false friends? empirical evidence on trade policy ... · pdf filefalse friends? empirical...

False Friends?

Empirical Evidence on Trade Policy Substitution

in Regional Trade Agreements

Magdalene Silberberger

Witten/Herdecke University

Frederik Stender*

Ruhr-University Bochum __________________________________________________________________________________

Abstract

In this paper, we examine the interplay of regional economic integration and the use of bilateral antidumping

(AD) measures. Our empirical analysis brings three central findings to light: (𝑖) We find that regional trade

agreements (RTAs) generally reduce the likelihood of AD activity among integration partners. (𝑖𝑖) In addition,

an improvement in tariff treatment of trading partners —regardless of whether expressed in absolute terms or as

a tariff margin— generally leads to a lower likelihood of bilateral AD activity. (𝑖𝑖𝑖) Regarding the interaction of

both events, however, a higher tariff margin that fellow integration partners reciprocally grant themselves com-

pared to third country trading partners leads to a higher likelihood of bilateral AD activity than a tariff margin in

equal amounts among non-integration trading partners. The latter effect seems to be primarily driven by those

RTAs with a participation of “South” countries.

Keywords: Antidumping trade liberalization regional economic integration

tariff margin panel data probit

JEL classification: F13; F14; F15

__________________________________________________________________________________

* Frederik Stender (corresponding author): Ruhr-University Bochum, Faculty of Management and Economics,

Chair of International Economics, 44780 Bochum/Germany, email: [email protected];

Magdalene Silberberger: Witten/Herdecke University, Alfred-Herrhausen-Straße 50, 58448 Witten/Germany,

phone: +49 (0) 2302 926-509, email: [email protected].

Acknowledgments: The authors deeply appreciate helpful comments and suggestions from Matthias Busse,

Ronald Davies, Maurizio Zanardi, the participants of the “Topics in International Trade 2016” conference held in

Dublin and the “Australasian Trade Workshop” in Brisbane in 2017, as well as from two anonymous referees.

mailto:[email protected]

- 1 -

1. Introduction

With cumbersome multilateral trade liberalization under World Trade Organization (WTO)

auspices, regional trade agreements (RTAs) have been experiencing a global revival as a sec-

ond-best for the past 25 years with around 80 percent of all existing RTAs coming into force

after 1989. While mutual tariff preferences appear as a core concession among integration

partners, the effect of RTAs on other —non-tariff— trade barriers, however, is less straight-

forward.

Concurrently with the proliferation of RTAs, the global trade community has witnessed

massive growth in antidumping (AD) activity. Originally intended to prevent or offset “un-

fair” price setting in international trade relations, i.e. selling an export good at a lower price

abroad than domestically, the increase of AD measures —especially through a growing num-

ber of emerging and developing market users since the 1990s— has given rise to the concern

that they are simply used as another protectionist instrument (Stiglitz, 1997; Aggarwal, 2004;

Prusa, 2005; Vandenbussche and Zanardi, 2008; Bown and McCulloch, 2012; Bienen et al.,

2014).

Figure 1 displays the temporal evolution of both trade policies over time. While only 26

RTAs were in force in 1991, their number has increased tenfold in the wave of “new regional-

ism” by 2014. A similar pattern can be found for annual AD activity where the number of

bilateral measures in force at the four-digit Harmonized System (HS) commodity level has

jumped from 346 in 1991 to 1,594 in 2014. Notably, those imposed among integration part-

ners recorded a share in total AD activity in the one-hundredth of a percent range in 1991 but

have grown significantly to above 20 percent within the following two decades. This devel-

opment may admittedly well be attributed to the mere circumstance of the increased number

of RTAs, but it should however be pointed out that RTA-growth achieved some 927 percent

- 2 -

from 1991 to 2014 whereas intra-bloc AD measures have seen an impressive five-digit expan-

sion rate over the same time span.

{Please insert Figure 1 about here}

Substantial research has been conducted on the determinants and the effects of both re-

gional economic integration and AD. Evidence on their interplay, however, is relatively

scarce. Some studies examine the general effect of tariff cuts on AD (Aggarwal, 2004; Fein-

berg and Reynolds, 2007; Moore and Zanardi, 2011; Ketterer, 2016) whereas others investi-

gate the question on how the conclusion of an RTA impacts the use of AD measures (Prusa

and Teh, 2010; Ahn and Shin, 2011). What is missing in the puzzle is research on the rela-

tionship between the depth of preferential liberalization and AD. Even though a common core

element across RTAs is the reduction of tariffs, the elimination of intra-bloc tariffs not only

occurs gradually, if at all, but the coverage of sectors negotiated as well as the actual extent of

tariff cuts in RTAs varies greatly (Estevadeordal et al., 2008; Bown and Tovar, 2016). An

important and interesting question would therefore be whether in the course of mutual tariff

concessions, member countries of RTAs use AD measures more or less frequently against one

another compared to third countries.

From a theoretical point of view, there are two possible effects: “Policy complementarity”

and “policy substitution” (Beverelli et al., 2014). On the one hand, AD could be used by poli-

cy makers to reinforce a pro-RTA bias and be directed the same way as tariff concessions

within the RTA. Consequently, mutual agreement on regional free trade gradually cuts intra-

bloc tariffs and presumably curbs the implementation and use of non-tariff barriers in view of

the benefits of the free movement of goods. In addition, less sheltered exporters’ home mar-

kets make price differentials more difficult to sustain and would therefore make price dump-

ing, and with this AD less frequent. As noted by te Velde and Bezemer (2006) and Limão

(2016), many RTAs include specific investment provisions that could also lead to more verti-

- 3 -

cally-motivated intra-bloc Foreign Direct Investment (FDI) and a shift to more production

networks among integration partners. As a result, intra-bloc price dumping and its counter-

measure would become less likely. Finally, –especially more recent– RTAs not only foster

trade liberalization but also entail various additional provisions such as the harmonization of

standards and intensified regional economic integration could entail a reluctance of trade au-

thorities to disturb relations with their partners by pursuing AD cases aggressively (Niels and

ten Kate, 2006).On the other hand, intra-bloc tariff cuts might expose member countries’ do-

mestic industries to more competitive pressure from one another that may call for protection,

particularly in industries that would have possibly been excluded in multilateral negotiations.

With the removal of intra-bloc tariffs, temporary trade barriers such as AD appear as one of

the few remaining legitimate tools to restore MFN-like treatment that member countries had

departed from through the implementation of the RTA (Bown et al., 2014). In this context,

one important factor is the institutional requirement of RTAs under GATT Article XXIV to

decrease preferential tariffs in across all sectors, regardless of a country’s preferences. In sec-

tion 3, we develop a simple framework which takes this circumstance into account and argue

that RTA members tend to substitute tariffs by AD measures to mitigate the costs associated

with preferential tariff reduction. However, Bown and Tovar (2016) argue that the notion to

liberalize all trade under the GATT Article XXIV is imprecise and thus the impact of the in-

stitutional constraint placed on countries through it remains an empirical one.

In this paper, we examine the interplay of regional economic integration and the use of bi-

lateral AD measures. As will be outlined more detailed in the literature review in section 2,

compared to the well-studied interplay of tariff liberalization in general and AD activity, to

the best of our knowledge, there are only two empirical studies (Prusa and Teh, 2010; Ahn

and Shin, 2011) that explicitly analyze the relationship between regional economic integration

and intra-bloc AD activity. Both, however, use a dummy-variable-approach to identify the

effect of regional economic integration on the use of AD measures

- 4 -

Our paper deviates from the existing literature in several aspects. Firstly, we extend the pe-

riod of investigation to more recent years (from 1991 through 2014). Secondly, mainly

through the exhaustive incorporation of a broad range of fixed effects we control for various

economic influences that may have an impact on bilateral AD activity apart from changes in

applied tariffs. Thirdly and most importantly, we explicitly consider that there may be differ-

ent intensities of regional economic integration with respect to the progress of intra-bloc tariff

liberalization. Not only do AD provisions vary widely across RTAs, some RTAs also reduce

almost all intra-bloc tariffs significantly while others yield only minor tariff preferences for

fellow integration partners. In addition, some products might be excluded from tariff reduc-

tions if they are considered to be sensitive and those are the ones that might also be prone to

competition policy through AD.

To account for this, we construct a variable that measures the extent to which RTA mem-

ber countries reciprocally grant themselves tariff margins in comparison towards third country

competitors and investigate how respective developments impact the intra-bloc use of AD.

More specifically, we speculate that an increase in the integration intensity of RTAs, that we

define as an amplification of the intra-bloc tariff margin, may prompt importing integration

partners to mitigate the effect of granted tariff preferences by the simultaneous introduction of

intra-bloc AD measures. While we are able to confirm the findings in Prusa and Teh (2010)

and Ahn and Shin (2011) with respect to a pure “RTA-effect”, our results also suggest a more

diversified conclusion regarding the relationship between regional economic integration and

AD measures which confirms our intuition that employing a dummy variable does not capture

the full effect of RTAs on AD activity.

What is more, we differentiate between various types of RTAs with respect to their com-

position of member countries in an extended model specification as existing literature points

towards potentially diverging effects regarding the interplay of tariff liberalization in general

and AD activity between traditional- and emerging and developing market AD users. We ac-

- 5 -

count for that by a comparison of North-North, North-South and South-South RTAs where

findings confirm variations in the extent of trade policy substitution depending on the mix of

countries of an RTA.

The remainder of this paper is structured as follows: Section 2 reviews the relevant litera-

ture. We then develop a simple model as a foundation for our empirical examination in sec-

tion 3. Our data and methodology are outlined in section 4. Section 5 presents and discusses

the main results whereas section 6 provides an extension of our basic model. Section 7 con-

cludes.

2. Literature Review

The relevant literature can roughly be divided in two strands: The first group of studies focus-

es on the relationship between tariff reductions in general and the use of AD measures where-

as the second group of literature investigates the relationship between regional economic inte-

gration and AD measure use.

A large part of empirical research in the first group of literature is motivated by a theoreti-

cal model of Anderson and Schmitt (2003) which is based on a reciprocal dumping model

developed in Brander and Krugman (1983). The authors find that once tariffs are reduced in

the course of market-opening, there is an incentive for governments to substitute previous

protection by other trade barriers where the first-best choice would be quotas. However, if the

systematic introduction of quotas is no longer possible AD activity may take their place.

Empirical studies dealing with a potential substitution effect between tariffs and AD

measures predominantly use a continuous tariff policy variable as key explanatory variable

and mostly find evidence for a respective substitution effect in developing countries or the

group of new-heavy, i.e. non-traditional, AD users. Exemplarily, Aggarwal (2004) studies AD

measure use within a wider array of macroeconomic factors and shows that for the group of

- 6 -

developing countries a decline in applied tariffs is associated with a substantial increase in

AD initiations in those countries. The author, however, does not find any evidence for a simi-

lar effect in the group of developed countries. In a related study, Feinberg and Reynolds

(2007) empirically investigate the role of tariff liberalization in explaining the pattern of a

significant increase in both, the number of AD initiations and the number of AD users. They

find that for the group of non-traditional AD users there is a positive effect of Uruguay Round

tariff concessions on the probability of a country filing an AD petition but, by contrast, that

the effect for traditional users is negative. In greater detail, for the group of non-traditional

AD users a decrease in a sector’s average tariff is estimated to be associated not only with an

increased probability of a country filing an AD petition but also with an increase in the num-

ber of petitions. Similarly, Moore and Zanardi (2011) find evidence for a substitution effect

only for a relatively small set of developing countries that heavily use AD protection. Lastly,

Ketterer (2016) explores that multilateral trade reforms undertaken by the European Union

(EU) in the course of the Uruguay Round have resulted in a substitution of tariffs by AD

measures.

Less attention has been given to the interplay of regional economic integration and intra-

bloc AD activity. In a theoretical paper, Copeland (1990) investigates the relationship be-

tween negotiable and non-negotiable trade barriers. He argues that RTAs contain loopholes

for protectionist measures which are used in the course of trade negotiations to substitute tar-

iffs for non-tariff measures. More specifically, once tariffs are lowered towards fellow mem-

ber countries, domestic producers’ demand for protection will force governments to use other

trade policy instruments which are comparatively more costly.

As concerns empirical studies, to the best of our knowledge, there have only been two pub-

lications to date, both of which employ a dummy-variable-approach to analyze the effect of

regional economic integration on bilateral AD activity. Closest related to our study is Ahn and

Shin (2011) who use a negative binominal quasi-maximum likelihood estimation with a count

- 7 -

variable that measures the number of AD actions for members of a Free Trade Area (FTA)

between 1995 and 2009 as a dependent variable. They generally conclude that the number of

AD investigations drops subsequent to the FTA enactment. In comparison, Prusa and Teh

(2010) employ a difference-in-difference approach and find both trade creation through a de-

creased incidence of intra-bloc AD filings, and trade diversion through the increased AD use

against non-Preferential Trade Agreement (PTA) members. The net effect of PTAs on total

AD filings is, however, rather small.

Lastly, there are two papers that are more loosely related to our research focus. Blonigen

(2005) examines whether a specific NAFTA-provision, more precisely its Chapter 19 dispute

settlement panel has contributed to a reduction of US AD- and countervailing activity against

Canada and Mexico and finds little evidence in this respect. By contrast, Bown and Tovar

(2016) provide empirical evidence that MERCOSUR’s preferential liberalization process has

led to increased extra-bloc import protection including AD for the cases of Argentina and

Brazil.

3. A Simple Model

As the contribution of our paper is mainly empirical we do not intend to establish a sophisti-

cated political-economy model on the interplay of regional economic integration and AD pol-

icy. Instead, in a more modest fashion, we briefly formalize the basic intuition behind differ-

ing effects of tariff preferences on AD between RTA member countries and external trading

partners.

To begin with, membership in an RTA entails both benefits and costs for individual sectors

but also for the economy as a whole. Benefits arise from free trade between member countries

due to the reduction of intra-bloc tariffs, and with this the exploitation of comparative ad-

vantage.The costs can be generalized as an increase in competition for domestic firms.

- 8 -

Whether or not a country joins an RTA depends on the ratio between expected benefits and

costs. Country 𝑖 will sign an RTA with country 𝑗 if the sum of benefits outweighs the costs

summed up over all sectors of 𝑖’s economy (𝑆):

∑ 𝐵 −

𝑆

∑ 𝐶 > 0

𝑆

(1)

The institutional fundament of RTAs is the substantial and sustainable removal of tariffs be-

tween member countries across all sectors, regardless of the participants’ preferences (Bown

and Tovar, 2016). Tariff concessions are thus likely to be negotiated and granted to fellow

member countries even in sectors which country 𝑖 would have otherwise protected due to

domestic interest.1 This means that without the conclusion of the RTA:

�̃�𝑁𝑂 𝑅𝑇𝐴 < 𝑆 (2)

where �̃� denotes the number of sectors in which tariff preferences are granted and S stands for

the total number of sectors in an economy. In contrast, the institutional framework of an

RTAs forces country i to grant tariff preferences to fellow member countries across all sectors

of the economy

�̃�𝑅𝑇𝐴 = 𝑆 (3)

Tariff concessions in sectors which would otherwise be protected increase the costs of an

RTA for country 𝑖 for reasons of increased competition, so that:

∑ 𝐶 >

�̃�𝑅𝑇𝐴

∑ 𝐶

�̃�𝑁𝑂 𝑅𝑇𝐴

(4)

Country 𝑖 will sign the RTA as long as:

1 There may be domestic interest in sensitive sectors in which a country may not hold a comparative advantage

but still intends to produce or export for political or strategic reasons or in sectors that are of particular im-

portance for a country in economic terms.

- 9 -

∑ 𝐵 >

𝑆

∑ 𝐶

𝑆

(5)

The net gain of an RTA, however, could be further increased if granted tariff concessions in

sensitive sectors were (partly) neutralized by the introduction of other trade barriers such as

AD. With this, there there is an incentive to substitute tariffs by AD measures. In particular

sectors that have had a relatively high level of tariff protection before the conclusion of the

RTAmay call for protection. Since tariffs are bound through mutual agreement between

member countries the use of other non-tariff measures –such as AD– is triggered (Martin and

Vergote, 2008).

4. Data and Methodology

For econometric implementation, we use a non-linear probit model framework. Our sample is

constructed as a bilateral panel of 9 developed and 22 developing countries including a binary

dependent variable that equals unity whenever an AD measure applied by importing country 𝑖

against country 𝑗 is in force in period 𝑡 at the HS four-digit commodity level 𝑘, zero otherwise

(see Appendix 1 for the countries in our sample).

Note that we treat the EU as a single country. As stated by the European Commission, alt-

hough producers in member countries of the EU can initiate AD investigations individually

AD policy actions must be undertaken corporately by the bloc.2 We pay careful attention to

the EU’s evolutionary enlargement of member states. To illustrate this point with an example,

consider the case of Austria, Finland, and Sweden, all of which joined the EU only in 1995.

For this reason. we disregard any AD activity originating from or targeting the three countries

before receiving EU membership status. Hence, neither do we count any AD case filed by

2 See http://ec.europa.eu/trade/policy/accessing-markets/trade-defence/actions-against-imports-into-the-eu/anti-

dumping/.

http://ec.europa.eu/trade/policy/accessing-markets/trade-defence/actions-against-imports-into-the-eu/anti-dumping/

http://ec.europa.eu/trade/policy/accessing-markets/trade-defence/actions-against-imports-into-the-eu/anti-dumping/

- 10 -

either one of the three before 1995 as a case filed by the EU, nor does any AD case filed

against either one of the three before 1995 represent a case filed against the EU.

The number of importers in our sample is determined by the availability of bilateral AD in-

formation that we extract from the World Bank’s Global Antidumping Database (GAD) col-

lected by Bown (2015). Importantly, we neither distinguish between different types of the

measure in force, nor do we make a distinction between preliminary and final AD measures as

both are de facto trade barriers that imply respective duties and have been found to nearly

equally impact trade flows (Staiger and Wolak, 1994; Chandra, 2016). The overall time span

we consider ranges from 1991 to 2014. Based on data limitations, however, several countries

enter our sample with later initial years.

We merge several explanatory variables with our AD information including annual effec-

tively applied trade-weighted bilateral tariff rates that we draw from the World Bank’s World

Integrated Trade Solutions (WITS) database across all HS four-digit commodities whenever

readily available. The WTO’s Regional Trade Agreement Information System (RTA-IS) is

used to obtain information of all RTAs in force that are relevant to our sample. While HS

four-digit annual bilateral import flows come from the UN Comtrade database, we collect

macroeconomic indicators such as GDP per capita at constant PPP International US-Dollars,

aggregate real GDP growth and current account balance as a percentage of GDP from the

World Bank’s World Development Indicators (WDI).

The sample finally used for estimation is further reduced in an additional aspect. As our re-

search interest lies upon analyzing as to whether or not bilateral AD activity is determined by

regional economic integration we keep only those commodities that have ever been subject to

the imposition of an AD measure by at least one importing country of our sample between

1991 and 2014. Put differently, we consider only those commodities in the empirical analysis

for which trade policy substitution would be feasible at all. Out of the 1,260 HS four-digit

commodities which have been traded between the countries of our sample in the period from

- 11 -

1991 to 2014, this condition is satisfied by 641 commodities (a share of 51 percent). Techni-

cally speaking, the procedure reduces “zeros” in our binary dependent variable. Nevertheless,

as can be seen in Appendix 3, respective data entries are still overwhelmingly present. Not

only does AD activity vary substantially across imposing countries but it does so likewise

across commodities. While the United States alone accounts for roughly 22 percent of all non-

zero data entries in our dependent variable, Ecuador, at the other extreme, records only five

different HS four-digit AD cases. The HS four-digit commodity sector that shows the greatest

AD activity in our sample is code 7208 (“Flat-rolled Products of Iron or Non-alley Steel, Hot-

rolled”), a finding that is in accordance with Durling and Prusa (2006).

Our baseline model specification reads as:

𝐴𝐷𝑖𝑗𝑘𝑡 = 𝛼0 + 𝛽1𝑅𝑇𝐴𝑖𝑗𝑡−1 + 𝛽2ln (1 + 𝑡𝑖𝑗𝑘𝑡−1) + 𝛽3ln (1 + 𝜏𝑖𝑗𝑘𝑡−1)

+ 𝛽4ln (1 + 𝜏𝑖𝑗𝑘𝑡−1𝑅𝑇𝐴 ) + 𝛾𝑚

′ 𝜲𝑖(𝑖𝑗)[𝑘]𝑡−1 + 𝑢𝑖𝑗𝑘𝑡

(6)

where 𝛼0 is a constant and 𝑢𝑖𝑗𝑘𝑡 is the error term that is described in greater detail further be-

low.

The first four explanatory variables are incorporated to capture the impact of trade policy

on bilateral AD activity. More precisely, 𝑅𝑇𝐴𝑖𝑗𝑡 is a binary variable that equals unity if 𝑖 and

𝑗 are member countries of the same RTA in period 𝑡 regardless of the bloc’s formal status of

economic integration, zero otherwise. In total, 70 different RTAs involving the countries of

our sample were in force between 1991 and 2014 where 66 of them are classified as FTAs and

only four have the formal status of a Customs Union (CU). Unilateral or partial scope agree-

ments are not taken into account. All 70 RTAs are considered for empirical analysis. While

among the entire group of RTAs 39 constitute North-South trade agreements, 17 are consid-

ered as South-South, and 14 as North-North trade agreements (see Appendix 2 for the com-

plete list of RTAs covered by our sample).

- 12 -

We control for annual sector-specific bilateral tariff treatment in two different ways. First-

ly, 𝑡𝑖𝑗𝑘𝑡 is a continuous variable indicating the absolute ad-valorem effectively applied tariff

of 𝑖 towards 𝑗 in commodity 𝑘 and period. In turn, 𝜏𝑖𝑗𝑘𝑡 is a continuous variable indicating the

tariff margin that i grants to j in sector k and period t, and could be referred to as the relative

tariff treatment that i imposes on j in sector k and period t compared to third country trading

partners. With slight distinction, we employ the definition from Carrère et al. (2010) and

compute 𝜏𝑖𝑗𝑘𝑡 as:

𝜏𝑖𝑗𝑘𝑡 = (𝑡𝑖𝑗𝑡𝑘

𝑏 − 𝑡𝑖𝑗𝑘𝑡

1 + 𝑡𝑖𝑗𝑘𝑡)

with: 𝑡𝑖𝑗𝑘𝑡𝑏 = ∑ (𝜃𝑖𝑠𝑡𝑘𝑡𝑖𝑠𝑡𝑘),𝑠

𝜃𝑖𝑠𝑘𝑡 = 𝐼𝑀𝑖𝑠𝑘𝑡 𝐼𝑀𝑖𝑘𝑡𝑇⁄ ; 𝐼𝑀𝑖𝑗𝑘𝑡 ∉ 𝐼𝑀𝑖𝑘𝑡

𝑇 ; 𝑗 ≠ 𝑠; ∑ 𝜃𝑖𝑠𝑘𝑡 = 1

𝑠

(7)

where 𝑡𝑖𝑗𝑘𝑡 is the above introduced absolute tariff rate country 𝑖 imposes on 𝑗 in commodity 𝑘

and in period 𝑡, and 𝑡𝑖𝑗𝑡𝑘𝑏 denotes the annual effective benchmark tariff that is specific to each

country-pair-commodity combination. It is computed as a trade weighted average tariff over

all of 𝑗’s competitors in our sample, 𝑠, in 𝑖’s commodity 𝑘 market, with 𝜃𝑖𝑠𝑡𝑘 representing 𝑠’s

share in 𝑖’s total commodity 𝑘 imports exclusive of those from 𝑗, and 𝑡𝑖𝑠𝑡𝑘 being the respec-

tive effectively applied tariff. Accordingly, the measure indicates above (below) average rela-

tive tariff treatment of 𝑗 in 𝑖’s market, and with this a positive (negative) tariff margin, if

𝜏𝑖𝑗𝑘𝑡 > 0 (−1 < 𝑇𝑖𝑗𝑡𝑘 < 0) so that an increase translates into a more preferential tariff treat-

ment. (Absatz komplett aus Mercosur-paper, IEEP)

One could argue that relative and absolute tariff treatment essentially measure the very

same from a trade policy perspective, and would thus be highly correlated with each other.

Consequently, one of the two variables would be redundant. In fact, however, relative and

absolute tariff treatment are conceptually different, as the former indicates the relation of im-

posed tariff policy among competitors, but the latter solely defines the level of the imposed

- 13 -

tariff burden. This view is further confirmed when inspecting the relatively low correlation

between the two tariff variables given below in Table 1.

{Please insert Table 1 about here}

Lastly, 𝜏𝑖𝑗𝑡𝑘𝑅𝑇𝐴 is our key variable of interest and denotes an interaction of the RTA dummy var-

iable and 𝜏𝑖𝑗𝑘𝑡. The variable captures the core element of regional economic integration where

fellow member countries are reciprocally granted tariff margins compared to third country

trading partners (aus Mercosur-paper, IEEP). It quantifies the impact on the probability of AD

measure use of a variation in the tariff margin that 𝑖 grants 𝑗 in sector 𝑘 conditioned for the

case that both are fellow member countries in an RTA in period 𝑡.

One potential issue with our interaction term could be a clustering of values for 𝜏𝑖𝑗𝑘𝑡 by

groups. Clearly, one would assume that tariff margins that are reciprocally granted among

RTA members are significantly higher than those among countries that are not tied together in

a common RTA. If this was the case, 𝜏𝑖𝑗𝑘𝑡 would be highly correlated with the RTA dummy

variable, resulting in an inherent double capture of our RTA dummy. The distribution of val-

ues for 𝜏𝑖𝑗𝑘𝑡 for both outcomes of the RTA dummy variable is given in Figure 2. As can be

seen in the lower panel, we indeed find 50 percent of all middle values of 𝜏𝑖𝑗𝑘𝑡 among integra-

tion partners to be grouped at positive values. Since one of the major motives for the conclu-

sion of an RTA is the prospect of mutual tariff preferences, nonetheless, this outcome is not

surprising. By comparison, 50 percent of all middle values of 𝜏𝑖𝑗𝑘𝑡 among non-integration

trading partners are found at negative values, naturally owing to the absence of mutual tariff

concessions. The upper panel, however, gives an indication that for both groups we still ob-

serve considerable variation in values for 𝜏𝑖𝑗𝑘𝑡. With this, our sample does well cover differ-

ent integration intensities of RTAs with respect to tariff margins observed among integration

partners.

- 14 -

{Please insert Figure 2 about here}

This view is further confirmed in Table 2 where we present descriptive information on val-

ues for 𝜏𝑖𝑗𝑘𝑡 among integration partners for selected RTAs. Exemplarily, MERCSOSUR

stands out not only with respect to the largest average 𝜏𝑖𝑗𝑘𝑡, but also with respect to the largest

maximum value. Speaking in non-technical terms, MERCOSUR’s member countries recipro-

cally grant themselves the highest tariff margins across the selection of RTAs in Table 2. An

explanation could be found in the comparatively high external tariff protection for all of

MERCOSUR’s member countries. The lowest tariff margin among MERCOSUR’s member

countries occurred in 2002 which corresponds well to the economic turmoil in the Southern

Cone of that year.

In comparison, the Canada-Chile FTA has on average provided the two countries with bet-

ter market access conditions (in terms of preferential tariff treatment) in each other’s market

to a much lower extent. Given substantial variation in intra-bloc tariff margins even across

RTAs (or put differently, integration intensities as defined in the introduction) the incorpora-

tion of the above described interaction term is reasonable from both an econometric and a

research question perspective.


Furthermore, vector 𝜲 defines a set of various control variables that we adopt from literature

including bilateral import growth (Δ 𝐼𝑀𝑖𝑗𝑘𝑡) and -share (𝐼𝑀_𝑆𝐻𝐴𝑅𝐸𝑖𝑗𝑘𝑡), both sector- and

time-varying. We expect higher imports to raise the incentive for countries to impose an AD

measure and thus a positive coefficient for both the level and the change in imports (Ag-

garwal, 2004; Blonigen, 2005; Feinberg and Reynolds, 2007; Moore and Zanardi, 2011; Ahn

and Shin, 2011; Ketterer, 2016). To account for macroeconomic conditions from the import-

ers perspective we introduce GDP growth (Δ 𝐺𝐷𝑃𝑖𝑡) and the current account balance as per-

- 15 -

centage of GDP (𝐶𝐴𝐵𝑖𝑡) which we both expect to be negative. Concerning the first variable,

we argue that a downgrade in economic activity presumably leads to increased incentives to

protect domestic industries by means of trade barriers while the second should be negative if

domestic firms assume that higher imports make the policy-makers more likely to impose an

AD measure (Knetter and Prusa, 2003; Prusa and Teh, 2010; Moore and Zanardi, 2011). To

measure a country’s economic development we employ annual per capita GDP (pc_𝐺𝐷𝑃𝑖𝑡). In

general, high average income is associated with better institutions and legal systems which are

necessary to administer complex AD procedures (Moore and Zanardi, 2011).

We additionally control for the complete absence of any legal framework regarding the use

of intra-bloc AD measures in RTAs by employing the binary variable 𝑁𝑂_𝑅𝑈𝐿𝐸𝑆𝑖𝑗𝑡 that

equals unity in this case. By contrast, the variable’s counterargument may attain three differ-

ent gradations of policy severity depending on the RTA under consideration, i.e. (𝑖) the strict

prohibition of intra-bloc AD activity, (𝑖𝑖) WTO-rule based AD use, or (𝑖𝑖𝑖) a custom-tailored

legal framework following WTO-rules in either milder or stricter implementation. Our sample

includes two RTAs that entirely prohibit the use of intra-bloc AD activity, namely the Cana-

da-Chile FTA and the Australia-New-Zealand Closer Economic Relations Trade Agreement

(ANZCERTA).3 The remaining RTAs divide into 6 with no legal framework, 36 with WTO-

rule based AD use, and 27 with a specific legal framework (Prusa, 2011; various official doc-

uments of RTAs that are considered for empirical analysis in this paper).4 . The elimination

3 Plausibly, the substitution of intra-bloc tariffs by imposing AD measures is not feasible in RTAs that entirely

prohibit the use of intra-bloc AD activity. The consideration of respective RTAs would thus as a consequence

curtail the detection of potential trade policy substitution. Excluding the Canada-Chile FTA and ANZCERTA

from our sample, however, would imply a sample selection bias that we assess to be more severe than a potential

downward-bias from both an economic and an econometric perspective. 4 Even if an RTA lacks specific rules with respect to intra-bloc AD use, AD activity undertaken by member

countries are subject to WTO-rules if they are WTO members. WTO guidelines, however, are vague and do not

provide detailed guidance on AD-initiation and usage (Blonigen and Prusa, 2001; Rey, 2012). We therefore

attribute more relevance to the fact of whether or not AD use is explicitly regulated the agreement. Our classifi-

cation of RTA according to their intra-bloc AD policy given in the text is in line with Prusa (2011) who identi-

fies three different groups of PTAs with regards to their treatment of AD: “disallow”, “no language”, “specific

rules”. Rey (2012) even maps five profiles: “reference to WTO rules”, “own rules replicating WTO rules”, “no

provision”, “own rules” and “prohibition”. With respect to empirical evidence, differentiating by type of AD

- 16 -

of intra-bloc tariffs in RTAs and with this, the grating of tariff preferences occurs rather grad-

ually instead of abruptly often following a negotiated scheme. For reason of modelling poten-

tial dynamics in tariff policy and following previous literature (Feinberg and Reynolds, 2007;

Moore and Zanardi, 2011; Ketterer, 2016), we introduce a variant specification and replace

the three tariff policy variables that we incorporated in equation (6) in terms of their levels by

their respective rate of change definitions:

𝐴𝐷𝑖𝑗𝑘𝑡 = 𝛼0 + 𝛽1𝑅𝑇𝐴𝑖𝑗𝑘𝑡−1 + 𝛽2Δ ln(1 + 𝑡𝑖𝑗𝑘𝑡−1) + 𝛽3Δ ln(1 + 𝜏𝑖𝑗𝑘𝑡−1)

+ 𝛽4Δ ln (1 + 𝜏𝑖𝑗𝑘𝑡−1𝑅𝑇𝐴 ) + 𝛾𝑚

′ 𝜲𝑖(𝑖𝑗)[𝑘]𝑡−1 + 𝑢𝑖𝑗𝑘𝑡

(8)

where Δ ln(1 + 𝑡𝑖𝑗𝑘𝑡) denotes the difference of the logarithm of (1 + 𝑡𝑖𝑗𝑘𝑡) compared to the

respective value in the previous period. A negative rate of change implies an improvement in

absolute tariff treatment within two consecutive periods. In comparison, Δ ln(1 + 𝜏𝑖𝑗𝑘𝑡) gives

the growth rate in relative tariff treatment computed as the difference of the logarithm of

(1 + 𝜏𝑖𝑗𝑘𝑡) compared to the respective value in the previous period. Because a higher level of

relative tariff treatment means a more preferential relative tariff treatment, by implication, a

positive value of its growth rate represents an improvement within two consecutive periods.

The same interpretation applies to the growth rate of conditioned relative tariff treatment.

We follow Ketterer (2016) and introduce all explanatory variables lagged by one period in

order to alleviate potential reverse causality. This procedure is relevant in particular with re-

gard to our tariff treatment variables as the use of AD measures may likewise be seen as

“safety valve” for pending tariff liberalization efforts (Niels and ten Kate, 2006; Moore and

Zanardi, 2009). In addition, it appears reasonable to assume AD implementation to be time

lagged when considered as a response to changes in the economic or trade environment of a

country.

provision, Prusa and Teh (2010) find a significant reduction of AD measures in PTAs that regulate AD rules,

whereas PTAs without any AD provisions are associated with a –statistically insignificant– increase in AD.

- 17 -

In terms of our model specifications given above, we are well aware that both may not en-

compass determinants of bilateral AD activity entirely. As a result, estimation results could be

subject to a classical omitted variables bias in case of a correlation of explanatory variables

neglected and those included. In this respect, one may think of further non-tariff measures that

find implementation complementarily or as a substitute to tariff policy and that are also corre-

lated to AD imposition. Ideally, we would thus control for these non-tariff barriers but bilat-

eral data coverage at the disaggregated sector level is often missing or poor. Therefore, ad-

dressing unobserved heterogeneity and thereof arising endogeneity issues, we incorporate

fixed effects into our model specifications and model equation (6)’s and (8)’s error terms

(𝑢𝑖𝑗𝑘𝑡) in two different fashions: Our benchmark estimation technique includes importer- (𝛼𝑖),

exporter- (𝛼𝑗), year- (𝜆𝑡), and two-digit commodity (𝜑𝑛) dummy variables so that:

𝑢𝑖𝑗𝑘𝑡1 = 𝛼𝑖1 + 𝛼𝑗1 + 𝜆𝑡1 + 𝜑𝑛1 + 𝜀𝑖𝑗𝑘𝑡 (9)

In comparison, a more robust version controls for importer-year- (𝜂𝑖𝑡), exporter-year- (𝜇𝑗𝑡),

and commodity effects so that:

𝑢𝑖𝑗𝑘𝑡2 = 𝜂𝑖𝑡2 + 𝜇𝑗𝑡2 + 𝜑𝑛2 + 𝜐𝑖𝑗𝑘𝑡 (10)

5. By contrast, we do not consider panel fixed effects as such, i.e. country-pair-commodity

dummy variables in the case of the dataset at hand. While their inclusion would allow to

control for historically established bilateral AD activity, it has been largely discussed

that it would in turn bias estimation results due to an incidental parameters problem in

fixed T non-linear binary response models (Greene, 2004).5Main Results

5 Although the use of a linear probability model is criticized from an econometric point of view due to the cir-

cumstance that the estimated probability of the model may fall outside the [0;1]-interval, we run both model

specifications using Ordinary Least Squares (OLS) as a robustness check. Unreported estimation results obtained

from the linear probability model including country-pair-commodity fixed effects are nonetheless very similar to

those reported and discussed below (can be provided upon request).

- 18 -

Table 2 displays average marginal probability effects of explanatory variables of our baseline

specification in columns (1) and (2), and our variant specification in columns (3) and (4). Cor-

responding fixed effects that are taken into account for estimation are given at the bottom end

of each column.6

Turning towards trade policy variables, estimation results indicate an increase in the likeli-

hood of bilateral AD activity with increasing ad-valorem tariffs, or put differently, with a de-

terioration in absolute tariff treatment (𝑡𝑖𝑗𝑘𝑡). The finding is mirrored when considering 𝜏𝑖𝑗𝑘𝑡,

i.e. a lower tariff margin is associated with an increase in the likelihood of observing bilateral

AD activity. With this, we find simultaneous discriminatory practice across trade policies.

This evidence contradicts theoretical predications and also to some extent results from the

abovementioned first group of studies that find a substitution effect between tariffs and AD.

There are, however, major differences across model specifications, country samples, and be-

tween previous literature and our study.7 Our results suggest that a trading partner which al-

ready experiences a comparatively unfavorable tariff treatment (i.e. high absolute tariffs

and/or a low tariff margins) will also be more likely to face more AD measures imposed.

From an industry perspective Moore and Zanardi (2011) argue that sectors that have success-

fully avoided liberalization in the past will be more likely to access the AD process to ensure

continuous protection.

In line with the findings of the abovementioned second group of literature, the coefficient

estimate for 𝑅𝑇𝐴𝑖𝑗𝑡 points towards a decreased likelihood of AD activity between member

6 The estimation with four-digit- instead of two-digit commodity effects only marginally alters the results. In an

additional robustness check, we estimate both baseline and variant specifications incorporating all explanatory

variables lagged by two periods instead of using their first lags as an augmented strategy to address potential

reverse causality problems. In all aspects, findings quantitatively and qualitatively confirm those reported, and

with this strengthen our original approach (both not reported, but can be provided upon request). 7 Ketterer (2016) uses the bound and applied MFN tariff change negotiated during the Uruguay round and relates

it to post-Uruguay AD initiations between 1996 and 2005 in the EU only, while Feinberg and Reynolds (2007)

investigate the impact of 24 importing countries’ tariff reduction during the Uruguay round on the number of AD

petitions filed by this country between 1996 and 2003. Closest to our study is Moore and Zanardi (2011) which

builds upon Feinberg and Reynolds (2007). They use a sample of 29 developing and six developed countries for

the period of 1991 through 2002 and initial tariff level as well as trade liberalization in applied tariffs and find a

substitution effect only for the “new heavy users”. When using a specification that is closest to ours in terms of

country sample they even observe a complementarity effect.

- 19 -

countries of the same RTA. It needs to be noted that our negatively signed coefficient esti-

mate could imply fewer newly initiated AD cases between fellow member countries com-

pared to non-integration trading partners but likewise the removal of existing AD measures

skewed towards fellow member countries. When interpreting the RTA dummy variable, how-

ever, one has to keep in mind that it does not capture any tariff effects of regional economic

integration, since we explicitly control for bilateral tariff treatment. Instead, the RTA dummy

variable measures an additional effect of RTAs, which Limão (2016) classifies as the breadth

of the agreement. The breadth of an agreement encompasses economic and non-economic

provisions as well as deeper policy dimensions which go beyond tariffs or even WTO matters

that are a major part of many, especially more recent RTAs. By this they can increase close-

ness between RTA partners and reduce trade policy uncertainty and thereby the number of

AD initiations (Prusa and Teh, 2010; Limão. 2016).

Albeit of deep interest from a policy perspective, the incorporation of interaction terms in

probit regressions is delicate from an econometric point of view. In a widely received contri-

bution, Ai and Norton (2003) address a prevalent misinterpretation of interaction effects as

marginal effects of interaction terms in non-linear regressions. In order to compute discrete

interaction effects adequately, we therefore compare the slopes of 𝜏𝑖𝑗𝑘𝑡 in bilateral AD activi-

ty using reference group contrast. More precisely, we compute the difference in the deriva-

tives of 𝐴𝐷𝑖𝑗𝑘𝑡 with respect to 𝜏𝑖𝑗𝑘𝑡 for the two possible outcomes of 𝑅𝑇𝐴𝑖𝑗𝑡, namely for the

case that either 𝑖 and 𝑗 are fellow member countries of the same RTA, and its failure. We find

the group contrast (denoted as “Integration Effect” in output tables) in 𝜏𝑖𝑗𝑘𝑡 always positively

signed and with the exception of column (4) statistically significant at least at the one-percent

level. Accordingly, a higher tariff margin among fellow integration partners leads to a higher

likelihood of bilateral AD activity than a tariff margin in equal amounts among non-

- 20 -

integration trading partners.8 This, however, must not necessarily imply that the conclusion of

an RTA overall increases the probability of an AD -measure on evidence of the fact that the

general RTA effect (see RTA dummy variable) is AD-reducing. Given that RTAs are subject

to a great variance of intra-bloc tariff concessions the net-effect of an RTA needs to remain

ambiguous at this point.

Focusing on the identification of potential trade policy substitution effects in RTAs, the

findings seem to support our theoretical intuition and the rather pessimistic view on the inter-

play of regional economic integration and AD use which seems to be dominated by increased

competitiveness concerns that lead to the substitution of tariffs with other non-tariff measures.

Additionally, from an implementation perspective the conclusion of an RTA leads to

stronger polictical/strategical/economic? ties between member countries.9 As indicated by

numerous empirical papers on the trade effects of RTAs (…), regional economic integrations

significantly increases trade volumes between member countries which, in turn, could entail

increased demand for protection because there is more at stake. As previously mentioned, for

many industries the conclusion of a trade agreement constitutes a threat to competitiveness.

This threat might be perceived as more prominent from a fellow member country. Notably,

this trade policy substitution effect might be a factors that dampens positive effects of RTAs.

Lastly, coefficient estimates for 𝐼𝑀_𝑆𝐻𝐴𝑅𝐸𝑖𝑗𝑘𝑡, pc_𝐺𝐷𝑃𝑖𝑡, and 𝐶𝐴𝐵𝑖𝑡 either show plausi-

ble signs or at least partly confirm those found in previous studies (Blonigen, 2005; Feinberg

and Reynolds, 2007; Moore and Zanardi, 2011; Ahn and Shin, 2011; Ketterer, 2016). Com-

pared to this, we would expect a positively signed coefficient estimate for Δ 𝐼𝑀𝑖𝑗𝑘𝑡 (intensi-

fied import relations would presumably increase the incentive to protect domestic producers)

8 Although preferential tariff policy that we incorporate into our model specifications as 𝜏𝑖𝑗𝑘𝑡 appears as the core

element of regional economic integration, we also estimate both baseline and variant specifications with an in-

teraction term of the RTA dummy variable and absolute tariff treatment, i.e. 𝑡𝑖𝑗𝑘𝑡. Estimation results are given in

Appendix 5. With the exception of column (4) where the “Integration Effect” is found to be positively signed yet

statistically insignificant, the interpretation of our original approach can be confirmed. 9 Although these ties may have existed before and could have been part of the reason to sign the RTA in the first

place, one can usually assume that an RTA intensifies them further.

- 21 -

and a negatively signed one for Δ 𝐺𝐷𝑃𝑖𝑡𝑖𝑡 (an economic downturn would presumably increase

the incentive to protect domestic producers) where in both cases we find the opposite. Alt-

hough significantly different from zero only in case of the former one (at the ten percent lev-

el), we would like to emphasize that its coefficient estimates across all columns are neverthe-

less relatively small, and with this point towards a negligible effect in economic terms.

Inspecting coefficient estimates for the binary variable 𝑁𝑂_𝑅𝑈𝐿𝐸𝑆𝑖𝑗𝑡, unsurprisingly, es-

timation results across all columns point towards an increased likelihood of bilateral AD pro-

tection in those RTAs that provide no legal framework with respect to intra-bloc AD activity.

With regard to concerns of double capture of the impact of regional economic integration on

bilateral AD activity through the simultaneous incorporation of 𝑅𝑇𝐴𝑖𝑗𝑡 and 𝑁𝑂_𝑅𝑈𝐿𝐸𝑆𝑖𝑗𝑡 ,

where the latter is constructed from the former from a methodological point of view, we also

run separate (unreported) estimations excluding either one of the variables. Findings, howev-

er, are quantitatively and qualitatively nearly identical to those reported.


6. Extension

As pointed out in the literature review in section 2 many studies find trade policy substitution

effects for developing countries only. One might therefore speculate that the impact of im-

provements in relative tariff treatment among integration partners on intra-bloc AD activity

could vary likewise depending on the mix of countries of an RTA. For this reason, comple-

menting our baseline model specification under equation (1), we run an extended version that

substitutes the single RTA dummy variable by three different dummy variables signaling var-

ious types of RTAs with respect to their composition of member countries. More precisely,

we decompose previously incorporated 𝑅𝑇𝐴𝑖𝑗𝑡 into (𝑖) 𝑅𝑇𝐴𝑖𝑗𝑡𝑁𝑆, (𝑖𝑖) 𝑅𝑇𝐴𝑖𝑗𝑡

𝑁𝑁, and (𝑖𝑖𝑖) 𝑅𝑇𝐴𝑖𝑗𝑡𝑆𝑆

that equal unity if 𝑖 and 𝑗 are member countries of the same (𝑖) North-South-, (𝑖𝑖) North-

- 22 -

North-, or (𝑖𝑖𝑖) South-South RTA, respectively, in period 𝑡 regardless of the bloc’s formal

status of integration, all zero otherwise. In addition, we interact each of the three new dummy

variables with 𝜏𝑖𝑗𝑘𝑡that allows us to estimate the impact on the probability of AD measure use

of a variation in tariff margins granted by 𝑖 towards 𝑗 conditioned for the case that both are

fellow member countries in one of the above introduced three types of an RTA in period 𝑡.10

As can be seen in Table 3, regarding our control and trade policy variables, previous find-

ings are confirmed across all types of RTAs. With respect to the interaction term, we apply

the abovementioned procedure in order to compute discrete interaction effects. Here, we find

that the group contrast is statistically significant only for those RTAs which involve a “South”

member. This may be interpreted in two ways. On the one hand, these results could admitted-

ly arise from the circumstance that the new heavy users of AD are in fact mostly emerging

and developing economies while trade relationships —including the implementation of AD

measures— have already been established between “North” countries. On the other hand, it

could likewise imply that AD measures are initiated because of competitiveness concerns,

either between developing or emerging economies that try to establish or maintain their posi-

tion in global markets producing a similar set of goods, or between “North” and “South”

countries where the “North” fears underpriced goods from the “South” and the “South” miss-

ing competitiveness towards the “North” in the trade of capital goods. This effect also sup-

ports our hypothesis that regional economic integration leads to increased competitiveness

pressure that is being met by substituting tariff reductions with other non-tariff measures, such

as AD.


10 For reasons of clarity, we only report estimation results including country-year- and commodity effects and

tariff treatment variables in levels. Those including country-, year-, and commodity effects or tariff treatment

variables in change definitions confirm reported findings and can be provided upon request.

- 23 -

7. Conclusion

In this paper, we examine the interplay of regional economic integration and the use of bilat-

eral AD measures. For the countries and commodities included, our empirical analysis brings

three central findings to light: (𝑖) We find that RTAs generally reduce the likelihood of AD

activity among integration partners. (𝑖𝑖) In addition, an improvement in tariff treatment of

trading partners —regardless of whether expressed in absolute or as a tariff margin terms—

generally leads to a lower likelihood of bilateral AD activity. (𝑖𝑖𝑖) Although the net effect of

regional economic integration on AD use may still be negative, the interaction of both events

indicates that a higher tariff margin that fellow integration partners reciprocally grant them-

selves compared to third country trading partners leads to a higher likelihood of bilateral AD

activity than a tariff margin in equal amounts among non-integration trading partners. The

latter effect seems to be primarily driven by those RTAs with a participation of “South” coun-

tries.

These results could be interpreted as an indication for trade policy substitution in RTAs.

They may likewise well be seen as another piece in the puzzle regarding the “stepping-stone”

versus “stumbling-block” discussion of RTAs. In this regard, it should be noted that in com-

parison to tariffs, other —non-tariff— protectionist measures are often more difficult to quan-

tify and might therefore have even unexpectedly strong trade depressing effects. With this,

RTAs may be attributed only sparse potential in paving the way to multilateral trade liberali-

zation efforts in view of limited trade creation prospects. In a more general context one could

argue that in order to reap the full benefits of regional free trade, mutual tariff concessions

among integration partners must not be offset by the simultaneous implementation of other —

non-tariff— trade barriers.

Based on our findings further research may be motivated. The questions that we consider

most important to address involve (𝑖) “why does this effect appear to be pronounced in partic-

- 24 -

ular in RTAs with “South” country involvement?”, and (𝑖𝑖) “what are the implications with

regard to the design of future and existing RTAs?”.

References

AGGARWAL, A. (2004), Macro Economic Determinants of Antidumping: A Comparative

Analysis of Developed and Developing Countries, World Development 32(6): 1043-57.

AHN, D. and SHIN, W. (2011), Analysis of Anti-dumping Use in Free Trade Agreements,

Journal of World Trade 45(2): 431-56.

AI, C. and NORTON, E.C. (2003), Interaction Terms in Logit and Probit Models, Economics

Letters 80(1): 123-29.

ANDERSON, S.P. and SCHMITT, N. (2003), Nontariff Barriers and Trade Liberalization, Eco-

nomic Inquiry 41(1): 80-97.

BEVERELLI, C., BOFFA, M. and KECK, A. (2014), Trade Policy Substitution: Theory and Evi-

dence from Specific Trade Concerns, WTO Staff Working Paper No. ERSD-2014-18.

BIENEN, D., CIURIAK, D. and PICARELLO, T. (2014), Does Antidumping Address “Unfair”

Trade?, The International Trade Journal 28(3): 195-228.

BLONIGEN, B.A. (2005), The Effects of NAFTA on Antidumping and Countervailing Duty

Activity, The World Bank Economic Review 19(3): 407-24.

- 25 -

BOWN, C.P. (2015), Global Antidumping Database, The World Bank,

http://econ.worldbank.org/ttbd/gad, accessed October, 2015 (various dates).

BOWN, C.P. and MCCULLOCH, R. (2012), Antidumping and Market Competition, Policy Re-

search Working Paper 6197, The World Bank.

BOWN, C.P. and TOVAR, P. (2016), Preferential Liberalization, Antidumping, and Safeguards:

Stumbling Block Evidence from MERCOSUR, Economics & Politics 28(3): 262-94.

BRANDER, J. and Krugman, P. (1983), A ’Reciprocal Dumping’ Model of International Trade,

Journal of International Economics 15(3-4): 313-21.

CARRÈRE, C., DE MELO, J. and TUMURCHUDUR, B. (2010), Disentangling Market Access Ef-

fects of Preferential Trading Arrangements with an Application for ASEAN Members un-

der an ASEAN–EU FTA, The World Economy 33(1): 42-59.

CHANDRA, P. (2016), Impact of Temporary Trade Barriers: Evidence from China, China Eco-

nomic Review 38: 24-48.

COPELAND, B.R. (1990), Strategic Interaction Among Nations: Negotiable and Non-

negotiable Trade Barriers, Canadian Journal of Economics 23(1): 84-108.

DURLING, J.P. and PRUSA, T.J. (2006), The Trade Effects Associated with an Antidumping

Epidemic: The Hot-Rolled Steel Market, 1996-2001, European Journal of Political Econ-

omy 22(3): 675-95.

FEINBERG, R.M. and REYNOLDS, K.M. (2007), Tariff Liberalisation and Increased Administra-

tive Protection: Is There a Quid Pro Quo?, The World Economy 30(6): 948-61.

GREENE, W. (2004), The Behaviour of the Maximum Likelihood Estimator of Limited De-

pendent Variables Models in the Presence of Fixed Effects, Econometrics Journal 7(1):

98-119.

KETTERER, T.D. (2016), EU Anti-dumping and Tariff Cuts: Trade Policy Substitution?, The

World Economy 39(5): 576-96.

http://econ.worldbank.org/ttbd/gad/

- 26 -

MOORE, M.O. and ZANARDI, M. (2009), Does Antidumping Use Contribute to Trade Liberali-

zation in Developing Countries?, Canadian Journal of Economics 42(2): 469-95.

MOORE, M.O. and ZANARDI, M. (2011), Trade Liberalization and Antidumping: Is There a

Substitution Effect?, Review of Development Economics 15(4): 601-19.

NIELS, G. and TEN KATE, A. (2006), Antidumping Policy in Developing Countries: Safety

Valve or Obstacle to Free Trade?, European Journal of Political Economy 22(3): 618-38.

PRUSA, T.J. (2005), Anti-dumping: A Growing Problem in International Trade, The World

Economy 28(5): 683-700.

PRUSA, T.J. (2011), Trade Remedy Provisions, in: CHAUFFOUR, J.-P. and MAUR, J.-C. (eds.),

Preferential Trade Agreement Policies for Development: A Handbook, The World Bank:

Washington, D.C.

PRUSA, T.J. and TEH, R. (2010), Protection Reduction and Diversion: PTAs and the Incidence

of Antidumping Disputes, NBER Working Paper 16276.

REY, J.-D. (2012), Antidumping Regional Regimes and the Multilateral Trading System: Do

Regional Antidumping Regimes Make a Difference?, WTO Staff Working Paper No.

ERSD-2012-22.

STAIGER, R.W. and WOLAK, F.A. (1994), Measuring Industry-Specific Protection: Antidump-

ing in the United States, Brookings Papers on Economic Activity: Microeconomics: 51-

118.

STIGLITZ, J.E. (1997), Dumping on Free Trade: The U.S. Import Trade Laws, Southern Eco-

nomic Journal 64(2): 402-24.

TE VELDE, D.W. and BEZEMER, D. (2008), Regional Integration and Foreign Direct Invest-

ment in Developing Countries, Transnational Corporations 15(2): 41-70.

UN COMTRADE (2015), International Trade Statistics Database, http://comtrade.un.org, ac-

cessed October, 2015 (various dates).

- 27 -

VANDENBUSSCHE, H. and ZANARDI, M. (2008), What Explains the Proliferation of Antidump-

ing Laws?, Economic Policy 23(53): 98-103.

WORLD BANK (2015a), World Development Indicators, http://data.worldbank.org/data-

catalog/world-development-indicators, accessed October, 2015 (various dates).

WORLD BANK (2015b), World Integrated Trade Solution, http://wits.worldbank.org/WITS,

accessed October, 2015 (various dates).

WORLD TRADE ORGANIZATION (2015), Regional Trade Agreement Information System

(RTA-IS), http://rtais.wto.org, accessed October, 2015 (various dates).

Figure 1. Temporal Evolution of Regional Trade Agreements

and Antidumping Activity, 1991 – 2014

Note: AD data based on countries considered for empirical analysis (see section 4 for details).

Data Sources: WTO (2015); Bown (2015).

0

50

100

150

200

250

300

0

200

400

600

800

1000

1200

1400

1600

1800

2000

Number of AD Measures in Force (Left-hand Scale) Number of Intra-bloc AD Measures in Force (Left-hand Scale)

Accumulated Number of RTAs in Force (Right-hand Scale)

- 28 -

Figure 2. Box-and-Whisker Plot of Relative Tariff Treatment

by Value of RTA Dummy Variable

Notes: Upper panel includes outside values whereas lower panel excludes outside values. Outside values are

displayed as red dots. Computation of relative tariff treatment based on countries considered for empirical analy-

sis (see section 4 for details).

Data Source: World Bank (2015b).

𝑅𝑇

𝐴𝑖𝑗

𝑡

𝜏𝑖𝑗𝑘𝑡

- 29 -

- 30 -

Table 1. Correlation Matrix of Explanatory Variables

VARIABLES 𝐴𝐷𝑖𝑗𝑡𝑘 ln (pc_𝐺𝐷𝑃𝑖𝑡−1) Δ 𝐺𝐷𝑃𝑖𝑡−1 𝐶𝐴𝐵𝑖𝑡−1 Δ 𝐼𝑀𝑖𝑗𝑘𝑡−1 ln (𝐼𝑀_𝑆𝐻𝐴𝑅𝐸𝑖𝑗𝑘𝑡−1) ln (1 + 𝑡𝑖𝑗𝑘𝑡−1) ln (1 + 𝜏𝑖𝑗𝑘𝑡−1) 𝑅𝑇𝐴𝑖𝑗𝑡−1 𝑁𝑂_𝑅𝑈𝐿𝐸𝑆𝑖𝑗𝑡−1

𝐴𝐷𝑖𝑗𝑡𝑘 1

ln (pc_𝐺𝐷𝑃𝑖𝑡−1) 0.0166 1

Δ 𝐺𝐷𝑃𝑖𝑡−1 0.0021 -0.2773 1

𝐶𝐴𝐵𝑖𝑡−1 -0.0244 -0.0343 0.1359 1

Δ 𝐼𝑀𝑖𝑗𝑘𝑡−1 -0.0002 -0.0009 0.0008 0.0005 1

ln (𝐼𝑀_𝑆𝐻𝐴𝑅𝐸𝑖𝑗𝑘𝑡−1) 0.0606 -0.1314 0.029 0.0334 0.0015 1

ln (1 + 𝑡𝑖𝑗𝑘𝑡−1) 0.0048 -0.3521 0.081 0.038 0.0005 0.0284 1

ln (1 + 𝜏𝑖𝑗𝑘𝑡−1) -0.0186 0.029 0.002 0.0071 -0.0022 0.0288 -0.3681 1

𝑅𝑇𝐴𝑖𝑗𝑡−1 -0.0183 -0.0013 -0.0117 0.0117 -0.0004 0.0966 -0.1765 0.2813 1

𝑁𝑂_𝑅𝑈𝐿𝐸𝑆𝑖𝑗𝑡−1 -0.0088 -0.0023 0.016 0.0899 -0.0001 0.0089 -0.0669 0.0408 0.3414 1

- 31 -

Table 2. Intra-bloc Tariff Margins of Selected Regional Trade Agreements

(Aggregated Trade), 1991 – 2014

Name of RTA

(year of entry into force)

Average 𝜏𝑖𝑗𝑘𝑡 among fel-

low integration partners

since entry into force

High of 𝜏𝑖𝑗𝑘𝑡 among fellow

integration partners since

entry into force (year)

Low of 𝜏𝑖𝑗𝑘𝑡 among fellow

integration partners since

entry into force (year)

ANDEAN Community (1988) 0.013 0.081 (2006) -0.046 (1994)

ANZCERTA (1991) 0.030 0.043 (2003) 0.012 (2011)

ASEAN-FTA (1992) 0.012 0.036 (1997) -0.008 (1999)

Canada-Chile (1997) 0.002 0.008 (2014) -0.011 (2006)

EU-Chile (2003) 0.012 0.023 (2014) -0.006 (2004)

Korea-Chile (2004) 0.018 0.029 (2006) 0.006 (2005)

NAFTA (1994) 0.014 0.036 (2002) -0.001 (1998)

MERCOSUR (1991) 0.045 0.104 (1995) 0.0002 (2002)

USA-Chile (2004) 0.007 0.016 (2014) -0.006 (2004)

Notes: Computed with tariff data of country sample considered for empirical analysis (see section 4 for de-

tails). Own calculation based on data from WITS (2015).

- 32 -

Table 3. Average Marginal Probability Effects on Antidumping Measure Use

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

VARIABLES ———— ———— ———— ————

𝑅𝑇𝐴𝑖𝑗𝑡−1 -0.00296*** -0.00361*** -0.00346*** -0.00414***

(0.000453) (0.000562) (0.000440) (0.000540)

ln (1 + 𝑡𝑖𝑗𝑘𝑡−1) 0.00664*** 0.00365 — —

(0.00167) (0.00228)

Δ ln (1 + 𝑡𝑖𝑗𝑘𝑡−1) — — 0.00424*** 0.00228

(0.00109) (0.00141)

Base Effect:

ln (1 + 𝜏𝑖𝑗𝑘𝑡−1) -0.0206*** -0.0277*** — —

(0.00305) (0.00432)

Δ ln (1 + 𝜏𝑖𝑗𝑘𝑡−1) — — -0.00199* -0.00150

(0.000903) (0.00149)

Integration Effect:

ln (1 + 𝜏𝑖𝑗𝑘𝑡−1) 0.0238*** 0.0306*** — —

(at 𝑅𝑇𝐴𝑖𝑗𝑡−1: 1 vs 0) (0.00482) (0.00634)

Δ ln (1 + 𝜏𝑖𝑗𝑘𝑡−1) — — 0.00465** 0.00207

(at 𝑅𝑇𝐴𝑖𝑗𝑡−1: 1 vs 0)

(0.00158) (0.00209)

Δ 𝐼𝑀𝑖𝑗𝑘𝑡−1 -2.92e-06* -3.17e-06* -2.92e-06* -3.12e-06*

(1.26e-06) (1.31e-06) (1.27e-06) (1.30e-06)

ln (𝐼𝑀_𝑆𝐻𝐴𝑅𝐸𝑖𝑗𝑘𝑡−1) 0.00227*** 0.00256*** 0.00227*** 0.00255***

(8.59e-05) (9.87e-05) (8.61e-05) (9.89e-05)

𝑁𝑂_𝑅𝑈𝐿𝐸𝑆𝑖𝑗𝑡−1 0.00505*** 0.00643*** 0.00524*** 0.00674***

(0.000806) (0.00112) (0.000812) (0.00112)

ln (pc_𝐺𝐷𝑃𝑖𝑡−1) 0.0191*** — 0.0189*** —

(0.00103)

(0.00103)

Δ 𝐺𝐷𝑃𝑖𝑡−1 1.24e-05 — -6.94e-06 —

(2.33e-05)

(2.30e-05)

𝐶𝐴𝐵𝑖𝑡−1 3.67e-05 — 5.54e-05* —

(2.78e-05)

(2.78e-05)

Observations 2,859,415 2,473,967 2,843,554 2,459,434

Pseudo-R2 0.279 0.271 0.278 0.27

Log likelihood -90,259.36 -86,425.33 -90,009.56 -86,291.30

Type of fixed effects:

Country- (𝛼𝑖, 𝛼𝑗) Yes No Yes No

Country-year- (𝜂𝑖𝑡, 𝜇𝑗𝑡) No Yes No Yes

Year- (𝜆𝑡) Yes No Yes No

Two-digit commodity- (𝜑𝑛) Yes Yes Yes Yes

Notes: Robust, clustered (at the country-pair-commodity level) standard errors in parentheses. Interaction effects are

computed as contrasts of average marginal probability effects using Stata’s “margins’ ” contrast operator (Version

14.0, StataCorp). Asterisks denote the level of statistical significance with *** p<0.001, ** p<0.01, * p<0.05, † p<0.1.

In the case of the integration effect, asterisks denote the level of statistical significance of the 𝜒[1]2 -test statistic of a

comparison of average marginal probability effects across reference groups.

- 33 -

Table 4. Average Marginal Probability Effects on Antidumping Measure Use

by Type of Regional Trade Agreements

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

VARIABLES —North-South— —North-North— —South-South— ——All——

𝑅𝑇𝐴𝑖𝑗𝑡−1𝑁𝑆 -0.00324*** — — -0.00331***

(0.000739)

(0.000742)

𝑅𝑇𝐴𝑖𝑗𝑡−1𝑁𝑁 — -0.00411** — -0.00451**

(0.00156)

(0.00155)

𝑅𝑇𝐴𝑖𝑗𝑡−1𝑆𝑆 — — -0.00396*** -0.00393***

(0.000927) (0.000924)

ln (1 + 𝑡𝑖𝑗𝑘𝑡−1) 0.00428† 0.00434

† 0.00356 0.00348

(0.00224) (0.00225) (0.00229) (0.00229)

Base Effect:

ln (1 + 𝜏𝑖𝑗𝑘𝑡−1) -0.0232*** -0.0248*** -0.0295*** -0.0283***

(0.00417) (0.00376) (0.00397) (0.00432)

Integration Effect:

ln (1 + 𝜏𝑖𝑗𝑘𝑡−1) 0.0135** — — 0.0171**

(at 𝑅𝑇𝐴𝑖𝑗𝑡−1𝑁𝑆 : 1 vs 0) (0.00678)

(0.00670)

ln (1 + 𝜏𝑖𝑗𝑘𝑡−1) — 0.000507 — 0.00321

(at 𝑅𝑇𝐴𝑖𝑗𝑡−1𝑁𝑁 : 1 vs 0)

(0.0108)

(0.0105)

ln (1 + 𝜏𝑖𝑗𝑘𝑡−1) — — 0.0480*** 0.0466***

(at 𝑅𝑇𝐴𝑖𝑗𝑡−1𝑆𝑆 : 1 vs 0)

(0.00636) (0.006415)

Δ 𝐼𝑀𝑖𝑗𝑘𝑡−1 -3.14e-06* -3.07e-06* -3.13e-06* -3.19e-06*

(1.30e-06) (1.28e-06) (1.29e-06) (1.31e-06)

ln (𝐼𝑀_𝑆𝐻𝐴𝑅𝐸𝑖𝑗𝑘𝑡−1) 0.00254*** 0.00247*** 0.00248*** 0.00255***

(9.83e-05) (9.71e-05) (9.74e-05) (9.88e-05)

𝑁𝑂_𝑅𝑈𝐿𝐸𝑆𝑖𝑗𝑡−1 0.00651*** 0.00400*** 0.00427*** 0.00669***

(0.00118) (0.00108) (0.00106) (0.00117)

Observations 2,473,967 2,473,967 2,473,967 2,473,967

Pseudo-R2 0.271 0.27 0.271 0.272

Log likelihood -86,503.37 -86,556.24 -86,493.82 -86,405.17


Country- (𝛼𝑖, 𝛼𝑗) No No No No

Country-year- (𝜂𝑖𝑡, 𝜇𝑗𝑡) Yes Yes Yes Yes

Year- (𝜆𝑡) No No No No







- 34 -

Appendix 1. Country Sample

Argentina, Australia, Brazil, Canada, Chile (1995), China (1997), Colombia, Costa Rica (1996), Ecuador (1998),

European Union, India (1992), Indonesia (1996), Israel, Jamaica (2000), Japan, Republic of Korea, Malaysia

(1995), Mexico, New Zealand (1995), Pakistan (2002), Paraguay (1999), Peru (1992), Philippines (1994), South

Africa, Taiwan, Thailand (1996), Trinidad and Tobago (1997), Turkey, United States, Uruguay (1997), Venezue-

la (1992).

Notes: The European Union is treated as a single country. Its evolutionary enlargement of member states is con-

sidered (see main text for details). Unless otherwise stated in parentheses, the initial year of the respective country

in our sample is 1991.

- 35 -

Appendix 2. List of Regional Trade Agreements Covered by Sample

North-North (14) North-South (39) South-South (17)

Australia-Chile (WTO-based Rules)

ASEAN-Australia-New Zealand (No

Rules)

ASEAN-FTA (AFTA) (WTO-based

Rules)

ANZCERTA (Disallowed) ASEAN-Japan (No Rules) ASEAN-China (WTO-based Rules)

Canada-Chile (Disallowed) ASEAN-Rep. Korea (No Rules) ASEAN-India (No Rules)

Canada-Israel (WTO-based Rules) Canada-Columbia (Specific Rules) China-Costa Rica (Specific Rules)

Chile-Japan (No Rules) Canada-Costa Rica (Specific Rules) Group of Three (Specific Rules)

EU-Chile (WTO-based Rules) Canada-Peru (WTO-based Rules) Costa Rica-Peru (WTO-based Rules)

EU-Israel (WTO-based Rules) Chile-China (WTO-based Rules) India-Malaysia (Specific Rules)

EU-Rep. Korea (Specific Rules) Chile-Columbia (WTO-based Rules)

Mexico-Central America (Specific

Rules)

Rep. Korea-Chile (WTO-based

Rules) Chile-Costa Rica (WTO-based Rules) Mexico-Uruguay (Specific Rules)

Rep. Korea-USA (Specific Rules) Chile-Malaysia (WTO-based Rules) Pakistan-China (WTO-based Rules)

Trans-Pacific Strategic Economic

Partnership (WTO-based Rules) Chile-Mexico (WTO-based Rules)

Pakistan-Malaysia (WTO-based

Rules)

USA-Australia (WTO-based Rules) China-New Zealand (Specific Rules) Peru-China (Specific Rules)

USA-Chile (WTO-based Rules)

CAFTA-Dom. Rep. (WTO-based

Rules) Peru-Mexico (Specific Rules)

USA-Israel (WTO-based Rules)

EU-CARIFORUM States (Specific

Rules) SAFTA (Specific Rules)

EU-Central America (Specific Rules)

ANDEAN Community (Specific

Rules)

EU-Columbia-Peru (Specific Rules)

CARICOM and Common Market

(Specific Rules)

MERCOSUR (Specific Rules)

EU-Mexico (WTO-based Rules)

EU-South Africa (WTO-based Rules)

EU-Turkey (Specific Rules)

India-Japan (WTO-based Rules)

Israel-Mexico (WTO-based Rules)

Japan-Malaysia (No Rules)

Japan-Mexico (WTO-based Rules)

Japan-Peru (WTO-based Rules)

Japan-Philippines (WTO-based

Rules)

Japan-Thailand (WTO-based Rules)

Rep. Korea-India (Specific Rules)

Rep. Korea-Turkey (Specific Rules)

Malaysia-Australia (Specific Rules)

New Zealand-Malaysia (Specific

Rules)

NAFTA (WTO-based Rules)

Peru-Chile (Specific Rules)

Peru-Rep. Korea (Specific Rules)

Thailand-Australia (WTO-based

Rules)

Thailand-New Zealand (WTO-based

Rules)

Turkey-Chile (WTO-based Rules)

Turkey-Israel (Specific Rules)

USA-Columbia (WTO-based Rules)

USA-Peru (WTO-based Rules)

Notes: ASEAN-FTA is treated as a South-South RTA as its only North-member country, i.e. Singapore, is not in-

cluded in our sample. According to the World Bank’s Atlas Method, Chile is classified as a high-income country

- 36 -

since 2011. In order to avoid changes in the type of RTAs as defined in section 6 including Chile, we treat Chile as a

North country throughout. Parentheses give information on the AD policy framework of the respective RTA where

(𝑖) “Disallowed” means that the use of AD measures among RTA member countries is entirely disallowed; (𝑖𝑖) “Spe-

cific Rules” means that the use of AD measures is generally allowed among RTA member countries but must be in

accordance with provisions constituted in the agreement (mainly replicates of GATT/WTO); (𝑖𝑖𝑖) “WTO-based

Rules” means that the use of AD measures is generally allowed among RTA member countries and the agreement

confirms/reaffirms rights and obligations constituted by GATT/WTO; (𝑖𝑣) “No Rules” means that the use of AD

measures is generally allowed among RTA member countries (under GATT/WTO guidelines if member) and there is

no explicit mention of provisions in agreement.

- 37 -

Appendix 3. Variable Summary Statistics

VARIABLE Observations Mean Std. Dev. Min Max

𝐴𝐷𝑖𝑗𝑡𝑘 4,479,186 0.007 0.081 0 1

𝑅𝑇𝐴𝑖𝑗𝑡−1 3,589,849 0.184 0.387 0 1

𝑅𝑇𝐴𝑖𝑗𝑡−1𝑁𝑆 3,589,849 0.082 0.274 0 1

𝑅𝑇𝐴𝑖𝑗𝑡−1𝑁𝑁 3,589,849 0.031 0.175 0 1

𝑅𝑇𝐴𝑖𝑗𝑡−1𝑆𝑆 3,589,849 0.079 0.27 0 1

ln (1 + 𝑡𝑖𝑗𝑘𝑡−1) 3,536,340 0.074 0.082 0 2.256

Δ ln (1 + 𝑡𝑖𝑗𝑘𝑡−1) 3,027,256 -0.003 0.036 -0.884 7.766

ln (1 + 𝜏𝑖𝑗𝑘𝑡−1) 3,412,488 -0.002 0.04 -1.957 1.375

Δ ln (1 + 𝜏𝑖𝑗𝑘𝑡−1) 2,872,292 0.0005 0.030 -0.820 5.496

Δ 𝐼𝑀𝑖𝑗𝑘𝑡−1 2,920,830 76.11 32942.79 -1.0 3.49e+07

ln (𝐼𝑀_𝑆𝐻𝐴𝑅𝐸𝑖𝑗𝑘𝑡−1) 3,465,811 -0.391 3.123 -17.919 4.605

𝑁𝑂_𝑅𝑈𝐿𝐸𝑆𝑖𝑗𝑡−1 3,589,849 0.026 0.160 0 1

ln (pc_𝐺𝐷𝑃𝑖𝑡−1) 3,589,849 9.70 0.728 7.576 10.845

Δ 𝐺𝐷𝑃𝑖𝑡−1 3,589,849 3.717 3.462 -13.127 21.829

𝐶𝐴𝐵𝑖𝑡−1 3,589,849 0.146 4.874 -15.928 38.787

- 38 -

Appendix 4. Variable Description and Data Source

VARIABLE Description Data Source

𝐴𝐷𝑖𝑗𝑡𝑘

Binary variable that equals unity whenever an antidumping measure

applied by importing country 𝑖 against country 𝑗 is in force in period 𝑡 at

the Harmonised System (HS) four-digit commodity level 𝑘, zero other-

wise.

Bown (2015)

𝑅𝑇𝐴𝑖𝑗𝑡−1 Binary variable set to unity if both trading partners were member coun-

tries of the same RTA in the previous period, zero otherwise.

Own computation based on

WTO (2015)

𝑅𝑇𝐴𝑖𝑗𝑡−1𝑁𝑆

Binary variable set to unity if both trading partners were member coun-

tries of the same North-South RTA in the previous period, zero otherwise.


WTO (2015)

𝑅𝑇𝐴𝑖𝑗𝑡−1𝑁𝑁


tries of the same North-North RTA in the previous period, zero otherwise.


WTO (2015)

𝑅𝑇𝐴𝑖𝑗𝑡−1𝑆𝑆


tries of the same South-South RTA in the previous period, zero otherwise.


WTO (2015)

𝑡𝑖𝑗𝑘𝑡−1 Annual trade-weighted effectively applied tariff rate of importing 𝑖 to-

wards exporting 𝑗 at the HS four-digit commodity level in the previous

period.

World Bank (2015b)

𝜏𝑖𝑗𝑘𝑡−1 Annual relative tariff treatment of importing 𝑖 towards exporting 𝑗 at the

HS four-digit commodity level in the previous period (see text for calcula-

tion details).

World Bank (2015b)

𝜏𝑖𝑗𝑡𝑘−1𝑅𝑇𝐴

Interaction term between a binary variable set to unity if both trading

partners are member countries of the same RTA and annual relative tariff

treatment of importing 𝑖 towards exporting 𝑗 at the HS four-digit com-

modity level, lagged by one period.


WTO (2015) and World Bank

(2015b)

𝜏𝑖𝑗𝑡𝑘−1𝑁𝑆


partners were member countries of the same North-South RTA and annual

relative tariff treatment of importing 𝑖 towards exporting 𝑗 at the HS four-

digit commodity level, lagged by one period.



(2015b)

𝜏𝑖𝑗𝑡𝑘−1𝑁𝑁


partners were member countries of the same North-North RTA and annual





(2015b)

- 39 -

𝜏𝑖𝑗𝑡𝑘−1𝑆𝑆


partners were member countries of the same South-South RTA and annual





(2015b)

Δ 𝐼𝑀𝑖𝑗𝑘𝑡−1 Percentage change in annual imports of 𝑖 from 𝑗 at the HS four-digit

commodity level compared to previous period, lagged by one period. UN Comtrade (2015)

𝐼𝑀_𝑆𝐻𝐴𝑅𝐸𝑖𝑗𝑘𝑡−1 Share of 𝑗 in 𝑖’s annual total imports at the HS four-digit commodity level

in the previous period. UN Comtrade (2015)

𝑁𝑂_𝑅𝑈𝐿𝐸𝑆𝑖𝑗𝑡−1


tries of the same RTA in the previous period and the RTA did not have a

legal framework regarding the use intra-bloc antidumping measures, zero

otherwise.

Rey (2012) and legal frame-

works of various agreements

(WTO, 2015)

pc_𝐺𝐷𝑃𝑖𝑡−1 GDP per capita of 𝑖 at constant PPP Int’l US-Dollars in the previous

period. World Bank (2015a)

Δ 𝐺𝐷𝑃𝑖𝑡−1 Aggregate GDP growth of 𝑖 in the previous period. World Bank (2015a)

𝐶𝐴𝐵𝑖𝑡−1 Current account balance as a percentage of GDP of 𝑖 in the previous

period. World Bank (2015a)

- 40 -

Appendix 5. Average Marginal Probability Effects on Antidumping Measure Use

(Interaction Term Variation)

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

VARIABLES ———— ———— ———— ————

𝑅𝑇𝐴𝑖𝑗𝑡−1 -0.00200*** -0.00228*** -0.00346*** -0.00414***

(0.000496) (0.000618) (0.000440) (0.000540)

ln (1 + 𝜏𝑖𝑗𝑘𝑡−1) -0.0154*** -0.0205*** — —

(0.00287) (0.00393)

Δ ln (1 + 𝜏𝑖𝑗𝑘𝑡−1) — — -0.00199* -0.00150

(0.000903) (0.00149)

Base Effect:

ln (1 + 𝑡𝑖𝑗𝑘𝑡−1) 0.00850*** 0.00652** — —

(0.00169) (0.00223)

Δ ln (1 + 𝑡𝑖𝑗𝑘𝑡−1) — — 0.00424*** 0.00228

(0.00109) (0.00141)

Integration Effect:

ln (1 + 𝑡𝑖𝑗𝑘𝑡−1) -0.0129*** -0.0167*** — —

(at 𝑅𝑇𝐴𝑖𝑗𝑡−1: 1 vs 0) (0.00312) (0.00412)

Δ ln (1 + 𝑡𝑖𝑗𝑘𝑡−1) — — -0.00435* 0.000271

(at 𝑅𝑇𝐴𝑖𝑗𝑡−1: 1 vs 0)

(0.00176) (0.00217)

Δ 𝐼𝑀𝑖𝑗𝑘𝑡−1 -2.89e-06* -3.13e-06* -2.92e-06* -3.12e-06*

(1.26e-06) (1.30e-06) (1.27e-06) (1.30e-06)

ln (𝐼𝑀_𝑆𝐻𝐴𝑅𝐸𝑖𝑗𝑘𝑡−1) 0.00228*** 0.00256*** 0.00227*** 0.00255***

(8.59e-05) (9.87e-05) (8.61e-05) (9.89e-05)

𝑁𝑂_𝑅𝑈𝐿𝐸𝑆𝑖𝑗𝑡−1 0.00491*** 0.00622*** 0.00524*** 0.00674***

(0.000802) (0.00111) (0.000812) (0.00112)

ln (pc_𝐺𝐷𝑃𝑖𝑡−1) 0.0195*** — 0.0189*** —

(0.00104)

(0.00103)

Δ 𝐺𝐷𝑃𝑖𝑡−1 1.08e-05 — -6.94e-06 —

(2.34e-05)

(2.30e-05)

𝐶𝐴𝐵𝑖𝑡−1 3.65e-05 — 5.54e-05* —

(2.78e-05)

(2.78e-05)

Observations 2,859,415 2,473,967 2,843,554 2,459,434

Pseudo-R2 0.279 0.271 0.278 0.27

Log likelihood -90,262.19 -86,420.61 -90,009.56 -86,291.35


Country- (𝛼𝑖, 𝛼𝑗) Yes No Yes No

Country-year- (𝜂𝑖𝑡, 𝜇𝑗𝑡) No Yes No Yes

Year- (𝜆𝑡) Yes No Yes No





- 41 -



false friends? empirical evidence on trade policy ... · pdf filefalse friends? empirical...

Documents