Trade Policy and Firm Boundaries

ULB -- Universite Libre de Bruxelles, Working Papers ECARES 01/2010;
Source: RePEc
ABSTRACT
We develop an endogenous growth model with R&D spillovers to study the long-run consequences of offshoring with firm heterogeneity and incomplete contracts. In so doing, we model offshoring as the geographical fragmentation of a firm's production chain between a home upstream division and a foreign downstream division. While there is always a positive correlation between upstream bargaining weight and offshoring activities, there is an inverted U-shaped relationship between these and growth. Whether offshoring with incomplete contracts also increases consumption depends on firm heterogeneity. As for welfare, whereas with complete contracts an R&D subsidy is enough to solve the inefficiency due to R&D spillovers, with incomplete contracts a production subsidy is also needed. Copyright © The editors of the "Scandinavian Journal of Economics" 2009 .

Full-text

Available from: Paola Conconi
Trade Policy and Firm Boundaries
Laura Alfaro
Harvard Business School and NBER
Paola Conconi
Universit´e Libre de Bruxelles (ECARES) and CEPR
Harald Fadinger
§
University of Vienna
Andrew F. Newman
Boston University and CEPR
January 2010
Abstract
We examine how trade policy affects firms’ organizational choices. We embed a model
of firms’ vertical integration decisions into a standard perfectly-competitive international
trade framework. In the model, integration decisions are driven by a trade-off between the
pecuniary benefits of coordinating production decisions and the managers’ private benefits
of operating in preferred ways. The price of output is a crucial determinant of this choice,
since it affects the size of the pecuniary benefits: higher prices lead to more integration.
Through its effect on product prices, trade policy also has an impact on firm boundaries. We
use a unique dataset that allows us to construct firm-level indexes of vertical integration
for a large set of countries. In line with the predictions of our model, we obtain three
main results. First, higher tariffs lead to higher levels of vertical integration. Second,
differences in ownership structure across countries, measured by the distance in sectoral
vertical integration indexes, are smaller in sectors with similar levels of protection. Finally,
ownership structures are more alike for members of regional trade agreements.
JEL classifications: D23, F13, F23.
Keywords: Firm Organization, Tariffs, Regional Trade Agreements.
Preliminary draft. We wish to thank for their comments participants at the 2009 ETSG conference in Rome
and the 2010 AEA conference in Atlanta. We are particularly grateful to Holger Breinlich, Maria Guadalupe,
Emanuel Ornelas and Maurizio Zanardi for their valuable suggestions. Research funding from the FNRS and the
European Commission through the PEGGED project is gratefully acknowledged by Paola Conconi. We thank
Francisco Pino for excellent research assistance.
Email: lalfaro@hbs.edu; Phone: 1-617-495-7981.
Email: pconconi@ulb.ac.be; Phone: 32(0)2-650-3426.
§
Email: harald.fadinger@univie.ac.at; Phone: 43(0)1-4277-37415.
Email: afnewman@bu.edu; Phone: 1-617-358-4354.
Page 1
1 Introduction
The boundaries of a firm mediate the way its employees trade off their private and collective
goals. In a highly integrated firm with a single headquarters that owns many links in the
supply chain, major production decisions can be well-coordinated to accomplish organizational
goals such as profit, but can also impose high costs on subordinate managers. A less-integrated
enterprise, with decision-making spread among several firm heads, may economize on private
costs, but may also leave the decisions ill-coordinated, harming profit.
A vast theoretical literature has studied these and other trade-offs affecting the choice of firm
boundaries, and of ownership and control structure more generally.
1
But it has largely left open
the question of what determines how this trade-off is actually resolved in the market. Several
recent papers have begun to address this question.
2
Market conditions are bound to affect the
value of the enterprise’s objective, the way its members make their trade-offs, and therefore the
ownership structure that best mediates it. In particular, as pointed out by Legros and Newman
(2009) and Conconi, Legros and Newman (2009), even in perfectly competitive environments
there will be a systematic relationship between the boundaries of the firm and the equilibrium
price level in the product market. In its starkest form, the prediction is that the higher the
market price, the more integrated firms will be.
The reason for this predicted relationship is very simple. The primary decision makers—the
“managers”—trade off the organizational goals (revenue, profit) against their private benefits
(doing things their way, or their view of the best way, or the easiest way).
3
When different parts
of the organization are not integrated, managers make their decisions independently, taking
significant account of their private benefits and rather less of the organization’s, resulting in
poor coordination. Integration puts the decisions in the hands of a single headquarters who has
strong incentives to coordinate, maximizing the organizational objective at the expense of the
other managers. Non-integration therefore favors high private benefits and low coordination;
integration generates high coordination but also high private costs. When organization value
1
The “incomplete contracts” approach we follow begins with Grossman and Hart (1986) and Hart and Moore
(1990), which emphasize the hold-up problem. The tradeoff we have delineated is the focus of Hart and Holmstr¨om
(2009). What they have in common is that the firm’s boundaries are identified with the extent of decision rights
over assets and associated operations.
2
Theoretical contributions include McLaren (2000) on hold-up and market thickness; Legros and Newman
(2008) on control structures and the terms of trade in supplier markets; and Marin and Verdier (2008) on
delegation and product demand elasticity. Another literature has examined the question of whether goods are
sold within or across firm boundaries in the global economy. See, for example, Antras (2003), Antras and Helpman
(2004), and Helpman (2006) for an overview.
3
As pointed out by Hart and Holmstr¨om (2009), private benefits may arise from various sources. For example,
employees often have their human capital tied to particular technologies and like to work with technologies with
which they are familiar. Also, their future career prospects may depend on how well their human capital fits the
firm’s needs, so strategic choices concerning technology will have significant private consequences. Differences
in the way of doing things (e.g., engineers and marketing departments) can make coordination difficult. Van
den Steen (2005) stresses the importance for organization design of conflicting private benefits stemming from
different corporate cultures and/or managerial vision.
1
Page 2
(market price) is high, this trade-off is made in favor of integration, since the organizational goal
is relatively more valuable than the private ones; at low prices, the trade-off goes the other way,
favoring non-integration. Thus anything that affects equilibrium prices will have an indirect
effect on the degree of integration.
Of course, in an industry composed of several enterprises facing this organizational design
trade-off, the market price will depend on the choices of ownership structure made by all the
enterprises: if integration is more productive, and all enterprises integrate, prices will be lower.
Market-clearing equilibrium in the product market will jointly determine quantity, price and
ownership structures. What these models show is that the increasing relationship between inte-
gration and price that pertains to a single price-taking firm will also obtain for the relationship
between the average degree of integration in the industry and the equilibrium price level.
Trade policy provides a natural proving ground for examining the effects of prices on organi-
zation, since it generates a plausibly exogenous source of equilibrium price variation: the degree
of trade protection will obviously affect equilibrium prices; however, as we shall argue below, it
is likely to be independent of firms’ boundary choices. The first-order effect of the imposition
of a tariff on an import-competing good is to raise its price. Thus, all else equal, the higher
the tariff, the more integrated firms in the industry should be.
4
By the same token, if tariffs
in the same industry in two countries are close, equilibrium prices and their ownership struc-
tures should be similar. Thus the theory predicts convergence in ownership structure between
countries with similar levels of protection. Moreover, if two countries are members of a regional
trade agreement in which internal tariffs have been completely eliminated, all else equal, firms
in those two economies should be equally integrated.
Empirical analysis on the effects of trade policy on organizational choices has been limited,
largely by the absence of an international dataset sufficiently comprehensive to support studies
of firm organization across a wide range of countries. We overcome this limitation by using a
new dataset from Dun and Bradstreet (D&B). This contains both listed and unlisted plant-level
observations in more than 200 countries for 2004. The data include ownership information about
the firm’s family members (number of family members, its domestic parent and its global parent)
to link multi-plant firms. The dataset enables us to study the differential effects of trade policies
on firms’ organization structure. Over the last decades, barriers to trade have fallen in developed
countries and diminished considerably in many developing countries. Despite the recent trends,
restrictions to trade are still quantitatively important for many countries and sectors, allowing
for comparative analysis.
In order to explore the predictions of our theory, we combine the D&B dataset with U.S.
input-output tables to construct measures of vertical integration at the firm level. These in-
4
This statement can be interpreted as a statement about intensive margins—more parts of the supply chain
should be part of a single firm as the price for the final good increases, or about extensive margins—a greater
fraction of firms are integrated at higher prices, assuming some heterogeneity among them.
2
Page 3
dexes represent the opportunity for vertical integration between related industries.
5
Despite its
limitations, this methodology allows us to analyze a large set of countries and industries, thus
overcoming an important constraint in the literature (we also do not have to worry about the
value of intra-firm activities being affected by transfer pricing).
We obtain data on applied most favored nation (MFN) tariffs at the 4 digit SIC level for all the
WTO members for which this information is available. We also collect systematic information on
all regional trade agreements (RTAs) that were in force in 2004. In order to account for various
alternative factors that affect vertical integration that have been emphasized in the literature, we
control for a number of country- and sector-specific variables (rule of law, financial development,
capital intensity, relationship-dependence, external credit dependence).
6
We also use number of
bilateral variables (distance, common border, common colonial relationship, as well as income
and income per capita).
We first examine the relation between tariffs and organizational structure. Consistent with
the predictions of our theoretical model, we find that higher tariffs lead to more vertical inte-
gration at the firm level. The impact of tariffs on vertical integration is sizable: in our preferred
estimation, a 100 percent tariff increase leads to a 1.44 percent increase in the vertical integra-
tion index; this implies that reducing tariffs from their mean level of 5.4 percent to 1 percent
reduces vertical integration by over 6 percent. Our results are robust to different specifications
and subsamples.
The theoretical framework also suggests that trade policy should affect the degree of or-
ganizational convergence across countries through its effect on prices. That is, convergence in
corporate organization, i.e., a tendency of industries to be characterized by the same ownership
structure across countries, may not only result from global cultural transmission or technological
diffusion, but also from standard neoclassical market forces, namely the law of one price (see
also Conconi, Legros and Newman, 2009). In line with our predictions, we find that for a given
country-pair differences in sectoral vertical integration indexes are significantly (at least at the
5 percent level) larger in those sectors in which differences in MFN-tariffs are larger.
We then examine the relation between the degree of sectoral organizational convergence and
common membership in a RTA. Our theoretical model suggests that, everything else equal, full
liberalization of product markets between two countries should result in the convergence of firms’
ownership structure within industries. Our empirical results show that ownership structures are
indeed more alike for members of RTAs. The impact on organizational convergence is stronger
for older trade agreements, which are more likely to have fully eliminated trade barriers among
5
We build on the methodology of Acemoglu, Johnson and Mitton (2008), who use the 1992 U.S. input-
output tables to calculate the opportunity for vertical integration for every pair of industries, by computing the
dollar value of one industry required to produce a dollar’s worth of the other industry. They then combine this
information with data from WorldBase for the year 2002, to construct measures of vertical integration. Section
3.3 describes in detail the empirical methodology.
6
See e.g., Acemoglu, Aghion, Griffith, and Zilibotti (2009), Acemoglu, Johnson, and Mitton (2009), Legros
and Newman (2008), McMillian and Woodruff (1999), and Rajan and Zingales (1998).
3
Page 4
member countries. As it is possible that countries that are more similar are also more likely to
form a RTA, we use a number of controls for common relationship. We find the difference in
vertical integration indexes to be around 13 percent smaller in country pairs engaged in a RTA
than for a country pair without one.
As mentioned above, our empirical analysis relies on exogenous price variation induced by
trade policy. In particular, we exploit the cross-country and cross-sectoral variation in MFN
tariffs and the existence of regional trade agreements. MFN tariffs are negotiated at the
GATT/WTO level over long periods of time and, as stressed by a vast political economy litera-
ture starting from the seminal paper by Finger, Hall and Nelson (1984), they are less “political”
than unilateral trade barriers: protectionist pressure is usually applied to administrative mea-
sures for the regulation of imports (e.g., safeguards, anti-dumping and countervailing duties).
7
Regional trade agreements, such as free trade areas or customs unions, are also negotiated over
long periods of time and regulated by GATT/WTO rules. Previous papers in the literature take
RTAs as being exogenous to firms’ decisions. See, for example, Trefler (2004) on the impact of
the Canada-U.S. Free Trade Agreement (CUSFTA) on industry- and plant-level labor produc-
tivity, and Bustos (2009) on the impact of the Southern Common Market (MERCOSUR) on
technology upgrading by Argentinean firms.
Our paper contributes to an emerging literature on general equilibrium models with endoge-
nous organizations, and in particular to a nascent stream of the empirical work examining firms’
organizational choices in a global economy. Acemoglu, Johnson, and Mitton (2009) (henceforth
AJM) study the determinants of vertical integration using data from D&B in 93 countries. The
authors find no evidence that contracting costs and financial development have significant effects
on vertical integration. However, they find greater vertical integration in countries that have
both greater contracting costs and greater financial development. They also find that countries
with greater contracting costs are more vertically integrated in more capital-intensive industries.
Acemoglu, Aghion, Griffith, and Zilibotti (2009) use detailed data on all British manufactur-
ing plants from the UK Census of Production combined with input-output tables to study the
determinants of backward vertical integration. Bloom, Sadun and Van Reenen (2009) study de-
centralization patterns in US, Europe, and Asia and find social capital variables and competition
to be associated with more decentralization.
Guadalupe and Wulf (2009) investigate the effect of the 1989 CUSFTA agreement, which
eliminated tariffs and other barriers between the U.S. and Canada, on hierarchies in large US
firms. The authors find that competition leads firms to flatten their hierarchies. Breinlich
7
Indeed, most papers in the empirical political economy literature focus on unilateral protectionist measures.
For example, Goldberg and Maggi (1999) and Gawande and Bandyopadhyay (2000) use data on 1983 non-tariff
barrier (NTB) coverage ratios for the U.S. manufacturing sector to test the lobbying model by Grossman and
Helpman (1994). Some studies find that firm size and industry concentration affect U.S. non-tariff barriers
through their impact on lobbying contributions (e.g., Mitra, 1999; Bombardini, 2008). In our empirical analysis,
we control for firm size. We also argue that firms’ ownership structures are unlikely to have a systematic impact
on trade policy in general, and on MFN tariffs in particular.
4
Page 5
(2008) also reveals a significant increase in the level of M&A activity in Canada (but not the
U.S.) following CUSFTA. Other studies have stressed the impact of trade liberalization on the
reallocation of resources across individual plants and firms (see Goldberg and Pavcnik (2004) for
an overview) or in work practices (Schmitz, 2005).
Another strand of the literature has focused on how organizational design can explain the
observed patterns of intra-firm trade and the location of multinational subsidiaries or suppliers
(Antras, 2003; Antras and Helpman, 2004; and Grossman and Helpman, 2004). Ornelas and
Turner (2008) examine how trade policy affects hold-up problems through its effect on a foreign
supplier’s investment incentives.
The paper is organized as follows. Section 2 presents the theoretical framework and discusses
the empirical implication or our model. Section 3 describes our data and the methodology to
construct vertical integration indexes. Section 4 presents and discusses the results on tariffs and
vertical integration. Section 5 analyzes the effect of trade policy (tariffs and RTAs) on the degree
of organization convergence within sectors across countries. Section 6 analyzes the robustness
of the results. The last section concludes.
2 The Model
Our model is similar to a standard specific-factor trade model between many small countries,
in which trade is the result of differences in the endowments of specific factors. We will first
describe its building blocks in its closed-economy form, before looking at international trade and
at the effects of trade policy.
2.1 Setup
In each economy, there are K + 1 sectors/goods, denoted by 0 and k = 1, . . . , K; good 0 is a
numeraire. The representative consumer’s utility can be written as
u(c
0
, . . . , c
K
) c
0
+
I
X
k=1
u
k
(c
k
), (1)
where c
0
represents the consumption of the numeraire good, and c
k
represents consumption of
the other goods. The utility functions u
k
(·) are twice differentiable, increasing, strictly concave,
and satisfy the Inada conditions lim
c
i
0
u
0
k
(c
k
) = and lim
c
i
→∞
u
0
k
(c
k
) = 0. Domestic demand
for each good k can then be expressed as a function of its own price alone, D
k
(p
k
).
Production of good k requires the cooperation of two types of input suppliers, denoted A and
B
k
. B
k
suppliers generate no value without being matched with an A; by contrast, A suppliers
can either match with any B
k
or engage in stand-alone production of the numeraire good 0.
Many interpretations of the A and B
k
firms are possible. For example, A firms may represent
5
Page 6
light assembly plants or some basic inputs, such as energy or various business services (e.g., IT,
retailing, logistics), which can be used to produce basic consumer goods or can be combined
with other inputs (B
k
suppliers) to produce more complex goods.
All goods are sold under conditions of perfect competition. Good 0 is the numeraire, with
price equal to 1. We assume that aggregate supply of A’s exceeds that of the B
k
’s so that a
positive amount of good 0 is produced in equilibrium.
So far, we have described a standard specific-factor model, in which A supplier firms represent
the mobile factor, while B
k
firms are the specific factor of production. As discussed below, the
crucial novelty of our model is that production inputs are run by managers, who trade off
the pecuniary benefits of coordinating their decisions with the private benefits of making these
decisions in their preferred way.
An equilibrium in the supplier market consists of matches between each B
k
firm and an A
firm, along with a surplus allocation among all the managers. Such an allocation must be stable,
in the sense that no (A, B
k
) pair can form an enterprise that generates payoffs to each manager
that exceed their equilibrium levels.
All A suppliers are equally productive both when matched with one of the B
k
’s. A stand-
alone A produces α units of the numeraire good. Since the price of the numeraire is equal to
unity, this also pins down the outside option for all A’s.
Individual firms
Our model of the firm relies on two key features. First, managers in each firm enjoy monetary
profits as well as private non-transferable benefits associated with the operations of the firm;
different managers view these operations differently and so their private benefits come into
conflict. For instance, a standardized production line could be convenient for the sectorally-
mobile A suppliers, but may not fit the specific design needs of the B
i
suppliers. Second, some
firm decisions (e.g., choosing production techniques, deciding on marketing campaigns, etc.)
cannot be agreed upon contractually; only the right to make them can be transferred through
changes of ownership.
Consider a firm composed of an A and a B
k
. For each supplier, a non-contractible decision
is rendered indicating the way in which production is to be carried out. Denote the A and
B
k
decisions respectively by a [0, 1] and b
k
[0, 1]. For efficient production, it does not
matter which particular decisions are chosen, as long as there is coordination between the two
suppliers. More precisely, the enterprise will succeed with a probability 1 (a b
k
)
2
, in which
case it generates R > 0 units of output; otherwise it fails, yielding 0. Output realizations are
independent across firms.
Overseeing each supplier firm is a risk-neutral manager, who bears a private cost of the
decision made in his unit. The A manager’s utility is y
A
(1 a)
2
, while the B
k
manager’s
6
Page 7
utility is y
k
b
2
k
, where y
A
, y
k
0 are their respective incomes, and (1 a)
2
and b
2
k
are the costs.
Observe that A’s most preferred action is 1, while B
k
’s is 0, so the managers disagree about the
direction in which decisions should go. Since the primary function of managers is to implement
decisions and convince their units to agree, they continue to bear the cost of decisions even if
they don’t make them.
While decisions themselves are not contractible, the right to make them can be contractually
reassigned (e.g. via a sale of assets). This assignment of decision rights is the organizational
design problem in this model. Managers can remain non-integrated, in which case they retain
control over their respective decisions. Alternatively, they can integrate by engaging a headquar-
ters (HQ), transferring to it the power to decide product standards (a and b
k
) and a share of
the realized revenue in exchange for a fixed payment. HQ is motivated only by monetary con-
siderations, and incurs no costs from the decisions a and b
k
; it will therefore wish to maximize
the income of the integrated firm.
Before production, B
k
managers match with A managers, at which time they sign contracts
specifying an ownership structure and a payment scheme. For simplicity, we assume that this is
accomplished via a fixed payment T from the B
k
to the A.
8
For each match (A,B
k
), total revenue in case of success is given by R times the product
market price, p
k
, which is taken as given by firms when they take their decisions and sign their
contracts. Since the A’s are in excess supply, they must all receive α in equilibrium. Thus T will
just cover A’s anticipated private cost incurred during production together with his opportunity
cost α.
After contract signing, managers (or HQ) make their production decisions, output is realized,
product is sold, and revenue shares are distributed.
Integration
HQ’s are elastically supplied at a cost normalized to zero. After paying its acquisition fee,
and receiving its compensating share of revenue, HQ’s continuation payoff is proportional to
(1 (a b
k
)
2
)Rp
k
.
9
HQ decides both a and b
k
, and since its incentive is to maximize the
expected revenue of the integrated firm, it chooses a = b
k
. Among the choices in which a = b
k
,
the Pareto-dominant one is that in which a = b
k
= 1/2, and we assume HQ implements this
choice. The private cost to each supplier manager is then
1
4
and the payoffs to the A and B
k
managers are equal to α and Rp
k
α
1
2
, respectively (thus T = α +
1
4
).
8
In general, the B
k
may prefer to give the A a positive contingent share of revenue; this complicates notation
but does not change any qualitative conclusion regarding the dependence of integration on price – see Legros and
Newman (2009).
9
The size of the HQ’s share is indeterminate and could be pinned down in many ways that we do not model
here; all that matters for our purposes is that it is positive.
7
Page 8
Non-integration
Under non-integration, each manager retains control of his activity. The decisions chosen are
the (unique) Nash equilibrium of the game with payoffs T (1 a)
2
for A, who chooses a, and
(1 (a b
k
)
2
) Rp
k
b
2
k
T for B
k
, who chooses b
k
. These are a = 1 and b
k
= Rp
k
/(1 + Rp
k
),
with resulting expected output 1
1
(1+Rp
k
)
2
. Notice that output increases with the price: as p
k
becomes larger, the revenue motive becomes more important for B
k
managers and this pushes
them to better coordinate with their A partners. The equilibrium transfer under non-integration
is T = α, and the payoffs are α for the A’s and
(Rp
k
)
2
1+Rp
k
α for the B
k
’s.
Choice of organizational form
To determine the choice of organization that the managers make, we must compare their payoffs
under integration and non-integration. Notice that A suppliers obtain α in both cases. B
k
suppliers obtain a higher payoff under integration if and only if Rp
k
1
2
>
(Rp
k
)
2
1+Rp
k
, or p
k
> 1/R.
Thus managers’ organizational choices depend on product prices. At low prices, despite
integration’s better output performance, revenues are still small enough that the B
k
managers
are more concerned with their private benefits and so remain non-integrated. At higher prices,
B
k
managers prefer integration, which allows them to achieve coordination at lower private costs.
Product market equilibrium and the OAS curve
Equilibrium for the economy comprises a general equilibrium of the supplier and product markets.
We have already characterized the supplier markets: some A suppliers produce by themselves α
units of the numeraire good, while others are matched with B
k
’s for the productions of goods
k = 1, . . . , K and receive α.
In product market k, the large number of firms implies that with probability one the supply
is equal to the expected value of output given p
k
; equilibrium requires that this price adjust so
that the demand equals the supply.
To derive industry supply, suppose R is distributed in the population according to some
continuous c.d.f. G(R) with mean 1 and support [R, R]. Since all enterprises in industry k
with R < 1/p
k
remain non-integrated, while the remaining ones integrate, total supply at price
p
k
[1/R, 1/R] is (recall that n
k
is the measure of B
k
suppliers)
S(p
k
) = n
k
"
Z
1/p
k
R
R(1 (
1
1 + Rp
k
)
2
)dG(R) +
Z
R
1/p
k
RdG(R)
#
. (2)
(If p
k
< 1/R, supply is n
k
R
R
R
R(1 (
1
1+Rp
k
)
2
)dG(R); if p
k
> 1/R, it is n
k
.)
This “Organizationally Augmented Supply” (OAS) curve incorporates the ownership struc-
ture decisions of the industry’s enterprises as well as the usual price-quantity relationship. When
8
Page 9
p
k
< 1/R, no firm is integrated, but supply increases with price, since every non-integrated firm’s
expected output does. As price rises above 1/R, the most productive enterprises integrate, and
those that remain non-integrated produce more, so that output rises further. Once p
k
reaches
1/R, all firms are integrated and industry output is fixed at n
k
(since the mean R is 1) for prices
higher than that threshold.
In the absence of trade, an equilibrium in the product market of good k is a price and a
quantity that equate supply and demand: D
k
(p
k
) = S(p
k
). The degree of integration of the
industry (i.e., the fraction 1 G(1/p
k
) of firms that integrate) is a nondecreasing function of the
equilibrium price, strictly increasing on [R, R].
2.2 Trade Policy and Firms’ Organization
The world consists of C small countries, indexed by c, which have identical demands and tech-
nologies in the production of all goods k = 1, . . . , K. Trade is the result of endowment differ-
ences between countries. In particular, we assume that the countries can be divided into two
homogeneous groups: a “Home” set of countries that are relatively more endowed in the spe-
cific factors necessary to produce goods k {m + 1, . . . , K}; and a “Foreign” set of countries
(denoted with a “*”) that are relatively more endowed in the specific factors necessary to pro-
duce goods k {1, . . . , m}. We thus have that n
k
< n
k
for k {1, . . . , m} and n
k
> n
k
for
k {m + 1, . . . , K}. Good 0, the numeraire, is always traded freely across countries. We choose
units so that the international market-clearing and the domestic price of good 0 in each country
are equal to unity.
Each country c imposes an exogenously-given ad valorem tariff t
c
k
0 on import-competing
good k. In sectors k {1, . . . , m} domestic prices are thus equal to p
c
k
= (1 + t
c
k
)P
k
in Home
countries, and to p
c
k
= P
k
in Foreign countries, where P
k
denotes the international price. This
is the solution to the following market-clearing condition:
X
c
M
c
k
(1 + t
c
k
)P
k
=
X
c
X
c
k
(P
k
), (3)
where M
c
k
= D
(1 + t
c
k
)P
k
S
(1 + t
c
k
)P
k
denotes Home imports and X
c
k
= S(P
k
) D(P
k
)
denotes Foreign exports. For goods k {m + 1, . . . , K} the market-clearing condition is
X
c
M
c
k
(1 + t
c
k
)P
k
=
X
c
X
c
k
(P
k
). (4)
From (3) and (4) we can derive an expression for international equilibrium prices as a function
of the tariffs applied by all countries, i.e., P
k
(t
c
k
) for k {1, . . . , m}, and P
k
(t
c
k
) for k
{m + 1, . . . , K}.
9
Page 10
The trade balance condition for a Home country requires
m
X
k=1
P
k
M
c
k
(1 + t
c
k
)P
k
K
X
k=m+1
P
k
X
c
k
(P
k
) + Z
0
= 0, (5)
where Z
0
denotes the net transfer of the numeraire good to settle the trade balance. A similar
condition must hold for a Foreign country.
In our model, trade policies have an effect on organizational choices through their impact on
product prices. In particular, our analysis of the OAS implies that an increase in t
c
k
leads to an
increase in the domestic price of good k; a firm with productivity R will choose integration over
non-integration if this price exceeds 1/R. Comparing two otherwise identical countries c and c
0
,
with t
c
k
> t
c
0
k
, the domestic price and therefore the degree of integration in industry k will be
higher in c than in c
0
.
Our theoretical framework can also be used to examine how trade policy affects the degree
of organizational convergence across countries. In particular, for a given country pair cc
0
, the
difference in the degree of integration within a sector k will depend on the differences in their
applied tariffs: the more similar are t
c
k
and t
c
0
k
, the smaller the difference between p
c
k
and p
c
0
k
and
the more similar firms’ organizational choices within industry k. Finally, consider a country pair
c, c
0
that has signed a RTA, eliminating all tariffs between each other. This implies that their
domestic prices will be equal in all sectors, leading to full convergence in organizational choices.
For the purpose of our empirical analysis, we can reformulate the predictions of our theoretical
model as follows:
1. Higher tariffs lead to a higher degree of vertical integration within sectors.
2. Country pairs have more similar ownership structures in sectors with closer levels of pro-
tection.
3. RTAs lead to organizational convergence among member countries.
3 Data and Descriptive Statistics
3.1 The WorldBase Database
We use data from WorldBase, a database of public and private plant-level observations in more
than 200 countries and territories compiled by Dun & Bradstreet (D&B) for 2004.
10
The leading
U.S. source of commercial credit and marketing information since approximately 1845, D&B
10
The dataset is not publicly available but was released to us by Dun and Bradstreet. For more information
see: http://www.dnb.com/us/about/db database/ dnbinfoquality.html.
10
Page 11
presently operates in the different countries and territories either directly or through affiliates,
agents, and associated business partners.
WorldBase is the core database with which D&B populates its commercial data products
including Who Owns Whom
TM
, Risk Management Solutions
TM
, Sales & Marketing Solutions
TM
,
and Supply Management Solutions
TM
. These products provide information about the “activities,
decision makers, finances, operations and markets” of the potential customers, competitors and
suppliers to the clients of D&B. D&B compiles their data from a wide range of sources. Sources
include partner firms in dozens of countries, telephone directory records, websites, and self-
registering firms. All information is verified centrally via a variety of manual and automated
checks.
Early uses of the D&B data include Caves’ (1975) size and diversification pattern comparisons
between Canadian and U.S. domestic plants as well as subsidiaries of U.S. multinationals in
Canada, and Lipsey’s (1978) comparisons of the D&B data with existing sources and observations
regarding the reliability of the data for U.S. More recently, Harrison, Love, and McMillian
(2004) use D&B’s cross-country foreign ownership information. Other research that has used
D&B data includes Black and Strahan’s (2002) study of entrepreneurial activity in the United
States, and Acemoglu, Johnson, and Mitton’s (2009) cross-country study of concentration and
vertical integration and Alfaro and Charlton’s (2009) analysis of vertical and horizontal activity
by multinationals.
In our view, D&B’s WorldBase, while not without problems, is the best database to analyze
our question. In particular it has four main advantages over most other sources. First, the data
include public and private plants and information to aggregate plants to the firm level. Second,
data sources restricted to Europe such as Amadeus are not useful for our purposes because
they do not have broad coverage of countries and in particular of developing countries, with
different levels of trade barriers. WorldBase by contrast has data in more than 200 countries
and territories. Third, Dun & Bradstreet compile their data from a wide range of sources,
whereas other databases collect primarily from national firm registries. All information is verified
centrally via a variety of manual and automated checks. The wide variety of sources from which
Dun & Bradstreet collects data reduces the likelihood that the sample frame will be determined
by national institutional characteristics. Finally, over its many years in business, D&B has
devised many methods of checking its data and reliability of their dataset.
11
11
See Alfaro and Charlton (2009) for a more detailed discussion of the WorldBase data and comparisons with
other data sources. To give some sense of the coverage of the Dun & Bradstreet sample used in this study,
we compare our data with that collected by the U.S. Census Bureau, Statistics of U.S. Businesses. The U.S.
2001-2002 business census recorded 24,846,832 establishments. Our data include 6,185,542 establishments (from
which we exclude establishments in the total sample without the year started). About three quarters of all
U.S. establishments have no payroll. Most are self-employed persons operating unincorporated businesses that
might or might not be the owner’s principal source of income. The U.S. census records 7,200,770 ‘employer
establishments’ with total sales of $22 trillion. Our data include 4,293,886 establishments with more than one
employee with total sales of $17 trillion. The U.S. census records 3.7 million small employer establishments
(fewer than 10 employees). Our data include 3.2 million U.S. firms with more than one and fewer than 10
11
Page 12
3.2 The Sample
We use data from the 2004 WorldBase file excluding records lacking primary industry and year
started (these restrictions were imposed by cost considerations) for a total of more than 24 mil-
lion observations. The unit of observation in WorldBase is the establishment (a single physical
location where business is conducted or services or industrial operations are performed) rather
than the firm (one or more domestic establishments under common ownership or control). Es-
tablishments, which we also refer to as plants, have their own addresses, business names, and
managers, but might be partly or wholly owned by other firms. The data base allows linking
plants to firms by using information on its domestic parent and its global parent using the DUNS
numbers. Our analysis is at the firm level, that is, we consider all plants connected by the same
global or domestic parent as one unit (see discussion below).
The paper uses four categories of data which WorldBase record for each establishment:
1. Industry information: the 4-digit SIC code of the primary industry in which each estab-
lishment operates and for most countries the SIC codes of up to 5 secondary industries,
listed in descending order of importance.
12
2. Ownership information: information about the firm’s family members (number of family
members, its domestic parent and its global parent).
13
3. Location information including the country, state, city, and street address of each family
member. We use the country location information to link establishments within a family
to the relevant tariff data.
4. Basic operational information: sales and employment.
We excluded countries and territories with fewer than 80 observations and those for which
the World Bank provides no data. We further restricted the sample to Word Trade Organization
(WTO) members for which we have data on tariffs/regional trading arrangements (see discussion
below). Our final sample includes data for 101 countries. Table A-2 in the Appendix lists the
employees. In our data, 6.1 percent of establishments are new (we define as new an establishment having a year
started date less than two years previous). The U.S. Census reported 12.4 percent of establishments to be new
in 2001-2002, for firms with 1-4 employees this was 15.9 percent, for firms with more than 500 employees 11
percent. Comparison by sectors (excluding a number of individual industries, such as religious organizations,
certain government-owned establishments and others which are hard to map given the different classification or
compare) show similar patterns.
12
D&B uses the United States Government Department of Commerce, Office of Management and Budget,
Standard Industrial Classification Manual 1987 edition to classify business establishments. In 1963, the firm
introduced the Data Universal Numbering System—The D&B D-U-N-S Number—used to identify businesses
numerically for data-processing purposes. The system supports the linking of plants and firms across countries
and tracking of the history of plants including potential name changes.
13
D&B also provides information about the firm’s status (joint-venture, corporation, partnership) and its
position in the hierarchy (branch, division, headquarters).
12
Page 13
countries included in our dataset and the sample frame. As a robustness check, we also exclude
countries for which we have less than 1000 plants that are part of firms with at least 20 employees
(see also Klapper, Laeven and Rajan, 2006).
We focus on manufacturing firms (i.e., firms with a primary SIC code between 2000 and
3999), to which our theory of vertical integration fits best. We exclude government/public
sector firms, firms in the service sector (for which we have no tariff data) or in agriculture (due
to the existence of many non-tariff barriers), as well as firms producing primary commodities
(i.e., mining and oil and gas extraction).
We exclude firms with less than 20 employees, since our theory does not apply to self-
employment or to very small firms with little prospect of vertical integration (see also Acemoglu,
Aghion, Griffith, and Zilibotti, 2009).
14
We focus on firms that only operate in one country, since this provides a cleaner analysis
of the effects on tariffs and RTAs on firms’ ownership structure. This is because the degree of
vertical integration of these firms depends only on the prices of the country in which they are
located. In the case of multinational corporations (MNCs), on the other hand, it is harder to
identify the relevant prices and tariffs. Moreover, by focusing on national firms, we can abstract
from issues having to do with the strategic behavior of multinationals across different markets
(e.g., transfer pricing, tariff jumping).
15
We include multinationals in the robustness analysis.
There, we split MNCs into separate entities - one for each country. The reason is that we
need to link organizational structure to domestic tariffs. We define multinational firms as those
foreign-owned establishments belonging to the same domestic ultimate.
3.3 Vertical Integration Indices
Constructing measures of vertical integration is difficult, as the exercise is highly demanding in
terms of data, requiring firm-level information on the sales and purchases of inputs by various
subsidiaries of a firm. Such data are generally not directly available and, to the best of our
knowledge, there is no data available for a wide sample of developed and developing countries.
To measure the extent of vertical integration of a given firm, we build on the methodology
used by AJM (2009). We combine information on plants’ activities and ownership structure from
the WorldBase dataset with input-output data to determine related industries and to construct
vertical integration coefficients V
f
j
, where j is any sector in which firm f is active. Notice that
the sample in AJM is restricted to a maximum of the 30,000 largest records per country in the
2002 WorldBase file (a limit imposed by cost constraints).
16
We instead have information on a
14
Restricting the analysis to firms with more than 20 employees also allows to correct for possible differences
in the the collection of small firms data across countries.
15
We describe an establishment as foreign-owned if it satisfies two criteria: (1) it reports to a global parent
firm, and (2) the parent firm is located in a different country. Parents are defined in the data as entities that
have legal and financial responsibility for another firm.
16
For many countries, this restriction is not binding. For countries with more than 30,000 observations, AJM
13
Page 14
broader sample of more than 24 million establishments in the 2004 WorldBase file. As discussed
below, this allows us to link establishments to firms.
Given the difficulty of finding input-output matrices for all the countries in our dataset, we
follow AJM (2009) and use the U.S. input-output tables to measure vertical linkages within
firms. As the authors note, input-output tables from the U.S. should be informative about input
flows across industries to the extent that these are determined by technology.
17
The input-output data come from the Bureau of Economic Analysis (BEA), Benchmark IO
Tables, which include the make table, use table, and the direct and total requirements coefficients
tables. We use the Use of Commodities by Industries after Redefinitions 1992 (Producers’ Prices)
tables. While the BEA employs six-digit input-output industry codes, WorldBase data use the
SIC industry classification. The BEA website provides a concordance guide between both, but
it is not a one-to-one key.
18
We match the 4-digit SIC codes of each plant in each firm with the 6-digit IO codes, using
the BEA’s concordance guide. For the codes for which the matching was not one-to-one, we
randomized between possible matches to chose one in order not to overstate vertical linkages.
The multiple matching problem, however, is not particularly relevant when looking at plants
operating only in the manufacturing sector (the key is almost one-to-one).
For every pair of industries, i, j, the input-output accounts allow one to calculate the dollar
value of i required to produce a dollar’s worth of j. Combining the SIC information for each plant
in each firm, the matching codes, and the input-output information for the US, we construct
the input-output coefficients for each firm f, IO
f
ij
. Here, IO
f
ij
IO
ij
I
f
ij
, where IO
ij
is the
input-output coefficient for the sector pair ij, stating the cents of output of sector i required to
produce a dollar of j, and I
f
ij
{0, 1} is an indicator variable that equals one if and only if firm
f owns plants in both sectors i and j. A firm that produces i as well as j will be assumed to
supply itself with all the i it needs to produce j; thus the higher is IO
ij
for an i-producing plant
owned by the firm, the more integrated in the production of j the firm will be measured to be.
By adding up the input-output coefficients IO
f
ij
for all inputs i, we arrive at the firm’s degree of
vertical integration in j.
To illustrate the procedure, consider the following example from AJM (2009). They consider
a Japanese establishment, which according to WorldBase has one primary activity, automo-
biles (59.0301), and two secondary activities, automotive stampings (41.0201) and miscellaneous
plastic products (32.0400). The IO
ij
coefficients in the three activities for this plant are:
select the 30,000 largest, ranked by annual sales. They include all industries, except those operating only in
“wholesale trade” and “retail trade”.
17
Note that the assumption that the U.S. IO structure carries over to other countries can potentially bias our
empirical analysis against finding a significant relationship between vertical integration and prices. On the other
hand, it also mitigates the possibility that the IO structure and control variables are endogenous.
18
This concordance is available upon request. The BEA matches its six-digit industry codes to 1987 U.S. SIC
codes http://www.bea.gov/industry/exe/ndn0017.exe.
14
Page 15
Output (j)
Input (i)
Autos Stampings Plastics
Autos 0.0043 0.0000 0.0000
Stampings 0.0780 0.0017 0.0000
Plastics 0.0405 0.0024 0.0560
SUM 0.1228 0.0041 0.0560
The table is a restriction of the economy-wide IO table to the set of industries in which
this establishment is active (i.e., it contains all of the positive IO
f
ij
values). For example, the
IO
ij
coefficient for stampings to autos is 0.078, indicating that 7.8 cents worth of automotive
stampings are required to produce a dollar’s worth of autos. Furthermore, this plant has the
internal capability to produce those stampings, and we therefore assume that it produces all the
stampings it needs for automobiles itself.
19
The bottom row shows the sum of the IO
f
ij
for each
industry: for example, 12.3 cents worth of the inputs required to make autos can be produced
within this plant. We would then say that the degree of vertical integration for this plant is 12.3
in autos.
Rather than the plant, however, our main unit of observation are all plants that belong to the
same firm, i.e., all plants that report to the same headquarters. For example, if the plant in the
example above is reported as being the headquarters of another Japanese plant (subsidiary), we
consider the activities of both plants when constructing a measure of vertical integration for the
firm. In the case of multi-plant firms, restricting the analysis to plant level may underestimate
the number of activities carried out within the boundaries of a firm.
We can now describe the methodology we used to construct construct the firm-level vertical
integration indexes. For a given firm f with primary sector is k located in country c, we define
the integration index in activity j as
V
f,k,c
j
=
X
i
IO
f,k
ij
, (6)
the sum of the IO coefficients for each industry in which the firm is active. Our main measure
of vertical integration is based on the firm’s primary activity:
V
f,k,c
= V
f,k,c
j
, j = k. (7)
In the case multi-plant firms (plants that are connected by the same global ultimate or head-
quarters), we consider the main activity of the headquarters or the domestic parent.
19
Many industries have positive IO
ij
coefficients with themselves; for example, miscellaneous plastic products
are required to produce miscellaneous plastic products; any firm that produces such a product will therefore be
measured as at least somewhat vertically integrated.
15
Page 16
As an alternative measure, we also construct an index based on all the firm’s activities:
V
f,k,c
=
1
N
f
X
j
V
f,k,c
j
, (8)
where N
f
is the number of industries in which firm f is active. This represents the average
opportunity for vertical integration in all lines of a business in which the firm is active.
Our approach for identifying vertical integration suffers from the data limitation that we do
not observe intra-firm transactions. Instead, we infer it from information about the goods pro-
duced in each of the firm’s establishments and the aggregate input-output relationship between
those goods. The advantage of our method is that we have a large amount of data for many
countries and industries and do not have to worry about the value of intra-firm activities being
affected by transfer pricing. Hummels, Ishii, and Yi (2001) argue that another advantage of
using I-O tables is that they avoid the arbitrariness of classification schemes that divide goods
into “intermediate” and other categories. However, our index represents the opportunity of ver-
tical integration. Firms may exercise this opportunity in different ways, as they may still, for
example, purchase or sell inputs from/to third parties.
20
Table 1 presents summary statistics for our vertical integration indexes. Appendix table A-1
compares the indices across the different samples. Our main sample is column 6. It consists of
257,000 manufacturing plants that are part of domestic firms with at least 20 employees located
in 101 countries.
21
3.4 Trade Policy
A further challenge in empirically assessing the impact of market prices on ownership structure
is the simultaneous determination of the two—prices should affect ownership structure but at
the same time ownership structure has an influence on market prices. We use trade policy to
deal with this endogeneity problem and argue that most-favored-nation (MFN) tariffs and RTAs
offer a plausibly exogenous source of price variation to the boundaries of the firm. Of course,
one might still worry about the political economy determinants of these policies. However, as
argued in the introduction, MFN tariffs are negotiated at the multilateral level over long periods
of time and are less “political” than unilateral forms of protection such as anti-dumping duties.
20
Hortacsu and Syverson (2009) combine Census data and the Commodity Flow survey (a random sample of
an establishment shipments in each four weeks during the year, one in each quarter) and ZIP code information
to measure intra-firm trade. They find that shipments from firms’ upstream units to their downstream units are
surprisingly low. This result is at odds with international trade studies, which show that intra-firm trade accounts
for roughly one-third of international shipments (e.g., Bernard, Jensen, Redding and Schott, 2008); however, as
a robustness check, we also perform the analysis using plant-level vertical integration measures.
21
Differences in methodology and samples restrict comparisons with AJM. However, the authors report mean
of 4.87 and a median of 3.34 in their vertical integration index. For the average vertical integration index, the
mean ranges from 3.5 to 6.6 and the median from 1.5 to 5.1. For the primary sector vertical integration index,
the mean ranges from 3.5 to 7.1 and the median from 1.2 to 5.1.
16
Page 17
Some papers in the literature have pointed out that industry concentration and firm size may
affect lobbying contributions and trade policy outcomes (e.g., Mitra, 1998; Bombardini, 2008).
These studies are based on US non-tariff barriers in the manufacturing sector. In our empirical
analysis, we control for both firm size and industry concentration. We also argue that firms’
ownership structure is unlikely to have a systematic impact on the determination of trade policies
in general, and on MFN tariffs in particular.
22
Concerning regional trade agreements such as free
trade areas or customs unions, these are also negotiated over long periods of time and regulated
by GATT/WTO rules. Previous papers in the literature (e.g., Trefler, 2004; Bustos, 2009) take
RTAs as being exogenous to firms’ decisions.
3.4.1 Tariffs
We collect applied MFN tariffs at the 4 digit SIC level for all the WTO members for which
this information is available. We restrict the set of countries to WTO members, which are
constrained under Article I of the GATT by the MFN principle of non-discrimination: each
country c must apply the same tariff t
c
k
to all imports in sector k originated in other WTO
members; preferential treatment is only allowed for imports originating from members of RTAs
or from developing countries.
The data source for MFN tariffs is the World Integrated Trade Solution (WITS) database,
which combines information from the UNCTAD TRAINS database (default data source) with
the WTO integrated database (alternative data source). Tariffs are for 2004 unless they are not
available for that year. In this case tariff data are chosen as the closest available data point
in a five year window around 2004 (2002-2006) giving priority to earlier years.
23
The original
classification for tariff data is the harmonized system (HS) 6 digit classification. Tariffs are
converted to the more aggregate SIC 4 digit level using internal conversion tables of WITS.
Here, SIC 4 digit level MFN tariffs are computed as simple averages over the HS 6 digit tariffs.
We also construct for each 4-digit SIC sector and every country the fraction of imports to
which MFN tariffs apply. To do this, we use information on RTAs and subtract from total
sectoral imports those that originate in countries with which the importer has a common RTA.
Bilateral import data at the 4-digit SIC level for 2004 come from the COMTRADE database.
3.4.2 RTAs
We collect information on regional trade agreements (RTAs) that were in force in 2004 from the
WTO Regional Trade Agreements Information System (RTA-IS).
24
The database includes all
22
There is no theory relating a firm’s boundaries with its incentives to form a lobby group. Even if one allows
that lobbying can play a role in determining MFN tariffs, it is not obvious how the direction of the political
pressure (pro or anti trade) and its extent (e.g., the size of the campaign contributions) could be systematically
related to firms’ organization decisions across a very large set of countries and sectors.
23
I.e., if data for 2003 and 2005 are available, but not 2004, 2003 is chosen.
24
Available online (http://rtais.wto.org/UI/PublicMaintainRTAHome.aspx).
17
Page 18
RTAs in force.
25
We construct a dummy that equals one whenever two countries are members
in a common RTA. The legal basis for the creation of RTAs can be found in Article XXIV of the
GATT/WTO (for agreements involving developed member countries) and in the Enabling Clause
(for agreements amongst developing countries only). Under Article XXIV, member countries can
form free trade areas (FTAs) or customs unions (CUs) covering ‘substantially all trade’, requiring
complete duty elimination and fixed timetables for implementation. The conditions contained
in the Enabling Clause are much less stringent, so RTAs between developing member countries
may effectively involve less trade liberalization. Therefore, we construct a second indicator
variable that only includes free trade agreements and customs union but excludes a number of
preferential trade agreements under the enabling clause that do not imply the full elimination
of trade barriers.
We also construct a variable RTAage that equals the age of the trade agreement in years,
since we expect older trade agreements to have a larger impact on firms’ organizational structure.
3.5 Other Controls
In order to control for alternative factors that explain vertical integration emphasized by the
literature we also collect a number of country- and sector-specific variables.
In terms of country-specific variables, the empirical and theoretical literature have studied
the role of institutional characteristics as well as the level of financial development.
26
We use
the variable “rule of law” from Kaufmann, Kraay, and Mastruzzi (2003) as a measure of the
Legal quality of a country’s institutions. This is a weighted average of a number of variables
(perceptions of the incidence of crime, the effectiveness and predictability of the judiciary, and
the enforceability of contracts) that measure individuals’ perceptions of the effectiveness and
predictability of the judiciary and the enforcement of contracts in each country between 1997
and 1998. The variable ranges from 0 to 1 and is increasing in the quality of institutions.
We also use private credit by deposit money banks and other financial institutions as a
fraction of GDP in 2004 taken from Beck, Demigurc-Kunt and Levine (2006) as a measure of a
country’s Financial development.
The literature stresses as well differential effects across industries. We combine these country-
specific measures with sector-specific information from the US, to proxy for exogenous variation
in sector characteristics.
First, we construct sectoral Capital intensity at the 4-digit-SIC level for the US. Data comes
from the NBER-CES manufacturing industry database (Bartelsmann and Gray, 2000). Following
Acemoglu, Johnson and Mitton (2006), capital intensity is defined as the log of total capital
expenditure relative to value added averaged over the period 1993-1997.
25
Note that the dataset does not include trade preferences under the Generalized System of Preferences (GSP),
such as the African Opportunity Act program of the US or the Everything but Arms program of the EU.
26
As AJM note, the effect of these variables, however, may be ambiguous.
18
Page 19
Second, we use Nunn’s (2008) measure of Relationship dependence for the US, which proxies
for the severity of hold-up problems. For each sector this variable measures the fraction of inputs
that are not sold on an organized exchange or reference priced. We convert the data for 1997
from the BEA’s input-output classification to 4-digit US-SIC.
27
Third, we follow Rajan and Zingales (1998) and construct the variable External dependence,
which measures sectoral dependence on external credit for the US as the fraction of investment
that cannot be financed with internal cash flows. The authors identify an industry’s need for
external finance (the difference between investment and cash generated from operations) under
two assumptions: (i) that U.S. capital markets, especially for the large, listed firms they analyze,
are relatively frictionless enabling us to identify an industry’s technological demand for external
finance; (ii) that such technological demands carry over to other countries. Following their
methodology, we construct similar data for the period 1999-2006.
28
To control for domestic industry concentration we use Herfindahl-indices. The variable is
constructed for each country-sector using sales of all plants in that sector.
We also use a number of bilateral variables from CEPII: bilateral distance measured as the
simple distance between the most populated cities in km, dummies for contiguity, for common
official or primary language, and for a common colonial relationship (current or past).
Finally, we use information on GDP and GNI per capita for the year 2004 from the World
Development indicators 2008.
Table 1 presents summary statistics for our control variables and Table A-4 in the Appendix
is the correlation table.
4 Tariffs and Vertical Integration
In this section, we assess the empirical validity of the first prediction of our theoretical model,
examining whether higher tariffs lead to more vertical integration at the firm level. To do so,
we estimate the following panel regression model:
log(V
f,k,c
) = α + β
1
log(t
k,c
) + β
2
log(Employment
f
) + β
3
X
k,c
+ δ
k
+ δ
c
+
f,c
. (9)
Note that, for the variables expressed in logs, the estimated coefficients can be interpreted as
elasticities. We take logs of tariffs (which are already expressed in ad-valorem terms) in order
to mitigate problems with outliers. While the distribution of tariffs is extremely skewed, log
27
Nunn’s dataset is available under http://www.economics.harvard.edu/faculty/nunn.
28
An industry’s external financial dependence is obtained by calculating the industry median of external fi-
nancing of U.S. companies using data from Compustat calculated as: (Capex-Cashflow)/Capex, where Capex is
defined as capital expenditures and Cashflow is defined as cash flow from operations. Industries with negative
external finance measures have cash flows that are higher than their capital expenditures.
19
Page 20
tariffs are approximately normally distributed.
29
The dependent variable is the (log) vertical
integration index of firm f located in country c, with primary sector k, as defined in (7). In
alternative specifications, we also use the average vertical integration index defined in (8). Our
main regressor of interest is the log MFN tariff applied by country c in sector k (log(t
k,c
)). The
set of explanatory variables includes a firm’s number of employees in logs (log(Employment
f
)),
which allows us to control for the relation between firm’s size and ownership structure. The vector
X
k,c
consists of different interactions between sector and country characteristics (e.g., interaction
between a sector’s capital intensity and a country’s level of financial development). We also
include sector fixed effects at the 4-digit level (δ
k
), which allows us to capture cross-industry
differences in technological or other determinants of vertical integration (e.g., a sector’s capital
intensity). Finally, we add country fixed effects (δ
c
), which capture cross-country differences in
institutional determinants of vertical integration (e.g., a country’s level of financial development
and the quality of its contracting institutions) and also control for country-specific differences
in the way firms are sampled.
30
Given that tariffs vary only at the sector-country level, while
the dependent variable varies at the firm level, we cluster standard errors at the sector-country
level.
The results are reported in Table 2. Consider first the left-hand panel, which reports the
results of the regressions using the main vertical integration index (V
f,k,c
). Column (1) presents
the results of the basic specification, which only includes the MFN tariff, firm size, and country
and sector fixed effects. In line with what predicted by our theoretical model, tariffs have
a positive and significant effects on a firm’s level of vertical integration. The estimate for β
1
implies that a 100 percent tariff increase leads to a 1.44 percent increase in the vertical integration
index. In terms of economic magnitudes this implies that reducing tariffs from their mean level
in manufacturing of 5.4 percent to 1 percent (a 440 percent decrease) reduces vertical integration
by 0.0144*440=6.16 percent. Hence, the impact of tariffs on vertical integration is sizable.
In columns (2)-(3) we add different sets of controls, to account for other determinants of
vertical integration, as suggested by the literature. In column (2) we include two interaction
terms: one between Capital intensity and Financial development and one between Capital in-
tensity and Legal quality. Notice that the tariff coefficient becomes larger—a 100 percent tariff
increase now leads to a 2.3-2.4 percent increase in the vertical integration index—and is now
significant at the 1 percent level. The estimates for the interaction terms are also highly sig-
29
See Table 1 for summary measures of the distribution of MFN tariffs. Our main sample corresponds to
column 6 in Table 1. Taking logs of variables removes some observations. However, the vertical integration index
equals zero only in 1439 cases. since more observations are lost as a combination of zeros of dependent and
explanatory variables that are zero (many tariffs are zero) we do not use a Tobit analysis. Below we present
results using log (1+ variables) obtaining similar results.
30
D&B samples establishments in the formal sector (and their are, of course, differences in the size of the formal
sector across rich and poor countries). In the robustness checks we try an alternative way to control for this by
restricting the sample to countries for which we have at least 1000 plants that are part of firms with at least 20
employees.
20
Page 21
nificant and indicate that more capital intensive sectors are more integrated in countries with
more developed financial markets, and less integrated in countries with better legal institutions.
In column (3), we include two alternative interaction terms: that between External dependence
and Financial development; and that between Relationship specificity and Legal quality. Again,
tariffs are positive and highly significant, while the interaction terms are insignificant.
31
The right-hand panel of Table 2 reports the results of regressions in which we used a firm’s
average vertical integration as our dependent variable (V
f,k,c
). The results on tariffs are consis-
tent, but somewhat less significant. This is not surprising, since in our regressions we consider
the effects of MFN tariffs applied to the primary activity of the firm, rather than to all of them.
As a first robustness check, we check if the results of the effect on tariffs on vertical integration
(9) is affected by using log(1 + variable) for the dependent and the explanatory variables, which
allows to keep the observations with zero tariffs. From Table 3 we see that results remain hardly
affected even though we add more than 50.000 observations to the sample. The magnitude of the
tariffs coefficient drops slightly but it remains positive and significant at the 1 to 5 percent level
in all specifications. Results are again slightly weaker when using the average vertical integration
index.
As mentioned, the determination of MFN tariffs is arguably “less political” than unilateral
forms of protection. There is also no theory relating a firm’s boundaries with its incentives to
form a lobby group. However, one may still worry, for example, that large firms, which are
more likely to vertically integrate, lobby for tariffs (leading to higher MFN tariffs). That is, one
may be concerned that MFN tariffs may be correlated with other (omitted) characteristics of
the firm associated with both the potential to lobby and the incentive to vertically integrate.
Our regression analysis controls for the size of the firm (proxied by employment) in addition to
controlling for industry and country effects and series of industry-country variables stressed in the
literature.
32
As an additional test, the regressions in Table 4 include Herfindahl indices to control
for the possibility of high concentration leading both to high tariffs and vertical integration.
33
As seen in the table, the point estimates for the tariff coefficient remain very similar both in
terms of magnitude and in terms of significance. The Herfindahl indices, however, are never
significant. This may seem surprising, since industry concentration is likely to have also a direct
impact on prices and therefore on organizational choice. Note, however, that Herfindahl indices
are not a good proxy for the level of competition in manufacturing sectors because manufactures
31
These results are broadly consistent with the theoretical framework described by AJM. In their empirical
analysis they do not find a significant effect for the interaction between Capital intensity and Financial develop-
ment.
32
Our dataset contains different numbers of firms from different countries. This variation in the selection of
samples of firms could be a source of variation in vertical integration. The main source of the problem would be
potential correlation between vertical integration and firm size (combined with differential selection on firm size
across countries). Controlling for firm size partially alleviates this.
33
Because there may be an issue of endogeneity (concentration being caused by vertical integration), we do not
include this variable in the main analysis.
21
Page 22
are highly tradable. Hence, tariffs are a much better measure of the level of competition that
Herfindahls. Table 5 performs similar analysis using log(1 + variable) for the dependent and the
explanatory variables, obtaining similar results.
5 Trade Policy and Organization Convergence
The theoretical framework discussed in Section 2 suggests that trade policy should affect the
degree of organizational convergence across countries through its effect on prices. The focus
of this section is on cross-country differences in ownership structure at the sectoral level. For
each country, we thus construct an industry measure of vertical integration by estimating the
following regression model:
V
f,k,c
= β log(Employment
f
) + V
kc
+
f,c
. (10)
The estimate for the sector-country dummy V
kc
gives us a measure of the average level of vertical
integration of industry k in country c, controlling for the effect of firm size on the average level
of vertical integration in that industry-country pair.
5.1 Tariff Differences
We verify first whether cross-country differences in sectoral organizational structure are affected
by differences in tariffs. Our model predicts that, for a given country-pair cc
0
, organizational
differences should be smaller for those sectors characterized by similar levels of protection. To
verify this, we estimate the following model:
log |
ˆ
V
k,c
ˆ
V
k,c
0
| = α + β
1
log |t
c
kc
t
c0
k
| + β
2
log |X
k,c
X
k,c
0
| + δ
k
+ δ
cc
0
+
k,c,c
0
. (11)
The dependent variable is the log of the absolute difference between countries c and c
0
in
the estimated vertical integration indexes for sector k (from equation (10) above). The main
regressor of interest is the log of the absolute difference between these countries’ MFN tariffs
in sector k. The term |X
k,c
X
k,c
0
| captures differences in other sector-country characteristics
that may affect the degree of organizational convergence. Notice that, since we are including
dyad fixed effects (δ
cc
0
), β
1
is identified by the cross-sectoral variation in the tariff difference for
a given country pair.
In the first column of Table 6 the only explanatory variable is the log-difference in tariffs.
In line with our predictions, we find that for a given country-pair differences in sectoral vertical
integration indexes are significantly (at the 5 percent level) larger in those sectors in which
differences in MFN-tariffs are larger. A 100 percent increase in the difference in MFN tariffs
leads to a roughly 0.8 percent increase in the difference in vertical integration indexes.
22
Page 23
The second column adds the product of the sectoral import shares to which MFN tariffs
apply and the interaction of this product with the difference in MFN tariffs. The positive
effect of differences in MFN tariffs on differences in vertical integration seems to be larger if
MFN tariffs affects a larger fraction of trade for both countries. The marginal effect of log
MFN tariffs evaluated at the mean of the log product of MFN import shares is 0.01, which is
strongly significant.
34
The third column adds interactions between capital intensity and the log
difference in legal quality and the log difference in financial development. The coefficient on the
difference in MFN tariffs increases in magnitude becomes significant at the one percent level.
The interaction term of capital intensity with the difference in financial development is positive
and strongly significant, while the interaction term of capital intensity with legal quality is
surprisingly negative and strongly significant. Finally, column four includes as alternative control
variables the log difference in financial development interacted with external dependence and the
log difference in legal quality interacted with relationship-dependence. Again, the coefficient on
the difference in MFN-tariffs is not affected and is significant at the one percent level, while both
interaction terms are positive but not very significant.
In the right hand panel of Table 6 we repeat the same specifications with the average (over
all firm activities) sectoral vertical integration index. Results on the impact of tariff differences
remain robust to using this alternative measure of vertical integration, but are slightly less
significant.
5.2 Regional Trade Agreements
In the remaining of this section, we examine the relation between the degree of sectoral organiza-
tional convergence and common membership in a regional trade agreement. Note however that,
unlike for the previous regressions, it is harder to give a causal interpretation to these regression
results, since it is possible that countries that are generally more similar are also more likely to
form a RTA.
To assess the validity of our third empirical prediction, we explore how the existence of a RTA
between two countries affects the extent to which these countries have similar vertical integration
structures at the industry level.
log |
ˆ
V
k,c
ˆ
V
k,c
0
| = α + β
1
RT A
cc
0
+ β
2
AgeRT A
cc
0
+ β
3
X
cc
0
+ δ
k
+ δ
c
+ δ
c
0
+
k,c,c
0
. (12)
The dependent variable is as in model (11). The main regressor of interest is now a RT A
cc
0
,
a dummy that equals one if countries c and c
0
are member of the same regional trade agreement.
We also include the age of the trade agreement to capture the effect that older RTAs are likely to
have a larger impact on difference in organizational structure. The vector X
c,c
0
captures a series
34
The t-statistic for the marginal effect evaluated at the mean of the log product of MFN import shares is 2.87.
23
Page 24
of bilateral controls, such as dummies for contiguity, common language, colonial relationship, as
well as variables capturing the distance between the two countries, the difference in legal quality,
differences in financial development, differences in GDP and in GDP per capita. Finally, we
include sector fixed effects (δ
k
) and country fixed effects (δ
c
and δ
c
0
).
Table 7 presents the results for this regression. In the first column of the left panel, we
include only a dummy for regional trade agreements. Indeed, the coefficient of RTA is negative
and significant at the one percent level.
35
It implies that if a country pair has a RTA the
difference in vertical integration indices is around 13.6 percent smaller than for a country pair
without a RTA. The second column adds the age of the RTA as an additional control variable.
The coefficients for RTA and for age are both negative and significant at the one percent level.
Thus, as expected, country pairs with older RTAs have a more similar organizational structure
than countries with young RTAs. The coefficients imply that country pairs that have a RTA with
an average age (13.7 years) have a roughly 16 percent smaller difference in vertical integration
indices than country pairs without a RTA (0.06 0.0385 log(13.7) 0.16).
In the third column, we add a series of bilateral control variables that may have an influence
on the similarity of organizational structure. The coefficient of RTA is reduced somewhat in
size but remains significant at the one percent level. Contiguity and common language have
a significant negative effect on the distance in vertical integration indices and so has distance,
while the dummy for common colony is insignificant. Differences in legal quality, in GDP and
in GDP per capita have a significant positive effect on the difference in vertical integration,
while differences in financial development surprisingly seem to have a negative effect. The
fourth column presents results for a stricter definition of RTAs, which includes only free trade
agreements and custom unions (notified under GATT Article XXIV) but excludes weaker forms
of preferential trade agreements (notified under the Enabling Clause). Again, results remain
very similar and the coefficient is significant at the one-percent level.
6 Robustness checks
First, we repeat the three sets of regressions (9), (11), and (12) for the sample of countries for
which we observe at least 1000 plants that are part of firms with at least 20 employees. Table
A-5 presents the results for specification (9) in logs. It is obvious that the results remain almost
unchanged. The point estimates for the tariff coefficient remain very similar in magnitude and
also the significance of the estimates is not affected by restricting the sample of countries. Table
A-6 repeats the same regressions using log(1+variable), which exploits also the observations with
zero tariffs. Coefficients for tariffs drop slightly in magnitude but remain strongly significant.
35
SE are clustered by sector. Clustering at the country-pair level, which would be appropriate here, is not
possible because the panel is strongly unbalanced across sectors so that the clustered variance-covariance-matrix
becomes numerically singular.
24
Page 25
Turning to the results on organizational convergence, we find that tariff differences continue
to have a significant positive effect on differences in vertical integration in three out of four
specifications when using our main vertical integration index (see Table A-7). Results for the
average vertical integration index also remain stable. Again, they are slightly weaker than for
the primary index (see right panel of Table A-7). Finally, results for regional trade agreements
are also very robust, as is apparent from Table A-8. Having a RTA reduces differences in vertical
integration by roughly 5 percent when controlling for sector and country effects. The estimate of
the coefficient for RTA is robust across specifications and always significant at the one percent
level.
In a second set of regressions we include multinational firms in the sample. As we have noted
in the text, given that multinational firms have plants in different countries, it is not clear what
is the relevant product price and what tariffs are distorting it. We have used the primary activity
of the respective domestic ultimate to which the plant belonging to a multinational reports.
36
We
first report results for regression (9). In Table A-9 one sees that, multinational firms are much
more vertically integrated (multinationals have a 10% higher level of vertical integration than
non-multinationals). Moreover, the impact of tariffs on vertical integration drops by an order
of three to two in the first specification and when controlling for other sector-country variables,
respectively. The coefficient on tariffs remains significant when using the primary sector index.
In Table A-10 we repeat the regressions using log(1 + variable). The coefficient for tariffs now
becomes insignificant in most specifications.
When considering the impact of tariff differences on differences in vertical integration indices
in Table A-11, we find that the impact of tariff differences on differences in vertical integration
remains highly significant in most of our specifications. Finally, having a RTA continues to
reduce differences in vertical integration by around 5 percent, as one can see from A-12 and
the significant impact of RTA remains robust and significant at the one percent level across
specifications.
7 Conclusions
This paper describes a simple model in which firm boundaries depend on the prices of the
products they sell: the higher are prices, the more integrated firms will be. More generally, when
equilibrium prices converge across economies, so do ownership structures. The reason behind
these predictions is that ownership structure is chosen to mediate managerial tradeoffs between
private benefits against organizational goals. Integration has a comparative advantage relative
to non-integration in coordinating firms’ operating decisions, but does so at higher private costs.
As prices rise, the relative value of this coordinating role increases, favoring integration.
36
Note also that multinationals are usually active in many sectors and therefore the primary SIC code of their
global ultimate is not necessarily a good measure of their primary activity.
25
Page 26
To examine the validity of these results empirically, we use a new dataset from Dun and
Bradstreet (D&B) that contains both listed and unlisted plant-level observations in more than
200 countries. The dataset enables us to construct firm-level vertical integration indexes and to
study the link between product prices and firms’ ownership structure. In particular, we exploit
the cross-country and cross-sectoral variation in MFN tariffs and the existence of regional trade
agreements, which provide a source of price variation that is plausably exogenous to firms’
ownership decisions. In line with the predictions of the model, our empirical results shows that
higher prices, as proxied by higher MFN tariffs, lead to more vertical integration at the firm level.
Moreover, convergence in prices, as measured by more similar MFN tariffs and the membership
in RTAs, leads to convergence in organizational structure.
8 References
Acemoglu, D., P. Aghion, R. Griffith, F. Zilibotti (2009) “Vertical Integration and Technology:
Theory and Evidence,” Journal of European Economic Association, forthcoming.
Acemoglu, D., S. Johnson, and T. Mitton (2009) “Determinants of Vertical Integration: Finan-
cial Development and Contracting Costs,” Journal of Finance 63, 1251-1290.
Alfaro, L., and A. Charlton (2009) “Intra-Industry Foreign Direct Investment,” American Eco-
nomic Review 99, 2096-2119.
Antras, P. (2003). “Firms, Contracts, and Trade Structure,” Quarterly Journal of Economics
118, 1375-1418.
Antras, P., and H. Helpman (2004). “Global Sourcing,” Journal of Political Economy 112,
552-580.
Beck T., A. Demigurc-Kunt, and R. Levine (2006). A New Database on Financial Development
and Structure (1960-2006).
Bernard, A. B., J. B. Jensen, S. J. Redding, and P. K. Schott (2008). “Intra-Firm Trade and
Product Contractability”, mimeo.
Black, S. E., and P. E. Strahan (2002). “Entrepreneurship and Bank Credit Availability,”
Journal of Finance 57, 2807-2833.
Bloom, N., R. Sadun and J. Van Reenen, H. (2007). “The Organization of Firms Across
countries, NBER Working Paper No. 15129.
Bombardini, M. (2008). “Firm Heterogeneity and Lobby Participation,” Journal of International
Economics 75, 329-348.
Breinlich, H. (2008). “Trade Liberalization and Industrial Restructuring through Mergers and
Acquisitions,” Journal of International Economics 76, 254-266.
26
Page 27
Bustos, P. (2009). “Trade Liberalization, Exports and Technology Upgrading: Evidence on the
impact of MERCOSUR on Argentinean Firms,” American Economic Review, forthcoming.
Caves, R. E., (1975). “Diversification, Foreign Investment and Scale in North American Manu-
facturing Industries,” Ottawa: Economic Council of Canada.
Conconi, P., Legros, P., and A. F. Newman (2009). “Trade Liberalization and Organizational
Change”, CEPR Discussion Paper No. 7028.
Djankov, S., R. La Porta, F. Lopez-de-Silanes, and A. Shleifer (2003). “Courts”, Quarterly
Journal of Economics 118, 453-517.
Finger, J. M., H. K. Hall, and D. R. Nelson (1982). “The Political Economy of Administered
Protection,” American Economic Review 72, 452-466.
European Commission (Directorate-General for Economic and Financial Affairs) (1996). “Eco-
nomic Evaluation of the Internal Market.” European Economy No.4.
Goldberg, P. and N. Pavcnik, (2004). “Trade, Inequality, and Poverty: What Do We Know?
Evidence from Recent Trade Liberalization Episodes in Developing Countries,” NBER
Working Papers No. 10593.
Grossman S. J., and O. D. Hart (1986). “The Costs and Benefits of Ownership: A Theory of
Vertical and Lateral Integration,” Journal of Political Economy 94, 691-719.
Grossman, G. M., and E. Helpman (1995). “Protection for Sale,”American Economic Review
84, 833-850.
Grossman, G. M., and E. Helpman (2004). “Managerial Incentives and the International Orga-
nization of Production,” Journal of International Economics 63, 237-262.
Guadalupe, M., and J. Wulf (2008). “The Flattening Firm and Product Market Competition:
the Effect of Trade Liberalization,” Harvard Business School Working Paper 09-067.
Gwartney J., and R. Lawson (2003), “Economic Freedom in the World: 2003 Annual Report,”
mimeo, Fraser Institute.
Harrison, A. E., I. Love, and M. S. McMillian (2004). “Global Capital Flows and Financing
Constraints,” Journal of Development Economics 75, 269-301.
Hart, O. D., and Holmstr¨om, B. R. (2009). “A Theory of Firm Scope,” Quarterly Journal of
Economics, forthcoming.
Hart, O. D., and J. H. Moore (1990). “Property Rights and the Nature of the Firm,” Journal
of Political Economy 98, 1119-1158.
Helpman, E. (2006) “Trade, FDI, and the Organization of Firms,” Journal of Economic Liter-
ature 44,589-630.
27
Page 28
Hortacsu, A. and C. Syverson (2009). “Why do Firms Own Production Chains”, mimeo.
Klapper, L., L. Laeven, and R. Rajan (2006). “Entry Regulation as a Barrier to Entrepreneur-
ship,” Journal of Financial Economics 82, 591-629.
Legros, P., and A. F. Newman (2008). “Competing for Ownership,” Journal of the European
Economic Association 6, 1279-1308.
Legros, P., and A. F. Newman (2009). “A Price Theory of Vertical and Lateral Integration,”
mimeo, Boston University.
Lipsey, R. E. (1978). “The Creation of Microdata Sets for Enterprises and Establishments,”
Annales de l’INSEE 30 31, 395-422.
Marin, D., and T. Verdier (2008). “Power Inside the Firm and the Market: A General Equilib-
rium Approach,” Journal of the European Economic Association 6, 752-788.
McLaren, J. (2000). “Globalization and Vertical Structure,” American Economic Review 90,
1239-1254.
Mitra, D. (1999). “Endogenous Lobby Formation and Endogenous Protection: A Long-Run
Model of Trade Policy Determination,” American Economic Review 89, 1116-1134.
Nunn, N. (2007). “Relationship-Specificity, Incomplete Contracts, and the Pattern of Trade,”
Quarterly Journal of Economics 122, 569-600.
Ornelas, E., and J. Turner (2008). “Trade Liberalization, Outsourcing, and the Hold Up Prob-
lem,” Journal of International Economics 74, 225-241.
Rajan, R. and L. Zingales (1998). “Financial Dependence and Growth,” American Economic
Review 88, 559-586.
Schmitz, J. A. (2005), “What Determines Productivity? Lessons From the Dramatic Recovery
of the U.S. and Canadian Iron Ore Industries Following Their Early 1980s Crisis,” Federal
Reserve Bank of Minneapolis Staff Report No. 286.
Trefler, D. (2004). “The Long and Short of the Canada-U.S. Free Trade Agreement,” American
Economics Review 94, 870-895.
Van den Steen, E. (2005). “Organizational Beliefs and Managerial Vision,” Journal of Law,
Economics, and Organization 21, 256-283.
World Bank (2004). “Doing Business in 2004: Understanding Regulation”. Washington D.C.:
World Bank and Oxford University Press.
Woodruff, C. (2002). “Non-contractible Investments and Vertical Integration in the Mexican
Footwear Industry, International” Journal of Industrial Organization 20, 1197-1224.
28
Page 29
Table 1: Summary Statistics
Mean Median Min Max Std. Dev. N
Tariffs 5.40% 2.84% 0% 2553% 8.870 256,915
Herfindahl 0.153 0.068 0 1 0.203 173,467
Log (difference Tariffs) 1.688 1.897 -4.605 7.845 1.301 225,758
Log (difference Vertical integration index, primary sector) 0.385 0.544 -13.377 4.194 1.386 225,758
Log (difference Vertical integration index, average) 0.234 0.409 -10.795 3.842 1.296 225,758
Log (Capital intensity) -2.902 -2.857 -4.994 -1.354 0.458 387
Log (Relationship specificity) -0.526 -0.456 -3.449 -0.014 0.356 387
Financial dependence -0.524 -0.756 -15.523 53.628 3.058 387
Legal quality 0.670 0.663 0.210 0.972 0.189 90
Financial development 0.704 0.711 0.030 1.821 0.444 90
Log (GDP) 26.093 26.161 20.482 30.085 1.525 90
Log (GNI per capita) 9.486 9.811 6.346 10.671 0.927 90
Mean log (difference Legal quality) 0.191 0.178 0.000 0.566 0.124 90
2
Log (difference Financial Development) 0.414 0.409 0.000 0.956 0.254 90
2
Log (difference GDP) 26.539 26.591 12.743 30.085 1.812 90
2
Log (difference GNI per capita) 9.094 9.398 0.000 10.657 1.185 90
2
Log (distance) 8.620 9.011 4.665 9.892 0.971 90
2
RTA 0.394 0.000 0.000 1.000 0.449 90
2
FTA 0.262 0.000 0.000 1.000 0.440 90
2
Age RTA 13.697 15.000 1.000 44.000 8.605 90
2
Contiguity 0.041 0.000 0.000 1.000 0.197 90
2
Colonial Relationship 0.033 0.000 0.000 1.000 0.178 90
2
Common Language 0.121 0.000 0.000 1.000 0.326 90
2
Notes: Vertical integration indices constructed using plant level data from 2004 WorldBase, Dun & Bradstreet. Tariff data from TRAINS/WTO. Info on
regional trade agreements (RTAs) from WTO. Financial development from Beck, Demigurc-Kunt and Levine (2006). Legal quality from Kaufmann, Kraay,
and Mastruzzi (2003). GDP and GNI from World Bank. Capital Intensity from NBER-CES manufacturing industry database. Relationship specificity from
Nunn (2008). Financial dependence from Compustat following Rajan and Zingales (1998).
Page 30
Table 2: Tariffs and Vertical Integration, Firm Level Analysis
(1) (2) (3) (4) (5) (6)
Vertical integration index, primary sector Vertical integration index, average
log (Tariff) 0.0144* 0.0228*** 0.0241*** 0.0076 0.0159** 0.0171**
(0.0074) (0.0072) (0.0073) (0.0072) (0.0071) (0.0072)
log (Employment) 0.0341*** 0.0383*** 0.0382*** 0.0612*** 0.0748*** 0.0748***
(0.0054) (0.0069) (0.0069) (0.0048) (0.0059) (0.0059)
log (Capital intensity) x log (Financial development) 0.0555** 0.0484*
(0.0253) (0.0293)
log (Capital intensity) x log (Legal quality) -0.249*** -0.218**
(0.0921) (0.105)
External dependence x log (Financial development) -0.0009 -0.0006
(0.0027) (0.0026)
log (Relation specificity) x log (Legal quality) 0.0455 0.0409
(0.0413) (0.0399)
R2 0.059 0.065 0.064 0.04 0.047 0.047
# Observations 200093 170867 170867 200541 171252 171252
Number of SIC1 387 386 386 387 386 386
Sector Fixed Effect YES YES YES YES YES YES
Country Fixed Effect YES YES YES YES YES YES
Cluster Country- Country- Country- Country- Country- Country-
Sector Sector Sector Sector Sector Sector
Notes: Robust standard errors in parentheses denoting *** 1%, **5%, and *10% significance. Dependent variable: log of the vertical integration index of
firm f located in country c with primary sector k. In columns (1)-(3), the vertical integration index considers only the primary sector; in columns (4)-(6),
all manufacturing sectors in which the firm is involved. Sample includes firms 20 employees, no MNCs, tariff data.
Page 31
Table 3: Tariffs and Vertical Integration, Firm Level Analysis
(1) (2) (3) (4) (5) (6)
Vertical integration index, primary sector Vertical integration index, average
log (1+Tariff) 0.0094** 0.0180*** 0.0185*** 0.0012 0.0087* 0.0093**
(0.00446) (0.0049) (0.0051) (0.0047) (0.0045) (0.0045)
log (Employment) 0.0308*** 0.0350*** 0.0350*** 0.0511*** 0.0597*** 0.0597***
(0.0081) (0.0093) (0.0094) (0.0079) (0.0080) (0.0080)
log (Capital intensity) x log (Financial development) 0.0393*** 0.0369**
(0.0137) (0.0165)
log (Capital intensity) x log (Legal quality) -0.138** -0.134**
(0.0557) (0.0592)
External dependence x log (Financial development) -0.0002 -0.0005
(0.0012) (0.0012)
log (Relation specificity) x log (Legal quality) 0.0121 0.0193
(0.0293) (0.0275)
R2 0.060 0.064 0.064 0.041 0.046 0.046
# Observations 256889 226174 226174 256889 226174 226174
Number of SIC1 387 387 387 387 387 387
Sector Fixed Effect YES YES YES YES YES YES
Country Fixed Effect YES YES YES YES YES YES
Cluster Country- Country- Country- Country- Country- Country-
Sector Sector Sector Sector Sector Sector
Notes: Robust standard errors in parentheses denoting *** 1%, **5%, and *10% significance. Dependent variable: log of one plus the vertical integration
index of firm f located in country c with primary sector k. In columns (1)-(3), the vertical integration index considers only the primary sector; in columns
(4)-(6), all manufacturing sectors in which the firm is involved. Sample includes firms 20 employees, no MNCs, tariff data.
Page 32
Table 4: Tariffs, Concentration and Vertical Integration, Firm Level Analysis
(1) (2) (3) (4) (5) (6)
Vert. Int. Index, Primary Sector Vert. Int. Index, Average
log (Tariff) 0.0185** 0.0304*** 0.0308*** 0.00844 0.0190** 0.0193**
(0.0078) (0.0078) (0.0079) (0.0077) (0.0078) (0.0079)
log (Employment) 0.0836*** 0.105*** 0.105*** 0.0981*** 0.127*** 0.127***
(0.0054) (0.0068) (0.0068) (0.0054) (0.0065) (0.0065)
Herfindahl 0.00324 0.0025 -0.00068 -0.00078 -0.0157 -0.0186
(0.0328) (0.0430) (0.0431) (0.0292) (0.0366) (0.0367)
log (Capital intensity) x log (Financial development) 0.0823*** 0.0787**
(0.0308) (0.0361)
log (Capital intensity) x log (Legal quality) -0.305*** -0.299**
(0.1080) (0.1260)
External dependence x log (Financial development) 0.00856 0.0104
(0.0486) (0.0490)
log (Relation specificity) x log (Legal quality) -0.00145 -0.00209
(0.0033) (0.0037)
R2 0.094 0.105 0.104 0.063 0.077 0.076
# Observations 173467 145990 145990 173711 146184 146184
Number of SIC1 387 385 385 387 386 386
Sector Fixed Effect YES YES YES YES YES YES
Country Fixed Effect YES YES YES YES YES YES
Cluster Country- Country- Country- Country- Country- Country-
Sector Sector Sector Sector Sector Sector
Notes: Robust standard errors in parentheses denoting *** 1%, **5%, and *10% significance. Dependent variable: log of the vertical integration index of
firm f located in country c with primary sector k. In columns (1)-(3), the vertical integration index considers only the primary sector; in columns (4)-(6),
all manufacturing sectors in which the firm is involved. Sample includes firms 20 employees, no MNCs, tariff data.
Page 33
Table 5: Tariffs, Concentration and Vertical Integration, Firm Level Analysis
(1) (2) (3) (4) (5) (6)
Vertical integration index, primary sector Vertical integration index, average
log (1+Tariff) 0.0116** 0.0225*** 0.0227*** -0.00022 0.00935** 0.00950**
(0.0048) (0.0060) (0.0062) (0.0048) (0.0046) (0.0047)
log (Employment) 0.0761*** 0.0922*** 0.0922*** 0.0864*** 0.105*** 0.105***
(0.0121) (0.0117) (0.0117) (0.0117) (0.0102) (0.0102)
Herfindahl 0.00924 0.0112 0.0104 0.00968 0.00601 0.00543
(0.0186) (0.0228) (0.0229) (0.0180) (0.0215) (0.0222)
log (Capital intensity) x log (Financial development) 0.0455** 0.0493***
(0.0196) (0.0189)
log (Capital intensity) x log (Legal quality) -0.138* -0.157**
(0.0788) (0.0726)
External dependence x log (Financial development) -0.0121 -0.00616
(0.0344) (0.0327)
log (Relation specificity) x log (Legal quality) -0.00025 -0.00153
(0.0015) (0.0017)
R2 0.104 0.113 0.113 0.074 0.087 0.087
# Observations 225066 196290 196290 225066 196290 196290
Number of SIC1 387 387 387 387 387 387
Sector Fixed Effect YES YES YES YES YES YES
Country Fixed Effect YES YES YES YES YES YES
Cluster Country- Country- Country- Country- Country- Country-
Sector Sector Sector Sector Sector Sector
Notes: Robust standard errors in parentheses denoting *** 1%, **5%, and *10% significance. Dependent variable: log of one plus the vertical integration
index of firm f located in country c with primary sector k. In columns (1)-(3), the vertical integration index considers only the primary sector; in columns
(4)-(6), all manufacturing sectors in which the firm is involved. Sample includes firms 20 employees, no MNCs, tariff data.
Page 34
Table 6: Tariffs and Organizational Convergence, Country-Industry Differences in Vertical Integration
(1) (2) (3) (4) (5) (6) (7) (8)
Log (diff. Vertical integration index, primary sector) Log (diff. vertical integration index, average)
log (diff. Tariffs) 0.00786** 0.00771* 0.00986*** 0.00961** 0.00750*** 0.00335 0.00696** 0.00288
(0.00309) (0.00401) (0.00338) (0.00445) (0.00276) (0.00359) (0.00307) (0.00398)
log (product Import shares) 0.00168 0.00528**
(0.00264) (0.0025)
log (diff. Tariffs) x log (Import shares) 0.00599 0.0141**
(0.00814) (0.00707)
log (Capital int.)x log (diff. Financial dev.) 0.0198*** 0.00159
(0.00641) (0.00562)
log (Capital int.) x log (diff. Legal quality) -0.0135** -0.0155***
(0.00628) (0.00589)
log (External dep.) x log (diff. Financial dev.) 0.000465 -0.0004
(0.000739) (0.000759)
log (Relation spec.) x log (diff. Legal quality) 0.0123* 0.00207
(0.00687) (0.00628)
R2 0.085 0.086 0.089 0.108 0.135 0.141 0.135 0.165
# Observations 225758 196681 182474 99234 225781 196700 182494 99243
# Country pairs 4404 3532 3454 3185 4404 3532 3454 3186
Sector Fixed Effect YES YES YES YES YES YES YES YES
Diadic Fixed Effect YES YES YES YES YES YES YES YES
Cluster Country - Country - Country - Country - Country - Country - Country - Country -
Pair Pair Pair Pair Pair Pair Pair Pair
Sample All All All All All All All All
Notes: Robust standard errors in parentheses denoting *** 1%, **5%, and *10% significance. Dependent variable: log of the absolute difference between
countries c and c
0
in the estimated vertical integration index in the primary sector k. In columns (1)-(4), the vertical integration index considers only the
primary sector; in columns (5)-(8), all manufacturing sectors in which the firm is involved. Sample includes firms 20 employees, no MNCs, tariff data.
Page 35
Table 7: RTAs and Organizational Convergence, Country-Industry Differences in Vertical Integration
(1) (2) (3) (4) (5) (6) (7) (8)
Log (diff. Vertical integration index, primary sector) Log (diff. Vertical integration index, average)
RTA -0.136*** -0.0606*** -0.0462*** -0.116*** -0.0828*** -0.0366***
(0.00864) (0.0115) (0.00902) (0.008) (0.01) (0.00868)
Log (Age RTA) -0.0385*** -0.0171***
(0.00473) (0.00392)
Contiguity -0.140*** -0.118***
(0.015) (0.0145)
Colonial relation 0.0439** 0.0337
(0.0216) (0.0212)
Common language -0.0438*** -0.0493***
(0.00961) (0.00867)
log (distance) 0.00975** 0.0161***
(0.00488) (0.00546)
log (diff. Legal quality) 0.806*** 0.698***
(0.0492) (0.0494)
log (diff. Financial dev.) -0.116*** -0.119***
(0.0187) (0.0182)
log (diff. GDP) 0.0215*** 0.0203***
(0.00296) (0.00269)
log (diff. GNI per capita) 0.0210*** 0.00746**
(0.0035) (0.00308)
Free Trade Agreements -0.101*** -0.0861***
(0.00986) (0.00888)
R2 0.079 0.079 0.088 0.078 0.047 0.047 0.052 0.046
# Observations 324671 324671 260750 324671 324730 324730 260799 324730
# Sectors 458 458 458 458 458 458 458 458
Sector Fixed Effects YES YES YES YES YES YES YES YES
Country Fixed Effects YES YES YES YES YES YES YES YES
Cluster Sector Sector Sector Sector Sector Sector Sector Sector
Notes: Robust standard errors in parentheses denoting *** 1%, **5%, and *10% significance. Dependent variable: log of the absolute difference between
countries c and c
0
in the estimated vertical integration index in the primary sector k. In columns (1)-(4), the vertical integration index considers only the
primary sector; in columns (5)-(8), all manufacturing sectors in which the firm is involved. Sample includes firms 20 employees, no MNCs, tariff data.
Page 36
Table A-1: Sample Comparisons: Vertical Integration Indices
All Manufacturing
all plants firms firms firms firms firms firms
>0 empl 20 empl. 20 empl. 20 empl. 20 empl. 20 empl.
count. with no MNCs, count. With
1000 obs. 1000 obs.,
no MNCs
# of plants (thousands) 24,698 20,533 2,162 441 435 257 245
# of connected plants (thousands) 1,378 1,145 741 108 108 31 30
# of connected firms (thousands) 375 298 120 25 25 11 11
# of firms (thousands) 23,695 19,685 1,542 358 352 238 226
# of MNCs 28,662 28,207 24,631 15,317 15,232 0 0
Mean, Vertical integration index, primary sector 0.0359 0.0357 0.0575 0.0717 0.0717 0.0631 0.0629
Median, Vertical integration index, primary sector 0.0121 0.0100 0.0257 0.0509 0.0478 0.0439 0.0439
Min, Vertical integration index, primary sector 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
Max, Vertical integration index, primary sector 0.8333 0.8333 0.8333 0.8333 0.8333 0.8333 0.8333
St. dev., Vertical integration index, primary sector 0.0540 0.0551 0.0741 0.0749 0.0748 0.0633 0.0629
Mean, Vertical integration index, average 0.0355 0.0352 0.0551 0.0665 0.0665 0.0575 0.0572
Median, Vertical integration index, average 0.0151 0.0130 0.0334 0.0506 0.0506 0.0439 0.0437
Min, Vertical integration index, average 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
Max, Vertical integration index, average 0.5349 0.5349 0.5349 0.5198 0.5198 0.5198 0.5198
St. dev., Vertical integration index, average 0.0504 0.0512 0.0603 0.0588 0.0588 0.0527 0.0522
Mean, Employment 441 530 5010 3949 3966 252 245
Median, Employment 1 2 52 60 60 45 45
Min, Employment 0 1 20 20 20 20 20
Max, Employment 349,980 349,980 349,980 349,980 349,980 72,144 72,144
St. Dev., Employment 6,242 6,843 20,550 19,055 19,113 1,518 1,512
Notes: Plant- and firm-level data from 2004 WorldBase data, Dun & Bradstreet.
36
Page 37
Table A-2: Sample Frame
WB Code Freq. Percent Cum. WB Code Freq. Percent Cum.
AGO 8 0.00 0.00 KOR 3,103 1.21 65.91
ALB 4 0.00 0.00 KWT 66 0.03 65.94
ARE 442 0.17 0.18 LTU 221 0.09 66.02
ARG 1,031 0.40 0.58 LVA 137 0.05 66.08
AUS 5,911 2.30 2.88 MAR 610 0.24 66.31
AUT 1,819 0.71 3.59 MDG 19 0.01 66.32
BEL 1,399 0.54 4.13 MEX 3,081 1.20 67.52
BEN 4 0.00 4.13 MLI 13 0.01 67.53
BFA 9 0.00 4.14 MOZ 18 0.01 67.53
BGD 6 0.00 4.14 MRT 3 0.00 67.53
BGR 380 0.15 4.29 MUS 50 0.02 67.55
BOL 56 0.02 4.31 MWI 2 0.00 67.55
BRA 6,062 2.36 6.67 MYS 3,560 1.39 68.94
CAF 1 0.00 6.67 NER 1 0.00 68.94
CAN 8,141 3.17 9.84 NGA 134 0.05 68.99
CHE 1,508 0.59 10.42 NIC 22 0.01 69.00
CHL 469 0.18 10.61 NLD 1,940 0.76 69.76
CHN 28,487 11.09 21.69 NOR 1,522 0.59 70.35
COG 9 0.00 21.70 NPL 1 0.00 70.35
COL 563 0.22 21.92 NZL 1,110 0.43 70.78
CRI 183 0.07 21.99 OMN 70 0.03 70.81
CZE 2,008 0.78 22.77 PAK 4 0.00 70.81
DEU 21,420 8.34 31.11 PAN 70 0.03 70.84
DNK 1,011 0.39 31.50 PER 896 0.35 71.19
DOM 226 0.09 31.59 PHL 355 0.14 71.32
ECU 188 0.07 31.66 PNG 6 0.00 71.33
EGY 613 0.24 31.90 POL 470 0.18 71.51
ESP 2,363 0.92 32.82 PRT 5,764 2.24 73.75
EST 170 0.07 32.89 PRY 50 0.02 73.77
FIN 782 0.30 33.19 ROM 655 0.25 74.03
FRA 16,623 6.47 39.66 RWA 2 0.00 74.03
GAB 3 0.00 39.66 SAU 329 0.13 74.16
GBR 8,611 3.35 43.01 SEN 47 0.02 74.17
GEO 7 0.00 43.02 SGP 864 0.34 74.51
GHA 82 0.03 43.05 SLV 133 0.05 74.56
GRC 2,234 0.87 43.92 SVK 321 0.12 74.69
GTM 93 0.04 43.95 SVN 518 0.20 74.89
HND 80 0.03 43.99 SWE 1,833 0.71 75.60
HRV 171 0.07 44.05 TGO 4 0.00 75.60
HTI 4 0.00 44.05 THA 508 0.20 75.80
HUN 2,510 0.98 45.03 TTO 81 0.03 75.83
IDN 238 0.09 45.12 TUN 996 0.39 76.22
IND 2,629 1.02 46.15 TUR 2,691 1.05 77.27
IRL 676 0.26 46.41 TZA 26 0.01 77.28
ISR 1,838 0.72 47.13 UGA 40 0.02 77.29
ITA 8,965 3.49 50.61 URY 115 0.04 77.34
JAM 47 0.02 50.63 USA 57,929 22.55 99.89
JOR 148 0.06 50.69 VEN 256 0.10 99.99
JPN 35,862 13.96 64.65 ZAF 1 0.00 99.99
KEN 139 0.05 64.70 ZMB 17 0.01 99.99
ZWE 18 0.01 100.00
Total 256,915 100
Notes: Data from 2004 WorldBase data, Dun & Bradstreet. (Sample 20 employees, tariff data, employment,
NO MNCs.
Page 38
Table A-3: Sample Frame: Restricted Sample
WB code Freq. Percent Cum.
ARG 1,031 0.42 0.42
AUS 5,911 2.41 2.84
AUT 1,819 0.74 3.58
BEL 1,399 0.57 4.15
BRA 6,062 2.48 6.63
CAN 8,141 3.33 9.95
CHE 1,508 0.62 10.57
CHN 28,487 11.64 22.20
CZE 2,008 0.82 23.02
DEU 21,420 8.75 31.77
DNK 1,011 0.41 32.19
ESP 2,363 0.97 33.15
FRA 16,623 6.79 39.94
GBR 8,611 3.52 43.46
GRC 2,234 0.91 44.37
HUN 2,510 1.03 45.39
IND 2,629 1.07 46.47
ISR 1,838 0.75 47.22
ITA 8,965 3.66 50.88
JPN 35,862 14.65 65.53
KOR 3,103 1.27 66.80
MEX 3,081 1.26 68.06
MYS 3,560 1.45 69.51
NLD 1,940 0.79 70.30
NOR 1,522 0.62 70.92
NZL 1,110 0.45 71.38
PRT 5,764 2.35 73.73
SGP 864 0.35 74.08
SWE 1,833 0.75 74.83
TUN 996 0.41 75.24
TUR 2,691 1.10 76.34
USA 57,929 23.66 100.00
Total 244,825 100
Notes: Data from 2004 WorldBase data, Dun & Bradstreet. (Sample 20 employees, tariff data, no MNCs.
Page 39
Table A-4: Correlation Table
Log (dist Log (dist. RTA Log (dist.) Log (dist. Log (dist. Log (dist. log (dist.
Vert. Ind., Vert. Int. Fin. dev.) Legal GDP) GNI
Prim. Sect.) Average) quality) per cap.)
Log (dist. Ver. Int., Prim. Sect.) 1.0000
Log (dist. Vert. Int., Average) 0.6201 1.0000
RTA -0.0376 -0.0334 1.0000
Log (dist.) -0.0275 -0.0048 -0.4791 1.0000
Log (dist. Fin. dev.) 0.0072 0.0055 -0.1510 0.0964 1.0000
Log (dist. Legal qual.) 0.0764 0.0778 -0.2583 0.1804 0.6031 1.0000
Log (dist. GDP) -0.0345 -0.0482 -0.1681 0.0533 0.2916 0.0049 1.0000
log (dist. GNI per capita) 0.0586 0.0629 -0.2626 0.1231 0.5508 0.7203 0.2031 1.0000
Notes: Plant and firm level data from 2004 WorldBase data, Dun & Bradstreet.
Page 40
Table A-5: Robustness 1: Tariffs and Vertical Integration, Firm-Level Analysis, Countries 1000 plants
(1) (2) (3) (4) (5) (6)
Vertical integration index, primary sector Vertical integration index, average
log (Tariff) 0.0141* 0.0231*** 0.0253*** 0.00782 0.0166** 0.0186**
(0.0079) (0.0078) (0.0078) (0.0075) (0.0076) (0.0075)
log (Employment) 0.0340*** 0.0378*** 0.0379*** 0.0618*** 0.0759*** 0.0760***
(0.0058) (0.0075) (0.0075) (0.0051) (0.0063) (0.0063)
log (Capital int.) x log (Financial dev.) 0.0707** 0.0658*
(0.0303) (0.0353)
log (Capital int.) x log (Legal quality) -0.325*** -0.300**
(0.12) (0.137)
External dep. x log (Financial dev.) 0.0787 0.0714
(0.0563) (0.0557)
log (Relation spec.) x log (Legal quality) 0.00137 0.00107
(0.0031) (0.0031)
R2 0.060 0.065 0.065 0.047 0.046
# Observations 188931 161700 161700 189353 162065 162065
Number of SIC1 384 383 383 384 383 383
Sector Fixed Effect YES YES YES YES YES YES
Country Fixed Effect YES YES YES YES YES YES
Cluster Country- Country- Country- Country- Country- Country-
Sector Sector Sector Sector Sector Sector
Notes: Robust standard errors in parentheses denoting *** 1%, **5%, and *10% significance. Dependent variable: log of the vertical integration index of
firm f located in country c with primary sector k. In columns (1)-(3), the vertical integration index considers only the primary sector; in columns (4)-(6),
all manufacturing sectors in which the firm is involved. Sample includes firms 20 employees, countries 1000 plants, no MNCs, tariff data.
Page 41
Table A-6: Robustness 2: Tariffs and Vertical Integration, Firm-Level Analysis, Countries 1000 plants
(1) (2) (3) (4) (5) (6)
Vertical integration index, primary sector Vertical integration index, average
log (1+tariff) 0.00971** 0.0194*** 0.0205*** 0.0019 0.0103** 0.0115**
(0.0047) (0.0054) (0.0057) (0.0049) (0.0048) (0.0048)
log (Employment) 0.0310*** 0.0352*** 0.0352*** 0.0517*** 0.0606*** 0.0607***
(0.0082) (0.0096) (0.0096) (0.008) (0.0079) (0.0079)
log (Capital int.) x log (Financial dev.) 0.0465*** 0.0465**
-0.0166 (0.0185)
log (Capital int.) x log (Legal quality) -0.175** -0.180**
(0.0792) (0.0777)
External dep. x log (Financial dev.) 0.0287 0.0363
(0.042) (0.0393)
log (Relation spec). x log (Legal quality) 0.00133 0.000689
(0.0014) (0.0015)
R2 0.060 0.064 0.064 0.040 0.046 0.046
# Observations 244803 216320 216320 244803 216320 216320
Number of SIC1 387 387 387 387 387 387
Sector Fixed Effect YES YES YES YES YES YES
Country Fixed Effect YES YES YES YES YES YES
Cluster Country- Country- Country- Country- Country- Country-
Sector Sector Sector Sector Sector Sector
Notes: Robust standard errors in parentheses denoting *** 1%, **5%, and *10% significance. Dependent variable: log (1+ the vertical integration index of
firm f located in country c with primary sector k). In columns (1)-(3), the vertical integration index considers only the primary sector; in columns (4)-(6),
all manufacturing sectors in which the firm is involved. Sample includes firms 20 employees, countries 1000 plants, no MNCs, tariff data.
Page 42
Table A-7: Robustness: Tariffs and Organizational Convergence, Country-Industry Differences in Vertical Integration, Countries
1000 plants
(1) (2) (3) (4) (5) (6) (7) (8)
Log diff. Vertical integration index, primary sector Log diff. Vertical integration index, average
log (diff. Tariffs) 0.00735** 0.00789* 0.00806** 0.0066 0.00771** 0.00264 0.00677* -0.00172
(0.0037) (0.0048) (0.004) (0.0052) (0.0034) (0.0043) (0.0036) (0.0046)
log (Product Import shares) -0.00196 0.00965***
(0.0033) (0.003)
log (diff. Tariffs) x log (Import shares) -0.00514 0.0106
(0.0101) (0.0086)
log (Capital int.) x log (diff. Financial dev.) -0.00252 -0.00641
(0.0082) (0.0073)
log (Capital int.) x log (diff. Legal quality) -0.0088 -0.00627
(0.0074) (0.007)
log (External dep.) x log (diff. Financial dev.) 0.000322 -0.00033
(0.0008) (0.0008)
log (Relation spec.) x log (diff. Legal quality) 0.0158** 0.00619
(0.0078) (0.0074)
R2 0.106 0.106 0.109 0.130 0.145 0.146 0.148 0.180
# Observations 147642 135310 125751 70302 147652 135320 125761 70305
# country pairs 956 863 809 809 956 863 809 809
Sector Fixed Effects YES YES YES YES YES YES YES YES
Diadic Fixed Effects YES YES YES YES YES YES YES YES
Cluster Country - Country - Country - Country - Country - Country - Country - Country -
Pair Pair Pair Pair Pair Pair Pair Pair
Sample All All All All All All All All
Notes: Robust standard errors in parentheses denoting *** 1%, **5%, and *10% significance. Dependent variable: log of the vertical integration index of
firm f located in country c with primary sector k. In columns (1)-(3), the vertical integration index considers only the primary sector; in columns (4)-(6),
all manufacturing sectors in which the firm is involved. Sample includes firms 20 employees, countries 1000 plants, no MNCs, tariff data.
Page 43
Table A-8: Robustness: RTAs and Organizational Convergence, Country-Industry Differences in Vertical Integration, Countries
1000 plants
(1) (2) (3) (4) (5) (6) (7) (8)
Log diff. Vertical integration index, primary sector Log diff. Vertical integration index, average
RTA -0.0533*** -0.0382*** -0.0243*** -0.0576*** -0.0454*** -0.0173***
(0.0044) (0.0057) (0.0046) (0.0041) (0.0052) (0.0040)
Log (Age RTA) -0.0382*** -0.00599***
(0.0057) (0.0018)
Contiguity -0.0623*** -0.0648***
(0.0076) (0.0065)
Colonial Relation -0.0351** -0.00207
(0.0145) (0.0137)
Common Language -0.0231*** -0.0295***
(0.0051) (0.0046)
log (distance) 0.00321 0.0123***
(0.0026) (0.0025)
log (diff. Legal quality) 0.323*** 0.377***
(0.0257) (0.0265)
log(diff. Financial dev.) -0.0571*** -0.0839***
(0.0083) (0.0079)
log (diff. GDP) 0.00602*** 0.00583***
(0.0012) (0.0010)
log (Diff GNI per capita) -0.00022 -0.00048
(0.0016) (0.0014)
Free Trade Agreements -0.0244*** -0.0287***
(0.0044) (0.0039)
R2 0.083 0.084 0.084 0.083 0.059 0.059 0.063 0.058
# Observations 223887 223887 191459 223887 223887 223887 191459 223887
# Sectors 458 458 458 458 458 458 458 458
Sector Fixed Effects YES YES YES YES YES YES YES YES
Country Fixed Effects YES YES YES YES YES YES YES YES
Cluster Sector Sector Sector Sector Sector Sector Sector Sector
Notes: Robust standard errors in parentheses denoting *** 1%, **5%, and *10% significance. Dependent variable: log of the vertical integration index of
firm f located in country c with primary sector k. In columns (1)-(3), the vertical integration index considers only the primary sector; in columns (4)-(6),
all manufacturing sectors in which the firm is involved. Sample includes firms 20 employees, countries 1000 plants, no MNCs.
Page 44
Table A-9: Robustness: Tariffs and Vertical Integration, Firm Level Analysis, Multinationals
(1) (2) (3) (4) (5) (6)
Vertical integration index, primary sector Vertical integration index, average
log (Tariff) 0.0051 0.0165* 0.0184** -0.0003 0.0109 0.0125
(0.0102) (0.0089) (0.0090) (0.0095) (0.0085) (0.0086)
log (Employment) 0.149*** 0.168*** 0.167*** 0.151*** 0.170*** 0.169***
(0.0066) (0.0072) (0.0072) (0.0057) (0.0062) (0.0062)
Multinational 0.101*** 0.0680*** 0.0680*** 0.203*** 0.175*** 0.175***
(0.0209) (0.0202) (0.0203) (0.0169) (0.0167) (0.0167)
log (Capital int.) x log (Financial dev.) 0.0715** 0.0672**
(0.0298) (0.0327)
log (Capital int.) x log (Legal quality) -0.362*** -0.335***
(0.1090) (0.1160)
External dep. x log (Financial dev.) 0.0805* 0.0542
(0.0467) (0.0450)
log (Relation spec.) x log(Legal Quality) -0.00226 -0.00158
(0.0028) (0.0023)
R2 0.139 0.150 0.150 0.162 0.182 0.181
# Observations 242997 211712 211712 243484 212140 212140
Number of SIC1 387 387 387 387 387 387
Sector Fixed Effect YES YES YES YES YES YES
Country Fixed Effect YES YES YES YES YES YES
Cluster Country- Country- Country- Country- Country- Country-
Sector Sector Sector Sector Sector Sector
Notes: Robust standard errors in parentheses denoting *** 1%, **5%, and *10% significance. Dependent variable: log of the vertical integration index of
firm f located in country c with primary sector k. In columns (1)-(3), the vertical integration index considers only the primary sector; in columns (4)-(6),
all manufacturing sectors in which the firm is involved. Sample includes firms 20 employees, MNCs, tariff data.
Page 45
Table A-10: Robustness: Tariffs and Vertical Integration, Firm Level Analysis, Multinationals
(1) (2) (3) (4) (5) (6)
Vertical integration index, primary sector Vertical integration index, average
log (1+tariff) -0.00198 0.0101 0.0111* -0.00753 0.00352 0.00438
(0.0066) (0.0062) (0.0064) (0.0068) (0.0060) (0.0060)
log (Employment) 0.133*** 0.146*** 0.145*** 0.128*** 0.140*** 0.141***
(0.0121) (0.0108) (0.0108) (0.0090) (0.0072) (0.0075)
Multinational 0.0861*** 0.0620*** 0.0620*** 0.163*** 0.144*** 0.144***
(0.0165) (0.0178) (0.0178) (0.0102) (0.0110)
(0.0110) log (Capital int.) x log (Financial dev). 0.0411** 0.0495***
(0.0170) (0.0183)
log (Capital int.) x log (Legal quality) -0.189*** -0.212***
(0.0694) (0.0635)
External dep. x log (Financial dev.) 0.0682** 0.0419
(0.0299) (0.0.0300)
log (Relation spec.) x log (Legal quality) -0.00202 -0.00179
(0.0013) (0.0012)
R2 0.182 0.194 0.193 0.219 0.238 0.238
# Observations 313914 280951 280951 313914 280951 280951
Number of SIC1 387 387 387 387 387 387
Sector Fixed Effect YES YES YES YES YES YES
Country Fixed Effect YES YES YES YES YES YES
Cluster Country- Country- Country- Country- Country- Country-
Sector Sector Sector Sector Sector Sector
Notes: Robust standard errors in parentheses denoting *** 1%, **5%, and *10% significance. Dependent variable: log of one plus the vertical integration
index of firm f located in country c with primary sector k. In columns (1)-(3), the vertical integration index considers only the primary sector; in columns
(4)-(6), all manufacturing sectors in which the firm is involved. Sample includes firms 20 employees, MNCs, tariff data.
Page 46
Table A-11: Robustness: Tariffs and Organizational Convergence, Country-Industry Differences in Vertical Integration, Multinationals
(1) (2) (3) (4) (5) (6) (7) (8)
Log diff. Vertical Integration index, primary sector Log diff. Vertical Integration index, average
log (diff. Tariffs) 0.00978*** 0.0112*** 0.0102*** 0.00651 0.00998*** 0.00640* 0.0102*** 0.006506
(0.0027) (0.0034) (0.0031) (0.0042) (0.0026) (0.0033) 0.0031 0.0042
log (product Import shares) -0.002 -0.00224
(0.0023) (0.0023)
log (diff. Tariffs) x log (Import shares) -0.00589 0.0113*
(0.0068) (0.0067)
log (Capital int.) x log (diff. Financial dev.) 0.0147*** -0.00401
(0.0054) (0.0042)
log (Capital int.) x log (diff. Legal quality) -0.0194*** -0.00898*
(0.0068) (0.0048)
log (External dep.) x log (diff. Financial dev.) -0.00027 -0.00170**
(0.0007) (0.0007)
log (Relation spec.) x log (diff. Legal quality) 0.0129** 0.0138***
(0.0060) (0.0051)
R2 0.075 0.079 0.081 0.094 0.101 0.104 0.107 0.126
# Observations 248573 216544 199850 108982 248611 216574 249211 136403
# country pairs 4587 3577 3605 3223 4587 3577 3850 3487
Sector Fixed Effect YES YES YES YES YES YES YES YES
Diadic Fixed Effect YES YES YES YES YES YES YES YES
Cluster Country - Country - Country - Country - Country - Country - Country - Country -
Pair Pair Pair Pair Pair Pair Pair Pair
Sample All All All All All All All All
Notes: Robust standard errors in parentheses denoting *** 1%, **5%, and *10% significance. Dependent variable: log of the vertical integration index of
firm f located in country c with primary sector k. In columns (1)-(3), the vertical integration index considers only the primary sector; in columns (4)-(6),
all manufacturing sectors in which the firm is involved. Sample includes firms 20 employees, MNCs, tariff data.
Page 47
Table A-12: Robustness: RTAs and Organizational Convergence, Country-Industry Differences in Vertical Integration, Multinationals
(1) (2) (3) (4) (5) (6) (7) (8)
Log diff. Vertical Integration index, primary sector Log diff. Vertical Integration index, average
RTA -0.0796*** -0.0267*** -0.0362*** -0.0758*** -0.0297*** -0.0290***
(0.0050) (0.0060) (0.0046) (0.0046) (0.0055) (0.0043)
Log (Age RTA) -0.0273*** -0.0238***
(0.0025) (0.0025)
Contig. -0.0642*** -0.0535***
(0.0075) (0.0061)
Colonial Relationship 0.00884 0.0209**
(0.0098) (0.0087)
Common Language -0.0196*** -0.0304***
(0.0048) (0.0046)
Log (Distance) 0.00412 0.00804***
(0.0028) (0.0025)
Log (diff. Legal quality) 0.423*** 0.391***
(0.0275) (0.0242)
Log (diff. Financial dev.) -0.0304*** -0.0269***
(0.0096) (0.0102)
Log (Diff GDP) 0.00972*** 0.0129***
(0.0019) (0.0016)
Log (Diff GNI per capita) 0.0121*** 0.00960***
(0.0017) (0.0015)
Free Trade Agreements -0.0612*** -0.0570***
(0.0058) (0.0050)
R2 0.068 459 0.08 0.067 0.056 0.056 459 0.054
# Observations 358363 358363 285726 358363 358363 358363 285726 358363
# Sectors 459 0.069 459 459 459 459 0.068 459
Sector Fixed Effects YES YES YES YES YES YES YES YES
Country Fixed Effects YES YES YES YES YES YES YES YES
Cluster Sector Sector Sector Sector Sector Sector Sector Sector
Notes: Robust standard errors in parentheses denoting *** 1%, **5%, and *10% significance. Dependent variable: log of the vertical integration index of
firm f located in country c with primary sector k. In columns (1)-(3), the vertical integration index considers only the primary sector; in