Bargaining Dynamics Across Dispute Types

Since 1995, the WTO has provided its member states with a highly structured process for settling international trade disputes. The WTO's dispute settlement mechanism is used when one or more member countries believe that trade practices by other WTO members (for example, tariffs, import restrictions, export subsidies) violate existing trade rules. Whereas the dispute settlement procedure of the General Agreement on Tariffs and Trade (GATT) system – the predecessor of the WTO until 1995 – left accused countries with many ways to delay or stall dispute settlement proceedings, the WTO's DSM provides for a multistage process with clearly defined time lines and virtually no opportunities for accused countries to block the process.Footnote ¹³

The WTO's DSM is crucial for the functioning of the global trading system. In the WTO's own words:

Empirical research confirms that the WTO DSM plays a major role in promoting co-operation and trade.Footnote ¹⁵ However, different mechanisms can lead to this outcome. The statement above appears to allude to both the enforcement and managerial perspectives. But it does not tell us exactly what we should look for in real WTO dispute settlements in order to better understand the mechanisms through which dispute settlement relates to international co-operation.

In the next two sections, we elaborate on both the enforcement (political stakes) and managerial (complexity) perspectives to derive distinct empirical implications to be tested in the subsequent sections. In particular, we examine theoretically how dispute resolution dynamics should look if WTO dispute settlement primarily addresses enforcement or managerial concerns. We argue that enforcement problems mostly arise when high domestic political stakes are involved. Managerial difficulties in international trading relations generally originate from a high degree of complexity and uncertainty regarding the exact rights and obligations associated with international trade rules. As we will see below, the two perspectives have fundamentally different implications for expected dispute settlement dynamics and outcomes. Furthermore, the two distinct theoretical arguments, which we derive from the enforcement and managerial perspectives, distinguish between two scenarios. First, we discuss how both domestic political stakes and complexity affect dispute escalation when governments behave non-strategically. Secondly, we relax this assumption and argue how the dynamics of dispute escalation should look when governments behave strategically; that is, when they are forward looking in the sense that they take their opponents’ possible future actions into account.

WTO Dispute Settlement as Interest Group Politics

According to the enforcement perspective, many international co-operative efforts have the structure of mixed-motive games.Footnote ¹⁶ Countries are interested in the benefits of co-operation, but they also have incentives to shirk or cheat. This is mainly the case when implementing an international commitment is costly or if free riding is associated with large benefits. In the logic of the enforcement perspective, dispute settlement is considered part of the monitoring and enforcement process that protects co-operating countries from non-compliance by others and deters countries from violating their commitments in the first place. Dispute settlement thus promotes co-operation by deterring non-compliance, since it increases the costs associated with non-cooperation.Footnote ¹⁷ When do countries engage in non-compliance, even though it increases the risk of a costly dispute settlement process?

Even a casual look at news media reporting on WTO trade disputes quickly reveals that some disputes are associated with great political stakes, in the sense that they attract a great deal of attention and are associated with extensive and controversial public debates in the countries involved. Examples include EU-US disputes over the use of growth hormones in beef production and genetically modified (GM) crops. Strong political attention, in turn, limits the freedom of policy makers to strike bargains – in particular, bargains that are perceived as unfavorable by domestic producers, interest groups or the general domestic audience.Footnote ¹⁸ Other disputes are hardly noticed by the news media and the domestic public. Examples include a dispute between the United States and Indonesia on ‘Measures Affecting the Production and Sale of Clove Cigarettes’ and a dispute between Chile and Mexico over ‘Measures Affecting the Import of Matches’. In such cases, policy makers have more room to maneuver and negotiate with the opponent(s) in a dispute and make concessions.

The domestic political stakes associated with a disputed trade policy affect governments’ motivations and thus their willingness and ability to co-operate with trading partners. Defendant governments are more likely to resist demands from complainants in disputes with greater domestic political stakes, because domestic constituencies will punish them for backing down and exposing domestic industries to more economic competition if the disputed trade-restricting policy is weakened or removed. Similarly, to the extent that a dispute is associated with greater political stakes, the complainant government faces higher domestic political costs if it gives in and accepts that the defendant makes only some concessions or does not lift the disputed trade-inhibiting policy at all.

To identify the political stakes of a dispute ex ante – that is, independently of the dynamics of a trade dispute once it has been launched – the existing literature has focused on economic sector characteristics because disputes concerning some sectors attract stronger political attention than others. Davis and Shirato, for instance, show that sector-specific factors influence government decisions during trade conflicts.Footnote ¹⁹ The magnitude of domestic political stakes in any given sector should then be a function of the importance of that sector for the respective country.Footnote ²⁰

However the political stakes may also depend on other factors, notably each sector's ability to organize politically, form alliances with other stakeholders, publicize the dispute, and lobby for or against trade-restricting policies or practices.Footnote ²¹ In the case of the EU-US dispute over GM crops, a heterogeneous but very large and well-organized political coalition of European farmers, consumer organizations, environmental groups and left parties has emerged. Although the importance of the issue in purely economic terms is quite modest, the high political stakes associated with the dispute have made it virtually impossible for the EU and its member states to accommodate US requests to open Europe's market to GM crops.Footnote ²²

Under conditions of higher domestic political stakes, it should thus be less likely that a defendant government will give in and accept an adverse ruling. In general, a formal WTO ruling increases political pressure on the defendant country, if it is found guilty, to remove the disputed trade restrictions because failure to comply with WTO rules would adversely affect the defendant country's international reputation.Footnote ²³ Such negative reputation effects could, in turn, negatively affect the defendant country's ability to reach advantageous trade or other agreements in the future. These negative reputation effects, however, could be offset by domestic political pressure when a very politically sensitive sector is affected. Hence we should expect trade disputes to be more prone to escalation if domestic political stakes are high.

So far, we have not considered that governments behave strategically when settling disputes. Strategic behavior is important because the exact implications of the enforcement perspective differ when we assume that governments are forward looking (that is, that they behave strategically). When governments decide whether or not to request the establishment of a dispute settlement panel, which constitutes a major step towards dispute escalation in the WTO, they are likely to pay close attention to the other country's expected behavior in subsequent phases of the process.Footnote ²⁴ We therefore need to discuss how the effect of domestic political stakes on the likelihood of escalation may vary at different stages of the dispute settlement process. This is necessary because governments not only consider immediate gains and losses when deciding what to do at a particular point in the dispute process, but also base their decision on what they expect to happen in the future.

We submit that the escalation dynamics – that is, the likelihood of dispute escalation at a given stage of the dispute settlement process – of strategic governments are different than if governments act ‘naively’ and fail to take into account their opponent's future responses. When forward-looking governments consider whether or not to request a panel ruling in a given dispute, they know that the domestic political stakes associated with the dispute will become more relevant if the conflict escalates beyond the consultation phase of the WTO dispute process. This is because with every additional stage of the DSM process, the dispute is likely to receive more attention from the domestic public. Once a WTO panel has been established and a ruling is issued, the case becomes more visible to a larger public and audience costs become more important as well.Footnote ²⁵ Forward-looking governments know that a dispute with high political stakes will become more difficult to control and settle when it escalates to the panel stage.

Anticipating such loss of control, governments will think twice before asking for a panel in the first place, for it will become more and more difficult to settle a dispute with greater political stakes the further it has moved through the WTO dispute process. A forward-looking government will therefore try to solve potentially high-stakes disputes at the consultation phase, given the reduced likelihood of solving such disputes once they have passed beyond the first phase of the dispute settlement process. In other words, strategic, forward-looking behavior mediates the effect of domestic political stakes at the first stage by reducing the likelihood of escalation to the panel stage (but not at subsequent stages).Footnote ²⁶ Two implications follow from these considerations. First, if observable indicators suggest that a dispute is associated with higher domestic political stakes, we should not see a substantially higher escalation risk from the consultation to the panel phase, compared to disputes in which observable indicators suggest lower domestic political stakes. Secondly, the escalation risk of such disputes should be substantially higher once they have reached the panel stage. This leads to the following hypothesis:

The WTO DSM as A Complexity-Reducing Device

In contrast to the enforcement perspective, the managerial perspective views dispute settlement procedures primarily as information and rule clarification devices. International rules are sometimes ambiguous or so complicated that governments might violate them inadvertently because they do not fully understand whether their current policies and practices are in line with those rules. Hence according to the managerial perspective, dispute settlement is part of a compliance management process under conditions of incomplete contracts.Footnote ²⁷ The basic modus operandi of dispute settlement mechanisms, according to this perspective, pertains to helping states implement their commitments by reducing complexity and providing information, rather than detecting and punishing violations of international rules. Correspondingly, the decisive characteristic of trade conflicts is dispute complexity, rather than domestic political stakes.

Complexity pertains to the content of a dispute and reflects the varying degrees of uncertainty about the legitimacy of disputed trade restrictions. Some discriminating trade instruments and policies may be obviously inconsistent with WTO rules, but the WTO conformity of other trade-inhibiting measures may be difficult to assess. Disputes over tariffs or quotas, for example, mostly deal with non-complex issues because countries usually understand whether or not a given tariff violates WTO rules. In contrast, disputes over environment, health and safety issues can be very complex, since it is often unclear whether environmental, health or safety concerns are sufficient to justify a trade barrier according to WTO rules.

Similar to high domestic political stakes, a high level of complexity is likely to hamper the prospects for co-operation in an international system without central authority, albeit in different ways. In the context of WTO dispute settlement, more complex disputes can be viewed as situations with less complete information. All parties incur transaction costs in such situations, either because WTO rules are vague or ambiguous or because the disputing parties hold different interpretations of which rules are applicable and how those rules should be interpreted.Footnote ²⁸ Under such conditions, both the complainant and the defendant will find it difficult to establish whose interpretation of the relevant WTO rules is correct.

Dispute complexity increases the likelihood that countries will disagree about their specific rights and obligations under WTO law. If there is a high degree of uncertainty, governments will often have inconsistent expectations about winning a conflict.Footnote ²⁹ When one or both sides in a dispute are overly optimistic about being right, early dispute settlement is difficult to achieve: the bargaining range (that is, the range of potential agreements that both parties should prefer to an escalation) decreases or even disappears when parties have conflicting and high expectations that they will win the case.Footnote ³⁰

The dispute between the United States and Japan over measures restricting the import of apples (DS No. 245, a similar dispute has also occurred between Australia and New Zealand) is such an example of variation in dispute complexity. This dispute, which may appear rather trivial at first sight, is relatively complex. It concerns the rather complicated question of whether Japan's health and safety regulations for the importation of apples are consistent with the WTO Agreement on the Application of Sanitary and Phytosanitary Measures (SPS Agreement) and the Agreement on Agriculture. In this case, dispute settlement needs to clarify whether (from scientific, economic and other viewpoints) Japan's practices meet particular environmental and health objectives – in particular to avoid transmission of the fire blight bacterium. A settlement also needs to clarify how those objectives should be weighed against losses in the exporting country.

In comparison, the WTO dispute on cigarette taxes, which was initiated by Chile against Peru in 2001 (DS No. 227), is arguably less complex. Peru's obligation to end its discriminatory policies was quite obvious to most trade specialists. In contrast to the apples dispute, there was virtually no legal uncertainty about whether the defendant country's trade-restricting policies were compatible with the relevant WTO rules. The principal purpose of legal action by Chile against Peru was not to obtain clarification on how WTO rules should be interpreted in this case, but to increase pressure on Peru in order to obtain the desired concessions.

When WTO member countries face a dispute over a complex issue, the role of dispute settlement differs fundamentally from what the enforcement perspective assumes. As illustrated by the US-Japan apples case, WTO dispute settlement serves primarily to reduce complexity and clarify rules.Footnote ³¹ This second (managerial) perspective subscribes to key assumptions of the legalization literature in international relations, which views international courts as institutions that help countries reduce uncertainty and transaction costs and clarify legal obligations.Footnote ³² This view, in turn, is based on the more general argument that international institutions help countries solve international co-operation problems by providing information, making rules explicit and pointing out deviations from officially recognized rules. WTO disputes, viewed from this perspective, escalate because international contracts are incomplete, there is considerable uncertainty over whether the disputed trade measure is legal and both parties believe that they are correctly interpreting the law.

The main instruments for establishing an authoritative interpretation of applicable trade law under conditions of uncertainty are the WTO's panel and appellate body.Footnote ³³ According to the managerial perspective, settling more complex disputes during the consultation phase should be less likely. Such disputes are more likely to escalate to the panel or appellate body stage because uncertainty about whether the disputed trade measures are legal under WTO law motivates both parties to hold out, hoping to obtain a verdict that supports their position. Trade disputes over complex issues should, in most cases, be resolved once the respective panel and (if an appeal is made) the appellate body have issued their verdicts. The reason is that the DSM's verdict has, at that point, reduced the legal uncertainty. This argument reflects the common assumption in legal research that court trials would not be needed in a world with zero transaction costs and full information. According to this argument, absent transaction costs and uncertainty, the parties in a dispute should always come to a mutually agreed solution without going to court, since adjudication always entails direct and indirect costs.Footnote ³⁴

Again, anticipation is likely to play a significant role in governments’ decisions. Both sides in a conflict should anticipate that the conflict will end once the court has issued a verdict if the procedure is mostly about rule clarification. They should therefore be willing to escalate a dispute to the court level, because both can benefit from reduced transaction costs and the lower level of uncertainty that results from the adjudication process. Compared to the enforcement argument, anticipation in the managerial logic works in the opposite direction: it increases, rather than decreases, the escalation of a dispute to the panel stage. Once the panel resolves the issue, however, governments should try to restore compliance with WTO law. Since governments know that compliance will most likely arise after the panel ruling, they are (according to the managerial argument) more inclined to ask for a ruling. Hence from the managerial perspective, forward-looking behavior increases the already existing likelihood of dispute escalation to the panel stage under conditions of complexity, whereas the opposite is true according to the enforcement perspective.

The principal empirical implication of the complexity (managerial) perspective on dispute settlement is that disputes over more complex issues are more likely to escalate, but only up to the level of the panel or appellate body ruling. Escalation to that level (but not beyond) presumably clarifies the legal situation and helps countries restore compliance and solve the problem. Hypothesis 2 reflects these arguments.

To summarize, we expect fundamentally different dispute resolution dynamics depending on whether the causal mechanism proposed by the enforcement or the managerial perspective dominates. According to the enforcement perspective, we should expect that a panel verdict is unlikely to resolve the conflict if a dispute's domestic political stakes are high. The key driver of dispute settlement dynamics from this perspective – domestic political stakes – initially reduces the probability of reaching the panel stage (assuming that governments are forward looking), but increases the probability of escalation after a panel ruling. In contrast, the managerial perspective implicates that dispute escalation to the panel stage can be an important step towards dispute resolution. The key driver of dispute settlement dynamics from this perspective – issue complexity – has an effect that clearly differs from what one would expect from an enforcement perspective: it increases the probability of dispute escalation to the panel stage, but reduces the probability of further escalation thereafter.

Empirical Design

The theoretical arguments presented above take into account anticipatory (forward-looking) behavior of states during the bargaining process between the complainant and defendant in WTO dispute settlement. The assumption of anticipatory behavior has important implications for the empirical analysis: when actors are forward looking, their choices at the early stages of the bargaining process are influenced by their expectations about their opponent's future choices. Failure to account for anticipation may lead to biased estimates of the probability that an actor will make a particular choice at the outset and during the bargaining process.Footnote ³⁵ We develop a strategic model that reflects the essential steps in the WTO DSM. Based on this model, we can deduce the likelihood that the complainant and the defendant will take specific actions during the WTO dispute settlement process. The empirical model is then estimated using statistical backwards induction.Footnote ³⁶

The strategic empirical approach used for this analysis differs in important ways from the traditional statistical view on selection dynamics. Specifically, our strategic empirical approach assumes that selection occurs because governments are forward looking, which motivates them to escalate some disputes but not others. Taking into account such forward-looking behavior requires the analyst to start at the end of the decision tree, which is the last stage of the dispute settlement process, and work backwards towards the beginning of the game. The decision to study strategic versus non-strategic behavior then implies that we cannot use traditional statistical selection models, which start at the beginning of the decision tree and hence do not capture strategic and forward-looking behavior. The choice between the traditional and the strategic approaches ultimately depends on theoretical considerations that specify which underlying processes lead to selection. Given the importance of forward-looking behavior in our theoretical discussion, we opted for the strategic approach. The results discussed below will show how important it is to properly model governments’ forward-looking behavior in our analysis.

Strategic Model

The game shown in Figure 1 represents the most important stages of the WTO DSM, starting with the consultation phase. If the complainant is not satisfied with the consultation outcome, he can request a panel ruling (P). We define the WTO panel and appellate body process as a single stage in order to simplify the analysis.Footnote ³⁷ The dispute ends at the consultation stage if no panel ruling is requested (∼P). After the panel ruling, the defendant can accept the verdict (A) or not (∼A). We assume that the game ends if the defendant implements the panel verdict, because there is no reason for the complainant to escalate the dispute further once it has reached its goal and the disputed trade barrier is removed. If the defendant refuses to implement the panel verdict, the complainant can decide to back down (BD) or escalate the conflict further; escalation leads to an implementation dispute (∼BD).Footnote ³⁸

Fig. 1

The dispute bargaining game

The game has four outcomes: (1) Y ₄ is observed if the dispute escalates to the implementation dispute stage; (2) Y ₃ occurs if the defendant refuses to change its trade policy after a panel verdict, but the complainant decides not to pursue the issue further; (3) Y ₂ is observed if a panel issues a verdict and the defendant complies with the ruling and implements the requested policy changes; and Y ₁ occurs if, after consultations with the defendant, the complainant does not request a panel. U_i (Y_j ) is actor i's utility of outcome j, and e_i,j is the error assigned to this utility. The parameters p ₁ through p ₃ denote the probabilities that the respective move is realized.

We use quantal response equilibrium (QRE) to derive the outcome probabilities.Footnote ³⁹ It follows from Figure 1 which outcomes will occur in equilibrium. The complainant will choose ∼BD if its utility of ∼BD is greater than the utility of BD, $$$ U_{C}^{\ast} {\rm{(}}{{Y}_{\rm{4}}}{\rm{)}}\: \, \gt \:U_{C}^{\ast} {\rm{(}}{{Y}_{\rm{3}}}{\rm{)}} $$$ and vice versa, where $$$ U_{i}^{\ast} {\rm{(}}{{Y}_j}{\rm{)}}\, = \,{{U}_i}{\rm{(}}{{Y}_j}{\rm{)}}\, + \,{{e}_{i,j}}. $$$ The defendant will opt for A if its utility of A is greater than the expected utility of ∼A, $$$ U_{D}^{\ast} {\rm{(}}{{Y}_{\rm{2}}}{\rm{)}}\, \gt \,{{p}_{\rm{3}}}U_{D}^{\ast} {\rm{(}}{{Y}_{\rm{4}}}{\rm{)}}\, + \,{\rm{(1}}\,{\rm{ - }}\,{{p}_{\rm{3}}}{\rm{)}}U_{D}^{\ast} {\rm{(}}{{Y}_{\rm{3}}}{\rm{)}}{\rm{.}} $$$ In the same vein, the complainant will not demand a panel if its utility of ∼P is greater than the expected utility of P, $$$ U_{C}^{\ast} {\rm{(}}{{Y}_{\rm{1}}}{\rm{)}}\, \gt \,{\rm{(1}}\,{\rm{ - }}\,{{p}_{\rm{2}}}{\rm{)}}U_{C}^{\ast} {\rm{(}}{{Y}_{\rm{2}}}{\rm{)}}\, + \,{{p}_{\rm{2}}}{{p}_{\rm{3}}}U_{C}^{\ast} {\rm{(}}{{Y}_{\rm{4}}}{\rm{)}}\, + \,{{p}_{\rm{2}}}{\rm{(1}}\,{\rm{ - }}\,{{p}_{\rm{3}}}{\rm{)}}U_{C}^{\ast} {\rm{(}}{{Y}_{\rm{3}}}{\rm{)}} $$$ .

Utilities, Variables and Data

We use several explanatory and control variables to measure governments’ utilities. The political stakes reflect the audience costs that governments face at the domestic level. These audience costs should differ across economic sectors and in the extent to which various sectors can effectively organize to voice their concerns. Following the existing literature, we expect disputes involving the manufacturing sector to involve lower domestic political stakes than those related to agriculture, resource extraction and mining, and services.Footnote ⁴⁰ The reasoning is that especially the agricultural sector has a particularly strong ability to organize politically, form alliances with other stakeholders, publicize disputes, and lobby for trade-restricting policies or practices. According to the seminal work by Olson, we should thus observe that due to this high level of organization, the agricultural sector should be able to politicize disputes and thus to pressure governments to take its side.Footnote ⁴¹ While the importance of the issue at stake might often be modest in purely economic terms, the high political stakes associated with those disputes often makes it very difficult for the defendant country to accommodate the complainant's requests, as the EU-US dispute over GM crops illustrates.Footnote ⁴² Hence we include a dummy variable that measures whether a dispute involves the manufacturing sector, assuming that manufacturing is associated with lower domestic political stakes.

To measure complexity, we use a dummy variable that captures whether a dispute refers to a complex trade-protective instrument – such as technical barriers to trade, anti-dumping, subsidies or various other instruments – in contrast to tariffs and quotas, which we assume to be less complex. Coding tariff and quota cases as less complex follows a similar logic as the coding by Guzman and Simmons.Footnote ⁴³ They argue that tariff and quota cases are easier to resolve since they refer to a continuous issue, in contrast to issues that have an ‘all or nothing’ character.

In additional analyses, we use more disaggregated measures that distinguish between more sectors and trade instruments. The results are largely similar. Since the number of disputes decreases with each additional stage in the settlement process, the results for the more disaggregated sector and instrument variables are based on few observations for each sub-sector or sub-instrument. This poses problems for statistical inference, and the use of more aggregated measures ensures that our conclusions are more reliable.

The empirical model also includes several control variables. The utilities of escalating the dispute to the implementation stage should depend on the retaliatory power of the two countries. Countries with less economic power will gain less from imposing countermeasures against a guilty country, and vice versa. Our measure of retaliatory power is the difference in economic size of the two countries.Footnote ⁴⁴ We also add the number of complainants involved in a dispute, because a higher number of complainants may increase the pressure to comply with a WTO verdict due to higher international reputation costs. Finally, the defendant's utility of compliance depends on the content of the ruling. To capture the extent to which the panel ruling strengthens the complainant's or the defendant's position, we include a variable that measures the ratio between the number of legal claims accepted by a panel and the total number of legal claims made by the complainant, as well as the square term of this ratio. The logic behind these two variables is that we expect compliance to be more easily achieved if only a few of the legal claims put forward by the complainant are upheld by the panel. Similarly, if the panel supports a higher proportion of the complainant's legal claims, the potential reputation costs of non-compliance should be higher. When, however, only a portion of the legal claims made is accepted by the panel, the defendant country should have the strongest incentive not to comply. Table 1 summarizes how the explanatory variables relate to the theoretical concepts and our own theoretical discussion, and how they are measured.

Table 1

Theoretical Concepts and Explanatory Variables

To complete the model-building process, we need to specify which variable affects an actor's utility at what stage in the WTO DSM. In other words, we need to assign the empirically observable measures of the utilities, U_i (Y ₁) through U_i (Y ₄), to estimate the probability of a particular outcome. Each of the utilities is a linear function of explanatory variables. Equations 1 through 7, define which variables are assigned to which particular utility. This structure follows from our theoretical considerations and is largely consistent with previous formal analyses of conflict bargaining.Footnote ⁴⁵

$$\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!{{U}_C}{\rm{(}}{{Y}_{\rm{1}}}{\rm{)}}\, = \,{\rm{0}}$$

$$ \!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!{{U}_C}{\rm{(}}{{Y}_{\rm{2}}}{\rm{)}}\, = \,{{\beta }_{C,{\rm{4}}}}{\rm{Complex}} $$

$$ {{U}_D}{\rm{(}}{{Y}_{\rm{2}}}{\rm{)}}\, = \,{{\beta }_{D,{\rm{3}}}}{\rm{Stakes}}\, + \,{{\beta }_{D,{\rm{4}}}}{\rm{Complex}}\, + \,{{\beta }_{D,{\rm{4}}}}{\rm{Strength}}\, + \,{{\beta }_{D,{\rm{4}}}}{\rm{Strengt}}{{{\rm{h}}}^{\rm{2}}} $$

$$ \!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!{{U}_C}{\rm{(}}{{Y}_{\rm{3}}}{\rm{)}}\, = \,{{\beta }_{C,{\rm{0}}}} $$

$$ \!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!{{U}_D}{\rm{(}}{{Y}_{\rm{3}}}{\rm{)}}\, = \,{{\beta }_{D,{\rm{0}}}} $$

$$ \!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!{{U}_C}{\rm{(}}{{Y}_{\rm{4}}}{\rm{)}}\, = \,{{\beta }_{C,{\rm{1}}}}{\rm{DifGD}}{{{\rm{P}}}_{CD}}\, + \,{{\beta }_{C,2}}{\rm{NumComp}}\, + \,{{\beta }_{C,{\rm{3}}}}{\rm{Stakes}} $$

$$ \!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!\!{{U}_D}{\rm{(}}{{Y}_{\rm{4}}}{\rm{)}}\, = \,{{\beta }_{D,{\rm{1}}}}{\rm{DifGD}}{{{\rm{P}}}_{DC}}\, + \,{{\beta }_{D,{\rm{2}}}}{\rm{NumComp}} $$

We normalize the complainant's utility of outcome Y ₁ to 0, as shown in Equation 1. Following the complexity argument, we assign the variable measuring dispute complexity to outcome Y ₂, as shown in Equations 2 and 3, because settling the dispute after a clarifying panel ruling should be valuable for both disputants. The variable measuring domestic political stakes (Manu) is added to the complainant's utility of Y ₄, as shown in Equation 6. If the complainant does not follow through with the complaint even though the defendant refuses to implement the panel verdict, the complainant incurs an audience cost for backing down. The audience cost should be smaller for disputes involving the manufacturing sector. The defendant incurs the same audience cost if it accepts an unfavorable panel verdict, as specified in Equation 3. Similarly, the content of the panel verdict influences the defendant's utility of compliance, and therefore the strength variable is added to the same equation. Finally, the number of complainants and amount of power affect the actors’ utilities in an implementation dispute, as shown in Equations 6 and 7.

It is important to keep in mind that, due to the assumed forward-looking behavior of the two actors, explanatory variables that appear in the utility function at a later stage of the WTO DSM will, indirectly, also appear in the utility functions at an earlier stage. Using a strategic model therefore has important implications for understanding the effect of the different variables. When relying on regular regression models, the effect of each variable is denoted by its regression coefficient. In a strategic model, however, the effect of one variable has both direct and indirect components: direct as displayed by its regression coefficient and indirect through entering the probabilities that affect the earlier stages in the backwards induction model.Footnote ⁴⁶ For example, the dummy variable for manufacturing that measures domestic political stakes should influence the utility of the complainant at the last stage U_C (Y ₄) and thus his decision whether to launch an implementation dispute. Due to the assumption of forward-looking behavior, the manufacturing variable will also influence its decision at the first stage, when the actor decides whether to ask for a panel ruling.Footnote ⁴⁷

The expected signs of the coefficients on the variables follow from the discussion above. A variable that has a negative effect on a country's utility from a specific outcome has a negative sign, and vice versa. These expectations generate the following implications for the probabilities of a particular move. As suggested by our theoretical arguments, and in line with the enforcement perspective, the probability of a panel ruling should increase if the dispute has greater domestic political stakes if we do not expect the government to be forward looking, but should decrease with greater domestic political stakes if we expect the government to be forward looking.

In both scenarios, however, disputes associated with greater domestic political stakes are more likely to continue and escalate further after the panel ruling. The probabilities that the defendant does not accept the ruling, p ₂, and that the complainant will not back down, p ₃, increase because the audience costs of these two choices are high for both parties. In line with the managerial perspective, complex cases should generally display a higher probability of escalating to a panel ruling. In contrast to disputes with greater domestic political stakes, the probability that a dispute is settled after a panel or appellate body ruling – that is, the probability of observing outcome Y ₂ (compliance after a ruling) – should increase as the complexity of the disputed issue increases: both the probability that the defendant will accept the ruling, 1−p ₂, and the probability that the complainant will request the establishment of a panel, p ₁, increase. The reason is that complexity positively affects both the complainant's and the defendant's utility of outcome Y ₂.

The data on trade disputes was collected from WTO documents. For each dispute, we obtained information on the defendant and complainants, and at which stage in the WTO DSM the dispute ended. Furthermore, we assigned a corresponding economic sector to each dispute and identified which trade policy instrument was used to protect this industry (for example, tariffs or subsidies). The first disputes entered the WTO system in 1995. We included all disputes for which a request for consultations was filed until 2006. Table 2 shows the frequency distribution of the dependent variable at the three dispute stages (Y ₁, Y ₂ and Y ₄). The table also lists, for each stage, how many disputes are, in our empirical definition, more complex cases and how many disputes involve higher domestic political stakes (the non-manufacturing sector). For example, the table shows that more complex cases end less often with a panel ruling (eighty cases) than with consultations (106 cases). Furthermore, more complex cases end more often with non-compliance (thirty-eight cases) and implementation disputes (twenty-four cases) than with compliance (thirty-seven cases) or backing-down by the complainant (fourteen cases).Footnote ⁴⁸ Disputes with lower domestic political stakes also end more often with panel rulings (fifty-three) than with consultations (forty-seven). Compliance occurs more often in these cases (thirty-four) than non-compliance (eighteen), and the number of implementation disputes is ten. Monadic and dyadic country-level trade data were taken from Barbieri, Keshk and Pollins and are log-transformed.Footnote ⁴⁹ Table 3 presents the summary statistics.

Table 2

Descriptive Statistics: Dispute Characteristics

Table 3

Descriptive Statistics: Other Independent Variables