Depth, Flexibility and International Cooperation: The Politics of Trade Agreement Design Leonardo Baccini, LSE Andreas D¨ ur, University of Salzburg Manfred Elsig, University of Bern

Abstract

1

States design some international institutions more flexibly than others. What explains this variation? Focusing on preferential trade agreements, we argue that different aspects of institutional design are interdependent. In particular, we posit that deep agreements create an incentive for states to add more flexibility, which can take the form of transitional flexibility or provisions that serve as safety valves in the long term. Both types of flexibility mechanisms increase with depth and are complementary. We also argue that states are concerned about the stability of an agreement and as a consequence introduce ex-ante constraints against exploiting the future application of flexibility. An original dataset on the design of 587 trade agreements signed between 1945 and 2009 allows us to test our arguments. Descriptive evidence, multivariate statistics and instrumental variable models all support the theoretical expectations. The paper contributes to the literatures on the design of international institutions and preferential trade agreements. Key Words: international institutions, preferential trade agreements, institutional design, depth, flexibility. Version: 24 July 2013

1

Special thanks to Peter Rosendorff for helpful comments on multiple drafts of this manuscript. We are also grateful to Tobias B¨ ohmelt, Yoram Haftel, Raymond Hicks, Yuch Kono, Yotam Margalit, Helen Milner, Johannes Urpelainen, the participants at the 2011 IPES conference, 2012 PEIO conference, 2012 and 2013 ECPR Joint Sessions and a seminar at Columbia University for helpful comments on earlier versions of this paper. Moreover, we acknowledge support from the Swiss National Science Foundation’s NCCR Trade Regulation and thank our research assistants for help in collecting the data for this paper.

Introduction International institutions strongly vary in their design. Some international agreements foresee deep cooperation with many commitments, whereas others are left shallow. Some feature multiple flexibility measures that allow states to temporarily withdraw concessions, whereas others are rigid and allow few opt-outs. Observing this variation across different dimensions of institutional design, we ask: are these different dimensions interdependent, that is, does depth make states design more flexible agreements? And what explains the type of flexibility instruments that states use? We focus on international economic cooperation, and in particular on bilateral and regional trade agreements, in responding to these questions. Since World War II, states have concluded several hundred of these agreements, with the number of agreements increasing particularly rapidly for the last twenty years.2 These preferential trade agreements (PTAs) are an attractive focus for our study as they are a worldwide phenomenon. In addition, we observe large variation across agreements in terms of design, and evidence is accumulating that this variation in design matters for variables as diverse as trade flows, foreign direct investments and compliance with human rights.3 In focusing on PTAs, our study nicely complements the existing literature on the design of international trade institutions that so far has mainly focused on the World Trade Organization (WTO) and its forerunner the General Agreement on Tariffs and Trade (GATT).4 Drawing on existing research, we develop an original argument that suggests that deep agreements create a demand from import-competing groups for more flexibility. States respond to this demand by including two different types of flexibility provisions in PTAs, namely transitional flexibility measures (that allow for transitional periods in respect to tariff liberalization) and escape flexibility measures (that can be invoked 2

D¨ ur et al. 2013. For the effect of trade agreement design on trade flows, see Egger and Wamser 2013; for foreign direct investments see B¨ uthe and Milner 2011; and for human rights compliance see Hafner-Burton 2009. 4 Rosendorff 2005; Rosendorff and Milner 2001; Kucik and Reinhardt 2008; Pelc 2009. 3

2

well into the future, such as safeguard measures, and anti-dumping and anti-subsidies provisions). Finally, states impose constraints on the use of these additional flexibility provisions as concerns for stability rise among exporting interests and the governments pushing for deeper agreements. We test these expectations on a new dataset on the design of 587 trade agreements signed between 1945 and 2009.5 This is the most comprehensive dataset on the design of trade agreements in terms of both agreements covered and detail of the coding. The coding of the texts of these agreements allows us to establish measures of depth and flexibility for each of them. Descriptive evidence and the results from ordered probit, tobit and instrumental variable models all offer clear-cut support for our theoretical expectations. The findings are highly robust to changes in estimation method and the operationalization of variables. We thus provide the first large-n study of the relationship between depth and flexibility in respect to trade agreements. Our study makes three major contributions to the literature on the design of international institutions.6 First, in providing evidence that deeper agreements require more flexibility, we establish that different aspects of institutional design are interdependent.7 The dataset is uniquely well suited to making this point, as it is the first to contain data on both the depth and flexibility of a large number of agreements. Treating an aspect of institutional design in isolation from other aspects seems problematic in view of this finding. Second, we not only show that states use different types of flexibility instruments, but that these different types of flexibility are complements rather than substitutes.8 Whereas much of the existing literature expects that as one flexibility instrument becomes more important, others become less prevalent, we see two types of flexibility that both go up in response to greater depth.9 Finally, while the desire for stability may indeed motivate states to add flexibility, clearly states also feel that too much un5

D¨ ur et al. 2013. Koremenos et al. 2001. 7 See also Downs et al. 1996; Johns 2012; Haftel 2013. 8 For a list of different flexibility measures, see for example Helfer 2013. 9 For the substitution expectation, see Koremenos 2005; Pelc 2010. 6

3

constrained flexibility may undermine the stability of an agreement in the long term.10 Flexibility thus is less of an easy fix for cooperation problems than often expected. The paper also speaks to the literature on PTAs by shedding some light on the origins of the wide variation in their design.11 Whereas for some time the quantitative PTA literature treated all PTAs as equal, more recently authors have started to pay attention to how diverse such agreements are in terms of design. With some agreements being as short as one or two pages and others covering more than a thousand pages, it is unlikely that the reasons for signing PTAs and the consequences of PTAs are equal across all of them. Both our theoretical argument and the empirical evidence contributes to this emerging literature. For example, our finding that depth and flexibility go hand-in-hand should make future research on the consequences of PTAs cautious about testing the effects of one of these two dimensions without also considering the other. In the following sections, we first discuss the literature on flexibility (Section 1) and then develop our argument (Section 2). Section 3 explains how we operationalize the key variables and also puts forward some descriptive evidence. We then test our arguments using ordered probit and tobit models (Section 4). Section five presents results from an instrumental variable approach and the final section concludes.

The Concept of Flexibility Flexibility provisions are generally understood as devices included in an agreement that allow states to anticipate and respond to domestic contingencies or to adjust their policies for other purposes without violating the terms of an agreement. Flexibility clauses then provide for legally accepted opt-outs without leading to a de jure breach of an agreement.

10

Rosendorff 2005 links flexibility to the stability of an agreement. For some key studies in this field, see Mansfield et al. 2002; Chase 2005. World Trade Organization 2011 offers a recent review of this literature. 11

4

A rich literature studies flexibility in international cooperation.12 The flexibility instruments discussed in this literature include membership rules and voting procedures, soft as opposed to hard law provisions, degrees of delegation, exit options, duration and re-negotiation clauses, reservations, escape clauses, and withdrawal clauses.13 Barbara Koremenos and co-authors classify these instruments into two types, namely transformative and adaptive flexibility.14 The former allows states to significantly change the terms of obligations (e.g. re-negotiations or exit), the latter makes possible an individual escape as a time-limited exception which however does not call into question the stability of an agreement. A significant proportion of the flexibility literature focuses on the WTO and the GATT. The focus has been on adaptive flexibility, as trade institutions are less likely to include transformative flexibility features such as exit options that may endanger overall stability. Krzysztof Pelc shows how GATT escape clauses have changed over time, with different flexibility measures serving different purposes.15 In another contribution, he provides evidence that the degree of flexibility (defined as the amount of overhang in tariff levels) is a function of how countries can use currency depreciation to react to crises and whether they have access to trade remedies.16 He argues that more options to react to future challenges will lead to less flexibility. This work suggests that flexibility depends on the perceived possibilities for using it in the future and that there are important interaction effects between flexibility tools. The latter is in line with the work by Koremenos, who focused on how various flexibility tools may act as substitutes.17 Little work has been done on flexibility in PTAs, with the notable exception of Jeffrey Kucik who focuses on variation in flexibility provisions in relation to trade remedies.18 He argues that the political struggle between exporters and import12

See Koremenos et al. 2001; Helfer 2013. See Goodman and Jinks 2004; Shaffer and Pollack 2010; Alvarez 2005; Helfer 2005; Koremenos 2005; Neumayer 2007; Rosendorff and Milner 2001; and Koremenos and Nau 2010. 14 Koremenos et al. 2001, 773. 15 Pelc 2009. 16 Pelc 2010. 17 Koremenos 2005; see also Helfer 2013. 18 Kucik 2012. 13

5

competing industries influences the flexibility of PTAs. While this literature addresses a large number of important questions, so far our understanding of why we see variation across agreements in terms of flexibility is limited. Some studies suggest that issue characteristics are important for the degree of flexibility granted in agreements; others focus on either member state characteristics or on societal interests. When looking at PTAs, however, we see that within the same issue area agreements signed by states with similar characteristics vary widely in their flexibility. To account for this variation, we argue that the depth of an agreement is a crucial driver in determining the flexibility of an agreement. Moreover, we highlight multiple forms of flexibility that states can use and suggest that states may view unchecked flexibility as a threat to the stability of an agreement. We develop this argument below in detail.

The Argument The depth of PTAs Existing research suggests that signing PTAs, which exchange domestic for foreign market access, can increase the odds of political decision-makers’ survival in office.19 The argument is that signing a PTA sends a credible signal to the median voter that the government is attempting to increase general welfare.20 Building on this argument, we assume that governments are favorably disposed towards signing new PTAs. Exporters that demand improved foreign market access or want the government to respond to discrimination in export markets can further amplify this government preference.21 Governments, however, do not face a dichotomous choice between signing or not signing a PTA, but a choice among many options ranging from no agreement to a very deep agreement. Depth here is understood as the extent to which PTAs liberalize 19

See Hollyer and Rosendorff 2012; Mansfield and Milner 2012. Mansfield and Milner 2012. 21 Manger 2009; D¨ ur 2010; Baccini and D¨ ur 2012. 20

6

trade.22 An important determinant of the depth of an agreement are average tariff cuts. PTAs, however, can also contribute to liberalizing trade relations between states by opening some services sectors to foreign competition or by allowing foreign companies to bid for tenders in government procurement. Similarly, the protection of foreign direct investment can substantially enhance market access for exporters, as foreign investments can be a precondition for trade. Some agreements also go beyond protecting investments against arbitrary provisions and explicitly liberalize foreign investments.23 Other behind-the-border obstacles that a PTA may remove are burdensome technical standards, sanitary or phytosanitary measures, inadequate protection of intellectual property rights, and competition rules that discriminate against foreign traders. Some PTAs, for example, foresee the mutual recognition of international product standards or the harmonization of technical regulations.24 This reduces transaction costs and increases market integration. The strengthening of intellectual property rights is important for enabling exporters to pursue a long-term market penetration strategy. Absence of patent or trademark protection directly translates into less exports and less technology transfer in the case of investments directed at foreign markets.25 Finally, cooperation on competition policy is important to address unfair competition by state enterprises and private firms. Similar to intellectual property rights it creates a regulatory environment that strengthens the rule of law. This provides important incentives for exporting firms to decide to pursue active market expansion in the PTA partner’s home economy. We argue that variation in the extent to which PTAs include these types of provisions – and thus variation in the depth of agreements – depends on how eager the government is to credibly signal a specific economic policy to the median voter and the pressure that exporters exert on governments to negotiate a deep agreement: the greater the demand for trade liberalization, the deeper the agreement. 22

This is similar to Downs et al. (1996,383) who define depth as “the extent to which [an agreement] requires states to depart from what they would have done in its absence”. 23 See also World Trade Organization 2011. 24 Piermartini and Budetta 2009. 25 Maskus and Penubarti 1995.

7

Flexibility in PTAs At the same time as governments strive to improve market access for exporters through reciprocal deals with foreign governments, we expect import competitors to ask for protection.26 We assume that once states start negotiating, import competitors find it very difficult to oppose the conclusion of such agreements, which are valued by governments as credible signals and supported by exporter interests. Instead, we expect import-competitors to demand the inclusion of flexibility provisions that soften the impact of an agreement. In particular, they ask for two types of flexibility instruments in PTAs. First, they demand provisions that can be used when states are faced with sudden changes in the economic environment, such as economic downturns. Such safety valves allow for a legal breach limited in time in case imports surge as a result of increased trade flows between PTA partners.27 These include balance of payments (BoP) exceptions or general or specific safeguard provisions. The Treaty of Rome (1957), for example, allowed member states of the European Economic Community to suspend part of their liberalization in case of BoP difficulties (Art. 109). Or, in the EU-South Korea PTA negotiations, the European car industry demanded (and received) a special safeguard related to cars. In addition, safety valves exist to address anti-competitive behavior by other states (excessive use of subsidies to boost exports) or by firms (dumping practices). These exist in the form of anti-dumping and countervailing duty provisions. We summarize this group of flexibility measures with the term “escape flexibility”. Second, import competitors can also ask for “transitional flexibility”, namely transition periods for the implementation of an agreement. Transitional flexibility gives domestic industries that are threatened by more competition through imports additional time to prepare for liberalization.28 It is a flexibility tool which is negotiated ex ante and addresses the period immediately following entry into force of an agreement. 26

Grossman and Helpman 1994. See also Kucik 2012. 28 Chase, 2003. 27

8

Most tariff elimination schemes foresee a gradual lowering of tariffs while each year the amount of flexibility decreases until tariffs are fully eliminated. In short, the longer the transition period, the greater the flexibility offered to import-competing interests. We assume that while governments are favorably disposed towards PTAs and act as catalysts for exporters’ interests, they cannot ignore import-competing industries’ demands for flexibility in negotiating PTAs. This is because governments that want to remain in office are sensitive to the preferences of politically mobilized interest groups.29 The underlying assumption is that governments depend on business support for reelection (or, in the case of autocratic governments, to stay in power) and the formulation and implementation of policies.30 Thus, rather than adhere to a “winner-takes-all” logic, governments try to satisfy all sides in a policy debate. Whereas exporters receive foreign market access, governments are likely to grant import-competitors some forms of flexibility.

The relationship between depth and flexibility So far, we have argued that the depth of an agreement is a function of how much governments need a PTA as a credible signal to the median voter, and the amount of exporter pressure for deep agreements. At the same time, governments give some flexibility to import competitors. But what explains the overall balance between depth and flexibility and the design of any flexibility measures? Our key argument is that the degree of flexibility granted increases together with the expected depth of an agreement. The reason for this is that deeper agreements have a greater positive trade flow effect than shallow agreements.31 This effect is not only driven by tariff cuts. In fact, the impact of trade-related provisions on trade flows is sometimes greater than the effect of tariff cuts, as they provide new market access (extensive margins) in addition to improved market access (intensive margins) for 29

Milner 1988; Grossman and Helpman 1994; Gilligan 1997. See, for example Wright 1996; Hall and Deardorff 2006. 31 Baier and Bergstrand 2007; D¨ ur et al. 2013. 30

9

exporters.32 Deeper agreements thus will lead to more extensive lobbying from importcompeting groups and increased demand for protection. Governments that want to remain in office, in turn, will feel compelled to make increasingly large concessions to the import-competing groups. As described above, governments can rely on multiple ways of providing flexibility that addresses the concerns of import competitors. We posit that states want to show to import-competing groups that they are addressing their legitimate concerns in both the short-term and the long-term. The expectation thus is for all types of flexibility mechanisms to be increasing in line with the potential for market integration. Therefore, these tools are complementary rather than substitutes. This postulated relationship between depth and flexibility instruments leads to the following two hypotheses: Hypothesis 1 The greater the depth of an agreement, the more escape flexibility it will include. Hypothesis 2 The greater the depth of an agreement, the more transitional flexibility it will include. An example illustrates the plausibility of these hypotheses: when eleven Caribbean countries signed up to the Caribbean Free Trade Area in 1965, they designed a rather shallow agreement that only covered trade in goods (the aim of liberalizing trade in services was mentioned in the agreement, but without asking member states to take any concrete steps). The agreement can be considered rather rigid, as it only included a limited number of escape clauses. A few years later (1973), when the Caribbean states decided to create the Caribbean Community, a deeper agreement, they also made the agreement more flexible by adding a provision explicitly allowing the imposition of antidumping duties. Moreover, whereas the 1965 agreement came with a transition period of six years, the 1973 agreement foresaw a transition period of eight years. The revised agreement from 2001, which was deeper still, includes an additional escape clause to increase the agreement’s flexibility even further. 32

D¨ ur et al. 2013.

10

While import-competing groups demand both flexibility mechanisms, escape flexibility poses risks for both governments and exporters alike. Low triggers for using these instruments and potential regulatory capture in the future may make the use of these instruments more uncertain and open up the space for potential overuse. In particular, investigating domestic authorities may use ample discretion in ways not really envisaged by the treaty negotiators. Governments thus face a classical time-inconsistency problem in designing escape flexibility. Economic actors anticipate that governments will be tempted to use such clauses in the future.33 Both governments and exporters face the challenge that unrestricted recourse to these opt-outs jeopardizes the overall benefits of the agreement (and its stability) in the long run. If foreign governments use flexibility mechanisms without restraint, the benefits of market access may be nullified. Exporters that highly value market access oppose lax rules on flexibility which would allow foreign governments or importers to demand too much protection.34 Even if exporters do not have the clout to completely avoid the inclusion of escape flexibility, they will try to restrict its use. Therefore, governments have a strong incentive ex-ante to define limits to the use of escape flexibility and to agree on procedural constraints to control the application of escape tools as a sort of tying hand approach to address the time-inconsistency problem. Various possibilities exist for making flexibility measures more rigid. In terms of safeguards and anti-dumping, states may limit the duration of an anti-dumping duty, restrict the upper level of the trade remedy imposed, or make reference to the GATT/WTO legal framework that prescribes a number of procedural and substantive obligations that shield against abuse. In the case of the South Korea-US negotiations, for example, the Korean car industry lobbied hard to get changes in US trade remedy laws that would make the imposition of anti-dumping duties more difficult. In terms of subsidies, treaties may push governments to decrease or eliminate them or demand 33 34

Kydland and Prescott 1977. Kucik 2012, 98.

11

active cooperation among authorities to address competition-related negative spillovers. For instance the trade agreement between Japan and Switzerland foresees that no export subsidies shall be introduced or maintained on agricultural products for which tariff liberalization has been agreed. We call these measures “flexibility strings”. Our prediction is that the use of escape flexibility is conditioned by strings attached to its use, which in turn decreases overall escape flexibility. Governments will seek an optimal degree of flexibility that allows for temporary breach and adaptation if necessary, but that is restricted through a set of tools. From the above discussion, we derive the following hypothesis: Hypothesis 3 As escape flexibility increases, we expect more flexibility strings. We do not test this conjecture in relation to transitional flexibility, as transition periods define clear rules for future behavior. There are no opportunities for importcompeting groups to stop the gradual elimination of tariff concessions without this leading to a clear breach of agreement. Moreover, the GATT/WTO defines an upper limit on the number of years generally accepted for implementation. Article 24 GATT states that transition periods should exceed 10 years only in exceptional cases. In the case of the EU-South Korea PTA, for example, the European car industry received a transition period of five years for tariff cuts on certain types of cars. In some agreements, however, governments have accepted a transition period of 15 to 20 years for a limited number of products. Despite some wiggle room in the use of transitional flexibility, the GATT/WTO international obligation constrains countries from stretching this flexibility tool beyond a certain limit (external string).

Descriptive Evidence For the empirical analysis, we rely on an original dataset on the design of 587 PTAs signed between 1945 and 2009.35 The list of agreements contains 358 bilateral and 229 35

This is the subset of more than 700 PTAs that were identified for which the full text of the agreement could be identified. We expect most of the agreements for which no full text could be found

12

plurilateral agreements, and substantially goes beyond the list of agreements registered with the WTO (only 60 per cent of them also feature in the WTO list). It is the result of the integration of several existing lists (including those maintained by the Organization of American State’s Foreign Trade Information System and the World Bank) and the systematic search of government web pages. The agreements in the dataset are relatively young (no fewer than 73 per cent of them have been signed since 1990); were signed by developing countries (67 per cent are of a South-South type); and are full free trade agreements (60 per cent). The agreements were manually coded for a total of eight market-access related sectors of cooperation that may be included in PTAs, encompassing goods, services, investments, intellectual property rights, competition, public procurement, standards, and trade remedies. For each of these sectors, a significant number of items were coded, meaning that we have about 100 data points for each agreement. To ensure the reliability of the data, all agreements were double-coded, and any differences between the two sets of data were then resolved.36 For nearly all variables, inter-rater agreement as measured by Cohen’s kappa is higher than 0.75. Moreover, cross-checks of the dataset against existing ones that partially overlap with it confirm the reliability of the data.37 The large number of agreements coded and the level of detail included in the coding mean that this dataset offers a unique opportunity to test our argument. In the following, we discuss how we use this dataset to measure depth, escape flexibility, transitional flexibility and flexibility strings. We use two different measures of depth below. On the one hand, we created an additive index of 48 items in our dataset (Depth (index)). We only included items in the index that theoretically are related to depth and weighted each item equally to be shallow and rigid, which means that losing them should bias results against our argument. 36 For more details, see D¨ ur et al. 2013. 37 For these other datasets, see the contributions in Estevadeordal et al. 2009 and World Trade Organization 2011. Estevadeordal et al. 2009 includes up to 70 agreements that also feature in the present dataset (varying depending on contribution). World Trade Organization 2011 analyzes the contents of 96 agreements.

13

(the appendix contains a list of all of these items). While theoretically this measure can range from 0 to 48, in practice we observe values in the range from 0 to 40. On the other hand, we relied on item response analysis on the same set of variables to arrive at a measure of depth (Depth (IR)). Latent trait analysis is a technique that is similar to factor analysis, with the advantage that it is applicable to binary data.38 The specific model that we apply, which is known as the Rasch model, assumes that all items capture one underlying latent dimension. The items, however, contribute more or less to this latent dimension (that is, they have more or less discriminatory power). Using this operationalization, provisions that are relatively rare contribute more to depth than provisions that are ubiquitous. After rescaling to remove negative values, this variable ranges from 0 to 3.3. The two measures are highly correlated (r=0.94). For both depth measures, we find that over time agreements have become deeper, with the trend towards deep agreements starting in the second half of the 1980s. We also observe that North-South agreements are substantially deeper than both North-North and South-South agreements in our dataset. North-North agreements are only slightly deeper than South-South agreements. The deepest agreements in the dataset are signed by the United States (for example, Australia-US and Colombia-US), Japan (for example, Japan-Switzerland) and the European Free Trade Association (for example, European Free Trade AssociationColombia). On both measures of depth, 147 agreements score 0, including the African Common Market and a large number of Arab agreements. Hypothesis 1 refers to a variable called Escape flexibility. We measure this variable using a simple additive index of the presence or absence of four provisions in PTAs: a provision allowing for the suspension of tariff cuts in case of balance of payments problems, a general safeguard provision, a provision allowing for the imposition of countervailing duties, and a provision allowing for the imposition of anti-dumping duties.39 In the absence of these provisions, a country that suspends its tariff cuts 38

Bartholomew et al. 2011. As a robustness check, we also used item response analysis equal to the approach described for depth above to combine these four items into a weighted index. The resulting variable is highly 39

14

or imposes antidumping and countervailing duties for goods covered by the agreement is in breach of the agreement. These four provisions thus serve as escape clauses. The index can range from 0 to 4. In our dataset, 73 agreements score 0 and 260 agreements score 4 on this variable. Over time, we see an increasing number of agreements including several of these escape provisions. Following Hypothesis 1, this variable should be positively related to depth. To examine this proposition, we show boxplots based on a recoded depth variable that ranges from 0 (very shallow) to 4 (very deep). The resulting graph (Figure 1) shows a strong positive relationship between the two variables. Figure 1 About Here Hypothesis 2 stipulates a relationship between depth and Transitional flexibility. We operationalize transitional flexibility relying on the maximum number of years that countries are allowed in which to achieve the liberalization of tariffs envisaged in the agreement.40 The longer the transition period, the more flexibility exists for import-competing groups to adjust to increased competition. Phase-out periods for tariff liberalization range between 0 years (all tariffs are liberalized at the date of entry into force of an agreement) and 25 years (usually for a selected number of sensitive products). We took the median for the member countries of a PTA if the transition periods vary across countries.41 In Figure 2 we show the relationship between depth and transitional flexibility. As expected in Hypothesis 2, we see a positive relationship between the two variables. As depth increases, also transitional flexibility increases. The figure shows that US agreements are not only unusually deep, but several of them also feature very long transition periods. By contrast, the African Economic Community (1991) is a stark outlier in the sense that it is very shallow and has one of the longest transition periods in the dataset. positively correlated with the additive index that we use (rs = 0.90). 40 Agreements sometimes also include transition periods for the “trade and” provisions. As such transition periods can only be found in deep agreements (because shallow ones tend to focus on tariffs, and thus do not include “trade and” provisions), relying on these other transition periods would make our test tautological. 41 If we take the mean or the minimum value, we obtain similar results (available upon request).

15

Figure 2 About Here Finally, hypothesis 3 refers to the constraints on escape flexibility. We operationalize this variable (flexibility strings) using six items that capture limits in respect to the imposition of antidumping duties, the provision of subsidies, and the use of the safeguard provisions.42 These capture provisions that impose WTO rules on the use of the antidumping instrument, the safeguard provision and the provision of subsidies, stipulate that the safeguard provision is only valid during the transition period, create a common policy on subsidies, and stipulate a minimum dumping margin (the appendix contains more information on these items). The variable potentially ranges from 0 to 6, depending on the number of provisions included in a trade agreement that restrict the imposition of antidumping duties and the use of subsidies and the safeguard provisions. In practice, the variable ranges from 0 (195 agreements) to 5 (4 agreements).43 Figure 3 offers a first test of our argument concerning a relationship between escape flexibility and flexibility strings. The relationship between the two variables is positive as expected following Hypothesis 3. As escape flexibility increases, countries make its use more difficult. The relationship is slightly non-linear, however: at the maximum of escape flexibility, flexibility strings are slightly less pronounced than when escape flexibility takes the value of three. As can be seen in the Appendix (Figure A-4), this non-linearity is driven by a decline in the number of WTO restrictions on safeguard provisions in highly flexible agreements. Figure 3 About Here

42

This variable partly overlaps with Kucik’s (2012) operationalization of “rigidity”. We also consider rules concerning subsidies, however, and use references to WTO rules as a short-cut to many of the constraints included in Kucik’s measure. 43 Again, this variable is highly correlated with one that relies on Item Response Analysis (rs = 0.99).

16

Explaining Flexibility in PTAs Going beyond the descriptive evidence, we estimate multivariate regression models that explain the three variables Escape Flexibility, Transitional Flexibility and Flexibility Strings. As our dependent variables are either ordinal (Escape flexibility and Flexibility strings) or censored (Transitional Flexibility), we use ordered probit and tobit estimations, respectively. We choose PTA as unit of analysis and thus consciously depart from previous PTA research that mainly used dyad-year or PTA-country-year as unit of analysis.44 The reasons for our choice are two-fold. First, the design of a PTA does not vary across member countries in our dataset. By using dyad-year or PTA-country-year as units of analysis, therefore, we would multiply the values of our dependent variable in plurilateral agreements. Second, with the exception of a small number of regional agreements (in particular, the European Union), the design of PTAs does not vary over time in our dataset. By dropping the time dimension, therefore, we do not lose any information. In fact, using PTA as unit of analysis is a conservative choice to avoid inflating the number of observations and artificially reducing standard errors.

Control Variables In the following multivariate models, we include several control variables that allow us to deal with potentially confounding factors. Since our unit of analysis is the PTA, we take the mean of each variable across member countries for continuous variables.45 In terms of economic variables, we use GDP per capita (GDP pc) and total GDP (GDP ) to capture the income level and economic importance of a country. Larger and richer countries might be able to negotiate agreements with a higher degree of depth and less flexibility than poorer or smaller countries. Moreover, we add T rade, which is the mean of the value of exports and imports across all the members of a PTA. As is the 44

For the former, see Mansfield et al. 2002. For the latter, see Kucik 2012. Our results are similar if we use the median or minimum value across member countries. See the robustness checks below. 45

17

practice in the empirical trade literature, we take the log of this value. We expect the amount of trade to positively influence the demand for agreements with a high degree of both depth and flexibility. Previous studies have shown that domestic institutions are important determinants of the design of international institutions and international organizations.46 We thus include a democracy score (Regime), which comes from Polity IV.47 Since this variable is ordinal, we use the median instead of the mean across PTA members. Our expectation is that democracies should sign deeper and more rigid agreements than autocracies. Furthermore, we add a variable that captures whether a country has recently undergone a transition from autocracy to democracy (Democratization). This variable scores one if at least one country has become a democracy over the past ten years using Polity IV as indicator. The expectation is that democratizing countries require more flexibility since they face high levels of uncertainty about future states of the world. We also add a dummy that scores one if every PTA member is a member of the WTO (W T O). WTO members tend to implement trade policies that differ from those of countries that are not part of this international organization.48 They also have ad hoc flexibility provisions upon which they can rely. Finally, we include the number of member countries of a PTA (N o. M embers). On the one hand, N o.M embers accounts for the broader-deeper trade-off argument in international agreements.49 On the other hand, Koremenos et al. argue that the size of membership is a crucial driver of flexibility.50 Table 1 summarizes the descriptive statistics of the dependent and independent variables. Table 1 About Here

46

Mansfield et al. 2008. Results do not change if we use other measures of democracy such as Cheibub et al. 2010. 48 Mansfield and Reinhardt 2003. 49 Gilligan 2004. 50 Koremenos et al. 2001 47

18

Statistical Models Since Escape Flexibility and Flexibility Strings are ordinal variables, we use ordered probit to estimate the models including them as dependent variables. The “omodel ” test in Stata 12 confirms that the proportional odds assumption holds and that an odered probit regression hence is appropriate. We rely on a tobit model in the equations that have Transitional Flexibility as dependent variable, since this is an interval variable that is truncated at 0. We note that a tobit model is particularly well-suited in our case, since observations cluster at the limiting value, i.e. 50% of our observations score zero . Tobit models produce unbiased estimates under these circumstances.

Findings Table 2 shows the results of the baseline models with the findings supportive of our hypotheses. Indeed, the coefficients of both Depth and Escape Flexibility are positive and statistically significant at the conventional level. Thus, deepening a PTA has a positive effect on escape flexibility and on transitional flexibility. Similarly, increasing the escape flexibility of a PTA has a positive impact on its flexibility strings. Paired with the descriptive evidence provided above, these models offer encouraging support for our argument.51 Table 2 About Here In Table 3, we summarize the magnitude of the substantive effects of Model 2. Moving Depth from its minimum to its maximum value increases by 60 percent the probability of having the highest value of Escape Flexibility, i.e. four. The results of Model 4 indicate that moving Depth from its minimum to its maximum value increases the tariff transition period by 13 years. The transition period increases by 6 years if Depth moves from a standard deviation below the mean to a standard deviation above the mean. These substantive effects are so sizeable that in explaining the flexibility 51

In the multilevel analysis we lose some PTAs since our control variables have incomplete coverage.

19

of PTAs a model that neglects the role of Depth suffers from serious missing variable problems. Finally, in Table 4, we summarize the magnitude of the substantive effects of Model 6. Moving Escape Flexibility from its minimum to its maximum value decreases by 88 percent the probability of having the lowest value of Flexibility Strings, i.e. zero. Similarly, moving Escape Flexibility from its minimum to its maximum value increases by 50 percent the probability of observing values of Flexibility Strings that are 3 or higher.52 Table 3 and Table 4 About Here Importantly, control variables have the expected sign when they are statistically significant at the conventional level. Thus, using PTA as unit of analysis does not seem to distort the impact of monadic and dyadic variables on the design of PTAs. We perform several tests to check the robustness of our findings. First, we replace the indicators of depth, escape flexibility, and flexibility strings obtained by summing provisions with the indicators of depth, escape flexibility, and flexibility strings using item response analysis (see the discussion above). Second, we use the minimum values of all the covariates to capture the weakest link among PTA members. Third, we use multinomial logistic regression to explain escape flexibility and flexibility strings, which allows us to see how the impact of our main covariates and control variables changes for different values of the two outcome variables. Fourth, we take the natural logarithm of transitional flexibility to improve the normality of this outcome variable. We do this because tobit models are very sensitive to a potential non-normal distribution. Fifth, we drop from the analysis EU treaties, EU PTAs with third countries, and US PTAs to check whether the results are driven by a few agreements with particular features. Finally, we use binary probit regressions having transformed the flexibility variables into dummy variables. Since our flexibility variables are ordinal, we are ultimately interested to see the impact of depth on low- and high-flexibility PTAs. For 52

In our dataset, only very few PTAs have values that are equal to 4 or 5.

20

all these checks, the results, available in the Appendix, are similar to the ones reported above.

Instrumental Variables Approach While the previous section has shown that our results are very robust, it may still be argued that states decide on the depth and flexibility of international agreements, simultaneously. That is, it may be that not only does depth influence flexibility, but also flexibility impacts on depth. To account for this possible endogeneity, we implement an instrumental variable approach. As already mentioned, our main variables are ordinal and interval. As such, we cannot rely on a classic IVREG estimation, which would produce biased and inconsistent coefficients. Thus, we use conditional recursive mixed process estimators as developed by David Roodman.53 Mixed process implies that different equations can have different kinds of dependent variables (response types), i.e. interval-censored and ordered probit. That grants the flexibility to regress an interval variable on an endogenous ordinal one, for instance. Recursive implies that we are only able to fit sets of equations with clearly defined stages, whereas we are unable to estimate simultaneous causation. In particular, we rely on a limited-information (LIML) estimator, in which only the final stage’s coefficients are structural, while at the first stage we only have reduced-form parameters.54 Formally, we estimate the following three sets of equations:

Escape F lexibilityi = α1 + β1 (Depthi = Z1i + Z2i ) + β2 Xi + 1 .

(1)

T ransitional F lexibilityi = α2 + β3 (Depthi = Z1i + Z2i ) + β4 Xi + 2 .

(2)

53

Roodman 2011. We use the command ‘cmp’ in Stata 12. The estimation relies on the Geweke-HajivassiliouKeane algorithm for estimating cumulative normal densities above dimension (command ‘ghk2’ in Stata 12). Differently from GLLAMM, ‘cmp’ uses simulation. That is, it makes many draws from the hypothesized normal distributions of the effects, computing the implied likelihood for each set of draws before averaging. See Train 2009; Greene 2011, chap 15. 54

21

F lexibility Stringsi = α3 + β5 (Escape F lexibilityi = Z3i + Z4i ) + β6 Xi + 3 .

(3)

where Escape Flexibility, Transitional Flexibility and Flexibility Strings are dependent variables. Depth is the main independent variable in equations 1 and 2, and Escape Flexibility is the main independent variable in equation 3. Z1i , Z2i , Z3i , and Z4i are instruments, Xi are vectors of control variables, β1 , β2 , . . . , β6 are the coefficients, α1 , α2 , and α3 are constants, and 1 , 2 , and 3 are the error terms. Equation 1 is a tobit in the first stage and an ordered probit in the second stage. Equation 2 is a tobit model in both first and second stage and Equation 3 is an ordered probit in both first and second stage.

Identification Strategy The key challenge for this approach is to find suitable variables to instrument Depth and Escape Flexibility. Useful instruments need to be good predictors of Depth (equations 1 and 2) and Escape Flexibility (equation 3), but unrelated with the error terms of Escape Flexibility and Transitional Flexibility once we include the control variables that account for alternative causal paths. To instrument Depth we use the depth of GATT/WTO agreements signed by member countries in the various negotiation rounds (GATT/WTO Depth). Since the GATT/WTO is the largest and most important trade agreement, it is safe to assume that provisions bargained for and included in the GATT/WTO treaties inspired the design of bilateral and plurilateral agreements. For instance, the signature of agreements concerning intellectual property rights and trade-related investment measures in the WTO triggered the inclusion of similar provisions in PTAs. Equally, the services provisions included in PTAs are very similar to those of the General Agreement on Trade in Services.55 To empirically measure the depth of GATT/WTO agreements, we use the same operationalization as for Depth (IR).56 Importantly, as described above, 55

World Trade Organization 2011, 134 Results (available upon request) do not differ when using GAT T /W T O Depth (index), though this instrument is slightly weaker than GAT T /W T O Depth (IR). 56

22

this operationalization only includes items related to the depth of agreements; this ensures that GATT/WTO Depth does not have a direct impact on the flexibility of PTAs. We also instrument the depth of PTAs using the score of the deepest EU PTA (EU Depth).57 The reason for this choice is that the EU provides “the standard model for regional integration”.58 In other words, the EU could serve a similar purpose to the one outlined above in respect to the GATT/WTO. The correlation between W T O Depth and the error term is 0.1, whereas the correlation between EU Depth and the error term is -0.04.59 To instrument Escape Flexibility, we rely on a variable that captures the level of transparency of PTA member countries, since earlier studies show that transparency affects the supply side of flexibility.60 The reasoning is that domestic political changes, adjustment costs, and distributional problems constitute private information. Countries then have an incentive to misrepresent their private information in order to freeride on cooperation. Since each country has an incentive to overstate its costs of compliance at any time, in the absence of transparency trade partners cannot distinguish between the use of an escape clause due to serious difficulties or due to opportunistic behavior. This follows naturally from Bayesian updating, as the sources of any given defection can be seen as coming from either forced emergency measures or opportunism, and is in line with previous studies in the field.61 For instance, the US-Vietnam PTA (2000) is less flexible than would be expected from its level of depth. Vietnam’s low 57

We again use the operationalization of depth relying on item response analysis. The results, which are available upon request, are similar, though slightly weaker, if we use Depth (index). Also as before, the operationalization only includes items related to the depth of agreements; this ensures that EU Depth does not have a direct impact on the flexibility of PTAs. 58 B¨ orzel and Risse 2012, 196. 59 If we run classic instrumental variable regressions in Model 2 and Model 4, the Cragg-Donald Wald F test of the first stage is 56.85 and 56.54, respectively. Thus, there is no concern that our instruments are weak. Moreover, results from the Sargan-Hansen test raise no concern of overidentification (that is, p > 0.1), as there is no relationship between the instruments and the residuals, which means that the instruments are valid. In the appendix we show that the correlation between the residuals of the second stage and GATT/WTO Depth as well as the residuals of the second stage and EU Depth are very low for both Model 2 and Model 4. 60 Svolik 2006. 61 Svolik 2006.

23

level of transparency might account for this ‘missing’ flexibility. Operationalizing transparency is difficult. We rely on the amount of information that states are willing (and able) to release. Since measures of flexibility are invoked due to economic difficulties, the capacity of communicating on the status of a country’s economy seems to us a more direct way of capturing transparency than the sole presence of a democratic regime. Concretely, we use the measure of transparency developed by Hollyer et al. who look at the availability (or absence) of policy-relevant data in the World Development Indicators database.62 Because the resulting variable (T ransparency) has limited variation, we use a linear regressor and the quadratic regressor, which are both statistically significant at the conventional level. In theory, T ransparency could be endogenous to Depth. Including a democracy score on the right-hand side, however, should account for the impact of domestic institutions on Depth. For instance, Svolik conceptualizes transparency with the presence of a democratic regime in general, and elections, in particular.63 In other words, the residuals of Depth should be uncorrelated with T ransparency and T ransparency 2 , granting the validity of the exclusion restriction. Indeed, Figure A-5 in the Appendix shows that n the Appendix we show that the correlation between the residuals of the second stage and T ransparency as well as the residuals of the second stage and T ransparency 2 is very low, i.e. ρ = −0.06 and ρ = −0.09 for T ransparency and T ransparency 2 respectively.64

Findings from the instrumental variables approach The results showed in Table 5 again support our expectations. First, the coefficient of Depth has the expected positive sign in every model where Escape Flexibility and Transitional Flexibility are the dependent variables and is statistically significant at the 99 percent confidence level. Second, the coefficient of Escape Flexibility has the 62

Hollyer et al. 2011; World Bank 2012. Svolik 2006. 64 We take the error term from the extended model (Model 6). 63

24

expected positive sign when Flexibility Strings is the dependent variable and is statistically significant at the conventional level. The findings thus are completely in line with our argument that emphasizes the importance of the endogenous determinants of the design of PTAs. Table 5 About Here Several factors lend confidence to this approach. First, the instruments have the expected sign and are statistically significant at the conventional level in the first stage. This supports our identification strategy. Second, the results on ρ lead us to reject the hypothesis that the two equations are independent. Third, control variables have the expected sign when they are statistically significant. Finally, the instrumental variables approach is robust if we use biprobit regressions, which rely on dummies capturing escape flexibility and flexibility strings. The results of these models are available in the Appendix and are very similar to the ones reported above.

Conclusion Trade agreements vary strongly in both their depth and their flexibility. We have argued that these two dimensions of institutional design are interdependent, as flexibility increases with depth. In addition, we have reasoned that states use various flexibility measures (transitional and escape flexibility) for the same purpose, namely to buy off certain societal groups (in our case, import-competing interest groups). Finally, we have conjectured that concerns about the agreement’s stability lead governments to explore various strings to control the unfettered use of the flexibility tools. We have tested these arguments drawing on an original dataset on the design of 587 trade agreements. PTAs are an ideal testing ground for our arguments, as they allow us to keep issue area characteristics constant, which may also explain variation in flexibility. Moreover, looking at only one type of agreement made it possible to operationalize the depth of agreements, which would be very difficult in datasets that 25

feature agreements from very different issue areas.65 Using this dataset, we have found support for our arguments. The depth of agreements is positively related to both types of flexibility. In addition, the results show that as escape flexibility increases, we see greater restrictions on the use of the escape flexibility provisions. These results are robust to changing the operationalization of our key variables and using instrumental variables. Our findings show that different design features of international agreements are not independent of each other. Flexibility is to a large extent determined by the depth of the international agreement. To explain the flexibility of international agreements, therefore, it is not sufficient to look at characteristics of the member states or issue characteristics. In addition, the results suggest that flexibility is an important ingredient when it comes to the overall stability of trade agreements. Flexibility mechanisms may contribute to agreements being self-enforcing.66 This might help explain the lack of dispute settlement cases under PTA laws notwithstanding more or less legalized systems to settle disputes. As to the question whether governments strive for “optimal obfuscation”, meaning that they try to substitute transparent trade barriers such as tariffs with less transparent ones, such as escape clauses, the paper suggests some caution.67 While greater depth involves tariff cuts, states design selective strings on some non-tariff barriers. Therefore, flexibility tools do not simply allow for new forms of unfettered protection. The relationship between tariff liberalization and other forms of behind-the-border discrimination in the PTA context is clearly more nuanced.

65

The “Continent of International Law” dataset, for example, contains agreements from four issue areas: economics, environment, human rights, and security. See Koremenos 2013. It is difficult to see how depth could be operationalized across these four issue areas. 66 Rosendorff 2005. 67 For the “optimal obfuscation” argument, see Kono 2006.

26

References Alvarez, Jose E. 2005. International Organizations as Law-Makers. Oxford: Oxford University Press. Baccini, Leonardo and Andreas D¨ ur. 2012. “The new regionalism and policy interdependence.” British Journal of Political Science 42(1): 57–79. Baier, Scott L. and Jeffrey H. Bergstrand. 2007. Do Free Trade Agreements Actually Increase Members’ International Trade? Journal of International Economics 71(1):72–95. Bartholomew, David J., Martin Knott, and Irini Moustaki. 2011. Latent Variable Models and Factor Analysis: A Unified Approach. Chichester: John Wiley and Sons. B¨orzel, Tanja A. and Thomas Risse. 2012. When Europeanisation Meets Diffusion: Exploring New Territory. West European Politics 35(1):192–207. B¨ uthe, Tim and Helen V. Milner. 2011. Institutional Diversity in Trade Agreements and Foreign Direct Investment: Credibility, Commitment, and Economic Flows in the Developing World, 1970-2007. Manuscript. Chase, Kerry A. 2003. “Economic Interests and Regional Trading Arrangements: The Case of NAFTA.” International Organization 57(1): 137–174. Chase, Kerry A. 2005. Trading Blocs: States, Firms, and Regions in the World Economy. Ann Arbor: University of Michigan Press. Cheibub, Jose A., Jennifer Gandhi and James R. Vreeland. 2010. Democracy and Dictatorship Revisited. Public Choice 143(1-2):67–101. Downs, George W., David M. Rocke and Peter N. Barsoom. 1996. Is the Good News about Compliance Good News about Cooperation? International Organization 50(3):379–406. D¨ ur, Andreas. 2010. Protection for Exporters: Power and Discrimination in Transatlantic Trade Relations, 1930-2010. Ithaca: Cornell University Press. D¨ ur, Andreas, Leonardo Baccini, Manfred Elsig, and Karolina Milewicz. 2013. The

27

Design of International Trade Agreements: Introducing a New Database. Review of International Organizations. Egger, Peter and Georg Wamser. 2013. Effects of the Endogenous Scope of Preferentialism on International Goods Trade. Manuscript. Estevadeordal, Antoni, Kati Suominen and Robert Teh, eds. 2009. Regional Rules in the Global Trading System. Cambridge: Cambridge University Press. Gilligan, Michael J. 1997. Lobbying as a Private Good with Intra-Industry Trade. International Studies Quarterly 41(3):455–474. Gilligan, Michael J. 2004. Is There a Broader-Deeper Tradeoff? International Organization 58(3):459–484. Goodman, Ryan and Derek Jinks. 2004. How To Influence States: Socialization and International Human Rights Law. Duke Law Journal 54(3):621–704. Greene, William H. 2011. Econometric Analysis, 7th ed. Harlow: Pearson Education. Grossman, Gene M. and Elhanan Helpman. 1994. Protection for Sale. The American Economic Review 84(4):833–850. Hafner-Burton, Emilie M. 2009. Forced to Be Good: Why Trade Agreements Boost Human Rights. Ithaca: Cornell University Press. Haftel, Yoram Z. 2013. Commerce and Institutions: Trade, Scope, and the Design of Regional Economic Organizations. Review of International Organizations, forthcoming. Hall, Richard L. and Alan V. Deardorff. 2006. Lobbying as Legislative Subsidy. American Political Science Review 100(1):69–84. Helfer, Laurence R. 2005. Exiting Treaties. Virginia Law Journal 91(7):1579–1648. Helfer, Laurence R. 2013. Flexibility in International Agreements. In Interdisciplinary Perspectives on International Law and International Relations, edited by Jeffrey Dunoff and Mark Pollack, 175–196. Cambridge: Cambridge University Press. Hollyer, James R. and Peter B. Rosendorff. 2012. Leadership Survival, Regime Type, Policy Uncertainty and PTA Accession. International Studies Quarterly 56(4):748-

28

764. Hollyer, James R., Peter B. Rosendorff and James Raymond Vreeland. 2011. “Democracy and Transparency.” The Journal of Politics 73(4):1191–1205. Kono, Daniel. 2006. Optimal Obfuscation: Democracy and Trade Policy Transparency. American Political Science Review 100(3):369–384. Johns, Leslie. 2012. Depth versus Rigidity in the Design of International Trade Agreements. Manuscript, UCLA, Los Angeles, CA. Koremenos, Barbara. 2005. Contracting around International Uncertainty. American Political Science Review 99(4):549–565. Koremenos Barbara and Allison Nau. 2010. Exit, No Exit. Duke Journal of Comparative and International Law 81(1):81–120. Koremenos, Barbara, Charles Lipson and Duncan Snidal. 2001. The Rational Design of International Institutions. International Organization 55(4):761–799. Koremenos, Barbara. 2013. The Continent of International Law. Journal of Conflict Resolution 57(4): 653–681. Kucik, Jeffrey. 2012. The Domestic Politics of Institutional Design: Producer Preferences over Trade Agreement Rules. Economics & Politics 24(2):95–118. Kucik, Jeffrey and Eric Reinhardt. 2008. Does Flexibility Promote Cooperation? An Application to the Global Trade Regime. International Organization 62(3):477– 505. Kydland, Finn E. and Edward Prescott. 1977. Rules Rather than Discretion: The Inconsistency of Optimal Plans. Journal of Political Economy 85(3):473-490. Manger, Mark. 2009. Investing in Protection: The Politics of Preferential Trade Agreements Between North and South. Cambridge: Cambridge University Press. Mansfield, Edward D. and Helen V. Milner. 2012. Votes, Vetoes, and the Political Economy of International Trade Agreements. Princeton: Princeton University Press. Mansfield, Edward D. and Eric Reinhardt. 2003. Multilateral Determinants of Re-

29

gionalism: The Effects of GATT/WTO on the Formation of Preferential Trading Arrangements. International Organization 57(4):829–862. Mansfield, Edward D. and Eric Reinhardt. 2008. International Institutions and the Volatility of International Trade. International Organization 62(4):621–652. Mansfield, Edward D., Helen V. Milner, and Jon C. Pevehouse. 2008. Democracy, Veto Players and the Depth of Regional Integration. The World Economy 31(1):67–96. Mansfield, Edward D., Helen V. Milner and B. Peter Rosendorff. 2002. Why Democracies Cooperate More: Electoral Control and International Trade Agreements. International Organization 56(3):477–513. Maskus, Keith E. and Mohan Penubarti. 1995. How Trade-Related Are Intellectual Property Rights? Journal of International Economics 39(3):227–248. Milner, Helen V. 1988. Resisting Protectionism: Global Industries and the Politics of International Trade. Princeton: Princeton University Press. Neumayer, Eric. 2007. Ratification: Explaining Reservations to International Human Rights Treaties. Journal of Legal Studies 36(2):397–430. Pelc, Krzysztof J. 2009. Seeking Escape: The Use of Escape Clauses in International Trade Agreements. International Studies Quarterly 53(2):349–368. Pelc, Krzysztof J. 2010. Eluding Efficiency: Why Do We Not See More Efficient Breach at the WTO? World Trade Review 9(4):629–642. Piermartini, Roberta and Michele Budetta. 2009 A Mapping of Regional Rules on Technical Barriers to Trade. In Bank Regional Rules in the Global Trading System, edited by Antoni Estevadeordal, Kati Suominen and Robert Teh, 250–315. Cambridge: Cambridge University Press. Roodman, David. 2011. Estimating Fully Observed Recursive Mixed-Process Models with cmp. Stata Journal 11(2):159–206. Rosendorff, B. Peter. 2005. Stability and Rigidity: Politics and Design of the WTO’s Dispute Settlement Procedure. American Political Science Review 99(3):389–400. Rosendorff, B. Peter and Helen Milner. 2001. The Optimal Design of International

30

Trade Institutions: Uncertainty and Escape. International Organization 54(4):829– 857. Shaffer, Gregory C. and Mark A. Pollack. 2010. Hard vs. Soft Law: Alternatives, Complements and Antagonists in International Governance. Minnesota Law Review 94(3):706–799. Svolik, Milan. 2006. Lies, Defection, and the Pattern of International Cooperation. American Journal of Political Science 50(4):909–925. Train, Kenneth. 2009. Discrete Choice Methods with Simulation, 2nd ed. Cambridge: Cambridge University Press. World Bank. 2012. World Development Indicators. World Trade Organization. 2011. World Trade Report – The WTO and Preferential Trade Agreements: From Co-existence to Coherence. Geneva: World Trade Organization. Wright, John R. 1996. Interest Groups and Congress: Lobbying, Contributions, and Influence. Boston: Allyn and Bacon.

31

4 3 2 0

1

Escape flexibility

0

1−5

6−10

11−25 26−40

Depth (index)

25

Figure 1: Depth versus escape flexibility.

Carif KR−US

20

AEC

CO−NT

15 10 5 0

Transitional flexibility

CL−VE

CAFTA CAFTA−DO PA−US

ECOTA CL−CO CL−PE CO−US EU−TR And−MERCOSUR CN−PE PE−SG WAEC CL−PE SAFTA Cent−CL RU−UA CL−MERCOSUR PE−US PICTA LAFTA JP−VNCA−CO GAFTA EE−FI ESAS EG−JO EFTA−FI CA−EFTA GT−TWJP−MYCL−JP BO−MX ECOWAS JO−SG JP−PH JP−TH CR−MX EU−CI ECCAS EU−PT1 ID−JP MX−NINI−TW EE−FO EG−TR JO−TNAR−BR CL−PA NAFTA EU−PT CARIFTA SV−HN−TW EU−SECARICOM−CU GB−IEPE−TH EU−IL CARICOM EU−GR SY−TR EG−JO JO−KW PA−SGCL−US AU−NZJO−MA EC−HE GCC JO−TR DZ−EU CU−MERCOSUR EU−FI EU−CHEU−TR1MK−TREFTA−JOEFTA−MA CL−CU ASEAN−AU−NZ JO−USAU−TH JP−MXKR−SGCO−EFTA FI−BG AR−BR CZ−EFTA MK−UA MA−TR EU−ZANZ−TH MX−UY CY−EU Gr3 BO−MERCOSUR CA−US JP−SG TPS CA−PE CL−KR AU−US BN−JP CL−MX CEMAC ECOWAS1 And EU−CL EUCL−CN MK−SI EU−MXCACM BG−EU EFTA−KR EFTA−MXMA−US BE−EU AT−EU MD−MK TN−TR IL−MERCOSUR CA−CR BH−US OM−US CARICOM−CO IN−KR CN−NZ ASEAN CEFTA LT−TR EFTA−MK BZ−GT PA−TW LT−PL IL−US COMESA AL−MD MERCOSUR−PE MY−PK CL−MX ACM HU−SIEFTA D8 AL−MKMK−RO BG−EFTA CL−TRAL−BG CA−CLMY−NZ AL−HRLV−PL EU−RS IL−PSHU−TRHR−RS AU−CL EU−IL AL−RO IL−MX YaoI EU−ISAR−CL ASEAN−KR1 CN−PK EU−HR FI−CSFR AR−CU AL−TRHR−TR EU−NO AU−NZ HR−EFTAEU−HUEU−PLAL−EU GCC1 OECS ACM CZ−TR TEA BA−TR EFTA−SI EU−SI EU−IL EFTA−GCCIN−SGFI−PL EU−ES EFTA−TN CZ−IL BG−ILME−TREFTA−RS PK−LK EFTA−ESCL−HK BA−MK IL−TRAL−BACZ−EU EU−SKCARICOM−CRJO−AE GT−MX BA−MD JO−SY CL−MO AL−BAAL−RSBA−HR Agd CA−IL BG−BA HR−SI CN−SG MU−PK HU−LT CZ−LV BA−RS BA−SI RO−RSFI−HU IN−LK EU−SY IL−SI CZ−SI ASEANBA−RO BG−MD EU−EG IL−PL BG−RSEU−RO AD−EU HR−MD UDEAC EU−TN IL−SK MERCOSUR GUAM LT−SI

0

10

20

30

Depth (index)

Figure 2: Depth versus transitional flexibility.

32

40

5 4 3 2 0

1

Flexibility strings

0

1

2

3

4

Escape flexibility

Figure 3: Escape flexibility versus flexibility strings.

33

Table 1: Descriptive statistics. Variables Escape Flexibility Transitional Flexibility Flexibility Strings Depth GDPpc GDP Trade Democracy Democrat. WTO Distance No. Members Transparency WTO Depth EU Depth

Mean 2.68 3.84 1.53 8.13 9.10 23.54 4.97 4.74 0.85 0.50 7.99 5.65 0.68 0.60 1.49

34

Std. Dev. 1.46 5.21 1.35 10.03 0.86 1.83 2.35 5.82 0.36 0.50 .99 8.65 0.30 0.85 0.61

Min 0 0 0 0 5.91 17.34 0.06 -10 0 0 4.47 2 0 -1.04 0.48

Max 4 25 5 40 10.83 27.80 12.03 10 1 1 9.87 91 1 1.20 2.35

Table 2: Baseline models. Covariates Depth

(1) Escape

(2) Escape

0.05*** (0.01)

0.05*** (0.01)

(3) (4) (5) Transitional Transitional Strings 0.36*** (0.36)

0.34*** (0.37)

EscapeFlexibility GDP

-0.06 (0.04) 0.36*** (0.08) -0.03 (0.03)

GDPpc Trade Regime Democratization GATTWTO NoMembers Cut1

0.75 (0.83) 1.29 (0.83) 1.79** (0.83) 2.29*** (0.83)

Cut2 Cut3 Cut4

-0.07 (0.05) 0.27*** (0.08) -0.03 (0.03) 0.04*** (0.01) 0.39*** (0.14) 0.05 (0.11) -0.01 (0.01) 0.15 (0.93) 0.71 (0.93) 1.22 (0.93) 1.75* (0.93)

-0.31 (0.36) -0.75 (0.61) 0.66** (0.26)

-0.48 (0.37) -0.71 (0.66) 0.64** (0.26) -0.06 (0.08) 0.70 (1.11) 2.77*** (0.90) 0.08 (0.05)

Cut5 Constant

Observations

565

8.42 (7.12)

9.97 (7.73)

497

494

561

0.76*** (0.05) 0.04 (0.05) 0.15* (0.08) 0.14*** (0.03)

4.54*** (0.87) 5.29*** (0.87) 6.15*** (0.88) 7.46*** (0.89) 8.49*** (0.91)

0.80*** (0.05) 0.06 (0.05) 0.13 (0.09) 0.11*** (0.03) -0.01 (0.01) -0.03 (0.16) 0.61*** (0.11) -0.01 (0.01) 4.95*** (0.97) 5.75*** (0.98) 6.64*** (0.99) 8.00*** (1.00) 9.07*** (1.01)

548

545

Models 1, 2, 5, and 6 are ordered probit. Models 3 and 4 are tobit models. Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.

35

(6) Strings

Table 3: Predictions for depth on escape flexibility (Model 2). Value 0 1 2 3 4

Min → Max C.I. -0.14 [-0.18, -0.11] -0.15 [-0.219, -0.11] -0.17 [-0.21, -0.13] -0.13 [-0.18, -0.09] 0.60 [0.47, 0.69]

36

Table 4: Predictions for escape flexibility on flexibility strings (Model 6). Value 0 1 2 3 4 5

Min → Max C.I. -0.88 [-0.93, -0.83] 0.09 [0.05, 0.13] 0.31 [0.26, 0.26] 0.40 [0.35, 0.46] 0.08 [0.05, 0.11] 0.01 [0.002, 0.02]

37

Table 5: Instrumental variable models. Covariates Depth

(7) Escape

(8) Escape

0.10*** (0.01)

0.10*** (0.01)

(9) (10) (11) TransitionalTransitional Strings 0.11* (0.06)

0.11* (0.06)

EscapeFlexibility GDP GDPpc Trade

-0.06 (0.04) 0.10 (0.09) -0.11*** (0.03)

Regime Democratization GATTWTO NoMembers Cut1 Cut2 Cut3 Cut4

-1.43* (0.85) -0.97 (0.86) -0.54 (0.87) -0.10 (0.88)

-0.06 (0.04) 0.07 (0.09) -0.10*** (0.03) 0.03*** (0.01) 0.35** (0.13) -0.18* (0.11) -0.01 (0.01) -1.48 (0.92) -1.00 (0.93) -0.55 (0.94) -0.09 (0.94)

0.13 (0.19) -0.46*** (0.39) 0.40** (0.16)

0.04 (0.20) -0.55 (0.40) 0.35** (0.16) -0.007 (0.04) -0.01 (0.59) 1.47*** (0.53) 0.04 (0.03)

Cut5 Constant

2.02 (4.30)

Transparency2

GATTWTODepth ρ

Observations

1.02*** (0.05) -0.07 (0.04) 0.62*** (0.08) -0.01 (0.03)

3.57*** (0.88) 4.25*** (0.90) 5.03*** (0.91) 6.23*** (0.94) 8.49*** (0.91)

1.07*** (0.05) -0.09** (0.05) 0.48*** (0.09) -0.00 (0.03) 0.04*** (0.01) 0.29* (0.15) 0.27** (0.11) -0.00 (0.01) 4.16*** (0.97) 4.87*** (0.98) 5.67*** (1.00) 6.89*** (1.02) 9.07*** (1.01)

-1.75** (0.70) 2.19*** (0.63)

-1.68** (0.62) 1.75*** (0.65)

3.69 (8.14)

Transparency

EUDepth

(12) Strings

3.98*** (0.97) 1.76*** (0.66) -0.55*** (0.07)

4.61*** (1.00) 1.26* (0.67) -0.51*** (0.08)

3.45*** (1.13) 2.13*** (0.75) 0.26*** (0.10)

4.10*** (1.14) 1.63** (0.63) 0.23*** (0.10)

-0.60*** (0.13)

-0.66*** (0.13)

571

567

497

494

565

561

Models 7 and 8 are tobit models in the first stage and ordered probit in the second stage. Models 9 and 10 are tobit models in both first and second stage. Models 11 and 12 are ordered probit in both first and second stage. All models are estimated using a conditional recursive mixed process. Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.

38

Appendix

Provision

Operationalizing depth Service chapter Monopoly IPR general TBT Cooperation Service continuous TBT WTO SPS WTO GATS reference SPS Cooperation TRIPS Services national treatment Procurement WTO IPR MFN International standards Competition chapter Services non−establishment Investment chapter Transfers and payments Paris Convention Establishment Investments national treatment Bern Convention Post−establishment operation Pre−establishment operation Acquisitions Movement of natural persons Standards of treatment Negative list approach Services MFN Non−distorting standards Investments MFN Procurement chapter Rome Convention Movement of natural persons (investments) Competition coordination Procurement national treatment Procurement transparency IPR substantive standards Geographic indications SPS harmonization WIPO copyright treaty WIPO Phonograms Treaty Mergers Pharmaceuticals in IPRs TBT harmonization National competition authority Common competition authority TRIM reference

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

0

10

20

30

40

50

Percent of agreements

Figure A-1: The items included in depth.

Services [Service chapter] Does the agreement contain a reference to the liberalization of trade in services? [GATS reference] Does the agreement contain a reference to the General Agreement on Trade in Services (GATS)? [MFN treatment] Does the agreement contain an MFN clause for services? [Negative list approach] Does the agreement foresee a negative list approach to services liberalization? [Services national treatment] Does the agreement contain a national treatment clause for services? [Services non-establishment] Does the agreement grant the right of non-establishment for service provision (that is, does it allow the provision of services without local presence)?

39

[Movement of natural persons] Does the agreement allow the movement of natural persons in the provision of services? [Service continuous] Does the agreement contain a review provision for the services provisions? Investments [Investment chapter] Does the agreement contain substantive investment provisions? [Standards of treatment] Does the agreement contain provisions that grant compensation to investors in case of strife and/or expropriation? [TRIM reference] Does the agreement contain a reference to the WTO agreement on trade-related investment measures? [Pre-establishment operation] Does the agreement contain non-discrimination provisions in relation to pre-establishment operations? [Establishment] Does the agreement contain non-discrimination provisions in relation to establishment (e.g. greenfield investments)? [Post-establishment operation] Does the agreement contain non-discrimination provisions in relation to post-establishment operation (e.g. the free movement of capital and resale)? [Mergers] Does the agreement contain non-discrimination provisions in relation to acquisition (e.g. mergers)? [Investments MFN] Does the agreement grant MFN treatment on investments? [Investments national treatment] Does the agreement grant national treatment on investments? [Transfers and payments] Does the agreement mention specific restrictions regarding transfers and payments? [Movement of natural persons (investments)] Does the agreement mention restrictions related to the temporary movement of business or natural persons? Intellectual Property Rights [IPR general] Does the treaty contain a provision on intellectual property rights (IPRs)? [IPR MFN] Does the treaty contain an MFN provision for IPRs? [Rome Convention] Does the treaty contain specific deadlines for acceding to the Rome Convention? [Paris Convention] Does the treaty contain specific deadlines for acceding to the Paris Convention? [Bern Convention] Does the agreement contain specific deadlines for acceding to the Bern Convention? [TRIPS] Does the agreement mention the TRIPS agreement? [WIPO copyright treaty] Does the agreement contain specific deadlines for acceding to the World Intellectual Property Organization (WIPO) copyright treaty? [WIPO Phonograms Treaty] Does the agreement contain specific deadlines for acceding to the WIPO Phonograms treaty? [IPR substantive standards] Does the agreement contain specific provisions in relation to substantive standards of protection?

40

[Pharmaceuticals in IPRs] Does the agreement contain references to pharmaceuticals? [Geographic indications] Does the agreement contain references to geographical indications? Public procurement [Procurement chapter] Does the agreement contain substantive provisions on public procurement? [Procurement national treatment] Does the agreement guarantee national treatment with respect to public procurement? [Procurement transparency] Does the chapter on public procurement include a transparency provision? [Procurement WTO] Does the agreement contain a reference to the WTO/GATT procurement agreements? Technical Barriers to Trade [TBT WTO] Does the agreement contain a reference to the WTO Agreement on technical barriers to trade (TBTs) (the GATT standards code)? [TBT Cooperation] Does the agreement call for cooperation and/or information exchange on TBTs? [Non-distorting standards] Does the agreement contain a requirement for standards to be least trade-distorting? [International standards] Does the agreement encourage the use of international standards? [SPS Cooperation] Does the agreement contain provisions calling for information exchange and technical cooperation on sanitary and phytosanitary (SPS) measures? [SPS WTO] Does the agreement contain a reference to the WTO SPS agreement? [TBT harmonization] Does the agreement contain provisions that stipulate the harmonization of standards? [SPS harmonization] Does the agreement contain provisions that stipulate the harmonization of SPS provisions? Competition [Competition chapter] Does the agreement contain a competition chapter? [National competition authority] Does the agreement contain a provision stipulating the establishment of a national competition authority? [Competition coordination] Does the agreement contain a provision stipulating coordination among national authorities? [Common competition authority] Does the agreement contain a provision stipulating the creation of a common authority/institution on competition? [Monopoly] Does the agreement contain a provision on monopolies and cartels? [Acquisitions] Does the agreement contain a provision on mergers and acquisitions?

41

Operationalizing escape flexibility and flexibility strings

Escape

Strings

AD allowed

●

CV allowed

●

BOP provision

●

Provision

WTO restriction AD

●

Safeguard

●

WTO restriction subsidy

●

WTO restriction safeguard

●

Common policy on subsidies

●

Safeguard only in transition period

●

De minimis dumping margin

●

0

20

40

60

0

20

40

60

Percent of agreements

Figure A-2: The items included in escape flexibility and flexibility strings.

Escape flexibility [AD allowed:] Does the agreement allow anti-dumping with specific provisions? [CV allowed:] Does the agreement mention the use of countervailing duties? [Safeguard:] Does the agreement foresee special safeguard provisions? [BOP provision:] Does the agreement include BOP provisions? Flexibility strings [WTO restriction AD:] Does the agreement make a reference to GATT/WTO antidumping provisions? [WTO restriction subsidy:] Does the agreement make a reference to GATT/WTO subsidy provisions? [WTO restriction safeguard:] Does the agreement make a reference to GATT/WTO safeguard provisions? [Common policy on subsidies:] Does the agreement foresee a common policy of parties on subsidies? [Safeguard only in transition period:] Does the agreement allow parties to use safeguard measures only during the transition period? [De minimis dumping margin:] Does the agreement foresee a de minimis dumping margin (or dumped volume) that differs from the GATT/WTO?

42

1.0

1.0

0.8 ●

0.4

0.6 0.4

0.0

0.2 0.0

●

1−5

6−10

26−40

0

1−5

6−10

26−40

1.0

Depth (index)

1.0

Depth (index)

0.8 0.6 ●

0.0

0.0

●

●

0.4

0.4

●

● ●

●

0.2

●

●

0.6

Countervailing (mean)

0.8

●

BOP (mean)

●

●

0.6

Safeguard (mean)

●

●

0

0.2

●

●

●

0.2

0.8

Anti−dumping (mean)

●

0

1−5

6−10

26−40

0

1−5

Depth (index)

6−10

26−40

Depth (index)

1.0 0.8 0.6 0.4

Safeguard transition

●

0.2

0.8 0.6 0.4

●

0.2

WTO safeguard

1.0

Figure A-3: Depth and the items included in escape flexibility.

●

●

0

1

●

●

●

0.0

0.0

●

2

3

4

●

●

0

1

3

4

1.0

Escape flexibility

1.0

Escape flexibility

2

0.8 0.6

De minimis

0.6

0.2

0.4 0.2

WTO AD

●

0.4

0.8

●

●

0.0

0.0

● ●

0

1

2

3

4

●

●

●

●

●

0

1

2

3

4

0

1

0.8 0.6 0.4 0.0

●

0.2

0.4 0.0

● ●

Common subsidy policy

0.6

● ●

0.2

WTO Subsidy

0.8

1.0

Escape flexibility

1.0

Escape flexibility

2

3

4

● ●

0

Escape flexibility

●

1

●

●

2

3

4

Escape flexibility

Figure A-4: Escape flexibility and the items included in flexibility strings.

43

Results from robustness checks Table A-1: Models with variables operationalized using item response analysis. VARIABLES Depth (IR)

(A1) Escape

(A2) Escape

(A3) Transitional

(A4) Transitional

0.29*** (0.03)

0.26*** (0.03)

3.94*** (0.49)

3.74*** (0.50)

Escape Flexibility (IR) GDP GDPpc Trade

-0.03 (0.02) 0.19*** (0.04) -0.04*** (0.02)

Regime Democratization GATTWTO NoMembers Constant

-0.75* (0.46)

-0.04* (0.02) 0.13*** (0.04) -0.04** (0.02) 0.03*** (0.01) 0.21*** (0.07) -0.06 (0.06) -0.00 (0.00) -0.42 (0.48)

-0.12 (0.37) -0.98 (0.63) 0.65** (0.26)

8.38 (7.28)

-0.29 (0.38) -0.76 (0.67) 0.62** (0.27) -0.11 (0.08) 0.36 (1.13) 2.76*** (0.92) 0.07 (0.05) 9.04 (7.88)

(A5) Strings

(A6) Strings

0.44*** (0.04) 0.02 (0.02) 0.12*** (0.04) 0.08*** (0.02)

0.45*** (0.04) 0.02 (0.02) 0.12*** (0.04) 0.06*** (0.02) -0.01 (0.01) 0.01 (0.07) 0.36*** (0.05) -0.00 (0.00) -1.99*** (0.45)

-1.99*** (0.43)

Observations 571 567 497 494 571 567 Models A1, A2, A5, and A6 are ordered probit. Models A3 and A4 are tobit models. Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.

44

Table A-2: Models with minimum values of each variable across member countries. VARIABLES Depth

(A7) Escape

(A8) Escape

(A9) Transitional

(A10) Transitional

0.05*** (0.01)

0.05*** (0.01)

0.36*** (0.04)

0.33*** (0.04)

EscapeFlexibility GDPmin GDPpcmin Trademin

-0.07* (0.03) 0.21*** (0.06) -0.01 (0.03)

Regimemin Democratizationmin GATTWTO NoMembers Cut1 Cut2 Cut3 Cut4

-0.69 (0.62) -0.15 (0.62) 0.33 (0.62) 0.83 (0.62)

-0.06 (0.04) 0.11* (0.07) -0.02 (0.03) 0.04*** (0.01) -0.06 (0.12) 0.02 (0.11) -0.00 (0.01) -1.25 (0.83) -0.71 (0.83) -0.20 (0.83) 0.31 (0.83)

-0.23 (0.29) -0.92** (0.47) 0.62*** (0.22)

-0.63* (0.35) -1.00** (0.51) 0.62*** (0.24) -0.01 (0.08) -0.50 (1.02) 3.17*** (0.93) -0.04 (0.06)

(A12) Strings

0.79*** (0.05) 0.12*** (0.04) 0.07 (0.06) 0.13*** (0.03)

0.79*** (0.05) 0.11** (0.04) -0.00 (0.07) 0.10*** (0.03) 0.01 (0.01) -0.31** (0.13) 0.51*** (0.12) 0.00 (0.01) 4.50*** (0.90) 5.30*** (0.91) 6.20*** (0.91) 7.57*** (0.93) 8.65*** (0.95)

5.31*** (0.74) 6.07*** (0.75) 6.94*** (0.75) 8.26*** (0.77) 9.30*** (0.80)

Cut5 Constant

(A11) Strings

7.20 (5.21)

15.22** (6.81)

Observations 565 561 497 493 548 545 Models A7, A8, A11, and A12 are ordered probit. Models A9 and A10 are tobit models. Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.

45

Table A-3: Multinomial logistic regressions. EQUATION 1

(A13)

(A14)

0.01 (0.07)

-0.03 (0.07)

Depth EscapeFlexibility Constant

2

Depth

-8.75*** (3.07) 0.12** (0.06)

-8.87** (3.63) 0.08 (0.06)

EscapeFlexibility Constant 3

Depth

-5.94** (2.92) 0.29*** (0.05)

-3.85 (3.38) 0.26*** (0.05)

EscapeFlexibility Constant 4

Depth

-7.67** (3.08) 0.29*** (0.05)

-7.55** (3.55) 0.25*** (0.05)

EscapeFlexibility Constant 5

(A15)

(A16)

1.00*** (0.14) -6.35** (2.76)

1.05*** (0.15) -8.48** (3.36)

2.37*** (0.24) -8.51*** (3.14)

2.48*** (0.27) -12.47*** (3.84)

1.99*** (0.20) -13.99*** (3.16)

2.22*** (0.23) -15.35*** (3.72)

2.06*** (0.34) -21.91*** (5.09) 2.15** (0.92) -15.52 (10.58)

2.27*** (0.36) -22.05*** (5.79) 2.64** (1.03) -18.99 (2,712.58)

VARIABLES

-8.48*** (2.70)

EscapeFlexibility Constant

-8.02** (3.18)

Observations 565 561 548 545 Coefficients for control variables are not shown. Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1

46

Table A-4: Dropping EU and US treaties. VARIABLES Depth

(A17) Escape

(A18) Escape

(A19) Transitional

(A20) Transitional

0.07*** (0.01)

0.07*** (0.01)

0.40*** (0.05)

0.39*** (0.05)

EscapeFlexibility GDP GDPpc Trade

-0.06 (0.05) 0.27*** (0.08) 0.00 (0.03)

Regime Democratization GATTWTO NoMembers Cut1 Cut2 Cut3 Cut4

0.14 (0.94) 0.76 (0.94) 1.29 (0.94) 1.83* (0.94)

-0.07 (0.05) 0.21** (0.09) -0.00 (0.03) 0.04*** (0.01) 0.25* (0.15) -0.00 (0.12) -0.01* (0.01) -0.39 (1.04) 0.23 (1.04) 0.78 (1.04) 1.35 (1.04)

-0.64 (0.40) -0.90 (0.64) 0.85*** (0.29)

-0.77* (0.42) -0.72 (0.68) 0.89*** (0.30) -0.05 (0.08) 0.68 (1.19) 1.40 (1.00) 0.10* (0.06)

(A22) Strings

0.76*** (0.05) 0.03 (0.05) 0.15* (0.08) 0.15*** (0.03)

0.81*** (0.06) 0.04 (0.05) 0.14 (0.09) 0.12*** (0.03) -0.02* (0.01) 0.04 (0.17) 0.79*** (0.12) -0.02 (0.01) 4.66*** (1.12) 5.33*** (1.12) 6.33*** (1.13) 7.76*** (1.15) 8.72*** (1.16)

4.26*** (1.00) 4.88*** (1.01) 5.79*** (1.02) 7.11*** (1.03) 8.02*** (1.05)

Cut5 Constant

(A21) Strings

7.64*** (0.43)

7.60*** (0.43)

Observations 477 473 412 409 462 459 Models A17, A18, A21, and A22 are ordered probit. Models A19 and A20 are tobit models. Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.

47

Table A-5: Probit models. (A23) EscapeFlexibity Dummy

VARIABLES

(A24) EscapeFlexibity Dummy

EscapeFlexibityDummy Depth GDP GDPpc Trade

0.10*** (0.01) -0.04 (0.06) 0.28*** (0.10) -0.01 (0.04)

Regime Democratization GATTWTO NoMembers Constant

Observations

-1.91* (1.05)

0.10*** (0.01) -0.04 (0.06) 0.27** (0.11) -0.02 (0.04) 0.02 (0.01) 0.37** (0.19) 0.12 (0.14) -0.02*** (0.01) -2.10* (1.20)

(A25) FlexibilityStrings Dummy

(A26) FlexibilityStrings Dummy

2.28*** (0.17)

2.29*** (0.18)

0.09 (0.06) 0.07 (0.11) 0.06 (0.04)

0.09 (0.07) 0.03 (0.12) 0.03 (0.03) 0.02 (0.01) 0.16 (0.22) 0.60*** (0.15) -0.02 (0.01) -4.61*** (1.34)

-4.62*** (1.18)

571 567 571 Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.

48

567

Table A-6: Biprobit models. VARIABLES Depth

(A27) Escape Dummy

(A28) Escape Dummy

0.14*** (0.01)

0.14*** (0.01)

EscapeFlexibityDummy GDP

-0.04 (0.05) 0.03 (0.10) -0.10*** (0.04)

GDPpc Trade Regime Democratization GATTWTO NoMembers Constant EUDepth GATTWTODepth

0.41 (1.02) 2.07*** (0.52) 1.49*** (0.42)

-0.04 (0.05) 0.07 (0.11) -0.09** (0.04) 0.01 (0.01) 0.33* (0.17) -0.14 (0.13) -0.02*** (0.01) -0.19 (1.14) 2.35*** (0.53) 1.19*** (0.43)

Transparency Transparency2 ρ

Observations

-0.71*** (0.12)

-0.67*** (0.12)

(A29) Strings Dummy

(A30) Strings Dummy

3.07*** (0.16) -0.05 (0.05) 0.67*** (0.10) 0.04 (0.04)

-3.19** (1.29)

3.21*** (0.15) -0.09 (0.05) 0.62*** (0.10) 0.03 (0.04) 0.02* (0.01) 0.24 (0.18) 0.39*** (0.13) -0.01 (0.01) -2.20 (1.34)

-7.17*** (2.59) 6.08*** (1.80) -1.02*** (0.32)

-6.34** (2.69) 5.01*** (1.87) -4.76 (37.32)

571 567 523 Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1.

49

519

−1.0

−0.5

0.0

0.5

1.00 0.90

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

1.0

0.5

1.0

1.5

−1.0

−0.5

0.5

1.0

3.5

● ●

2.5

●

●

●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ●

0.5

●●

●

● ● ● ● ● ● ● ● ● ●● ●● ●● ● ● ● ● ● ● ● ● ● ●●● ●● ● ● ● ● ● ●● ●● ● ●● ● ● ● ●●● ● ● ● ● ● ●● ●●● ● ● ● ● ● ●● ●● ● ● ● ●● ●● ●● ● ● ● ● ●●●● ● ● ● ● ●● ●●● ●● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●●●● ● ● ● ● ● ●●●● ● ● ● ● ● ● ●●● ●●● ●●● ● ● ●● ●● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ●● ●● ● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ●● ● ●● ● ●●● ● ●●● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ●● ● ●● ●● ●● ● ●●● ● ● ●● ● ● ● ●● ● ● ● ● ● ● ●● ● ● ●● ● ● ●●● ● ● ● ● ● ●● ● ● ●● ● ● ● ●● ● ● ● ●● ● ● ● ●● ● ● ● ●● ●● ● ●● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●●●●● ● ● ● ● ● ● ● ● ● ●

●

0.2

0.4

0.6

0.8

1.0

0.85 0.95 1.05 1.15

●

1.5

2.0

EU Depth

Residuals (Flexibility strings)

0.85 0.95 1.05 1.15

Residuals (Flexibility strings)

●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ●

1.0

WTO Depth

●

2.0

●

1.5

● ● ● ● ●● ● ●● ● ● ● ● ● ● ● ●● ● ● ●● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ●

0.0

● ● ●

●

0.5

● ● ● ● ● ● ● ● ● ● ● ● ●

Residuals (Transitional flexibility)

3.5 2.5 1.5 0.5

Residuals (Transitional flexibility)

●

● ● ●

●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

EU Depth

● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ●

●

WTO Depth

●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

●

0.80

● ● ● ● ● ● ●● ● ● ● ● ●● ● ● ● ● ● ●● ● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ●● ● ● ●● ● ● ● ● ● ● ● ●● ● ●● ● ● ●● ● ● ● ● ●● ● ●

Residuals (Escape flexibility)

1.00 0.90

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●

0.80

Residuals (Escape flexibility)

Checks on the instruments

●

●

●

0.0

Transparency

● ●● ● ● ● ● ● ● ● ● ● ● ●● ● ●● ●●●● ● ● ● ● ● ● ● ● ● ● ●●●●●● ● ● ● ●● ● ● ●●●● ●●●●● ● ● ●● ● ● ● ● ●●● ● ● ●●● ● ●● ● ● ● ● ● ●● ●● ●● ● ● ● ● ●●●● ● ● ● ● ● ●● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ●● ● ●● ● ● ● ● ●● ● ● ● ● ●● ●●● ● ● ● ● ●● ● ● ●● ● ● ● ● ● ● ● ● ●●● ● ● ● ●●●● ● ● ●● ● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ●● ● ●● ●● ● ●● ● ● ● ● ● ● ●●● ● ● ● ● ●● ●● ●● ●● ●●●● ●● ● ● ● ● ●● ●● ● ●●● ● ●● ●●●●●●●● ● ●●● ● ● ●● ● ●● ● ● ● ●● ●● ● ● ● ●● ● ● ●●● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ●●●● ● ● ● ● ● ●●● ●● ● ● ● ● ● ● ● ● ●● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ●●●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ●● ●● ●● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ●●●●●● ● ● ● ● ● ● ● ● ● ● ●

0.2

0.4

0.6

0.8

1.0

Transparency2

Figure A-5: Correlations between instruments and residuals

50