International Economic Agreements and the Activities of
Heterogeneous Multinational Firms⇤
Leonardo Baccini†
Pablo M. Pinto‡
Stephen Weymouth§
October 31, 2014
Abstract
The proliferation of international institutions defines the current wave of globalization. In this
paper we explore how international economic agreements influence the operations of the most
salient actors in global trade: multinational corporations (MNCs). Building on the insights
of the New New Trade Theory, we claim that preferential trade agreements (PTAs) and the
World Trade Organization (WTO) increase firm supply chain activities through the reduction
of trade costs. We argue that the largest, most productive firms are the principal beneficiaries
of the global shift toward preferential trade agreements. Using firm-level data covering the
near universe of U.S. multinationals and preferential tari↵ cuts at HS 6-digit level, we find that
the e↵ects of international economic institutions are sizable and skewed towards large firms.
Further, we show that the increase in supply chain activities is driven by lower tari↵s granted
by the U.S. to its trade partners and vice versa. We also find that employment concentration
among U.S. affiliates in host countries increases 12 percent on average after the formation of
a PTA with the US. Our evidence indicates that PTAs between the U.S. and host countries
increase vertical sales to the U.S. among the largest firms, whereas joining the WTO has no
discernable e↵ect on their vertical activities. Our paper sheds light on the stalled Doha Round
of the WTO and global shift toward preferential liberalization.
Keywords: global production networks, preferential trade agreements, WTO, heterogeneous
firms.
⇤ We
thank Chinchih Chen for excellent research assistance and Michael Bechtel, Andrew Kerner,
Ian Osgood, Erica Owen, Rachel Wellhausen, Oliver Westerwinter, Bill Zeile, Ka Zeng, seminar
participants at University of Saint Gallen, panel participants at 2014 MPSA, 2014 APSA, and 2014
ECPR for useful comments on previous drafts of this paper. The statistical analysis of firm-level
data on U.S. multinational companies was conducted at the Bureau of Economic Analysis, U.S. Department of Commerce, under arrangements that maintain legal confidentiality requirements. The
views expressed are those of the authors and do not reflect official positions of the U.S. Department
of Commerce. All errors are our own.
† London
School of Economics & Political Science; [email protected]
‡ University
of Houston; [email protected]
§ Georgetown
University; [email protected]
1
Introduction
The origins and consequences of international institutions are subjects of important scholarship
and debate (Abbott and Snidal, 1998; Keohane, 1984). One stream of research examines the e↵ects
of institutions regulating international trade (Rose, 2004; Tomz et al., 2007; Büthe and Milner,
2008). Another stream argues that international economic institutions do not benefit all countries
equally (Gowa and Kim, 2005; Goldstein et al., 2007). In this paper we explore the distributional
e↵ects of international economic agreements at the level of individual firms. Drawing on literature
examining the role of firms in international trade (Melitz, 2003; Bernard et al., 2007), we argue
that preferential trade agreements (PTAs) and the World Trade Organization (WTO) increase
firm supply chain activities through the reduction of trade costs. In explaining the distributional
implications of trade liberalization, our main contribution is to demonstrate through theory and
evidence that firm size matters: the biggest multinationals emerge as the clear winners from the
shift toward preferential agreements.
Our argument addresses two unprecedented and overarching transformations in the international trading system over the past two decades. One such transformation is the globalization
of production (Antras, 2010). Firms increasingly source from foreign countries (i.e. o↵shore) and
have strategically placed di↵erent stages of their supply chains in di↵erent countries to reap cost advantages that emerge from di↵erences in factor endowments or institutions (Feenstra and Hanson,
2005; Yeaple, 2006; Nunn, 2007). The formation of a global supply chain often involves the establishment of subsidiaries abroad for production of intermediates traded intra-firm with the parent
company (vertical FDI) and trade with participants in third countries (export-platform FDI). Firm
participation in the “trade-investment-service nexus” (Baldwin, 2012) has important implications
for the study of international political economy, which has tended to examine trade and investment
in isolation.
The second transformation involves the governance of global commerce, which has shifted
from multilateral institutions to (new) regional trade governance. This change is underscored by
the long-standing deadlock of the Doha Round of the WTO, and by the dramatic increase in
preferential trade agreements (PTAs). The ongoing difficulties in reaching a deal in the Doha
Round have attracted increasing attention from social scientists (Collier, 2006; Odell, 2009); and
1
similarly there is no shortage of explanations for the proliferation of PTAs (Baccini and Dür, 2012;
Mansfield and Milner, 2012; Baccini and Urpelainen, 2015). A related body of literature directly
links the formation of PTAs to increased foreign direct investment (Chase, 2003; Manger, 2009;
Büthe and Milner, 2008, 2014).
We depart from existing approaches examining aggregate data to study the e↵ects of international economic institutions on individual firms’ global supply chain activities. Figure 1 shows the
dramatic increase of vertical FDI and export-platform FDI by U.S. firms over the past two decades.
We see that the share of sales to third countries (export-platform) and the share of sales to the U.S.
(vertical) increased between 1989 and 2009, while the share of sales to host countries (horizontal)
declined. What role do international economic institutions play in explaining the growth in the
global supply chain activities of U.S. firms?
Figure 1
We argue that liberalizing institutions promote global supply chains principally through
the cost-cutting e↵ects of tari↵ reduction. Building on the insights of the New New Trade Theory (Melitz, 2003; Helpman et al., 2004), we posit that by lowering trade costs, PTAs and the
WTO promote supply chain activities di↵erentially across firms within industries, with the benefits
accruing to the largest and most productive firms. Furthermore, we distinguish between preferential (e.g. PTA) and multilateral (e.g. WTO) liberalizations. The key di↵erence between the two
types of liberalization hinges on the discriminatory versus non-discriminatory nature of trade costs
reduction, a↵ecting the decision of MNCs to expand operations in one host market over another.
Our theoretical framework leads to three clear predictions. First, given that preferential
liberalization allows for the concession of discriminatory tari↵ cuts to a select number of trade
partners, we expect that PTAs increase vertical FDI activities in PTA partner countries by lowering the costs of selling back to the home market. Yet, in line with Melitz (2003), the e↵ect of
preferential liberalization will vary across firms: only the largest, most productive MNCs benefit
2
from the cost advantage generated by the tari↵ cuts.1 Second, given that the MFN principle allows
only non-discriminatory tari↵ cuts, we expect that preferential liberalization has a larger e↵ect on
vertical FDI than does multilateral liberalization, since WTO member host countries do not gain
market access advantages over other WTO members. Third, we expect that both preferential and
multilateral liberalization will be strongly associated with more MNC trade with third countries,
since export-platform FDI involves complex networks in which goods cross borders several times.
We examine the empirical implications of our argument using confidential firm-level data
covering the near universe of U.S. FDI as well as preferential tari↵ data at HS 6-digit product
level, and our main results are as follows. We find that a PTA between the host country and the
U.S. increases affiliate exports back to the U.S. market, whereas joining the WTO has little to
no e↵ect on vertical FDI activities. Host country participation in the WTO, which both lowers
costs of importing and exporting goods, is associated with higher affiliate sales to third countries.
Based on our analysis of tari↵ data, our findings suggest that tari↵ reduction is the key mechanism
explaining both the increase in vertical FDI after a PTA, and the increase in export-platform FDI
after preferential and multilateral trade liberalization. Importantly, the e↵ects of both preferential
and multilateral trade liberalization scale with firm productivity, suggesting a reallocation of sales
from the least to the most competitive MNCs. Examining the China case, our results indicate that
global supply chain activities in that country boomed following its accession to the WTO. Finally,
to provide a concrete sense of the redistributive e↵ect associated with preferential liberalization,
we examine changes in the concentration of U.S. affiliates. We find that employment concentration
increases 12 percent on average in PTA partner countries post-PTA.
Our theory and empirical findings make several contributions to the study of international
institutions and global governance. Preferential liberalization – more than multilateral liberalization
– benefits global supply chain activities of the most productive firms, increasing their market power
at the expense of the least productive firms. In other words, in the current era of globalization, the
non-discriminatory MFN principle, a core component of GATT/WTO, does not seem to serve the
1 Specifically,
while PTAs lower costs that would allow new firms to enter a market, the increased
competition from new firm entry pushes the less productive firms out of the market, leading to a
reallocation of affiliate sales toward the most productive MNCs.
3
interests of the most powerful and productive actors.2 Our paper also speaks to broader debates
on the e↵ectiveness on international institutions (Gray, 2013; Baccini and Urpelainen, 2015) and
helps explain why less productive firms are more likely to oppose trade agreements, while larger and
more productive firms, which appear to disproportionately reap the benefits from PTAs, usually
rally in support of preferential liberalization (Manger, 2009).
Finally, our paper extends important recent work by Büthe and Milner (2008, 2014) on the
linkages between trade agreements and foreign direct investment.3 While Büthe and Milner (2014)
show that variation in the design of trade agreements matters for FDI flows, we demonstrate that
the impact of PTAs and the WTO varies across firms within industries. In particular, we show
that the e↵ect of economic institutions on FDI operates through the reduction of trade costs. Our
results are in line with Antras and Foley (2009) on the e↵ects of ASEAN on the activity of U.S.
MNCs, and Bernard et al. (2006), who find that as trade costs fall, high-productivity firms expand
at the expense of low-productivity firms.
The remainder of the paper proceeds as follow. In the next section we present our theory
and state our main hypotheses. In the third section we describe our data and empirical strategy.
The fourth section presents the baseline results at the country-level. The fifth section reports our
main firm-level results. The sixth section presents some extensions of our empirical analysis and
additional robustness checks. A final section concludes.
2 The
inadequacy of the multilateral trading system in the current phase of globalization has
been also noted by recent studies in economics (Antras and Staiger, 2012) and in policy papers
(Baldwin, 2012). However, to the best of our knowledge, we are the first to o↵er theory and
systematic empirical analysis of the panoply of international economic agreements, preferential and
multilateral alike.
3 Büthe
and Milner (2014) find that PTAs increase FDI in developing countries and that such a
positive e↵ect is larger when they establish dispute settlement mechanisms.
4
2
Theory
Building on the insights from recent developments in international trade theory – dubbed in the
specialized literature as the New New Trade Theory to distinguish it from the New Trade Theory
of the 1980s – we argue that: (1) the e↵ect of preferential and multilateral trade liberalization on
the expansion of MNC activities works through trade cost reduction, which we observe in tari↵
cuts; (2) the positive e↵ect of preferential and multilateral trade liberalization on FDI scales with
firm productivity; (3) preferential and multilateral trade liberalization have heterogeneous e↵ect
on MNCs activities.
This section proceeds as follows. We first present the main insights of the NNTT. We then
argue that discriminatory, preferential liberalizations reduce costs in ways that non-discriminatory,
multilateral trade liberalization does not. Finally, we discuss the empirical implications of the e↵ect
of PTAs (i.e. discriminatory trade liberalization) as well as the WTO (i.e. non-discriminatory trade
liberalization) on di↵erent types of MNCs activities.
2.1
Heterogeneous Firms and Trade Liberalization
Melitz (2003) is the seminal theoretical contribution to NNTT. His theory is motivated by the
observation that the neoclassical trade theory cannot account for the empirical regularity that
exporting firms are di↵erent from other firms operating in the same sector of the economy (Bernard
and Jensen, 1999). The Melitz Model (henceforth MM, 2003) captures this regularity by adding
heterogeneity in firm productivity into a Krugman (1979) type monopolistic competition framework.
The key features of MM (2003) are three-fold. First, each industry is populated by a continuum of
firms di↵erentiated by the varieties they produce and by di↵erences in their productivities. Second,
when firms make a costly investment decision to enter the domestic market, they face uncertainties
about their future productivity and profits. Third, firms face fixed production costs; such costs
imply increasing returns of scale.
To explore firm behavior in an open economy, MM (2003) adds costs associated with trade
(on trade costs see Anderson and Wincoop (2004)) to firms’ production function. Firms observe
their productivity before deciding whether to export or not: a firm will export if and only if export
5
profits are larger than the fixed costs of exporting. Because of the presence of export costs, the
exporting productivity threshold is higher than the threshold of doing business at home. In the
open economy equilibrium, only a subset of more productive firms will export. Helpman et al. (2004)
extends the MM (2003) model to allow for MNCs activities abroad. Setting up and managing a
foreign affiliate forces firms to incur additional costs. Helpman et al. (2004) show that MNCs should
be the most productive firms–even more productive than exporting firms, a proposition that finds
empirical support (Eaton et al., 2011).
Let’s check how the relevant properties of MM (2003) equilibrium change when trade liberalization occurs. First, falling trade costs lower the productivity threshold for export. This results
in entry of new firms into foreign markets, i.e. an increase in the extensive margin. Second, lower
trade costs increase competition in the liberalizing markets. The drop in price of foreign goods due
to lower tari↵s forces domestic firms to lose a share of their markets at the expense of the more
productive foreign firms, i.e. increase of intensive margin for firms that remain active. Third, since
highly productive firms export more and new firms can enter the domestic market, demand for labor (and other inputs) increases in the market that liberalizes. In turn, real wages and production
costs go up. The combination of decreasing profits and rising costs raises the domestic productivity
threshold, forcing less productive firms to exit the market. The punch line result here is that with
trade liberalization, market shares and profits increase for the most productive firms, including
firms already operating in the economy and new entrants, who are likely to be productive enough
to benefit from the lower costs and who can endure greater competition. The least productive
firms, on the other hand, see their market shares and profits shrink and, eventually, are forced out
of the market altogether.
2.2
Preferential versus Multilateral Trade Liberalization
International economic agreements generate two types of trade liberalization. To access the WTO,
countries implement multilateral trade liberalization; whereas the formation of a PTA produces
6
preferential trade liberalization. Both types of liberalization reduce trade costs through tari↵
reduction.4
To illustrate the magnitude of these cuts, we can compare the tari↵s o↵ered by the US to its
trading partners. Although 80% of today’s trade occurs under MFN tari↵s, for countries like Chile
and Mexico preferential tari↵s account for a large share of their imports (Baldwin, 2012). Figure 2
shows box plots of tari↵ reductions implemented by the US in all the preferential trade agreements
signed since 1990. In particular, tari↵ cuts capture the di↵erence between MFN tari↵s (prior the
formation of PTAs) and preferential tari↵s (PRF) in the first year in which the agreement is in
force. Data come from WITS and are disaggregated at HS-6 digit level. While US tari↵ cuts are
quite similar among trade partners, there is variation across the 20 PTAs.
Figure 2
Moreover, Figure 3 shows the average US tari↵ cut by type of products. The di↵erences in
tari↵ cuts are particularly stark when comparing intermediates versus final and mixed goods.5 The
figure shows that tari↵ cuts on intermediates are larger than tari↵ cuts on final and mixed goods.
Similarly, Figure A.1 in the appendix shows the average US tari↵ cut for low contract intensive
products versus high contract intensive products.6 Contract-intensive products are characterized by
4 PTAs
and the WTO include many trade-related provisions that increase market competition,
e.g. protecting of investment and intellectual property rights (IPRs). Such regulatory measures
are potentially as important as tari↵ reductions (Baldwin, 2012; Dür et al., 2014; Baccini and
Urpelainen, 2015). Table A.1 in the appendix shows the design of all the US PTAs. US PTAs
share a very similar template, including a large number of these additional trade-related provisions(Baccini and Urpelainen, 2015). Similarly, every country accessing the WTO has to sign onto
three agreements: TRIMS, TRIPS, and GATS, which regulate respectively investment, IPRs, and
services.
5 Data
on the type of good come from Francois and Pindyuk (2012) and Bekkers et al. (2012).
Goods are categorized in “intermediate consumption”, “final consumption”, and “mixed use”.
6 Data
come from Nunn (2007) and “high” refers to these products that have a score above the
75th percentile in Nunn (2007)’s variable The contract intensity measures we use are built at the
7
a large proportion of intermediate inputs that require relationship-specific investments. Firms are
more likely to engage in vertical FDI in contract intensive activities to mitigate hold up problems. In
Figure A.1 in the appendix we observe that high contract intensive products face larger (average)
tari↵ reductions than low contract intensive products. In sum, preferential trade liberalization
appears more pronounced in products that constitute the bulk of the global supply chain, i.e.
intermediates and contract intensive goods.7
Figure 3
Although both joining the WTO and forming a PTA reduce trade costs, there is a crucial
di↵erence between preferential tari↵s and MFN tari↵s. Preferential tari↵s are usually reciprocal
– PTA members o↵er proportional tari↵ reductions to each other – and discriminatory – PTA
members do not have to lower tari↵s with countries excluded from the agreement.8 In contrast,
MFN tari↵s are both reciprocal and non-discriminatory, i.e. any tari↵ cuts granted to a WTO
member have to be extended to the other members. For instance, after the US-Chile FTA (signed
in 2003) both Chile and the US cut tari↵s with each other without o↵ering similar concessions to
the remaining WTO members. However, when Croatia entered into the WTO in 2000, the US MFN
tari↵s with Croatia were set at the same level as US MFN tari↵s with the other WTO members.
Similarly, Croatia set MFN tari↵s with the US at the same level than MFN tari↵s with the other
WTO members. We argue that this distinction between discriminatory and non-discriminatory
tari↵s has important implications for firms’ supply chain activities and in particular for vertical
FDI.
3-digit level, whereas tari↵ cut are at the 6-digit level. We use the average tari↵ cut to move from
6-digit to 3-digit.
7 Figure
A.2 and A.3 in the appendix shows that a similar pattern holds for the WTO. There
is heterogeneity in MFN tari↵ cuts implemented by the US with new WTO entrants and cuts are
larger in intermediates compared to finished goods.
8 Article
XXIV of the GATT/WTO allows such a discriminatory trade policy also among
GATT/WTO members. There are also other forms on non-reciprocal tari↵ reductions under the
GATT/WTO regime, such as the Generalized System of Preferences, but they are not included in
our analysis.
8
The argument we have been developed so far can be distilled as follows. First, both PTAs
and the WTO have an impact on the activities of MNCs, but only the most productive and larger
firms reap the benefits from trade liberalization. Second, for both PTAs and the WTO such an
impact materializes through trade costs reduction e↵ected by tari↵ cuts. However, the type of trade
costs reduction, i.e. discriminatory versus non-discriminatory, have di↵erent e↵ects on the type of
MNC activities – vertical versus export-platform activities. Below we elaborate on this argument,
distinguishing between types of MNCs activities as well as types of liberalization. We also state
our testable hypotheses.
2.3
PTAs and FDI activity
In line with Antras and Foley (2009), we posit that the e↵ect of preferential liberalization depends
on the type of MNCs activities. The sharpest contrast is found in the comparison between vertical
and export platform sales. The key di↵erence between these two economic activities is that vertical
FDI is mostly a↵ected by tari↵ cuts implemented by the home country. On the contrary, exportplatform FDI is a↵ected by tari↵ reductions implemented by both the host country as well as by
tari↵ reduction implemented by third countries. In other words, bilateral trade relations shape
vertical FDI, whereas export-platform activities respond to trade relations among (at least) three
countries. We posit that this di↵erence drives the heterogeneous e↵ect of preferential liberalization
on MNC activities.
2.3.1
PTAs and Vertical FDI
By lowering trade costs, PTAs between the US and partner countries should have a positive e↵ect
on sales back to the US, and this e↵ect should scale with firm productivity. Indeed, because of lower
preferential tari↵s, shipping products back to the US is now cheaper than before the formation of
a PTA, relative to countries that are excluded from such a PTA. It is important to stress that
trade costs are reduced compared to other host markets without PTAs. Recall the example of the
Chile-US PTA. After this agreement was signed in 2003, it became cheaper for the US to import
products from Chile than from other countries that do not enjoy preferential tari↵s with the US.
Moreover, due to lower trade costs, competition should increase in host markets such as Chile
9
that sign into PTAs. In turn, increasing competition should raise wages and input costs; as the
productivity threshold rises, less productive domestic firms as well as less productive US affiliates
would be forced to reduce their sales and could get pushed out of the partner’s market. In line
with MM (2003) and Bernard et al. (2006) the big winners from preferential liberalization are the
most productive US affiliates that ship products back to the US.
Moreover, if PTAs reduce trade costs in the ways suggested above, we should observe
increases in vertical sales of U.S. MNC affiliates in products where the tari↵ cuts implemented by
the US become large. Indeed, preferential tari↵ cuts implemented by the US directly reduce trade
costs faced by affiliates serving the home market. For instance, for US firms, re-importing from
Chile is particularly cheap in those products where tari↵s cuts implemented by the US are sizable
as a result of the Chile–US PTA. Again, the e↵ect of tari↵ cuts on vertical sales should scale with
productivity, since only the most productive affiliates are able to pay raising wages and other input
costs, which stem from increasing competition. Increases in vertical sales scaling with tari↵ cuts
would lend support to our proposed mechanism, which hinges on trade costs reduction. In sum,
we expect that:
H1: Among the most productive U.S. MNCs, the formation of a PTA between the US
and partner countries increases affiliate sales to the home market through preferential tari↵s cuts
implemented by the US.
2.3.2
PTAs and Export-Platform FDI
The e↵ect of a US PTA on US company foreign affiliates that sell products to third countries is
less clear cut than the e↵ect of US PTAs on vertical FDI. On the one hand, there is no tari↵
reduction between the US and third countries that are not part of the PTA. Thus, trade costs do
not necessarily drop for affiliates serving third countries excluded by preferential tari↵s. On the
other hand, since a PTA between the US and country A lowers tari↵s for sales originating in the
parent company, inputs imported from the US to be assembled in country A and then sold to third
countries are now cheaper than before the formation of such a PTA relative to countries excluded
from it. For instance, for the US, exporting to Chile is particularly cheap in products where tari↵s
cuts implemented by Chile are sizable as a result of the Chile-US PTA. Such a discriminatory
10
reduction in trade costs in intermediate inputs a↵ects affiliates selling to a third country.9 In sum,
we expect that:
H2: Among the most productive U.S. MNCs, the formation of a PTA between the US and
partner countries increases affiliate sales to third markets through preferential tari↵ cuts implemented by partner countries.
2.4
WTO and FDI
By joining the WTO, countries are able to enjoy non-discriminatory MFN tari↵s. As previously
discussed, while vertical FDI activities involve bilateral trade between the host and the home
market, export-platform FDI involves more complex networks, including investment and trade
among at least three countries. Taken together, we argue that the e↵ect of non-discriminatory
trade liberalization through the WTO on vertical FDI di↵ers from the e↵ect of discriminatory
trade liberalization through PTAs. The e↵ect of multilateral liberalization on export-platform FDI
is similar to the e↵ect of preferential liberalization.
2.4.1
WTO and Vertical FDI
Entering into the WTO leads to lower MFN tari↵s, which in turn reduce trade costs. Products
exported from new WTO members to the home market become cheaper, so vertical FDI should
also increase after multilateral liberalization. Such an e↵ect is similar to the one produced by
preferential liberalization. However, the di↵erence rests on the non-discriminatory nature of MFN
tari↵s. Indeed, MFN tari↵s are set at the same level with all WTO members. For instance, when
Croatia entered into the WTO on the 30th of November 2000, the US extended to Croatia the same
MFN tari↵s that the other existing WTO members had already been enjoying. Therefore, following
Croatia’s accession, US MNC imports of products from Croatia were not cheaper than imported
products from all other WTO members, who also enjoy access to the MFN tari↵ in the US market,
9 We
note that trade-related provision included into the US-country A PTA could also positively
impact all the US affiliates that operate in country A, even these doing business with third countries
not covered by the PTA.
11
and particularly when compared with firms enjoying preferential tari↵s. In fact, for US MNCs,
re-importing products from Croatia was more expensive than re-importing products from Mexico
and Chile, since these two countries have been enjoying preferential tari↵s with the US since 1994
and 2003, respectively. In other words, given their non-discriminatory nature, the e↵ect of MFN
tari↵s on sales to the US is expected to be not as large as the e↵ect of discriminatory tari↵ cuts
produced by forming PTAs with the US. In sum we expect that:
H3: Among the most productive U.S. MNCs, host country membership in the WTO increases
affiliate sales to the home market through MFN tari↵ cuts implemented by the US.
As a corollary of this hypothesis, we expect that:
HP3a: For the same level of affiliate productivity, the positive e↵ect of non-discriminatory
MFN tari↵ cuts on affiliate sales to the home market should be smaller than the positive e↵ect of
discriminatory preferential tari↵ cuts.
2.4.2
WTO and Export-Platform FDI
MFN tari↵s should expand export-platform activities. Indeed, in joining the WTO, countries lower
their MFN tari↵s with the US. Therefore, affiliates of US MNCs face a trade cost reduction when
they export goods to the host market as a result of its lower MFN tari↵. Moreover, in entering into
the WTO countries lower MFN tari↵s with all the other WTO members. Thus, US affiliates face a
trade costs reduction when they ship goods to third markets, to the extent that these third countries
are covered by MFN tari↵s as WTO members. It is important to stress that while MFN tari↵s set
by a country such as Vietnam, which recently entered the WTO, and by third markets with all
the WTO members are necessarily the same, MFN tari↵s might vary across WTO members. That
is, for each WTO member MFN tari↵s might be higher or lower than the other WTO members
depending on each country’s bargain power at the moment of joining the WTO (Pelc, 2011). Since
goods used in export-platform FDI activities cross borders several times, non-discriminatory MFN
tari↵ cuts have a cumulative e↵ect in reducing trade costs and increasing competition. Again
building on the insights from NNTT, we hypothesize a redistribution of sales from the least to the
most productive affiliates through the labor market channel. In sum, we expect:
12
H4: Among the most productive U.S. MNCs, WTO MFN tari↵ cuts implemented by host
countries have a positive e↵ect on affiliate sales to third markets through MFN tari↵ cuts implemented by host countries.
We summarize our main argument in the following Table:
Type of activity
Vertical FDI
Export-Platform FDI
PTA x Productivity
Positive (strong) e↵ect through
discriminatory tari↵ cuts
implemented by the US (H1)
Positive e↵ect through
discriminatory tari↵ cuts
implemented by partner country (H2)
WTO x Productivity
Positive (weak) e↵ect through
non-discriminatory tari↵ cuts
implemented by the US (H3)
Positive e↵ect through
non-discriminatory tari↵ cuts
implemented by partner country (H4)
The key elements of our theory are the redistribution e↵ect produced by the formation of
international economic agreements and the mechanism through which such redistribution operates:
the reduction in trade costs. Simply put, we expect that the e↵ect of PTAs and the WTO scales
with the productivity of US affiliates and is triggered by preferential and MFN tari↵ cuts. Moreover,
we expect that US PTAs a↵ect vertical FDI more than joining the WTO, whereas both preferential
and multilateral liberalization will positively influence export-platform FDI.
3
Data
Multinational corporations are involved in complex operations (Bilir et al., 2013; Helpman et al.,
2004), driven in part by the evolution of global supply chains and the growing importance of
trade in intermediate goods (Jensen et al., 2015). Bilir et al. (2013) identify three main types of
MNCs activities: horizontal, vertical, and export-platform. Horizontal activities are those where
affiliates abroad sell products in the local market in which they are produced. Vertical activities
involve exchange between the affiliate and the headquarter firm. For the most part, transactions
associated with vertical MNC activity comprise intermediate goods, which are shipped back to the
home country to be assembled in finished products. Export-platform activities comprise sales by
foreign affiliates to third countries. Export-platform activities may comprise both intermediate and
finished products.
13
Our empirical analysis examines these diverse MNC activities using confidential firm-level
data covering the near universe of U.S. FDI collected by the U.S. Bureau of Economic Analysis
(BEA)10 as well as preferential tari↵ data at HS 6-digit product level. The BEA conducts Benchmark Surveys every five years capturing the universe of U.S. multinationals.11 Any U.S. person
with direct or indirect ownership of ten percent or more of the voting securities of a foreign business
during the benchmark fiscal year is a U.S. parent of the foreign business, which is termed its foreign
affiliate. The U.S. multinational is the combination of the U.S. legal entity that has established
or purchased the affiliate (i.e., the U.S. ‘parent’) and at least one foreign business enterprise (i.e.
the foreign ‘affiliate’). The International Investment and Trade in Services Survey Act requires
that owners of foreign affiliates detail the balance sheets, income statements and international
transactions of their affiliates. As a result of the confidentiality assurances and the penalties for
noncompliance, the coverage is considered nearly complete and the accuracy of the responses is
high. Our sample consists of all majority-owned affiliates of non-bank U.S. parent companies.
The BEA data allow us to construct a panel of firm-level data on the sales and operations
of U.S. parent firms and their foreign affiliates. In particular, we use affiliate-level data detailing
the distribution of affiliate sales to examine the e↵ect of international economic agreements on
MNC activities. Along with total affiliate sales, the survey reports sales to 1) the U.S., 2) the
host country, and to 3) other foreign countries.12 We use these data to construct our measures of
vertical, horizontal and export-platform sales, respectively.
Using these data, we are able to examine how the activities of U.S. MNCs change in response
to trade agreements in ways that address three potential shortcomings in the extant literature. First,
we are able to test a direct mechanism through which preferential and MFN liberalization operate:
tari↵ cuts. These e↵ects are usually confounded with other mechanisms related to specific provi10 In
a typical benchmark year, the survey covers over 99% of affiliate activity by total sales,
assets and U.S. FDI. For example, in 1994 participating affiliates accounted for 99.9 % of total U.S.
FDI.
11 The
BEA also conducts annual surveys on a subset of MNCs. To avoid sample selection bias,
we rely on the data from the universe of U.S. MNCs covered in the scope of the benchmark years.
12 Sales
figures broken down by destination are available for majority-owned affiliates only.
14
sions included in the design of trade agreements, e.g. provisions protecting investment or dispute
settlement mechanisms.13 Second, using US affiliates as the unit of analysis allows us to uncover
patterns in the data according to which international trade agreements reallocate activities among
firms based on their productivities. Third, we can explore how these two forms of liberalization
a↵ect di↵erent MNC activities, including vertical and export-platform FDI.
3.1
Dependent Variables
Following the literature (Bilir et al., 2013; Antras and Foley, 2009), we use data from the BEA
Benchmark Surveys to construct measures of MNC activities based on the destination of affiliate
sales.14 Given that our predictions that preferential and MFN liberalizations have di↵erent e↵ects
depending on the type of MNC global production activity, we concentrate mostly on vertical and
export platform sales by affiliates. As it is customary, in order to mitigate the impact of outliers
we take the natural logarithm of sales.15
3.2
Main Independent Variables
Since our main hypotheses are about the e↵ect of tari↵ cuts associated with MFN and preferential
liberalization on di↵erent MNC activities, we incorporate data measuring tari↵ cuts o↵ered by the
US and its partner countries under PTAs and under the WTO. We construct two variables aimed
at directly testing our main theoretical mechanism. The first variable is the di↵erence between
MFN and preferential tari↵s, as described above. We create two versions of this variable: Tari↵
13 Prior
studies have yielded interesting results by exploit the variance in the design of trade
agreements around the world. One challenge inherent in this approach is that the provisions often
do not vary dramatically across the PTA agreements of a single country, such as the U.S. Hence the
source of identifying variance is at the country level, which could confound the estimated e↵ects. By
looking at tari↵ reductions o↵ered by the US and its partners for specific sectors, we are attempting
to isolate the e↵ect of trade liberalization through proposed mechanism.
14 The
15 We
benchmark years are 1989, 1994, 1999, 2004, and 2009.
add one to the value of sales prior to the log transformation so as not to exclude affiliates
with zero sales.
15
Cut, which measures the nominal level of the tari↵ concession, and Tari↵ Cut (proportional), which
captures the proportional tari↵ reduction, i.e.
M F N P RF
.
MF N
These two variables score zero for
countries that have no PTA in force with the US.16 Both Tari↵ Cut and Tari↵ Cut (proportional)
capture liberalization implemented by the US, since we are mostly interested in estimating the
e↵ect of preferential trade liberalization on sales back to the US market; to save space, we report
results using Tari↵ Cut (proportional) only. The second variable WTO Cut captures the tari↵s
concessions implemented under WTO by the US and by the partner country. This variable is
constructed in similar fashion to Tari↵ Cut.17 Descriptive statistics of the relevant variables used
in our analyses are shown in Table 1.
To allow for comparison with earlier work, we also create a series of variables aimed at
capturing membership in the multilateral trade regime (WTO) and bilateral PTAs with the US.
The variable PTA with US is a dummy equal to one beginning with the first benchmark year after
a country signs the agreement with the U.S.; zero otherwise. In the case of Chile, for instance,
PTA with U.S. scores one in 2004 and 2009.
With country and year fixed e↵ects, the dummy PTA with U.S. is equivalent to a di↵erencein-di↵erences estimation in which the “treated units”—countries that at some point sign a PTA
with the US—score one for all years when a PTA with US is in existence. The PTA data come
from the DESTA database (Dür et al., 2014).18 In similar fashion, to assess the e↵ect of MFN
liberalization we create a dummy variable for periods after which a country joins the WTO.
Table 1
16 As
noted, data come from WITS (2014) and are disaggregated at HS 6-digit level. We create
a cross-walk to the NAICS and collapse the data to the 4-digit level to conform with the BEA
industry classifications.
17 As
with Tari↵ Cut, we create two versions: WTO Cut and WTO Cut (proportional). Re-
sults using either version of the WTO Cut variable are similar; we present those for WTO Cut
(proportional) in the tables below.
18 Data
are available at http://www.designoftradeagreements.org/.
16
We also include a number of other institutional variables in our main regressions. We control
for the presence of a BIT with the US, GATT membership and the number of PTAs to which the
partner country is a signatory during the period prior to the benchmark (Cumulative PTAs).19 We
also include the log of GDP/capita in all our our specifications. In robustness tests presented in the
Appendix and in our country-level analysis, we add a series of additional political and economic
control variables, as recommended by Büthe and Milner (2014).20
4
Baseline Analysis: Country-level Results
Our main tests of hypotheses 1-4 seek to estimate the e↵ects of the level of tari↵ cuts resulting
from MFN and preferential liberalization on vertical and export-platform sales at the firm-level.
However, previous studies have found that PTAs increase aggregate FDI (Büthe and Milner, 2008,
2014), and so we begin with an analysis at the country level using aggregated BEA data. The
purpose of the country-level analysis is to examine the relationship between PTAs (and WTO) and
MNC activity, ignoring all sectoral and firm-level sources of variation.
We begin with models of the counts of total affiliates and counts of each type of affiliates
(Horizontal, Export-Platform, and Vertical) in each country-benchmark year. We also estimate
models of total sales for each type of activity at the country-level. The models take the following
form:
Ait = ↵ P T A with U Sit +
19 For
W T Oit + Cit + &i + ⌧t + ✏it
benchmark year z, Cumulative P T Asit =
Pz
1
t=z 5 P T At .
(1)
This variable is similar to the
one used by Büthe and Milner (2008, 2014). Data come from DESTA (Dür et al., 2014). We note
that the number of PTAs included in DESTA dataset is substantively larger than the number of
PTAs included in Büthe and Milner (2008, 2014). This variable enters in di↵erent estimations in
log transformation or as dummies for each quartile of its distribution to allow for a more flexible
functional relationship.
20 To
capture economic and political conditions in the host country in years prior to and inclusive
of the benchmark year, our control variables are (5 year) lagged average values, inclusive of the
benchmark year.
17
where Ait is either the log of the number of total affiliates or the log of total affiliate sales aggregated
to the country i level in benchmark year t.21 PTA with USit and W T O capture the existence of a
PTA with the US and WTO membership respectively, Cit are the economic and political control
variables (including other international economic agreements), and &i and ⌧t are country and year
fixed e↵ects, respectively. The models are estimated using OLS and standard errors are clustered
at the country-level. The expectation from the extant literature is that PTAs should promote
investment, reflected in a positive coefficient corresponding to PTA with USit .
We report the estimates of the number of U.S. company foreign affiliates on left-side panel
of Table 2. The dependent variable in column 1 is the total number of affiliates in each countrybenchmark year, while columns 2-4 break down the number of affiliates by type of activity.22 The
results indicate that PTAs with the US are not associated with number of affiliates. PTAs with
the US are however associated with higher horizontal and vertical sales at the 90% confidence level.
The results in column 8 indicate that a PTA with the U.S. is associated with a large increase in
exports to the U.S.: on average, a PTA with the U.S. increases horizontal and vertical sales by 65%
and 170% respectively. WTO membership does not seem to be associated with MNC activity in
any of the count or the sales models; and similar results pertain to GATT. Finally, signing a BIT
with the US is positively associated with more affiliates and greater sales; yet the coefficient is only
statistically significant for total sales and horizontal sales.23
The only systematic relationship unveiled by the country level results is the positive association between the cumulative number of PTAs entered into by the partner country and aggregate
affiliate activities, including total affiliates and total affiliate sales. This result is consistent with
21 We
add one to the count and sales variable prior to taking the natural log to preserve country-
year observations with no affiliates sales in the sample.
22 The
categories are not mutually exclusive. An affiliate with positive sales in a particular
category (horizontal, vertical, and export-platform) is counted as an affiliate in that category.
23 Column
1 also indicates a positive and statistically significant relationship between economic
development and the logged number of U.S. MNC affiliates. A one-standard deviation increase in
GDP per capita (equivalent to about a 19% increase) is associated with 9% more affiliates. We
also find that trade openness is strongly positively associated with affiliate presence.
18
Büthe and Milner (2008, 2014). The substantive e↵ect is not trivial: a 10% increase in cumulative PTAs is associated with a nearly 2% increase in number of affiliates. As shown in columns 2
and 3, total PTAs are also strongly associated with the number of horizontal affiliates and with
export-platform affiliates. Yet we find that cumulative PTAs are associated with smaller increases
in the number of vertical affiliates. There is also an association between cumulative PTAs and the
aggregate sales of U.S. MNCs. The relationship holds for horizontal and export sales, but there is
no evidence that cumulative PTAs increase vertical sales back to the US.
Table 2
Together these results suggest that preferential liberalization (through PTAs) and MFN
liberalization relate to US MNC activity in di↵erent ways: contrary to expectations, the number
of PTAs signed by the host country is not systematically associated with higher FDI in all types of
activities, which might be expected if participation in these agreements were to provide reassurances
to investors. The positive and significant relationship between cumulative PTAs and sales by US
MNC affiliates to the local market is consistent with credible commitment arguments24 , while the
(lack of) e↵ect on vertical activities is not. The signing of bilateral agreements with the US (PTAs
or BITs) or participation in multilateral agreements provide less conclusive results at the countrylevel. Note, however, that the degree of aggregation of the data in these country-level analyses
are not ideal for probing our hypotheses. Instead, in the ensuing empirical analysis we will rely
on firm-level sales data disaggregated by activity, and fine-grained data on the tari↵ cuts resulting
from MFN or PTAs.
5
Main Analysis: Firm-level Results
We next examine the relationship between trade liberalization and affiliate-level sales Sajit .
Sajit = ↵P T A with U Sit + W T Oit + Cit + 'i + &j + ⌧t + ✏ajit
24 The
(2)
positive relationship between cumulative PTAs and export platform sales could be the
result credibility or of lower tari↵s abroad
19
All models include controls for host country’s GDP per capita Cit , industry- &j , country'i , and year ⌧t fixed e↵ects. The industry fixed e↵ects &j absorb omitted industry-specific determinants of affiliate activity, including technological requirements and factor-intensities, average firm
productivity, size and concentration levels, and institutional and policy preferences. The models
are estimated using OLS and standard errors are clustered at the country-level.
The ensuing sections analyze sequentially the link between international economic agreements and vertical and horizontal sales. For each dependent variable we present two sets of results:
the first set examines PTAs and WTO membership measured with country-level dummy variables, as is customary in the extant literature. The second sets of results directly examine the
mechanism—tari↵ cuts—around which our theory is built. We present models with and without
interactions with firm size (and productivity).
5.1
Vertical Sales
Table 3 reports the estimates of models of vertical sales from an affiliate in the host country
to its parent company in the US. In the baseline model presented in Column 1, we find that
preferential liberalization (PTA with the US) and MFN liberalization (WTO membership) are
positively associated with vertical sales, but the estimated coefficients are not statistically di↵erent
from zero.25
Table 3
Our theory predicts that the largest and most productive firms should gain the most from
preferential liberalization. Hence, we expect the e↵ects of PTAs on vertical sales to scale with firm
productivity. In columns 3–5 of Table 3 we probe this relationship by adding an interaction term
between PTA with U.S. and three di↵erent proxies for firm-level productivity: affiliate size (Column
3), affiliate physical plant and equipment assets (PPE, column 4), and headquarter productivity
(column 5). We are unable to compute productivity at the affiliate-level, so we use parent-company
25 GATT
membership and a BIT with the U.S., on the other hand, are strongly associated with
higher sales. However, the results are not robust: when we control for affiliate size (natural log of
employment), BIT is the only relationship that survives (Column 2).
20
productivity as a proxy.26 Since size and physical assets are highly correlated with productivity
(Bernard et al., 2007), we rely principally on these variables, available at the affiliate level, for the
most direct test of our theory at the level of the individual firm.
We fit the following models:
Sajit = ↵ P T A with U Sit +
P roductivityaji + ! P T A with U Sit ⇥ P roductivityaji + (3)
+ Cit + 'i + &j + ⌧t + ✏ajit
As before, all models include controls for partner country’s GDP per capita Cit , industry&j , country- 'i , and year ⌧t fixed e↵ects.
The results are consistent with our expectations. We find that the marginal e↵ect of PTAs
on vertical sales is stronger for larger (in terms of employment, column 3, and assets, column 4) and
more productive firms (HQ productivity, column 5). The relationships are strong in substantive
and statistical terms. To ease the interpretation of the results, we graph the marginal e↵ect of
U.S. PTA across the full range of employment in Figure 4. We observe that the positive marginal
e↵ect of U.S. PTA becomes statistically significant for firms with more than 90 employees. Figure
5 graphs the predicted sales under a PTA and without a PTA for affiliates at the mean level of
employment (55 employees), one standard deviation (445 employees) and two standard deviations
(4,900 employees) above the mean. The results are similar if we use assets as the measure of affiliate
size (Column 4).
Figures 4 and 5
Measured in terms of headquarter productivity (column 5), we find that PTAs appear to
have a strong and positive e↵ect on the vertical activities of the most productive firms. Specifically,
column 5 shows that a standard deviation change in productivity increases the marginal e↵ect of
26 Following
Bilir (2014), we compute Solow residuals as our measure of parent firm productivity.
See the appendix for further details.
21
a PTA with the U.S. on affiliate vertical sales by around 11%.27 WTO membership, on the other
hand, has no relationship with vertical sales back to the US, suggesting that preferential and MFN
liberalizations a↵ect firm activities in di↵erent ways.
5.1.1
Exploring the Mechanism: Reduction in Trade Costs
Recall that our main mechanism focuses on the reduction of trade costs arising from preferential
and MFN liberalization. Moreover, the e↵ects should vary depending on the level of the cuts, and
the type of MNC activities. In the theory section, we showed that even though the US has already
low MFN tari↵s (on average), tari↵ reductions implemented by the US with PTA partners are not
trivial. Moreover, we demonstrated that much of these tari↵ cuts occur in intermediates and high
contract intensive goods, which make the bulk of the global value chain.
We now place Tari↵ Cut (proportional) on the right hand-side of our model to estimate
the e↵ect of cuts on di↵erent types of MNCs activity. A positive and statistically significant
relationship between US tari↵ cuts and Vertical sales (from the affiliate to the US parent) would
support Hypothesis 1. Cuts o↵ered by the US to all WTO members on MFN terms, on the other
hand, should have a weaker e↵ect on sales back to the US: specific affiliates operating in the
WTO entrant face strong competition from other potential sources – also enjoying MFN and/or
preferential tari↵s.28 We expect preferential liberalization to scale with affiliate size.
27 Models
A.5.9–A.5.10 in Appendix A.5 add headquarter fixed e↵ects to models 3 and 7 in Table
3, yielding substantively similar results. Models A.5.11–A.5.12 add country-year fixed e↵ects aimed
at absorbing all country characteristics that a↵ect vertical sales that are time-varying (as well as
those that are time-invariant). These could include, but are not limited to, economic performance,
tax rates, capital controls, changes in the bilateral relationship with the U.S. and a host of other
policy and institutional features, along with shocks such as military conflict or financial crises. The
estimated marginal e↵ects of productivity and size are nearly identical to those reported on Table
3.
28 We
should note that cuts in tari↵s in goods for intermediate use by the US partner through
preferential or MFN liberalization could lead to export platform sales, by reducing the costs of final
goods. We explore this relationship in the next section.
22
In columns 6-8 of Table 3, we examine the relationship between tari↵ cuts and vertical sales.
Column 6 shows that tari↵ cuts o↵ered by the US under preferential agreements (PTA Tari↵ Cut
(US)) are not associated with vertical activities of the average affiliate; MFN Cuts o↵ered by the US
through WTO are weakly associated with vertical sales. Column 7 reports the interaction between
size and US preferential Tari↵ Cuts. As expected, the e↵ect of preferential trade liberalization on
vertical sales positively scales with firm size. Figure 6 illustrates the marginal e↵ect of a tari↵ cut
along the range of affiliate size. US tari↵ cuts reduce vertical sales from small affiliates; the marginal
e↵ect becomes positive and statistically significant at the level of 148 employees. Multilateral
liberalization through the WTO is also non-linear in size as reported in column 8: the relationship
between WTO tari↵s reductions is positive for small firms and turns negative for bigger firms, but
these relationships do not attain statistical significance at conventional levels.
Figure 6
In sum, we find strong support for Hypotheses 1: preferential liberalization increases sales
back to the US. The increase in sales is related to the size of the cut in preferential tari↵s o↵ered
by the US to its trading partners, and the e↵ect scales with firm size and productivity.
5.2
Export Platform Sales
In Table 4 we examine the relationship between trade liberalization and export-platform sales.29
In Column 1 of Table 4 we present evidence consistent with a positive relationship between liberalization by partner countries (preferential and MFN) and export platform sales by affiliates of
US MNCs. First, a US PTA is associated with a 72% increase in export-platform sales; and second, the estimated increase in export platform sales following WTO membership is 86% (column
1). Controlling for affiliate size a↵ects the magnitude of the coefficients on PTA with the US and
29 Unfortunately
we cannot test systematically the liberalization e↵ect of the tari↵s cuts o↵ered to
the partner by its PTAs with third countries. The BEA data does not record the specific destination
of the third-country exports, so we are not able to identify whether those exports go to a particular
PTA participant. As a control we use cumulative PTAs as a proxy for reductions in the cost of
exporting to third countries through preferential trade agreements.
23
WTO membership, but the associations remain strong in substantive and statistical terms (column
2). We also find that the relationship between PTA with the US and export platform sales is not
linear: in this case the increase in exports to third countries associated with the presence of a
PTA is larger for smaller firms (columns 3 and 4), and does not seem to scale with headquarter
productivity (column 5).30
5.2.1
Exploring the Mechanism: Reduction in Trade Costs
Beginning with column 6 of Table 4, we examine the e↵ects of tari↵s cuts implemented under
preferential and MFN agreements. We find that for the average firm (irrespective of size), host
country PTA and MFN cuts are associated with higher exports to third markets. We also find that
the e↵ects of both preferential and MFN tari↵ cuts implemented by the partner country on export
platform activities appear to scale with affiliate size (preferential in column 7 and WTO in columns
8 and 10). To illustrate the substantive e↵ects we graph these results. Figure 7 shows the marginal
e↵ect of cuts o↵ered by US PTA partners as firm size increases. Figure 8 shows that MFN cuts
o↵ered at time of accession to WTO increase export platform sales by the largest firms only. Both
results provide strong support to Hypotheses 2 and 4.
Figures 7 and 8
6
Additional Evidence
In this section we examine additional implications of our theory. In particular, we show that the
impact of tari↵ cuts implemented under PTAs with the US and accession to the WTO on affiliates’
vertical and export platform sales are stronger for industries that use intermediate goods more
intensively. Moreover, we endogenize preferential tari↵ cuts to mitigate concerns about reverse
30 Cumulative
PTAs with third countries is also associated with export platform sales. A standard
deviation increase in the natural log of cumulative PTAs – equivalent to the partner participating
in three additional PTAs– is associated with a 6% increase in export-platform sales. These results
are robust to dividing the number of PTAs in to quartiles and entering the quartile dummies into
the model in place of cumulative PTAs.
24
causality. We then examine the China case, focusing on changes in affiliate supply chain activities
following China’s accession to the WTO. Finally, we estimate the e↵ect of PTAs on market concentration to provide further evidence of the redistribution e↵ect from the least to the most productive
firms.
We also perform a number of robustness tests and report the results in the appendix. First,
we show that our results are robust to entropy balancing, which allows us to better account for
firm-level characteristics. Second, we implement a placebo test using local sales of affiliates, i.e.
horizontal FDI activities, as the outcome variable. Finally, we show that our findings are robust to
additional fixed e↵ects31 and di↵erent transformations of the main variables.
6.1
Cuts in Intermediate Goods
We expect sharper increases in vertical sales following higher tari↵ cuts in intermediate products,
which constitute the bulk of global supply chain trade. We explore relationships between intermediates trade and liberalization in Table 5 using BEA data on affiliate trade in intermediates.32
In column 1, we model intermediate goods imported by the foreign affiliate from the U.S. parent.
The results reproduced in Column 1 suggest that tari↵ cuts implemented by the partner country
under a US PTA have a strong e↵ect on intermediate sales to the affiliate; the result is significant
at 95%, and substantive large: a 10% increase in preferential tari↵s cuts would result in a 3.25%
increase in headquarter sales of intermediate goods. Tari↵ cuts o↵ered by the partner on MFN
terms under WTO, on the other hand, result in a 1.4% drop in headquarter sales of intermediate
goods to the affiliate. The di↵erence between preferential and MFN tari↵ cuts becomes apparent:
under preferential liberalization, the parent company faces a preferential cut advantage leading to
31 We
introduce parent firm-level fixed e↵ects to control for unobserved heterogeneity at the parent
firm level. We also added country-year dummy variables, which absorb all country characteristics
that a↵ect vertical sales that are time-varying (as well as those that are time-invariant). Finally, we
estimate models including country- and industry-specific time trends, which capture slow moving
unobserved confounders a↵ecting affiliate sales that vary by partner country or industry.
32 Following
Hanson et al. (2005), we capture trade in intermediates using the BEA measure of
affiliate imports of “goods intended for further processing, assembly, or manufacture.”
25
higher intermediate sales; when the partner country lowers tari↵s on MFN terms the affiliate may
procure from the home country or from third parties.
Next we explore whether tari↵ cuts o↵ered by the partner under WTO have stronger e↵ects
on export platform activities among firms in industries with high intermediate input intensities, a
result that would follow from our argument. Column 2 in Table 5 interacts tari↵ cuts with industry
intermediate goods intensity.33 We find the positive relationship between tari↵ cuts and export
platform sales increases with the intensity of intermediate goods imports. The results are shown
graphically in Figure 9. In column 3, we find that the result holds to the inclusion of country-year
fixed e↵ects.
6.2
Instrumental Variables Estimation
The results thus far indicate that tari↵ cuts increase affiliates sales to the home market. However,
if time-varying firm-level characteristics are correlated with affiliates sales and with tari↵ cuts, our
models would not be correctly identified and our estimates would be biased. In other words, there
is the possibility that tari↵ cuts are endogenous to affiliates sales. For instance, large firms might
lobby for extensive tari↵ cuts, since they are less threatened by trade liberalization compared to
smaller firms. While it is perhaps unlikely that any one individual firm will influence policy, such
activity could bias our estimates upward. Note that since we find an impact of tari↵ cuts on sales
for intermediates goods, concerns about reverse causality are assuaged somewhat. Indeed, should
tari↵ cuts be endogenous to sales, the relationship should hold for both intermediates and finished
goods. In any case, in this section we further address concerns about endogeneity, relying on an
instrumental variable approach.34
33 To
construct intermediate goods intensities at the industry level, we compute the share of
intermediates in total sales for each industry-benchmark year, 1994-2009. The question was not
included in the 1989 benchmark survey, so we are not able to include that year.
34 Ours
is not the first study to recognize that trade liberalization could be endogenous to firm-
level characteristics (Trefler, 1993; Goldberg and Pavcnik, 2005). Trefler (1993) uses industry
characteristics such as market concentration to instrument for non-tari↵ barriers. This approach
has been followed by other studies (Trefler, 2004; Amiti and Konings, 2007). Goldberg and Pavcnik
26
Our identification strategy follows Cheng (2012). In particular, we use tari↵ cuts implemented by other countries that form PTAs with the same US PTA partner as instruments for U.S.
PTA cuts. For instance, we use tari↵ cuts implemented by Canada as a result of its PTA with Costa
Rica to instrument for tari↵ cuts implemented by the US in its PTA with Costa Rica. The intuition
is that the US tries to negotiate the same (preferential) tari↵ deal agreed by other countries that
compete in the same markets in order to level the playing field with potential competitors. We
include PTAs negotiated either at the same time or before the US PTAs.35 We label the instrument
Competitor Cut.
Unfortunately, we are able to instrument only a sub-sample of the PTAs formed by the US
for three reasons. First, we are unable to instrument the PTAs that have been signed but that they
are not in force by 2009 (with the exception of the US-Korea PTA), the last benchmark year in the
BEA data. Second, we are unable to instrument Canada and Mexico since we do not have data
on PTAs formed before NAFTA.36 Third, we are unable to instrument tari↵ cuts for some PTAs
since data for some developing countries are not available, or are only very sparsely available, from
WITS. We are left with six instrumented PTAs: Australia, Chile, Costa Rica, Peru, Singapore,
(2005) use as instruments the level of tari↵s that are in place before trade liberalization. Both
approaches share the same problem: the instruments are correlated with tari↵s and with the
outcome variables, and therefore likely violate the exclusion restriction.
35 Before
starting negotiations, trade partners establish a joint study group composed of high-level
officials and experts from both sides. Such a group has the goal to assess the potential for enhanced
trade relations and suggest tari↵ reduction in specific industries. When the joint study group ends
its work, formal negotiations begin. In all the PTAs used as instruments the establishment of joint
study groups, informal and formal negotiations overlap with these of the PTAs instrumented. Note
that treaties can be amended between signature and ratification.
36 Canada
formed PTAs with Portugal and Spain in 1954, with Australia in 1960, and with New
Zealand in 1980. None of them have been ratified by the WTO and they are all inactive except
the PTA with Australia. Mexico formed several PTAs with other Latin American countries in the
1980s. None of them has been ratified by the WTO and they are now all inactive.
27
and South Korea. (For the full list of instrumented PTAs and their instruments, see Table A.2 in
the appendix.)
Recall that our key variable is the interaction between tari↵ cuts and productivity and so
we also need to instrument this interaction term. Following previous studies (Park et al., 2010),
we use the interaction between Competitor Cut and our measures of productivity, i.e. number of
employees and assets (PPE), to instrument for the interaction term in our main regressions. More
formally, we estimate two stages (Wooldridge, 2012). The first-stage models are the following:
Cutij,t
1
=
1 Comp.
Cutij,t
+ ' j + & i + ⌧t
Cutij,t
1
⇥ P rodij,t
1
1
+
1 + ⌘ij,t
2 P rodij,t 1
+
3 Comp
Cutij,t
1
⇥ P rodij,t
1
+
4 Xj,t 1 +
(4)
1
=
1 Comp
Cutij,t
+
4 Xj,t 1
+ 'j + & i + ⌧t
1
+
2 P rodij,t 1
1
+ ⇣ij,t
+
3 Comp
Cutij,t
1
⇥ P rodij,t
1+
(5)
1
The second-stage model is:
Sij,t =
d
1 Cutij,t 1
+
2 P rodij,t 1
+
\
⇥ P rodij,t
3 Cut
1
+
4 Xj,t 1
+ ' j + & i + ⌧t
1
+ ✏ij,t
1
(6)
Armed with our instruments, our identification strategy is sound if three conditions are
satisfied. First, tari↵ cuts implemented by competitors should not impact affiliate sales to the US.
Since vertical FDI is a↵ected almost exclusively by the level of tari↵s with the home country, such
a possibility seems remote. However, it might be the case that PTAs formed by US competitors
increase economic activities of the affiliates of firms from those competitors, which in turn raise
the demand of labor and other inputs in the partner countries. Such increases in wages and input
costs may also a↵ect sales of US affiliates operating in these host countries through increasing costs
of production. To mitigate this concern, we select countries (1) that are relatively small and/or
less developed compared to the US (when data are available); (2) that negotiated PTAs at about
the same time than the US did so that any e↵ects through the labor market have no time to
28
materialize.37 Table A.2 in the appendix reports which PTAs we use to instrument Competitor
Cut.38
Second, our instruments have to be strong predictors of Competitor Cut. The correlation
between our instrument and Competitor Cut is equal to 0.45. All the diagnostics (reported in the
Table 6) shows that our instrument is strong and there are no concerns about under-identification.
Third, our instruments should not be correlated with (time-varying) industry characteristics. This might be the case if US MFN tari↵s (pre-PTA) are correlated with MFN tari↵s of US
competitors that form agreements with the same host markets. Indeed the level of tari↵s before the
formation of a PTA may be a proxy of industry characteristics, which are in turn correlated with
our outcome variable. Formally, Cov(M F NU S , M F NU SCompetitor ) = 0. Indeed, the correlation
between US MFN and US Competitors MFN is very low, i.e. ⇢ = 0.08.
Table 6 reports the results of the instrumental variable estimations. Instrumenting tari↵s
cuts implemented under a PTA signed with the US by the cuts implemented by the partner with
third countries yields results in line with those presented in Table 3: reciprocal liberalization through
PTAs lead to lower vertical sales by smaller affiliates, and higher sales by larger ones irrespective
of whether we measure size in terms of assets (Column 3) or employment (Column 6). In the IV
estimation the marginal e↵ect of tari↵ cuts turns positive for firms that employ more than 218
employees.
Table 6
Importantly, both instruments are positive and statistically significant in the first stage
(reported in Table 6). Regarding the diagnostics, (1) the Kleibergen-Paap Wald rk F statistic
shows that our models are not weakly identified; (2) the Kleibergen-Paap rk LM statistic shows that
our models are not under-identified; (3) the Anderson-Rubin Wald test shows that orthogonality
conditions are valid. In sum, results from our instrumental variable estimations (paired with
37 In
this spirit, we exclude EU PTAs from our instrument.
38 Figure
A.5 in the appendix shows that the correlation between residuals of the second stage
and our instrument is equal to zero.
29
all the other analyses) seem to indicate that the arrow of causality goes from preferential trade
liberalization to affiliates sales to the US and not the other way around.
6.3
The Case of China
We examine the China case since it is the largest economy to join the WTO, and among the
emerging markets, China attracts a large amount of foreign direct investment and is central to
global supply chain activities (Branstetter and Foley, 2010; Wang and Wei, 2010). Therefore, we
re-run our main analyses focusing only on U.S. affiliates in China. Table A.8 shows both vertical
and export-platform activities increased dramatically after the accession of China into the WTO
(for ease of interpretation see figures A.7 and A.8). As China gained access to US MFN tari↵ cuts
the smallest (least productive) affiliates experience a sharp drop in their sales back to the US; access
to US MFN tari↵ did not seem to a↵ect vertical sales among the largest (and most productive)
affiliates operating in China. The export platform sales results suggest that when comparative
advantages are very large, the positive e↵ect of international trade agreements on affiliate activities
is magnified.
6.4
Preferential Trade Liberalization and Market Concentration
We argued that preferential liberalization through PTAs benefits the largest and most productive MNCs, and the results in the previous section provide evidence suggesting that this is indeed the case. In this section we further probe the implications of the theory to examine the
e↵ects of preferential (and multilateral) trade liberalization on market concentration. We compute Herfindahl-Hirschman Indices of employment and sales concentration of U.S. affiliates at the
country and industry levels, and over time, and examine how these indices change following host
country participation in trade agreements.
First we examine levels of employment and sales concentration in the twenty countries
that entered into a PTA with the United States. Table 7 presents HHI of sales and employment
concentration, as well as average employment and assets of US affiliates in PTA partner countries,
before and after the PTA is in e↵ect. The bottom rows of the table shows the average pre- postPTA change in all PTA countries. We see that concentration tends to increase post-PTA, both
30
in terms of sales and employment. Moreover, affiliates tend to be larger (in terms of assets and
employment) after preferential liberalization.
Table 7
To analyze these patterns more systematically, we report the results of regressions of employment (columns 1-6) and sales (columns 7-12) HHI in Table 8. Models 1, 2, 7 and 8 are
Herfindahl-Hirschman Indices of concentration at the country level; all other models are of HHI
computed at the NAICS 4-digit level. In models 5, 6, 11 and 12, we restrict the sample to industries
with positive sales to the US. The results from these analyses are quite revealing: PTAs are associated with higher employment and sales concentration in all models39 ; the results are substantively
and statistically significant. Multilateral liberalization, on the other hand, leads to mixed and less
robust results: GATT tends to be associated with lower concentration, whereas liberalization under
WTO coincides with more concentrated markets.
Table 8
7
Conclusion
International economic institutions are a prominent feature of the post-WWII economic order
and of the current wave of globalization. It is no surprise that there is a heated debate on the
e↵ect of international institutions on important economic outcomes (Büthe and Milner, 2008, 2014;
Mansfield and Reinhardt, 2008). Such a debate is fuelled by the ongoing deadlock in the Doha
round (Kessie, 2013), by the controversial impact of recent IMF interventions (Pop-Eleches, 2009;
Stone, 2002), and by the dramatic proliferation of trade agreements (Mansfield and Milner, 2012).
In this paper we analyzed the impact of international economic institutions on the activities of most
productive and powerful economic actors in the market: multinational corporations. In particular,
we focuses on the global supply chain activities of US MNCs.
39 Signing
a Bilateral Investment Treaty with the US is associated with lower concentration in
the industry models, but not at the country level.
31
Our theory argued that preferential and multilateral liberalization should have heterogeneous e↵ects on MNC activities. The source of that heterogeneity depends on the type of MNC
activity, the size of the tari↵ cuts, and where those cuts occur. Cuts o↵ered by the US on preferential terms should promote sales back to the US, while preferential tari↵s o↵ered by the partner
would result in higher sales to third parties. MFN liberalization by the partner country, on the
other hand, should mainly a↵ect sales to third countries, but have a limited e↵ect on vertical sales
(back to the US). When the host country engages in preferential liberalization with third countries,
we should also observe an increase in export platform sales by US affiliates. However, this preferential opening up of third markets should not a↵ect vertical sales. We further argued, drawing
on insights from international trade theory, that the main beneficiaries are the largest and most
productive firms.
In the analyses presented in the previous sections we find strong support to our hypotheses. The analyses shows that: (1) MFN and preferential liberalization impact global production
activities in di↵erent ways: WTO/MFN’s e↵ect is primarily observed on export-platform sales;
preferential liberalization on the other hand a↵ects vertical sales and export platform sales; (2) the
mechanism operates through the reduction of trade costs; (3) the largest/most productive firms
disproportionately reap the benefits from liberalization.
Our paper makes three contributions to the literature. First, our paper explains the ways
in which the shift in trade governance from a multilateral to a bilateral/regional setting contributes
to firms’ global supply chain activities. To our knowledge, our paper is the first to demonstrate the
empirical link between tari↵ cuts through PTAs and the WTO and firms’ trade-related investment
activities. Second, we show that international economic institutions matter, but only for a relatively
small number of firms. Therefore the paradox of this current wave of globalization is that the
proliferation of institutions establishing regulatory provisions among countries is accompanied with
a reduction in the number of economic actors benefiting from such regulations. Third, our paper
suggests that debates over the e↵ects of international institutions on economic and policy outcomes
are best informed through evidence at the micro-level. Thus, it is natural to change the unit of
analysis from countries to companies.
32
Finally, our findings have two important policy implications. First, international economic
institutions are responsible for the increasing market concentration, leading to the expansion of
larger and more productive firms. Since the reallocation of sales from the least to the most productive firms occurs through the labor market channel, international institutions are likely to contribute
to income inequality within countries. The design of such institutions—PTAs in particular—should
tackle this risk of increasing inequality through provisions that establish adequate safety nets. Moreover, since PTAs increase market concentration, preferential liberalization should be paired with
careful domestic regulations and robust competition policies to avoid the occurrence of monopolies
or cartels. Without such devices, one of the main goals of trade liberalization—enhanced consumer
welfare—may not be achieved.
33
References
Abbott, K. W. and D. Snidal (1998). Why states act through formal international organizations.
Journal of Conflict Resolution 42 (1), 3–32.
Amiti, M. and J. Konings (2007). Trade liberalization, intermediate inputs, and productivity:
Evidence from indonesia. American Economic Review 97 (5), 1611–1638.
Anderson, J. E. and E. V. Wincoop (2004). Trade costs. NBER Working Paper 10480.
Antras, P. (2010). International trade and organizations. NBER Reporter, Number 2.
Antras, P. and C. F. Foley (2009). Regional trade integration and multinational firm strategies.
NBER Working Paper.
Antras, P. and R. W. Staiger (2012). O↵shoring and the role of trade agreements. American
Economic Review 102, 3140–3183.
Baccini, L. and A. Dür (2012). The new regionalism and policy interdependency. British Journal
of Political Science 42, 57–79.
Baccini, L. and J. Urpelainen (2015). Cutting the Gordian Knot of Economic Reform: How International Institutions Promote Liberalization. New York: Oxford University Press.
Baldwin, R. (2012). Wto 2.0: Global governance of supply chain trade. CEPR Working Paper No.
64.
Bekkers, E., J. Francois, and M. Manchin (2012). Import prices, income and inequality. European
Economic Review 56 (4), 848–69.
Bernard, A. B. and J. B. Jensen (1999). Exceptional exporter performance: Cause, e↵ect, or both?
Journal of International Economics 47 (1), 1–25.
Bernard, A. B., J. B. Jensen, S. J. Redding, and P. K. Schott (2007). Firms in international trade.
Journal of Economic Perspectives 21 (3), 105–130.
Bernard, A. B., J. B. Jensen, and P. K. Schott (2006). Trade costs, firms and productivity. Journal
of Monetary Economics 53, 917–937.
Bilir, L. K. (2014). Patent laws, product lifecycle lengths, and mulinational activity. American
Economic Review . forthcoming.
Bilir, L. K., D. Chor, and K. Manova (2013). Host country financial development and mnc activity.
Working Paper.
Branstetter, L. and C. F. Foley (2010). Facts and fallacies about us fdi in china. In R. C. Feenstra
and S.-J. Wei (Eds.), China’s Growing Role in World Trade, pp. 513–539. University of Chicago
Press.
Büthe, T. and H. V. Milner (2008). The politics of foreign direct investment into developing
countries: Increasing fdi through international trade agreements? American Journal of Political
Science 52 (4), 741–762.
34
Büthe, T. and H. V. Milner (2014). Foreign direct investment and institutional diversity in trade
agreements: Credibility, commitment, and economic flows in the developing world, 1971–2007.
World Politics 66 (1), 88–122.
Chase, K. (2003). Economic interests and regional trading arrangements: The case of nafta. International Organization 57 (1), 137–74.
Cheng, W. (2012). Tari↵s and employment: Evidence from chinese manufacturing industry. Working Paper, London School of Economics.
Collier, P. (2006). Why the wto is deadlocked: andwhat can be done about it. International
Organization 29 (10), 1423–1449.
Dür, A., L. Baccini, and M. Elsig (2014). The design of international trade agreements: Introducing
a new database. Review of International Organization.
Eaton, J., S. Kortum, and F. Kramarz (2011). An anatomy of international trade: Evidence from
french firms. Econometrica 79 (5), 1453–1498.
Feenstra, R. C. and G. 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.
Francois, J. and O. Pindyuk (2012). Mapping of hs6, bec, gtap, and wiod product categories. In The
World Input Output Database (WIOD): Contents, Sources and Methods. European Commission.
Goldberg, P. K. and N. Pavcnik (2005). Trade, wages and the political economy of trade protection:
Evidence from colombian trade reforms. Journal of International Economics 66 (1), 75–105.
Goldstein, J. L., D. Rivers, and M. Tomz (2007). Institutions in international relations: Understanding the e↵ects of the gatt and the wto on world trade. International Organization 61 (1),
37–67.
Gowa, J. and S. Y. Kim (2005). An exclusive country club. World Politics 57 (4), 453–478.
Gray, J. (2013). The Company States Keep: International Economic Organizations and Investor
Perceptions. Cambridge: Cambridge University Press.
Hainmueller, J. (2012). Entropy balancing for causal e↵ects: A multivariate reweighting method
to produce balanced samples in observational studies. Political Analysis 20 (1), 25–46.
Hanson, G. H., R. J. Mataloni Jr, and M. J. Slaughter (2005). Vertical production networks in
multinational firms. Review of Economics and statistics 87 (4), 664–678.
Helpman, E., M. J. Melitz, and S. R. Yeaple (2004). Export versus fdi with heterogeneous firms.
American Economic Review 94 (1), 300–316.
Henisz, W. J. (2000). The institutional environment for economic growth. Economics and Politics 12 (1), 1–31.
Jensen, J. B., D. P. Quinn, and S. Weymouth (2015). The influences of foreign direct investments, intrafirm trading, and currency undervaluation on u.s. firm trade disputes. International
Organization. forthcoming.
35
Keohane, R. O. (1984). After Hegemony: Cooperation and Discord in the World Political Economy.
Princeton: Princeton University Press.
Kessie, E. (2013). The future of the doha development agenda. European Yearbook of International
Economic Law 4 (1), 481–494.
Krugman, P. R. (1979). Increasing returns, monopolistic competition, and international trade.
Journal of International Economics 9 (4), 469–479.
Manger, M. S. (2009). Investing in Protection: The Politics of Preferential Trading Agreements
between North and South. New York: Cambridge University Press.
Mansfield, E. D. and H. V. Milner (2012). Votes, Vetoes, and the Political Economy of International
Trade Agreements. Princeton: Princeton University Press.
Mansfield, E. D. and E. Reinhardt (2008). International institutions and the volatility of international trade. International Organization 62 (4), 621–652.
Melitz, M. (2003). The impact of trade on intra-industry reallocations and aggregate industry
productivity. Econometrica 71 (6), 1695–1725.
Nunn, N. (2007). Relationship-specificity, incomplete contracts, and the pattern of trade. The
Quarterly Journal of Economics 122 (2), 569–600.
Odell, J. S. (2009). Breaking deadlocks in international institutional negotiations: the wto, seattle,
and doha. International Studies Quarterly 53 (2), 273–299.
Park, A., D. Yang, X. Shi, and Y. Jiang (2010). Exporting and firm performance: Chinese exporters
and the asian financial crisis. The Review of Economics and Statistics 92 (4), 822–842.
Pelc, K. (2011). Why do some countries get better wto accession terms than others? International
Organization 65 (4), 639–672.
Pop-Eleches, G. (2009). From Economic Crisis to Reform: IMF Programs in Latin America and
Eastern Europe. Princeton: Princeton University Press.
Rose, A. K. (2004). Do we really know that the wto increases trade? American Economic Review 94 (1), 98–114.
Stone, R. W. (2002). Lending Credibility: The International Monetary Fund and the PostCommunist Transition. Princeton: Princeton University Press.
Tomz, M., J. Goldstein, and D. Rivers (2007). Do we really know that the wto increases trade?
comment. American Economic Review 97 (5), 2005–2018.
Trefler, D. (1993). Trade liberalization and the theory of endogenous protection: An econometric
study of u.s. import policy. Journal of Political Economy 101 (1), 138–160.
Trefler, D. (2004). The long and short of the canada-u.s. free trade agreement. American Economic
Review 94 (4), 870–895.
Wang, Z. and S.-J. Wei (2010). What accounts for the rising sophistication of china’s exports?
In R. C. Feenstra and S.-J. Wei (Eds.), China’s Growing Role in World Trade, pp. 63–104.
University of Chicago Press.
36
Wooldridge, J. (2012). Econometric analysis of cross section and panel data. MIT University Press.
Yeaple, S. R. (2006). O↵shoring, foreign direct investment, and the structure of us trade. Journal
of the European Economic Association 3 (4), 602–11.
37
-.1
0
Percentage change
.1
.2
.3
.4
Figure 1: Changes in Affiliate Activities of U.S. Multinationals, 1989-2009
Local sales
Export-Platform
Vertical sales
Note: The figure reports changes in the average affiliate sales shares to three destinations: local
(sales to the host country), export-platform (sales to third countries), and vertical (sales to the
U.S. parent company). The data are from the U.S. Bureau of Economic Analysis.
Figure 2: Tari↵ reductions in US PTAs since 1990
Australia
Bahrain
Canada
Chile
Colombia
Costa Rica
Dominican Republic
El Salvador
Guatemala
Honduras
Jordan
Korea, Rep.
Mexico
Morocco
Nicaragua
Oman
Panama
Peru
Singapore
Vietnam
0
10
20
30
Tariff Cut (MFN-PRF)
40
Note: The figure displays the distribution of tari↵ cuts (MFN-PRF) for 20 PTAs signed by the US
after 1990. Data come from WITS (2014) and are at the HS 6-digit tari↵ line.
2.8
2.9
3
3.1
3.2
Figure 3: Mean tari↵ reductions in US PTAs by product use
Final & Mixed Goods
US Tariff Cut
Intermediates
90% CI
Note: The figure displays the di↵erences between MNF tari↵s prior to the formation of PTAs and
preferential tari↵s (PRF) after a PTA is in force by type of product classified as intermediate or
consumption and mixed use. The categorization of products come from Francois and Pindyuk
(2012) and Bekkers et al. (2012). The whiskers represent 90% confidence intervals.
0
-.5
0
Marginal effects
.5
1
5
10
Ln Employment (% observations)
1.5
2
15
Figure 4: Marginal E↵ect of PTA with US on Vertical Sales by Firm Size
0
1
2
3
4
5
6
7
8
Ln Employment (affiliate)
9
10
11
12
Note: Marginal e↵ects (and 90% confidence intervals) of PTA with US on affiliate sales the US
based on Results from Column 3 in Table 3. See Tables 1 and 3, and text for description of the
sales and productivity measures.
0
1
Predicted Ln Vertical Sales
5
4
3
2
6
Figure 5: Di↵erences in Predicted Vertical Sales by Size and US PTA
Mean (4.0)
+ 1 sd (6.3)
Ln Employment (affiliate)
No PTA
+ 2 sd (8.5)
PTA with US
Note: Predicted vertical sales (and 95% confidence intervals), holding all other variables at their
means, based on results from Column 3 in Table 3.
15
0
-5
0
Marginal effect
5
5
10
Ln Employment (% observations)
10
Figure 6: Marginal E↵ect of US Preferential Tari↵ Cuts on Vertical Sales by Firm Size
0
1
2
3
4
5
6
7
8
Ln Employment (affiliate)
9
10
11
12
Note: Marginal e↵ects (and 90% confidence intervals) of US PTA cuts based on Results from
Column 7 in Table 3.
0
-2
0
Marginal effect
2
4
5
10
Ln Employment (% observations)
6
15
Figure 7: Marginal E↵ect of Host PTA Tari↵ Cuts on Export-Platform Sales by Firm Size
0
1
2
3
4
5
6
7
8
Ln Employment (affiliate)
9
10
11
12
Note: Marginal e↵ects (and 90% confidence intervals) of partner PTA cuts on affiliate sales to third
markets based on Results from Column 7 in Table 4.
0
-1
0
Marginal Effects
1
2
5
10
Ln Employment (% observations)
3
15
Figure 8: Marginal E↵ect of Host MFN Tari↵ Cuts on Export-Platform Sales by Firm Size
0
1
2
3
4
5
6
7
8
Ln Employment (affiliate)
9
10
11
12
Note: Marginal e↵ects (and 90% confidence intervals) of partner MFN cuts on affiliate sales to
third markets based on Results from Column 11 in Table 4.
80
0
0
Marginal effects
10
20
20
40
60
Intermediate intensity (% observations)
30
Figure 9: Marginal E↵ect of Host MFN Tari↵ Cuts on Export-Platform Sales by Intensity of Use
of Intermediate Goods
0
.2
.4
.6
.8
Intensity of use of intermediate goods
1
Note: Marginal e↵ects (and 90% confidence intervals) of partner MFN cuts on affiliate sales to
third markets based on Results from Column 2 in Table 5.
Table 1: Summary Statistics
Firm level variables
Observations
Variable
82,946
Ln Total Sales
82,946
Ln Horizontal Sales
82,946
Ln Export Platform Sales
82,946
Ln Vertical Sales (to US)
82,946
Ln Employment (affiliate)
82,946
Ln PPE Assets (affiliate)
74,394
Productivity (headquarter)
.. Omitted to preserve anonimity of reporters
Country level variables
Variable
Ln GDP/capita (partner)
GATT Only (partner)
WTO (partner)
BIT with US
Ln Cumulative PTA (partner)
PTA with US
Tariff Cut (US)
Tariff Cut Proportional (US)
Tariff Cut PTA (partner)
Tariff Cut WTO (partner)
Tariff Cut WTO Proportional (US)
Observations
708
708
708
708
708
708
697
697
680
708
707
Average
9.534
8.532
3.440
2.154
4.038
7.428
0.178
Std. Dev.
3.116
3.819
4.569
3.731
2.245
3.579
0.610
Min
..
..
..
..
..
..
..
Max
..
..
..
..
..
..
..
Average
8.177
0.250
0.500
0.189
3.168
0.049
0.115
0.030
0.009
0.001
0.002
Std. Dev.
1.593
0.433
0.500
0.392
1.063
0.217
0.656
0.162
0.087
0.016
0.020
Min
4.451
0
0
0
0
0
0
0
0
0
0
Max
11.851
1
1
1
5.352
1
5.784
1
1
0.272
0.245
Note: The maximum and minimum values of the firm-level variables are suppressed to avoid disclosure of confidential information.
Total
Affiliates
0.517***
(0.187)
0.004
(0.007)
-0.295
(0.319)
0.141
(0.225)
-0.043*
(0.026)
0.004**
(0.002)
0.012
(0.093)
0.006
(0.113)
0.069
(0.120)
-0.079
(0.092)
0.203***
(0.071)
Y
Y
772
0.974
169
(2)
(3)
Affiliate Counts
ExportHorizontal
Platform
Affiliates
Affiliates
0.490***
0.216
(0.170)
(0.177)
0.004
0.010
(0.006)
(0.007)
-0.292
-0.034
(0.296)
(0.277)
0.176
0.009
(0.215)
(0.219)
-0.042
-0.044
(0.026)
(0.031)
0.004***
0.003**
(0.002)
(0.001)
0.010
0.062
(0.092)
(0.113)
0.039
-0.050
(0.106)
(0.120)
0.045
0.059
(0.113)
(0.141)
-0.050
0.041
(0.098)
(0.099)
0.214***
0.173**
(0.067)
(0.082)
Y
Y
Y
Y
772
772
0.976
0.962
169
169
Vertical
Affiliates
0.479**
(0.211)
0.009
(0.006)
-0.224
(0.273)
-0.082
(0.200)
-0.062*
(0.035)
0.003**
(0.001)
-0.087
(0.102)
-0.048
(0.088)
0.056
(0.117)
0.020
(0.106)
0.103*
(0.059)
Y
Y
772
0.964
169
(4)
Total Sales
1.404*
(0.802)
0.016
(0.038)
0.578
(1.822)
-0.212
(1.115)
-0.039
(0.079)
0.023***
(0.008)
0.258
(0.516)
-0.859
(0.803)
1.238*
(0.633)
0.472
(0.310)
0.985***
(0.351)
Y
Y
772
0.886
169
(5)
Horizontal
Sales
1.364**
(0.643)
0.011
(0.035)
0.234
(1.740)
0.115
(1.115)
-0.050
(0.076)
0.023***
(0.007)
0.285
(0.505)
-0.262
(0.817)
1.318**
(0.594)
0.503*
(0.292)
0.809**
(0.322)
Y
Y
772
0.879
169
(8)
Export
Platform Sales Vertical Sales
0.250
2.637**
(1.023)
(1.172)
0.065
0.111***
(0.051)
(0.038)
1.655
-1.170
(1.995)
(1.695)
0.066
0.285
(1.291)
(1.341)
-0.114
-0.131
(0.088)
(0.147)
0.003
0.010
(0.010)
(0.011)
0.370
-0.132
(0.683)
(0.612)
0.585
-0.558
(0.914)
(0.725)
0.861
0.465
(0.968)
(0.701)
0.751
0.995*
(0.746)
(0.538)
0.982**
0.406
(0.472)
(0.324)
Y
Y
Y
Y
772
772
0.849
0.865
169
169
(6)
(7)
Ln Aggregate Sales
Note: The dependent variables are the logged sum of total affiliates and the logged sum of total affiliate sales in each country-year based
on affiliate-level data from the BEA. All models include country and year fixed e↵ects. Robust standard errors adjusted for country-level
clustering. *** p < 0.01, ** p < 0.05, * p < 0.10.
Country FE
Year FE
Observations
R-squared
Countries
Ln Cumulative PTAs (partner)
PTA with US
BIT with US
WTO member (partner)
GATT only (partner)
Trade/GDP
Political Instability
Political Constraints (partner)
Ln Population (partner)
GDP growth (partner)
Dependent Variable
Ln GDP/capita (partner)
(1)
Table 2: International Economic Agreements and U.S. MNC Activities(Country-level)
0.307
(0.223)
0.376*
(0.198)
0.268
(0.258)
0.271**
(0.135)
0.048
(0.031)
0.154
(0.107)
0.201
(0.184)
0.270
(0.169)
0.204
(0.245)
0.231**
(0.105)
0.021
(0.026)
0.120
(0.097)
0.484***
(0.021)
(3)
(4)
(5)
(6)
(7)
(8)
82946
0.107
165
0.211
(0.183)
0.283*
(0.168)
0.202
(0.243)
0.227**
(0.106)
0.021
(0.027)
-0.466***
(0.120)
0.458***
(0.023)
0.141***
(0.043)
82946
0.0917
165
0.232***
(0.014)
0.142***
(0.041)
0.344*
(0.194)
0.261
(0.172)
0.194
(0.263)
0.212**
(0.108)
0.002
(0.028)
-0.981***
(0.224)
74394
0.0482
163
0.003
(0.038)
0.166**
(0.064)
0.245
(0.220)
0.397**
(0.181)
0.296
(0.232)
0.300**
(0.124)
0.065**
(0.032)
0.149
(0.120)
73736
0.114
163
0.548*
(0.296)
1.088
(0.715)
0.500***
(0.032)
0.144
(0.127)
-0.003
(0.046)
0.050
(0.325)
0.168
(0.153)
73736
0.117
163
0.547*
(0.289)
-2.248***
(0.362)
0.688***
(0.172)
0.480***
(0.022)
0.145
(0.125)
-0.001
(0.043)
0.081
(0.307)
0.177
(0.147)
-0.100
(0.118)
73736
0.114
163
1.093
(0.830)
1.088
(0.715)
0.501***
(0.032)
0.142
(0.127)
-0.002
(0.046)
0.051
(0.323)
0.166
(0.154)
Note: The dependent variable is the log of total affiliate sales to the US based on affiliate-level data from the BEA. Robust standard
errors adjusted for country-level clustering. All models include country, year, and industry fixed e↵ects. *** p < 0.01, ** p < 0.05, *
p < 0.10.
Observations
82946
82946
R-squared
0.0471
0.106
Countries
165
165
All models include year, industry and country fixed effects
WTO Cuts (US) x Ln Employment
PTA Tariff Cuts (US) x Ln Employment
PTA Tariff Cuts (US)
WTO Cut (US)
PTA x Productivity
Productivity (headquarter)
PTA x Ln Assets (PPE)
Ln Assets (PPE, affiliate)
PTA x Ln Employment
Ln Employment (affiliate)
PTA with US
Ln Cumulative PTAs (partner)
BIT with US
WTO member (partner)
GATT only
Ln GDP/capita
(2)
Vertical Sales Vertical Sales Vertical Sales Vertical Sales Vertical Sales Vertical Sales Vertical Sales Vertical Sales
(1)
Table 3: International Economic Agreements and U.S. MNC Affiliate Vertical Sales, 1989-2009 Benchmarks (Affiliate-level)
4.845*
(2.483)
82946
0.154
165
3.408
(2.262)
(2)
Export
Platform
Sales
-0.608**
(0.297)
0.636***
(0.173)
0.546**
(0.257)
0.336
(0.275)
0.221***
(0.080)
0.503***
(0.102)
0.578***
(0.042)
82946
0.157
165
3.382
(2.236)
(3)
Export
Platform
Sales
-0.630**
(0.299)
0.610***
(0.180)
0.550**
(0.263)
0.344
(0.272)
0.221***
(0.078)
1.733***
(0.308)
0.631***
(0.035)
-0.297***
(0.053)
82946
0.144
165
2.521
(2.066)
0.347***
(0.021)
-0.156***
(0.038)
(4)
Export
Platform
Sales
-0.439
(0.279)
0.582***
(0.172)
0.518*
(0.287)
0.306
(0.270)
0.181**
(0.075)
1.648***
(0.364)
74394
0.100
163
5.507**
(2.445)
0.179***
(0.035)
-0.046
(0.052)
(5)
Export
Platform
Sales
-0.623*
(0.336)
0.763***
(0.193)
0.720***
(0.234)
0.380
(0.274)
0.276***
(0.078)
0.588***
(0.120)
71558
0.148
164
0.940
(1.894)
0.558**
(0.261)
1.100***
(0.282)
0.634***
(0.036)
0.529**
(0.247)
0.196***
(0.070)
(6)
Export
Platform
Sales
-0.341
(0.258)
0.546***
(0.157)
71558
0.148
164
0.955
(1.893)
0.559**
(0.261)
-0.502
(0.332)
0.318***
(0.106)
0.633***
(0.036)
0.529**
(0.247)
0.196***
(0.070)
(7)
Export
Platform
Sales
-0.343
(0.258)
0.547***
(0.156)
71558
0.148
164
0.234***
(0.071)
0.921
(1.893)
-0.782**
(0.338)
1.099***
(0.282)
0.632***
(0.036)
0.530**
(0.247)
0.196***
(0.069)
(8)
Export
Platform
Sales
-0.338
(0.258)
0.549***
(0.157)
82931
0.153
165
2.844
(2.500)
0.992**
(0.396)
0.578***
(0.042)
0.375
(0.243)
0.200**
(0.083)
(9)
Export
Platform
Sales
-0.535
(0.325)
0.444***
(0.154)
82931
0.153
165
0.319***
(0.075)
2.819
(2.499)
-0.835***
(0.308)
0.577***
(0.042)
0.376
(0.243)
0.199**
(0.083)
(10)
Export
Platform
Sales
-0.531
(0.325)
0.448***
(0.153)
Note: The dependent variable is the log of total affiliate sales to third countries based on affiliate-level data from the BEA. Robust
standard errors adjusted for country-level clustering. All models include country, year, and industry fixed e↵ects. *** p < 0.01, **
p < 0.05, * p < 0.10.
Observations
82946
R-squared
0.0966
Countries
165
All models include year, industry and country fixed effects
Constant
WTO Cuts (partner) x Ln Employment
PTA Tariff Cuts (partner) x Ln Employment
PTA Tariff Cuts (partner)
WTO Cut (partner)
PTA x Productivity
Productivity (headquarter)
PTA x Ln Assets (PPE)
Ln Assets (PPE, affiliate)
PTA x Ln Employment
Ln Employment (affiliate)
PTA with US
Ln Cumulative PTAs (partner)
BIT with US
WTO member (partner)
GATT only
Ln GDP/capita
(1)
Export
Platform
Sales
-0.481
(0.337)
0.762***
(0.196)
0.622**
(0.267)
0.383
(0.246)
0.253***
(0.081)
0.544***
(0.112)
Table 4: International Economic Agreements and U.S. MNC Affiliate Export Platform Sales, 1989-2009 Benchmarks (Affiliate-level)
Table 5: Tari↵ Cuts and Export-Platform Sales by Intensity of Use of Intermediates
(1)
Imported
Intermediates
Dependent Variable
from HQ
0.003
Ln GDP/capita
(0.152)
0.345*
GATT only
(0.196)
-0.151**
WTO member (partner)
(0.063)
0.240
BIT with US
(0.160)
-0.022
Ln Cumulative PTAs (partner)
(0.026)
0.335**
PTA Tariff Cuts (partner)
(0.154)
WTO Cuts (partner)
Intermediate intensity
WTO Cuts (partner) x Intermediate intensity
Constant
Observations
R-squared
Countries
Fixed effects
2.369**
(1.186)
51824
0.0624
158
Country, Year
(2)
(3)
Export Sales
Export Sales
-0.063
(0.448)
0.527**
(0.229)
0.272
(0.307)
0.496
(0.424)
0.227**
(0.092)
-0.505
(0.436)
-0.205
(0.745)
22.662***
(3.643)
1.620
(3.262)
69988
0.0906
164
Country, Year,
Industry
-0.506
(0.389)
-0.100
(0.738)
21.796***
(3.353)
5.364***
(0.223)
69988
0.207
164
Country-year,
Industry
Note: The dependent variable in Column 1 is the log of the sales of goods for further processing
from the US parent company to the affiliate. The dependent variable in Column 2 is the log of
sales by US affiliates to third countries. Intermediate intensity is defined in the text. *** p < 0.01,
** p < 0.05, * p < 0.10.
Table 6: Preferential Cuts and U.S. MNC Affiliate Vertical Sales: IV Regression
(1)
Dependent Variable
Ln GDP/capita
GATT only
WTO member (partner)
BIT with US
Ln Cumulative PTAs
Ln Assets (PPE, affiliate)
(2)
(3)
2nd Stage
First Stage
PTA Tariff
PTA Tariff
Cuts (US) x Vertical Sales
Cuts (US)
Ln Assets
0.002
0.023
0.412**
(0.009)
(0.087)
(0.189)
-0.001
-0.011
0.300
(0.002)
(0.023)
(0.191)
-0.001
-0.008
0.138
(0.005)
(0.044)
(0.257)
-0.001
-0.006
0.135
(0.002)
(0.015)
(0.093)
-0.003
-0.027
0.017
(0.002)
(0.022)
(0.032)
0.00003
0.0006*
0.242***
(0.00002)
(0.0004)
(0.013)
Ln Employment (affiliate)
Instruments
Competitor Cut
Competitor Cut x Ln Assets
0.901***
(0.048)
0.011*
(0.006)
-0.123
(0.227)
1.024***
(0.123)
Competitor Cut x Ln Employment
Instrumented
PTA Tariff Cuts (US)
PTA Tariff Cuts (US) x Ln Assets
(4)
(5)
First Stage
PTA Tariff
PTA Tariff
Cuts (US) x Vertical Sales
Cuts (US)
Ln Empl.
0.002
0.013
0.309*
(0.009)
(0.046)
(0.167)
-0.001
-0.007
0.318*
(0.002)
(0.012)
(0.183)
-0.001
-0.005
0.146
(0.005)
(0.023)
(0.235)
-0.001
-0.003
0.183*
(0.002)
(0.008)
(0.094)
-0.003
-0.014
0.034
(0.002)
(0.012)
(0.029)
-0.00004
(0.00005)
0.0002
(0.0002)
0.907***
(0.072)
-0.294***
(0.065)
0.020***
(0.007)
1.077***
(0.106)
-2.914***
(0.349)
0.285***
(0.048)
68444
150
0.181
0.475***
(0.022)
-1.993***
(0.664)
PTA Tariff Cuts (US) x Ln employment
Observations
68444
68444
Countries
150
150
R-squared
0.896
0.889
Kleibergen-Paap Wald rk F statistics
47.27***
Kleibergen-Paap rk LM statistics
3.96**
Anderson-Rubin Wald test
43.56***
All models include benchmark year and industry fixed effects
(6)
2nd Stage
68444
68444
150
150
0.896
0.891
51.71***
4.07**
10.17***
0.370**
(0.147)
68444
150
0.197
Australia
Bahrain
Canada
Chile
Colombia
Costa Rica
Dominican Republic
El Salvador
Guatemala
Honduras
Jordan
Mexico
Morocco
Nicaragua
Oman
Panama
Peru
Singapore
South Korea
Vietnam
Average change
HHI Sales
Change
Percentage
-0.009
-2.6%
0.116
13.9%
0.075
39.2%
-0.022
-3.7%
0.077
13.3%
0.041
6.2%
0.054
7.6%
-0.037
-4.6%
-0.049
-7.0%
0.145
21.2%
-0.109
-10.9%
0.005
1.5%
0.142
17.6%
0.016
1.6%
0.225
29.0%
0.027
4.3%
-0.042
-6.3%
0.023
6.0%
-0.035
-6.8%
0.028
3.5%
0.034
6.2%
HHI Employment
Change
Percentage
0.103
28.7%
0.020
2.3%
0.063
27.8%
-0.009
-1.4%
0.052
9.1%
0.038
5.7%
0.065
9.3%
-0.021
-2.8%
-0.019
-2.7%
0.165
24.6%
0.079
9.8%
0.073
23.5%
0.095
11.8%
0.083
9.1%
0.282
39.3%
0.136
24.1%
0.017
2.7%
0.089
24.7%
0.013
2.5%
-0.062
-7.7%
0.063
12.0%
Ln(Employment)
Change
Percentage
-0.359
-8.6%
0.017
0.6%
-0.080
-1.9%
0.083
2.0%
0.135
3.0%
-0.084
-1.8%
-0.531
-10.4%
0.207
4.9%
-0.985
-23.0%
0.532
10.7%
2.879
100.7%
-0.830
-14.8%
0.836
18.2%
0.339
7.9%
1.582
43.0%
-0.417
-15.0%
-0.521
-12.9%
-0.310
-7.7%
0.155
3.5%
1.273
43.8%
0.196
7.1%
Ln(Assets - PPE)
Change
Percentage
-0.337
-4.4%
0.092
1.7%
-0.593
-7.2%
0.235
3.1%
0.496
6.2%
0.693
8.9%
0.207
2.5%
0.602
8.2%
0.160
2.2%
0.658
8.3%
3.899
66.1%
-0.288
-3.5%
1.089
13.0%
0.563
7.4%
2.118
25.0%
-0.212
-3.5%
0.061
0.8%
-0.454
-6.1%
0.315
3.9%
1.624
20.5%
0.546
7.7%
Table 7: Change in HH Indices of Employment and Sales Concentration Before and After PTA
All
Countryyear
Unit of analysis
Sample
719
169
0.630
Country,
Year
(1)
0.096***
(0.030)
-0.106*
(0.063)
0.120
(0.115)
Observations
Countries
R-squared
Fixed effects
BIT with US
WTO
GATT Only
Ln (population)
Ln (GDP/capita)
PTA with US
Dependent Variable
Herfindahl-Hirschman Index Employment
(2)
(3)
(4)
(5)
0.100*** 0.030*** 0.028**
0.037**
(0.030)
(0.011)
(0.011)
(0.014)
-0.112* -0.121*** -0.133*** -0.183***
(0.060)
(0.040)
(0.033)
(0.042)
0.144
0.002
0.039
0.122*
(0.121)
(0.051)
(0.052)
(0.065)
-0.006
-0.038
(0.047)
(0.024)
0.042
0.039
(0.054)
(0.024)
-0.029
-0.070***
(0.041)
(0.020)
710
19770
19597
8430
166
169
166
138
0.631
0.311
0.317
0.454
Country, Industry, Industry, Industry,
Year
Country, Country, Country,
Year
Year
Year
Country- Industry- Industry- Industryyear
country- country- countryyear
year
year
All
All
All
Positive
vertical
sales
(6)
(7)
0.035**
0.068**
(0.015)
(0.030)
-0.204*** -0.072**
(0.043)
(0.034)
0.160**
0.149**
(0.076)
(0.072)
-0.010
(0.048)
0.076**
(0.035)
-0.085***
(0.027)
8356
719
136
169
0.464
0.835
Industry, Country,
Country,
Year
Year
Industry- Countrycountryyear
year
Positive
All
vertical
sales
Herfindahl-Hirschman Index Sales
(8)
(9)
(10)
(11)
0.067**
0.029**
0.027**
0.036**
(0.029)
(0.013)
(0.013)
(0.016)
-0.070** -0.136*** -0.145*** -0.158***
(0.032)
(0.048)
(0.040)
(0.046)
0.171**
0.080
0.110**
0.099
(0.072)
(0.052)
(0.055)
(0.070)
-0.065**
-0.034
(0.031)
(0.027)
-0.020
0.036*
(0.030)
(0.021)
0.014
-0.031*
(0.027)
(0.018)
710
19770
19597
8430
166
169
166
138
0.834
0.442
0.445
0.491
Country, Industry, Industry, Industry,
Year
Country, Country, Country,
Year
Year
Year
Country- Industry- Industry- Industryyear
country- country- countryyear
year
year
All
All
All
Positive
vertical
sales
Table 8: Concentration of Employment and Sales: Regression Analysis
(12)
0.036**
(0.018)
-0.182***
(0.045)
0.138*
(0.079)
-0.005
(0.052)
0.112***
(0.037)
-0.046**
(0.023)
8356
136
0.494
Industry,
Country,
Year
Industrycountryyear
Positive
vertical
sales
Appendix A
Data on Multinational Activity
Our study relies on confidential data on the activity of U.S. multinational firms abroad collected by
the U.S. Bureau of Economic Analysis (BEA). These data were accessed on site at BEA through
a special-sworn-status research arrangement with the Bureau. The data include detailed financial
and operating information at the level of the foreign affiliate and the level of the U.S. parent.
The affiliate sales information used in this study was extracted from the BEA’s data files for
each benchmark survey year, then merged with the parent firm information to create a complete
parent-affiliate-year panel. The sample includes all majority-owned affiliates; we exclude values:
1) that were imputed based on previous survey responses; 2) from firms in the financial sector; 3)
corresponding to form rejected by the BEA due to inaccuracies.
The analysis relies primarily on affiliate-level sales data from the quinquennial Benchmark
Surveys. The benchmark years are 1989, 1994, 1999, 2004, and 2009. The data are based on
affiliate-level sales data, which are disaggregated in the survey according to the destination of the
buyer. We characterize horizontal sales as those to the host country; vertical sales are sales to the
U.S., export-platform sales are directed to other countries.
In Tables 3 and 4 we interact preferential trade agreements with a measure of firm productivity. Following Bilir (2014), productivity is measured at the parent-firm level based on a simple
Solow residual, which we calculate for each parent firm-year by regressing firm-level log of value
added on firm-level physical assets, employment, and industry and year dummies. The residuals of
this regression are our time-varying measures of firm productivity.
Appendix B
Data on Tari↵s
Data on MFN and preferential tari↵s come from WITS (2014) and rely on HS trade categorization.
U.S. HS codes are based on the Harmonized System established by the World Customs Organization
(WCO). The WCO assigns 6-digit codes for general categories, and countries adopting the system
then define their own codes to capture commodities at more detailed levels. In the United States,
the most detailed level of disaggregation is ten digits by Pierce and Schott (2010). Since US HS
system is rooted on WCO 6-digit HS, we construct concordance between 6-digit HS combined and
4-digit NAICS from 1996 to 2009 by two steps. First, based on concordance between 10-digit US HS
and 7-digit NAICS provided by Pierce and Schott (2010), we construct the concordance between
6-digit US HS and 4-digit NACIS. Second, we use WITS concordances between HS combined and
other HS systems (H1, H2 and H3) to match 6-digit US HS codes over time.
Appendix C
Additional Robustness Checks
We implement other robustness checks to further assess the validity of our main findings. First,
we show that our results are robust to entropy balancing, which allows us to better account for
firm-level characteristics. Second, we implement a placebo test using local sales of affiliates, i.e.
horizontal FDI, as outcome variable. Finally, we show that our findings are robust to additional
fixed e↵ects and varying transformations of the main variables. Below we discuss these robustness
checks in detail.
C.1
Entropy Balancing
It might be the case that firms in countries with a US PTA di↵er significantly from firms in countries
without a US PTA. For instance, it is plausible that the US picks trade partners that have productive
firms to reap the largest benefit from preferential integration. In econometric terms, observations
are unbalanced with respect to the treatment variable PTA. This poses a threat to our conclusions
if these observed di↵erences are also correlated with di↵erences in vertical sales, or if they proxy for
unobserved di↵erences that might drive the correlation. To overcome this issue, we rely on entropy
balancing Hainmueller (2012). This technique is similar to propensity matching, but it has the
welcoming feature that unbalanced observations are not dropped from the analysis.40
Specifically, by using entropy balancing observations are re-weighted with respect to the
treatment (i.e. PTA) so that all the relevant covariates are balanced (i.e. they have the same mean).
In econometric terms, entropy balancing reweights the observations to statistically generate a region
of common support where firms in countries with a US PTA and firms in countries without a US
PTA are comparable on structural covariates. Entropy balancing does this by directly incorporating
covariate balance into the weight function that is applied to the sample units. The net result is
that we can compare firm in countries with a US PTA to a comparable counterfactual of firms in
countries without a US PTA.
Table A.3 in the appendix shows the means of firm-level covariates before and after running
‘ebalance’. By using entropy balancing the di↵erence in means between treatment and control firms
is substantially reduced and is never statistically significantly di↵erent from zero. Then we re-run
our main models using the weights obtained from entropy balancing. Our main results reproduced
in Table A.4 remain unchanged, adding plausibility to initial our econometric strategy: Preferential
liberalization increases vertical sales of the larger and more productive affiliates. GATT and WTO
e↵ects are stronger in export-platform sales.
C.2
PTAs and Horizontal FDI
The e↵ect of both PTAs and the WTO on horizontal FDI should be weak. Since horizontal FDI
is established to service local markets in the presence of trade costs, a reduction on tari↵s creates
incentives to replace local sales with foreign exports. Firms could concentrate production at home,
allowing them to reap the benefits from establishment of specific economies of scale. If this is the
case, we should see a reduction of horizontal FDI after preferential and multilateral liberalization.
40 We
use the command ‘ebalance’ in Stata 12. We adjust the covariates using the first moment,
i.e. we set target equal to one.
Tari↵ cuts implemented by the US, on the other hand, should have no e↵ect on horizontal FDI. 41
Horizontal FDI can be a↵ected by competition-enhancing regulations. For instance, if PTAs open
the services sector and improve regulations of government procurement, US affiliates serving the
local market could still benefit from this more business-friendly environment. If that is the case, US
PTAs and the WTO could have a positive impact on local sales. We explore these relationships in
Table A.7: we find that there is no association between Preferential and MFN tari↵ cuts o↵ered by
the US and horizontal sales as expected (Columns 1-5); figure A.6 presents the results from column
4. Moving to the partner country, neither preferential tari↵s nor MFN tari↵ cuts are robustly
associated with horizontal sales (Columns 6-8). We do, however, find that preferential cuts are
associated with lower local sales of the smaller affiliates, and increases the sales of the largest
affiliates.
C.3
Other Specifications
In this section, we report a number of robustness tests. We show that our results hold if we include
a battery of fixed e↵ects to mitigate concerns on an omitted variable problem. We include countryand industry-specific time trends, which capture slow moving unobserved confounders a↵ecting
affiliate sales that vary by partner country or industry. We also introduce parent firm-level fixed
e↵ects to control for unobserved heterogeneity at the parent firm level. We added country-year
dummy variables, which absorb all country characteristics that a↵ect vertical sales that are timevarying (as well as those that are time-invariant). These could include, but are not limited to,
economic performance, tax rates, capital controls, changes in the bilateral relationship with the
U.S. and a host of other policy and institutional features, along with shocks such as military
conflict or financial crises.
Further, we augment the model specification with a battery of political and economic control
variables in Cit , as recommended by Büthe and Milner (2014). The economic variables include:
economic performance (GDP growth); the host country’s market size (the log of host country
population); trade openness (import and export as percentage of GDP). As political controls, we
include political constraints (Henisz, 2000) and political stability (CITE). The coefficients on the
independent variables of interest remain the same in substantive and statistical terms. The control
variables are (5 year) lagged average values, inclusive of the benchmark year j.
Importantly, our findings are not sensitive to the operationalization of our key variables,
and di↵erent transformations of the dependent variable. Our findings hold if we use tari↵ cuts
instead of percentage change in tari↵ cuts. In additional specifications, we modeled the share of
sales directed to the U.S. in total affiliate sales. Consistent with the results reported here, we find
that high productivity firms direct a greater share of sales to the U.S. during years in which the
host country participates in a PTA with the U.S..42
41 Given
that tari↵ cuts are orthogonal to horizontal FDI, this can be considered a placebo test.
That is, a null relationship between US tari↵ cuts and horizontal sales would provide further
evidence on the validity of our theory linking the e↵ect of economic agreements to trade costs
reduction.
42 All
results are available from authors upon request.
Appendix D
Additional Figures and Tables
2.5
3
3.5
4
4.5
Figure A.1: Mean tari↵ reductions in US PTAs by contract intensity of product
Low Contract Intensive Industries
US Tariff Cut
High Contract Intensive Industries
90% CI
Note: The figure displays the di↵erences between MNF tari↵s prior to the formation of PTAs and
preferential tari↵s (PRF) after PTA is in force, by type contract intensity. The measure of contract
intensive product comes Nunn (2007). The whiskers represent 90% confidence intervals.
Figure A.2: MFN Tari↵ Reductions in pre- and post-WTO Accesion
Armenia
Cambodia
Cape Verde
China
Ecuador
Estonia
Georgia
Kyrgyz Republic
Latvia
Lithuania
Macedonia, FYR
Moldova
Nepal
Oman
Saudi Arabia
Taiwan, China
Ukraine
Vietnam
0
10
20
30
Tariff Cut (MFNpre-MFNpost)
40
Note: The figure displays the distribution of MFN tari↵ cuts after accession to WTO. Data come
from WITS (2014) and are at the HS 6-digit tari↵ line.
0
.2
.4
.6
.8
Figure A.3: Mean MFN Tari↵ Reductions after WTO Accession by Product Use
Final & Mixed Goods
MFN tariff cut
Intermediates
90% CI
Note: The figure displays the di↵erences between MNF tari↵s prior after a WTO Accession by type
of product classified as intermediate or consumption and mixed use. The categorization of products
come from Francois and Pindyuk (2012) and Bekkers et al. (2012). The whiskers represent 90%
confidence intervals.
0
-1
0
Marginal effects
1
2
5
10
Ln Employment (% observations)
3
15
Figure A.4: Marginal E↵ect of US MFN Tari↵ Cuts on Vertical Sales by Firm Size
0
1
2
3
4
5
6
7
8
Ln Employment (affiliate)
9
10
11
12
Note: Marginal e↵ects (and 90% confidence intervals) of US MFN Cuts based on Results from
Column 8 in Table 3.
-.5
0
.5
1
Figure A.5: Correlation between Residuals of the Second Stage and Preferential Tari↵ Cut (proportional) Implemented by US Competitors. ⇢ = 0
0
.2
.4
.6
.8
Preferential Tariff Cut Implemented by US Competitors -- Instrument
Residuals
Fitted values
Note: Residuals are estimated from Model 6 in Table 6.
1
0
-1
-.5
Marginal effect
0
.5
5
10
Ln Employment (% observations)
1
15
Figure A.6: Marginal E↵ect of US PTA Tari↵ Cuts on Horizontal Sales
0
1
2
3
4
5
6
7
8
Ln Employment (affiliate)
9
10
11
12
Note: Marginal e↵ects (and 90% confidence intervals) of US PTA Cuts based on Results from
Column 4 in Table A.7.
0
-6
-4
Marginal effects
-2
0
2
4
6
8
Ln Employment (% observations)
2
10
Figure A.7: E↵ect of China’s WTO Accession on Vertical Sales
0
1
2
3
4
5
6
7
8
Ln Employment (affiliate)
9
10
11
12
Note: Marginal e↵ects (and 90% confidence intervals) of US MFN Cuts on affiliate sales to the US.
0
-10
-5
Marginal effect
0
5
2
4
6
8
Ln Employment (% observations)
10
10
Figure A.8: E↵ect of China’s WTO Accession on Export-Platform Sales
0
1
2
3
4
5
6
7
8
Ln Employment (affiliate)
9
10
11
12
Note: Marginal e↵ects (and 90% confidence intervals) of China MFN Cuts on US affiliate sales to
third markets.
Table A.1: Design of U.S. PTAs
PTA
Year
Services
Investment
IPRs
Competition
Government
Procurement
Depth
US-Australia
2004
yes
yes
yes
yes
yes
3.19
US-Bahrain
2004
yes
yes
yes
no
yes
3.01
US-CAFTA-DR
2004
yes
yes
yes
no
yes
3.13
US-Canada
1988
yes
yes
no
no
yes
1.90
US-Canada
1992
yes
yes
yes
yes
yes
2.74
US-Chile
2003
yes
yes
yes
no
yes
2.90
US-Colombia
2006
yes
yes
yes
yes
yes
3.40
US-Jordan
2000
yes
yes
yes
no
yes
2.59
US-Korea
2007
yes
yes
yes
yes
yes
3.26
US-Mexico
1992
yes
yes
yes
yes
yes
2.74
US-Morocco
2004
yes
yes
yes
no
yes
3.19
US-Oman
2006
yes
yes
yes
no
yes
3.19
US-Panama
2007
yes
yes
yes
No
yes
3.19
US-Peru
2006
yes
yes
yes
yes
yes
3.33
US-Singapore
2003
yes
yes
yes
yes
yes
3.01
US-Vietnam
2000
yes
yes
yes
no
no
2.69
Note:“Yes” means that a specific section regulating each trade-related issue is included in the treaty.
Depth is built using a latent trait analysis on 48 dummy variables related to trade-related issues
(Dür et al., 2014).
Table A.2: PTAs used to build our instrument used for IV Regressions
PTA Instrumented
Signature
Ratification
PTA used as instrument
Signature
Ratification
US-Australia
US-Chile
US-South Korea
US-Costa Rica
US-Peru
US-Singapore
18 May 2004
6 June 2003
30 June 2007*
5 August 2004
12 April 2006**
6 May 2003
1 January 2005
1 January 2004
15 March 2012
1 January 2009
1 February 2009
1 January 2004
Thailand-Australia
South Korea-Chile
India-South Korea
Canada-Costa Rica
Canada-Peru
Japan-Singapore
5 July 2004
15 February 2003
7 August 2009
23 April 2001
29 May 2008
13 January 2002
1 January 2005
1 April 2004
1 January 2010
1 November 2002
1 August 2009
30 November 2002
* Amended on December 3, 2010.
** Ratified with amendments on February 1, 2009.
Table A.3: Balance of covariates before and after weighting
Original sample
Variable
Ln Employment
Ln Assets (PPE)
Exporter
Positive Sales to US
Agr. and Mining
Manufacturing
Telecomm.
Wholesale
Services
After entropy weighting
mean
4.293
7.713
0.483
0.960
0.040
0.404
0.031
0.218
0.299
Treatment
variance
5.726
13.450
0.250
0.038
0.039
0.241
0.030
0.171
0.210
skewness
-0.418
-0.842
0.068
-4.698
4.675
0.390
5.436
1.363
0.878
mean
3.996
7.381
0.478
0.947
0.045
0.375
0.021
0.250
0.305
Control
variance
4.916
12.690
0.250
0.050
0.043
0.234
0.020
0.188
0.212
skewness
-0.408
-0.807
0.090
-4.009
4.393
0.515
6.704
1.154
0.850
Variable
Ln Employment
Ln Assets (PPE)
Exporter
Positive Sales to US
Agr. and Mining
Manufacturing
Telecomm.
Wholesale
Services
mean
4.293
7.713
0.483
0.960
0.040
0.404
0.031
0.218
0.299
Treatment
variance
5.726
13.450
0.250
0.038
0.039
0.241
0.030
0.171
0.210
skewness
-0.418
-0.842
0.068
-4.698
4.675
0.390
5.436
1.363
0.878
mean
4.293
7.712
0.483
0.960
0.040
0.404
0.031
0.218
0.299
Control
variance
4.681
11.790
0.250
0.038
0.039
0.241
0.030
0.171
0.210
skewness
-0.493
-0.901
0.068
-4.697
4.675
0.391
5.436
1.363
0.878
Table A.4: International Agreements and U.S. MNC Activities (Matched sample)
Ln GDP/capita
GATT only
WTO
BIT with US
Ln Cumulative PTAs (partner)
PTA with US
Productivity (headquarter)
PTA with US x Productivity
Ln Assets (PPE, affiliate)
PTA with US x Ln Assets
(1)
(2)
(3)
(4)
Vertical
Sales
Vertical
Sales
Vertical
Sales
Vertical
Sales
0.161
(0.237)
0.348*
(0.190)
0.356
(0.258)
0.344*
(0.184)
0.060
(0.039)
0.045
(0.103)
0.102
(0.228)
0.351**
(0.172)
0.377*
(0.227)
0.369**
(0.180)
0.083*
(0.046)
0.029
(0.109)
-0.003
(0.041)
0.157**
(0.061)
0.195
(0.204)
0.242
(0.167)
0.275
(0.257)
0.264*
(0.149)
0.030
(0.036)
-0.967***
(0.252)
-0.031
(0.266)
0.284
(0.178)
0.289
(0.238)
0.240
(0.165)
0.053
(0.052)
-0.455***
(0.157)
(5)
ExportPlatform
Sales
-0.828*
(0.475)
0.638**
(0.265)
0.858**
(0.369)
0.168
(0.278)
0.325***
(0.118)
0.479***
(0.130)
(6)
ExportPlatform
Sales
-1.018**
(0.466)
0.472*
(0.257)
0.793**
(0.375)
0.090
(0.311)
0.306**
(0.127)
1.901***
(0.310)
82946
0.267
165
0.654***
(0.037)
-0.322***
(0.053)
82946
0.309
165
0.218***
(0.021)
0.130***
(0.042)
Ln Employment (affiliate)
PTA with US x Ln Employment
Observations
82946
74394
82946
R-squared
0.261
0.267
0.302
Countries
165
163
165
All models include benchmark year and industry fixed effects
0.435***
(0.033)
0.127***
(0.048)
82946
0.310
165
Note: The dependent variable is the log of affiliate sales to the U.S.. Robust standard errors
adjusted for country-level clustering. All models include country, year, and industry fixed e↵ects.
*** p < 0.01, ** p < 0.05, * p < 0.10.
0.490*
(0.282)
-2.063***
(0.389)
0.683***
(0.187)
-2.198
(6.432)
0.477***
(0.022)
77620
0.108
152
69014
0.120
150
Industry,
Industry,
country, year country, year
-7.730
(6.180)
0.124
(0.262)
0.001
(0.011)
-0.077
(0.392)
0.016
(0.196)
0.027
(0.025)
0.004
(0.003)
0.211*
(0.116)
0.250
(0.181)
0.001
(0.011)
0.252
(0.390)
-0.049
(0.212)
-0.000
(0.015)
0.002
(0.002)
0.299**
(0.149)
0.229
(0.210)
0.171
(0.111)
0.019
(0.033)
-0.480***
(0.135)
0.453***
(0.022)
0.144***
(0.045)
0.054
(0.146)
-0.016
(0.048)
Vertical
Sales
Vertical
Sales
(4)
0.475***
(0.022)
0.003
(0.336)
-0.037
(0.090)
-0.296
(0.308)
-0.009
(0.010)
0.215
(1.957)
0.489**
(0.233)
0.029
(0.021)
0.002
(0.004)
0.110*
(0.064)
Vertical
Sales
(5)
0.277*
(0.162)
0.008
(0.010)
0.304
(0.385)
0.028
(0.200)
0.007
(0.013)
0.002
(0.002)
0.340*
(0.182)
0.233
(0.198)
0.187
(0.121)
-0.028
(0.035)
-0.497***
(0.125)
0.452***
(0.021)
0.147***
(0.048)
Vertical
Sales
77620
0.110
152
Country
Industry,
country
69014
0.123
150
Country
Industry,
country
77620
0.117
152
Industry
Industry,
country
0.325
(0.228)
-2.853***
(0.510)
0.656***
(0.199)
-147.900** -239.313*** -55.668***
(70.314)
(78.231)
(9.424)
-0.104
(0.274)
-0.010
(0.010)
-1.635
(1.434)
0.339*
(0.189)
0.021
(0.018)
0.002
(0.003)
0.326
(0.214)
0.210
(0.222)
0.097
(0.319)
-0.028
(0.079)
-0.806***
(0.117)
0.451***
(0.022)
0.150***
(0.044)
Vertical
Sales
(3)
(7)
-0.090
(0.287)
-0.009
(0.011)
-1.545
(1.433)
0.335*
(0.192)
0.017
(0.016)
0.001
(0.003)
0.292
(0.196)
0.217
(0.201)
0.094
(0.307)
-0.021
(0.081)
-0.775***
(0.116)
0.451***
(0.022)
0.153***
(0.047)
Vertical
Sales
(8)
0.476***
(0.022)
0.013
(0.332)
-0.031
(0.093)
-0.279
(0.308)
-0.007
(0.010)
0.374
(1.781)
0.442*
(0.244)
0.027
(0.019)
0.001
(0.004)
0.110
(0.071)
Vertical
Sales
(9)
-0.180
(0.257)
-0.011
(0.010)
-1.926
(1.268)
0.478***
(0.172)
0.017
(0.013)
0.002
(0.003)
0.317*
(0.175)
0.276
(0.178)
0.056
(0.268)
-0.004
(0.078)
-0.765***
(0.178)
0.479***
(0.023)
0.142**
(0.063)
Vertical
Sales
(10)
0.502***
(0.022)
-0.056
(0.297)
-0.002
(0.095)
-0.358
(0.289)
-0.007
(0.010)
0.382
(1.682)
0.592**
(0.242)
0.028
(0.018)
0.000
(0.003)
0.056
(0.060)
Vertical
Sales
69014
0.128
150
Industry
Industry,
country
Country,
Industry
77620
0.120
152
Industry,
country
Country,
Industry
69014
0.132
150
Industry,
country
Country,
Industry
77620
0.214
152
Country,
Industry
69014
0.222
150
Industry,
Industry,
HQ, country HQ, country
0.731***
0.523**
0.205
(0.271)
(0.241)
(0.251)
-2.107***
-2.875***
-2.560***
(0.405)
(0.597)
(0.477)
0.695***
0.667***
0.589***
(0.199)
(0.211)
(0.195)
-73.309*** -214.685*** -296.312*** -295.471*** -393.935***
(14.461)
(81.102)
(80.592)
(101.962)
(107.950)
0.475***
(0.021)
0.159
(0.155)
-0.061
(0.040)
0.119
(0.254)
0.009
(0.010)
0.130
(0.362)
0.175
(0.196)
0.035
(0.022)
0.003
(0.002)
0.153**
(0.074)
Vertical
Sales
(6)
Table A.5: Robustness Tests: Vertical Sales
(11)
82946
0.212
165
Industry
-4.892***
(0.272)
Industry,
Country/
year
0.457***
(0.023)
0.149***
(0.041)
Vertical
Sales
(12)
73736
0.194
163
Industry
0.046
(0.221)
-3.735***
(0.959)
0.637***
(0.183)
0.340***
(0.022)
Industry,
Country/
year
0.480***
(0.022)
Vertical
Sales
Note: The dependent variable is the log of affiliate sales to the U.S.. Robust standard errors adjusted for country-level clustering. ***
p < 0.01, ** p < 0.05, * p < 0.10.
Observations
R-squared
Countries
Trends
Fixed effects
Constant
Tariff Cuts (US) x Ln Employment
PTA Tariff Cuts (US)
WTO Cut (US)
PTA x Ln Employment
Ln Employment (affiliate)
PTA with US
Ln Cumulative PTAs (partner)
BIT with US
WTO member (partner)
GATT only
Trade/GDP
Political Instability
Political Constraints
Ln Population
GDP growth
Ln GDP/capita
(2)
(1)
67152
0.149
151
77605
0.152
152
Industry,
Industry,
Country, Year Country, Year
-30.058***
(8.615)
-1.075***
(0.346)
0.321***
(0.080)
-34.611***
(9.330)
0.577***
(0.044)
0.185
(0.275)
0.187**
(0.073)
(2)
Export
Platform
Sales
-0.256
(0.257)
-0.007
(0.017)
2.375***
(0.644)
0.056
(0.428)
0.034
(0.030)
0.004
(0.005)
0.128
(0.178)
0.569***
(0.044)
-0.179
(0.262)
-0.199**
(0.098)
(4)
Export
Platform
Sales
-0.395
(0.819)
-0.010
(0.033)
-4.803
(3.486)
0.346
(0.558)
0.137
(0.093)
0.017**
(0.008)
1.880***
(0.181)
(5)
Export
Platform
Sales
-0.687
(0.716)
-0.016
(0.028)
-4.416
(3.971)
0.537
(0.484)
0.146
(0.093)
0.007
(0.007)
1.356*
(0.740)
-0.553
(0.781)
-0.175
(0.245)
-0.185**
(0.093)
-1.557***
(0.523)
0.630***
(0.037)
0.280***
(0.098)
0.572***
(0.044)
-0.170
(0.242)
-0.163*
(0.085)
(6)
Export
Platform
Sales
-0.498
(0.649)
-0.015
(0.027)
-3.599
(2.906)
0.439
(0.488)
0.124
(0.080)
0.013*
(0.007)
1.730***
(0.177)
(7)
Export
Platform
Sales
-0.687
(0.716)
-0.016
(0.028)
-4.416
(3.971)
0.537
(0.484)
0.146
(0.093)
0.007
(0.007)
1.356*
(0.740)
-0.553
(0.781)
-0.175
(0.245)
-0.185**
(0.093)
-1.557***
(0.523)
0.630***
(0.037)
0.280***
(0.098)
0.572***
(0.044)
-0.170
(0.242)
-0.163*
(0.085)
(8)
Export
Platform
Sales
-0.498
(0.649)
-0.015
(0.027)
-3.599
(2.906)
0.439
(0.488)
0.124
(0.080)
0.013*
(0.007)
1.730***
(0.177)
67152
0.143
151
Country
Industry,
Country
77605
0.148
152
Country
Industry,
Country
Industry,
Country
Country,
Industry
67152
0.157
151
Industry,
Country
Country,
Industry
77605
0.161
152
Industry,
Country
Country,
Industry
67152
0.157
151
Industry,
Country
Country,
Industry
77605
0.161
152
-0.707*
-0.813**
-0.813**
(0.378)
(0.325)
(0.325)
0.324***
0.303***
0.303***
(0.081)
(0.077)
(0.077)
-913.572*** -775.154*** -831.243*** -773.873*** -831.244*** -773.835***
(245.596)
(140.017)
(234.908)
(129.084)
(234.907)
(129.039)
(3)
Export
Platform
Sales
-0.734
(0.906)
-0.011
(0.037)
-5.103
(4.642)
0.453
(0.554)
0.172
(0.111)
0.011
(0.008)
1.425*
(0.822)
-0.628
(0.853)
-0.226
(0.270)
-0.230**
(0.104)
-1.389**
(0.580)
0.629***
(0.037)
0.289***
(0.095)
-1.141**
(0.438)
0.376***
(0.092)
-461.271***
(147.413)
0.631***
(0.042)
-0.092
(0.270)
-0.139*
(0.077)
(10)
Export
Platform
Sales
-0.403
(0.833)
-0.009
(0.033)
-4.863
(3.077)
0.686
(0.483)
0.129
(0.084)
0.014*
(0.008)
1.845***
(0.150)
1.494***
(0.033)
-1.833*
(1.055)
0.631***
(0.037)
0.275***
(0.092)
(11)
Export
Platform
Sales
-1.369***
(0.269)
0.335***
(0.075)
-3.613***
(1.339)
0.574***
(0.043)
(12)
Export
Platform
Sales
HQ, Industry, HQ, Industry, Industry,
Industry,
Country
Country
Country/year Country/year
Country,
Country,
Industry
Industry
Industry
Industry
67152
77605
71573
82931
0.269
0.260
0.267
0.252
151
152
164
165
-522.231**
(234.285)
(9)
Export
Platform
Sales
-0.554
(0.928)
-0.010
(0.038)
-6.131
(4.287)
0.712
(0.504)
0.156
(0.097)
0.007
(0.007)
1.596**
(0.739)
-0.398
(0.781)
-0.153
(0.281)
-0.152*
(0.085)
-0.851
(0.682)
0.686***
(0.036)
0.168
(0.124)
Note: The dependent variable is the log of affiliate sales to third countries. Robust standard errors adjusted for country-level clustering.
*** p < 0.01, ** p < 0.05, * p < 0.10.
Observations
R-squared
Countries
Trends
Fixed effects
Constant
WTO Cuts (partner) x Ln Employment
WTO Cut (Partner)
PTA Cuts (partner) x Ln Employment
Ln Employment (affiliate)
PTA Cuts (Partner)
Ln Cumulative PTAs (partner)
BIT with US
WTO member (partner)
GATT only
Trade/GDP
Political Instability
Political Constraints
Ln Population
GDP growth
Ln GDP/capita
(1)
Export
Platform
Sales
-0.066
(0.226)
-0.012
(0.016)
1.951***
(0.570)
0.091
(0.417)
0.016
(0.026)
0.0001
(0.004)
0.329*
(0.192)
0.224
(0.246)
0.405
(0.265)
0.198***
(0.072)
-0.659*
(0.357)
0.637***
(0.037)
0.315***
(0.107)
Table A.6: Robustness Tests: Export Platform Sales
Table A.7: International Economic Agreements and U.S. MNC Affiliate Horizontal Sales
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
Horizontal Horizontal Horizontal Horizontal Horizontal Horizontal Horizontal Horizontal
Sales
Sales
Sales
Sales
Sales
Sales
Sales
Sales
Ln GDP/capita
GATT only
WTO member (partner)
BIT with US
Ln Cumulative PTAs (partner)
PTA with US
0.755**
(0.359)
0.541**
(0.240)
0.121
(0.191)
0.112
(0.210)
0.052
(0.071)
0.100
(0.176)
Ln Employment (affiliate)
0.613*
(0.323)
0.400**
(0.196)
0.036
(0.165)
0.059
(0.234)
0.017
(0.065)
0.055
(0.182)
0.644***
(0.031)
Tariff Cuts (US)
0.826***
(0.245)
0.420***
(0.149)
0.829***
(0.246)
0.421***
(0.149)
0.617**
(0.308)
0.381**
(0.176)
0.786***
(0.262)
0.418***
(0.153)
0.785***
(0.262)
0.419***
(0.153)
0.779***
(0.260)
0.410***
(0.152)
0.120
(0.245)
-0.018
(0.055)
0.120
(0.246)
-0.017
(0.055)
0.062
(0.228)
0.013
(0.060)
0.120
(0.244)
-0.026
(0.055)
0.120
(0.244)
-0.026
(0.055)
0.119
(0.244)
-0.024
(0.054)
0.668***
(0.023)
-0.185
(0.161)
0.666***
(0.024)
-0.449
(0.284)
0.054
(0.047)
0.644***
(0.032)
0.665***
(0.024)
0.664***
(0.024)
0.668***
(0.023)
-0.105
(0.276)
0.156
(0.231)
-0.105
(0.276)
-1.308**
(0.539)
0.291***
(0.086)
3.019***
(0.393)
0.158
(0.231)
Tariff Cuts (US) x Ln Employment
0.160
(0.225)
WTO Cut (US)
WTO Cut (Partner)
Tariff Cuts (Partner)
Tariff Cuts (Partner) x Ln Employment
WTO Cuts (Partner) x Ln Employment
Constant
Observations
R-squared
Countries
-0.139
(2.574)
82946
0.0659
165
-1.742
(2.339)
82946
0.193
165
-3.253*
(1.677)
73743
0.198
163
-3.266*
(1.681)
73743
0.198
163
-1.767
(2.224)
82939
0.193
165
-2.960
(1.802)
71558
0.199
164
-2.947
(1.799)
71558
0.199
164
-0.546***
(0.066)
-2.917
(1.790)
71558
0.199
164
Note: The dependent variable is the log of total affiliate sales to the host country based on affiliatelevel data from the BEA. Robust standard errors adjusted for country-level clustering. *** p < 0.01,
** p < 0.05, * p < 0.10.
-3.803*** -6.398***
(1.285)
(1.924)
0.281***
(0.050)
0.336*
(0.193)
0.147***
(0.036)
0.459**
(0.183)
-1.860**
(0.773)
0.339***
(0.047)
(5)
(6)
-0.479
(3.411)
2048
0.316
0.259
(1.169)
0.424***
(0.054)
-0.095
(3.355)
2048
0.320
-4.000**
(1.836)
0.692***
(0.245)
0.296***
(0.057)
1.127***
(0.172)
-0.084
(3.288)
2048
0.332
-10.670***
(1.945)
0.190***
(0.040)
Export
Export
Export
Platform Sales Platform Sales Platform Sales
(4)
Note: The dependent variable is the log of sales to the US and to third countries by affiliates of US MNCs operating in China. Robust
standard errors adjusted for country-level clustering. *** p < 0.01, ** p < 0.05, * p < 0.10.
Constant
-3.864*** -3.678*** -3.589***
(0.744)
(0.727)
(0.733)
2055
2055
2055
Observations
0.290
0.291
0.284
R-squared
Models include industry and year fixed effects
WTO Cuts (China) x Ln Assets, PPE
WTO Cuts (China) x Ln Employment
WTO Cut (China)
WTO Cuts (US) x Ln Assets, PPE
Ln Assets (PPE, affiliate)
WTO Cuts (US) x Ln Employment
Ln Employment (affiliate)
WTO Cut (US)
Vertical
Sales
Vertical
Sales
Vertical
Sales
(3)
(2)
(1)
Table A.8: U.S. MNC Activities in China