DOI 10.1007/s11558-013-9179-8
The design of international trade agreements:
Introducing a new dataset
Andreas D¨ur·Leonardo Baccini·Manfred Elsig
© Springer Science+Business Media New York 2013
Abstract Preferential trade agreements (PTAs) have been proliferating for the last
twenty years. A large literature has studied various aspects of this phenomenon.
Until recently, however, many large-N studies have paid only scant attention to variation
across PTAs in terms of content and design. Our contribution to this literature
is a new dataset on the design of trade agreements that is the most comprehensive
in terms of both variables coded and agreements covered. We illustrate the dataset’s
usefulness in re-visiting the questions if and to what extent PTAs impact trade flows.
The analysis shows that on average PTAs increase trade flows, but that this effect is
largely driven by deep agreements. In addition, we provide evidence that provisions
that tackle behind-the-border regulation matter for trade flows. The dataset’s contribution
is not limited to the PTA literature, however. Broader debates on topics such
as institutional design and the legalization of international relations will also benefit
from the novel data.
Keywords Preferential trade agreements · New regionalism · Institutional design ·
Dataset · Trade flows · Gravity model
JEL Classifications F140 · F150 · F5
Electronic supplementary material The online version of this article
(doi:10.1007/s11558-013-9179-8) contains supplementary material, which is available
to authorized users.
A. D¨ur
University of Salzburg, Salzburg, Austria
L. Baccini
LSE, London, United Kingdom
M. Elsig ( )
University of Bern, World Trade Institute, Hallerstrasse 6, Bern, CH-3012, Switzerland
e-mail: manfred.elsig@wti.org
Rev Int Organ (2014) 9:353–375
Published online: 31 October 2013
A. D¨ur et al.
1 Introduction
Since the early 1990s, countries across the world have signed a large number of
preferential trade agreements (PTAs). PTAs thus have become an important instrument
of foreign economic policy-making for nearly all governments. A substantial
literature has emerged that studies various aspects of this phenomenon, known as
new regionalism (Mansfield and Milner 1999). Until recently, however, much of this
literature failed to take account of the differences in substance and content across
PTAs. Some agreements, such as the European Union (EU) or the North American
Free Trade Agreement (NAFTA), are very broad and contain provisions that
lead to a far-reaching liberalization of markets. By contrast, other agreements are
narrow and exhibit modest commitments. Only over the last few years, have some
scholars started to pay greater attention to the scope and depth of these agreements
(Estevadeordal et al. 2009; Haftel 2010; Hicks and Kim 2012; Kucik 2012; Mansfield
and Milner 2012).
We contribute to this emerging literature by presenting a new dataset on the design
of trade agreements (DESTA) that to our knowledge is the most comprehensive in
terms of both items coded and number of agreements included. The World Trade
Organization (WTO)’s dataset on the “anatomy” of PTAs, for example, contains
information on 131 agreements with respect to 52 items (World Trade Organization
2011). The chapters in Estevadeordal et al. (2009) rely on an even smaller number of
agreements. For our dataset, by contrast, we coded 587 agreements for more than 100
items. For instance, our dataset contains twelve variables coding intellectual property
rights; the WTO dataset one. The scope and coverage of DESTA is a major asset
when tackling questions where looking only at the most prominent agreements leads
to selection bias. This applies to both studies of the formation of PTAs and analyses
of the consequences of PTAs.
To illustrate DESTA’s potential, we revisit the literature on the PTA–trade nexus,
that is, the questions if and to what extent PTAs impact trade flows. Our analysis
produces three major findings: first, we present the so far most sophisticated operationalization
of the concept of depth of agreements and show that deep agreements
matter significantly more than shallow ones. Second, we corroborate findings that
PTAs have an anticipatory, short-term and long-term effect on trade flows. Third, not
only tariff cuts, but also other market access and trade-related provisions in PTAs
concerning topics such as investments and intellectual property rights matter for trade
flows. In the conclusion we look beyond the literature on the PTA–trade nexus and
suggest that the dataset is of much broader relevance for debates in International
Relations, including on the effects of legalization in international politics and the
design of international institutions.
2 The political economy of PTAs
With the surge of the new regionalism in the early 1990s, a vibrant literature on PTA
formation developed (for reviews of this literature, see Mansfield and Milner 1999;
World Trade Organization 2011). Broadly, two types of research interests have dom-
354
The design of international trade agreements: Introducing a new dataset
inated the field: first, on why countries negotiate and sign PTAs, and second, more
recently, on the potential effects of being party to one or more PTAs.
The political economy literature has offered various explanations for why states
engage in PTAs. Much work has focused on arguments drawing on competition
effects among important trading nations and/or interest group mobilization and industry
and market characteristics (Baldwin 1993; Baccini and D¨ur 2012; Chase 2003;
D¨ur 2007; Grossman and Helpman 1995; Manger 2009; Milner 1997). Other prominent
arguments stress the role of democratization, the distribution of power and
alliances, ambitions to use international trade institutions as instruments to lockin
or credibly commit to specific policies, domestic veto players within political
systems, bureaucratic interests, electoral concerns, foreign direct investments, or
forum-shopping as a result of lack of progress in multilateralism (see for example
B¨uthe and Milner 2008; Gowa and Mansfield 1993; Hollyer and Rosendorff 2012;
Maggi and Rodr´ıguez-Clare 2007; Mansfield and Milner 2012; Mansfield et al. 2002,
2007; Mansfield and Reinhardt 2003).
Over time, the focus has shifted from formation to effects. PTAs as a type
of trade institution may yield effects on a multitude of economic, political and
social phenomena. For trade economists, the trade flow implications have been the
key concern (Baier and Bergstrand 2007). Political scientists have focused on a
broader set of outcomes ranging from studying trade volatility, inducing and sustaining
domestic economic reforms, addressing behind-the-border protectionism, or
allowing for non-trade effects, such as upholding human rights protection in PTA
signatories or reducing conflicts between PTA members (see for example, HafnerBurton
2005; Rickard and Kono 2013; Mansfield and Pevehouse 2000; Mansfield
and Reinhardt 2008).
While there is thus no shortage of explanations for PTA formation and selected
outcomes, so far most of this literature suffers from lack of data on the design of
PTAs. Many studies conceptualize PTAs as a dichotomous variable, namely whether
countries sign an agreement or not, and hence treat PTAs as if they were all equal
in purpose and effect. As a proxy for design differences, some work has controlled
for different levels of trade integration (e.g., Magee 2008, customs union, free trade
agreements, partial scope agreements). Yet, few studies focus on selected design differences
(an exception is Hafner-Burton 2005), study regional specifications (Hicks
and Kim 2012) or explain functional differences in design, for example with respect
to dispute settlement (Smith 2000) and flexibility provisions (Kucik 2012).
The limited attention paid to differences in the design of trade institutions is
problematic given that PTAs clearly vary in terms of overall ambitions and commitments
reflected in depth of concessions and flexibility clauses or opt-outs.
The logics of signing deep or shallow agreements will differ, as the former may
cause important losses for some sectors of society. And a very narrow and shallow
agreement is unlikely to have the same consequences as a broad and deep
one. Incorporating design differences in our models should thus assist us in better
understanding both why states sign PTAs and what effects these PTAs can be
expected to have. Our goal in collecting and disseminating systematic data on the
design of trade agreements is to allow the PTA literature to take into account these
differences.
355
A. D¨ur et al.
3 DESTA: Description of the dataset
We identified a total of 733 PTAs signed between 1945 and 2009 and that include
concrete steps towards the preferential liberalization of trade in goods and/or services.
The number of PTAs that we found is substantially larger than the number
of agreements covered by comparable datasets. The list maintained by the WTO
includes 356 of our agreements; 507 form part of a list held at the World Trade
Institute.1 We identified the remaining agreements via systematic searches of the
web pages of foreign ministries and other governmental institutions.2 The number
of initial memberships in these agreements is 3,659 (thus an average of 18 signed
agreements for the 201 countries covered by the dataset); moreover, in 419 instances
countries acceded after the initial signing of the agreements.
Regrettably, we have not been able to find full texts for all of the agreements
and ended up coding 587 agreements with 3,318 (initial) members for a total of 10
broad sectors of cooperation, encompassing market access, services, investments,
intellectual property rights, competition, public procurement, standards, trade remedies,
non-trade issues, and dispute settlement. For each of these sectors, we coded a
significant number of items, meaning that we have well over 100 data points for each
agreement. The coding has been carried out manually by two independent coders,
with any differences resolved by a referee for the final dataset. Inter-rater agreement
as measured by Cohen’s kappa is higher than 0.75 for nearly all variables (with
any value higher than 0.60 considered as substantial degree of agreement) and as
high as 0.85 for some variables such as those capturing the depth of services provisions.
Moreover, cross-checks against other datasets that have been put together
independently from ours have confirmed the reliability of our data.3
The resulting DESTA dataset is, to the best of our knowledge, the most ambitious
attempt at measuring the design of PTAs in terms of agreements and sectors
covered. In the following we present selected information from the dataset to illustrate
both the range of issues covered in DESTA and its potential use and application.
The discussion also offers a description of the phenomenon of regionalism that has
characterized the international political economy for some time.
Our dataset confirms the commonly held view that countries have multiplied their
efforts to sign and ratify PTAs in recent years (see Fig. 1). The tipping point was in
the early 1990s. This surge of new agreements is in particular related to the efforts of
European Union and European Free Trade Agreement (EFTA) states to stabilize trade
relations with new European democracies after the end of the Cold War. Competition
for market access motivated other countries to follow suit (Baccini and D¨ur 2012;
Baldwin and Jaimovich 2012) with many of the more recent agreements being signed
between emerging economies and developing countries. As a result, while European
1In contrast to these lists, we did not count accession agreements, and services agreements that are signed
at the same time as goods agreements, as separate PTAs. This explains why our count of agreements
notified to the WTO is smaller than the one indicated by the WTO itself.
2The additional agreements that we found were not notified to the WTO, and thus have not made it into
many of the datasets on PTAs that are based on the WTO’s PTA inventory.
3For these other datasets, see Estevadeordal et al. (2009) and World Trade Organization (2011).
356
The design of international trade agreements: Introducing a new dataset
1950 1960 1970 1980 1990 2000 2010
Year
0
100
200
300
400
CumulativenumberofPTAsinforce
0
10
20
30
40
Agreementssignedbyyear
Fig. 1 PTAs over time
states are still the top signatories of PTAs, most countries in the world now are part
of several active agreements (see Fig. 2). In fact, with the exception of Mongolia, all
but a few tiny (island) countries have signed at least one PTA since World War II.
Overall, we count 393 agreements that were in force as of 2009.
The large majority of agreements (60 %) signed since 1945 are free trade agreements
that liberalize tariffs on a majority of goods; 29 % are partial agreements
that either focus on a sector or a few products or require only limited tariff cuts;
9 % are customs unions that aspire not only to the dismantling of trade barriers and
the free flow of products but also to the establishment of a common external tariff
system; and 1 % are pure services agreements (see Fig. 3). In terms of actor constellation,
the most common are bilateral agreements (67 %, e.g., Australia–New
Zealand). Region–country agreements (e.g., EU–Mexico) account for 17 % of the
0 1−5 6−10 11−15 16−20 21−25 26+0 1−5 6−10 11−15 16−20 21−25 26+
Fig. 2 Number of PTAs signed, by country (1945–2009). The map shows the number of PTAs that a
country signed over the period 1945–2009 and that were still active as of 2009
357
A. D¨ur et al.
Fig. 3 Types and regional composition of PTAs. We show absolute frequencies to the right. We only
coded the level of integration for PTAs for which we have a full text
agreements, plurilateral agreements (e.g., NAFTA) for 14 %, and interregional agreements
(e.g., EFTA–South African Customs Union) for 2 %. As to the geographic
dimension, agreements are still predominantly regional. Nevertheless, Fig. 3 also
indicates that 27 % of treaties are concluded between countries or regions that are
located in different continents (e.g., Singapore–Chile).
Importantly, DESTA supports its own raison d’ˆetre: the data clearly reveal that
PTAs differ considerably in terms of their contents. In Fig. 4, we show the percentage
of agreements in our dataset that contain a selected number of provisions that we
coded.4 The figure shows that while most agreements foresee safeguard provisions,
only few agreements substantively regulate intellectual property rights, foreign direct
investments, or procurement by public authorities. The differences are even more
pronounced when looking at specific provisions. In competition policy, for example,
70 % of agreements stipulate that member states may not distort competition, but only
two percent foresee the establishment of a common competition authority. No fewer
than 87 percent of agreements include some type of dispute settlement provision; and
6 percent foresee the creation of a standing legal body to adjudicate cases. Overall,
PTAs exhibit major differences in design and contents.
The data presented above also suggest that agreements differ in their “depth,”
defined as “the extent to which (an agreement) requires states to depart from what
they would have done in its absence” (Downs et al. 1996, 383). We use two different
measures to operationalize depth.5 On the one hand, we produce an additive index
4In the Online Appendix, which is available at this journal’s website, we also show variation by level of
development (Fig. A-1) and across time (Fig. A-2) in the inclusion in PTAs of a few key provisions.
5A third measure, relying on tetrachoric factor analysis and the Thurstone method to calculate factor
scores, is highly and positively correlated with the two measures we present here and produces the same
substantive findings (r=0.64 and r=0.63, respectively).
358
The design of international trade agreements: Introducing a new dataset
Fig. 4 Variation in agreement design. The figure shows the percentage of agreements that contain the
various provisions. To the right, we show the absolute frequencies. Note that the denominator for calculating
the percentages slightly varies as a result of missing values for trade remedies and dispute settlement
variables
that combines seven key provisions that can be included in PTAs (see Table 1). The
first provision captures whether the agreement foresees that all tariffs (with limited
exceptions) should be reduced to zero (that is, whether the aim is to create a full
free trade area). The other six provisions capture cooperation that goes beyond tariff
reductions, in areas such as services trade, investments, standards, public procurement,
competition and intellectual property rights. For each of these areas, we code
whether the agreement contains any substantive provisions. A substantive provision,
for example, is a national treatment clause in the services chapter. A statement that
the contracting parties desire to open their services markets, by contrast, does not
count as a substantive provision.
Table 1 Operationalization of depth (additive index)
Variable Value
More than a partial scope agreement? 0/1
Substantive provision on services? 0/1
Substantive provision on investments? 0/1
Substantive provision on standards? 0/1
Substantive provision on public procurement? 0/1
Substantive provision on competition? 0/1
Substantive provision on intellectual property rights? 0/1
Total range 0/7
359
A. D¨ur et al.
0123456
1945−1949
1950−1954
1955−1959
1960−1964
1965−1969
1970−1974
1975−1979
1980−1984
1985−1989
1990−1994
1995−1999
2000−2004
2005−2009
Mean(Index)
0123
1945−1949
1950−1954
1955−1959
1960−1964
1965−1969
1970−1974
1975−1979
1980−1984
1985−1989
1990−1994
1995−1999
2000−2004
2005−2009
Mean(Latenttrait)
Fig. 5 Depth over time. The vertical bars show the standard errors of the means
On the other hand, we rely on latent trait analysis on a total of 48 variables
that theoretically are related to the depth of an agreement (these variables pertain
to such aspects as services liberalization, trade-related investment measures, intellectual
property rights and standards) to arrive at a measure of depth.6 Latent trait
analysis is a type of factor analysis for binary data (Bartholomew et al. 2011). Concretely,
we apply the Rasch model that assumes that all items capture one underlying
latent dimension, but with different discriminatory power.7 Doing so allows us to
deal with highly correlated data and to account for the fact that not all items are of
equal importance in establishing the extent of countries’ commitments.
The two variables are highly correlated (r=0.90, t=48.80) and both show a similar
picture: the depth of PTAs remained relatively stable for half a century after the end
of World War II (with considerable variation across agreements), but has increased
significantly over the last twenty years (see Fig. 5).8 In fact, the thirteen agreements
that receive the maximum score of seven on the additive index measure of depth
all have been signed since 2000. At the same time, however, major variation across
agreements exists at any point in time. The 1985 US–Israel agreement, for example,
has a depth index score of four, whereas two agreements that Venezuela signed with
Paraguay and Uruguay as recently as 2008 both score zero. Our data thus support the
common wisdom that agreements have become deeper over the past twenty years;
but this trend only explains a part of the variation in depth across agreements.
Substantial variation also exists in the depth of agreements signed by different
countries. Figure 6 shows the average depth by country of all agreements that were
6We thus leave out all flexibility and enforcement provisions. The Online Appendix contains a list of all
the variables used in the latent trait analysis.
7The Online Appendix discusses this approach and shows the extent to which each variable is related to
the latent trait.
8For the following figures we rescaled the variable resulting from latent trait analysis to have a minimum
of 0 (rather than a mean of 0).
360
The design of international trade agreements: Introducing a new dataset
0 1 2 3 4 5 60 1 2 3 4 5 6
Fig. 6 Depth across countries. The map shows the average depth (index) by country of the agreements
that were still active as of 2009. The map looks virtually the same when using the depth measure that relies
on latent trait analysis
still active as of 2009. It shows that the United States and Japan sign the deepest
agreements, whereas African countries sign the shallowest.
In Fig. 7 we offer some more micro-evidence from our dataset. Figure 7a shows
the depth of a selected sample of plurilateral agreements, using the index scores of
depth (left) and the measure based on latent trait analysis (right). The panel shows
only moderate change over time, with most plurilateral agreements being quite shallow.
In Fig. 7b, we plot the depth of all agreements signed by Chile, a country that has
been at the forefront of the new regionalism for the last twenty years. Interestingly,
for a single country we find large variation in the depth of agreements. Whereas most
of Chile’s agreements with other Latin American countries are shallow, the transcontinental
agreements signed by Chile include significant obligations for market
opening.
4 Agreement design and the PTA–trade nexus
A large literature addresses the impact of trade agreements on bilateral trade flows.
Going back to Jacob Viner, much of the early research tried to measure the extent
to which individual PTAs create or divert trade (Viner 1950; Frankel 1997). Judith
Goldstein, Michael Tomz and Douglas Rivers find that PTAs tend to increase trade
(concretely, they lead to a 34 % increase in trade, with this effect increasing over
time), but also recognize that “A natural extension of the research would be to code
not only the existence but also the strength of PTAs” (Goldstein et al. 2007, 51).
Scott Baier and Jeffrey Bergstrand, for their part, stress that PTAs are not exogenous
to trade flows and other variables that may impact trade flows (Baier and
Bergstrand 2007). When controlling for this endogeneity, the trade effect of PTAs
turns out to be much larger than previously estimated. In fact, they conclude that
PTAs double the bilateral trade of two members over a ten-year period. Baier and
361
A. D¨ur et al.
Fig. 7 The depth of selected agreements
Bergstrand confirm this result after using matching to account for the selection effect
of dyads with large trade flows signing PTAs (Baier and Bergstrand 2009).
More recently, attempts have been made to differentiate the effects by types of
trade agreements (partial scope agreements vs. free trade agreements vs. customs
unions, see Baier et al. 2011). Relying on WTO-notified agreements, Magee (2008)
shows that the long-run cumulative effects of customs unions (up to 18 years) are
around 129 %, whereas the cumulative effects for free trade agreements are around
66 %. Partial scope agreements have no statistically significant effects on trade flows.
Similary, Roy (2010) shows that customs unions increased trade by 90 % after five
years in existence, whereas free trade agreements increased trade flows on average by
25 %. Magee (2008) further provides evidence that anticipatory effects are notable in
particular in the four years leading up to the actual entry into force of the agreement.
We build on this literature to show the importance of considering the design of
trade agreements when estimating PTAs’ trade effects. Our analysis, however, relies
362
The design of international trade agreements: Introducing a new dataset
on a substantially higher number of agreements and a more refined measurement of
the depth of agreements (beyond simple categorizations). Even within the category of
free trade agreements, there are important differences as to the provisions to allow for
new or increased market access. Following from this observation, we conjecture that
various dimensions of design might matter for the PTA–trade nexus. Most obviously,
agreements differ in the extent to which they reduce tariffs. Lower tariffs make goods
cheaper and therefore more competitive in countries that are members of the same
trade agreement. As such, the larger the tariff reduction agreed upon in a PTA, the
greater the expected increase in trade.
Other PTA provisions can have a similar effect. A rule that grants national treatment
to service providers from the other member state(s), for example, should
facilitate trade in services. As much trade in goods depends on the provision
of services, services liberalization may also have a positive effect on trade in
goods. Similarly, the liberalization of government procurement policies, by enabling
exporters to compete for public contracts, should increase trade in goods and services.
Some agreements also foresee the mutual recognition of standards or the adoption
of international standards. In the EU–South Korea trade negotiations, for example,
the EU chief negotiator persuaded his counterpart to end the practice of demanding
national and U.S. standards in the production of cars and instead to rely completely
on international standards (Elsig and Dupont 2012). By reducing the costs of trade,
such provisions should lead to an increase in trade.
Some PTAs also liberalize investment policies or offer protection to foreign direct
investments. By doing so, they may allow member countries to attract additional
foreign direct investments.9 These investments increase vertical intra-industry trade,
that is, trade in similar goods produced by the same industry, but differentiated by
the unit value of the goods. Moreover, the provisions protecting intellectual property
rights included in some PTAs may stimulate trade to the extent that they curtail the
domestic production of counterfeited goods. Finally, even competition-related provisions
can affect trade if they facilitate foreign direct investments or reduce state aid
to domestic companies. Overall, therefore, the design of an agreement should matter
for the trade effect of PTAs. Stated in the form of a hypothesis, the deeper a PTA, the
larger its positive impact on trade flows between member countries.
4.1 Econometric strategy
The gravity model is the workhorse model to estimate the effect of a series of variables
on trade flows between two countries. In its basic form, this model assumes
the amount of trade between two countries to increase along with economic size,
as measured by gross domestic product (GDP), and decrease in direct proportion to
increases in the cost of transportation between the two countries, as measured by the
geographical distance between them. Over time this basic form has been enriched by
several other variables capturing political characteristics, international institutions,
9On the effect of PTAs on investments, see, for example, B¨uthe and Milner (2008).
363
A. D¨ur et al.
and cultural factors. We follow many previous studies in relying on this model to estimate
the effect of PTA design on trade flows (Rose 2004; Carr`ere 2006; Baier and
Bergstrand 2007; Goldstein et al. 2007).
The analysis covers the design of 536 PTAs signed between 1945 and 2009 (we
lose some agreements because of missing values for covariates). Our unit of observation
consists of up to 22,690 directed dyads comprising the 179 countries for which
we were able to obtain data.10 We disaggregate the EU to the member-state level to
evaluate the effect of the EU’s PTAs on trade flows between each EU member state
and all the other countries in the dataset. Formally, we estimate the following model:
lnT radeij,t = α + β1PT Aij,t−1 + β2Depthij,t−1 + β3Xij,t−1 + γij + θt + . (1)
where lnTrade is the dependent variable, PTA is a dummy variable capturing whether
two countries form part of a PTA at t − 1, and Depth is the main independent variable.
Xij are vectors of control variables, β1, β2 and β3 are the coefficients. α is
the constant and is the error term. γ and θ are respectively directed dyad fixed
effects (which among other things control for distance and contiguity) and year fixed
effects. While in most specifications we use yearly data, we rely on five-yearly data
when including country-year fixed effects, as otherwise the number of right-hand
side variables becomes too large.
lnT rade is measured as the log of the value of exports between two countries in
the dyad. We rely on a combination of two sources for these data, namely the International
Monetary Fund’s Direction of Trade Statistics (DOTS) and the dataset put
together by Kristian Gleditsch, to minimize the number of missing values (Direction
of Trade Statistics 2010; Gleditsch 2002). For Depth, we mainly rely on the operationalization
via the index described above, but also cross-check the results using the
latent trait measure.11 Some dyads form more than one PTA; in these cases we also
analyze the impact on trade flows of the second and any subsequent PTA. Depth
thus is time varying for some dyads.
Our design variable enters the model at the date of signature of an agreement. Opting
for the date of entry into force instead does not affect our results, because the large
majority of agreements enter into force after a relatively short period where states
seek domestic ratification12 As control variables we use those commonly included in
the gravity model. Importantly, we always include the variable Depth together with
the dummy variable PT A, which captures the average effect of PTAs on trade. Since
our unit of analysis is the directed-dyad-year, we include monadic variables for each
country in the dyad. Table A-1 in the Online Appendix summarizes the descriptive
statistics and sources for these variables.
10Also Baier and Bergstrand (2007) and Goldstein et al. (2007) use directed dyads. The results do not
change when using undirected dyads.
11We rescaled the latent trait measure to have positive values for all dyads that signed a PTA.
12Of 617 agreements for which we have this information, 459 entered into force within a year of signature
and another 81 within two years of signature. Moreover, some agreements are provisionally applied
immediately after signature.
364
The design of international trade agreements: Introducing a new dataset
5 Baseline analysis
Table 2 shows six models. The first model only includes the dummy variable PT A,
which does not distinguish among different designs, and a few covariates. The sign
for PT A is positive and statistically significant, confirming that on average PTAs
increase trade between members.13 The second and third models include Depth, first
measured via the index and then via latent trait analysis. The signs of the coefficients
are positive and statistically significant at the 99 percent level, indicating that, as
expected, the design of agreements matters. The deeper a PTA, the larger its effect
on trade flows between member countries. While in the following we only show
results with Depth (index), all substantive findings remain the same when using
Depth (latent) instead. In the Online Appendix (see Table A-3, Models A6 and
A7), we also report models with five-year and ten-year lags of the treatment PT A.
While the total effect remains positive, the coefficient of PT A turns negative and
statistically significant. This happens because PT A and its lags are highly collinear
with Depth.
In Model 4 we replace Depth with eight dummy variables. For all dyads
with a value of Depth (index) that is higher than the median across all dyads,
these dummies are coded one in the five years prior to the signature of the PTA
(Deep anticipatory); in the five years after the signature of a PTA (Deep short −
term); between five and 15 years after the signature of a PTA (Deep middle −
term); and 15 and more years after the signature of a PTA (Deep long − term).
In doing so, we are able to estimate the effect of Depth at different points in time.
Moreover, we include Deep anticipatory as a test to see if unobserved trends drive
our results.14 Similarly, we replaced Depth (index) with four dummy variables that
are operationalized as discussed, with the exception that they are coded one for dyads
that have a value of Depth (index) that is lower than or equal to the median across
all dyads (Shallow ...).
Seven of the eight coefficients for these dummies are positive and statistically significant
(at the 99 percent level). The negative and statistically significant coefficient
for Shallow anticipatory suggests that shallow agreements are signed by countries
that trade with each other less than expected by the gravity model (indicating that
non-economic reasons may play a role in countries’ decisions to sign shallow agreements)
and that exporters find it difficult to anticipate these PTAs. While the signs
and levels of significance of control variables are in line with the results of other
studies using the gravity model, the coefficient of PT A is negative, though it is not
statistically significant. Moreover, as for Models 1, 2, and 3, we report a model with
13If we add five-year and ten-year lags, the coefficients for PT A and the two lags are positive and statistically
significant. The total effect, i.e., the sum of the coefficients of PT A and its lags, is 0.34. Results are
reported in Table A-3 (Model A5) in the Online Appendix. If we add five-year and ten-year lags without
including Depth, the total effect of PT A and its lags is 0.50.
14For a similar approach, see Kuziemko and Werker (2006). We also include other leads to capture an
anticipatory effect between six and ten year before the signature of a PTA as well as between 11 and 15
years before the signature of a PTA. Whereas our main results do not change, 15-year leads are statistically
significant. Results are shown in the Online Appendix (Model A1 in Table A-2).
365
A. D¨ur et al.
Table 2 Baseline models
VARIABLES (1) (2) (3) (4) (5) (6)
ln(Trade) ln(Trade) ln(Trade) ln(Trade) ln(Trade) ln(Trade)
PT A 0.265*** 0.053*** 0.100*** −0.007 0.209*** 0.205***
(0.015) (0.017) (0.016) (0.022) (0.015) (0.021)
Depth(index) 0.155*** 0.125***
(0.006) (0.006)
Depth (latent) 0.260***
(0.013)
Shallow anticipatory −0.100*** 0.040***
(0.012) (0.014)
Shallow short − term 0.087*** −0.014
(0.021) (0.022)
Shallow medium − term 0.175 0.107***
(0.022) (0.022)
Shallow long − term 0.286*** 0.122***
(0.031) (0.030)
Deep anticipatory 0.269*** 0.212***
(0.015) (0.012)
Deep short − term 0.452*** 0.236***
(0.020) (0.019)
Deep medium − term 0.363*** 0.384***
(0.019) (0.020)
Deep long − term 0.207*** 0.372***
(0.034) (0.031)
Regimei 0.000 −0.000 0.000 0.000
(0.001) (0.001) (0.001) (0.001)
Regimej 0.001 0.001 0.002* 0.001
(0.001) (0.001) (0.001) (0.001)
Conf lictij −0.594*** −0.575*** −0.579*** −0.593***
(0.128) (0.128) (0.128) (0.128)
Conf licti −0.124*** −0.133*** −0.133*** −0.125***
(0.010) (0.010) (0.010) (0.010)
Conf lictj −0.073*** −0.084*** −0.083*** −0.076***
(0.010) (0.010) (0.010) (0.010)
lnGDPi 0.551*** 0.545*** 0.541*** 0.547***
(0.012) (0.012) (0.012) (0.012)
lnGDPj 0.424*** 0.420*** 0.418*** 0.421***
(0.011) (0.011) (0.011) (0.011)
GAT T/WT O 0.042*** 0.041*** 0.046*** 0.040***
(0.011) (0.011) (0.011) (0.011)
366
The design of international trade agreements: Introducing a new dataset
Table 2 (continued)
VARIABLES (1) (2) (3) (4) (5) (6)
ln(Trade) ln(Trade) ln(Trade) ln(Trade) ln(Trade) ln(Trade)
Constant −20.642*** −20.423*** −20.287*** −20.502*** 0.128 0.052
(0.361) (0.360) (0.361) (0.358)
Dyad FE yes yes yes yes yes yes
Exporter-year FE no no no no yes yes
Importer-year FE no no no no yes yes
Year FE yes yes yes yes yes yes
Observations 821,676 821,676 821,676 821,676 230,571 230,571
R-squared 0.419 0.424 0.423 0.426 0.590 0.591
Number of id 22,690 22,690 22,690 22,690 28,859 28,859
Robust standard errors clustered by dyad in parentheses
*** p<0.01, ** p<0.05, *p<0.1
Models (1) to (4) are calculated using annual data; Models (5) and (6) rely on 5-yearly data
five-year and ten-year lags of the treatment PT A in the Online Appendix (Model
A8 in Table A-3). Including these lags does not affect our finding with respect to the
depth of agreements. The coefficient for PT A, however, turns out to be negative and
statistically significant, even if the combined effect of PT A and the two lags is positive.
This negative sign for PT A again is best explained as a result of the collinearity
between PT A and Depth.
Table 3, which shows the effects of the main variables in Models 1 and 4 on trade,
presents several interesting results.15 First, the Deep ... dummies clearly outperform
the Shallow ... dummies with the exception of the long-term effects, which are not
statistically distinct one from another. We likely underestimate the long-term effect of
deep PTAs because many of the deepest agreements have only been signed recently,
making it impossible to estimate their long-term effects. We therefore re-ran Model
4 only considering agreements that are at least 15 years old (see Model A2 in Table
A-2 in the Online Appendix). In this model specification, as expected Deep long −
term by far outperforms Shallow long − term. Specifically, in the long term trade
increases by 106 % for dyads with deep PTAs. Deep agreements thus increase trade
more than shallow ones.
15Following Goldstein et al. (2007), we use arc elasticity, which is the appropriate way to calculate the
effect of dummies on the response variable. The arc elasticity is defined as the elasticity of one variable
with respect to another between two given points. As the two points get closer together, arc elasticity
approaches point elasticity. More formally, the arc elasticity is defined as T rade/T rade
P T A/P T A (Goldstein et al.
2007, 47).
367
A. D¨ur et al.
Table 3 The effect of PTAs on trade
Main variables Increase in trade ( %) C.I.
Model 1 (PTA & GATT/WTO) & Model 4 (design variables)
Deep anticipatory 31 [27, 35]
Deep short − term 57 [51, 63]
Deep middle − term 44 [39, 49]
Deep long − term 23 [15, 32]
Shallow anticipatory −10 [−12, −7]
Shallow short − term 9 [5, 14]
Shallow middle − term 19 [14, 24]
Shallow long − term 33 [25, 41]
PT A 31 [27, 34]
GAT T/WT O 4 [2, 7]
Model 6
Deep anticipatory 24 [21, 27]
Deep short − term 27 [22, 31]
Deep middle − term 47 [41, 53]
Deep long − term 45 [36, 54]
Shallow anticipatory 4 [1, 7]
Shallow short − term −1 [−5, 3]
Shallow middle − term 11 [7, 16]
Shallow long − term 13 [7, 20]
PT A 23 [18, 28]
Note: the table reports the estimated percentage increase in trade for a dyad, relative to when neither
country participates in a PTA or in the GATT/WTO. Each effect is calculated as an arc elasticity, e
ˆβ − 1,
where ˆβ is the appropriate parameter estimate from Model 1 (for PT A and GAT T/WT O), Model 4, and
Model 6 (for Deep ... and Shallow ...). 95 % confidence interval in parentheses
Second, Model 4 also offers evidence of anticipatory effects for deep agreements.
Third, the estimated effect of the GAT T/WT O is small when compared to the effect
of PTAs. We may, however, underestimate the actual effect of the GATT/WTO as
our analysis does not take into account the existence of “nonmember participants”
and formal members that opted out from obligations (Goldstein et al. 2007). Most
importantly, however, the findings show that it is not possible to correctly identify
and estimate the effect of PTAs on trade flows if deep and shallow PTAs are lumped
together.
In Models 5 and 6 in Table 2, we follow Baier and Bergstrand (2007) to better
account for the endogeneity coming from omitted variables. In our case, this is clearly
an issue since one of the leads was positive and statistically significant in Model 4.
First, in Model 5 we estimate a model including Depth, PT A, and a battery
of dummies on the right hand-side of the equation. Specifically, we include timevarying
exporter fixed effects (i.e., it), time-varying importer fixed effects (jt), as
368
The design of international trade agreements: Introducing a new dataset
well as dyads fixed effects (i.e., ij) as in the previous models.16 While the coefficient
for PT A remains positive and statistically significant (which confirms Baier
and Bergstrand’s (2007) results), the coefficient for Depth is also positive and statistically
significant, highlighting the importance of agreement design for trade flow
effects.17
Second, in Model 6 we estimate a model including the eight dummies of Model
4 as well as it, jt, and ij dummies. We report the arc elasticities for this model
in Table 3. In line with Model 4, the Deep ... dummies clearly outperform the
Shallow ... dummies. This is true also for the long-term effect. Moreover, differently
from Model 4, the middle- and long-term effects are larger than the short-term effect.
All in all, there is robust and consistent evidence that deep agreements increase trade
substantially more than shallow agreements.
6 Additional evidence
6.1 Other model specifications
As the gravity model suffers from several shortcomings, we implement some checks
to make sure that our results also hold under different specifications. Results are
shown in Table 4. First, we follow Santos Silva and Tenreyro (2006) and estimate
a pseudo-Poisson model to control for the large number of zeros in the outcome
variable.18 In this case we do not use the logarithmic transformation of T rade, but its
raw value. We include all the standard covariates of the gravity model, e.g., distance
and contiguity, as well as exporter and importer fixed effects. Our main results are
confirmed, showing that the selection bias coming from a large number of zeros does
not affect our findings (see Model 7 in Table 4).
Second, Helpman et al. (2008) addressed the selection bias issue relying on a
natural logarithm transformation and using several model specifications. Unfortunately,
we are unable to estimate the non-linear estimator developed by Helpman
et al. (2008) because we have too many fixed effects (many more than Helpman
et al. 2008). In particular, the estimation of the second stage regression produces
problematic results in a large sample because of the large number of exporter and
16We also implement first-differenced panel gravity equation estimates as suggested by Baier and
Bergstrand (2007). Specifically, we take the first differences of T rade, PT A, and Depth and regress
them on it, jt, and ij dummies. Then, we get the residuals for these estimations. Finally, we run a simple
ordinary least squares regression with robust standard errors in which T rade residuals are the dependent
variable and PT A and Depth residuals are the independent variables. The coefficients of both PT A and
Depth remain positive and statistically significant. Results are reported in Model A3 in Table A-2 in the
Online Appendix.
17The results (available upon request) do not change if we include one-year and two-year lags of PT A
and Depth in the five-year dataset. Specifically, the coefficient of these lags are positive and statistically
significant.
18We use the command PPML, written by Santos Silva and Tenreyro (2006) in STATA 12.
369
A. D¨ur et al.
Table 4 Baseline models
Variables (7) (8) (9) (10)
ln(Trade) ln(Trade) ln(Trade) ln(Trade)
PT A 0.119*** 0.327*** 0.231***
(0.019) (0.024) (0.017)
Depth (index) 0.115*** 0.077*** 0.124***
(0.0075) (0.009) (0.006)
PT A − WT O only 0.236***
(0.018)
Depth (index) − −WT O only 0.121***
(0.006)
Distance −0.614*** −0.704***
(0.009) (0.025)
Contiguity 0.404*** 0.439***
(0.020) (0.057)
CommonLanguage 0.000 0.227***
(0.018) (0.024)
CommonColony 0.213*** 0.320***
(0.047) (0.031)
CommonLegalSystem 0.288*** 0.147***
(0.012) (0.015)
CommonCurrency 0.256*** 0.141***
(0.024) (0.041)
Regimei 0.026*** 0.023***
(0.002) (0.002)
Regimej 0.024*** 0.002
(0.002) (0.002)
Conf lictij −1.764*** −0.841***
(0.147) (0.219)
Conf licti 0.010 −0.083***
(0.019) (0.017)
Conf lictj 0.005 −0.359***
(0.020) (0.045)
lnGDPi 1.010*** 0.526***
(0.018) (0.016)
lnGDPj 0.929*** 0.460***
(0.016) (0.017)
GAT T/WT O 0.313*** 0.218***
(0.022) (0.016)
T ransition −0.001
(0.001)
Constant −36.486***
(0.360)
370
The design of international trade agreements: Introducing a new dataset
Table 4 (continued)
Variables (7) (8) (9) (10)
ln(Trade) ln(Trade) ln(Trade) ln(Trade)
Exporter FE yes yes no no
Importer FE yes yes no no
Exporter-year FE no no yes yes
Importer-year FE no no yes yes
Year FE yes yes yes yes
Observations 749,763 615,152 230,571 227,282
R-squared 0.878 0.669 0.590 0.590
Number of id 21,295 21,295 28,859 28,755
Robust standard errors clustered by dyad in parentheses.
*** p<0.01, ** p<0.05, * p<0.1
Note: Model (7) is a pseudo-Poisson model. Models (9) and (10) rely on 5-yearly data.
importer dummies required. Nevertheless, we estimate another model specification
implemented by Helpman et al. (2008), known as the bins approach.19
Specifically, in the selection equation we predict whether a dyad has no trade
using a probit model whose dependent variable scores one if T rade equals zero.20 In
line with Helpman et al. (2008), we include all the covariates in Model 7 in addition
to a dummy for common religion (across dyads) as the excluded variable.21 Next,
we partitioned the obtained predicted probability ( ˆρij , using Helpman et al. (2008)
notation) into a number of bins with equal numbers of observations and assign an
indicator variable to each bin. Finally, we include this set of indicator variables in
the second stage and we report results with 50 bins.22 In this case, too, our main
results hold (see Model 8 in Table 4).23 In models 7 and 8, the coefficients of Depth
are smaller than the ones in models 2 and 5, but the coefficients are not statistically
significant different one from another.
Third, Model 9 reports the results for the subsample of PTAs that have been notified
to the WTO. This reduces the number of PTAs included in our analysis from 536
19This specification does not allow to separate the selection effect from the firm heterogeneity effect.
However, we can “obtain our key results for the intensive-margin contribution of the various trade barriers”
(Helpman et al. 2008, 465).
20Any cumulative distribution function instead of the normal distribution would work here. We estimate a
probit model following Helpman et al. (2008).
21Data on the variable religion come from the CIA World Factbook. This variable is positive and
statistically significant in the first stage.
22If we use 100 bins, we obtain the same results, which are available upon request.
23Differently from Helpman et al. (2008), the coefficient of PT A is positive and statistically significant.
That might be explained by the fact that our sample is much larger than the one of Helpman et al. (2008).
371
A. D¨ur et al.
to 348. The purpose of doing so is to make sure that our results are not driven by the
inclusion of a large number of shallow PTAs that do not feature in other studies. The
reduction in the number of PTAs included, however, does not affect the main results.
Finally, our results are similar if we drop dyads from the analysis that signed one of
the 145 PTAs that we identified but could not code because we were unable to locate
their full texts (or if we control for these dyads by adding a dummy).24
6.2 Trade-related sectors vs. tariff reduction
Our general argument is that design matters for a PTA’s effect on trade flows between
member countries. Specifically, we argue that not only tariff cuts, but also traderelated
provisions are crucial instruments for boosting trade. Previous results support
this hypothesis. A possible objection to our analysis so far, however, is that the depth
of a PTA captured by looking at trade-related provisions is correlated with the magnitude
of tariff reductions. Deep PTAs that include provisions liberalizing services,
and protecting investments and intellectual property rights are likely to be the ones
that also implement the largest tariff cuts between member countries. If that is the
case, the underlying factor leading to an increase in trade might not be trade-related
provisions, but rather the tariff reductions.
To address this concern, in the absence of a more direct measure of the extent of
tariff cuts, we include a variable capturing the length of tariff transition, that is, how
many years it takes for the tariff cuts to be fully implemented. We label this variable
T ransition.25 Tariff transition is correlated with the magnitude of tariff reduction,
i.e., the larger the tariff reduction, the longer the tariff transition.26 Indeed, if tariff
reductions are large, so are adjustment costs, which are spread out over several
years to make cooperation possible. Illustratively, the North American Free Trade
Agreement, which cut tariffs to zero on basically all products, has one of the longest
transition periods in our database, namely 15 years. This variable should thus effectively
control for the magnitude of tariff cuts envisaged by a PTA. Model 10 in
Table 4 shows that Depth remains positive and statistically significant after including
T ransition. The coefficient for T ransition is negative, though it is not statistically
significant.
7 Conclusion
We have presented a new dataset on the design and contents of PTAs. DESTA contains
information on a total of 733 PTAs and a detailed analysis of the design of 587
24Results are reported in Model A4 (Table A-2) in the Online Appendix.
25Tariff transition data might be different for country i and country j. However, since the correlation
between T ransitioni and T ransitionj is 0.98, we take the minimum of these two values to avoid
multicollinearity problems.
26In fact, in our dataset full free trade agreements have an average transition period of 5.7 years as compared
to 1.7 years for partial trade agreements. Customs unions also have a relatively long transition period
of 4.5 years.
372
The design of international trade agreements: Introducing a new dataset
PTAs. To our knowledge, it is the most comprehensive dataset on the design of trade
agreements currently available. We thus expect that DESTA will be a major asset
in addressing several long-standing questions relating to the new regionalism and
international institutions more broadly.
The novel data has allowed us to revisit the literature on the PTA–trade nexus,
with the results corroborating earlier studies that showed that PTAs increase trade.
Our additional insight is that this effect is driven by deep agreements, whereas shallow
agreements have a substantially smaller impact on trade flows. Of particular
interest is the result that provisions included in PTAs that do not directly concern
tariffs – such as those liberalizing services trade or protecting investments and intellectual
property rights – have a significant impact on trade. We do not claim that
this analysis resolves the question of the PTA–trade nexus once and for all. Future
studies should look more precisely at questions such as which provisions increase
trade by how much; what type of trade – inter-industry or intra-industry – is affected
most; and which sectors see the largest increase in trade. The central finding that
design matters, however, is very robust to changes in operationalization and model
choice.
Future research on PTAs thus should further open the black box of trade agreements
and concentrate on variation across PTAs in design and content rather than
treat all PTAs as if they were the same. Studies taking PTA design seriously may
address new questions such as, which design elements bring about significant distributional
consequences in participating member states? What type of obligations
coupled with what type of enforcement mechanisms are likely to induce domestic
policy change? How do certain design features in isolation or in conjunction
with other variables affect implementation? And do design features travel from one
agreement to another, that is, is there a diffusion of institutional design?
Of course, PTAs are only one among many types of international institutions,
albeit expanding in number and depth over time. Studying the design of PTAs will
allow the PTA literature to better engage with the broader literature on international
cooperation and international organizations, creating possibilities to contribute
actively to ongoing debates and advancements in research programmes as diverse as
legalization (Abbott et al. 2001), rational design (Koremenos et al. 2001; Rosendorff
and Milner 2001), diffusion (Simmons and Elkins 2004; Braun and Gilardi 2006), or
overlapping regimes (Drezner 2006; Busch 2007). Many international organizations
have been extensively studied in terms of design. Connecting PTAs more tightly to
non-trade institutions would be beneficial for situating the role of trade institutions
more broadly. The DESTA dataset will be of major use in developing this research
agenda.
Acknowledgments All authors have contributed equally to the paper. We are grateful to the NCCR
Trade Regulation (www.nccr-trade.org) for financial support and our research assistants for help in collecting
the data for this paper. Special thanks go to Karolina Milewicz for her support in the larger project.
Todd Allee, Richard Baldwin, Tim B¨uthe, Jappe Eckhardt, Jeff Kucik, Marcelo Olarreaga, Ron Rogowski,
Michael Z¨urn and participants of the ECPR Joint Sessions of Workshops at the University of Antwerp
provided helpful comments on previous versions of this paper. We also wish to thank Raymond Hicks
for kindly assembling the gravity model data. For more information on the DESTA dataset, see www.
designoftradeagreements.org.
373
A. D¨ur et al.
References
Abbott, K., Keohane, R., Moravcsik, A., Slaughter, A.-M., Snidal, D. (2001). The concept of legalization.
International Organization, 54(3), 401–19.
Baccini, L., & D¨ur, A. (2012). The new regionalism and policy interdependence. British Journal of
Political Science, 42(1), 57–79.
Baier, S., & Bergstrand, J. (2007). Do Free trade agreements actually increase members’ international
trade? Journal of International Economics, 71(1), 72–95.
Baier, S., & Bergstrand, J. (2009). Bonus Vetus OLS: a simple method for approximating international
trade-cost effects using the gravity equation. Journal of International Economics, 77(1), 77–85.
Baier, S., Bergstrand, J., Feng, M. (2011). Economic integration agreements and the margins of
international trade. Working Paper.
Baldwin, R. (1993). A domino theory of regionalism. NBER working paper no. 4465.
Baldwin, R., & Jaimovich, D. (2012). Are free trade agreements contagious? Journal of International
Economics, 88(1), 1–16.
Bartholomew, D.J., Knott, M., Moustaki, I. (2011). Latent variable models and factor analysis: a unified
approach. Chichester: Wiley.
Braun, D., & Gilardi, F. (2006). Taking Galton’s problem seriously. Towards a theory of policy diffusion.
Journal of Theoretical Politics, 18(3), 298–322.
Busch, M. (2007). Overlapping institutions, forum shopping, and dispute settlement in international trade.
International Organization, 61(4), 735–61.
B¨uthe, T., & Milner, H. (2008). The politics of foreign direct investment into developing countries: increasing
FDI through international trade agreements? American Journal of Political Science, 52(4), 741–
62.
Carr`ere, C. (2006). Revisiting the effects of regional trade agreements on trade flows with proper
specification of the gravity model. European Economic Review, 50(2), 223–247.
Chase, K. (2003). Economic interests and regional trading arrangements: the case of NAFTA. International
Organization, 57(1), 137–74.
Direction of Trade Statistics (2010). Global insight direction of trade. In Direction of trade. Available:
DRI WEFA’s global insight.
Downs, G.W., Rocke, D.M., Barsoom, P.N. (1996). Is the good news about compliance good news about
cooperation? International Organization, 50(3), 379–406.
Drezner, D. (2006). The viscosity of global governance: when is forum-shopping expensive? Paper
presented at the annual meeting of the american poltitical science association. Philadelphia.
D¨ur, A. (2007). EU trade policy as protection for exporters: the agreements with Mexico and Chile. Journal
of Common Market Studies, 45(4), 833–55.
Elsig, M., & Dupont, C. (2012). European Union Meets South Korea: bureaucratic interests, exporter
discrimination and the negotiations of new trade agreements. Journal of Common Market Studies,
50(3), 492–507.
Estevadeordal, A., Suominen, K., Teh, R. (Eds.) (2009). Regional rules in the global trading system.
Cambridge: Cambridge University Press.
Frankel, J. (1997). Regional trading blocs in the world economic system. Washington, D.C.: Institute for
International Economics.
Gleditsch, K. (2002). Expanded trade and GDP data. Journal of Conflict Resolution, 46(5), 712–24.
Goldstein, J., Tomz, M., Rivers, D. (2007). Institutions in international relations: understanding the effects
of the GATT and the WTO on world trade. International Organization, 61(1), 37–67.
Gowa, J., & Mansfield, E. (1993). Power politics and international trade. American Political Science
Review, 87(2), 408–20.
Grossman, G., & Helpman, E. (1995). The politics of free-trade agreements. American Economic Review,
85(4), 667–90.
Hafner-Burton, E. (2005). Trading human rights: how preferential trade agreements influence government
repression. International Organization, 59(3), 593–629.
Haftel, Y. (2010). Ratification counts: US investment treaties and FDI flows into developing countries.
Review of International Political Economy, 17(2), 348–377.
Helpman, E., Melitz, M., Rubenstein, Y. (2008). Estimating trade flows: trading partners and trading
volumes. The Quarterly Journal of Economics, 123(2), 441–487.
374
The design of international trade agreements: Introducing a new dataset
Hicks, R., & Kim, S.Y. (2012). Reciprocal trade agreements in Asia: credible commitment to trade
liberalization or paper tigers? Journal of East Asian Studies, 12(1), 1–29.
Hollyer, J., & Rosendorff, P. (2012). Leadership survival, regime type, policy uncertainty and PTA
accession. International Studies Quarterly, 56(4), 748–764.
Koremenos, B., Lipson, C., Snidal, D. (2001). The rational design of international institutions. International
Organization, 55(4), 761–99.
Kucik, J. (2012). The domestic politics of institutional design: producer preferences over trade agreement
rules. Economics and Politics, 24(2), 95–118.
Kuziemko, I., & Werker, E. (2006). How much is a seat on the security council worth? Foreign aid and
bribery at the United Nations. Journal of Political Economy, 14(5), 905–930.
Magee, C.S.P. (2008). New measures of trade creation and trade diversion. Journal of International
Economics, 75, 349–362.
Maggi, G., & Rodr´ıguez-Clare, A. (2007). A political-economy theory of trade agreements. American
Economic Review, 97(4), 1374–1406.
Manger, M. (2009). Investing in protection: the politics of preferential trade agreements between north
and south. Cambridge: Cambridge University Press.
Mansfield, E., & Milner, H. (1999). The new wave of Regionalism. International Organization, 53(3),
589–627.
Mansfield, E., & Milner, H. (2012). Votes, Vetoes, and the political economy of international trade
agreements. Princeton: Princeton University Press.
Mansfield, E., & Pevehouse, J. (2000). Trade blocs, trade flows, and international conflict. International
Organization, 54(4), 775–808.
Mansfield, E., & Reinhardt, E. (2003). Multilateraldeterminants of regionalism: the effects of GATT/WTO
on the formation of preferential trading arrangements. International Organization, 57(4), 829–62.
Mansfield, E., & Reinhardt, E. (2008). International institutions and the volatility of international trade.
International Organization, 62(4), 621–52.
Mansfield, E., Milner, H., Rosendorff, P. (2002). Why democracies cooperate more: electoral control and
international trade agreements. International Organization, 56(3), 477–514.
Mansfield, E., Milner, H., Pevehouse, J. (2007). Vetoing Co-operation: the impact of veto players on
preferential trading arrangements. British Journal of Political Science, 37(3), 403–32.
Milner, H. (1997). Industries, governments, and the creation of regional trade blocs. In E. Mansfield, &
H. Milner (Eds.) The political economy of regionalism (pp. 77–106). New York: Columbia University
Press.
Rickard, S., & Kono, D. (2013). Think globally, buy locally: International agreements and government
procurement. Review of International Organizations. doi:10.1007/s11558-013-9177-x.
Rose, A. (2004). Do we really know that the WTO increases trade? American Economic Review, 94(1),
98–114.
Rosendorff, P., & Milner, H. (2001). The optimal design of international trade institutions: uncertainty and
escape. International Organization, 55(4), 829–58.
Roy, J. (2010). Do customs union members engage in more bilateral trade than free-trade agreement
members. Review of International Economics, 18(4), 663–81.
Santos Silva, J.M.C., & Tenreyro, S. (2006). The log of gravity. Review of Economics and Statistics, 88
641–658.
Simmons, B., & Elkins, Z. (2004). The globalization of liberalization: policy diffusion in the international
political economy. American Political Science Review, 98(1), 171–189.
Smith, J.M. (2000). The politics of dispute settlement design: explaining legalism in regional trade pacts.
International Organization, 54(1), 137–80.
Viner, J. (1950). The customs union issue. New York: Carnegie Endowment for International Peace.
World Trade Organization. (2011). World trade report 2011: the WTO and preferential trade agreements:
from co-existence to coherence. Geneva: WTO.
375