Seal the Deal: Bargaining Positions, Institutional Design, and the Duration of Preferential Trade Negotiations

ABSTRACT

Multilateral negotiations at the World Trade Organization have stalled. This has contributed to a steep rise in preferential trade agreements (PTAs). At the same time, negotiations for PTAs have not always proven quick and painless: While some treaties are sealed within a few months or days only, other agreements are preceded by protracted bargaining processes in trade and trade-related issue areas. In this article, we provide a theoretical explanation for this empirical variation. More specifically, we argue that PTA negotiations take longer the greater the distance between the prospective partners’ initial bargaining positions. Moreover, we contend that negotiation processes become more protracted the higher the relative ambition of the prospective PTA. Due to the limited links to the domestic political arena in autocracies, we expect this latter effect to play out for groups of democratic bargaining partners only. We test these two hypotheses for 198 preferential trade negotiations using novel measures for bargaining templates and the ambition of PTA clauses. In our two-stage survival models, we find support for our argument. In line with qualitative evidence from recent preferential trade initiatives, our models indicate that services, investment and intellectual property rights are particularly sticky agenda items for democratic leaders at the international bargaining table.

KEYWORDS:

• Institutional design
• international negotiations
• international organizations
• preferential trade agreements

Multilateral negotiations at the World Trade Organization (WTO) have stalled. This has contributed to a steep rise in preferential trade agreements (PTAs), with countries seeking to expedite the process of reciprocal trade liberalization in a more limited setting (Hoekman 2015; Mansfield and Reinhardt 2003).

Interestingly, while PTAs on average are associated with shorter negotiation spells than WTO rounds,¹ there is substantial variation in the duration of preferential trade negotiations: Whereas in 1998 India and Sri Lanka inked their PTA after 26 days of negotiations only, South Korea and Australia negotiated over 14 years until they signed their joint agreement in 2014. A more recent and prominent example includes the Comprehensive Economic and Trade Agreement (CETA) between the European Union and Canada, which had been preceded by a protracted 5-year bargaining process in the public spotlight. Understanding the duration of trade negotiations matters in the real world. Some leaders bank their future on being able to negotiate PTAs quickly. Moreover, businesses base their investment and export decisions on market access. Negotiation delay in international trade additionally inflicts costs upon voters-as-consumers in terms of foregone economic benefits, and risks to undermine public and private trust in the ability to arrive at an agreement through a given international institution (Fearon 1998, 277).

What then helps to explain the observed variation in the duration of preferential trade negotiations?

Our argument, which comes in two parts, stipulates that the design of historic and actual PTAs explains variation in the duration of PTA negotiations.² The first part of the argument builds on the assumption that negotiations do not occur in a “vacuum” (Odell and Tingley 2013, 154), but rather evolve around initial bargaining positions based on which countries exchange offers and counter-offers. Thus, we expect that the distance between these bargaining positions can be regarded as a determinant of how long it takes countries to converge on a final agreement text. Second, this convergence process can additionally be driven by the relative ambition of the agenda items under discussion, or what we term the relative scope of a planned treaty: Certain PTAs depart little if anything from previously contracted obligations and are relatively “easy” to commit to. Other agreements, in turn, lead countries to make more ambitious commitments or venture into new issue areas, potentially triggering adverse reactions in member states’ domestic political arenas and resulting in bargaining delay. Due to the stronger link to the domestic political arena, we expect this effect to play out for groups of democratic PTA partners only.

To test these arguments, we use originally collected data on 198 PTA negotiations and novel variables for the design templates with which countries enter into bargaining processes. From the latter, we additionally obtain issue-area specific indicators for how ambitious a new trade agreement is for its members. In our two-stage selection models, we find that the initial bargaining positions of countries matter for explaining both their likelihood to enter negotiations in the first place and the time they take to seal a trade deal. Moreover, our results robustly show that members are reluctant to accept ambitious commitments in existing or new issue areas, but that this effect only holds for democratic PTA partners.

After unpacking our variable for the relative scope of commitments, we further find that investment, intellectual property rights (IPR), and services constitute particularly thorny agenda items for democratic leaders in trade negotiations.

With our study we contribute to the literature on international negotiations and cooperation (Christiansen and Tangen 2002; Fearon 1994, 1998; Mansfield et al. 2002; North 1992). More specifically, we show that both the shadow of the past (previous commitments) and the shadow of the future (relative ambition of a planned PTA) matter in explaining bargaining dynamics. Rigorous empirical tests suggest that democratic countries bargain harder if a prospective treaty departs substantially from previously signed agreements. Furthermore, our novel measures for the relative scope of commitments allow us to pin down sticky agenda items in trade negotiations at a high level of detail. The findings presented in this context can contribute to the general debate about the challenges of and potential complementary measures to liberalization in trade and trade-related issue areas.

In the following sections, we will first offer a brief review of the existing literature on the duration of international negotiations, followed by the discussion of our theoretical argument. Subsequently, we will present our empirical research design and the main findings from the empirical analysis. In the concluding section, we will identify avenues for further academic research and elaborate on the policy-making implications of our findings.

The duration of preferential trade negotiations

Negotiations can be defined as “a process in which explicit proposals are put forward ostensibly for the purpose of reaching agreement on an exchange or the realization of common interest where conflicting interests are present” (Iklé 1976, 3–4). What then explains the duration of negotiation processes? In very general terms, incentives or costs to agree describe bargaining dynamics and hence the pace of negotiations (Jeong 2016).

One potential prolonging factor are high bargaining costs. Bargaining costs capture the degree of difficulty countries encounter in the convergence process toward an agreement (Fearon 1998; North 1992). This concept entails the idea that parties enter negotiations with fixed preferences over an optimal institutional design. During the bargaining process, countries try to find a compromise between their preferences. Depending on the level of compatibility between these preferences, negotiations might take longer or shorter.

Countries might also bargain harder, and thus longer, if enforcement costs are high. Such costs appear if compliance with an agreement will be well monitored and enforced (Fearon 1998). Similarly, negotiations could last longer if countries expect punishment by their specific audience in case they back-down from previous commitments (audience costs; Fearon 1994).

With regard to positive incentives to conclude an agreement, international treaties might become more attractive to policy makers if they serve credible commitment purposes (Mans-field and Milner 2015; Mansfield et al. 2002). This means that by signing treaties actors commit to certain policies in the long-term and reduce uncertainties for domestic stakeholders. In this context, political leaders are prone to receive rewards (votes or payments) by committing to long-term goals. Then, the probability of cooperation and the willingness to make concessions increase, with an expediting effect on the bargaining process.

Against the background of these theories, we argue that one central aspect has yet been ignored in the literature: countries’ track record of earlier treaty commitments. The regime of international trade is a complex one, with countries committing to the same treaty type several times over time. A first potential upshot from this fact is that initial bargaining positions might be endogenous to the institutional design of previously signed agreements. Put differently, previous commitments or the shadow of the past are likely to affect initial bargaining positions, the ease of convergence between these positions, and consequently negotiation duration.

Second, countries’ previously signed treaties might help inform their willingness to commit to a certain level of ambition in an envisaged agreement. The degree to which a future treaty departs from previously accepted clauses is likely to go hand in hand with the level of familiarity of citizens with the new provisions in a treaty and consequently their support for or resistance to the planned agreement. This, in turn, will influence leaders’ willingness to seal a deal quickly, and consequently, influence negotiation duration. Below, we will explain these arguments in greater detail.

Endogeneous bargaining costs

The trade regime is complex with countries signing multiple PTAs with a variety of treaty partners. Countries enter all of these negotiations with preferences over their optimal institutional design. For instance, in the PTA negotiations between Singapore and Australia, the former country insisted on a positive list approach to services trade liberalization, while the latter argued for a negative list approach. After protracted negotiations, Singapore eventually gave in to Australian demands.

We expect that bargaining positions are sticky. This means that preferences of one country in year t should be correlated with preferences in year t–1 (Castle and Pelc 2017). In short, current preferences are endogeneous to past preferences. Indeed, there is empirical evidence that preferences over provisions are heavily influenced by the agreements already concluded by prospective PTA partners. By means of illustration, the United States has strategically leveraged the provisions in its existing PTAs to set a precedent for the terms in new treaties (Feinberg 2003, 1028). These sticky preferences can be understood as institutional design templates. Crump (2007) calls the accepted formulas ˝tailored to individual issue areas˝ that negotiators tend to work with templates.

Since preferential trade negotiations involve conflicting preferences over the institutional design to select, what is the relationship between differences in the proposed templates and negotiation duration?

In line with Fearon (1998), we expect that substantive differences in the initial positions on the “optimal” institutional design can be viewed as implying high bargaining costs in the bargaining process. As just discussed, in trade negotiations, countries bargain based on specific templates. In this context, one could conjecture that the more different countries’ templates are, the longer it will take them to converge on a common denominator based on an exchange of offers and counter-offers.

In situations of high bargaining costs, there are two scenarios: Either countries refrain from creating a potential institution or they will engage in protracted discussions about which models to use for trade negotiations and how to split the gains and costs emanating from trade liberalization (Keohane 1988, 387). If states opt for the latter option, more time has to be invested to reach an agreement. As a result, the negotiation process is prolonged.

The example of the trade negotiations between the European Union and South Africa further elucidates this mechanism: In 1994, the two parties entered their trade negotiations with a large discrepancy in terms of institutional design propositions. Whereas the European Union insisted on an extension of market access in a strict reciprocal way, South Africa emphasized the economic development dimension of trade and pushed for a more non-reciprocal form of liberalization as enshrined in its prior agreements (Bilal and Laporte 2004). It took the two trading blocs over four years to converge on a final legally binding text in 1999 (Baccini and Urpelainen 2014, 201). In light of this qualitative evidence, we formulate our first hypothesis:

Hypothesis 1: The more different the design templates of the prospective PTA partners, the longer the negotiation process for trade liberalization.

Commitment costs

Discrepancies in institutional design templates between countries define endogenous bargaining costs. However, not only can we observe variation in institutional design templates across countries, but also across time for individual countries. In short, the very same country might commit to different treaty provisions over time. In this section, we argue that a substantial deviation in design from previous agreements leads to protracted PTA negotiations among countries where leaders receive direct feedback from domestic stakeholders, that is democratic regimes.

Whereas some PTAs imply little if any modification in institutional design, other treaties contain ambitious provisions which depart significantly from previously signed agreements. Departure from previous provisions can take two forms: (i) an increase in the depth of commitments in a previously liberalized issue area or (ii) an extension of commitments to new issue areas. We subsume both of these adjustments under the concept of the relative scope of a PTA. The relative scope of a PTA therefore captures the upward adjustments it requires member states to make relative to their established templates through more ambitious commitments in existing issue areas or new commitments in new issue areas. What are the consequences of large commitment costs, that is high relative scope, for bargaining dynamics? How do policy makers react to unfamiliar provisions in trade agreements? Are they bargaining harder and thus, longer in the context of a high relative scope?

We argue that the answers to these questions depend on the degree of accountability leaders have. Our argument is that relative scope is costly only in situations of high accountability. Decision makers that lack such accountability are expected to care less about policy change. In the following paragraphs, we will explain how the relative scope of a PTA interacts with domestic political institutions to affect the duration of trade negotiations.

Increasing the relative scope of a PTA can trigger resistance in the domestic political arena for the following reason: During negotiations over an ambitious trade agreement, citizens may face difficulties in anticipating the direction and magnitude of the economic and political ramifications of the planned PTA. This uncertainty can relate to the effects of trade liberalization on market shares and employment, as well as legal uncertainty over the degree to which the government ties its hands by signing a PTA (Zahrnt 2007, 376–377).

Previously concluded agreements are indicative of the domestic reforms a country has already undertaken. The more ambitious a new PTA, the higher the odds that citizens have to cope with uncertainty over the precise consequences of the agreement. Committing to an agreement with ambitious provisions might thus risk to generate backlash in the domestic political arena of the member states. Domestic backlash, in turn, would only matter to leaders in democratic countries. In democratic regimes, voters have the possibility to directly punish or reward leaders for agreements made at the international level. In the absence of this feedback loop (for example in less democratic countries), commitment costs should be lower.

At this point, one might object that even if ambitious trade agreements trigger an adverse reaction among certain groups of voters, the median voter in democratic regimes would still unequivocally favor free trade. As a result, democratic leaders could also leverage these agreements as credible commitment devices. If anything, they would benefit even more from inking an ambitious PTA because of the sizable potential welfare gains stemming from such an agreement. Indeed, data from the Pew Research Centre on Global Attitudes & Trends shows that the median voter is generally in favor of free trade and liberal trade policies adopted by the government.

However, recent studies by the Pew Research Centre introduce nuances into this account: While citizens worldwide still view free trade and low tariffs favorably, some new issues on the liberalization agenda are met with skepticism. For instance, when asked about the project of a Transatlantic Trade and Investment Partnership between the United States and the European Union, citizens in Germany and the United States exhibit strong reservations concerning investment and standards. Fifty-six percent of citizens in the United States believe that granting European companies the unconditional right to buy domestic companies would mostly hurt their economy, while only 31% are of the view that their economy would benefit. The contrast is even more pronounced in Germany, where 73% are opposed to American companies on German soil, and only 19% are in favor of foreign takeovers. On standards, citizens from both sides of the Atlantic think that the standards adopted by their government ensure a higher quality than the foreign versions: This pattern holds across the board, be it with respect to auto, food, or environmental safety standards (Pew Research Centre 2014).

What is the implication of this evidence for the process of PTA negotiations? We argue that since trade policy preferences are multidimensional, democratic leaders cannot neglect citizens’ concerns when negotiating trade agreements with an ambitious scope. Sealing such a PTA quickly would not produce the expected electoral benefits through a credible commitment to free trade, but opposition from voters suffering from uncertainty about the PTA provisions. How, then, can a democratic government respond to the demands in the domestic political arena? We argue that it will not readily commit to an overhaul of the country’s PTA template in a dramatic manner, but rather seek one of the following two options:

i) Flexibility: The government is willing to subscribe to far-reaching PTA provisions, provided that there is flexibility in issue areas with wide-ranging provisions. Flexibility instruments offer a certain relief to import-competing industries during the process of trade liberalization. In addition, transitional flexibility provisions provide scope for citizens to learn about the effects of a PTA in different states of the world (Koremenos 2001). Importantly, for our analysis and in line with previous research, we assume that the details of flexibility provisions are negotiated among PTA partners after the scope of provisions from a trade deal have already been agreed (Baccini et al. 2015, 3). We further conjecture that negotiations over flexibility are time-consuming. In this regard, the negotiations between the United States, Canada, and Mexico over the North American Free Trade Agreement (NAFTA) in the early 1990s are illustrative: Even though the three countries were generally committed to a high level of liberalization, stand-offs on certain agenda items determined the fate of the bargaining process even in the final hours. Canada insisted on a provision that grants it flexibility in screening foreign investment above a certain threshold. Experts have noted that this demand expressed the country’s wariness to open up to foreign investment without any room for manoeuvre in regulating the expected increase in capital inflows. Investment constituted a thorny issue in the negotiations between Canada and the United States until the final days of the free talks (Cameron and Tomlin 2002, 112–113).

ii) Policy space carve-outs³: An alternative scenario is that for certain detailed provisions of the agreement democratic leaders are not willing to make commitments due to policy space and distributional considerations. Even if the governments seek a comprehensive agreement in terms of the number and depth of issue areas covered, they may be wary of giving up all the policy space in those issue areas. The safeguarding of policy space, in turn, would respond to a larger demand from society to provide certain public goods (for example high public health standards or a clean environment). For instance, in late 2015, 12 Pacific-Rim countries concluded the Trans-Pacific Partnership Agreement (TPP).⁴ In the TPP negotiations, provisions on the protection of IPR constituted a sticking point even in the final hours. Even though there was a strong commitment among all countries to move beyond existing WTO obligations, a majority persistently refused to grant the United States extensive periods of exclusivity for biological test data. Long periods of exclusivity preclude producers of generic medicines from obtaining regulatory approval without going through clinical trials themselves and curb governments’ policy autonomy (Shadlen 2005, 767). The United States faced a particularly staunch opposition from Australia and New Zealand (Chan 2015). The position of these two democracies was guided by the overarching goal to maintain policy space within the area of public health. The topic of test data exclusivity had already caused a delay in the bargaining process for the US–Australia trade agreement (Crump 2007, 132). In a similar vein, during the TPP negotiations Australia’s Minister for Trade and Investment was quoted affirming:

As I have made clear repeatedly, the government will not support outcomes that would increase the prices of medicines for Australians or adversely affect our health system more generally; end of story. (…) Nor would we accept outcomes that undermine our ability to regulate or legislate in the public interest in areas such as health. (Australian Department of Foreign Affairs and Trade 2015)

The examples above may hint at more general dynamics in international trade negotiations. If the relative scope of a new PTA is substantial, there will be a domestic demand for in-built flexibilities and a safeguarding of policy space in sensitive issue areas. This demand will travel to the international bargaining table, where it prolongs negotiations. From this reasoning we derive our second hypothesis:

Hypothesis 2: The larger the relative scope of a prospective PTA, the longer the negotiation process among democratic PTA partners.

Should we expect similar dynamics to play out irrespective of whether countries negotiate over a national treatment clause for services trade or a sub-state level commitment for the liberalization of public procurement markets? Recent evidence on mega-regional trade initiatives suggests a nuanced account: For instance, leaked bargaining documents of the European Union’s trade initiatives with the United States and Canada reveal investment as a particularly sticky agenda item between the trading powers (Castle and Pelc 2017). In a similar vein, public discontent on both sides of the Atlantic has given rise to protracted bargaining over the mutual recognition and harmonization of standards. Provisions on trademarks, copyrights and patents are also highly sensitive for both sides at the transatlantic bargaining table. In contrast, competition rules and public procurement provisions have been largely absent from the public debate about PTA negotiations (New 2016).

The empirical analysis presented in the next section will allow us to examine whether these patterns can be generalized to a higher number of preferential trade initiatives: For this purpose, we will present a fine-grained measure for the relative scope of commitments in a planned PTA to identify which issue areas consistently lead to protracted trade negotiations.

Figure 1 summarizes our argument graphically, where both the distance in initial bargaining positions and the relative scope of a planned PTA (as institutional design deviation from past PTAs) explain negotiation duration. We expect particularly long negotiations in situations with high bargaining and commitment costs (top right of the panel) and especially short trade talks in situations with low bargaining and commitment costs (bottom left of the panel).

Figure 1. Negotiation duration as function of bargaining and commitment costs.

Empirical analysis

The aim of the statistical test is to assess whether differences in design templates and the relative scope of a planned PTA protract preferential trade negotiations. Since the duration of PTA negotiations is characterized as time data, we estimate survival models. In our context, survival refers to not concluding trade negotiations. When estimating survival models, one can choose between a range of parametric and semi-parametric model specifications. Formal tests show that the pH-assumption, which must be met for the workhorse semi-parametric Cox model, is violated for a number of covariates. From the range of available parametric models the Weibull model performs best.⁵ To facilitate the output interpretation, we follow Mölders (2015) in estimating a Weibull model in accelerated failure time (AFT) form: Positive signs imply a prolonging, negative signs an accelerating effect on PTA negotiations.

One concern in our estimation is about a selection effect underpinning PTA negotiations. There are strong theoretical reasons to believe that the start of bargaining processes is non-random. To tackle the resulting risk of selection bias, we estimate a duration model with selectivity (Boehmke et al. 2006). In the first stage we endogenize the decision to start negotiations. We do so by creating dyad-year observations for which we do not observe PTA negotiations or existing agreement ties. Subsequently, we compare these observations without treatment to our sample of countries involved in trade negotiations. In the second stage, we analyze the duration of bargaining processes, taking into account the selectivity of our observations. Below, we explain variables included in the first and second stage.

Data and operationalization

For the empirical analysis of our two hypotheses we rely on an original dataset covering 198 PTA negotiations in the post-1990 period.

The dependent variable Negotiation duration is operationalized as the time which elapses between the start of trade negotiations and the official signature date of a PTA. More precisely, we measure the number of days between the two events. To retrieve this information, we followed a two-step approach: First, we applied web scraping to collect data on PTA negotiation duration from government and regional organization websites. In a second step, we reverted to the softwares Factiva and LexisNexis to systematically screen newspaper articles and other media contributions (for example transcripts from news channels) for information on preferential trade negotiations after 1990. We decided to limit our empirical scope to PTAs concluded in the post-1990 period because digitized information on trade agreements pre-1990 is scarce. In Factiva and LexisNexis, we used codes to search for news articles in four languages: English, Spanish, German, and French. The codes are based on keywords and Boolean operators. For instance, to obtain information of the signature dates of a bilateral PTA involving English-speaking countries, the following search code was used in Factiva:

“(PTA member 1 AND PTA member 2 AND sign AND trade agreement) OR (PTA member 1 AND PTA member 2 ink AND trade agreement) OR (PTA member 1 AND PTA member 2 AND complete AND trade agreement)”

If we found media contributions indicating different dates for the same event, we opted for the earliest announcement. Our approach for identifying negotiation start and signature dates is aligned with the empirical strategy by Mölders (Mölders 2015, 9). We also crosschecked our results with signature dates listed in treaty texts. In case of a mismatch, we gave preference to official documents over media reports. In addition, we excluded internal European Union agreements (for example the Amsterdam Treaty of 1996) due to the peculiarities of intra-EU decision-making processes (Weinberg 2016).

In total we were able to identify 198 data points on negotiation duration.⁶ Figure 2 shows substantive variation in the dependent variable: For example, whereas Australia and Korea negotiated nearly 14 years (5072 days) until they signed their joint agreement, India and Sri Lanka concluded their bargaining process within 26 days only. The median duration of PTA negotiations is just above 645 days.

In light of the dependent variable’s nature and to avoid data multiplication, the unit of analysis is at the PTA-level. Consequently, measures on the right-hand side of the regression model will be aggregated to the undirected dyadic (or for plurilateral agreements to an even higher) level. Except indicated otherwise, we follow Baccini et al. (2015) by taking the mean across PTA members.

Turning to our first main independent variable, we measure Difference in design templates as the difference in the proposed templates of the prospective PTA member states. To measure each country’s design template, we proceed in the following manner: We rely on two datasets on PTA design by Dür et al. (2014, Design of Trade Agreements database) and Lechner (2016). The datasets contain information on PTA design in seven issue areas: services, investment, IPR, public procurement, standards, competition policy, and non-trade issues (NTIs: economic, social, civil and political rights, environmental protection, and security issues). The two datasets provide us with 152 variables⁷ for each trade agreement covered. Each of the 152 variables is coded in a binary (0 or 1) fashion, with a coding of 1 implying a deeper commitment. We assume that if a country has included a liberalization provision in at least one previous PTA, the cost of doing so again are lower than if the provision has never been included before. Moreover, prospective PTA partners will take note of the commitments already made by the reference country and use them as a benchmark for evaluating concessions in the bargaining phase (Crump 2007). Consequently, the template of country i at point in time t consists of the maximum commitments it has made until t. If a country is not yet a member of any PTA, we construct a fictional design template, where we set each design aspect to 0 and subsequently calculate the relative scope. The underlying assumption here is that a country has to adjust in every design aspect for its very first PTA.⁸ In case a new PTA is more shallow than a country’s template on any specific provision, we assume zero scope widening. To derive our measure for the difference in initial bargaining positions, we compute the distance in design templates between prospective PTA partners at the start of negotiations. For this purpose, we calculate the Cohen kappa index of distance (Cohen 1968) between the parties’ design templates. Cohen kappa takes into account the possibility of (dis)agreement by chance.⁹ This is especially relevant for our context since a data- point of zero might mean that countries have been against a particular issue or that they have simply not yet considered regulating the issue, but might want to do so in future. The Cohen kappa measure of difference provides a cautious approach for such a situation, where a certain fraction of the data points overlap by chance. The resulting variable ranges from 0 to 1, with 0 indicating completely congruent templates and 1 indicating completely dissimilar templates. We include this variable in the first and second stage. Not only could the difference in design templates prolongate negotiations, but also decrease the probability of a negotiation start. One could hypothesize that two trading partners with similar templates are more likely to find a common denominator in negotiations and hence exhibit higher odds of starting PTA negotiations in the first place (Keohane 1988, 387). Figure 3 shows the distribution of this variable.

Figure 2. Negotiation duration in days.

Figure 3. Difference in design templates.

Our second regressor Relative scope reflects the degree to which a new PTA requires countries to make upward adjustments to their existing laws and regulations, with ramifications for the domestic society at large. Critical for our analysis is the assumption that a country’s existing PTA track record proxies for its laws and regulations in force. To create this variable, we rely on the country-year-specific PTA design templates calculated for the first independent variable. We then subtract these design templates from the design of any new PTA concluded by a given country. The subtraction is performed on every single variable: For instance, if a new trade agreement features a liberalization clause in the education services sector, we compare this to the maximum education services commitments in past PTAs concluded by the prospective agreement partners. If the education services clause is novel, the difference would be 1 (1–0 = 1). If it matches earlier commitments, a value of 0 is obtained (1–1 = 0). To define a cutoff point for comparison, we take the negotiation start date of the new PTA as a reference. Subsequently, we create our variable for the relative scope in four steps: First, we sum the differences across the 152 variables for each PTA partner. Second, we divide these sums by the number of variables covered in the respective dimension, to avoid an overemphasis on issue areas with more extensive coding. Third, we calculate the average across these seven standardized values. Fourth, we average these means across the PTA partners. Accordingly, a value of 0.5 for the variable Relative scope means that a planned trade agreement on average requires countries to make upward adjustments on 50% of all design aspects of their existing PTA templates.

Figure 4 plots the distribution of the relative scope variable. Three PTAs stand out as particularly ambitious scope-wise: NAFTA (concluded in 1992), Australia-Korea (2014), as well as Mexico-Nicaragua (1997). NAFTA is often portrayed as the first PTA with truly novel and deep provisions. More recent PTAs also imply a relatively wide scope for individual countries: The Australia-Korea treaty features ambitious provisions implying upward commitments on as much as 80% of the template clauses in the issue areas of investment, procurement, and standards.

Figure 4. Relative scope of PTAs.

We expect that the protracting effect of relative scope on negotiation duration is only present for democratic regimes. Hence, we interact a democracy-dummy with the relative scope variable. The democracy variable is “one” if member states are above the threshold of 6 of the Polity 2 score developed by Marshall et al. (2016). This threshold is in line with previous research (Poast and Urpelainen 2013). We chose the Polity 2 score as basis for our democracy measure, because it reflects the competitiveness of political participation, constraints on chief executives, as well as the openness and competitiveness of executive recruitment in a country of interest. It therefore allows us to proxy for the degree of electoral challenges in a given country. Whereas a low score means that the electorate has no possibility to punish or reward the government for its trade policy, a high score indicates strong feedback from the public. Given its high validity, this measure has been used repeatedly in empirical contributions on the relationship between regime type and trade policy (Manger and Pickup 2014; Mansfield et al. 2002; Milner and Kubota 2005).

Figure 5 shows the relationship between negotiation duration and relative scope for groups of less democratic and democratic PTA partners, respectively. The scatter plots indicate that while relative scope and negotiation duration correlate positively for democracies, a greater scope does not go along with longer negotiations for non-democratic partners.

Figure 5. Relative scope and negotiation duration for democracies and autocracies.

Beyond our main explanatory variables, we integrate a set of control variables in our regressions in order to account for potentially confounding factors.

First, in addition to the PTA track record of the bargaining parties, we include three design features of the jointly negotiated agreement: Enforcement, flexibility, and the number of member states. As regards the first design dimension, we rely on the variable Enforcement created by Allee and Elsig (2015) as a 0–9 ordinal index for the strength of PTA dispute settlement provisions. If a planned agreement exhibits a high degree of enforcement, countries might be more circumspect in negotiations given the sovereignty costs at stake. Second, we include two indices for PTA flexibility provisions by Baccini et al. (2015): The first variable reflects whether countries are allowed to resort to Escape flexibility tools (for example anti-dumping duties) under a PTA, while the second measures the strings attached to the use of these tools (Rigidity on flexibility). These covariates allow us to bring new empirical evidence into the debate on the impact of different flexibility and rigidity provisions on negotiation duration (Bearce et al. 2015; Rosendorff and Milner 2001). Finally, we control for the number of Member states in a planned PTA. Both Moser and Rose (2012) and Mölders (2015) find that it becomes more difficult to conclude trade negotiations once the number of national delegations sitting at the negotiation table increases.

We also add a variable for power asymmetries among negotiation partners. Power asymmetries are proxied by a dummy labeled Power asymmetry, which indicates the presence of an agreement between a major economic market – namely Australia, Canada, EFTA, the EU, Japan, New Zealand, and the United States – and any weaker country or countries. One could reasonably conjecture that weak states are eager to quickly secure preferential access to the vast markets of more economically powerful partners (Manger and Shadlen 2014). This mechanism might also be at stake with regard to the first stage. Thus, we include Power-asymmetry in the first and second stage of our model. Next, we account for domestic institutional constraints in the form of Veto players (Henisz 2000): Veto players have a final say regarding the ratification of a trade agreement, yet may also impact the bargaining process of forward-looking governments (Mölders 2015). Similarly, veto players might decrease the probability of negotiation start in the first place. As a further political variable, we include a measure for the quality of the Rule of law among PTA partners (Kaufmann et al. 2016): Baccini (2014) finds that dyads in which at least one country is characterized by poor domestic institutions experience negotiation delay due to persistent uncertainty over the implementation of the agreement. Again, these variable could play a role in the first stage too.

The variable Log Trade flows between countries proxies for the economic importance of prospective treaty partners. We take the measure from the Correlates of War project. High trade flows are expected to increase the likelihood that countries are entering negotiations and concluding treaties fast.

Regarding negotiation capacity, we include a dummy for GATT/WTO membership of all PTA partner countries (WTO membership): Membership in the multilateral trade club can be seen as a proxy for familiarity with the rules of the world trading system and could hence increase the likelihood of negotiations and speed up trade talks. The same logic applies to the variable PTA activity, which reflects the average number of prior agreements concluded by the new partners. The dummy First PTA, in turn, scores 1 if within the bargaining group at least one country within the negotiation group signs its very first trade agreement and could thereby slow down the bargaining process due to lack of experience.

Another driver for entering negotiations is Diffusion: Our model covers a variable measuring the number of treaties signed by countries located in the same region. To round up our model, we include controls for Geographic distance, Common language, Common border, Colonial past, and temporal fixed effects based on the negotiation start year in the first and second stage model.

Table 1 displays the descriptive statistics for the variables used in the empirical analysis.

Table 1. Descriptive statistics.

Statistic	N	Mean	St. dev.	Min	Max
First stage:
Polity2 Score	307,347	11.97	5.69	0.00	20.00
Difference in design templates	276,00	0.75	0.20	0.02	1.00
Power-asymmetry	348,523	0.42	0.20	0.00	1.00
Veto players	294,436	0.24	0.65	0.00	0.69
Rule of law	324,071	−0.17	0.65	−1.67	2.02
Log trade flows	295,858	1.09	3.76	−29.20	11.35
WTO membership	303,714	0.73	0.45	0.00	1.00
PTA activity	336,394	6.18	3.87	0.00	36.60
Geographic distance	313,087	12,039.43	7,556.75	15.00	25,13
Common language	322,570	0.14	0.34	0.00	1.00
Contiguity	322,570	0.01	0.06	0.00	1.00
Colonial Past	322,570	0.00	0.04	0.00	1.00
Diffusion	348,523	26.14	16.22	7.00	94.00
Second stage:
Negotiation duration	242	911.50	804.80	26	5072
Democracy	242	0.57	0.50	0.00	1.00
Relative scope	244	0.14	0.14	0.00	0.62
Difference in design templates	224	0.62	0.20	0.06	1.00
Enforcement	230	6.32	1.09	4.00	9.00
Escape flexibility	231	4.23	1.23	0.00	5.00
Rigidity on flexibility	230	3.42	1.48	0.00	6.00
Member states	244	6.57	9.22	2	91
Power-asymmetry	244	0.47	0.50	0.00	1.00
Veto players	237	0.34	0.13	0.00	0.61
Rule of law	241	0.443	0.64	−0.97	1.75
Log of trade flows	201	14.25	5.04	−23.03	25.65
WTO membership	242	0.89	0.31	0	1
PTA activity	242	11.21	5.21	0	28
First PTA	244	0.17	0.38	0	1
Geographic distance	242	5,788.57	4,961.49	111.09	19,711.86
Common language	242	0.26	0.44	0	1
Year	242	2,002.63	5.77	1,99	2,01

Results

The baseline findings are displayed in Table 2, where the first column summarizes the results for the first stage (selection into PTA negotiations), and the second column shows the findings for the second stage (duration of PTA negotiations).

Table 2. Baseline results on PTA selection and duration.

	Start of PTA negotiations	Negotiation duration
	(First stage)	(Second stage)
Democracy		−0.32
		(0.22)
Relative scope		−0.35
		(0.62)
Democracy * Relative scope		2.08**
		(0.83)
Difference in design templates	−0.50***	0.52**
	(0.077)	(0.26)
Enforcement		0.01
		(0.08)
Escape flexibility		−0.18**
		(0.07)
Rigidity on flexibility		0.19***
		(0.05)
Member states		0.02*
		(0.01)
First PTA		0.53***
		(0.20)
Power-asymmetry	0.46***	−0.15
	(0.05)	(0.14)
Veto players	−0.01	−1.28***
	(0.13)	(0.48)
Rule of law	0.17***	0.22*
	(0.03)	(0.13)
Log trade flows	0.01***	0.01
	(0.00)	(0.01)
WTO membership	0.22***	0.65***
	(0.05)	(0.25)
PTA activity	0.07***	−0.01
	(0.05)	(0.01)
Geographic distance	−0.02***	−0.03***
	(0.00)	(0.00)
Common language	−0.085*	−0.28*
	(0.05)	(0.17)
Year	−0.02***	−0.05***
	(0.03)	(0.01)
Polity2 score	0.01
	(0.01)
Contiguity	0.50***
	(0.11)
Colonial past	0.32**
	(0.16)
Diffusion	0.00
	(0.00)
Constant	40.78***	110.3***
	(7.20)	(29.08)
Observations	191,255	198
Robust standard errors in parentheses.
***p < 0.01, **p < 0.05, *p < 0.1

We observe that the start of PTA negotiations is significantly influenced by agreement costs. Consistent with our expectations, we find that countries with more similar design templates are more likely to embark on PTA negotiations. In line with previous research, we further find that countries with a good rule of law are more likely to select themselves into PTA negotiations. Interestingly, the coefficient for the variable Polity2 score is positively signed as expected, but statistically insignificant. Additional regressions show that the variable Rule of law takes away the statistical significance of Polity2 score, confirming the omitted variable dynamics detected by Baccini (2014). As expected, WTO membership and high PTA activity increase the probability of countries entering trade negotiations. PTA activity of regional peers has no influence on the eagerness to sign trade deals. Colonial past and common borders have a positive impact on the start of PTA negotiations.

Moving on to the discussion of our main regressors in the second stage, we find support for our first hypothesis. Indeed, a greater Difference in design templates correlates with longer negotiations. Countries with a difference score of 1 negotiate 145 days longer than countries with an average difference score of 0.622. Also, our second hypotheses is confirmed: relative scope increases negotiation duration, but only for democratic countries. Our model predicts that an increase of 0.1 on the relative scope scale results in an average negotiation delay of 279 days. This applies to democracies only. The effect is insignificant for countries with a Polity2 score that is lower than 6.

In addition to confirming the two main hypotheses, we highlight two interesting corollary findings. First, the design variable Escape flexibility exerts a negative and statistically significant effect on negotiation duration, while Rigidity on flexibility has the opposite effect. These results suggest that flexibility provisions help to prevent a war of attrition in international bargaining (Rosendorff and Milner 2001), rather than leading to protracted negotiations due to stabilized long-term prospects for cooperation (Bearce et al. 2015). Second, and perhaps surprisingly, we observe that even though they have a higher likelihood of selecting themselves into PTA negotiations, WTO members take more time to effectively negotiate a trade agreement than non-members. One tentative explanation could lie in the better fallback option of WTO members compared to non-members: In case negotiations fail, WTO members still benefit from the multilateral system. They might therefore be less dependent on concluding PTA negotiations.¹⁰ Common language and a short geographic distance are further factors that speed up trade talks.

In summary, the results from our baseline analysis confirm our two central hypotheses: Countries with more different design templates are slower in concluding trade deals compared to countries with more similar templates. Additionally, PTAs with a high relative scope are preceded by protracted negotiations in the case of democratic treaty partners.

Sticking points in PTA negotiations

The variable Relative Scope is a measure for the average relative scope of a planned PTA in seven issue areas: competition policy, IPR, investment, NTIs, public procurement, services, and standards. Similarly, the Difference in design templates variable captures the difference in design between countries in all of these issue areas. Building on the previous analysis, we now unpack these two variables to identify for each PTA how ambitious and how different the provisions are in a given area. The resulting descriptive statistics are displayed in Table 3.

Table 3. Descriptive statistics on separate issue-areas .

Statistic	N	Mean	St. dev.	Min	Max
Relative scope variables:
RS competition	244	1.48	1.53	0.00	7.00
RS IPR	244	1.51	1.91	0.00	7.00
RS investment	244	1.90	2.71	0.00	10.00
RS NTIs	244	5.14	5.96	0.00	29.33
RS procurment	244	1.27	1.30	0.00	5.00
RS services	244	1.45	2.01	0.00	12.00
RS standards	244	2.11	2.15	0.00	10.00
Difference in design templates variables:
Difference in design templates (competition)	189	0.11	0.25	0.00	1.00
Difference in design templates (IPR)	118	0.03	0.20	0.00	1.00
Difference in design templates (investment)	185	0.04	0.15	0.00	0.74
Difference in design templates (NTIs)	200	0.12	0.26	0.00	1.00
Difference in design templates (procurement)	119	0.01	0.24	0.00	1.00
Difference in design templates (servcies)	172	0.09	0.23	0.00	1.00
Difference in design templates (standards)	188	0.10	0.24	0.00	1.00

To examine which issue areas are particularly challenging in PTA negotiations, we estimate seven separate models. Separate estimations are preferred over a global model with seven regressors due to the risk of multicollinearity between the issue-area-specific scope variables, a concern for which the variance inflation factors lend further support. However, in order to control for issue linkage and the general level of ambition in a trade agreement, we include the overall scope variable in the estimation.

In Table 4, we see that the differences in design templates for investment and procurement matter for negotiation duration. All other specific Difference in design templates variables fail to meet the standard level of statistical significance. With regard to our second hypothesis, ambitious commitments in the areas of IPR, investment, and services are associated with protracted negotiations. Interestingly, non-democratic countries appear to conclude PTA bargaining processes more quickly the more ambitious the clauses on procurement.

Table 4. Baseline results on negotiation duration using issue areas separately.

	Competition	IPR	Investment	NTIs	Procurement	Services	Standards
Democracy	−0.026	−0.27	−0.24	−0.111	−0.24	−0.13	0.040
	(0.19)	(0.18)	(0.17)	(0.180)	(0.20)	(0.18)	(0.19)
RS competition policy	−0.43
	(0.74)
RS IPR		−0.48
		(0.38)
RS investment			−0.16
			(0.34)
RS NTI				−0.60
				(0.69)
RS procurement					−0.56***
					(0.18)
RS services						−0.51
RS standards						(0.74)	0.60
							(0.43)
Democracy * RS (specific)	0.37	1.42**	1.07**	2.05	0.74*	1.68*	0.36
	(0.77)	(0.55)	(0.45)	(1.691)	(0.43)	(0.91)	(0.59)
Difference in design templates (specific)	0.28	0.28	1.42***	0.16	0.49***	0.23	0.31
	(0.22)	(0.27)	(0.49)	(0.21)	(0.19)	(0.26)	(0.21)
Relative scope (overall)	0.68	0.40	−0.00	0.37	0.98	0.34	−0.55
	(0.73)	(0.62)	(0.68)	(0.635)	(0.74)	(0.83)	(0.79)
Difference in design templates (overall)	0.56*	0.52*	0.53**	0.50*	0.54*	0.53**	0.34
	(0.31)	(0.28)	(0.26)	(0.28)	(0.28)	(0.26)	(0.32)
Enforcement	0.02	0.04	0.04	0.04	0.02	0.03	0.02
	(0.08)	(0.08)	(0.07)	(0.0768)	(0.08)	(0.08)	(0.08)
Escape flexibility	−0.23***	−0.21***	−0.19***	−0.22***	−0.22***	−0.20***	−0.22***
	(0.07)	(0.07)	(0.07)	(0.07)	(0.07)	(0.07)	(0.07)
Rigidity on flexibility	0.20***	0.20***	0.20***	0.19***	0.19***	0.20***	0.20***
	(0.05)	(0.05)	(0.05)	(0.05)	(0.05)	(0.05)	(0.06)
Member states	0.01	0.02*	0.02*	0.0138	0.01*	0.01*	0.01
	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)
Power	−0.07	−0.16	−0.07	−0.07	−0.12	−0.10	−0.03
	(0.14)	(0.15)	(0.13)	(0.13)	(0.13)	(0.17)	(0.14)
Veto players	−1.33***	−1.17**	−1.24***	−1.25**	−1.11**	−1.39***	−1.39***
	(0.49)	(0.49)	(0.47)	(0.50)	(0.50)	(0.54)	(0.48)
Rule of law	0.18	0.23*	0.11	0.15	0.20*	0.19	0.19
	(0.13)	(0.13)	(0.12)	(0.13)	(0.12)	(0.13)	(0.13)
Log trade flows	0.015	0.02	0.02	0.02	0.02*	0.01	0.01
	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)
WTO membership	0.65**	0.71***	0.62**	0.66**	0.55**	0.66**	0.66***
	(0.27)	(0.25)	(0.25)	(0.26)	(0.27)	(0.26)	(0.25)
PTA activity	−0.00	−0.00	−0.01	−0.01	−0.01	−0.01	0.00
	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)
First PTA	0.62***	0.57***	0.51***	0.55***	0.66***	0.52**	0.44**
	(0.21)	(0.20)	(0.19)	(0.20)	(0.20)	(0.22)	(0.22)
Geographic distance	−0.03***	−0.03***	−0.03**	−0.03**	−0.03**	−0.03***	−0.03**
	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)
Common language	−0.33*	−0.33**	−0.34**	−0.34*	−0.27	−0.33*	−0.26
	(0.18)	(0.16)	(0.15)	(0.19)	(0.17)	(0.17)	(0.17)
Year	−0.05***	−0.06***	−0.06***	−0.05***	−0.04***	−0.05***	−0.05***
	(0.01)	(0.02)	(0.01)	(0.01)	(0.01)	(0.01)	(0.01)
Constant	112.90***	125.30***	123.30***	113.70***	92.91***	113.30***	107.40***
	(28.60)	(31.53)	(27.31)	(28.24)	(27.62)	(30.60)	(28.65)
Observations	170	118	185	198	119	170	188

Robust standard errors in parentheses.

***p < 0.01, **p < 0.05, *p < 0.1

Based on our empirical evidence, we can confirm that the lessons learned from the prominently discussed mega-regional trade negotiations TPP, TTIP, and CETA fit into a broader picture. Since especially investment and IPR are part of a widespread debate, our analysis reveals a consistent link between the degree to which a respective issue mobilizes domestically and the pace of negotiations internationally.

Robustness checks

In addition to our baseline estimations, we perform six sets of sensitivity checks to examine whether our results are robust to alternative specifications¹¹:

First, we re-run our survival model with alternative variables for regime type. More specifically, we measure regime type based on the Unified Democracy Scores, which are composite indices for regime type based on 10 extant scales (Pemstein et al. 2010). A score greater or equal to 1 classifies countries as democracies. The counterintuitive positive effect of RS standards in less democratic countries is no longer significant. All other results remain stable.

Second, we alternate our metric for the relative scope in PTA design in two ways. First, we take the absolute instead of the relative scope to estimate a placebo effect. If our mechanism is correct, the absolute scope should not have any significant effect on negotiation duration. The results confirm this expectation: all relative scope coefficients are insignificant.

Third, we try to tackle the concern that the degree to which ambitious PTA provisions matter in the domestic political arena is influenced by the economic size of the trading partner: Agreeing to wide-ranging provisions in a PTA with a small economy may be fundamentally different from doing the same in a treaty with a large economy due to different levels of (prospective) import penetration. For this purpose, in our construction of templates, we consider only those PTAs as constitutive of the template where the largest partner country exhibits a similar economic size to the largest current bargaining partner. We do so by dividing our sample into three equal, percentile-based categories for economic size. Subsequently, we match past and current PTA partners. The results again confirm our main hypotheses.

Fourth, we use an alternative metric to calculate the Difference in design templates measure. Instead of the Cohen kappa, we take the Rogers and Tanimoto dissimilarity, which double weights disagreement. The results prove to be even more in favor of our hypothesis: All Difference in design templates coefficients are positive and significant.

Fifth, we estimate a more extensive model featuring two additional political variables. As a first variable, we include a regressor for government partisanship. It could turn out that countries governed by left-wing parties approach negotiations with greater care than countries with right-wing governments. To measure governing party positions, we apply the vanilla method by Gabel and Huber (2000) to manifesto data. With our second variable we account for agricultural trade liberalization as a potential sticking point on the agenda. Opening up agricultural markets has proven to be difficult given the strength of well-organized and protection-seeking farmers. In this context, we rely on information on WTO coalitions provided by the multilateral trade organization to examine whether in PTA negotiations there is at least one member from the Cairns group (offensive interests on agriculture, such as Chile) and one country from the G-10 (defensive interests, such as Switzerland). The results remain unchanged.

Sixth, we conduct a subsample analysis, where we drop the Australia-Korea agreement. Recall from above that Australia and Korea negotiated for more than 14 years and have a high relative scope. The findings show that this outlier is not driving our results.

In a seventh set of tests, we use two variables as substitute and more trade-related measures for the degree of ambition in a PTA. First, we rely on a variable for average dyadic trade flows between PTA partners and interact it with Democracy. The intuition is that high trade flows could result in a situation where larger segments of domestic society have to deal with the ramifications of a proposed trade agreement than under circumstances of negligible trade levels. Second, we use the average Grubel-Lloyed index as a measure for the level of intra-industry trade (IIT) between the prospective agreement partners (Grubel and Lloyd 1971). Defenders of the New Trade Theory have argued that IIT is less adjustment-cost-intensive than inter-industry trade due to monopolistically competitive market structures (Balassa 1966). We find that as the level of trade flows among PTA partners increases, democracies start to exhibit negotiation delay. Somewhat counter-intuitively, we observe the same result with the Grubel-Lloyed index, indicating longer PTA negotiations in the presence of high levels of IIT. Two explanations are conceivable for this result: First, it has been argued that while IIT is less adjustment-cost-intensive than inter-industry trade, lobbying under the former is more pronounced and hence protectionism more likely because individual firms have greater incentives to engage in private action (Gilligan 1997). Second, IIT has been shown to correlate positively with deep integration commitments in trade agreements, making it difficult to disentangle the effects of trade flow structures and PTA design on negotiation duration (Kim 2015).

Conclusions and future research

Against the background of stalled multilateral trade talks, preferential trade negotiations have become a pervasive feature of the landscape of international cooperation. At the same time, PTAs differ starkly in terms of negotiation duration. What explains this variation?

We observe two robust patterns regarding the drivers of bargaining processes for trade liberalization: First, differences in initial positions matter both for whether national delegations convene at the bargaining table in the first place and the duration of their trade talks. Second, democracies can seal relatively ambitious PTAs only after protracted negotiations. Our empirical analysis reveals services, investment and IPR as particularly sensitive agenda items for democratic leaders in trade negotiations.

Our paper contributes to the empirical literature on trade agreement design (Baccini et al. 2015; Dür et al. 2014). Instead of treating all PTAs in an undifferentiated manner, scholars are increasingly cognizant of differences across agreements in terms of the breadth of issues covered and the depth of commitments enshrined therein. We hypothesize that this variation in institutional design can inter alia translate into different bargaining dynamics at the international level, depending on countries’ level of ambition relative to previously signed agreements. Moreover, our results indicate that commitments at the international level are not necessarily beneficial to democratic leaders, but might even induce costs, if they extend well beyond previously signed treaty clauses.

With this article, we also make a case for embedding PTA bargaining processes, and international negotiations more generally, in a broader context of cooperation. Negotiations do not occur in a “vacuum” (Castle and Pelc 2017; Odell 2000). Rather, national delegations convene at the bargaining table with a track record of previously signed agreements, which will inform their initial bargaining positions and concession management throughout the process (Crump 2007). Recently, scholars have started undertaking efforts to develop measures for individual countries’ bargaining positions.¹² In this paper, we were able to show that differences in bargaining positions matter for the likelihood of countries entering PTA negotiations and the duration of their bargaining processes.

Further research in this regard could improve our understanding of international cooperation and provide an empirical complement to existing formal models of bargaining processes and strategies.

Against the background of the widely discussed mega-regional trade projects, our analysis also has implications for the policy-making world: Rather than finding that trade negotiations are completely under the radar of domestic politics and unconditionally pursued by democratic leaders, we observe that certain design features can lead to protracted negotiations at the international level. If democratic leaders want to maintain the momentum for preferential trade negotiations, they might wish to engage more strongly with domestic political actors expressing resistance to ambitious market liberalization. Gradualist approaches to market opening or pro-active communication strategies about the relative scope of a planned PTA might be worth considering in this regard. A further, normative discussion of these dynamics lies beyond the scope of this paper. However, we hope that the results presented in this study invite scholars and policy makers alike to shed more light on the nexus between domestic politics and international trade negotiations.

Acknowledgments

We are grateful to Todd Allee, Leonardo Baccini, Matthew Castle, Stephen Chaudoin, Andreas Dür, Manfred Elsig, Andrew Lugg, Gabriele Spilker, participants at the PEIO 2017, attendants of workshops in Salzburg and Bern, the anonymous referees, and the editors of International Interactions for comments on earlier versions of this paper.

Additional information

Funding

This work was supported by the Swiss National Science Foundation, NCCR Trade Regulation.

Notes

¹ The Urugay round took more than 7 years and the Doha round has been negotiated for more than 16 years and is still not concluded. The average PTA negotiation duration in our sample is 2.5 years.

² To develop our argument, we build on important research on the economics of bargaining dynamics (Cameron and Tomlin 2002; Fearon 1998; Heron 2011; Limão 2007; Meunier 2007; Syropoulos 2002). Our study is consistent with previous work. However, whereas these studies often relied on formal modeling or case studies, we provide large-n quantitative evidence. In addition, we offer an argument and new evidence on the effect of the multidimensionality of international institutions on bargaining dynamics. Most scholars have so far focused on tariff cuts in PTAs and paid less attention to the multiplicity of issue areas addressed in these agreements.

³ We understand the protection of national sovereignty to regulate a respective issue area as the maintenance of policy space. Policy space carve-outs, then define the attempt of countries to preserve national sovereignty over an issue area.

⁴ After the withdrawal by the US, the TPP has been modified content-wise and relabeled as the Comprehensive and Progressive Agreement for Trans-Pacific Partnership (CPTPP) among the remaining 11 Pacific-Rim countries.

⁵ For further information on the model choice criteria see Appendix C.

⁶ For a detailed list of the treaties see Appendix A.

⁷ See Appendix B for a detailed list of the variables.

⁸ Our results are robust to the exclusion of PTAs where at least one member states negotiates its first agreement.

⁹ , where is the observed agreement and is the expected random agreement (product of marginal totals).

¹⁰ We thank one of our anonymous reviewers for this interpretation.

¹¹ Detailed output tables can be found in Appendix D.

¹² For an example in international investment see Allee and Lugg (2016).