Abstract
Coalition formation is considered an important tool to leverage bargaining power in GATT/WTO negotiations. While most of the literature has focused on developing countries, we show that sizable economies are the primary users of coalitions at the GATT/WTO. We also find evidence that middle powers do not exhibit distinctive collectivist behavior at the WTO. There is a linear and strong relationship between countries’ economic power—measured as real GDP—and coalition participation within the GATT/WTO system. We explain these results, presenting evidence that large economies—countries that have greater trade negotiations power—join coalitions more often because they are better equipped to absorb transaction costs and more prepared to deal with the uncertainty of WTO negotiations. We also found a relationship between coalition entry and trade openness, with countries more open to trade joining coalitions more often.
Keywords
Introduction
Our primary goal here is to understand why countries join coalitions in the General Agreement on Tariffs and Trade (GATT) and the World Trade Organization (WTO). Then, we show that coalitions are part of the bargaining strategy of the most powerful countries in the GATT/WTO and that there is a strong relationship between countries’ economic power and coalition participation within the GATT/WTO system. Therefore, although coalitions are a vital tool for weak countries to increase their bargaining power, countries that have higher trade bargaining power are better equipped to absorb transaction costs and more prepared to deal with the uncertainty of the WTO negotiations.
Coalitions are a much-researched field in international political economy, yet the literature on coalitions in the GATT/WTO is still scarce. Besides, most of the previous studies have only analyzed a few cases of bargaining coalitions (Hamilton and Whalley, 1989; Higgott and Cooper, 1990; Narlikar, 2003; Narlikar and Tussie, 2004). There is a consensus that coalitions are one of the few instruments available to weak countries to increase their bargaining power (Braithwaite, 2004; Drahos, 2003; Narlikar, 2003). Hence, the formation of coalitions is a valuable tool to leverage bargaining power, and weaker countries are those most in need of forming coalitions (Keohane, 1969: 295). As Narlikar points out: “United we stand, divided we fall. The reasoning of the weak in their dealings with the strong is simple and direct” (Narlikar, 2003: 2).
Narlikar’s (2003) seminal contribution has two main concerns: (a) to understand coalition formation, and (b) their strategies to reach their goals. Studies of coalition formation matter because the pooling of bargaining resources allows greater negotiating weight and may influence policy outcomes. However, Narlikar is less concerned with the rationale that leads countries to form or opt to participate in coalitions. She assumes that weak countries need to unite to have their demands met in the negotiations, while strong states do not need to perform costly intra-coalition bargaining to get their views taken into account (Narlikar, 2003: 2). Our work about what drives countries to join coalitions complements Narlikar’s research.
Arguably, there is a relationship between the size of countries’ economies and their participation in bargaining coalitions. Unfortunately, there is no consensus on the direction of the effect (positive or negative) or even about the functional form of this relationship (e.g., linear or inverted U-shaped) (Drahos, 2003; Higgott and Cooper, 1990).
While the literature focuses on developing countries, our article shows that coalition formation is a strategy used mainly by powerful economies. First, we argue that the powerful countries—those with higher real GDP—have more interest in building coalitions since they have more interests in different areas of trade negotiations. Second, to negotiate either in coalitions or bilaterally is highly expensive. There is a large set of transaction costs that, together with uncertainty about the final results of the bargaining process, raise insurmountable barriers to weaker countries. Specifically, high transaction costs and uncertainty over outcomes promote greater participation of the relatively powerful economies more capable of paying for coalition formation costs and taking risks. Uncertainty is a characteristic of negotiations, both bilateral and in coalitions. Smaller economies do not have the resources to deal with uncertainties over the outcome of high-cost international negotiations. The problem is not that coalitions are a worse strategy for small economies, but that even a potentially less costly strategy is expensive for them. Some smaller members could benefit from coalition formation, but they are incapable of negotiating collectively. Additionally, we argue that the increasing costs of negotiating in the WTO outweighs the potentially decreased costs of bargaining coalitions. Hence, although coalitions are essential for weaker countries, powerful nations take advantage of coalitions more often.
Our article shows that smaller economies join fewer coalitions, middle powers cooperate more, and big economies participate the most. Other findings are that countries more open to trade participate more in coalitions and democracies do not join coalitions more than dictatorships. Additionally, large and complex economies form coalitions more often because they are better equipped to absorb transaction costs and to deal with the uncertainties of trade negotiations.
Coalitions in the GATT/WTO system
While there were few coalitions in GATT’s early years, the number of coalitions increased substantially from the 1980s (Narlikar, 2003: 34). Arguably, GATT was a “rich man’s club,” where tariff protection and United States (US) domination regarding international trade were the rules. Despite their relatively large proportion, developing countries maintained a low profile in the GATT (Narlikar, 2003: 35).
A few years before the Uruguay Round, there were a number of changes in world trade negotiations. The Reagan administration (1981–1989) changed its trade policies regarding US trade partners (Evans, 1989), and US trade partners were pushed to avoid domestic protectionism. At the same time, the US started to advocate deep reforms in the international trade system. The Uruguay Round (1986–1994) was, in large part, pushed forward by pressures from the US, the WTO being the most critical result of the trade talks.
Affected countries usually react to outside pressures. Many times, they form coalitions. For this reason, coalitions increasingly became active in the Uruguay Round. During the negotiations, most coalitions wanted to block proposals pushed forward by the US and the European Community (EC) (Narlikar, 2003).
The G-10 was a visible developing country coalition, and Brazil and India played a prominent role in it. The G-10 was composed of countries with diverse interests, and this group could not maintain its unity to the end of the Uruguay Round. The pre-launch phase of the Uruguay Round also saw the involvement of developing countries in the Café au Lait group—a “crossover” coalition combining developed and developing countries (Narlikar, 2003: 39).
Developed nations also formed a coalition. The Quad (Canada, EC, Japan, and the US) pushed their liberalization agenda in the Uruguay Round while being reluctant to push the opening of agricultural markets in the developed world. During the Uruguay Round, the EC acted as if it were a coalition in the agricultural negotiations in opposition to the Cairns Group—formed by competitive agricultural exporting countries.
After the failure of the Seattle Ministerial Conference in 1999, there was increased demand to change the WTO. Public riots and demonstrations raised concerns about the internal legitimacy of the WTO. Representatives from the developing world complained about their marginalization from crucial decision-making processes (Levi and Murphy, 2006; Schott, 2000). As in the Uruguay Round, developing countries saw that they could achieve more by joining forces through coalitions (Narlikar et al., 2012: 172).
During the Doha Round, developing country coalitions were successful in defending a narrow or issue-based agenda (Narlikar, 2003). With the Doha Declaration in 2001, they were successful in arguing that the lives of individuals around the world—especially those living with HIV/AIDS—were more important than the intellectual property rights of the pharmaceutical industry (Odell, 2009; Sell and Prakash, 2004). Again, developing countries advocating this agenda were not the economically weakest countries in the world; they were at least regional powers. Brazil, India, and South Africa were the foremost leaders and winners of the Doha Declaration (Odell, 2009; Sell and Prakash, 2004).
In this article, we test whether major economies are more likely to enter coalitions than weak or middle powers in the GATT/WTO negotiations. We propose that larger economies join coalitions more often for two different reasons.
First, states with the greatest number of interests to protect and promote tend to participate in a higher number of coalitions; smaller economies have no interest in most of the negotiation themes, so they have fewer incentives to join coalitions.
Second, we explain the rationale for participating in coalitions mainly as a transaction cost problem (Williamson, 1979). The economically powerful countries are better equipped to pay the transaction costs involved in bargaining together in the GATT/WTO. Specifically, entering into a coalition involves considerable cost, since the negotiations within the GATT/WTO system are complex and comprehensive. Countries have, for instance, to maintain diplomats in Geneva. They also have to build the necessary expertise and an adequate infrastructure to negotiate. These costs are prohibitive for countries with scarce resources. In sum, the causal mechanism proposed is the following: higher GDP (greater diversity of interests) → higher ability to pay transaction costs → higher participation in coalitions. 1
The WTO literature documents many of these transaction costs. Tussie (2009) argues that developing countries have fewer resources to gather and analyze information. Thus, coalitions work as a division of labor, with stronger members providing information for their trade partners in exchange for their support. However, it is not clear by how much coalitions can decrease the transaction costs for the weak. Baccini (2014) argues that as the length of negotiations increases, transaction costs grow accordingly. The weak suffer most of the burden when negotiating in GATT/WTO becomes more expensive.
The number of countries involved in a negotiation is also related to transaction costs (Olson, 1965), as more countries with different trade preferences need to be satisfied with the final agreement. The negotiations have become more and more involved in each round as the GATT/WTO system has increased the number of countries involved in negotiations and as these countries have distinct interests (Hoekman and Kostecki, 2010).
Larger economies have diversified trade interests and stronger capabilities to cover expansive trade costs. As a result, they either obtain material gains or shape international rules and practices more often than weaker countries. Notwithstanding, smaller economies have less diversified interests and are less likely to reap monetary benefits directly from the negotiations. Collective negotiations are not only costly but also risky. At the outset, no one knows what the results of trade rounds will be or the unintended consequences of trade agreements. Therefore, larger economies negotiate more effectively because they have enough resources to pay substantial costs and take risks.
Cooperation in the GATT/WTO system
Some practitioners and scholars argue that the GATT/WTO empowers developing countries in negotiations with developed nations. Former Brazilian Minister of Foreign Affairs and former Ambassador to the WTO Celso Lafer (1998) argues that the WTO offers more opportunities to middle-income countries than bilateral negotiations with powerful states. Other authors suggest that coalitions are an effective strategy for weaker countries bargaining in the international trading system (Braithwaite, 2004; Drahos, 2003).
Keohane (1969: 295) proposes a typology based on the role that each type of country has in the international system. The first type is the imperial power in a unipolar system or two major powers in a bipolar system. They are large states that play a central role in shaping the international system by themselves. Second, some states do not dominate the system individually but can significantly influence it through unilateral or multilateral action. Third, middle powers are not able to affect the system in isolation but may be relevant when they collaborate. Finally, there are small states that cannot influence the international system, except in large groups.
Following Keohane, some authors argue that middle powers exert an influence in the international arena only through collective action and, therefore, they suggest that middle powers exhibit collectivist behavior (Higgott and Cooper, 1990). Hence, middle powers should join coalitions and international institutions more often than either weak or powerful countries. While weak countries are mostly irrelevant, powerful countries are capable of acting unilaterally without having to make concessions to other coalition members. Nonetheless, Cepaluni et al. (2012) show that larger coalitions are more successful in reaching their goals because the GATT/WTO is a consensus-based institution and the trade system informally penalizes countries that isolate themselves in negotiations.
According to Diego-Fernández (2008: 435), coalitions matter for at least three reasons: (a) working in groups increases the possibility of advancing one’s position; (b) it helps resource pooling; and (c) it increases the possibility of being represented in green-room-type of processes. Increasing the chances to advance one’s position is probably the main reason for building coalitions. In this respect, (a) a proposal has greater legitimacy if presented by many countries, and (b) it puts more pressure on the other side to accept it, scale down its demands, or to trade favors. While large economies have more room to maneuver than smaller ones, this does not allow them to do what they want. Support from other countries gives them legitimacy, and legitimacy helps even powerful countries to make a better deal (Pelc, 2010).
Here, we assume that countries and coalitions are relatively consistent in their behavior when pursuing their broadly defined interests. Countries participate in the WTO negotiations to “reap monetary benefits,” minimize negative impacts on their economies, or even pursue “non-trade public policy goals”—see, for instance, the pro-human rights advocacy of the TRIPS/Public Health group. Therefore, the assumption that states and coalitions seek their interests does not rule out the possibility that they may espouse normative principles. We shy away from inferring the real drivers behind countries’ or coalitions’ actions. Normative principles and strategic actions converge in many situations, and it is hard to disentangle these different motivations.
Another critical benefit of forming coalitions is the pooling of resources in monetary terms or in the division of labor and assistance for meetings. As negotiations become more intense or technical, the pooling of resources becomes more relevant, and many delegations are not in a position to react to every single proposal because of their limited resources. However, resource pooling is not only crucial for weak states: “participating in a coalition may allow a country to be more active in issues in which its interest may not be that fundamental, but is important enough to keep an eye on” (Diego-Fernández, 2008: 440). Another resource-related benefit of participating in coalitions relates to information gathering. Since the WTO Secretariat is small, most of the research has to be done by members themselves: participating in a coalition where members can gather useful information, draft proposals and statements, organize meetings, and contact the WTO Secretariat becomes a valuable asset.
The green room is the WTO Director-General’s conference room, and it cannot hold more than a few dozen people (Diego-Fernández, 2008: 443). Many delegations that are not invited to the green room want at least to be represented by like-minded delegations. As Wolfe (2015: 15–16) argues, the coordinator of coalitions represents most members in green room meetings—usually an economically powerful developing or developed country, such as Brazil, Indonesia, US, EU, Japan, Australia, or China.
Our argument indicates the prevalence of the use of coalitions by the most economically powerful countries. Larger economies are sensitive to a broader range of issues since they are more diversified and complex. Besides, they have a stronger capacity to cope with the uncertainties of lengthy negotiations. Negotiating in the GATT/WTO requires paying high transaction costs and participating in coalitions is a form of reducing them—for instance, due to a division of labor between member states. Nevertheless, the negotiations within the GATT/WTO system are complex and comprehensive, and there are remaining costs involved that are more easily paid by the powerful.
The increase over the last decades in the number of actors involved in the negotiations and of the issues on the table makes the transaction costs much larger, ceteris paribus, reducing incentives for smaller countries to participate in coalitions. An implication of the argument is that the predominance of larger economies in the use of coalitions should be greater within WTO negotiations than at the GATT, as negotiations have become increasingly complex and include a higher number of themes and actors. Specifically, the increasing transaction costs of overall trade negotiations—especially after the creation of the WTO and the consequent expansion and complexification of the trade system—are outweighing the potentially decreased costs of negotiating through coalitions. 2
Alternatively, US trade threats can explain the prevalence of powerful countries in coalitions. Smaller economies would join coalitions, but they fear the consequences of doing so, as the US might sanction them. The Special 301 legislation allows the US Trade Representative (USTR) to identify countries that deny adequate protection for intellectual property rights or deny fair and equitable market access to US exports. During the Uruguay Round, for example, almost all the G-10 countries leading the resistance to the US agenda were targeted by the US under its 301 legislation. Therefore, fear of reprisal might enter a smaller country’s coalitional calculus. 3
Data and variables
Dependent variables
Our primary source of information is the WTO website. The link “Groups in Negotiation” has information on groups participating in the negotiations such as coalition names, member-country names, issues under negotiation, coalitions’ websites, and the nature of the coalitions (e.g., customs union, regional, or broad interests). Arguably, this is the best data set available since it is an official and specialized source. However, we supplemented and checked the WTO data with information from the literature.
Narlikar (2003: 29) says that there is a disagreement on the definition of coalitions, and “alliances, coalitions, and alignments are often used interchangeably.” Therefore, we follow her broad definition of coalition “as an umbrella term that encompasses different types of alignment categories.” This definition excludes incidental policy convergence among states because a coalition only exists when its members are involved in conscious coordination, irrespective of their political purposes. Similarly, the section “Groups in Negotiation” states: “these groups often speak with one voice using a single coordinator or negotiating team.”
The chief information used to build our dependent variable consists in coalition names, their members, the year of coalition formation and their scope—see Online Appendix. We have data on coalitions from 1982 to 2008. Countries became part of our data set from the moment they entered the GATT/WTO system. 4
To mitigate problems related to measurement error in our dependent variable, we measure coalition entry in two ways: (a) all coalitions that have participated at the GATT/WTO (coaltrade), and (b) coalitions created only for the GATT/WTO negotiations (coalwto). For coalwto, we exclude coalitions that negotiate both inside and outside the GATT/WTO system—for example, the African, Caribbean, and Pacific Group, APEC, CARICOM and regional trade alliances such as European Union, Mercosur, and NAFTA. We classified “regional integration” as non-WTO coalitions. Our analyses with two different dependent variables (coaltrade and coalwto) are similar. The results were slightly weaker in the second measurement (coalwto), since we increased the number of zeros in our dependent variable. 5
Independent and control variables
Economic power is our primary explanatory variable, and we measure it is as the log of real GDP. We use GDP as a proxy for countries’ relative power, because countries’ GDP reflect many economic resources related to the abstract notion of power (Wood, 1988). We collected data on GDP from the Penn World Table (PWT 6.3). A country is a major power, middle power, or small power according to its relative position in the global hierarchy of power in a given year. In this sense, our data and analysis in a panel form capture the historical changes in countries’ status during the time frame. Operationally, countries fall into one of these power categories according to their place in the distribution of our data. Countries with the highest GDP are at the top of the distribution, countries with intermediate GDP are in the middle of the distribution, and countries with smaller GDP are at the bottom of the distribution.
Other interesting findings concern the relationship between coalition entry and three possible confounding variables: trade openness, level of development, and political regimes. We collected both trade openness and the level of development from PWT 6.3: the sum of import and export flows divided by GDP for each country (annually) measures the former, and the real GDP per capita measures the latter. We use two measures of democracy: a dichotomous (Alvarez et al., 1996; Cheibub et al., 2010), and a continuous index (Polity IV). The Polity IV democracy score varies from −10 (complete dictatorship) to +10 (full democracy).
We used a broad set of control variables. These variables are: the size of government (government expenditure/GDP) and the growth rate of GDP per capita from PWT 6.3; and data on battle-related deaths, electric power consumption (kWh per capita), exports (% of GDP) and imports (% of GDP) from World Development Indicators (WDI). Most of these controls are statistically insignificant. 6
We use the size of members’ delegation in Geneva (delegation) as a proxy for the ability to pay the transaction costs that countries incur when negotiating at the GATT/WTO. Arguably, the number of delegates in Geneva should correlate with coalition activity. Therefore, we expect that larger states have more coalition activity since they have, on average, large delegations. The idea of transaction costs is difficult to measure accurately, more so the ability to pay them. Following Sattler and Bernauer (2011), we employ the number of delegates in Geneva as a mediating variable between economic power and coalition entry. The number of delegates is, then, a proxy of the bureaucratic capacity of countries bargaining at the GATT/WTO (Bown, 2005; Guzman and Simmons, 2005).
In other words, if a country has a sizeable bureaucratic capacity, it is also more likely to pay the high costs involved in negotiating in GATT/WTO. The WTO used to publish an annual phone directory for internal use, which listed each member’s delegates in Geneva (Busch et al., 2009: 562). Unfortunately, this data is only available from 1994 to 2004. The number of delegates to the WTO in Geneva has been used in previous studies and captures the resources invested in bureaucratic capacity more directly than other measures, such as GDP per capita (Sattler and Bernauer, 2011: 10).
Lastly, we collected data on all US threats under its Special 301 legislation. The International Intellectual Property Alliance has compiled country-level data on this since 1989. Therefore, we created a variable indicating whether the US threatened these countries or not from 1989 to 2008. The most severe threat is investigation under the 301 legislation because this could result in short-term trade sanctions. However, the US also puts countries on a “Priority Watch List” and “Watch List,” signaling that the US might investigate these countries in the future.
We present all the variables listed in Table 1.
Descriptive statistics for the main and control variables.
Methodology
We employ both panel and cross-section (pooled) non-linear models. The goals of employing this diversified set of statistical models are: (a) to test the robustness of our empirical findings and theoretical mechanisms, and (b) to explore the functional form of the relationship between our dependent variable and independent variables. In panel form, we present results from ordinary least squares (OLS), logistic, and negative binomial regressions. In cross-sectional form, we present graphical results from the non-linear generalized additive models (GAMs) that are better suited to address the functional form of the hypothesis. To explore our proposed causal mechanism, we present a mediation analysis (Imai et al., 2011).
Panel data techniques offer a series of advantages over cross-section analyses, such as increasing the estimation accuracy and controlling for unobserved heterogeneity. Here we want not only to verify a relationship between coalition entry and our independent variables but also to visualize their functional form. GAMs allow the relationship to vary locally over the range of dependent and independent variables, unlike linear models that assume a globally linear relationship between dependent and independent variables (Beck and Jackman, 1998; Hastie and Tibshirani, 1990). We present the results graphically and in probability form as it facilitates interpretation.
Mediation analysis examines causal mechanisms that underlie a relationship between an independent and a dependent variable through the inclusion of a third mediation variable. Rather than suppose a direct causal relationship, a causal mediation model shows that an independent variable influences a dependent one indirectly through a mediator. The indirect effect (average causal mediation effect—ACME) represents the expected causal mechanism, and the direct effect (average direct effect—ADE) represents all the other mechanisms. The total effect (TE) is the sum of the ACME and the ADE (Imai et al., 2011). We employ a two-stage mediation method, as proposed by Imai et al. (2011). We use multilevel linear regressions to estimate the effect of the log of GDP on the number of delegates and to estimate the effect of the log of GDP mediated by delegates on coalition entry. We re-estimated the same model but using data on US trade threats. At both models, the intercept varies by countries.
Results
Panel data
Economic power increases the likelihood of a country joining a coalition across the two different ways we measure our dependent variable. Table 2 presents generalized linear models (GLM) in which coalition entry includes both all coalitions that participate in the GATT/WTO (Panel A) and the ones created for negotiating at the GATT/WTO (Panel B). Columns 1–6 present OLS, logit, and negative binomial models with all regressors included. Columns 1–3 have as a covariate the dichotomous measure of democracy and columns 4–6 the continuous one. 7
OLS, logit, and negative binomials regressions with many controls.
Full linear (OLS), logistic, and negative binomial regression models with many control variables. All specifications include country fixed effects and have bootstrap standard errors clustered at the country level. Bootstrap standard errors in parenthesis. +p < :1; *p < :05; **p < :01; ***p < :001.
We tested if the size of the government, the growth rate of GDP per capita, battle-related deaths, electric power consumption (kWh per capita), exports (% of GDP), and imports (% of GDP) affect our two measures of coalition entry. None of these variables has a significant or a substantive effect in the six models presented in Table 2. Democracy is also not a determinant of coalition entry. However, we will keep analyzing this variable closely for theoretical reasons, as many studies claim that democratic regimes are more likely than dictatorships to engage in free-trade agreements (Dai, 2002; Mansfield et al., 2000).
At Table 2, economic power—or real GDP—remains statistically significant in all model specifications. The same happens for trade openness. Level of development is also a significant variable at the logit and negative binomial models, but it loses significance in the OLS model. 8
Columns 1 and 3 (Panel A) show that, ceteris paribus, a country that increases its real GDP by 1% enters coalitions one time more often. For Panel B the effect is smaller, but it is also close to 1. Columns 2 and 5 show that an increase in real GDP increases the log-odds of a country entering a coalition. The effects of the negative binomial models are also relevant in substantive terms.
Hence, our results show that GDP is the most crucial variable. It presents a substantial effect, and it usually reaches high statistical significance. Trade openness is also a significant predictor of coalition entry, and the same happens for the level of development. Finally, both measures of democracy are weak predictors of a country joining a coalition. They are rarely statistically significant, and their effects are small.
Non-linear models with GAM
GAM allows estimation of smooth functions rather than linear coefficients for each independent variable. Here we analyze the results from Figure 1(a)–(d) that only take into account coalitions built for negotiations within the GATT/WTO (coalwto). In all graphs, we display probabilities at the y-axis scale.

Generalized additive models (GAMs)—(coalwto). (a) Marginal effects of the log of real GDP on coalition entry (coalwto). (b) Marginal effects of the log of trade openness on coalition entry (coalwto). (c) Marginal effects of democracy (Polity IV) on coalition entry (coalwto). (d) Marginal effects of the log of real GDP per capita on coalition entry (coalwto). Effects of economic power (1a), trade openness (1b), polity IV (1c), and level of development (1d) on coalition entry (coalwto). Each graph shows the estimated value of the degrees of freedom for each smooth function. A 95% confidence interval for the cubic smoothing splines is shaded.
Figure 1a shows the relationship between the log of GDP (economic power) and the probability of coalition entry. As we can see, the relation between economic power and participation in coalitions is quasi-linear, even adopting a flexible statistical model. Keeping all other variables constant, the higher the economic power, the more likely a country will join a coalition. The probability that one of the economically weakest countries in our data set joins a trade coalition is less than 20%. Middle powers present a probability of around 50%. Economically powerful countries join coalitions with more than 80% probability.
Figure 1b presents the relationship between the log of trade openness and coalition entry. Again, even using GAM, we find a linear effect. Figure 1c shows a local linear effect on coalition entry when the most closed dictatorships open their political regimes. After that, there is almost no effect of higher levels of democratization on coalition entry. These results suggest that only dictatorships that become increasingly democratic cooperate more. After the initial liberalization, more democracy does not translate into membership in coalitions. Figure 1d presents a different perspective. There is no linear pattern and no substantive difference between the least and the most developed countries. The results are not robust, considering that we observe large confidence intervals. 9
Mediation analysis and causal mechanisms
The mediation analysis developed here has two steps. First, we use linear multilevel regression to estimate the effect of the log of GDP on the number of delegates in Geneva. Second, we use another linear multilevel regression to estimate the effect of the log of GDP on coalition entry mediated by delegates and controlling for democracy, trade openness, and level of development. As we mentioned before, we use a shorter time horizon as there is no data available on the number of delegates for a more extended period.
Our mediation models for transactional costs in Table 3 have 1451 observations and present the results for two dependent variables: coaltrade and coalwto. We present the ACME, ADE, and TE for both dependent variables. For coaltrade, the ACME is 0.046 and is significant. The ADE decreases coalition entry by −0.001, and it is not statistically significant. Total effects add up to 0.045. Our proposed mechanism (ACME) corresponds to more than the entire total effect. For coalwto, the ACME is just 0 .043, and it is statistically different from zero. The ADE of the log of GDP increases coalition entry by 0.018 but, again, it is not statistically significant. Total effects add up to 0.061, and it is highly significant. Our proposed mechanism (ACME) corresponds to a proportion of 70% of the total effect. Hence, with an increase from 15% to 20% on the country level of GDP, there is, on average, one more entry on coalitions.
Multilevel linear models for mediation analysis.
Multilevel linear models for mediation analysis. Controls include democracy, trade openness, and level of development. All models present bootstrap standard errors. +p < :1; *p < :05; **p < :01; ***p < :001.
The ACME for coaltrade and coalwto indicates the significance of our proposed mechanism. The number of delegates in Geneva is only a proxy of some of the transaction costs for a country when negotiating collectively at the WTO. However, both estimates of ACME and ADE are statistically significant even under circumstances that are likely to underestimate the real effect. 10
Lastly, fear of reprisals enters into countries’ coalitional calculus. However, the results—for both coaltrade and coalwto—do not challenge the transaction cost mechanism. The mediation models in Table 3 show that US threats directly influence countries’ decision to join a coalition. Nonetheless, the fear of reprisal is not mediated by GDP, unlike the number of diplomats in Geneva, a proxy of bureaucratic capacity. In other words, fear of reprisal has a direct effect on coalitional calculus. Nonetheless, the impact of US threats is independent of countries’ economic sizes.
Conclusion
We find strong and robust evidence that larger economies and countries more open to trade tend to join coalitions more often than smaller ones. There is a linear and robust relationship between GDP and coalition participation within the GATT/WTO system. The magnitude of the size effect is substantial. We obtain similar results with different statistical methods, several model specifications, and different ways of measuring our dependent variable. Also, the relationship between GDP and coalition entry remains almost linear when we use flexible regression models. Therefore, middle powers do not present a distinctive collectivist behavior if we compare them to large economies.
Countries more open to trade also participate in coalitions more often in contrast with closed economies. Surprisingly, we find little evidence that democracies join coalitions less than dictatorships. As should be clear, we only select countries that are members of the GATT/WTO, and we cannot generalize our findings on political regimes to other domains. As a caveat, the positive sign of the panel regression coefficients and the local linearity in GAM models do not discard the democratic effect altogether, since it is possible that there is substantial heterogeneity in the democratic effect on trade cooperation. Nonetheless, the nonlinear relationship shows that the “democratic peace theory” is more complicated than a linear relationship assumes (Beck and Jackman, 1998).
Our results regarding larger economies and trade openness are comparable to Sattler and Bernauer’s (2011) findings. For them, larger economies get involved more in trade disputes because of the greater diversification of their economies. Moreover, their greater market size also makes them targets that are more attractive in these negotiations. The transaction costs literature helps us make sense of our main finding. Economically powerful countries tend to participate more in coalitions because they have enough resources to pay for the high transaction costs of collectively participating in lengthy, complicated negotiations with many countries. Bargaining in the GATT/WTO is also a risky business. As coalitions are groups competing in negotiations with other countries and coalitions, victory is not guaranteed. Economic gains from participating in the GATT/WTO are also debatable and probably favor wealthy developed countries, thereby discouraging weaker economies from bargaining collectively in the GATT/WTO negotiations.
Finally, Olson (1965) remarked that there is a tendency to the “exploitation” of the great by the small in the payment of the costs for the provision of collective benefits. Powerful countries willing to bear the costs of group formation can, however, overcome obstacles to collective action. It remains for future research to explore how the political and economic relations between countries influence the formation of coalitions.
Supplemental Material
Supplemental_material,_for_'United_we_stand_and_divided_we_fall – Supplemental material for United we stand and divided we fall: Coalitions in the GATT/WTO negotiations
Supplemental material, Supplemental_material,_for_'United_we_stand_and_divided_we_fall for United we stand and divided we fall: Coalitions in the GATT/WTO negotiations by Gabriel Cepaluni and Ivan Filipe Fernandes in International Political Science Review
Footnotes
Acknowledgements
We are grateful to Silvia Ferrari, Athos Damiani, and Julio Trecenti, who advised us on statistical matters in earlier stages of the project. Thomas Sattler shared his data on the number of delegates in Geneva. Matthew Winters, Diego Correa, and Robert Wolfe provided us with generous suggestions. We also thank participants in the seminars of the Department of Statistics at the University of São Paulo (USP), at the 23rd International Political Science Association (IPSA) World Congress of Political Science, and at the Grupo de Economia Política (GEP) from the IRI/USP (Flávio Pinheiro, Feliciano Guimarães, Gustavo Araújo, Pedro Feliú, and Rafael Magalhães). Finally, we are grateful to the anonymous reviewers and editors for their constructive comments.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
Supplemental material
Supplemental material for this article is available online.
Notes
Author biographies
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
