2.1 Cooperation Events, Bargaining Space, and Preference Similarity

The theoretical framework is based on the definition of three key concepts: cooperation events, bargaining space, and preference similarity. Of these, cooperation events are central. Cooperation is the process in which participants adjust their behavior to the actual or to the anticipated preferences of others. Such cooperation often implies a process of policy coordination, which can take place if two countries take a common stance in WTO negotiations (Keohane, Reference Keohane2005). We conceptualize common stances as cooperative behavior resting on three reasons: First, from a communal point of view, these international multi- party and multi-issue negotiations are highly complex, and some form of cooperation and coordination is needed to ensure successful outcomes (Olson, Reference Olson1971). By forming alliances, the complexity of the negotiations can be reduced, increasing the likelihood of successfully completing the negotiations (Dupont, Reference Dupont and Zartman1994, Reference Dupont1996). Second, from the perspective of the countries entering coalitions, cooperation allows them to increase their power by pooling their (hard and soft) negotiation resources (Betzold et al., Reference Betzold, Castro and Weiler2012). Cooperation increases the power of the individual countries, which have incentives to form cooperative ties and which would otherwise be rather powerless in large negotiations. Third, cooperation implies that two countries have common interests which they represent together. In general, we assume that cooperation hinges on countries having similar preferences on those interests, since preference similarity aids finding an agreement in negotiation settings (Hinich and Munger, Reference Hinich and Munger1997). In a multi-party setting, when a general agreement is hard to achieve, preference similarity makes countries natural allies and leads to coalition formation. Therefore, we define the instance in which two countries make a joint statement or they register mutual support in the WTO as a cooperation event. We are able to analyze this form of cooperation, which the WTO facilitates, empirically while recognizing that other types of cooperation also exist. WTO talks are a form of international, multi-party, and multi-issue negotiations (Gray, Reference Gray2011). During such negotiations, the preferences of the parties are crucial and they determine the outcome of negotiations as shown by Moravcsik (Reference Moravcsik1997). The reason for this is that preferences define and determine the bargaining space, i.e. the distance between the various negotiating parties. Intuitively, agreement is facilitated if the preferences among the actors are relatively similar, the bargaining space is therefore quite narrow, and possible solutions are feasible and are not too far away from the preferences of all parties (Hinich and Munger, Reference Hinich and Munger1997; Hopmann, Reference Hopmann1996). The importance of this mutuality of interests has long been established by IR scholars for two player games (Axelrod, Reference Axelrod1967; Jervis, Reference Jervis1978).

Our proposed theory rests on the assumption that in international negotiations preference similarity leads to cooperation (Hinich and Munger, Reference Hinich and Munger1997; Hopmann, Reference Hopmann1996). Shared interests lead to similar preferences, which make countries natural allies. Multi- party, multi-issue negotiations such as the WTO trade talks are highly complex. One way of reducing this complexity is the formation of (formal or informal) cooperative ties based on preference similarity (Dupont, Reference Dupont1996). By bringing forward such similarities regarding individual negotiation issues, the bargaining space becomes less crowded and more easily manageable, which increases the likelihood of finding an agreement. As an additional incentive, pooling positions also increases the bargaining power – and thus the influence – of negotiation parties, since multiple parties carry more weight than single parties (Chasek and Rajamani, Reference Chasek and Rajamani2003; Dupont, Reference Dupont1996; Laatikainen and Smith, Reference Laatikainen and Smith2020). Pooling of positions and interest-based cooperation, is thus a precondition for both a successful conclusion of multi-party negotiations, but also the bargaining success of individual countries. For us, the key question is: What are these similarities which can lead to cooperation in negotiations between countries? The premise here is that shared interests lead to preference similarity. These similarities are expressed as common stances, which are one form of WTO cooperation in the negotiation process.

2.2 Domestic Actors and WTO Cooperation

We scrutinize the factors that underpin the formation of cooperation at the dyadic level in international bargaining situations, a necessary condition for the success of such negotiations. Domestic structures inherent to different political regimes condition international cooperation (De Mesquita et al., Reference De Mesquita, Morrow, Siverson and Smith2002; Odell, Reference Odell2009; Weiler, Reference Weiler2012). We expect domestic political and economic actors to be relevant in the process. The widely shared assumption in the literature that democratic regimes act, on average, more cooperatively in international negotiations rests on the idea that such negotiations often focus on the provision of a specific public good (such as free trade, international security, or environmental protection), and that democracies have a tendency to provide more of such public goods (De Mesquita et al., Reference De Mesquita, Morrow, Siverson and Smith2002; Lake and Baum, Reference Lake and Baum2001; Rosendorff, Reference Rosendorff2006). We build on this idea that democracies are more cooperative in general, but add a qualifier, based on new evidence from domestic trade policy that this higher propensity to cooperate is not based on the median voter’s preference, but on lobbying efforts of trading industries (Betz, Reference Betz2017; Betz and Pond, Reference Betz and Pond2019). This leads to the expectation, explained in more detail below, that democracies are only more cooperative in dyads with larger trade relationships, since this is what interlinks the preferences of the countries in question.

The expectation that democracies are more cooperative regarding trade is based on the median voter theorem (Congleton, Reference Congleton2004). Highly simplified, research in trade economics resting on the Heckscher–Ohlin theory argues that free trade is the best policy most of the time (Baldwin, Reference Baldwin2008). Countries can specialize in areas of the economy in which they have a comparative advantage, i.e. where they are able to produce at lower (relative) opportunity costs compared to their competitors. Through free trade, exporters can provide their goods and services to a (much) larger market, while consumers benefit from more choice, better quality, and lower prices. The standard assumption in the economic literature is therefore that the average voter-cum-consumer benefits from free trade, and as a consequence supports free trade. Based on this theory, all countries involved in international trade should benefit overall. From this perspective, the position of negotiating parties should on average lean towards liberalization if the negotiating country aims to maximize overall economic benefit. Such trade liberalization within countries can, however, also produce losers, i.e. the (prior to liberalization) scarce factor of production (Rosendorff, Reference Rosendorff2006). In other words, there is still the potential of significant proportions of the economy to lose (Kapstein, Reference Kapstein2000). However, if the average person wins, as the standard model argues, more trade should be in the common interest and also be the preferred solution. Trade, according to this model, will lead to enhanced welfare for at least certain segments of the population and as the gains outweigh the losses, the losing segments of society can be compensated. The overall net gains should lead to support for free trade by the general population. An increasingly rich body of research tests these micro-level expectations derived from macro-economic theory. They rely on population-based experiments. Bearce and Moya (Reference Bearce and Moya2020) find that concerns about job loss outweigh awareness of lower consumer prices. Baccini, Guidi, and Poletti (Reference Baccini, Guidi and Poletti2022) ask whether citizens would prefer more redistribution or less trade in times of crises. They find that citizens from Italy, France, and Germany support politicians who advocate social spending and redistribution rather than trade restriction. Therefore, citizens are aware of the costs of trade, such as potential job losses, but they expect to be compensated domestically rather than giving up on the consumer benefits of international trade.

Building on this expectation of general support of international trade by the population, selectorate theory (De Mesquita et al., Reference De Mesquita, Morrow, Siverson and Smith2002) then posits that the leader of a political regime controls the policy output, but they are constrained by the selectorate. If the selectorate is large, as in a democracy where the electorate equals the selectorate, the winning coalition, i.e. those who define the policy outcome, should therefore be driven by the will of the majority of citizens of voting age. However, in non-democracies this influence of the average citizen is less pronounced. Thus, the outcome of the cooperative behavior depends more directly on who is affected (positively or negatively) by trade, and the power these stakeholders have within a given non-democratic political system.

Little is known as to why some autocratic leaders are more pro-trade than others. But the volatility of autocratic trade policies is well-documented (Chow and Kono, Reference Chow and Kono2017). Chow and Kono argue that the mode of entry of autocrats can explain their propensity to cooperate on international trade. Leaders who have come into power through extralegal means require public support to consolidate their power, and use trade openness to gain this support. Once their power is consolidated, they often revert to restricting trade. Therefore, the support for trade cooperation of autocrats is driven by a desire to increase the welfare of the population – but only if their position is not secure. Based on this established theory, we could simply derive the expectation that more democratic countries, interested in providing a public good, are acting more cooperatively in international trade negotiations. However, some recent developments in the scholarly debate have cast doubt on this conclusion. For instance, Bättig and Bernauer (Reference Bättig and Bernauer2009) have demonstrated that while democracies have a higher level of political commitments (i.e. they promise to provide more of the public good) for the climate change negotiations, their actual policy outputs in terms of emission reductions are not clear. More related to free-trade policy, Betz and Pond (Reference Betz and Pond2019) do not find evidence that democratic countries offer lower tariffs on goods with high consumption shares, casting considerable doubt on the validity of the selectorate theory and the median voter theorem presented above, at least for the trade realm. Instead, what Betz (Reference Betz2017) posits is that lobbying and reciprocity of interests are much more important for determining free-trade outcomes and cooperation. In sectors with strong exporting industries, strong lobbying efforts arise for free trade, which leads to domestic tariff cuts in exchange for liberalization abroad. In the same vein, research focusing on firms argues that trade divides industries and that larger firms are the most vocal advocates for globalization (Kim and Osgood, Reference Kim and Osgood2019). Export lobbies can influence the lowering of tariffs but their influence is limited by governments’ greater freedom in taxing or subsidizing imports (Levy, Reference Levy1999). Research analyzing the response of businesses to the trade war between China and the US during Donald Trump’s presidency shows that businesses have been rather quiet (Dolan et al., Reference Dolan, Kubinec, Nielson and Zhang2025). This may suggest that during periods of relative political stability, lobbying in favor of free trade occurs largely in the background, attracting limited public attention. However, when trade policy becomes highly salient, lobbying efforts may lose effectiveness, which may result in lobbyists adopting a more restrained approach. On the basis of this new evidence, we propose to update earlier theories on democratic (compared to non-democratic regimes) cooperativeness with respect to international trade negotiations. Instead of simply claiming that democracies are more cooperative in international trade negotiations, we postulate that this is only the case conditional on specific interest group efforts within the country (or country pair) in question. In other words, we still expect more cooperation between democratic dyads, but only when there is sufficient trade between such partners that the affected trading industries step up their lobbying game sufficiently. Therefore, this can be another driver of international trade liberalization.

Of course, the question can (and should) be asked as to why lobbying industries would not also lobby their governments under autocratic rule. Autocracies, as Przeworski (Reference Przeworski2022, 1) states, also ‘collect garbage, regulate traffic, issue dog licenses, and fill street holes: they govern’. One could add that they also negotiate (bilateral and international) trade agreements, and that all such government dealings directly affect the people living under autocratic rule. Hence, some form(s) of lobbying should also be expected in autocracies.

Unfortunately, most studies of interest groups and lobbying focus on democracies, with some notable recent exceptions. While some scholars clearly find that (general) lobbying activity is higher in democracies (Kanol, Reference Kanol2016; Bearce and Roosevelt, Reference Bearce and Roosevelt2023), the question remains: Why? In a recent book, Grömping and Teets (Reference Grömping and Teets2023) develop an interesting framework for analyzing lobbying under authoritarian rule. The authors propose that, while societal groups cannot simply be seen as uninterested bystanders in autocratic regimes, there are overall fewer opportunities for them to engage in due to stricter societal control, and leaders are less interested in what such groups have to say, given that the accountability of democracies is absent. Thus, there are fewer access points for policymakers, because they demand less information and feedback from their constituency. As a consequence, societal groups – whether they represent citizen groups or business interests – have more difficulty in being represented by government action in autocracies than in democracies.

For our trade negotiations, this means that in such a setting with monopolized access to decision makers (Boix and Svolik, Reference Boix and Svolik2013), even industrial groups have a hard time getting the ear of their governments. This is particularly the case since, as shown by Betz and Pond (Reference Betz and Pond2019), governments might be more interested in the revenue generated by tariffs than in other societal interests. Moreover, while in democracies strong lobbies can find ways to sway their governments, autocrats can ignore lobbyists because these have fewer access points and lower information demand than those in democracies. In autocracies, lobbyists can even be prevented from organizing and expressing their interests due to strict societal controls (Grömping and Teets, Reference Grömping and Teets2023).

In our specific case of WTO negotiations, we argue that a high dyadic trade relationship directly links the two countries’ interests in question on the trade issue under negotiation, and is thus a source of also interlinking their preferences. Following the discussion from above, the more democratic the two countries in question are, the more the industries in question will lobby for trade cooperation within these dyads, which in turn should lead to intensified cooperation during the WTO talks. This idea of dyadic cooperation of democracies (but also non-democracies) hinges on the notion of preference similarity, i.e. countries with more similar interests are more likely to cooperate (Hinich and Munger, Reference Hinich and Munger1997; Hopmann, Reference Hopmann1996). In autocratic countries, such preference similarity based on trade relationships is not well represented due to the lack of lobbying opportunities, and thus has no, or a much weaker, effect than it does in democracies. Thus, we hypothesize that for low trade relationships, the likelihood of cooperation is unrelated to regime type; i.e. other than the long-standing assumption in the literature that democracies are more cooperative by default, we do not expect to see such differences when specific societal interests driving such behavior are absent. Once these interests come into play due to more prominent trade relationships, we expect democracies to be more responsive to societal (lobbying) demands. Hence, dyads including democracies (and particularly those with two democracies) are expected to act more cooperatively at higher levels of bilateral trade.

So far, we have focused the discussion on the difference between democratic and autocratic regimes, but autocratic regimes are of course not a monolithic block of countries. Thus, we now discuss how we expect cooperation to differ among distinct autocratic regime types, based on the classification of Geddes et al. (Reference Geddes, Wright and Frantz2014).

In non-democratic regimes (which often coincide with less-developed economies) the labor force is the abundant factor according to the Heckscher–Ohlin model (Wu, Reference Wu2015). Thus, the large population in such countries would stand to benefit the most from trade liberalization. Reducing inequality and increasing the income of the poorest segments of society via trade can help autocratic governments to prevent demands for regime change, and as a consequence ensure regime stability (Acemoglu and Robinson, Reference Acemoglu and Robinson2006). From this classical perspective, some level of cooperativeness should also be expected by less democratic regimes in the WTO negotiations. However, as we have seen above, when it comes to trade policy, even in democratic countries governments do not necessarily behave in line with such median voter models of policymaking. Instead, they follow other interests, such as revenue generation or protectionist policies by erecting trade barriers. Overcoming such trade barriers and reciprocity in trade cooperation requires lobbying from affected trading industry groups.

The question, then, is: How do various autocratic regime types differ in terms of allowing and enabling such lobbying efforts? Regimes are defined according to ‘the rules that identify the group from which leaders can come and determine who influences leadership choice and policy’ (Geddes et al., Reference Geddes, Wright and Frantz2014). The selection of leadership in different political regimes and the way in which they take into consideration the interests of societal groups will lead to a different representation of these interests at the international stage (Geddes et al., Reference Geddes, Wright and Frantz2014, 315).

How easily groups can form and are able to lobby autocrats depends again on how much information can flow in such regimes from societal groups. This is, in turn determined by the number of access points to the regime (possibility of lobbying), but also by how strict the social control of a given regime is (formation of groups). Grömping and Teets (Reference Grömping and Teets2023, 23) wrote ‘that information demands differ across autocratic regime types. For instance, more competitive party-based regimes likely have higher information demands than personalist or military regimes without electoral pressure.’’ We agree with this assessment and add that regimes more tailored to a single leader (such as monarchies and personalist regimes), not only have lower information demands, but also tend to be more repressive than party-based autocracies.

Party-centered regimes share some characteristics, such as greater accountability and transparency, with democracies (Dai et al., Reference Dai, Snidal and Sampson2010; Mattes and Rodriguez, Reference Mattes and Rodriguez2014). Indeed, China, the most economically successful party-centered autocracy, has an internally contested economic governance process with various layers of authority and no convergence of preferences within the state (Tan, Reference Tan2020; Tarlea, Reference Tarlea2025). In such settings, societal groups and demands have, in our view, the best chance of being heard among the various regime types in question. Personalist leaders have been shown to be the least likely to be removed from power because they are in charge of the security apparatus and of political appointments (Mattes and Ro-driguez, Reference Mattes and Rodriguez2014). The leader can impede elite coordination against his rule and thus the threat of removal from office is extremely low (Weeks, Reference Weeks2008). At the same time, personalist rulers have been argued to be dependent on international trade because they use international revenues to fund their patronage networks (Escribà-Folch and Wright, Reference Escribà-Folch and Wright2010). The institutions that allow for repression and for the collection of taxes in personalist regimes tend to be weaker in comparison to military or to party systems. Therefore, personalist leaders depend on international revenues which they use for domestic political payoffs (Escribà-Folch and Wright, Reference Escribà-Folch and Wright2010), and this, we contend, opens an entry point for lobbying trade associations.

When bilateral trade is low among autocratic dyads, we do not expect to observe differences among them in trade cooperation negotiations at the WTO. However, as trade relations become more important, and lobbying efforts to cooperate in the specific dyad increase, we posit that this is easier and more successful in party-centered and personalist regimes than in military regimes and monarchies (but less so than in democracies).

To quickly summarize our expectations: when dyads have no, or very low, trade relationships, we do not expect regime types to play a big role in how cooperative they are within the specific dyad. This, we argue, is also true for democracies, and in this way we depart from the long-standing expectation that democracies are by default more cooperative than other regimes. Once the trade relationship between two countries intensifies, and lobbying efforts to facilitate trade and to cooperate pick up, lobby groups’ ease of mobilization and access to policy-makers influence how cooperative governments behave. In democratic countries, lobbying is least restricted and therefore dyads that include democracies are expected to become more cooperative the more they trade, followed by dyads including personalist and party-centered regimes, where some lobbying regarding trade cooperation should be expected, while in military regimes and monarchies we expect the least increase in cooperation as trade relations grow.

3.1 Dependent Variable

The WTO officially commenced on 1 January 1995 under the Marrakesh Agreement, signed by 123 nations on 15 April 1994, replacing the General Agreement on Tariffs and Trade (GATT). It is the largest international economic organization in the world. This paper relies on data on bilateral cooperation during these WTO negotiations focusing on various issues and collected by Günther (Reference Günther2016).

This dataset on cooperation among WTO members during the Doha Round negotiations is built around undirected dyads for all WTO members during the years between 2000 and 2011. Because cooperative behavior, as operationalized in this paper, refers to the frequency of interactions between dyads of WTO members without focusing on the initiator of such cooperation, non-directional dyads are used. In other words, for the frequency count of cooperation between country A and country B, it does not matter whether the former or the latter established the contact and started the cooperative interaction. Thus, the dependent variable in this study, WTO member’s cooperative behavior in the Doha Round negotiations, refers to the frequency of cooperation between pairs of WTO members.

Cooperation among WTO members is seen as two members having authored and submitted joint proposals, statements, or other kinds of negotiation documents. To generate the data for WTO member’s cooperative behavior in the 12 year period, of negotiations, a coding procedure was applied to joint proposals and other written submissions. These documents are publicly available via the WTO’s official online database, in contrast to the negotiation meetings and minutes of these meetings which are not public. From the publicly available documents, data were coded regarding the authors of the document (i.e. WTO members), the date the document was created, and which negotiation issue it related to. If a document had more than a single country as its author, the count of cooperative ties for the given year increased by one for all the countries listed as contributors.

This coding procedure was applied to all dyads where the countries held full WTO membership negotiating independently during any year over the research period from 2000 to 2011. Overall, there were 138 such individual WTO members holding full membership, which constitute 9,453 undirected dyads. Due to the research period of 12 years, the analysis should include 113,436 dyad-years. However, if in a given year one member (or both) of such a dyad has not been an independently negotiating full WTO member, a missing value was assigned for this dyad and for the year in question. We made an additional adjustment for the EU, since the EU member states are not considered independently negotiating WTO members and instead act as a ‘single customs union with a single trade policy and tariff’ (WTO, 2015) with a strong executive representing the EU in international negotiations (Tarlea et al., Reference Tarlea, Bailer, Kudrna and Wasserfallen2024). Thus, if a country was an EU member in a given year, no dyads for this country are recorded, and instead this country is considered to be part of the dyads the EU formed with other (non-EU) countries in the negotiations. The 15 EU member countries that joined before the onset of the Doha round are therefore not part of the dataset at all. The 12 countries that joined the EU during the study period (Cyprus, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Malta, Poland, Slovak Republic and Slovenia in 2004; Bulgaria and Romania in 2007) only formed cooperative ties for the years prior to becoming EU members.

Overall, we thus include 94,530 dyads of active, independently negotiating WTO members pairs over the 12 years included in the study. For each such dyad, we record a value for the number of cooperative interactions, i.e. our dependent variable. This count variable ranges from zero to 34 in our overall count of all negotiation issues, with about two-thirds of our observed dyads (i.e. 63,048 of the 94,530 observations) recording zeros, thus indicating the non-existence of cooperation among the respective WTO members. The maximum value of 34 was recorded in the year 2005 for the two Asian WTO members Japan and Taiwan (which officially operates under the unwieldy name Separate Customs Territory of Taiwan, Penghu, Kinmen, and Matsu) in the WTO in order to avoid conflict with the People’s Republic of China).

3.2 Independent and Control Variables

Regime type: To capture a country’s regime type, we rely on the classification of autocratic countries by Geddes et al. (Reference Geddes, Wright and Frantz2014). To operationalize regime type for E1, we simply merge all non-democratic regimes into a single autocratic category. Democracies are not classified by Geddes and colleagues; hence, we recode the countries with missing values in the dataset as democracies, and cross-check this classification with data provided by Freedom House (Freedom House, 2020). Since we need dyadic data, we code country pairs, and form categories for pairs with two democracies (1), for mixed dyads (2), and for autocratic dyads, and (3), because we use dummies for the different pairs in the models below (with democratic dyads as the baseline), the ordering of the variables does not affect the results.

For (E2), we need a finer measure of the dyadic regime type combination. Geddes et al. (Reference Geddes, Wright and Frantz2014) differentiate autocratic regimes into party-centered autocracies, military autocracies, personalist regimes and monarchies. Since there are only few monarchies left in the set of countries that are WTO members today, we reclassify and merge this category with personalist regimes, as these are in our view the closest categories. Again, countries not classified by the authors are coded as democracies. Thus, this second variable capturing regime type has four categories: democracy, party regime, personalist regime, and military regime. We code country pairs as follows: (1) two democracies, (2) democracy-party, (3) democracy-personalist, (4) democracy-military, (5) two party regimes, (6) party-personalist, (7) party-military, (8) two personalist regimes, (9) personalist-military, (10) two military regimes. Again, due to our modeling approach, the ordering of the variables does not affect the results. Based on the two independent variables capturing regime type, we test the hypotheses described above (interacted with trade flows). To avoid simultaneity issues, we lag the two independent variables and other time-varying independent variables by one year.

Trade flows: Data for bilateral trade flows between WTO members (in current US millions of dollars) come from the Correlates of War project (Barbieri et al., Reference Barbieri, Keshk and Pollins2012, Reference Barbieri, Keshk and Pollins2009), and are based on trade data provided by the IMF. The two trade flows within each dyad, i.e. exports and imports, are added to create a measure that represents the strength of the trading relationship among all dyads in the dataset. The trade flow values for the EU are calculated as the sum of the trade flows of all its member states in a given year. The Correlates of War project does not contain data on Hong Kong, Macao, and Liechtenstein, which results in missing trade flow values for those dyads consisting of one of these WTO members. This variable tests the assumption that higher trade flows among dyads lead to an increased likelihood of cooperation. In some models, we also interact this variable with the main independent variable, i.e. regime type, to check whether the strength of the trading relationship has a significant impact on the various regime combinations and how they cooperate. As trade flows are time-varying, and of particular concern when it comes to potential reverse causality, the variable is lagged by one year.

Geographic proximity: We assume proximity to play a role both for a closer trade relationship, and hence also for the likelihood of cooperation during WTO negotiations. To capture proximity, we rely on the cshapes-library of the R statistical computing environment (Weidmann et al., Reference Weidmann, Kuse and Gleditsch2010), and use it to calculate the minimum distance between any country pairs of WTO members in our dataset. Thus, a value of zero indicates that two countries are directly bordering each other, which greatly facilitates trade for such dyads.

Cultural proximity: Cultural proximity, such as a common language, can also greatly facilitate trade and thus increase the likelihood of cooperation during the WTO negotiations. Whether two countries use the same official language is captured via data provided by Mayer and Zignago (Reference Mayer and Zignago2011).

GDP: We measure GDP using the data from the World Development Indicators (World Bank, 2020). Again, we need dyadic data for our models. First, we use a homophily measurement, i.e. the (logged) difference between the GDP of the two countries in question. We assume that if the GDP difference increases, the power disparity between the countries of the dyads grows and the likelihood of cooperation decreases. In addition, by adding together the two GDP values of the countries in question, we also include a measure of total (logged) GDP of the dyad. In doing so, we control for whether more powerful countries are more likely to cooperate. Both GDP variables are again lagged by one year.

Colonial ties: Finally, we include a colonial ties dummy, indicating whether two countries are ‘bound’ by a colonial history, which should, we argue, increase their propensity to cooperate in the WTO negotiations. The data are taken from the Quality of Government Standard Dataset (Teorell et al., Reference Teorell, Dahlberg, Holmberg, Rothstein, Khomenko and Svensson2016).

Negotiations coalitions: Multilateral international negotiations are complex processes in which smaller states or countries from the Global South can often easily become sidelined, with powerful countries dominating the talks. Joining coalitions has proven effective at facilitating the participation of these states (Rickard, Reference Rickard2024; Weiler and Castro, Reference Weiler and Castro2020). Being in the same group in a coalition is an additional observational implication of common interests and how these interests can influence cooperation. Therefore, we expect that countries that share a negotiation group also cooperate more. We expect a substantial part of cooperation within the WTO to be driven by working together in coalitions. Although statements by coalitions are excluded from our analysis, being in such coalitions is still an important expression of preference similarity, and countries in coalitions are used to working closely together. Thus, when controlling for coalition groups it is important not to overestimate the regime type trade effect we want to isolate.

3.3. Empirical Strategy

So far an often unexplored issue in international negotiations is whether the decision to cooperate is a one-shot game, or rather represents a sequential decision-making process. Scholars have thus far mostly theorized cooperation in international forum as the consequence of a single decision: negotiators (or negotiation teams) decide either to include or exclude other country delegations as potential cooperation partners. If both sides decide that cooperation is mutually beneficial, a cooperative tie is formed. In contrast, we model this cooperation game as a sequential decision-making process. In a first step, the decision to cooperate can be quite broad, and can be an indication of a potentially wide coalition to which country delegations want to keep the door open. However, in a second step, negotiation teams then decide on the amount of cooperation with different partners, and this second step separates the core negotiation alliances from those in the periphery, which are less relevant from the point of view of the negotiation team of the country in question.

In order to capture this sequential decision-making process, and to test the relationship between our independent and dependent variables, we rely on Hurdle models, sometimes also referred to as Cragg models. The basic idea is that a Bernoulli probability governs the binary outcome of whether a count variable has a zero or positive realization. If the realization is positive, the hurdle is crossed, and the conditional distribution of the positives is governed by a truncated-at-zero model (Cragg, Reference Cragg1971). Applied to our cooperation data, we argue that there are two stages that contribute to the intensity of cooperation among dyads in the WTO negotiations. First, a selection stage, which decides whether two dyads cooperate at all, and second an allocation stage, in which it is determined how intensely they cooperate. The two stages are estimated separately, but the second stage can only be meaningfully interpreted conditional on the first stage (Clist, Reference Clist2011).

Such Hurdle models are appropriate in the case of sequential decision-making processes, as we have argued our cooperation decision to be. Thus, the first stage is modeled using binary (logit) models and the dependent variable captures whether cooperation exists (1) or not (0). The second stage models the actual number of cooperative ties in a given year only for those dyads that have been selected, i.e. for which a ‘1’ was observed, in the first stage. The models capturing the amount of cooperation in the allocation stage thus have count data (integers), i.e. the number of cooperative ties, as the dependent variables, and are used in this second stage of the modeling process. In addition, as robustness checks, we first run a (full) model combining the two stages, using simply the count variable of cooperative ties as the dependent variable. Second, because of the dyadic and annual nature of the data, there are a large number of zeros in our dataset for the cooperation variable (about 64% of observations). We therefore run an additional robustness check by combining the observations, for all dyads, across the entire time period.Footnote ¹ This means we only have a single entry for each of the dyads. For the cooperation variable, we sum up all cooperation events, i.e. we count how often they cooperate overall. For the trade and GDP variables, we take the mean. When coded in this manner, only about 24% of dyads do not cooperate at all, implying a much lower number of zeros.