American Economic Association
Trade, FDI, and the Organization of Firms
Author(s): Elhanan Helpman
Source: Journal of Economic Literature, Vol. 44, No. 3 (Sep., 2006), pp. 589-630
Published by: American Economic Association
Stable URL: http://www.jstor.org/stable/30032346
Accessed: 21-06-2017 09:54 UTC
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted
digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about
JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms
American Economic Association is collaborating with JSTOR to digitize, preserve and extend access to
Journal of Economic Literature
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
Journal of Economic Literature
Vol. XLIV (September 2006), pp. 589-630
Trade, FDI, and the Organization
of Firms
ELHANAN HELPMAN*
New developments in the world economy have triggered research designed to better
understand the changes in trade and investment patterns, and the reorganization of
production across national borders. Although traditional trade theory has much to
offer in explaining parts of this puzzle, other parts required new approaches.
Particularly acute has been the need to model alternative forms of involvement of
business firms in foreign activities because organizational change has been central in
the transformation of the world economy. This paper reviews the literature that has
emerged from these efforts. The theoretical refinements have focused on the individual
firm, studying its choices in response to its own characteristics, the nature of the
industry in which it operates, and the opportunities afforded by foreign trade and
investment. Important among these choices are organizational features, such as
sourcing strategies. But the theory has gone beyond the individualfirm, studying the
implications of firm behavior for the structure of industries. It provides new explanations
for trade structure and patterns of foreign direct investment, both within and
across industries, and has identified new sources of comparative advantage.
1. Introduction
International trade and foreign direct
investment (FDI) have been among the
fastest growing economic activities around
* Harvard University, Tel Aviv University, and CIAR. I
thank the National Science Foundation and the
U.S.-Israel Binational Science Foundation for financial
support. Parts of this paper were used in my European
Investment Bank Lecture, the European University
Institute, Florence. I am grateful to the Department of
Economics at the European University Institute for its
hospitality. Much of the work for this paper was done
when I was Sackler Visiting Professor at Tel Aviv
University. Pol Antras, Roger Gordon, Gene Grossman,
Marc Melitz, Stephen Redding, and three referees provided
useful comments.
the world. In 2003, world merchandise
exports were close to 7.3 trillion dollars;
world exports of commercial services were
close to 1.8 trillion dollars; and world FDI
inflows were close to 560 billion dollars.1
However, between 1990 and 2001 sales by
foreign affiliates of multinational corporations
1 FDI inflows reached a peak of 1.4 trillion dollars in
2000, but declined from 2000 to 2003; see UNCTAD
(2004). According to UNCTAD (2002), foreign affiliates of
multinational corporations accounted for 11 percent of
world GDP and 35 percent of world trade in 2001. In the
1990s, merchandise exports grew at an annual rate of 6.4
percent in real terms while merchandise production grew
at an annual rate of 2.5 percent only (see World Trade
Organization 2004).
589
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
590 Journal of Economic Literature, Vol. XLIV (September 2006)
expanded much faster than exports of goods
and nonfactor services.2 A striking feature of
this growth has been an unprecedented
expansion of FDI in services; the inward
stock of FDI in services increased from 950
billion dollars in 1990 to 4 trillion in 2002.3 In
2001-02, services accounted for two-thirds of
FDI inflows.
These remarkable figures mask equally
remarkable changes in the nature of trade
and FDI flows. The fast expansion of trade
in services has been accompanied by fastgrowing
trade in intermediate inputs.4
Moreover, the growth of input trade has
taken place both within and across the
boundaries of the firm, i.e., as intrafirm
and arm's-length trade.5 In the United
States, the latter has grown particularly
fast. Many studies have documented the
growth of international vertical specialization,
as reflected in the flows of inputs
across national borders for further processing
and final assembly.6 These trends are
closely related to the growing fragmentation
of production, in which multinational
corporations play a central role.
Technological change, such as computeraided
design and computer-aided manufacturing,
contributed to this process. The
2 According to UNCTAD (2002), by almost 7 percent
per year.
3 See UNCTAD (2004).
4 See Alexander J. Yeats (2001).
5 See Robert C. Feenstra (1998) and Maria Borga and
William J. Zeile (2004). According to Borga and Zeile
(2004), exports of U.S. parent companies to their foreign
affiliates for further processing have increased from 8.5
percent of total U.S. exports of goods in 1966 to 14.7
percent in 1999, and from 39.3 percent of total exports
of goods by U.S. parents to their foreign affiliates in
1966 to 64.7 percent in 1999. These shares vary substantially
across industries; they are particularly large in electronic
and other electric equipment as well as in
transportation equipment, and particularly small in
petroleum manufacturing as well as in food and kindred
products.
6 See, for example, Jose Campa and Linda S.
Goldberg (1997) for the United States, United
Kingdom, and Canada; Vanessa Strauss-Kahn (2003) for
France; and Hummels, Rappoport, and Yi (1998) and
Hummels, Jun Ishii, and Yi (2001) for other OECD
countries.
same technological changes also contributed
to growing outsourcing within and
across national borders.7
In addition to these broad trends, new
data sets enable researchers to uncover previously
unobserved patterns of trade and
FDI flows. Especially important is the finding
that a systematic relationship exists
between the characteristics of business
firms and their participation in foreign
trade and investment. Exporting firms are
not a random sample of the population of
firms in an industry, and neither are firms
engaged in FDI. Only a small fraction of
firms export, they are larger and more productive
than firms that serve only the
domestic market, and more firms export to
larger markets.8 A small fraction of firms
engage in FDI, and these firms are larger
and more productive than exporting firms.
A lot of within-industry heterogeneity
exists, and the distribution of firms by size
or productivity varies substantially across
industries.9
Sourcing strategies of business firms
have become more complex than ever
before, and so have the integration strategies
of multinational corporations.10 As a
result, the traditional classification of FDI
into vertical and horizontal forms has
become less meaningful in practice. Large
multinationals invest in low-cost countries
to create export platforms from which they
serve other countries around the world,
and the large flows of FDI across industrial
7 See Katharine G. Abraham and Susan K. Taylor
(1996) and Ann P. Bartel, Saul Lach, and Nachum
Sicherman (2005) on outsourcing trends in the United
States.
8 See Jonathan Eaton, Samuel Kortum, and Francis
Kramarz (2004). They report that only 17.4 percent of
French firms in manufacturing industries export, and they
export 21.6 percent of the aggregate manufacturing output.
These numbers hide large variations across industries,
however. In food and tobacco industries, for example, only
5.5 percent of the firms export, while in chemicals 55.4
percent of the firms export.
9 See Andrew B. Bernard, J. Bradford Jensen, and
Peter K. Schott (2005) for a portrait of U.S. firms.
10 See UNCTAD (1998).
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
Helpman: Trade, FDI, and the Organization of Firms 591
countries cannot be satisfactorily classified
as horizontal FDI.11
New theories have been developed to
explain these changes. While the new theories
do not replace comparative advantage
explanations of intersectoral trade and FDI
flows or replace imperfect competition
explanations of intraindustry trade, they do
bring to trade theory a new focus: the organizational
choices of individual firms. By
focusing on the characteristics of individual
firms, the theory can address new questions:
Which firms serve foreign markets?
And how do they serve them, i.e., which
choose to export and which choose to serve
foreign markets via FDI? How do they
choose to organize production, do they outsource
or integrate? Under what circumstances
do they outsource in a foreign
country rather than at home? And if they
choose integration, under what circumstances
do they choose to integrate in a foreign
country, via FDI, rather than to
integrate at home?12
I discuss this literature in two sections.
Section 2 examines insights from models of
heterogeneous firms in which the internalization
decision, i.e., outsourcing versus integration,
is put aside. This proves to be a
useful simplification because the resulting
predictions go a long way toward explaining
why firms sort into exclusive domestic producers,
exporters, or foreign direct investors,
and the structure of complex integration
strategies. Naturally, these models cannot
explain why some firms outsource while
others integrate. This issue is taken up in
11 See Karolina Ekholm, Rikard Forslid, and James R.
Markusen (2004) and Susan E. Feinberg and Michael P.
Keane (2003). See also section 2.5 for more details.
12 I attach traditional meanings to the terms "outsourcing"
and "integration." That is, outsourcing means the acquisition
of an intermediate input or service from an
unaffiliated supplier, while integration means production of
the intermediate input or service within the boundary of the
firm. These choices are distinct from the choice of country
in which to engage in these activities, because outsourcing
can be carried out in the home country of the firm, or in any
number of foreign countries, and similarly for integration.
section 3, which examines the implications
of the theory of incomplete contracts for
internalization and offshoring decisions. The
result is a trade theory with rich sourcing
patterns.13
Various studies emphasize different tradeoffs
in the decision to internalize or offshore,
and no model integrates all considerations
into a single framework. But the studies discussed
in section 3 all build on a common
assumption, namely that some inputs are
highly specific to a final product and that
their supply is not fully contractible. This
assumption is enough to study (1) the impact
of variations across industries in the intensity
of inputs that suffer from agency problems;
(2) Ricardian-type comparative advantage
that arises when legal systems of different
quality interact with sectoral differences in
contract dependency; (3) the impact of different
degrees of contract incompleteness,
which may vary across countries; (4) the role
of matching between buyers and sellers of
intermediate inputs, and the resulting "thick
market" effect; and (5) the interaction
between within-industry heterogeneity with
incomplete contracts, which yields joint predictions
about internalization and offshoring.
In particular, it predicts the relative prevalence
of the four main organizational forms:
integration at home, outsourcing at home,
integration abroad, and outsourcing abroad.
While the main purpose of this article is
to review the theoretical literature, I report
empirical evidence wherever possible. The
interplay between theory and empirics is
13 Some of the issues examined in section 3 are discussed
in Barbara J. Spencer (2005). I have chosen to focus
on incomplete contracts, thereby not covering the work on
managerial incentives, such as Gene M. Grossman and
Elhanan Helpman (2004) and Dalia Marin and Thierry
Verdier (2005). The reason for this choice is that there is a
lot of common ground in the approaches reviewed in section
3, while the papers on managerial incentives are
somewhat idiosyncratic. I also do not review earlier work
on incomplete contracts, such as Spencer and Larry D.
Qiu (2001) and Qiu and Spencer (2002), which have a narrow
focus, such as Keiretsu-type organizations, and have
no obvious implications for the broader issues discussed in
the introduction.
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
592 Journal of Economic Literature, Vol. XLIV (September 2006)
particularly important here because many
of these theoretical studies have been
motivated by evidence. As one would
expect, the theoretical models deliver new
empirical implications that can be confronted
with data. I report empirical studies
that do that, but other empirical
implications have not yet been tested.
Some will undoubtedly be tested in the
near future, while others will have to wait
because they require data that are not yet
available. These issues are discussed in the
closing section of the paper.
2. Heterogeneous Productivity
In the 1980s, trade theory introduced
within-industry heterogeneity resulting
from product differentiation and monopolistic
competition. Heterogeneity in these
studies was not designed, however, to
explain asymmetries across firms in productivity
or size. Not because it was not known
at the time that firms differ along these
dimensions, but rather because the aim was
to explain large volumes of trade between
countries with similar factor compositions
and large volumes of intraindustry trade.
For this purpose, differences in productivity
or size were not considered to be important.
As a result, the models assumed (for
the most part) symmetry across firms within
an industry in terms of the available
technology, which implied in turn similar
productivity levels and similar participation
in foreign trade. The monopolistic competition
models implied that all firms export to
all countries unless there is pressure for the
formation of multinational corporations.14
14 See Helpman and Paul R. Krugman (1985, chapters
7 and 12). Differential incipient pressure on factor prices
across countries can lead to the formation of multinational
corporations despite the prevalence of factor price equalization.
Under these circumstances, some firms become
multinationals while others do not. This produces asymmetries
in the organizational forms of different firms in the
same industry, and different trading patterns, but these
firms do not differ in productivity or size.
Detailed empirical studies of exporting
firms have led to a recognition of the limitations
of the symmetry assumption. As new
firm-level data became available, it became
clear that not all firms within an industry
export, nor are exporting firms a random
sample of the population of firms in an
industry. This evidence accumulated in the
1990s and showed that only a small fraction
of firms export and that exporters are
larger and more productive than nonexporters.15
Eaton, Kortum, and Kramarz
(2004) find, for example, that in the mid-
1980s only 17.4 percent of French firms in
manufacturing industries exported, that
they exported only 21.6 percent of their
output, and that both averages hide wide
variations across industries. And Helpman,
Marc J. Melitz, and Stephen Ross Yeaple
(2004) report that, in a large 1996 sample of
U.S. firms, exporters had a 39 percent labor
productivity advantage over nonexporters.16
Finally, there appear to exist
large sunk costs of exporting in developed
and developing countries alike.17
In view of these findings, Melitz (2003)
developed a theoretical model of monopolistic
competition with heterogeneous firms that
15 See Sofronis K. Clerides, Lach, and James R. Tybout
(1998) for Colombia, Mexico, and Morocco; Bernard and
Jensen (1999) for the United States; Bee-Yan Aw, Sukkyun
Chung, and Mark J. Roberts (2000) for Taiwan; Miguel A.
Delgado, Jose C. Farifias, and Sonia Ruano (2002) for
Spain; and John R. Baldwin and Wulong Gu (2003) for
Canada.
16 A detailed account of the characteristics of U.S. firms
that trade in goods is provided by Bernard, Jensen, and
Schott (2005). In their data too only a small fraction of
firms export and they export a small fraction of their own
output.
17 See Roberts and Tybout (1997) for Colombia and
Bernard and Jensen (2004) for the United States. While
these studies only report large persistence in exporting status,
Sanghamitra Das, Roberts, and Tybout (2005) use a
structural model to estimate the size of sunk exporting
costs for Colombian firms in three industries. They find
that the sunk costs for small producers are between
$412,000 and $430,000, and for large producers between
$344,000 and $402,000. Moreover, they find that fixed
exporting costs are important for at least some of these
firms.
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
Helpman: Trade, FDI, and the Organization of Firms 593
was designed to explain these features of the
data.18 His model has become the cornerstone
of a growing literature that examines the
role of heterogeneity in international trade
and foreign direct investment.19 The success
of Melitz's model derives from the fact that,
when combined with old and new approaches
to trade theory, it yields rich predictions that
can be confronted with data, and so far the
model has performed admirably well.
The main insights from Melitz's model are
derived from an interaction between productivity
differences across firms and fixed
costs of exporting. The fixed export costs are
interpreted as distribution and servicing
costs in foreign markets, and a firm has to
bear them in every country to which it
exports. As a result, the total fixed export
costs are larger the more foreign countries
the firm chooses to serve.20
To illustrate the nature of these interactions,
consider an industry supplying a differentiated
product, in which each of a
continuum of firms manufactures a different
brand. The demand function for firm j's
brand is x(j) = Ap(j)-e, where x is the quantity
and p is the price, A is a measure of the
demand level, and e=- 1/(1- a) is the
demand elasticity. The demand elasticity is
18 Other, related models of this type are developed in
Catia Montagna (2001) and S6bastien Jean (2002). In addition,
Bernard, Eaton, Jensen, and Kortum (2003) propose
a model of heterogeneous firms with a different market
structure (i.e., Bertrand competition instead of monopolistic
competition) in order to address similar questions. I
focus on Melitz (2003) because his model has proved to be
most adaptable to a wide range of applications, including
integration with the literature on incomplete contracts and
the international organization of production (see section
3). Richard E. Baldwin (2005) provides an alternative discussion
of this model.
19 Melitz (2003) builds on the work of Hugo A.
Hopenhayn (1992), who studied the entry and exit dynamics
of firms in an industry.
20 Earlier studies, including Baldwin (1988), Baldwin
and Krugman (1989), Dixit (1989), and Roberts and
Tybout (1997), used sunk costs of exporting, yet only
Roberts and Tybout touch upon some of the issues
addressed by Melitz. Their model, which is designed to
estimate the impact of sunk costs on export decisions, is
not as useful, however, for dealing with the wide range of
issues to which Melitz's model has been applied.
assumed to be constant, with 0 < a < 1,
which implies E > 1.21 Although the demand
level A is endogenous to the industry, it is
treated as exogenous by producers because
every producer is of negligible size relative
to the size of the industry.
Firmj discovers its productivity 6(j) only
after it enters the industry. Let clO(j) be its
variable production cost per unit of output
and let cf, be its fixed cost, where c measures
the cost of resources (e.g., the wage rate
when there is only labor input); and fD is a
measure of fixed production costs in terms
of resources. Then, if the firm chooses to sell
the product, its profit-maximizing strategy is
to charge p (j) = c/aO(j), which yields the
operating profits ir(j) = 6(j)e-1B - cfD,
where B - (1- a)A(c/a)1-e.
Figure 1 depicts these profits as a function
of the productivity measure =- =E-1. The
firm index j is dropped because profits do
not depend on the identity of the firm, only
on its productivity level; firms with higher
productivity have higher profits. The profit
function in the figure is:
(1) irD(O) = OB - cfD.
As is evident from the figure, firms with
productivity levels below 0D choose not to
produce because, for these firms, variable
profits do not cover their fixed cost, while
firmns with higher productivity supply their
brands to the market. Given a productivity
distribution G(O), we can calculate the fraction
of firms that serve the domestic market
as the fraction of firms with productivity
above the cutoff ED.
2.1 Export
Now interpret the profit function 7rD(e)
as applying to sales in the domestic market,
so that A is the demand level in the domestic
21 As is well known, this form of demand function can
be derived from a constant elasticity of substitution (CES)
utility or production function. In this event A = E/fjjj
p(j) dj, where E is total spending on these products and
J is the set of available brands.
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
594 Journal of Economic Literature, Vol. XLIV (September 2006)
Figure 1. Producing and Nonproducing Firms
market. And assume that firms can sell
their products in country f as well, which
has the demand function x(j) = Aep(j)-E.
That is, the demand elasticity is the same in
the two markets but the demand level is not
necessarily the same at home as in country
f. In addition, there are melting iceberg
trading costs for the shipment of every
brand of the product from home to f, such
that r > 1 units have to be shipped for one
unit to arrive, and there are fixed export
costs cfx. The variable trading costs typically
include transport costs, insurance, fees,
duties, and other impediments that may
stem from language barriers, differences in
the legal systems, and the like.22 Under
these circumstances, a firm that chooses to
sell in the domestic market, i.e., one with
productivity 0 > 08, can make additional
profits
22 See James E. Anderson and Eric van Wincoop (2004)
for estimates of the size of these costs.
(2) cf(O)
from export sa
o)AW(c/a)1-E.
Figure 2 depicts b
for the case in wh
and E-'fX >fD. Whe
els are the same, rD
result of the tradin
tion on the relative size of the fixed costs
then ensures O > ED. It follows that lowproductivity
firms, with 0 < OD, still
choose to close down, because they lose
money from domestic sales as well as from
exporting, while firms with productivity
above OD make money from serving the
domestic market. Now, however, high-productivity
firms, with 0 > 01, also make
money from exporting. Such firms choose,
therefore, to serve the domestic market as
well as the market in t. Firms with intermediate
productivity levels, between OD
and 0 attain the highest profits by serving
the domestic market only, i.e., they
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
Helpman: Trade, FDI, and the Organization of Firms 595
Figure 2. Exporting and Nonexporting Firms
choose not to export. The sorting pattern
depicted in this figure implies that exporting
firms are more productive than nonexporters
and that they are bigger. The last
implication follows from the fact that
more-productive firms sell more in the
domestic market and they sell in the foreign
country as well. Evidently, this
model's predictions are consistent with the
data, in which exporters are larger and
more productive than nonexporters.
Next observe that we can add as many
profit functions from exporting as there are
foreign countries [. Assuming that the foreign
countries differ only in market size, Af,
would then imply a negative correlation
between market size and the export cutoff
Of. That is, the smaller the foreign country 1
the larger its cutoff Oe. For simplicity, suppose
that min,(xW> OD.23 In this event, all
23 This should be true in big countries but may not be
true in small countries. In any case, the analysis can be carried
out without this assumption.
exporting firms sell in the domestic market
too, and there exist firms, with productivity
between 0D and mine(, which serve the
domestic market but do not export. All firms
with productivity levels above min[OjX
export. In this multicountry world, the positive
correlation between productivity and
export status is preserved. In addition, we
obtain a new prediction which is consistent
with the data: there exists a positive correlation
between the size of an export market
and the number of firms that export to it.24
Naturally, this correlation may not hold
when the trading cost 7 is not the same with
every foreign country. Nevertheless, it
should still hold once we control for the
cross-country variation in trading costs.
2.2 Turnover
I described a static version of Melitz's
(2003) model. This is sufficient for the issues
24 See the evidence in Eaton, Kortum, and Kramarz
(2004) for French firms.
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
596 Journal of Economic Literature, Vol. XLIV (September 2006)
discussed above as well as for a number of
other issues to be discussed below. Yet, the
original formulation of the model is dynamic,
shedding light on entry, exit, and turnover of
firms. In the dynamic version of the model,
the fixed production and export costsfD andfx
have to be borne every period. There also
exists an entry costfE that is a capital cost; it
has to be borne only once, at entry. Moreover,
there is a constant probability of death 6, of
every firm, irrespective of its productivity. In
this setting, free entry requires the expected
present value of profits to equal the entry
cost. In a steady state, firms constantly leave
the industry, as a fraction 6 die every period.
At the same time, there is a constant inflow of
new firms, and a fraction of these firmsthose
whose productivity is above the cutoff
OD-remain in the industry. In the steady
state equilibrium, the inflow equals the outflow,
so that the number of firms remains
constant in every productivity category. As a
result, the ratio of new entrants per period to
the stock of active firms, a measure of
turnover, equals 6/[1- G(OD)], where G(.) is
the cumulative distribution of 0.25 This setup
can be used to study the determinants of
turnover, which I illustrate in the next section
with a discussion of trade liberalization.
2.3 Trade Liberalization
Consider multilateral trade liberalization,
which leads to a proportional reduction of
trading costs r in all countries. On impact,
this reduction in trading costs raises the
profits of exporters and reduces the cutoff
Of. As a result, a larger proportion of firms
choose to export. But the presence of a larger
number of exporters in a market reduces
the demand facing every supplier, which
cuts into the profits of exporters and nonexporters
alike.26 After allowing the general
25 Let N be the stock of active firms and let nE be the
flow of new entrants per period. Then [1 - G(uDnllE is
the inflow of active firms and 6N is the outflow. In steady
state the two are equal. Therefore nE/N= 6/[1 - G(eD)].
26 See the determinants of the demand level A in footnote
21.
equilibrium effects to work themselves out,
the final outcome is a lower export cutoff xOf
(although not as low as one would predict
from the impact effect) and a higher domestic
cutoff OD. It follows that trade liberalization
leads to higher average productivity,
since only the more-productive firms survive
entry, and output is reallocated toward more
productive firms. These are interesting
implications, which illustrate important
issues that this model can address, and
which could not be addressed by earlier
models of international trade. Moreover,
Daniel Trefler (2004) finds that both of
these predictions are consistent with the
impact of the Canada-U.S. Free Trade
Agreement on Canadian industries.27
2.4 Horizontal FDI
Melitz's (2003) model can be generalized
to handle horizontal foreign direct investment.
The traditional classification of FDI
has been into horizontal and vertical FDI,
where the former concerns subsidiaries that
serve the local market in the host country
while the latter concerns subsidiaries that
add value to products that are not destined
(necessarily) for the host country market
(more on this in the next section).28
Following Helpman, Melitz, and Yeaple
(2004), suppose that a home-country firm
can build a (second) production facility in
country [, at cost cf, that will enable it to
produce its brand of the product in country
Sat unit cost c/IO, where 0 is the firm's productivity.
Then if the firm exports to country
1c, its profits from exporting are given by (2),
while if it chooses to serve the foreign market
via FDI, the firm's profits from FDI are
(3) jr7(E) - OBf-
27 Tybout and M. Daniel Westbrook (1995) also fin
important market share reallocations from low to high pr
ductivity plants in response to trade liberalization in Mexi
28 According to the BEA data, the destination of sales
U.S. subsidiaries are distributed as follows: 65 percent
the host country markets, 11 percent in the United State
and 24 percent in other countries (see J. Steven Landef
and Raymond Mataloni 2004).
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
Helpman: Trade, FDI, and the Organization of Firms 597
where Bfc= (1- a)Af(cf/a)1-e. Comparing
(2) with (3) we note that, as long as f, >fx
and ce < c', the firm faces a proximity-concentration
trade-off, for which Brainard
(1997) provides empirical evidence.
Namely, by choosing FDI instead of exporting
the firm gives up concentration of production,
which raises its fixed costs, but
saves on variable unit costs by avoiding
trade costs (and possibly on unit production
costs). Figure 3 describes this trade-off for
the case in which c= c, Be= B, (i.e., the
demand level is the same in the two countries),
and f,> r--fx>fD. Under these circumstances,
0>> 0ex> OD. It follows that
the most productive firms, with 0> Of,
serve the foreign market via subsidiary
sales; lower productivity firms, with
Of < E < Of, serve the foreign market via
export; and still lower productivity firms,
with D < ( < (x, serve only the domestic
market. Evidently, this sorting pattern is
consistent with the empirical evidence that
multinational corporations are more productive
than exporters who are not multinationals,
and exporters who are not
multinationals are more productive than
firms who serve only the domestic market.29
Since more productive firms produce
more output, this sorting pattern also
implies that multinational firms are larger
than exporters, and exporters are larger
than firms who serve only the domestic mar-
ket.
Helpman, Melitz, and Yeaple (2004) also
show that when the distribution of productivity
6 is characterized by a Pareto distribution,
the size distribution of firms also is
Pareto, and the model then predicts more
29 Helpman, Melitz, and Yeaple (2004) find that, in
1996, U.S. firms that engaged in FDI had a 15 percent
labor productivity advantage over exporters who did not
engage in FDI, and the latter had a 39 percent labor productivity
advantage over firms who engaged in neither
export nor FDI. See also Keith Head and John Ries (2003)
for evidence from Japan; Sourafel Girma, Holger Girg,
and Eric Strobl (2004) for evidence from Ireland; and
Girma, Richard Kneller, and Mauro Pisu (2005) for evidence
from the United Kingdom.
subsidiary sales relative to export sales in
sectors with greater productivity (and therefore
size) dispersion. This is a particularly
interesting implication because it suggests
that heterogeneity can be a source of comparative
advantage. The use of a Pareto distribution
is compelling in this case because
the actual size distribution of firms is well
approximated by such a distribution (see
Robert L. Axtell 2001). Helpman, Melitz,
and Yeaple also show that the shape parameter
of a Pareto distribution can be precisely
estimated in almost every one of fifty-two
sectors for which they have data, and these
estimates exhibit large variations in the
degree of dispersion across sectors. Using
these measures of dispersion, as well as nonparametric
measures, they estimate the
impact of heterogeneity on the ratio of subsidiary
sales to export sales of U.S. firms in a
sample of twenty-seven countries, and a
broader sample of thirty-eight countries,
both in 1994. Their estimates, which control
for the variation in fixed costs and other relevant
variables, are precise and consistent
with the theory. Moreover, the estimates are
large economically; they compare in size to
the impact of freight, tariffs, and measures
of fixed costs on the ratio of export to subsidiary
sales, which have been routinely used
in studies of the proximity-concentration
trade-off.30
2.5 Technology Adoption
Paula Bustos (2005) introduces a technology
choice into Melitz's (2003) model in order
to study the impact of trade liberalization on
technology upgrading in Argentina. For this
purpose, suppose that a firm located in
Argentina can serve the domestic market or it
can serve the domestic market and also export
to a foreign market (we disregard the FDI
option). But unlike the Melitz model, now,
upon entry, and after learning its productivity
6, the firm can choose to use an advanced
30 The comparability in size is of beta, or standardized,
coefficients.
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
598 Journal of Economic Literature, Vol. XLIV (September 2006)
Figure 3. Multinationals, Exporting, and Nonexporting Firms
technology H or a traditional technology L, as
in Yeaple (2005). The advanced technology
requires higher fixed costs, so that fDH >fDL
and fxH >fxL, but its advantage is that is has
lower variable costs, so that variable unit costs
are caH/6 if the firm uses technology H and
caL/O if the firm uses technology L, a, < aL.
With suitable restrictions on these parameters,
a firm with productivity below a cutoff
ED exits the industry because all choices
afford it negative operating profits, a firm
with productivity between OD and 0XL uses
technology L to serve the domestic market
only, a firm with productivity between OxL
and 0XH uses technology L to serve the
domestic market and to export, and a firm
with productivity above OxH uses technology
H to serve the domestic market and to export.
Naturally, OXH > OXL> OD. In other words,
more productive firms use the more
advanced technology but some lowproductivity
exporters use the traditional
technology. This sorting pattern is consistent
with Bustos's data.
Now consider multilateral trade liberalization,
which reduces trading costs to
Argentinian firms. This raises the operating
profits of all exporters, but proportionately
more so from the use of the advanced technology
if an exporter's productivity is close to
OXH. As a result, OXH declines, and some
exporters who used technology L switch to
H, while exporters who used the better technology
have no incentive to switch to technology
L. Firms that serve only the domestic
market also have no incentive to switch technologies,
and they keep using technology L.
The model therefore predicts that only firms
with intermediate productivity levels
upgrade their technology in response to
trade liberalization. And indeed, Bustos
finds an inverted U shaped relationship
between productivity and technology
upgrading in Argentina.31
31 Bustos (2005) also examines skill upgrading, which is
positively correlated with the upgrading of technology. She
finds that of the 17 percent rise in the demand for skilled
workers after trade liberalization, 15 percent took place
within firms in each of three skill categories: production,
nonproduction, and R&D workers. Since she has data on
the education level of workers within each one of these
skill categories, she measures skill upgrading as the rise in
average years of schooling.
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
Helpman: Trade, FDI, and the Organization of Firms 599
2.6 Complex Integration Strategies
Although horizontal FDI of the type
described in the previous section is prevalent,
the evidence points to a growing importance
of more complex integration strategies
by multinational corporations. Feinberg and
Keane (2003) find, for example, that among
U.S. multinationals with affiliates in Canada,
only 12 percent are of the purely horizontal
type (i.e., they have negligible intrafirm
flows of intermediate inputs) and only 19
percent are of the purely vertical type (i.e.,
they have negligible intrafirm flows of intermediate
inputs in one direction only). The
remaining 69 percent of the firms pursue
more complex integration strategies.32
Yeaple (2003) provides the first analysis of
such complex strategies, identifying an
important complementarity between the
two types of FDI. In what follows, I briefly
discuss insights from Grossman, Helpman,
and Adam Szeidl (forthcoming) who combine
heterogeneity features from Melitz
(2003) with the modelling of the two types of
FDI from Yeaple (2003) in order to explore
patterns of FDI in an environment that
offers a rich choice of integration strategies.
The model has a simple structure. There
are two symmetric countries in the North
and one country in the South. Every
Northern country has a population of firms
who know how to produce varieties of a differentiated
product. A typical firm has a production
function OF(m, a), where 0 is (as
before) a firm-specific productivity level and
F(.) is a concave constant-returns-to-scale
production function, common to all firms; m
represents intermediate inputs and a represents
assembly. That is, every final good is
produced with a combination of intermediate
inputs and assembly. The elasticity of
substitution between m and a is smaller than
one. And in this model there are no fixed
manufacturing costs fD nor fixed exporting
costs fx.
32 See also UNCTAD (1998), where the term "complex
integration strategies" was coined.
Intermediate inputs and assembly are produced
from a bundle of primary inputs at cost
c per unit, where c is higher in the North than
in the South. As a result, there is a cost advantage
to locating these activities in South,
unless other costs enter the calculus. To
introduce a tradeoff in the location decision,
it is assumed that no fixed costs are borne by
a firm that locates both activities in the
Northern country in which it is headquartered,
but that such a firm has to bear a fixed
cost g if it locates the production of intermediates
in a different country and a fixed cost f
if it locates assembly in a different country.
The firm may also incur transport costs for
either intermediate inputs or final goods. In
combination, this cost structure induces a
nontrivial decision problem in which the
optimal integration strategy depends on
these cost parameters as well as on the
demand levels in the three countries. The
demand function is Ap(j)-E (as before), and A
is higher in a Northern country than in South.
First consider the case in which there
are no transport costs. Then, given the
fixed cost f of FDI in assembly, there are
four integration strategies that may be chosen
by a firm in equilibrium, depending on
the fixed cost of FDI in intermediates g
and the firm's productivity. They are
depicted in figure 4. Region {S, H} describes
a strategy whereby the firm manufactures
intermediates in South and
assembles final goods in the home count
i.e., the country in which the firm is h
quartered. The other regions have simi
interpretations; the first letter denotes
location of intermediate inputs while
second letter denotes the location of
assembly. The fixed cost of FDI in inter
mediate inputs varies along the vertical ax
while the productivity measure =e-0-1
varies along the horizontal axis.
We see that for low fixed costs g the leastproductive
firms perform both activities
home, intermediate-productivity firms pro
duce intermediates in South and assemble
final goods at home, and high-productivity
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
600 Journal of Economic Literature, Vol. XLIV (September 2006)
Figure 4. Optimal Integration Strategies
firms perform both activities in South. That
is, the least-productive firms do not engage
in FDI; they produce intermediates and perform
assembly in the home country and
export the final product to the other
Northern country and to South. Firms with
intermediate productivity engage in partial
FDI; they produce intermediate inputs in
South, import them to the home country,
assemble them there into a final product,
and then export the final product to the
other Northern country and to South.
Finally, the most-productive firms engage in
FDI to the greatest possible extent; they
produce intermediate inputs in South and
assemble there the final product. The final
product is then exported to the two
Northern countries, i.e., the South serves as
an export platform to the rest of the world.33
The figure also shows that for an intermediate
range of FDI costs g there are only two
optimal integration strategies; low productivity
firms do everything at home while high
33 See also Ekholm, Forslid, and Markusen (2004) on
export-platform FDI.
productivity firms do everything in South.
Finally, for high values of g, low productivity
firms do everything at home, the highest
productivity firms do everything in South,
and firms in between produce intermediates
in the home country and assemble final
goods in South.
It is also clear from the figure that, given a
distribution of 0, the fraction of firms that
do both activities at home is rising with g
while the fraction of firms that do both activities
in South is declining with g. Moreover,
as shown by the broken lines, the fraction of
firms that assemble final goods in South
declines with g. That is, FDI in intermediates
and in assembly are complementary; as
the fixed cost of FDI in intermediate goods
increases, the fraction of firms assembling in
South declines.34
34 This is what Grossman, Helpman, and Szeidl (fothcoming)
call unit-cost complementarity, which has its origin
in Yeaple (2003). It arises from the fact that when
intermediates are produced in South at lower unit cost, it
becomes more attractive to assemble final goods there
because the larger final good sales make it easier to cover
the fixed cost of FDI in assembly.
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
Helpman: Trade, FDI, and the Organization of Firms 601
In the absence of trading costs, horizontal
FDI has no economic justification. And
indeed, figure 4 shows no instance in
which a firm in one Northern country
chooses to perform assembly in the other
Northern country. At most there is vertical
FDI (region {S, H}) and complex integration
(region {S, S}). But horizontal FDI
becomes a viable option when trade in final
goods is costly. So consider a modified version
of this model with melting iceberg
transport costs of final goods (but still free
trade in intermediate inputs). For low
transport costs, the equilibrium integration
strategies are the same as in figure 4. But
for intermediate levels of such transport
costs and relatively low demand in South,
the multinationals pursue different integration
strategies for high values of g. The
least-productive firms perform both activities
in the home country while the mostproductive
firms perform both activities in
South. However, firms with productivity
between these extremes produce intermediate
inputs in the home country but
choose different strategies for serving foreign
markets depending on how productive
they are within this range; the less-productive
firms choose subsidiary sales in the
other Northern country and export to
South, while the more-productive firms
choose subsidiary sales in both foreign
countries. As a result, all these firms engage
in horizontal FDI except that the moreproductive
firms do not export at all; they
serve every market with local subsidiary
sales. In this case too there is complementarity
between the two forms of FDI; as g
increases, a smaller fraction of firms engage
in subsidiary sales in foreign countries.35 As
35 The composition of FDI in assembly is now driven
by an additional source of complementarity, what
Grossman, Helpman, and Szeidl (forthcoming) call
source-of-components complementarity, which stems
from the fact that, for moderate transport costs of final
goods, a Northern market is cheaper to serve from assembly
lines in South if and only if intermediate inputs are
also produced in South.
Grossman, Helpman, and Szeidl (forthcoming)
show, this type of complementarily is
robust in the sense that it holds also for
high transport costs of final goods and for
high transport costs of intermediate
inputs.36
2.7 Variable Markups
The constant-elasticity demand function
that was used above has been the workhorse
of monopolistic competition studies
in economics, including international trade.
It is a convenient tool in many applications
and it is easily derived from either CES
preferences or a CES production function.
It has one particularly undesirable feature,
however: it implies that markups depend
neither on cost nor on demand levels.37 As
a result, the distribution of prices is a
scaled version of the distribution of marginal
costs, with no impact of market size or
the number of competitors on the shape of
the price distribution. Yet empirical evidence
on regional markets in the United
States suggests that higher demand, as
measured by market density, reduces
markups and price dispersion.38 Moreover,
with this type of demand, free entry implies
that total spending on the industry's products
has no effect on firm size because
higher spending raises the demand level A
but entry of new firms then reduces this
demand level, so that at the end of the
36 When transport costs of final goods are high and the
demand level in South is not very high, firms do not assemble
final goods in South for export to North; they either
export from the home country or they serve foreign markets
through subsidiary sales.
37 A firm with marginal cost c/O that faces the demand
function x(j)= Ap(j)-, where e= 1/(1- a)> 0, maximizes
profits by charging price p(j)= c/aO. Under these circumstances,
the ratio of price to marginal cost-which is a
measure of the markup-equals 1/a, and it does not
depend on marginal cost. Moreover, it does not depend on
the demand level A.
38 See Chad Syverson (2005) for a study of ready-mixed
concrete plants.
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
602 Journal of Economic Literature, Vol. XLIV (September 2006)
process A does not change.39 This too is
inconsistent with the evidence; market size
actually is positively correlated with firm
size.40 In order to accommodate these features
of the data, it is necessary to find an
alternative specification of demand in which
markups are endogenous. The theory will be
more consistent with the evidence when the
model implies that a larger market size
reduces a firm's markup because, in this
case, the firm also raises sales at constant
cost and productivity.
Although comparable evidence on variation
across countries does not exist, it is quite
likely that markups, prices, and firm size vary
across countries in similar fashion. To
address these issues, Melitz and Gianmarco
I. P. Ottaviano (2005) combine supply-side
features from Melitz (2003) with demand
side features from Ottaviano, Takatoshi
Tabuchi, and Jacques-Frangois Thisse (2002)
to construct a model of international trade
with variable markups in which market size
affects average prices, price dispersion, and
firm size. The model yields interesting predictions
concerning trade and the impact of
trade liberalization on productivity and price
distributions.
Ottaviano, Tabuchi, and Thisse (2002) use
the following quadratic, quasilinear utility
function:
u
-
39 For simplicity, conside
export opportunities. Using
p(j) = c/aO, a firm with prod
its that equal either B - cf
where B = (1- a)A(c/a)1implies
that the expected p
profits equals the entry co
depends on the cutoff ED
Together with the equation
the two equations uniquely
the value of A. It follows
products is precisely offset
(brands) so that A is not aff
40 See Jeffrey R. Campbe
evidence on retail trade indu
where x0 i
that yields
set of bran
and j7 are
all brands
brands are less substitutable for each other
the larger is y.41 Let y > 0. Then, assuming
that the consumer has enough income to justify
positive consumption of the outside
good, his demand for brandj is:
(4) x(j)
where N is the number of products
sumes and p is the average price
products.42 This is a linear demand
in which the demand level is decre
the own price, increasing in the
price p (i.e., the competitors' pric
declining in the number of produc
That is, as the competitive pressure
fies, either because prices of com
products decline or the number of
ing products increases, the manufac
brand j faces lower demand.44 In an
my populated by Q such consumers
gate demand for the brand equa
Facing production unit cost c/O, th
facturer maximizes profits by chargin
(5) p
41 This utility function represen
goods: a homogeneous outside goo
product with consumption x(j) o
alized to many goods.
42 The consumer chooses to pu
p(j)
and in an equilibrium with these types of consum
set J consists only of such products.
43 Note that the inequality in the previous foo
together with (4) imply P > 1.
44 J. Peter Neary (2003) uses a related demand
ture in his model of general oligopolistic competi
which preferences are not quasi linear and
Nevertheless, in his case too the demand level of
product depends on the average price of other pr
because it affects the marginal utility of consumpti
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
Helpman: Trade, FDI, and the Organization of Firms 603
Under these circumstances, the markupdefined
as the ratio of price to marginal
cost-is increasing in the average price and
declining in the number of products.
Moreover, the markup is increasing in y,
implying that it is higher the less substitutable
are brands for each other.
In this model, firms with very low productivity
do not produce even if they bear no
fixed production costs fD, because demand
drops to zero at a finite price.45 It is then
possible to solve the entire model of a closed
economy by adding a free entry condition,
with or without a positive fixed cost fD. The
solution for 0D then implies that the cutoff
for positive production is declining in y and
in Q. That is, less-productive firms survive
entry the less substitutable the products are
for each other and the larger the market size
is. It follows that in sectors with less substitutability
there is more productivity and size
dispersion and average productivity is
lower.46
Melitz and Ottaviano (2005) assume that
countries differ only in their numbers of
consumers Q, that there are neither fixed
production nor fixed export costs, but that
there are variable trade costs 'r> 0. The trade
costs introduce a degree of market segmentation
that produces cross-country variation
in the number of consumed products that is
positively correlated with market size, as
measured by the number of consumers.
Moreover, average productivity is higher in
larger markets, because low-productivity
45 With fD = 0, demand is not negative for
p(j)<
Together with (5) it implies
C
Therefore the cutoff productivity level 6D satisfies
OCy+r/N
46 Syverson (2004) provides evidence of these effects
for a cross section of U.S. manufacturing industries.
firms find it harder to compete in larger
markets. This is similar to the result in
Melitz (2003) but for different reasons. In
Melitz (2003), trade raises the profits of
high productivity firms-which exportand
these exporters raise the demand for
domestic inputs. As a result, domestic producers
who do not export are hit by competition
from foreign exporters on the one
hand and by higher input prices on the
other, which forces the least productive of
them to leave the business. The cutoff 0D
increases, and so does average productivity.
In contrast, in Melitz and Ottaviano (2005),
trade does not change input prices, but it
reduces markups as a result of the increase
in competitive pressure from foreign
exporters, and this raises the cutoff 0D and
average productivity. The implication is
that not only do consumers in larger markets
have access to more products, they also
pay lower prices.47
Multilateral liberalization raises the number
of products in all markets, which raises
competition and cuts into markups. Only
more-productive firms survive this pressure,
resulting in higher productivity and lower
prices. Evidently, this sort of trade liberalization
is beneficial to all countries concerned.
In contrast, when only a subset of countries
liberalize trade amongst themselves, the
impact on the liberalizing countries differs
markedly from the impact on the excluded
countries. In the former countries average
47 The differences between Melitz (2003) and Melitz and
Ottaviano (2005) stem from two sources: they use different
shapes of demand functions (constant elasticity in one case
and linear in the other) and they make different assumptions
about the outside good (no outside good with constant
marginal utility in one case and the presence of such a good
in the other). The absence of an outside good in Melitz
(2003) generates impacts on input costs that are absent in
Melitz and Ottaviano (2005), where quasi-linearity fixes the
unit cost c. But this is distinct from the impacts of the shapes
of the demand functions for final goods, which are isoelastic
in one case and linear in the other. In the isoelastic case, the
demand level has no effect on the cutoff because any shift in
demand is offset by entry, in contrast to the case of linear
demand. And while the markup is constant in the isoelastic
case, it responds to demand and entry in the linear case.
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
604 Journal of Economic Literature, Vol. XLIV (September 2006)
productivity rises, markups and prices
decline, and the number of products
increases. The opposite takes place in the
excluded countries. Under these circumstances,
the liberalizing countries gain while
the other countries lose.
2.8 Factor Proportions
Although Melitz (2003) places the firm at
the center of analysis, his approach has
implications for trade flows at the sectoral
level. This is apparent from the fact that
sectoral average productivity levels are
endogenous, and they depend on the determinants
of the sectoral cutoffs, OD (or OD).
These endogenous productivity levels generate
Ricardian-type comparative advantage
that affects the sectoral patterns of trade
flows.
Bernard, Stephen Redding, and Schott
(forthcoming) have extended the Melitz
(2003) model to accommodate variable factor
proportions, producing a richer model
of trade in differentiated products than the
standard Helpman and Krugman (1985)
version. They consider a two-sector, twofactor
world with constant expenditure
shares on each sector's output, CES preferences
for varieties in every sector, and
Cobb-Douglas production functions for
activities that generate either fixed or variable
costs. And they achieve great simplicity
by assuming that the Cobb-Douglas
production functions have the same exponents
in all activities within a given sector,
while they vary across sectors. In a world
with no trading frictions, i.e., neither fixed
nor variable costs of exporting, the analysis
proceeds along the now familiar lines of
the integrated equilibrium approach, with
results similar to Helpman-Krugman. The
sectoral cutoffs OD are not affected by
trade, and therefore neither are sectoral
productivity levels. The intersectoral pattern
of trade is of the Heckscher-Ohlin
type: every country is a net exporter of
goods that use relatively more intensively
the input with which the country is better
endowed.48
Next they introduce melting iceberg variable
trade costs and fixed export costs,
where the sectoral fixed export cost, arising
from a Cobb-Douglas production function,
has the same factor intensity as the other
sectoral activities. These costs segment
markets across countries. Now trade has an
influence on the cutoffs 0D; they rise in
every country and every industry. This
means that trade raises average productivity
everywhere in the world. Importantly,
however, in every country it raises average
productivity proportionately more in the
comparatively advantaged industry, i.e., the
sector that is relatively intensive in the
input with which the country is relatively
well endowed. Under the circumstances,
the Heckscher-Ohlin-type comparative
advantage, which emanates from factor
composition, also produces Ricardian comparative
advantage; and the two forms of
comparative advantage are positively correlated.
This is an important result, because
the empirical evidence suggests that it is
necessary to control for TFP differences
across countries in order to estimate the
impact of factor proportions on trade
flows.49 In addition, trade increases firm
size, and relatively more so in sectors having
comparative advantage. Finally, trade
raises the rate of gross job destruction and
gross job creation, thereby raising turnover.
But net job creation rises in comparatively
advantaged industries and declines in the
other sectors. These are very interesting
predictions that will undoubtedly influence
empirical analysis.
2.9 Gravity Equation of Trade Flows
The gravity equation is a major tool for the
empirical analysis of trade flows. It has been
48 This result does not hold in a world in which different
activities within a sector have different factor propor-
tions.
49 See Trefler (1995) and Donald R. Davis and David
E. Weinstein (2001).
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
Helpman: Trade, FDI, and the Organization of Firms 605
used to study the impact on trade flows of
international borders, currency unions,
membership in the WTO, and other variables.
And it has been used outside trade for
instrumental variable estimation of the
impact of variables such as social infrastructure
or political institutions on measures of
economic success.50 In all these applications,
the standard procedure is to estimate a
gravity equation of bilateral trade flows on a
sample of countries that export to each
other. This selected sample of countries represents,
however, only about half of the
country pairs in large samples of countries;
in the majority of the other half of country
pairs, the countries do not trade with each
other; and in the remaining pairs, one country
exports to the other but not vise versa.51
These facts raise two questions: First, what
accounts for the absence of trade among so
many pairs of countries? And second, to
what extent are estimates of trade flows that
disregard the nontrading countries reliable?
Helpman, Melitz, and Rubinstein (2006)
show that a modified version of Melitz (2003)
can account for the lack of trade between
potential trading partners and that the modified
theoretical framework provides guidance
for an estimation procedure that
exploits the information contained in the
zero trade flows. In particular, they argue
that lack of trade is not random, but rather
arises from economic conditions, and that
therefore we should simultaneously explain
which countries trade bilaterally and,
amongst those that do, how much is traded.
The model suggests that the standard estimation
procedure introduces two types of biases:
a sample-selection bias and an
omitted-variable bias. The sample-selection
bias problem is well known and it can be corrected
for with standard methods. The omitted-variable
problem is novel, however. It
stems from the fact that, in addition to the
50 See the discussion in Helpman, Melitz, and Yona
Rubinstein (2006).
51 See Helpman, Melitz, and Rubinstein (2006).
intensive margin of trade, i.e., the response
of a firm's export to changing conditions,
there is an extensive margin, which consists
of the response of the number of exporting
firms to changing conditions. If determinants
of the number of exporting firms are not
accounted for in the estimation of trade
flows, the resulting estimates suffer from an
omitted-variable bias. Helpman, Melitz, and
Rubinstein propose a method for decomposing
the impact of covariates, such as distance
between countries, on trade flows into an
intensive and extensive margin, and they find
that the extensive margin is empirically quite
important.
The main ingredient of the modified
model is a cumulative distribution function
of productivity 6 that has finite support
[OL, OH], where 0L is the lowest
productivity level and OH < c is the highest.
It is evident from figure 2 that if
OH )'-1 )'-1 falls between the domestic
cutoff 0D and the export cutoff Of, then
home firms produce for the home market
but none of them finds it profitable to
export to country f. Moreover, OH can be
below the export cutoff of some countries
and above the export cutoff of other countries,
so that domestic firms may find it
profitable to export to some countries but
not to others. The export cutoff O is
smaller the larger is the market in f and
the lower are the fixed export or trading
costs with f. The variables that affect the
cross-country variation in Of therefore
explain to which foreign countries the
home country should export.
Using the firms' optimal pricing and sales
strategies together with the free-entry condition,
the model implies two equations for
every export flow, say from country j to
country i. One equation describes the log of
exports from countryj to country i, mi, when
exports are positive, as a function of standard
covariates, such as distance between the
countries and whether they share a common
language, as well as exporter and importer
fixed effects. But in addition, it includes a
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
606 Journal of Economic Literature, Vol. XLIV (September 2006)
variable wij which is an increasing function
of the fraction of countryj firms that export
to country i:
mij = o+-X+XIn
this equation, A) is countryj's fixed effect
as an exporter, Xi is country i's fixed effect as
an importer, di is the distance between the
two countries, and uY is an error term that
describes the unobserved variation across
country pairs in variable trade costs.52
Covariates other than distance are accommodated
in similar fashion. The second
equation describes a latent variable that is
positive if and only if j exports to i. It is
defined as the log of the ratio of variable
export profits for the most productive firm to
the fixed export cost, where the latter is the
same for all firms inj. When this latent variable,
zi, exceeds zero, some firms from
country j export to country i because the
most productive firm makes profits large
enough to cover the fixed export cost. The
resulting equation is
zj=
wherej is an export
importer fixed ef
variable that imp
exporting from j
term that combines the unobserved variation
across countries in variable trade costs
uij (which also appears in the trade flow
equation) and a variable vi that represents
unobserved variation across countries in
the fixed export costs. Evidently, r/j is correlated
with uj.
It is common to estimate the equation of
trade flows mj for country pairs with positive
trade flows, without controlling for the
impact of the fraction of exporting firms
through w These are the sources of the
selection and omitted-variable biases.
52 The need for separate importer and exporter fixed
effects has been well known; see, for example, Feenstra
(2004). The theoretical model provides, however, a clear
interpretation of the determinants of these fixed effects.
Helpman, Melitz and Rubinstein (200
show how to use standard methods to correct
for the sample selection bias, by estimating
the equation for zi and the equation
for mij. This bias turns out to be rather small
in their data. In addition, they show how to
account for the impact of w2 on the trade
flows. Their procedure recognizes the fact
that no data on w2 are available, nor are
there data on the fraction of exporting firms,
a variable that impacts wj.53 In particular,
they show that the estimated (selection)
equation for zij can be used to construct estimates
of the ws, which can then be used in
the estimation of the trade flow equation. In
this way one can separately identify the
impact of a variable, such as distance di, on
the intensive margin (via its direct impact on
mij) and the extensive margin (via its impact
on mj through wj). The theoretical underpinning
of this procedure derives from the
fact that fixed costs affect the latent variable
z2 directly, but affect the trade flow mj onlyindirectly via their impact on the fraction of
exporting firms.
It should be evident from this section that
productivity differences across firms in an
industry have important implications for
trade, trade policies, and FDI flows. Not only
does this new way of thinking shed light on a
host of substantive issues, it also helps in formulating
better empirical strategies for estimating
trade flows. To illustrate, this approach
suggests that countries seeking to integrate
into the world trading system or to join a free
trade area should expect substantial reallocations
across firms within industries, which will
impact sectoral productivity levels. In addition
to raising the profits of exporting firms, lower
protection raises the competitive pressure
from foreign enterprises, which cuts into the
domestic firms' markups, raises domestic
53 That is, there are no data on the fraction of exporting
firms for the large samples of countries used to estimate
gravity equations. Note also that even if data on w were
available, it could not be used directly because w is an
endogenous variable, and one would therefore need to
instrument it.
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
Helpman: Trade, FDI, and the Organization of Firms 607
factor costs, and drives the country's leastproductivity
firms out of business.
3. Incomplete Contracts
My discussion of trade and FDI has so far
focused on final products.54 Importantly,
none of the studies reviewed in the previous
section, where firms make FDI choices,
explicitly analyzes the internalization decision.
That is, it is assumed that foreign operations
are organized in foreign affiliates, be it
for the purpose of manufacturing final products
designated for the foreign market, manufacturing
components in a foreign country
to be assembled at home or in a foreign
country, or assembling final goods to be sold
in the home or foreign market. Yet the
choice of whether to manufacture components
inhouse or acquire them from an unaffiliated
firm is a key decision about
organizational form, as is the decision of
whether to source such components at home
or in a foreign country. The same applies to
assembly, which is just another activity in the
chain of tasks that need to be performed in
order to deliver a product to a final user. A
better understanding of these choices is
needed in order to explain the trends in trade
and FDI and their relation to the evolving
organization of production and distribution.
Two facts stand out that triggered a
major research effort into the international
organization of production. First, with the
advent of computer-aided design, computer-aided
manufacturing, and institutional
changes in labor markets, outsourcing has
rapidly expanded.55 This is true about both
54 The only exception being the discussion of the interrelationship
between FDI in components and FDI in
assembly in section 2.6.
55 I use "outsourcing" to mean the acquisition of an
input or service from an unaffiliated company. This is the
standard terminology used in industrial organization. A
narrower definition is used in some of the literature, e.g.,
Mary Amiti and Shang-Jin Wei (2005). A notable example
of a very narrow definition is Jagdish Bhagwati, Arvind
Panagariya, and T. N. Srinivasan (2004), who restrict the
term to outsourcing of services from foreign unaffiliated
companies. I find it preferable to use the traditional defi-
nition.
domestic and international outsourcing,
where rising domestic outsourcing means
an increase in the purchase of intermediate
goods and services from domestic unaffiliated
firms, and rising foreign outsourcing
means an increase in the purchase of intermediate
goods and services from foreign
unaffiliated firms. These trends have been
widespread across different sectors and
different inputs.56 Second, the sourcing of
inputs from foreign countries has increased
at a rapid pace, both via arm's-length trade
(outsourcing) and via intrafirm trade
(FDI), a phenomenon known as offshoring.57
In order to understand these
trends, we need to understand the twodimensional
decision problem of business
firms: whether to outsource or insource
(i.e., integrate), and whether to offshore or
not.58 This choice yields four possibilities:
insourcing at home, outsourcing at home,
insourcing abroad (FDI), and outsourcing
abroad. The first two organizational forms
do not involve foreign trade, while the latter
two do: intrafirm trade in the case of
56 See Edward J. Bardi and Michael Tracey (1991),
Elizabeth Gardner (1991), Susan Helper (1991), James
Bamford (1994), Abraham and Taylor (1996) and Bartel,
Lach, and Sicherman (2005) for evidence on the growth of
outsourcing in various industries. The Economist (1991)
provides an early overview.
57 Feenstra and Gordon H. Hanson (1996) find more
than a doubling of the share of imports in total purchases
of intermediates from 1972 to 1990 in the United States
(from 5.0 percent to 11.6 percent), while Campa and
Goldberg (1997) find similar trends in Canada and the
United Kingdom. And Hummels, Ishii, and Yi (2001) and
Yeats (2001) find that foreign trade in components has
grown faster than foreign trade in final goods. Finally,
Hanson, Mataloni, and Matthew J. Slaughter (2005) find
that intrafirm trade within U.S. multinationals has grown
very fast, although somewhat less than international outsourcing
by U.S. firms, and Feinberg and Keane (2005)
report that sales of U.S. parent firms to their Canadian
affiliates as a fraction of the affiliates' total sales, as well as
sales of the Canadian affiliates to their U.S. parents as a
fraction of the parents' total sales, have almost doubled
between 1984 and 1995.
58 Two additional decisions, which are equally important
but received only scant attention, concern the types of
inputs that should be acquired by means of each one of
these organizational forms, and if an input is to be sourced
aboard, to which country it should be offshored.
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
608 Journal of Economic Literature, Vol. XLIV (September 2006)
FDI, and arm's-length trade in the case of
outsourcing.59
An analysis of these issues by means of the
incomplete contracts approach to the theory
of the firm helps in understanding why some
companies source inputs abroad primarily
via FDI, while other companies source them
abroad primarily via outsourcing. It also
helps in understanding why intrafirm trade
as a fraction of total trade is positively correlated
with capital-labor ratios across U.S.
industries and capital-labor ratios across
countries from which the U.S. imports (see
Pol Antrhs 2003). Moreover, this approach
helps in understanding why differences
across countries in the quality of legal systems
generate comparative advantage, and
thereby impact the patterns of trade (see
Andrei A. Levehenko 2004 and Nathan
Nunn forthcoming). Finally, when combined
with productivity variation across firms within
industries, this approach helps in predicting
the relative prevalence of alternative
forms of the international organization of
production as a function of sectoral characteristics
and differences in features of the
trading partners.
3.1 The Incomplete Contracts Approach
To illustrate the incomplete contracts
approach to the theory of the firm, consider
the following example.60 A final good producer
makes profits r0 0 0 if she does not use
a specialized intermediate input. If, however,
she uses one unit of the specialized intermediate
input, her profits become r1 > r,
where zr does not include the cost of the
59 The segmentation of production across different
countries has become so large that it prompted the WTO
to describe in its 1998 annual report the detailed acquisition
of inputs by U.S. car manufacturers in different countries,
concluding that only 37 percent of a car's value was
generated in the United States. Rone Tempest (1996)
describes an equally global sourcing strategy of Mattel in
the manufacturing of Barbie dolls (see Feenstra 1998).
60 See Oliver D. Hart (1995) for a detailed discussion of
this approach, and Patrick Bolton and Mathias
Dewatripont (2005) for a textbook treatment.
input to the final good producer. For simplicity,
assume that the final good producer
can use only one unit of this input.
In order to acquire the input, the final
good producer needs to engage a supplier.
The supplier can produce the required
input at cost c. Importantly, however, the
final good producer and the supplier cannot
sign an enforceable contract that specifies
the nature of the specialized intermediate
input, but the final good producer can recognize
ex post, after the input is delivered,
whether the input has the requisite features.
For this reason the supplier can
choose the characteristics of the input, and
when delivering it to the final good producer
he can bargain with the final good producer
for payment. At the bargaining stage,
the cost c of the intermediate input is sunk,
and it therefore plays no role in determining
the bargaining outcome. But it does
play an important role in determining
whether the supplier chooses to manufacture
the requisite intermediate input in the
first place.
There are two stages to the game. In the
first stage, the supplier decides whether to
manufacture the intermediate input and, if
he does, whether to endow the input with
the special characteristics requested by the
final good producer. In stage two, the supplier
delivers the input and bargains for payminent.
As usual, the game is solved
backwards, starting from stage two.
Assume that whenever the final good producer
and the supplier bargain they reach an
agreement according to the Nash bargaining
solution, with the bargaining weight P/e
(0, 1) for the final good producer and 1-/3
for the supplier. In this event, their payoffs
are derived as follows. In case of a breakup
of the negotiation, the final good producer
has the outside option ro while the supplier
has the outside option r0 0 0. The size of oo
depends on how specialized the intermediate
input is. If, for example, it is so highly
specialized that no one else can use it, i.e., it
has no value outside the relationship, then
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
Helpman: Trade, FDI, and the Organization of Firms 609
0o = 0. If, however, the intermediate input
can be used by other manufacturers, then
o > 0. In either case, the payoff of the
supplier equals his outside option plus the
share 1- /P of the surplus from the relationship,
while the payoff of the final good
producer equals her outside option w0 plus
the share /3 of the surplus from the relationship.
The size of the surplus depends
on whether the intermediate input satisfies
the specifications needed by the final good.
If it does, the surplus equals i minus the
outside options of the two players; that is, it
equals r--o0-0. If it does not, it equals
-r0 - o0, because the input adds no value to
the relationship. Naturally, the latter case
does not arise in equilibrium. Therefore in
an equilibrium in which the supplier delivers
the requisite input the payoffs from the
bargaining game are
Pf= go+ P(g,- no-0-0)
for the final good producer and
P,=
for the supplier. Note
they split the profits
as the payment of th
the supplier for the
that her profits net o
equal her payoff Pf.
This solution to t
determines the ince
to engage in a busi
the final good produ
game). If P, > c, thi
a profitable deal; o
That is, in case Ps <
produce the spec
input.
In this example, there is a one-sided
holdup problem. The supplier is held up by
the final good producer because he makes a
relationship-specific investment. More generally,
however, the final good producer may
also be required to make a relationshipspecific
investment, in which case there 41will
be a two-sided holdup problem. Moreover,
the outside options of the two parties may
depend on the organizational form of the
business firm, e.g., whether the intermediate
input is produced in-house or outsourced. In
the former case, the supplier is an employee
of the final good producer while in the latter
case he is not (see Sanford J. Grossman and
Hart 1986 and Hart 1995). Finally, when
these interactions are placed in a general
equilibrium setup, the outside options
become endogenous, and they depend on
the nature of the technology and the organization
of the industry (see Grossman and
Helpman 2002).
I now turn to the application of this
approach to the main issues discussed in the
preamble to this section. I have ordered the
topics in a way that eases the exposition, and
not in the order in which they appeared in
the literature.
3.2 Contractual Input Intensity
Traditionally, input (or factor) intensity
refers to the relative requirement of various
inputs (or factors of production) in the manufacturing
of a good. But the theory of
incomplete contracts identifies another
important measure of input intensity: the
relative requirement of intermediate inputs
that are under direct control of the final
good producer, and intermediate inputs that
require the engagement of suppliers. The
importance of the second intensity measure,
which I term the contractual input intensity,
stems from the fact that intermediate
inputs under the direct control of the final
good producer suffer less form agency problems
than intermediate inputs that require
the engagement of suppliers. Naturally, the
two intensity measures can be correlated.
And when they are, the theory yields interesting
predictions about the structure of
trade.
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
610 Journal of Economic Literature, Vol. XLIV (September 2006)
The contractual input intensity impacts
the power of incentives that a final good
producer wants to give a supplier. In particular,
the more intensive the production
process is in intermediate inputs that are
controlled by suppliers, the more powerful
incentives she wants to give the suppliers.
Yet her most desired incentives are not
extreme. That is, she never wants to give
them the strongest possible or the weakest
possible incentives. Under the circumstances
her choice of organizational form,
such as outsourcing versus integration, is
determined to an important degree by its
effect on the incentives of suppliers.
To understand the role played by contractual
input intensity, consider an industry of a
differentiated product in which the demand
function is, as before, x(j)= Ap(j)-E, E= 1/(1a)>0.
Now, however, the production of brand
j requires two customized inputs, headquarter
services h(j) and components m(j). These
intermediate inputs are combined via a
Cobb-Douglas production function to produce
either brand j of the differentiated
product, x(j), or another intermediate input
of type j, say yo), which is used to assemble
x(j). In the latter case xU) = y().61 I express
this production relationship as
(6) z(j)-=O[
where z is either x or y, 0 represents productivity,
which for the time being is the
same for all firms in the industry, and 7
measures contractual input intensity. The
larger 11 is the more intensive the sector is
in headquarter services (but 17 does not
vary across firms in the industry). The critical
assumption is that h has to be supplied
by the final good producer while m
61 These two possibilities may seem to be an unnecessary
complication at this stage, but they provide a unified
treatment of distinct papers in the literature, as will
become clear below.
requires the engagement of a supplier,
which can take place either inside or outside
the firm. But in either case the supplier
controls min. In this event, the
internalization decision is only about the
intermediate input m, not about h. In a
simple version of the model, there is only
labor and both h and m are produced with
a fixed amount of labor per unit output.
More generally, there can be many factors
of production (primary inputs), and h and
m may be produced with different factor
proportions. It then follows that the overall
factor intensity of an industry is jointly
determined by its contractual input intensity
and by the factor intensities of
headquarter services and components.
Using the demand function x(j)= Ap(j)and
the production function (6), we can
calculate revenue as a function of the
inputs h and m, say R[h(j), m(j)]. The
assumption is that the final good producer
bears directly the cost of headquarter services
and decides the level of h, while the
supplier, who may be working for the final
good producer or be independent, chooses
m. Great simplification is attained by
assuming that the final good producer can
obtain as many suppliers as she wants by
offering a reward structure consisting of an
upfront payment and a share of the profits
at the bargaining stage. In this event competition
among suppliers leaves them with
no rents, and a supplier's total net income
(net of input cost) equals his opportunity
cost. At the bargaining stage, the distribution
of revenue R[h(j), m(j)] depends on
the bargaining weights, which are P3 for the
final good producer and 1-fP for the intermediate
good producer, and on organizational
form, which determines every party's
outside option.
Consider outsourcing. Under this organizational
form, the outside options at the bargaining
stage are zero for both parties
because one party owns h and the other
owns m, and both inputs have been customized
for product j to a degree that they
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
Helpman: Trade, FDI, and the Organization of Firms 611
have no value outside the relationship.62 As
a result, the final good producer receives the
fraction / of the revenue while the supplier
receives the fraction 1- P.
Next consider integration. Now both h
and m belong to the final good producer,
because the supplier is her employee. But,
following Grossman and Hart (1986),
assume that if the bargaining fails and the
supplier does not cooperate, then the final
good producer cannot deploy the inputs as
effectively as she can when the supplier
cooperates. In particular, without the cooperation
of the supplier she is able to produce
only a fraction 3 of the output in (6).
Under the circumstances the outside
option of the supplier at the bargaining
stage is zero, while the outside option of the
final good producer is fraction 6a of the
revenue R[h&), m(j)].63 As a result, in the
bargaining stage the final good producer
receives a fraction /v= 8a+ /3(1- 6a) of the
revenue R[h(), m(j)], and the supplier
receives a fraction 1- /3.
An important trade-off in the choice of
organizational form by the final good producer
is derived from a comparison of the
optimal distribution shares of revenue,
R[h(), m()], with the shares that arise under
outsourcing and integration. Let P* be the
final good producer's most preferred share,
which maximizes her profits. First note that
it cannot be zero, because if it were zero she
would have no incentive to provide headquarter
services, and in the absence of h,
62 The outside options need not be zero in this case.
For example, the final good producer may have the option
of using a generic intermediate input m instead of the specialized
variety, in which case her outside option will not
be zero. Similarly, the intermediate good producer may
have the option of selling the input m(j) to another firm,
which will provide it with a positive outside option.
Grossman and Helpman (2002) provide an analysis in
which the suitability of a component m to a final good producer
is measured by the distance between the supplier
and the final good producer in technology space, and suppliers
have the option to choose the location of their intermediate
inputs in this space. When bargaining breaks
down, final good producers and component suppliers
enter a secondary market. The secondary market equilibrevenue
equals zero. Second, note that
cannot be one, because if it were one the
supplier would have no incentive to provide
components, and in the absence of m, revenue
would equal zero. Evidently, P3* is
strictly positive and strictly smaller than one.
Moreover, it can be shown that P3* is an
increasing function of the intensity of headquarter
services, as measured by ij. The
shape of the relationship between P* and 7
is depicted in figure 5. /3P* equals zero when
17 = 0, it equals one when 77 = 1, and it rises
in between. Moreover, it is concave for low
values of 77 and convex for high values.
The figure also shows the distribution of
revenue shares under outsourcing and integration,
3 and P, respectively; they are
above the optimal share P* when an industry
is component-intensive, so that 77 is small
(such as i1M), and they are below P* when an
industry is headquarter-intensive, so that 77 is
large (such as 77H). The arrows show the
direction of rising profits; that is, profits rise
when the final good producer's share shifts
vertically toward 3*. This characterization
implies that there exists a cutoff fcl-not
drawn in the figure-with r7M < 71c < 7H, such
that the final good producer has higher profits
from outsourcing when 77 is below r17 and
higher profits from integration when 71 is
above iL. It follows that, based on the power
of incentives consideration alone, final good
producers prefer outsourcing in component
intensive industries and integration in headquarter
intensive industries. However, the
rium then determines the outside options in the primary
market. In the Grossman and Helpman technology space
a generic input is defined as the input that is equal-distant
from all final goods. Feenstra and Spencer (2005) also
develop a model of contracting in the presence of generic
inputs, and they use it to analyze the organization of
Chinese suppliers. The model discussed in the text focuses
on the simpler case of zero outside options.
63 The finding that the outside option of the final good
producer under integration is the fraction 3o of revenue
rather than the fraction 3 stems from the concavity of the
revenue function in the quantity sold. That is, revenue as a
function of the quantity x is proportional to xa, where
(recall) a determines the demand elasticity E = 1/(1- a)
and 0 < a < 1.
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
612 Journal of Economic Literature, Vol. XLIV (September 2006)
Figure 5. Optimal Bargaining Share
final verdict on whether to outsource or integrate
does not depend on these considerations
alone if there also exist cost differences
in running firms with different organizational
forms.64
Antrhs (2003) uses a variant of this model
in which there are no organization-specific
costs; there are fixed entry costs, but these
costs are independent of whether the firm
chooses to outsource or integrate. In this
event the power of incentives dominates the
integration decision. That is, given the
headquarter intensity measure 77 a firm
prefers integration if i > r1 and outsourcing
if 7" <7C. He assumes that h is capital intensive
and m is labor intensive, and that h and
m are not tradeable across borders. In this
event every final good producer has to
deploy h and m in the same country.
64 Feenstra and Hanson (2005) estimate a related
model of the firm from Chinese export-processing data. A
plant that processes imported inputs for sales to a foreign
firm can be owned either by a foreign firm or by a Chinese
entity. Similarly, imported inputs for further processing
can be owned by a foreign firm or by the processing plant.
In the latter case the inputs are controlled by the plant's
manager. The organizational form, which consists of the
ownership of the plant and the ownership of the imported
Moreover, these inputs are used to manufacture
an intermediate input y that can be
freely shipped across borders. The production
function is given by (6), where z = y.
The final good x is produced from y in the
destination country with one unit of y per
unit x. Finally, consumers spend fixed budget
shares on goods in every sector and they
have CES preferences across brands.
In a two-country, two-sector version of
this model, trade structure can be derived
from the integrated equilibrium, similarly
to Helpman and Krugman's (1985) analysis
of trade in differentiated products.
Assuming that one sector has i" above the
cutoff r7 and another sector has 7 below
this cutoff implies that firms are integrated
in the former sector and firms outsource in
inputs, is determined according to the property fights
approach, as in the model described in the text. Feenstra
and Hanson find that the prevalence of alternative organizational
forms varies across Chinese regions in accordance
with the model's prediction. This is a good case study,
because 55.6 percent of Chinese exports during the sample
period, 1997 to 2002, are of this nature (i.e., exportprocessing),
and the distribution of export-processing
exports into the four organizational forms has a nonnegligible
fraction in every regime.
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
Helpman: Trade, FDI, and the Organization of Firms 613
the latter sector. Moreover, since h is capital
intensive while m is labor intensive, the
sector with integrated firms is capital intensive
and the sector with outsourcing firms is
labor intensive. As a result, there is
intrafirm trade in the tradeable intermediate
inputs in the capital-intensive sector
and arm's-length trade in tradeable intermediate
inputs in the labor-intensive sector.
This implies a positive correlation between
capital intensity and the share of intrafirm
trade. A multicountry version of this model
also implies a positive correlation between
the share of intrafirm imports and the capital
abundance of the exporting country.
Antrais (2003) provides evidence supporting
these predictions. In U.S. data, intrafirm
imports as a fraction of total imports are
positively correlated with the capital intensity
across twenty-three manufacturing
industries, and intrafirm imports as a fraction
of total imports are positively correlated
with capital abundance across
twenty-eight exporting countries.
Antris (2005) applies a one-factor variant
of this model to product cycles. The two
countries are North and South. He assumes
that both headquarter services and final
goods can be produced only in North. In
addition to whether to integrate or outsource,
however, a final good producer has to decide
in which country to source the component m,
i.e., whether to offshore m or not. Integration
or outsourcing in North imply no trade in
components, integration in South implies
intrafirm trade, and outsourcing in South
implies arm's-length trade in components.
Contracts are complete in North but incomplete
in South. That is, the two countries differ
in the degree of contract incompleteness.
The main result is that there exist two cutoff
values of the contractual input intensity
measure i, mc and i,, > r which determine
the desired organizational form. When
headquarter intensity is above the upper
threshold lq,, final good producers source m
in North (the model is silent on whether
they outsource or integrate there, because
contracts are complete in North). For values
between 17, and r,, final good producers
invest in subsidiaries in South and source m
from their affiliates in South. And when
headquarter intensity is below the lower
cutoff ijc final good producers outsource in
South. Interpreting i as a feature of technology
that changes over time-so that 71 is
high for a new product and it declines over
time as experience in production is
gained-these results imply a product cycle
of the Raymond Vernon (1966) type: all
parts of the value chain of a new product are
produced in North, over time the production
of components is shifted to subsidiaries
in South, and as the product matures, the
components are outsourced to Southern
manufacturers.
3.3 Contractual Input Intensity and
Productivity Heterogeneity
A combination of variation in contractual
input intensity across sectors and variation in
productivity across firms within industries
generates equilibria in which all four organizational
forms-insourcing at home, outsourcing
at home, insourcing abroad, and
outsourcing abroad-coexist in an industry
and their relative prevalence varies across
industries as a function of sectoral characteristics.
Note that these four organizational
forms do not coexist in the previous models.
Following Antrhs and Helpman (2004),
assume that the production function (6)
applies to a typical industry, but that the
productivity level 0 varies across firms. As in
Melitz (2003), an entrant into the industry
obtains a productivity draw 0 after sinking
the entry cost. After entry, and knowing her
productivity, the final good producer has to
decide on her organizational form.
There are two countries, North and South,
with the wage rate in North exceeding the
wage rate in South. Labor is the only primary
input. All final good producers are located in
North, where they also produce headquarter
services h. The intermediate inputs m can be
produced either in North or South with the
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
614 Journal of Economic Literature, Vol. XLIV (September 2006)
same labor input per unit output. This makes
the variable costs of m lower in South. But
there are different fixed costs of sourcing in
North and South, and these fixed costs also
differ for outsourcing and integration. In particular,
Antrhs and Helpman (2004) focus on
the casefs >fs >fN >fN, wheref$ is the fixed
cost of integration in South (FDI), fs is the
fixed cost of outsourcing in South, fv is the
fixed cost of integration in North, and fN is
the fixed cost of outsourcing in North, all
measured in terms of Northern labor.65
Under these circumstances outsourcing
dominates integration in componentintensive
industries, because (1) outsourcing
has lower fixed cost; and (2) for low values of
rj outsourcing provides better incentives to
suppliers of intermediate input m (see figure
5). It follows that in component-intensive
industries all firms outsource, and the only
remaining trade-off is between domestic and
foreign outsourcing. In the offshoring decision,
the trade-off is between lower variable
cost in South and lower fixed cost in North.
This trade-off is depicted in figure 6, where
ir represents profits from outsourcing in
South and Ij represents profits from outsourcing
in North as a function of the productivity
measure 0 - Oe-1. The profit line
rs is steeper because variable costs are lower
in South. Evidently, firms with productivity
below OD exit the industry, high-productivity
firms-with 0 above v-import components
from unaffiliated producers in South,
and firms with productivity between ED and
Om acquire components from unaffiliated
firms in North. That is, among the active
firms low-productivity firms outsource at
home and high-productivity firms outsource
abroad.
A similar analysis of a headquarterintensive
sector shows that all four organizational
forms can coexist. The trade-off
65 They also provide a brief discussion of the implications
of other orderings of fixed costs. See also Grossman,
Helpman, and Szeidl (2005).
between outsourcing and integration in
North is depicted in figure 7, where zCN represents
the profits of an integrated producer
and irN represents the profits of an outsourcing
producer. The profit line iN is steeper
because integration in a headquarterintensive
sector provides better incentives to
suppliers of parts (see figure 5). In this case,
low-productivity firms-with 0 below ED)
exit the industry; high-productivity firmswith
O above O0-integrate; and firms with
intermediate productivity levels outsource.
Combining this analysis with a similar analysis
of the trade-off between outsourcing and
integration in South, and accounting for the
fact that offshoring has an advantage in terms
of variable costs but a disadvantage in terms
of fixed costs, we obtain the sorting pattern
depicted in figure 8. That is, the leastproductive
firms exit the industry while the
most-productive firms use FDI to produce
intermediate inputs in South. In between, the
less-productive firms outsource in North, the
more-productive firms outsource in South,
and firms with intermediate productivity
levels integrate in North.
Three interesting results emerge from a
comparative statics analysis of this model.
First, offshoring declines with headquarter
intensity 77. Second, more productivity dispersion
leads to more offshoring; in componentintensive
sectors it leads to more outsourcing
in South while in headquarter-intensive sectors
it leads to more integration plus outsourcing
in South.66 In addition, in
headquarter intensive sectors, where there is
both outsourcing and integration, more productivity
dispersion leads to more integration
and less outsourcing. These predictions apply
to variations across industries; e.g., the model
predicts more offshoring in sectors with higher
component intensity and sectors with more
productivity dispersion. Third, an improvement
in the competitive advantage of South,
66 Productivity dispersion is measured by the shape
parameter of the Pareto distribution.
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
Helpman: Trade, FDI, and the Organization of Firms 615
Figure 6. Importing and Nonimporting Firms in a Component-Intensive Sector
be it as a result of declining relative wages in
South or declining protection in North, raises
offshoring in all sectors; and in headquarterintensive
sectors, outsourcing of components
from foreign suppliers rises proportionately
more than purchases of intermediate inputs
from foreign affiliates.67
3.4 Matching and Thick Market Effects
In the previous models, final good producers
could attract suitable suppliers at
will as long as the suppliers could gain from
the relationship with a final good producer
as much as they expected to gain from
67 Grossman, Helpman, and Szeidl (2005) use a variant
of this model to examine the complementarity between
outsourcing and offshoring. In their model, the fraction of
offshoring firms is larger the smaller the fixed cost of outsourcing.
This is the sense in which offshoring is complementary
to outsourcing; as the fixed cost of outsourcing
changes, it generates a positive correlation between the
fraction of firms that outsource and the fraction of firms
that offshore.
alternative activities. In other words, there
was an infinitely elastic supply of component
producers. This, of course, is not
entirely realistic, because matching
between buyers and sellers is a complex
process that involves risks on both sides. In
particular, the quality of a match with a supplier
that a final good producer can expect
when she outsources depends on the number
of potential suppliers in the market and
on their expertise, i.e., whether their knowledge
and experience are suitable for the
manufacturing of the type of intermediate
inputs required by her brand.
One simple approach, which places
matching between buyers and sellers of
intermediate inputs at the heart of the
analysis, has been developed by John
McLaren (2000) and Grossman and
Helpman (2002). In this approach, potential
buyers of an intermediate input find it
more attractive to outsource the "thicker"
the market for the input is, in the sense
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
616 Journal of Economic Literature, Vol. XLIV (September 2006)
Figure 7. Insourcing and Outsourcing Firms in North in a Headquarter-Intensive Sector
that there exist more sellers to serve the
buyers' needs. And similarly, sellers of an
intermediate input find it more attractive
to operate the larger the number of potential
buyers is. Although there can be more
than one reason for this type of market
externality, both papers use an endogenous
probability of successful matching between
buyers and sellers as the main driving force
of this process. In this type of environment,
international trade (or "globalization,"
using McLaren's terminology) affects
the trade-off between outsourcing and
integration. In particular, in the presence
of economies of scale to matching, trade
encourages outsourcing.
Since Grossman and Helpman's (2002)
analysis is closer in form to what we have
seen in previous sections, and they also show
how to deal with these issues in general
equilibrium, I will use their framework to
illustrate this approach. To this end, consider
an industry supplying a differentiated
product, in which the demand for varietyj is,
as before, x(j)=Ap(j)-e, where e=1/(1a)
> 1 (i.e., 0 < a < 1). Now assume that, in
order to produce brandj, the manufacturer
of the final good needs to acquire an input
that is highly specific to this brand. As
before, assume that the input has to be
tailor-made for brandj, and once it has been
tailor-made for j it cannot be used for any
other brand, nor can it be put to any other
use. For simplicity, assume that one unit of
the intermediate input is needed per unit of
final good and that no other inputs are
required.68
First consider a closed economy in which
the producer of brand j has two organizational
options: she can produce the intermediate
good inhouse or outsource. If she
68 This is a special case of the production function (6),
with q = 0. It implies that all else equal, the final good producer
would like to give the supplier the most powerful
incentives possible. As noted in footnote 62, Grossman and
Helpman (2002) also develop a richer analysis in which the
quality of a match depends on the distance between the
final good producer and the supplier in technology space.
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
Helpman: Trade, FDI, and the Organization of Firms 617
outsource
in North
insource
in North
outsource
in South
insource
in South
(FDI)
Figure 8. Sorting Pattern in a Headquarter-Intensive Sector
produces inhouse, she needs 1/0>1 units
of labor for every unit of the tailor-made
intermediate input, where 0 is a measure
of productivity, common to all firms. In
addition, she has to bear a fixed labor costfy,
which includes her entry cost (the entry cost
covers the acquisition of the technology, the
cost of setting up shop, and the like). After
entry, her optimal pricing strategy generates
a profit level ry(A) which is increasing in
the demand level A. It follows that integration
is viable in a free entry equilibrium if
and only if the demand level equals Av, at
which the integrated firm breaks even; that
is, Irv(Av) =0. Obviously, the demand level
cannot be higher than Av, because this
would induce entry of additional integrated
final good producers, and if it were lower
than Av no final good producer would
choose to integrate.
Next consider a final good producer who
chooses to outsource. For this she needs to be
matched with a supplier of the intermediate
input, because inputs with her specialized
needs are not readily available in the market.
It is assumed that once she is matched
with a supplier, they cannot sign a contract
for the delivery of the brand-specific intermediate
input. In this event there exists a
holdup problem; the supplier can choose
how much of the input to produce, but then
he has to bargain with the final good producer
for payment. A specialized supplier of
inputs can produce them with one unit of
labor per unit output, which gives him a cost
advantage over the integrated firm (which
needs 1/0> 1 units of labor per unit output).
In the ensuing Nash bargaining, both parties
have zero outside options and the final
good producer gets a fraction P3 of the surplus.
Evidently, the distribution of the bargaining
power between the two parties
affects payoffs. Using these payoffs it is then
possible to calculate the expected profits of
a final good entrant who plans to outsource
and the expected profits of an intermediate
good producer. These expected profits
depend on the probabilities of being
exit
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
618 Journal of Economic Literature, Vol. XLIV (September 2006)
matched and on entry costs, in addition to
the payoffs at the bargaining stage.
Let M(N, M) be the matching function,
which describes the number of matches
that take place in a market with N outsourcing
final good producers and M producers
of intermediate inputs. This
function is increasing in both arguments
and M(N, M) < min{N, M}. Then the probability
of a final good producer being
matched is M(N, M)/N and the probability
of an intermediate good producer being
matched is i(N, M)/M. Once a supplier
and a final good producer have been
matched, the supplier manufactures an
intermediate input tailored to the specific
needs of the final good producer. In this
model, all suppliers are equally capable of
manufacturing every such input. I discuss
asymmetries in the qualifications of suppliers
below.
When the matching function exhibits
constant returns to scale M(N, M)/N= g(1,
M/N) and g(N, M)/M = u(N/M, 1). We can
then use the entry costs of final and intermediate
good producers to formulate two
free entry conditions: the expected profits
(before entry) of an outsourcing final good
producer equal her entry cost as an outsourcing
enterprise, and the expected profits
(before entry) of an intermediate good
producer equal his entry cost. These
expected profits are functions of the
demand level A and the ratio of entrants
M/N, that is, iCN(A, M/N) and iM(A, M/N).
Both expected profits are rising in A, but
the final good producer's profits rN(A, M/N)
are rising in M/N while the intermediate
good producer's profits iM(A, M/N) are
declining in M/N. Hence, there is complementarity
between entry of intermediate
good producers and entry of outsourcingoriented
final good producers. Other things
equal, an increase in M raises the expected
profits of final good producers while an
increase in N raises the expected profits of
intermediate good producers. It follows
that more entry of one type encourages
more entry of the other type.
Viability of outsourcing in the resulting
equilibrium requires zero expected profits
for both final and intermediate good producers;
that is, rN(A, M/N) =0 and rM(A,
M/N) = 0. The two free entry conditions are
satisfied for unique values of the demand
level and the ratio of entrants, say Ao and
ro= Mo/No.
Grossman and Helpman (2002) show that
an equilibrium with integrated firms only
always exists, but that it is not stable unless
Av< Ao. Namely, stability requires the
demand level that ensures zero profits of
integrated firms to be lower than the
demand level that ensures zero profits of
outsourcing final good producers and their
suppliers of parts. The reason that an equilibrium
with integrated firms only always
exists is that, in the absence of suppliers of
intermediate inputs, the final good producers'
optimal strategy is to enter as integrated
manufacturers; and in the absence of outsourcing
final good producers the optimal
strategy of intermediate good producers is
not to enter. This is one consequence of the
above-discussed entry complementarity.
And Grossman and Helpman also show that
there is no mixed equilibrium in which
some final good producers insource while
others outsource.69 Finally, they show that a
unique stable equilibrium exists, in which
final good producers integrate when Av< Ao
and outsource when Av> Ao.70 It follows
that structural features determine whether
integration or outsourcing prevails.
The analysis so far has focused on the equilibrium
organizational form, which does not
depend on the size of the economy. Together
with a resource constraint, our equilibrium
conditions determine the number of
69 An exception is a special case in which the parameters
of the economy are such that Av= Ao.
70 Recall the definitions of the demand levels Av an
Ao; at Av an integrated firm just breaks even, while at Ao
and a ratio M/N = ro of entrants an outsourcing firm just
breaks even.
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
Helpman: Trade, FDI, and the Organization of Firms 619
entrants. The main implication is that the
number of entrants rises proportionately
with the size of the economy. Namely, if,
say, the labor force doubles, so does the
number of entrants in an equilibrium with
integrated firms only, as well as in an outsourcing
equilibrium. Under these circumstances,
the opening of free trade between
two countries that differ only in size does
not change the equilibrium organizational
form and the number of entrants in the
world economy. This is a direct consequence
of the assumption that the matching
function exhibits constant returns to scale,
because with this sort of matching technology
the probabilities of finding suppliers or
buyers of intermediate inputs do not
depend on the number of entrants, only on
their ratio N/M, and therefore the critical
demand levels Av and Ao do not depend on
market size.
In the absence of constant returns to scale
in matching the probabilities of a match,
i(N, M)/N and u(N, M)/M, depend not only
on the ratio of entrants M/N but also on the
absolute number of entrants. As a result,
there is feedback from country size to organizational
form. When M(N, M) exhibits
increasing returns to scale two stable equilibria
can coexist: one with integration, the
other with outsourcing. Moreover, in this
case outsourcing is more likely the larger the
country is, because larger market size raises
the probability of a match for both buyers
and sellers of intermediate inputs for every
ratio of entrants N/M. This implies that
opening trade between two countries that
differ only in size makes outsourcing more
likely. In particular, it is possible to have a
situation in which every country in isolation
is too small to support an outsourcing equilibrium,
yet by opening to trade, the world
economy sustains an outsourcing equilibrium
(see also McLaren 2000). If the outsourcing
equilibrium is unique, it implies
that trade changes the organization of production
from integration to outsourcing.
More generally, increasing returns to
matching imply that market integration
encourages outsourcing through the thickmarket
effect.71
One drawback of this approach is that, in
an outsourcing equilibrium, international
trade in intermediate inputs results from the
random matches of buyers and sellers from
different countries. Although the volume of
trade in intermediate inputs is well determined
in both directions, it is not related to
a deliberate effort of final good producers in
one country to seek out suppliers of parts in
a different country. In other words, in this
model offshoring is not a strategic choice of
business firms. The approach described in
the next section makes explicit the offshoring
decision and introduces a role for
different degrees of contract incompleteness.
It also introduces natural asymmetries
into the matching of final good producers
and suppliers of intermediate inputs.
Grossman and Helpman (2003, 2005)
develop a different variant of organizational
choice under incomplete contracts, in
which technological proximity between
final good producers and suppliers of intermediate
inputs plays a key role. In this
model firms choose in which country to
search for an outsourcing partner, and
countries may differ in their degrees of
contract incompleteness. These modifications
introduce separate roles for variations
across countries in market thickness, legal
systems, and other institutional features, as
71 The above described model is special in many ways.
It clarifies, however, the role of market thickness in the
link between trade and the organization of production.
One of its stark implications is that all firms choose the
same organizational form. To avoid this outcome, one can
introduce heterogeneity. Thus, for example, one could
divide explicitly the fixed costs into entry and operating
fixed costs. By paying the fixed entry cost a final good producer
would find out its productivity 0, drawn from a
known distribution, as in Melitz (2003). After that the final
good producer would decide whether to outsource, integrate,
or leave the industry. Under these circumstances
outsourcing could coexist with integration, whereby lowproductivity
firms outsource while high-productivity firms
integrate.
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
620 Journal of Economic Literature, Vol. XLIV (September 2006)
determinants of the sourcing strategies of
business firms.
To understand the basic mechanism of
outsourcing in Grossman and Helpman (2003,
2005), consider a simplified version of a closed
economy in which integration is not an option.
An industry supplies a differentiated product
with an isoelastic demand function for every
brand x(j) = Ap(j)-e, e= 1/(1- a) > 1. A unit
of x(j) is produced with one unit of a tailormade
intermediate input that has no other
uses, and it takes one unit of labor to manufacture
one unit of the intermediate input by
specialized suppliers of parts.
There are N final good producers, each
one specializing in a different brand, and M
producers of intermediate inputs. Unlike
the previous model, however, in which N
and M were finite numbers, now M is a
finite number while N is a mass. In this formulation
each supplier serves many downstream
firms. The final good producers are
all located on the circumference of a circle
of length one. This circumference represents
a technology space; a point in this
space represents the expertise of an intermediate
good producer or the expertise
needed by a final good producer for her
intermediate input. The finite number of
intermediate good producers is symmetrically
spaced at distance 1/M from each
other, while the mass N of final good producers
is uniformly distributed with density
N at each point on the circumference. I will
shortly discuss how these firms found themselves
spaced in this way. For now, take
these locations as given.72
A final good producer cannot manufacture
her product without outsourcing its tailormade
input to a supplier. The cost of
72 The circumference of a circle is used in many applications
as a space for bilateral matching. In international
trade it has been used, for example, by Helpman (1981) for
matching buyers and sellers in the presence of monopolistic
competition, and by James E. Rauch and Vitor
Trindade (2003) for matching buyers and sellers in the
presence of informational frictions.
manufacturing an intermediate input has
two parts: a variable cost of one unit of labor
per unit output plus a fixed cost of customization
to the special needs of the final
good producer. The cost of customization is
proportional to the distance d in technology
space between the seller and buyer of the
input, say wvd, where w is the wage rate and
v is a cost parameter. That is, it is more costly
to customize the input when the seller and
buyer are far away from each other than
when they are close to each other. Under the
circumstances every final good producer
chooses to source her input from the closest
supplier, with the distance d varying
between zero and 1/2M.
It is assumed that the investment in customization
has to be made by the producer
of the intermediate input, and that this
investment is not contractible.73 Moreover,
once a final good producer and an intermediate
input supplier form a relationship, the
final good producer is bound to acquire her
input from this partner.74
After the investment in customization,
the two parties sign an order contract,
which stipulates the production of intermediates,
assembly of final goods, and the distribution
of profits from sales. At this stage
both parties seek to maximize joint profits.75
This generates a profit ro that is distributed
according to the Nash bargaining
weights, which are taken to be 1/2 for each
party.76 The profit r, determines the incentive
of the intermediate good producer to
customize the input. If zr/2 0 wvd, the
73 I will introduce partial contractibility shortly.
74 Here again other options are possible, such as the
existence of a secondary market or the use of generic
inputs.
75 Grossman and Helpman (2005) assume that at this
stage, after the customization costs are sunk, there is no
further agency problem, i.e., the only agency problem
exists at the customization stage. Naturally, one could
introduce at this stage too an agency problem in which
there is a role for incomplete contracts, similarly to the formulation
in sections 3.1 and 3.2.
76 It is not difficult to allow more general weights /and
1- /.
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
Helpman: Trade, FDI, and the Organization of Firms 621
intermediate good producer is willing to
customize the product, otherwise he is not,
because the expected payoff does not cover
investment costs.77
In view of these considerations, we can
calculate the aggregate postentry profits of
an intermediate input producer as the sum
(integral) of profits across all final good producers
who purchase from him intermediate
inputs, call it fIM. And we can calculate the
post-entry profits of a final good producer
who finds herself at distance d from the
nearest producer of intermediate inputs, call
it HN(d). Then the expected value of HN(d),
where the expectation is taken over the distance
d, determines the expected pre-entry
profits of a final good producer, call it IN. To
calculate this expectation, assume that when
a final good producer enters the industry she
is equally likely to be located at any point on
the circumference of the circle. And indeed,
ex post, the final good producers are uniformly
distributed in this technology space.
As for the intermediate good producers, they
can each choose their location in the technology
space. But in the Nash equilibrium of
the entry game they choose equal distances
from each other. Moreover, entry of intermediate
input producers proceeds until the
expected profits HlM equal the entry cost wfM,
and entry of final good producers proceeds
until the expected profits I'N equal the entry
cost wfN. These conditions together with the
resource constraint then determine the
equilibrium number of entrants, M and N.
Note that in this model too there is complementarity
between entry of the final
good and the intermediate input producers;
the more entry there is of one type the
more profitable is entry of the other type.
This is the thick market effect. And the
77 It follows that if 1/M > 7Co/2wv, then there exist final
good producers who cannot find suppliers for their specialized
intermediate inputs, and they exit the industry.
This is similar to the presence of an exit cutoff in the models
discussed in section 2. The discussion below proceeds
under the assumption that this is indeed the case.
more suppliers of parts enter the industry
the smaller the average distance between
buyers and sellers of parts is.
To introduce trade, Grossman and
Helpman (2005) consider a two-country
(North and South) world, in which final good
producers enter only in North and intermediate
input producers enter in both the
North and the South. As in the closed economy
described above, final good producers
have to outsource intermediate inputs. But
now they have to pay a fee for finding the
location of input suppliers in the technology
space, and this fee is separate for each country.
Therefore, when the search costs for
component suppliers are large enough, a
final good producer searches in one country
only, either in North or in South. This generates
segmentation of input markets across
countries, and introduces a deliberate decision
of where to search for a supplier. This
decision involves two considerations, in
addition to search costs. First, wages differ
across countries, making it attractive to
search in the low-wage South, where higher
profits can be shared. Second, the number
of suppliers of parts differs across countries,
making it attractive to search in the country
with a larger number of suppliers, where the
probability of finding a good match is higher.
It follows that if search costs are the same in
North and South, the outsourcing of intermediate
inputs in both countries can take
place only if the number of suppliers is
smaller in South.
Grossman and Helpman (2005) characterize
a general equilibrium of a trading world
of this type and analyze its determinants.
They find multiple equilibria. The positive
feedback that produces multiple equilibria is
the following. As more input suppliers enter
a particular country, the country becomes
more attractive to final good producers
searching for suppliers of parts, because the
suppliers are more closely packed there in
technology space, making it more likely for a
final good producer to find a supplier who
will undertake the requisite investment in
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
622 Journal of Economic Literature, Vol. XLIV (September 2006)
customization. Moreover, the larger the
number of final good producers searching in
a country, the more attractive it is for intermediate
input suppliers to set up shop there.
For this reason there can be one equilibrium
with intermediate inputs produced in both
countries, and another equilibrium with
intermediate inputs produced only in
North.78
Focusing on an equilibrium with suppliers
of parts in both countries, Grossman and
Helpman (2005) derive comparative statics
results. An increase in the size of South raises
the number of suppliers of intermediate
inputs in South and lowers their number in
North; the volume of outsourcing rises in
South and declines in North; the volume of
trade rises relative to income; and the wage
rises in South relative to North. That is,
unlike in a neoclassical world, here growth in
labor of one type does not reduce its relative
factor reward. In addition, uniform improvements
across countries in the customization
technology have no effect on the numbers of
input suppliers, the volumes of outsourcing,
the relative wage, or the volume of trade relative
to income. But improvements in customization
that are biased toward South
increase the entry of parts suppliers in
South, reduce their entry in North, and shift
outsourcing from North to South. Moreover,
such improvements in technology raise the
relative wage of South and the volume of
trade relative to income.
One may argue that computer-aided manufacturing
and computer-aided design have
reduced the cost of customization. If so,
then this analysis suggests that the observed
patterns of outsourcing and trade expansion
cannot be explained by this technological
improvement alone, unless we have reason
78 In a two-country world, the positive feedback is limited
by a relative wage response, which stems from the fact
that expanding economic activity in a country raises the
demand for its labor, which raises in turn its wage relative
to the wage in the other country. This general equilibrium
effect limits to some extent the concentration of economic
activity in one country only, despite the presence of a thick
market effect.
to believe that it has been particularly effective
in reducing customization costs in
South.
To discuss the impact of different degrees
of contract incompleteness, Grossman and
Helpman (2005) extend the model at the
customization stage. Instead of assuming
that the investment in customization is not
contractible, they assume that a fraction of
this investment is contractible, and that the
supplier of an intermediate input and its
potential buyer negotiate an investment contract,
which specifies an upfront payment for
the contractible part of the investment. As a
result, there exists a range of distances d in
which customization did not take place
before (in the absence of contractibility), but
takes place now, and this range is larger the
larger the fraction of contractible investment
is. It follows that contractibility enlarges the
set of active matches.
This generalization has a number of implications.
First, starting with no contractibility,
the introduction of a positive fraction of contractible
customization costs in North
increases the number of suppliers of parts in
North, reduces their number in South, and
raises the relative wage of North. As a result,
the volume of outsourcing rises in North,
declines in South, and trade relative to
income shrinks. Second, an improvement in
contracting institutions in South, which raises
the contractible fraction of customization
costs there, may not expand outsourcing in
South. When a significant fraction of customization
costs are contractible in North
but not in South, initial improvements in
contractibility in South raise outsourcing in
both South and North, raise South's relative
wage, and raise the trade volume relative to
income. But once the fraction of contractible
costs crosses a threshold, further improvements
in contractibility in South reduce outsourcing
there, further raise outsourcing in
North, reduce the relative wage of South,
and diminish the ratio of trade to income. In
other words, the response to better contracting
institutions in South is not monotonic,
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
Helpman: Trade, FDI, and the Organization of Firms 623
and it depends on how far the South's institutions
lag behind those of North. This nonmonotonicity
has no simple intuitive explanation;
it results from a complex interaction between
direct effects of changes in the degree of contractibility
and indirect effects that operate
through labor and product markets in the
general equilibrium.
The analysis has so far dealt with outsourcing,
where the choice is between the
acquisition of intermediate inputs at home
(in North) or abroad (in South). Grossman
and Helpman (2003) discuss a variant of the
same model in which a final good producer
can either outsource or integrate, but this
trade-off is analyzed at the expense of abandoning
the endogeneity of wages and the
trade-off between locating the activity in
North or South. In particular, they assume
that the production of intermediates takes
place in South, so that intermediates are offshored,
and a firm has to decide only
whether to produce its intermediates in a
subsidiary or acquire them at arm's-length
from an unaffiliated supplier. And they
assume constant wages in every country.
The trade-off is the following. As in
Grossman and Helpman (2002), an integrated
firm has a cost disadvantage in producing
intermediates. Therefore, while a specialized
supplier of parts needs only one unit of
labor per unit of intermediate input, a final
good producer needs 1/0>1 units of labor
per unit of intermediate input, where 0 is
common to all firms. But, the final good producer
has a cost advantage in customization;
his customization costs are zero while a specialized
supplier of parts bears customization
costs wvd, which are (as before)
proportional to the distance in technology
space between him and the producer of the
final good. As a result, a final good producer
who chooses integration makes profits Hv,
which can be calculated in the usual way.
A final good producer who chooses outsourcing
seeks out the closest supplier of
parts in technology space, and negotiates with
him an investment contract (to be followed by
an order contract after customization takes
place). The largest distance d that makes
such a relationship viable depends on the
degree of contract incompleteness: the larger
the contractible fraction of the investment
in customization, the larger this
distance. If follows that the profits of an outsourcing
final good producer depend on
how far she is from the closest supplier of
parts, HfN(d).
Figure 9 depicts the profits I, and Us(d)
as functions of the distance d. Naturally, HI
is flat, because it does not depend on this
distance. But FIN(d) is flat up to ds, and
declines gradually after a downward drop at
ds, where ds is defined as the largest distance
at which the supplier has an incentive
to customize the input without an investment
contract. The flat part results from the
fact that up to distance ds the supplier's
payoff from the order contract, which is
independent of the distance d, exceeds the
customization costs, in which case no investment
contract is signed and the supplier of
parts invests in customization nevertheless.79
Just slightly above ds, however, the
customization costs wvd exceed the supplier's
payoff from the order contract, in which
case the supplier of parts does not invest in
customization unless an investment contract
is signed, and the equilibrium investment
contract allocates the customization costs
equally between the supplier and the buyer
of intermediate inputs. The larger the distance
between the two parties the larger the
contribution of the final good producer to
the customization costs and the smaller her
profits.
Under these circumstances there exists a
critical distance do, which satisfies
FIN(do) = l-, such that all final good producers
with d < do prefer to outsource and
all final good producers with d > do prefer
to integrate (or exit if FIN(d) <0). Since d is
79 Let Ps be the supplier's payoff from the order contract,
the same for all d. Then at ds this payoff just equals
the customization costs wvds, i.e., wvds= Ps.
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
624 Journal of Economic Literature, Vol. XLIV (September 2006)
Figure 9. Choice of Integration Versus Outsourcing Depends on Technological Distance from
Input Suppliers
random before entry, we can use the uniform
distribution of location on the circumference
of the circle together with the number of
intermediate input producers to calculate the
expected profits of a final good producer who
enters the industry. Entry proceeds until
these expected profits, net of entry cost,
equal zero. We can similarly calculate the
free entry condition for intermediate good
producers.
Grossman and Helpman (2003) find that
outsourcing is more prevalent in larger markets,
and that the thick market effect is
responsible for the positive correlation
between market size and the fraction of outsourcing
firms and their market share. They
also find that better contracting institutions
in South, which render larger fractions of
the customization costs contractible,
increase the prevalence of outsourcing.
Analyzing the trade-off between outsoureing
at home or abroad and the trade-off
between outsourcing or integration, provides
useful insights but it gives only a partial
view of organizational choices. A complete
analysis requires simultaneously allowing
firms to choose between outsourcing and
integration in every country, thereby admitting
an interaction between the offshoring
decision and the internalization decision. No
such analysis exists for the class of models
discussed in this section, only for the model
discussed in section 3.3.
3.5 Ricardian Comparative Advantage
Differences across countries in legal systems
and institutions that shape the enforcement
of contracts, and thereby the degree of
contract incompleteness, have the potential
for influencing patterns of comparative costs
across countries. Ricardian comparative
advantage, as reflected in the cross sectoral
variation in productivity levels, can arise as a
result of institutional variation across countries
when the relative requirement of
contract-dependent inputs varies across sectors,
because institutions impact costs in
sectors with a larger need for contractThis
content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
Helpman: Trade, FDI, and the Organization of Firms 625
dependent inputs relatively more than in
sectors with less need for contractdependent
inputs.
Nunn (forthcoming) provides a detailed
empirical analysis of the impact of the
degree of contract incompleteness on international
trade flows.s0 As the main representative
of the degree of contract
incompleteness, he uses a measure of the
rule of law, which consists of a weighted
average of a number of variables that gauge
the effectiveness of the judiciary, its predictability,
and its enforcement of contracts.81
He finds that the results do not
change much when this variable is replaced
with other, more objective measures of the
efficacy of courts.82 To compute an index of
contract dependence for every final good
sector, Nunn uses U.S. input-output tables
to compute the proportion of intermediate
inputs used in every final good, and he classifies
intermediates into those that are traded
on an organized exchange, those that
have a reference price, and those that have
none of these.83 He assumes that a good is
more contract dependent the larger is the
fraction of its intermediate inputs that have
no organized exchange nor a reference
price, or alternatively, the larger is the fraction
of its intermediate inputs that have no
organized exchange only. The main empirical
finding is that countries with better legal
systems export relatively more in sectors
that are more intensive in contract-dependent
inputs. This finding is robust to controls
80 Levehenko (2004), who preceded Nunn (forthcoming),
makes related arguments. I focus on Nunn (forthcoming),
however, because he provides the sharper
empirical analysis. Both Levehenko and Nunn develop
simple theoretical models to guide their empirical work.
81 These variables are estimated from subjective perceptions;
see Daniel Kaufmann, Aart Kraay, and Massimo
Mastruzzi (2003).
82 These objective measures are from Simeon Djankov,
Rafael La Porta, Florencio Lopez-de-Silanes, and Andrei
Shleifer (2003) and they are available for a smaller sample
of countries.
83 The classification of inputs into those that have an
organized exchange, those that have a reference price, and
those that have no organized exchange and no reference
price is from Rauch (1999).
of other determinants of trade flows, alternative
specifications of the estimated equation,
and alternative estimation methods.
Moreover, not only has the quality of the
legal system a statistically significant impact
on trade flows, it also has a large economic
impact. In particular, its impact, as measured
by the beta coefficient, is of similar
magnitude to that of human capital and
physical capital combined. In other words,
contracting institutions are an important
source of comparative advantage.
Daron Acemoglu, Antris, and Helpman
(2006) propose a model in which Ricardian
comparative advantage emerges from the
interaction of contract incompleteness with
the deliberate choice of technology by final
good producers. In their model, a final good
producer can choose how to divide the production
process, so as to have many or few
intermediate inputs. Every intermediate
input has to engage a supplier. The supplier
of the input, who can be a worker in the firm
or an outside supplier, has to execute a set of
activities in order to produce it. A subset of
these activities are contractible, while the
others are not. The fraction of noncontractible
activities provides a measure of
contract incompleteness.
On the one hand, more sophisticated
technologies-that allow more intermediate
inputs in the production process-are more
costly to acquire, and they may involve larger
organizational costs. On the other hand,
more sophisticated technologies are more
productive. Using this trade-off, a final good
producer makes an optimal technology
choice, and this choice depends on features
of the industry and the degree of contract
incompleteness. Acemoglu, Antrhs, and
Helpman (2006) find that better contracting
institutions lead to the choice of more
sophisticated technologies, and that the
impact of contracting institutions on technology
choice is relatively larger in sectors
with lower elasticities of substitution across
intermediate inputs, because low substitutability
makes the sector more sensitive to
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
626 Journal of Economic Literature, Vol. XLIV (September 2006)
contractual frictions. As a result, countries
with better contracting institutions have a
relative productivity advantage, and therefore
comparative advantage, in sectors with
less substitutable inputs.
Arnaud Costinot (2005) proposes a different
model in which contracting institutions
interact with technological features to create
Ricardian comparative advantage. In his
model, every industry is characterized by a
set of tasks that have to be performed and
these sets are exogenous. Industries are
ordered by the complexity of their technology,
which is measured by the number of
tasks in their set. Workers are assigned to
tasks. A worker has to spend a fixed amount
of time to learn a particular task. As a result,
there are increasing returns to scale in the
performance of tasks. But a worker can shirk
and not perform his task. In the event of
shirking, no output is produced because
every task is essential. The degree of contract
incompleteness is measured by the
probability that a worker shirks, which is
exogenous and independent across workers.
When a team of workers produces a product,
it is efficient to assign every worker the
same number of tasks. Given the size of the
team, it is then possible to compute expected
output per worker. The resulting optimal team
size, which maximizes output per worker, is
larger in more complex industries and in
countries with better legal institutions, in
which contracts are enforced with higher
probabilities. In a competitive economy, better
institutions raise output per worker proportionately
more in more complex industries.
As a result, a country with better contractenforcing
institutions gains a comparative
advantage in more complex industries.
4. Concluding Remarks
New developments in the world economy
have called for new developments in the theory
of international trade and foreign directed
investment, designed to better
understand the shifts in trade and investment
patterns and the reorganization of production
across national boarders. Although traditional
trade theory has much to offer in
explaining parts of this puzzle, such as the
international fragmentation of production,
the theory had to be generalized in order to
provide a better understanding of the trends
in the data.84 Particularly acute has been the
need to model different forms and degrees of
involvement of business firms in foreign
activities because organizational change has
been a central element in the transformation
of the world economy. As a result, theoretical
refinements have focused on the individual
firm, studying its choices in response to its
own characteristics, the nature of the industry
in which it operates, and the opportunities
afforded by foreign trade and
investment. Important among these choices
are modes of serving foreign markets and
sourcing strategies.
But the theory went beyond the individual
firm, studying the implications of firm
behavior for the structure of an industry and,
by implication, structural differences across
industries. These variations deliver new
explanations for trade structure and patterns
of FDI, both within and across industries.
For example, they identify new sources of
comparative advantage, such as the degree
of heterogeneity within industries and the
quality of contracting institutions.
Heterogeneity plays a key role in this literature
in two ways. First, there is heterogeneity
as a result of productivity differences
across firms wi4thin industries, because some
firms happen to be luckier than others.s85
Second, there is heterogeneity in organizational
form. The two are related, however,
because differences in productivity induce
different choices for the organization of production
and distribution. In this theory,
trade and FDI patterns are jointly deter-
84 See Ronald W. Jones (2000).
85 The empirical literature supports the view that
causality goes from productivity to, say, exports, rather
than the other way around; see, for example, Bernard and
Jensen (1999) and Clerides, Lach, and Tybout (1998).
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
Helpman: Trade, FDI, and the Organization of Firms 627
mined with organizational structures, such
as sourcing and integration strategies.
Some implications of these models have
been tested empirically. Examples include
the sorting patterns of firms into exporters
and foreign direct investors. Other implications
have not been tested. These include
patterns of sorting into outsourcing at home,
integration at home, outsourcing abroad,
and integration abroad, because this cannot
be done with the available data sets. More
generally, hypotheses that require detailed
firm-level data about trade in different types
of products, such as intermediate inputs versus
final goods, and whether this trade takes
place within the boundary of the firm or at
arm's-length, cannot be examined. The theoretical
models point out, however, what
additional data need to be collected in order
to improve the empirical analysis. Since this
is still a lively area of research, we can expect
to see much more theoretical and empirical
work on these topics.
REFERENCES
Abraham, Katharine G., and Susan K. Taylor. 1996.
"Firms' Use of Outside Contractors: Theory and
Evidence." Journal of Labor Economics, 14(3):
394-424.
Acemoglu, Daron, Pol Antras, and Elhanan Helpman.
2006. "Contracts and Technology Adoption."
Harvard University. Mimeo
Amiti, Mary, and Shang-Jin Wei. 2005. "Fear of Service
Outsourcing: Is It Justified?" Economic Policy, 42:
307-39.
Anderson, James E., and Eric van Wincoop. 2004.
"Trade Costs." Journal of Economic Literature,
42(3): 691-751.
Antrkis, Pol. 2003. "Firms, Contracts, and Trade
Structure." Quarterly Journal of Economics, 118(4):
1375-1418.
Antr~is, Pol. 2005. "Incomplete Contracts and the
Product Cycle." American Economic Review, 95(4):
1054-73.
Antrks, Pol, and Elhanan Helpman. 2004. "Global
Sourcing." Journal of Political Economy, 112(3):
552-80.
Aw, Bee Yan, Sukkyun Chung, and Mark J. Roberts.
2000. "Productivity and Turnover in the Export
Market: Micro-Level Evidence from the Republic of
Korea and Taiwan (China)." World Bank Economic
Review, 14(1): 65-90.
Axtell, Robert L. 2001. "Zipf Distribution of U.S. Firm
Sizes." Science, 293(5536): 1818-20.
Baldwin, John R., and Wulong Gu. 2003. "ExportMarket
Participation and Productivity Performance
in Canadian Manufacturing." Canadian Journal of
Economics, 36(3): 634-57.
Baldwin, Richard E., 1988. "Hysteresis in Import
Prices: The Beachhead Effect." American Economic
Review, 78(4): 773-85.
Baldwin, Richard E., 2005. "Heterogeneous Firms and
Trade: Testable and Untestable Properties of the
Melitz Model." NBER Working Papers, no. 11471.
Baldwin, Richard E., and Paul R. Krugman. 1989.
"Persistent Trade Effects of Large Exchange Rate
Shocks." Quarterly Journal of Economics, 104(4):
635-54.
Bamford, James (1994), "Driving America to Tiers,"
Financial World, November 8, pp. 24-27.
Bardi, Edward J., and Michael Tracey. 1991.
"Transportation Outsourcing: A Survey of U.S.
Practices." International Journal of Physical
Distribution and Logistics Management, 21(3): 15-21.
Bartel, Ann P., Saul Lach, and Nachum Sicherman.
2005. "Outsourcing and Technological Change."
NBER Working Papers, no. 11158.
Bernard, Andrew B., Jonathan Eaton, J. Bradford
Jensen, and Samuel Kortum. 2003. "Plants and
Productivity in International Trade." American
Economic Review, 93(4): 1268-90.
Bernard, Andrew B., and J. Bradford Jensen. 1999.
"Exceptional Exporter Performance: Cause, Effect,
or Both?" Journal of International Economics,
47(1): 1-25.
Bernard, Andrew B., and J. Bradford Jensen. 2004.
"Why Some Firms Export." Review of Economics
and Statistics, 86(2): 561-69.
Bernard, Andrew B., J. Bradford Jensen, and Peter K.
Schott. 2005. "Importers, Exporters, and
Multinationals: A Portrait of Firms in the U.S. That
Trade Goods." NBER Working Papers, no. 11404.
Bernard, Andrew B., Stephen Redding, and Peter K.
Schott. Forthcoming. "Comparative Advantage and
Heterogeneous Firms." Review of Economic Studies.
Bhagwati, Jagdish, Arvind Panagariya, and T. N.
Srinivasan. 2004. "The Muddles over Outsourcing."
Journal of Economic Perspectives, 18(4): 93-114.
Bolton, Patrick, and Mathias Dewatripont. 2005.
Contract Theory. Cambridge and London: MIT Press.
Borga, Maria, and William J. Zeile. 2004. "International
Fragmentation of Production and the Intrafirm
Trade of U.S. Multinational Companies." Bureau of
Economic Analysis Working Paper WP2004-02.
Bustos, Paula. 2005. "Rising Wage Inequality in the
Argentinean Manufacturing Sector: The Impact of
Trade and Foreign Direct Investment on Technology
and Skill Upgrading." Harvard University. Mimeo.
Campa, Jose, and Linda S. Goldberg. 1997. "The
Evolving External Orientation of Manufacturing: A
Profile of Four Countries." Federal Reserve Bank of
New York Economic Policy Review, 3(2): 53-81.
Campbell, Jeffrey R., and Hugo A. Hopenhayn. 2005.
"Market Size Matters." Journal of Industrial
Economics, 53(1): 1-25.
Clerides, Sofronis K., Saul Lach, and James R. Tybout.
1998. "Is Learning by Exporting Important? MicroDynamic
Evidence from Colombia, Mexico, and
Morocco." Quarterly Journal of Economics, 113(3):
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
628 Journal of Economic Literature, Vol. XLIV (September 2006)
903-47.
Costinot, Arnaud. 2005. "Contract Enforcement,
Division of Labor, and the Pattern of Trade."
Department of Economics, Princeton University.
Mimeo.
Das, Sanghamitra, Mark J. Roberts, and James R.
Tybout. 2001. "Market Entry Costs, Producer
Heterogeneity, and Export Dynamics." NBER
Working Papers, no. 8629.
Davis, Donald R., and David E. Weinstein. 2001. "An
Account of Global Factor Trade." American
Economic Review, 91(5): 1423-53.
Delgado, Miguel A., Jose C. Farifias, and Sonia Ruano.
2002. "Firm Productivity and Export Markets: A
Non-Parametric Approach." Journal of International
Economics, 57(2): 397-422.
Dixit, Avinash K. 1989. "Hysteresis, Import
Penetration, and Exchange Rate Pass-Through."
Quarterly Journal of Economics, 104(2): 205-28.
Djankov, Simeon, Rafael La Porta, Florencio Lopezde-Silanes,
and Andrei Shleifer. 2003. "Courts."
Quarterly Journal of Economics, 118(2): 453-517.
Eaton, Jonathan, Samuel Kortum, and Francis
Kramarz. 2004. "Dissecting Trade: Firms, Industries,
and Export Destinations." American Economic
Review, 94(2): 150-54.
Ekholm, Karolina, Rikard Forslid, and James R.
Markusen. 2004. "Export-Platform Foreign Direct
Investment." University of Colorado. Mimeo.
Fan, Joseph P. H., and Larry H. P. Lang. 2000. "The
Measurement of Relatedness: An Application to
Corporate Diversification." Journal of Business,
73(4): 629-60.
Feenstra, Robert C. 1998. "Integration of Trade and
Disintegration of Production in the Global Economy."
Journal of Economic Perspectives, 12(4): 31-50.
Feenstra, Robert C. 2004. Advanced International
Trade: Theory and Evidence. Princeton and Oxford:
Princeton University Press.
Feenstra, Robert C., and Gordon H. Hanson. 1996.
"Globalization, Outsourcing, and Wage Inequality."
American Economic Review, 86(2): 240-45.
Feenstra, Robert C., and Gordon H. Hanson. 2005.
"Ownership and Control in Outsourcing to China:
Estimating the Property-Rights Theory of the Firm."
Quarterly Journal of Economics, 120(2): 729-61.
Feenstra, Robert C., and Barbara J. Spencer. 2005.
"Contractual versus Generic Outsourcing: The Role
of Proximity." University of California, Davis. Mimeo.
Feinberg, Susan E., and Michael P. Keane. 2003.
"Accounting for the Growth of MNC-Based Trade
Using a Structural Model of U.S. MNCs." Yale
University. Mimeo.
Feinberg, Susan E., and Michael P. Keane. 2005. "Tariff
Effects on MNC Decisions to Engage in Intra-firm
and Arms-Length Trade." Yale University. Mimeo.
Gardner, Elizabeth (1991), "Going on Line with
Outsiders," Modern Healthcare 21 pp. 35-47.
Girma, Sourafel, Holger Gtrg, and Eric Strobl. 2004.
"Exports, International Investment, and Plant
Performance: Evidence from a Non-Parametric
Test." Economics Letters, 83(3): 317-24.
Girma, Sourafel, Richard Kneller, and Mauro Pisu.
2005. "Exports versus FDI: An Empirical Test."
Review of World Economics/Weltwirtschaftliches
Archiv, 141(2): 193-218.
Grossman, Gene M., and Elhanan Helpman. 2002.
"Integration versus Outsourcing in Industry
Equilibrium." Quarterly Journal of Economics,
117(1): 85-120.
Grossman, Gene M., and Elhanan Helpman. 2003.
"Outsourcing versus FDI in Industry Equilibrium."
Journal of the European Economic Association,
1(2-3): 317-27.
Grossman, Gene M., and Elhanan Helpman. 2004.
"Managerial Incentives and the International
Organization of Production." Journal of International
Economics, 63(2): 237-62.
Grossman, Gene M., and Elhanan Helpman. 2005.
"Outsourcing in a Global Economy." Review of
Economic Studies, 72(1): 135-59.
Grossman, Gene M., Elhanan Helpman, and Adam
Szeidl. 2005a. "Complementarities between
Outsourcing and Foreign Sourcing." American
Economic Review, 95(2): 19-24.
Grossman, Gene M., Elhanan Helpman, and Adam
Szeidl. Forthcoming. "Optimal Integration
Strategies for the Multinational Firm." Journal of
International Economics.
Grossman, Sanford J., and Oliver D. Hart. 1986. "The
Costs and Benefits of Ownership: A Theory of
Vertical and Lateral Integration." Journal of Political
Economy, 94(4): 691-719.
Hanson, Gordon H., Raymond J. Mataloni Jr., and
Matthew J. Slaughter. 2001. "Expansion Strategies of
U.S. Multinational Firms." In Brookings Trade
Forum: 2001, ed. Susan M. Collins and Dani Rodrik.
Washington, D.C.: Brookings Institution, 245-94.
Hanson, Gordon H., Raymond J. Mataloni Jr., and
Matthew J. Slaughter. 2005. "Vertical Production
Networks in Multinational Firms." Review of
Economics and Statistics, 87(4): 664-78.
Hart, Oliver D., 1995. Firms, Contracts, and Financial
Structure. Clarendon Lectures in Economics.
Oxford and New York: Oxford University Press,
Clarendon Press.
Hart, Oliver D., and John Moore. 1990. "Property
Rights and the Nature of the Firm." Journal of
Political Economy, 98(6): 1119-58.
Head, Keith, and John Ries. 2003. "Heterogeneity and
the FDI versus Export Decision of Japanese
Manufacturers." Journal of the Japanese and
International Economies, 17(4): 448-67.
Helper, Susan. 1991. "Strategy and Irreversibility in
Supplier Relations: The Case of the U.S. Automobile
Industry." Business History Review, 65(4): 781-824.
Helpman, Elhanan. 1981. "International Trade in the
Presence of Product Differentiation, Economies of
Scale and Monopolistic Competition: A
Chamberlin-Heckscher-Ohlin Approach." Journal
of International Economics, 11(3): 305-40.
Helpman, Elhanan, and Paul R. Krugman. 1985.
Market Structure and Foreign Trade: Increasing
Returns, Imperfect Competition, and the
International Economy. Cambridge: MIT Press.
Helpman, Elhanan, Marc J. Melitz, and Yona
Rubinstein. 2006. "Trading Partners and Trading
Volumes." Harvard University. Mimeo.
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
Helpman: Trade, FDI, and the Organization of Firms 629
Helpman, Elhanan, Marc J. Melitz, and Stephen Ross
Yeaple. 2004. "Export versus FDI with
Heterogeneous Firms." American Economic Review,
94(1): 300-316.
Holmes, Thomas J. 1999. "Localization of Industry and
Vertical Disintegration." Review of Economics and
Statistics, 81(2): 314-25.
Holmstrom, Bengt, and Paul Milgrom. 1994. "The
Firm as an Incentive System." American Economic
Review, 84(4): 972-91.
Hopenhayn, Hugo A. 1992. "Entry, Exit, and Firm
Dynamics in Long Run Equilibrium." Econometrica,
60(5): 1127-50.
Hubbard, Thomas N. 2001. "Contractual Form and
Market Thickness in Trucking." RAND Journal of
Economics, 32(2): 369-86.
Hummels, David, Jun Ishii, and Kei-Mu Yi. 2001. "The
Nature and Growth of Vertical Specialization in
World Trade." Journal of International Economics,
54(1): 75-96.
Jean, S6bastien. 2002. "International Trade and Firms'
Heterogeneity under Monopolistic Competition."
Open Economies Review, 13(3): 291-311.
Jones, Ronald W. 2000. Globalization and the Theory
of Input Trade. Ohlin Lectures, Vol. 8. Cambridge
and London: MIT Press.
Kaufmann, Daniel, Aart Kraay, and Massimo Mastruzzi.
2003. "Government Matters III: Governance
Indicators for 1996-2002." World Bank. Mimeo.
Landefeld, J. Steven, and Raymond J. Mataloni Jr.
"Offshore Outsourcing and Multinational Companies,"
presentation at the Brookings Institution,
June 22, 2004.
Levchenko, Andrei A. 2004. "Institutional Quality and
International Trade." IMF Working Paper 04/231.
Marin, Dalia, and Thierry Verdier. 2005. "Corporate
Hierarchies and International Trade: Theory and
Evidence." University of Munich. Mimeo.
McLaren, John. 2000. "'Globalization' And Vertical
Structure." American Economic Review, 90(5):
1239-54.
Melitz, Marc J. 2003. "The Impact of Trade on IntraIndustry
Reallocations and Aggregate Industry
Productivity." Econometrica, 71(6): 1695-1725.
Melitz, Marc J., and Gianmarco I. P. Ottaviano. 2005.
"Market Size, Trade, and Productivity." NBER
Working Papers, no. 11393.
Montagna, Catia. 2001. "Efficiency Gaps, Love of
Variety and International Trade." Economica,
68(269): 27-44.
Neary, J. Peter. 2003. "Presidential Address:
Globalization and Market Structure." Journal of the
European Economic Association, 1(2-3): 245-71.
Nunn, Nathan. Forthcoming. "RelationshipSpecificity,
Incomplete Contracts, and the Pattern of
Trade." Quarterly Journal of Economics.
Ottaviano, Gianmarco I. P., Takatoshi Tabuchi, and
Jacques-Frangois Thisse. 2002. "Agglomeration and
Trade Revisited." International Economic Review,
43(2): 409-35.
Puga, Diego, and Daniel Trefler. 2002. "Knowledge
Creation and Control in Organziations." NBER
Working Papers, no. 9121.
Qin, Larry D., and Barbara J. Spencer. 2002. "Keiretsu
and Relationship-Specific Investment: Implications
for Market-Opening Trade Policy." Journal of
International Economics, 58(1): 49-79.
Rauch, James E. 1999. "Networks versus Markets in
International Trade." Journal of International
Economics, 48(1): 7-35.
Rauch, James E., and Vitor Trindade. 2003.
"Information, International Substitutability, and
Globalization." American Economic Review, 93(3):
775-91.
Roberts, Mark J., and James R. Tybout. 1997. "The
Decision to Export in Colombia: An Empirical
Model of Entry with Sunk Costs." American
Economic Review, 87(4): 545-64.
Spencer, Barbara J. 2005. "International Outsourcing
and Incomplete Contracts." Canadian Journal of
Economics, 38(4): 1107-35.
Spencer, Barbara J., and Larry D. Qiu. 2001. "Keiretsu
and Relationship-Specific Investment: A Barrier to
Trade?" International Economic Review, 42(4):
871-901.
Strauss-Kahn, Vanessa. 2003. "The Role of
Globalization in the Within-Industry Shift Away
from Unskilled Workers in France." NBER Working
Papers, no. 9716.
Syverson, Chad. 2004. "Product Substitutability and
Productivity Dispersion." Review of Economics and
Statistics, 86(2): 534-50.
Syverson, Chad. 2005. "Prices, Spatial Competition,
and Heterogeneous Producers: An Empirical Test."
University of Chicago. Mimeo.
Tempest, Rone. "Barbie and the World Economy," Los
Angeles Times, September 22, 1996, Al and A12.
The Economist (1991), "The Ins and Outs of
Outsourcing," August 31, pp. 54-56.
Trefler, Daniel. 1995. "The Case of the Missing Trade
and Other Mysteries." American Economic Review,
85(5): 1029-46.
Trefler, Daniel. 2004. "The Long and Short of the
Canada-U.S. Free Trade Agreement." American
Economic Review, 94(4): 870-95.
Tybout, James R., and M. Daniel Westbrook. 1995.
"Trade Liberalization and the Dimensions of
Efficiency Change in Mexican Manufacturing
Industries." Journal of International Economics,
39(1-2): 53-78.
United Nations Conference on Trade and
Development. 1998. World Investment Repo
Trends and Determinants. New York and Geneva:
United Nations Publications.
United Nations Conference on Trade and Development.
2002. World Investment Report: Transnational
Corporations and Export Competitiveness. New York
and Geneva: United Nations Publications.
United Nations Conference on Trade and Development.
2004. World Investment Report: The Shift towards
Services. New York and Geneva: United Nations
Publications.
Vernon, Raymond. 1966. "International Investment
and International Trade in the Product Cycle."
Quarterly Journal of Economics, 80(2): 190-207.
World Trade Organization. 1998. Annual Report 1998.
Geneva: World Trade Organization.
World Trade Organization. 2004. Annual Report 2004.
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms
630 Journal of Economic Literature, Vol. XLIV (September 2006)
Geneva: World Trade Organization.
Yeaple, Stephen Ross. 2003. "The Complex Integration
Strategies of Multinationals and Cross Country
Dependencies in the Structure of Foreign Direct
Investment." Journal of International Economics,
60(2): 293-314.
Yeaple, Stephen Ross. 2005. "A Simple Model of Firm
Heterogeneity, International Trade, and Wages."
Journal of International Economics, 65(1): 1-20.
Yeats, Alexander J. 2001. "Just How Big Is Global
Production Sharing?" In Fragmentation: New
Production Patterns in the World Economy, ed. Sven
W. Arndt and Henryk Kierzkowski. Oxford and New
York: Oxford University Press.
This content downloaded from 147.251.185.127 on Wed, 21 Jun 2017 09:54:23 UTC
All use subject to http://about.jstor.org/terms