containerization in ocean shipping
(Levinson 2006), the development of jet aircraft engines (Hummels 2007) and more recently
the ICT revolution. All these innovations promoted trade both between and within countries.
We also assume that borders obstruct trade. The large negative e¤ect of political borders
on trade volumes is well know at least since the work of McCallum (1995), who showed that,
controlling for distance and income, trade between two Canadian provinces is 20 times larger
than trade between a Canadian province and a U.S. state. While the exact magnitude of the
border e¤ect is still subject to debate, all existing studies coincide in …nding large e¤ects. For
instance, in a recent survey of the empirical literature on gravity equations, Head and Mayer
(2014) report that countries are typically found to trade 5 to 7 times more with themselves
than with any other country.
There is equally strong evidence that sharing economic regulations and signing economic
agreements promotes trade and reduces the border e¤ect. For instance, Head and Mayer
(2014) also report that sharing a common currency or being part of a free trade area is
associated on average with a doubling of the volume of trade. Likewise, Helpman, Melitz
and Rubinstein (2008) …nd that having a similar legal system increases the bilateral volume
of trade by more than 60 percent. Using a simple model, Anderson and van Wincoop (2004)
attempt a rough decomposition of the border e¤ect. They argue that the overall cost of
borders is equivalent to an ad valorem tari¤ of 44 percent, which can be broken down into
8 percent policy related barriers (including non-tari¤ barriers), 7 percent language barriers,
14 percent currency barriers, 6 percent information cost barriers, and 3 percent security
barriers.

2
 Our modeling assumption that technological barriers reduce trade along the intensive
margin while policy-induced barriers a¤ect the extensive margin is also grounded in empiri-
cal evidence. There is a wide consensus that transport costs a¤ect signi…cantly the intensive
margin of trade. On the other hand, Helpman, Melitz and Rubinstein (2008) and Dutt,
Mihov and Van Zandt (2013) …nd that free-trade agreements and WTO membership pre-
dominantly a¤ect the extensive margin.
Turning next to governments, our model follows the standard assumptions that underpin
the literature on federalism and the architecture of government since Oates (1972). Having
separate local governments enables better preference matching, but sharing a common gov-
ernment enables bene…cial policy coordination and reaps economies of scale. While these
assumptions originated as simple observations of real-world patterns, models of political econ-
omy have provided them with rigorous micro-foundations. Majority rule makes centralization
costly when localities have di¤erent preferences (Lockwood 2002; Besley and Coate 2003).
Frictions in bargaining between political leaders (Harstad 2007) and in their agency rela-
tionship to their constituents (Bo¤a, Piolatto and Ponzetto 2016) explain both why multiple
local government cannot fully coordinate their policies and why a single central government
cannot fully tailor public services to local preferences. Accordingly, Strumpf and Oberholzer-
Gee (2002) …nd empirically that U.S. states with more heterogeneous preferences are more
likely to decentralize policy-making, while Lassen and Serritzlew (2011) show that municipal
amalgamations cause citizens to experience a worse …t between what they want and what
they get from local government. Evidence from sub-national governments also con…rms the
existence of economies of scale in government administration, albeit— consistent with our
model— not in the provision of public goods more broadly (Reingewertz 2012; Blom-Hansen,
Houlberg and Serritzlew 2014; Blom-Hansen et al. 2016; Blesse and Baskaran 2016).
Our assumption of economies of scope in government is equally classic (Musgrave 1971;
Dahl and Tufte 1973; Alesina and Spolaore 2003). Marks and Hooghe (2004, p. 18) “em-
phasize the costs of decomposing authority” as a paramount concern in the analysis of
multi-level governance, especially in the international arena with its prevalence of intersect-
ing task-speci…c jurisdictions. Empirical evidence shows that multiplying administrative
tiers reduces their e¢ ciency, and is particularly associated with lower labor productivity and
excess government employment (Le Galès and John 1997; Andrews and Boyne 2009). This
cost is particularly pronounced for special-purpose governments in charge of a single task
(Berry 2009). These e¢ ciency losses re‡ect both the costs of administrative duplication and
economies of scope in political accountability. Bo¤a, Piolatto and Ponzetto (2016) show the-

3
oretically that dividing policy-making responsibilities across multiple levels of government
increases overall rent extraction by government o¢ cials. Fan, Lin and Treisman (2009) re-
port that across countries corruption increases with the number of administrative tiers: as
they rise from two to six, the probability of a …rm reporting that it is never expected to pay
bribes falls by 32 percentage points.33

A.3 Analytical Derivations

A.3.1 Computing Equilibrium Consumption

Locality l maximizes the objective function:

Z 1 Z 1
WlM = ln cl (m; i) dmdi, (A1)
0 0

subject to the budget constraint:

Z 1 Z 1
pl (m; i) [cl (m; i) ql (m; i)] dmdi 0, (A2)
0 0

where ql (m; i) and pl (m; i) are the production and price of input m of industry i in locality
l. The productions ql (m; i) must be consistent with available technology as described in
Section 2. Since individuals are atomistic, they take prices as given in their maximization
problems.
Equilibrium prices are as follows:
8
>
> if l = m
< 1
pl (m; i) = R (A3)
e if i 2 [0; ] and Il=m =0
>
>
: e if i 2 ( ; 1] or R
Il=m = 1 but l 6= m.

To prove this claim, normalize world income to unity (Y = 1). Note …rst that each locality
has unit density of expenditure on each input in each industry. We next examine production.
Consider …rst industries that require contract enforcement, i 2 [0; ]. Locality l employs unit
R
density of labor to produce each non-traded input m for which Il=m = 0. Thus, output of

33
The evidence about the importance of economies of scope has been gathered mostly at the sub-national
level. Casual observation suggests, however, that economies of scope also apply at the supra-national level.
An example is the European Central Bank, which aims at reducing the border e¤ect by eliminating currency
barriers. Its creation does not seem to have reduced the size or costs of national central banks, but instead
it seems to have just added to these costs.

4
each non-traded input has density e , so the value of output has unit density given price e .
R1 R
The remaining mass 0 Il=m dm of industry-i labor is employed to produce an identical mass
of output. Unit density of it is sold domestically at a unit price. The remainder is shipped
R
in identical amounts to other localities with Il=m = 1, each of which receives a density e
of imports, hence import value of unit density given price e . In industries that do not
require contract enforcement, i 2 ( ; 1], the whole unit mass of industry-i labor is employed
to produce the locality’s own input variety, which is sold in identical amounts to all other
localities in the world. Thus, the value of sales in each locality of each input in each industry
has unit density, just like expenditure. This proves our claim.34
With these prices at hand, we can compute the equilibrium productions and consumptions
described in Section 2.

A.3.2 Dealing with Integer Constraints

A world without unions consists of N 2 N countries. Country n consists of measure Sn > 0

P PN
of localities, such that N
n=1 Sn = 1. Utilitarian social welfare is W = F
n=1 Sn W (Sn ; Sn ).
The welfare function:

W F (Sn ; Sn ) = + (1 + Sn ) Sn (A4)
Sn

is concave in Sn and increasing at Sn = 0. Whenever > + it has a unique maximum

at Sn = S1 .
Pareto e¢ ciency then requires that either Sn S1 for all n = 1; 2; :::; N or Sn S1 for
all n. Otherwise some localities could leave a country with excessive size Sn > S1 and join
another with insu¢ cient size Sn < S1 , raising the welfare of every locality in both countries.
Utilitarian welfare maximization requires all countries to have the same size. If there are
two countries m and n such that Sm > Sn > S1 , then transferring the marginal locality from
m to n not only raises its welfare, but it also raises the welfare of Sm localities by more than
it lowers the welfare of Sn < Sm localities. Likewise if Sm < Sn < S1 .
Therefore, once integer constraints are taken into account, the utilitarian welfare optimum

34
It is straightforward to show that this equilibrium is unique. First, rule out variation in the prices of
traded inputs since this would generate excess demand (supply) of cheap (expensive) varieties. Second, rule
out that the relative prices of traded and nontraded varieties be above (below) = since this would lead to
an excess demand (supply) of nontraded inputs.

5
without unions is a partition of the world into a number:

N1 = arg max + 1 + N (A5)

N 2N N N

of identical countries. The objective function W has strictly decreasing di¤erences in (N; )
because for any H > L and and NH > NL :

H L H L
W (NH ; H) W (NH ; L) = < W (NL ; H) W (NL ; L) = . (A6)
NH NL

Thus, the welfare-maximizing number of countries N1 is decreasing in in the sense of

monotone comparative statics. It is likewise decreasing in , and increasing in and .
By the same reasoning, the utilitarian welfare optimum with unions is a world union
composed of a number:
N2 = arg max N (A7)
N 2N N
of identical countries.

A.3.3 Equilibrium Conditions for the General Model

The most general model in this paper is discussed in Section 5 when we assume that 0
and 0. The model of war and conquest of Section 4 applies in the limit as ! 0,
while the model of diplomacy in Section 3 applies in the limit as ! 0 and ! 0. In this
Appendix, we discuss the di¤erent possible equilibria of the general model.

Law and Diplomacy Assume …rst that the core decides to forego warfare and join the
free world, so F = 1. Then we can de…ne the welfare of core localities as:

W C (SC ; UC ) = +( + ) (1 ) + ( UC + min fUC ; g) SC ICU , (A8)

SC

and the welfare of periphery localities as:

W P (SP ; UP ) = + (1 + UP ) SP IPU , (A9)

SP

6
where ICU and IPU are indicator functions that take value 1 if SC = UC and SP = UP ,
respectively, and zero otherwise. Utilitarian social welfare for the entire world equals:

W (SC ; SP ; UC ; UP ) = W C (SC ; UC ) + (1 ) W P (SP ; UP ) . (A10)

There are three possible equilibrium political structures. The …rst is a single-level gov-
ernment structure with countries of optimal sizes:
s s
S1C = and S1P = , (A11)
( + )

and without any unions.35 Utilitarian world welfare under this …rst structure is:

W1 W (S1C ; S1P ; S1C ; S1P )

n p p o
2
= + + (1 ) 2 [ ( + )] + (1 ) ( ) . (A12)

The second potential equilibrium is a two-level government structure with countries of

optimal size: r
S2C = S2P = , (A13)

and with a world union: UC = UP = 1. Utilitarian world welfare under this second structure
is:
p
W2 W (S2C ; S2P ; 1; 1) = + + 2
2 . (A14)

The third potential equilibrium features a two-level government structure for the core
with countries of size S3C = S2C and a core union UC = , but a single-level government
structure for the periphery with countries of size S3P = S1P and no union (UP = S3P ).
Utilitarian world welfare under this third structure is:

W2 W (S2C ; S2P ; 1; 1)
h p p i
2 2
= + + (1 )+( + ) 2 + (1 ) ( ) . (A15)

The equilibrium political structure under diplomacy is the one that delivers the highest

35 2
Equation (A11) assumes that > + ( + ), so that there is enough preference heterogeneity to
ensure that the optimal core country size is always smaller than the entire core.

7
welfare: 8
>
> 1
max fW 2 ; W 3 g
< (S1C ; S1P ; S1C ; S1P ) if W
arg max W 3 (S2C ; S2P ; 1; 1) if W 2 max fW 1 ; W 3 g (A16)
>
>
: (S ; S ; ; S ) if W 3 max fW 1 ; W 2 g .
3C 3P 3P

The core union yields higher welfare than single-level government (W 3 > W 1 ) if:
p hp p i
< 1 ( ) ( + ) 2 ( + ) , (A17)

for an increasing and concave threshold (@ 1 =@ > 0 > @2 1 =@

2
) such that @ 1 =@ > 0,
@ 1 =@ > 0, @ 1 =@ < 0, @ 1 =@ > 0 and @ 1 =@ > 0.
The core union yields higher welfare than the world union (W 3 > W 2 ) if:
p p p
> 2( ) (1 + ) 2 , (A18)

for an increasing and concave threshold (@ 2 =@ > 0 > @2 2 =@

2
) such that @ 2 =@ > 0,
@ 2 =@ > 0, @ 2 =@ < 0, @ 1 =@ > 0 and @ 2 =@ = 0.
Single-level government yields higher welfare than the world union (W 1 > W 2 ) if:

> 1 + (1 ) 2. (A19)

The two functions 1 ( ) and 2 ( ) have a single crossing because:

@ 2 @ 1
1 (0) > 2 (0) = 0 and > . (A20)
@ @

In other words, the core union can follow but not precede single-level government and precede
but not follow the world union because:

@
W2 W3 = 1 2
(1 ) S3P > 0 (A21)
@

and
@
W3 W1 = ( S1C ) > 0. (A22)
@

War and Conquest If there are empires, the analysis is as essentially as it was in Section
4 for = 0. The welfare of core localities that form an imperial metropolis is given by:

8
 1
Wl = W E (E) = + (1 ) + + ln 1 + + ( + )E E !.
E
(A23)
The optimal size of empires is larger because so are gains from trade within the metropolis:
s
E = . (A24)
( + )

Thus, if core localities build empires their welfare is given by:

1 p
W E (E ) = + (1 ) + + ln 1 + 2 [ ( + )] !. (A25)

The free world contains a measure F < 1 of localities in the periphery. In this case, all
the analysis in Section 4 applies and, in particular, Condition (18) still determines whether
the free world has a single or two-level government structure.
When are empires formed? In the absence of empires, the welfare of core localities in the
welfare-maximizing political structure is given by:
8
>
> C 1 2 3
< W (S1C ; S1C ) if W > max fW ; W g
WCF = W C (S2 ; ) if W 2 > max fW 1 ; W 3 g (A26)
>
>
: W C (S ; 1) if W 3 max fW 1 ; W 2 g .
2

If W E (E ) < WCF , there are no empires, diplomacy prevails and the size of the free world is
F = 1. If instead W E (E ) WCF , there are = E empires of size E , and the size of the
free world is reduced to F = 1 = .
Core localities prefer empires to peaceful countries (W E (E ) > W C (S1C ; S1C )) if > E,

for a threshold E implicitly de…ned by:

p hp p i 1
2 ( E + ) ( E + ) =! (1 ) ln 1 + , (A27)

such that (recalling the de…nition of ) @ E =@ < 0, @ E =@ < 0, @ E =@ 0, @ E =@ 0,

@ E =@! > 0 and @ E =@ < 0.

9
 Core localities prefer empires to the peaceful core union (W E (E ) > W C (S2 ; )) if:

1 p hp p i
> E ( ) ! (1 ) ln 1 + + ( + ) 2 ( + ) ,
(A28)
for a concave threshold (@ 2 E =@ 2
< 0) with:

@ E @ 1
E ( E) = 1 ( E) and < , (A29)
@ @

such that (recalling the de…nition of ) @ E =@ < 0, @ E =@ > 0, @ E =@ 0,

@ E =@! = 1, @ E =@ < 0 and @ E =@ > 0.
Core localities prefer empires to the peaceful world union (W E (E ) > W C (S2 ; 1)) if:

1 p hp p i
> E1 ( ) ! (1 ) ln 1 + + ( + ) 2 ( + ) ,
(A30)
2 2
for an increasing and concave threshold (@ E1 =@ >0>@ E1 =@ ) with:

@ E @ E1 @ 1 @ 2
E1 ( )> E ( ) and < < + (1 ) , (A31)
@ @ @ @

such that (recalling the de…nition of ) @ E1 =@ < 0, @ E1 =@ > 0, @ E1 =@ 0, @ E1 =@! =

1, @ E1 =@ < 0 and @ E1 =@ > 0.

A.4 Extensions

A.4.1 Looting in All Industries

In Section 4 we assumed for simplicity that imperial powers loot from their colonies a fraction
2 [0; 1] of market goods in the industries that do not rely on contract enforcement (a
fraction 1 of the total). Analogous results obtain, at the cost of a greater algebraic
burden, if we assume looting can occur in all industries.
In contract-intensive industries, the metropolis then loots not only traded output, but
also output that would otherwise be non-traded. Such looting ine¢ ciently incurs at the
same time both the e¢ ciency costs of autarky and the technological costs of trade. As a
consequence, In a share (1 E) of industries, consumption in the metropolis is cl (m; i) =
e [1 + e (E M ) =M ].

10
 As a result, the welfare of a core locality that builds an empire is:

1
Wl = W E (E) = + (1 E) ln 1 + e

1
+ (1 + E) + ln 1 + E !. (A32)
E

The optimal size of an empire, from the perspective of the metropolis, is then:
v
u
u h i.
E =t (A33)
1 1
+ ln 1 + ln 1 + e

The di¤erence relative to our baseline is that the equilibrium size of empires is increasing
in their ability to loot. If an empire is larger, fewer goods need to be produced with autarky
technology. Thus, more looted goods would have incurred transport costs regardless of
looting. These economies of scale in looting result from the same transportation technology
that enables global trade. Thus they strengthen our baseline …ndings, making the size of
empires increase even more sharply in response to globalization.

A.4.2 Imperial Unions

In Section 4 we ruled out by a simplifying assumption the creation of extractive empires

within trade-promoting international unions. This alternative political structure would be
desirable if and only if the extractive bene…ts of colonialism were su¢ ciently high.
To consider such an extension, suppose that core localities can impose their preferred
political structure, including economic unions, onto the rest of the world. If they choose to
conquer empires but also create a union, they obtain welfare:

1
Wl = W2E (E; U ) = + (1 ) + ln 1 + + U E ! , (A34)
E

which is maximized by a world union and empires of size:

s
E2 = . (A35)

Intuitively, empires are smaller with a world union than without it (E2 < E ) because their
size no longer yields gains from market access. Yet they are larger than peaceful countries

11
within the union (E2 > S2 ) because their size still yields extractive gains.
There are then two possibilities. If Equations (23)–(25) are satis…ed, aggressive imperial-
ism is an intermediate stage. Colonialism emerges only when the opportunities for colonial
trade are large enough ( L > 0), and it endogenously disappears when economic unions are
created. If instead Equations (24) and (25) fail to hold, core localities always resort to war
and conquest. They never form peaceful countries ( L = 0), nor do they cease waging war
after creating unions. Results are otherwise analogous to our baseline scenario. An economic
E 36
union is created when trade opportunities cross a threshold value ( U ). Imperialism delays
E
the creation of international unions ( U > U ), especially when democratic values are less
E
widespread (@ U =@ 0) and great powers are less likely to clash with each other in their
E
colonial conquests (@ U =@ 0).

A.4.3 Shallow Unions

In Sections 3 and 4 we assumed that an international union provides all market regulation. In
reality, unions often di¤er in their depth. They may range from a roundtable for negotiating
tari¤s to a complex set of treaties regulating disparate issues such as product standards and
intellectual property rights.
To capture such di¤erences in the depth of international integration, we can extend our
model to include the possibility of forming a “shallow union:” that is, a union that enables
trade in a fraction (1 ) of the industries subject to the border e¤ects. A shallow union
may be easier to implement, so that its cost is reduced to < . Then, the utility of a
locality in a shallow union is:

W F (S2 ; 1) = + (1 + S2 ) S2 , (A36)
S2

where optimal country size becomes:

s
S2 = . (A37)

A shallow union leaves the border e¤ect in industries. As a result, countries in a shallow
union are larger than countries in a deep union, but smaller than in the case of no union.
This simple extension yields a rich set of possible outcomes. Expanding trade opportu-
q p
36 E E
The threshold is formally de…ned by: U +2 U = +2 .

12
nities make deep unions more attractive than shallow unions. However, shallow unions are
cheaper and may therefore be chosen …rst. Hence, a deep union may follow a shallow union,
suggesting that the creation of supranational institutions may be gradual (as in Section 5,
but along a di¤erent margin). Expanding trade opportunities still make deep unions more
attractive than empires. But whether they make shallow unions more attractive than em-
pires depends on which political structure generates a higher volume of trade. By completely
removing internal borders, empires promote a high volume of trade between their localities.
A shallow union, instead, promotes trade between all world localities, but removes the border
e¤ect only partially. To compare them, we compute how an expansion of trade opportunities
a¤ects welfare under the two alternative political structures:

@
W F (S2 ; 1) W E (E) = [1 E (1 S2 )] . (A38)
@

The expression in (A38) shows that expanding trade opportunities make a shallow union
more attractive than empires when the equilibrium size of empires is su¢ ciently small:
E <1 + S2 . However, the equilibrium size of empires grows with expanding trade
opportunities. This makes a non-monotonic evolution possible. At low levels of , expanding
trade opportunities may trigger a switch from empires to shallow unions. But shallow unions
may then be followed by another era of even larger empires for high levels of . This possibil-
ity may help explain why shallow forms of international cooperation appeared before World
War II, such the extensive European network of bilateral trade agreements built around the
1860 French-British treaty (Irwin 1993) or the League of Nations. It may also help explain
the decline in international cooperation during the interwar period, and to some extent the
backlash against globalization and the revived nationalism observed in recent times.

A.4.4 Federalism

In Section 3 we did not consider the possibility of creating a sub-national level of government
in addition to, or instead of, a supranational one. Sub-national federalism is a classic strategy
to improve preference-matching in the provision of public services, at the cost of sacri…cing
economies of scale and scope (Oates 1972).
To consider this possibility, we can extend our model and assume it is possible to create
a federal structure that subdivides a country into regional jurisdictions that provide some
public services. The cost of creating this new level of government is to forego economies of
scope equal to F. The bene…t is that the regional governments can provide a share (1 + ) =2

13
of public services while incurring only a share (1 ) =2 of …xed costs, for an index 2 [0; 1]
that de…nes the bene…ts of federalism. Each region in a country provides a bundle of public
services that contains equal amounts of the ideal varieties of the localities that form the
region, but not of the localities that form other regions in the same country. Under this
assumption, we can re-write Equation (11), describing the utility of a representative locality
l, as follows:

1 1+ 1+ 1 F U
Wl = + (1 + U) S+ F + F Il U Il ,
2 2 2S 2F
(A39)
where F if the size of regions and IlF is an indicator variable that takes value 1 if the country
of locality l has a federal structure (F < S), and zero otherwise (F = S).
In Section 3 there were two possible political structures: (i) centralized countries without
an international union (IlU = IlF = 0); and (ii) centralized countries within an international
union (IlU = 1, IlF = 0). The country sizes and welfares that these options deliver are
still those described in Section 3. The present extension adds two additional possibilities:
(iii) federal countries without an international union (IlU = 0, IlF = 1), which implies the
following sizes for countries and their regions:
s s
(1 + ) (1 )
Sf = >F = , (A40)
(1 ) 2 (1 + )

and welfare:
p q
2
Wl F ; Sf ; Sf = + (1 ) (1 + ) [ (1 ) 2 ] 1 F; (A41)

and (iv) federal countries with an international union (IlU = IlF = 1), which implies the
following sizes for countries and their regions:
s s
(1 + ) (1 )
S3 = >F = , (A42)
(1 ) (1 + )

and welfare q
2
Wl (F ; S3 ; 1) = + 2 1 F U. (A43)

The creation of a federal structure increases country size, both in the absence of an
international union (Sf > S1 ) and in its presence (S3 S2 ). The creation of an international

14
union reduces the size of federal countries just as it does the size of unitary countries (S3 <
Sf ). However, it does not change the size of federal regions (F ), since their optimal design
responds purely to considerations related to public-service provision, and not to market
regulation.
The bene…t of creating a federal structure is a reduction in the cost of preference mismatch
( ). Its cost is a sacri…ce of economies of scale ( ) and scope ( F ). Overall, a federal structure
is desirable for a country that does not belong to an international union if:
p q p
2 F
(1 + ) [ (1 ) 2 ]+ 1 +p 2 . (A44)

For a country that belongs to an international union, a federal structure is desirable if:
q
2 F
1 + p 1. (A45)
2

These conditions show that, if is large enough and F small enough, adopting a federal
structure can be an equilibrium outcome both with and without international unions.
This …nding does not change the main results of our model. As trade opportunities
expand, the increase in welfare is: @Wl =@ = U . Thus, it is smallest for a unitary country
outside a union (@Wl =@ = S1 ); it is intermediate for a federal country outside a union
(@Wl =@ = Sf ); it is greatest for countries inside a union, irrespective of their sub-national
structure (@Wl =@ = ). At the supranational level, the evolution of political structure
remains as explained in Section 3. The di¤erence is that the creation of international unions
may be delayed if countries adopt a federal structure, which makes them larger. At the sub-
national level, a federal structure may be an intermediate stage, becoming obsolete when an
international union is created, enabling a decline in country size.

A.5 Historical Narratives

A.5.1 European History

Since the late Middle Ages, European sovereign states on average grew in size until the end of
the nineteenth century, when this trend was dramatically reversed. For example, Kitamura
and Lagerlöf (2020) show that borders declined monotonically from 1500 to 1900, and then
started to increase. Medieval Europe was fragmented into hundreds of small states at a time
when trade was costly, insecure and limited to few commodities. The early modern period
saw important changes in both the economic and the political organization of the continent.

15
With the Commercial Revolution, trade began to ‡ourish and the feudal system started to
be replaced by a smaller number of countries of growing size. While in 1600 there were 112
sovereign states in Europe and the Near East, at the beginning of 1800 the number had
fallen to 79.
The Industrial Revolution gave trade an even more prominent role and triggered ma-
jor changes in socioeconomic conditions that ultimately made the rise of the nation state
possible. Trade expansion was enabled by the introduction of canals, improved roads and
railways. At the same time, the high degree of political fragmentation at the time of the
Congress of Vienna (1815) was followed by the uni…cation of Germany and Italy (1871) and
the further consolidation of other nation states. The economic rationale of building large
internal markets was especially evident in the case of German uni…cation, which started
with the formation of a customs union (Zollverein). This process of political centralization
culminated at the beginning of the twentieth century, when Europe was divided into just 28
independent states.
Yet, the twentieth century marks a turning point. Europe entered a stage of political
fragmentation, with the number of independent states growing to 58 in 2000.37 Simulta-
neously, the process of creating the European Union (EU) started. Our model in Section
5 shows that international unions start from countries that have closer economic ties and
expands outward as globalization increases the value of trade with more remote locations
too. This prediction is consistent with the history of the EU. In 1957, Belgium, France,
Italy, Luxembourg, the Netherlands and West Germany signed the Treaty of Rome estab-
lishing the European Economic Community (EEC) as a customs union. The EEC gradually
expanded to include nearby countries: Denmark, Ireland and the United Kingdom in 1973,
Greece in 1981, Portugal and Spain in 1986. In 1992, the Maastricht Treaty converted the
EEC into the EU and, in 1995, Austria, Finland and Sweden joined. In 2002, the Euro was
launched. Then the EU grew to include 28 member states in 2013. Over this period, the
union has grown in size and scope.
In June 2016, however, the United Kingdom voted in a referendum to leave the union and
there is a concern that other countries might follow in the future. While rising nationalism
may have played a role, it is instructive to look at these events from the lens of our model.
Three lessons can be learned. First, the value of union membership is proportional to the
economic ties between countries. These are stronger for countries located in the core of

37
The number of independent states in Europe and Near East is taken from Euratlas-Nüssli
(http://www.euratlas.com).

16
continental Europe. For instance, while EU members account for almost 80% of Belgium’s
exports, the same …gure is around 50% for the United Kingdom. Second, our model provides
a rationale for trade-promoting unions. As more power is shifted to the union in other areas
such as migration, tensions may arise, especially in countries with a strong national identity.
These two observations may explain why some British politicians have advanced the idea of
replacing the EU single market with a Commonwealth free-trade zone. Third, our model
suggests that the value of joining the union is proportional to its size. As a country exits,
the economic foundations of the union become more fragile.

A.5.2 The Rise and Fall of Colonial Empires

The pattern of an initial decline and subsequent increase in the number of countries is not
con…ned to Europe only. For instance, the number of African countries fell from 36 in 1816 to
4 in 1914, to rise again to 51 in 2000. Similarly, in South-East Asia, these numbers changed
from 37 to 4 and then 20 in the years 1816, 1914 and 2000, respectively.38 However, to better
interpret the political evolutions in these regions, it is important to bring con‡ict into the
picture.
In our model of war and conquest, empires are built to extract trade surplus from the
colonies and disappear when a world union is formed to foster free markets. According to
historians and in line with this view, one of the key driving forces behind colonial expansion
was the desire to secure trade and access to scarce resources in an era of revived commerce,
but when mercantilist practices where common. For much of the second millennium, states
deployed force to create markets (Findlay and O’Rourke, 2007). Due to the scarcity of land
and the desire to avoid powerful rivals, European great powers expanded by conquering
territories overseas. The role of colonial powers in enforcing trade within the empire but not
outside was very clear in the case of maritime commerce. On the one hand, large navies
were built to control and protect trading routes; on the other hand, privateers were often
authorized to capture merchant ships belonging to enemy nations. Despite some notable
setbacks, colonialism continued to grow prior to World War I and …nally collapsed after
World War II.
The sharp decline of empires started after the creation of international agreements aimed
at promoting economic cooperation. This is no coincidence. In the words of Rosecrance
(1986) and Spruyt (2005), empires dissolved because the gains through commerce displaced

38
The number of countries is taken from Butcher and Gri¢ ths (2013).

17
gains through territorial acquisition.39 There is also evidence that international organizations
played a direct role in the process of decolonization. For example, in 1960 the UN General
Assembly voted the Declaration on the Granting of Independence to Colonial Countries
and Peoples. Our model of regional unions is consistent with the very di¤erent patterns
of geographic expansions of empires and unions. As long as war is the dominant means of
territorial expansion, the European Great Powers try to avoid con‡ict with each other by
conquering far-away colonies and building global empires. Yet, once trade is enforced by
peaceful international unions, countries seek economic integration with proximate partners,
with whom the gains from trade are higher. The switch from global empires to regional
unions is also consistent with the increased regionalization of world trade patterns observed
in the data (e.g., Fouquin and Hugot 2016).

References

[1] Anderson, James E. and Eric van Wincoop (2004). “Trade Costs,”Journal of Economic
Literature 42(3): 691–751.

[2] Andrews, Rhys, and George Boyne (2009). “Size, Structure and Administrative Over-
heads: An Empirical Analysis of English Local Authorities,” Urban Studies 46(4): 739–
759.

[3] Berry, Christopher B. (2009). Imperfect Union: Representation and Taxation in Multi-
level Governments. Cambridge: Cambridge University Press.

[4] Besley, Timothy, and Stephen Coate (2003). “Centralized versus Decentralized Provision
of Local Public Goods: A Political Economy Approach,” Journal of Public Economic
87(12): 2611–2637.

[5] Blom-Hansen, Jens, Kurt Houlberg and Søren Serritzlew (2014). “Size, Democracy, and
the Economic Costs of Running the Political System,”American Journal of Political Sci-
ence 58(4): 790–803.

[6] Blom-Hansen, Jens, Kurt Houlberg, Søren Serritzlew and Daniel Treisman (2016). “Juris-
diction Size and Local Government Policy Expenditure: Assessing the E¤ect of Municipal
Amalgamation,”American Political Science Review 110(4): 812–831.

39
Bonfatti (2012) also attributes the fall of empires to the growing importance of trade between industrial
countries relative to trade with colonies.

18
[7] Blesse, Sebastian, and Thushyanthan Baskaran (2016). “Do Municipal Mergers Reduce
Costs? Evidence from a German Federal State,” Regional Science and Urban Economics
59: 54–74.

[8] Bonfatti, Roberto (2012).“Trade and the Pattern of European Imperialism, 1492–2000,”
Oxford University Department of Economics Discussion Paper No. 618.

[9] Dahl, Robert A., and Edward R. Tufte (1973). Size and Democracy. Stanford, CA: Stan-
ford University Press.

[10] Donaldson, Dave (2018). “Railroads of the Raj: Estimating the Impact of Transporta-
tion Infrastructure,”American Economic Review 108(4–5): 899–934.

[11] Dutt, Pushan, Ilian Mihov and Timothy Van Zandt (2013). “The E¤ect of WTO on
the Extensive and the Intensive Margins of Trade.” Journal of International Economics
91(2): 204–219.

[12] Fan, C. Simon, Chen Lin, and Daniel Treisman (2009). “Political Decentralization and
Corruption: Evidence from Around the World,” Journal of Public Economics 93(1–2):
14–34.

[13] Findlay, Ronald and Kevin H. O’Rourke (2007). Power and Plenty: Trade, War and
the World Economy in the Second Millennium. Princeton, NJ: Princeton University Press.

[14] Fouquin, Michel and Jules Hugot (2016). “Back to the Future: Trade Costs and the
Two Globalizations,”CEPII Working Paper No. 2016-13.

[15] Harstad, Bård (2007). “Harmonization and Side Payments in Political Cooperation,”
American Economic Review 97(3): 871–889.

[16] Head, Keith and Thierry Mayer (2014). “Gravity Equations: Workhorse,Toolkit, and
Cookbook,” in Handbook of International Economics, Vol. 4, edited by Gita Gopinath,
Elhanan Helpman and Kenneth Rogo¤, 131–195. Amsterdam: North Holland.

[17] Helpman, Elhanan, Marc Melitz and Yona Rubinstein (2008). “Estimating Trade Flows:
Trading Partners and Trading Volumes,” Quarterly Journal of Economics, 123(2): 441–
487.

[18] Hummels, David (2007). “Transportation Costs and International Trade in the Second
Era of Globalization”, Journal of Economic Perspectives 21(3): 131–154.

19
[19] Irwin, Douglas A. (1993). “Multilateral and bilateral trade policies in the world trading
system: an historical perspective,”in New Dimensions in Regional Integration, edited by
Jaime De Melo and Arvind Panagariya, 90-119. Cambridge: Cambridge University Press.

[20] Kitamura, Shuhei, and Nils-Petter Lagerlöf (2020). “Geography and State Fragmenta-
tion,”Journal of the European Economic Association 18(4): 1726–1769.

[21] Lassen, David D., and Søren Serritzlew (2011). “Jurisdiction Size and Local Democracy:
Evidence on Internal Political E¢ cacy from Large-Scale Municipal Reform,” American
Political Science Review 105(2): 238–258.

[22] Le Galès, Patrick, and Peter John (1997). “Is the Grass Greener on the Other Side?
What Went Wrong with French Regions, and the Implications for England,” Policy &
Politics 25(1): 51–60.

[23] Levinson, Marc (2006). The Box: How the Shipping Container Made the World Smaller
and the World Economy Bigger. Princeton, NJ: Princeton University Press.

[24] Lockwood, Benjamin (2002). “Distributive Politics and the Costs of Centralization,”
Review of Economic Studies 69(2): 313–337.

[25] McCallum, John (1995). “National Borders Matter: Canada-U.S. Regional Trade Pat-
terns,”American Economic Review 85(3): 615–623

[26] Musgrave, Richard A. (1971). “Economics of Fiscal Federalism,” Nebraska Journal of

Economics and Business 10(4): 3–13.

[27] Pascali, Luigi (2017) “The Wind of Change: Maritime Technology, Trade and Economic
Development,”American Economic Review 107(9): 2821–2854.

[28] Reingewertz, Yaniv (2012). “Do Municipal Amalgamations Work? Evidence from Mu-
nicipalities in Israel,”Journal of Urban Economics 72(2–3): 240–251.

[29] Rosecrance, Richard (1986). The Rise of the Trading State: Commerce and Conquest
in the Modern World. New York, NY: Basic Books.

[30] Spruyt, Hendrik (2005). Ending Empire: Contested Sovereignty and Territorial Parti-
tion. Ithaca, NY: Cornell University Press.

20
[31] Steinwender, Claudia (2018). “Real E¤ects of Information Frictions: When the States
and the Kingdom became United,”American Economic Review 108(3): 657–696.

[32] Strumpf, Koleman S., and Felix Oberholzer-Gee (2002). “Endogenous Policy Decentral-
ization: Testing the Central Tenet of Economic Federalism,”Journal of Political Economy
110(1): 1–36.

21