Precedent in the Trade Regime

As in the rest of international law, there is broad consensus over the absence of formal binding precedent at the WTO. The panel in US—Zeroing says as much in a recent ruling: “In our view, there is not a system of precedent within the WTO dispute settlement system and panels are not bound by Appellate Body reasoning.”Footnote ¹² The Appellate Body, the WTO’s higher court, has explicitly struck down claims by panel judges that argued the contrary. In one such instance, a panel of judges attempted to argue that past rulings comprised “subsequent practice” within the meaning of the Vienna Convention on the Law of Treaties—to which the WTO Agreement is subject—in which case past rulings would become a source of law. The AB forcefully disagreed, arguing that the contracting parties did not intend for panels’ rulings to become definitive interpretations of the rules.Footnote ¹³ Similar denials of binding precedent are also found elsewhere in WTO law. Presciently, Article 3.2 of the Dispute Settlement Understanding (DSU) reads “Recommendations and rulings of the DSB cannot add to or diminish the rights and obligations provided in the covered agreements.” The intent is unambiguous: judicial rulings should not change the meaning of countries’ commitments under the agreement.

Yet these same judges appear able to tolerate some disparity between legal status and practice. When the aforementioned panel in US—Zeroing declared that there was no system of precedent in the WTO, it went on in the same breath to cite the AB in a more liberal stance:

Common though it may be, whenever WTO judges invoke past rulings to justify their decision, they expose themselves to accusations of flouting DSU Article 3.2 and engaging in judicial rule-making. In the very same US—Zeroing dispute, the United States pushed back against the panel’s references to past rulings, reminding everyone that: “...the Panel is not bound to follow the reasoning of any prior report. According to the United States, the Panel is charged with making its own objective assessment of the matter before it.”Footnote ¹⁶ What is striking is how rare such explicit remonstrations on the part of countries have been.

Instead, it is legal scholars who have been most vocal in their concerns over the elevation of precedent at the WTO. These scholars are not against legal consistency per se. Rather, they recognize that an obligation to follow past decisions implies the power to impose an interpretation of the rules on the future. Precedent following cannot only be backward-looking (Schauer Reference Schauer1987). These scholars’ concern is that if rulings have effects beyond the case at hand, international courts are empowered in a way that runs into “deeper questions of legitimacy and accountability stemming from the ‘democratic deficit’ inherent in this process” (Barfield Reference Barfield2001, 43). Ragosta, Joneja and Zeldovich (Reference Ragosta, Joneja and Zeldovich2003) warn that the court might “abuse its binding nature to create WTO ‘common law,’ to which the Members never agreed.” They caution that binding precedent might empower WTO judges to create obligations where country negotiators never intended for there to be any. The result, they warn, may threaten support for additional liberalization. Some view this threat as sufficient grounds for exiting the WTO altogether (Ragosta, Joneja and Zeldovich Reference Ragosta, Joneja and Zeldovich2003). The greater point is that a commitment to legal consistency necessarily entails the empowerment of international courts beyond what some view as a legitimate threshold.

These views all support the notion that the meaning of the law cannot be derived unambiguously from the text. Law is made and changes incrementally by way of interpretation (Dworkin Reference Dworkin1982; Venzke Reference Venzke2012).Footnote ¹⁷ Strikingly, the WTO itself recognizes this intersubjective aspect of rules’ evolution, and it explicitly restricts the power of “definitive interpretation” to states.Footnote ¹⁸ Article IX:2 of the WTO Agreement states that “[t]he Ministerial Conference and the General Council [member states] shall have the exclusive authority to adopt interpretations of this Agreement and of the Multilateral Trade Agreement.”

In sum, there is considerable analytical ambiguity over the function of precedent. On the one hand, there is consensus as to precedent’s lack of formal authority in international law, and trade law in particular, and a group of observers who warn against straying from it. On the other hand, both judges and parties routinely invoke past rulings under the banner of legal consistency. It is this discrepancy between principle and practice that this article explores empirically. The underlying premise of the article is that if precedent changes state expectations over what policies the rules allow, it should have an observable effect on behavior, no matter what its formal status may be.

The empirical literature considering the implications of precedent on country behavior is limited. While many have followed in the footsteps of Bhala (Reference Bhala1998) in distinguishing between de jure and de facto stare decisis, few have suggested means of testing which one is more applicable in practice. Among these, Busch (Reference Busch2007) claims that the incentive to set a precedent multilaterally leads NAFTA members to file at the WTO instead of NAFTA, even when NAFTA rules offer a more favorable verdict on a case’s merits. More recently, Davis (Reference Davis2012) considers the spillover effects flowing from precedent, asking whether countries internalize rulings sufficiently in disputes to which they are not party to modify their own domestic legislation accordingly.

And some anecdotal evidence hints at countries employing the kind of strategy this article seeks to identify. As Simmons and Guzman (Reference Simmons and Guzman2005, 567) mention in passing, powerful countries are more likely to initiate WTO cases with “long term precedential value,” without going further into how we would know this was happening. Closest to this article’s line of argument, Porges (Reference Porges2003) claims that a “government may have as its objective the establishment of a legal precedent rather than any particular commercial outcome.” Here, I use a novel empirical approach, applied for the first time to the WTO setting, to evaluate whether these claims can be verified. Does precedent in international law affect state behavior?

The Example of European Safeguards Disputes

To illustrate both the theory and the approach taken to test it, I examine a series of disputes fought over the same legal issue: safeguards. Safeguards constitute the quintessential escape clause of the trade regime: they allow members to temporarily escape their obligations when confronted with difficult domestic circumstances. Together with antidumping and countervailing duties, they form one of the regime’s three trade remedies. Up to 1995, the exercise of safeguards fell under the GATT’s Article XIX, which was then superseded by the Agreement on Safeguards with the inception of the WTO.

With the creation of the WTO, safeguards became considerably easier to use. This reflected an effort to turn countries away from resorting to alternative policy instruments, which were usually considered to be more economically distortionary.Footnote ³¹ Compensation, for instance, was no longer required of members using the safeguard (Pelc Reference Pelc2009). As Bown (Reference Bown2002, 58) put it at the time, it is difficult to imagine by what means the WTO’s new Agreement on Safeguards could have made the escape clause any more attractive to WTO members, short of paying members to exercise it.

The European Union has long stood out among WTO members by how little it relies on safeguards. By 2001, which was, as this discussion goes on to show, a turning point in the use of safeguards, the European Union had not filed a single safeguard investigation on behalf of its industries. The United States, by comparison, had initiated ten important investigations for large industries. Emerging economies like India, which counted as many investigations as the United States in 2001, as well as Argentina, were already heavy users by this point, on their way to becoming the most frequent users of trade remedies in the system.

Faced with what was from its standpoint a costly rise in the use of safeguards by trade partners, Europe took on the issue in dispute settlement. Given its own low use, any legal constraints on the exercise of safeguards represented a net gain from Europe’s standpoint. Such constraints would come at little cost to its own policy space, and would decrease the extent to which its trade partners could erect barriers against European exports.

Yet Europe did not take on the issue of safeguards by challenging large markets, or countries, like the United States or India, that had a high vested interest in the issue. It began by challenging a small barrier in a tiny market, over Korean safeguards on skimmed powdered milk, in 1997. In this first dispute, the European Union made a bold legal claim: despite the WTO’s Agreement on Safeguards no longer making any mention of it, the European Union invoked a clause from GATT Article XIX which required that the WTO Member exercising a safeguard action find itself faced with developments it had not “foreseen” or “expected.” This clause had gone unenforced even under the GATT, and the WTO’s Agreement on Safeguards made no reference to it.

Before the panel had ruled in the dispute against Korea, the European Union filed another safeguards dispute, this time against Argentina, over footwear. The commercial value of this case for Europe, measured in bilateral trade flows, was even lower than Korea’s powder milk market had been. Here too, Europe made its legal claim about “unforeseen developments.” Europe’s efforts did not go unnoticed. As Porges (Reference Porges2003) put it at the time, speaking of the two concurrent cases, “the disputes concerned export markets of little or no commercial concern to European exporters [...]; the very lack of stakeholder interest left Commission litigators with a free hand to target their arguments.”

Europe got its way. In both disputes, the AB affirmed the authority of the “unforeseen developments” clause, as per the European Union’s arguments. In the second of its rulings, the AB went as far as to underline the continuity of its reasoning: “[w]e have recently confirmed this principle in our Report in [Korea–Dairy].”Footnote ³²

Europe obtained the language it was seeking. The precedent, as it were, was set. Every single subsequent safeguards case has cited these two rulings, net of appeal, to justify the continued relevance of the “unforeseen developments” clause, despite its absence in the Agreement on Safeguards. What some have described as judicial activism on the part of the AB was, by 1999, a fait accompli: since then, there exists a shared, WTO-wide understanding that safeguards flout the rules unless the domestic circumstances leading to their use are “unforeseen” (Goldstein and Steinberg Reference Goldstein and Steinberg2008; Pelc Reference Pelc2009). Despite the lack of formal binding precedent, for all intents and purposes, Europe obtained a rule change through litigation that it could never have obtained through negotiation. The centrality of these two disputes of negligible commercial importance to jurisprudence over safeguards comes across clearly in Figure 1. The figure shows the universe of safeguards disputes: all disputes that either cite or were cited by a ruling on safeguards. The size of the nodes is proportional to their in-degree, that is, to the number of other rulings that cite them. What is evident is that despite their commercial insignificance, Korea—Dairy (DS98) and Argentina—Footwear (DS121) have disproportional jurisprudential impact. Figure 1 also makes light of another fact: all EU-initiated disputes are shaded in red. The take-away is that all seminal safeguards cases have been filed by the European Union.

FIGURE 1.

Europe’s Seminal Safeguards Cases, EU filings vs. ROW

Indeed, after the rulings in Korea—Dairy and Argentina—Footwear, the European Union filed two more safeguards disputes in short order; these were no longer over small markets. In both US—Line Pipe and US–Steel Safeguards, the amount of trade at issue was more than 55 times greater than had been at stake in the disputes against Korea and Argentina. Both US—Line Pipe and US—Steel Safeguards belong in the top 5% of all WTO disputes according to trade value. In both cases, the industries in question were politically powerful, and frequently petitioned for safeguards. The second of these cases remains one of the most notorious disputes in the WTO. When the United States raised safeguards on steel in 2001, it was widely condemned across the membership, and the legal proceedings pit the United States against much of the rest of the world. The panel ruled against the United States on all claims, and did so by invoking the rulings in the two cases the European Union had filed three years prior. When the United States appealed, the AB upheld all the panel’s findings, including the “unforeseen developments” claim, by pointing to its finding in Korea–Dairy and Argentina–Footwear. The steel safeguards dispute marked a turning point. The United States ceased relying on safeguards almost entirely (Pelc Reference Pelc2009). By 2010, it had not filed a single other safeguard investigation.

Europe thus successfully suppressed the use of safeguards through litigation. It led to an interpretation of the rules which likely could never have been achieved through negotiation within a trade round (Goldstein and Steinberg Reference Goldstein and Steinberg2008). And it did so in a specific way, by setting the precedent through a series of low commercial value disputes, and then relying on this precedent in subsequent high value cases. The resulting pattern is shown in Figure 2, which illustrates the relevant citations network across time and commercial value. Because of the acyclic nature of citation network graphs, all unidirectional citation links go back in time—a case cannot cite a future ruling to justify its reasoning. What Figure 2 makes evident is the increase in commercial value over the life of this citation path. Europe shapes the WTO acquis in low commercial value cases, and later exploits that precedent in higher value cases. The positive slope of the links, in this case, is meaningful. It suggests that when it comes to building precedent, countries “buy low and sell high”: they invest in precedent in low stakes disputes, when they encounter least resistance, and reap the benefits in higher stakes cases.

FIGURE 2.

Charting the Commercial Value of a Precedent Chain Across Time

Comparing Figure 1 to Figure 2, it becomes evident that for the safeguards issue area, Europe’s legal strategy is apparent. As a result, commercial and precedential value are negatively correlated across this subset of disputes. The question is, to what extent does this pattern apply to the universe of WTO disputes? Do other countries systematically behave in a way similar to the European Union on safeguards, using low salience disputes to shape jurisprudence in a way that favors them subsequently?

Network Analysis

Most recent legal network studies aim to measure how “legally central” (Fowler, Johnson, Spriggs, Jeon and Wahlbeck Reference Fowler, Johnson, Spriggs, Jeon and Wahlbeck2007) or how “strongly embedded in case law” (Lupu and Voeten Reference Lupu and Voeten2010) a case is. Fowler, Johnson, Spriggs, Jeon and Wahlbeck (Reference Fowler, Johnson, Spriggs, Jeon and Wahlbeck2007) do so to show that reliance on precedent in the USSC has risen through time; that reversed Supreme Court decisions are more “important,” or legally central, than others; and that judges are more likely to ground such reversals in case law. Lupu and Voeten (Reference Lupu and Voeten2010), in turn, find that ECtHR cases where the value of persuading judges in domestic courts is higher tend to be more embedded in case law. Taking a further step in examining the development of legal doctrine, Clark and Lauderdale (Reference Clark and Lauderdale2012) add a useful innovation by considering the extent to which a ruling relies on past opinions to arrive at an alternative measure of legal significance.

The aim of this article is different. I am interested in the extent to which a given country files disputes that it then relies on—directly or indirectly—in subsequent commercially important disputes. The main departure from the above studies is that the level of analysis is at the complainant level: the analysis is concerned with the strategic behavior of countries pursuing their own interest through litigation. The second distinction is that I am concerned with an additional dimension, that of commercial value.

For these reasons, I depart from the prevalent method employed in the legal networks literature, which calculates centrality measures by relying on a recent advance in web search theory, called Kleinberg scores (Kleinberg Reference Kleinberg1999). Kleinberg scores measure the degree to which nodes serve as hubs and authorities. Hubs are those cases that cite many rulings; authorities are those cases that are cited by many rulings. Both have a mutually reinforcing effect: a case’s authority score grows when it is cited by many good hubs. Conversely, a case’s hub score grows as it cites many good authorities.

Yet the emphasis that Kleinberg scores put on hubs is a premise that travels poorly to the WTO context. Because of the relatively small size of the WTO network and the considerable expertise of judges, we obtain relatively little information about the “quality” of a ruling from its ability to cite good authorities. In this case, identifying and citing good authorities is neither costly, nor is it especially difficult, in a way that might tell us something about the quality of panelists’ rulings.Footnote ³³

Instead, the starting point of the approach used here goes back to an earlier and simpler network measure by Katz (Reference Katz1953). Katz centrality measures the influence of a given node in a directed acyclic network by measuring all the paths leading to it, but discounting indirect citations in proportion to how far on the path they are. In other words, a ruling C adds more to the centrality of A when it cites it directly, rather than when it cites a ruling B which in turn cites A. This discount rate, or what Katz called the attenuation factor, is a tunable parameter β that is set, within certain bounds, by the user. Setting β to 1 would mean no attenuation, so that indirect citations would be just as valuable as direct citations. Setting β to 0 would mean that only direct citations mattered. The analysis here employs an α of 0.35, though varying this measure from 0.2 to 0.5 has no discernible effect on the analysis.

Using the adjacency matrix A_ji , the centrality measure of node i, determined by whether node j links to it, in a network of (n + 1) nodes, is expressed as

To adapt the Katz centrality measure to the present context, it is possible to add a weight W_j to each dispute j, corresponding to the bilateral trade at stake in that dispute, such that a citation from a highly commercially significant dispute will add more to dispute i’s precedential value than a citation from a commercially trivial dispute: This can be expressed in matrix format, using a 1 × n row vector w for the weights: Finally, to formulate this at the country level, in a network containing m countries, one can imagine an m × n matrix such that W_gh is the logged commercial value of dispute h for country g. This produces C, an m × n matrix such that C_gh becomes the precedential value of dispute h for country g: Intuitively, what the measure C amounts to is picking all the citation paths for which the initial authority and the final hub are both cases initiated by the same WTO member, and weighting each node on this path that was initiated by that same member by its commercial weight, while discounting it in accordance to its distance from the initial authority, to arrive at the precedential value of that initial authority.

Precedential Value vs. Inward Citations

The measure of interest, precedential value, is thus quite distinct from a case’s in-degree, or the number of disputes that cite it directly. The key distinction is that precedential value considers both indirect citations, and the commercial value of citing cases, and is measured from complainant’s point of view. Across all WTO cases, the dispute with the highest in-degree is US—Wool Shirts and Blouses. This dispute, filed by India and ruled on in 1996, set a number of important precedents.Footnote ³⁴ Yet in terms of precedential value to the complainant, India, it ranks only 27th among the disputes in the system. A even starker example of this contrast is US—Gasoline, which was initiated by Venezuela. While it ranks as the 4th most cited dispute across the system, its precedential value is nil. What this means is that no subsequent dispute of positive commercial value initiated by Venezuela either cites US—Gasoline or cites another dispute (initiated by whomever) that in turn cites US—Gasoline. In other words, there exists no commercially significant hub filed by Venezuela that is on a citation path that counts US—Gasoline as an authority. If Venezuela set any useful precedent in the dispute, it did not exploit it subsequently.

It is nonetheless useful to consider the in-degree of disputes across the network, both as a description of the network, and as a test of the data. This distribution is graphed in Figure 3, where both the number of citations, on the horizontal axis, and frequency, on the vertical axis, are logged. It shows the power-law tail that we expect from scale-free networks, or networks where nodes join in accordance with preferential attachment. In this case, there are a few cases in the network that are subsequently cited a great number of times.

FIGURE 3.

Distribution of Inward Citations Across WTO Disputes

Considering instead precedential value as defined in (4), the dispute that ranks highest in the system is Japan–Alcoholic Beverages II, initiated by the European Union, and ruled on in 1996. What this means is that Europe went on to file a number of commercially important disputes that relied on the precedents set in Japan—Alcoholic Beverages II, either directly or indirectly.

Tests

Do countries strategically shape precedent? Do they file low commercial value cases to build up valuable precedent that they subsequently exploit? To answer this question, I calculate the precedential value, in the manner outlined above, of all cases since 1995 where a ruling was adopted in the WTO, which make up all the nodes in the WTO network. This measure of precedential value makes up my main dependent variable throughout the analysis. I also re-estimate the same model using the simpler in-degree measure as the dependent variable instead, and compare the results. The level of analysis is the complainant dispute, since I am interested in whether individual complainants observably invest in precedent. A few disputes include more than one complainant. The resulting sample counts 171 observations.

Both to code my main independent variable of interest, the Logged Commercial Value of every dispute, and to arrive at the weight w used in the calculation of the centrality measure above, I first identify the traded products at issue in every case. These are coded in the Harmonized System (HS) at the six-digit level, and are taken from the Horn and Mavroidis database of WTO disputes, hosted by the World Bank,Footnote ³⁵ which I complete and extend using WTO documents. I then rely on trade flows data from the World Integrated Trade Service (WITS), also hosted by the World Bank, to calculate the bilateral trade flows of the products at issue in the dispute in the year preceding its initiation.Footnote ³⁶

Additionally, I control for a range of other possible determinants of precedential value. The most important of these is time. The variable Panel Year is coded as the year in which the panel report was issued. We would expect this measure to be strongly negative: more recent cases are likely to have shorter citation paths, and lower precedential value. To show this simple bivariate relationship, I graph both the in-degree and out-degree of WTO disputes, according to the date of the panel ruling, in Figures 4 and 5, respectively. In both cases, the expected bivariate relationship is apparent. I also make sure that the results are robust to an inclusion of a more flexible time variable, using cubic splines, and illustrated by the fitted line in Figure 4.

FIGURE 4.

In-Degree of WTO Disputes Through Time

FIGURE 5.

Out-Degree of WTO Disputes Through Time

Next, I account for each complainant’s market size. This is in part to account for Simmons and Guzman (Reference Simmons and Guzman2005)’s concern that powerful countries are more likely to file disputes with long term precedential value. If richer countries have a greater ability to act strategically in this way, we should expect this coefficient to be positive. The corresponding Logged Complainant GDP variable is coded using World Development Indicators data.

I also control for some additional characteristics of the litigation process itself. Number of Third Parties is a count variable coded as the number of countries that joined the dispute as third parties. This is a proxy for how many other WTO members have a stake in the dispute. Number of Panel Claims is a count of how many legal claims the complainant made when filing the dispute. There is some anecdotal evidence suggesting that employing a “trash can approach” is a sign of low legal merit of the dispute. The last control variable I include in all estimations is a Non-Merchandise Dispute Indicator, which is coded as 1 if a dispute challenged no specific product, but rather an issue of domestic legislation, such as US Section 301 (Pelc Reference Pelc2010, Reference Pelc2013). These often relate to more opaque barriers, and tend to be highly politically sensitive, since they concern domestic statutes. By definition, these dispute do not have a measurable stake, that is, the Logged Commercial Value is zero. As such, including the variable on the right-hand side is a way of verifying whether any effect of Logged Commercial Value might not be reducible to this category of nonmerchandise disputes.

The variables above make up my base estimation, shown in the first column of Table 1. Here, as in models 2 and 3, I rely on a standard OLS estimation. Robust standard errors are clustered on the dispute combined throughout. Indeed, some disputes have multiple complainants in the same dispute (under the same DS number), while others count as two separate disputes (with separate DS numbers) that result in common litigation and a single panel and AB report. To account for this, I code all such disputes as belonging to a single Dispute Combined number, and cluster errors on that identifier.

TABLE 1.

Do Countries Strategically Shape WTO Precedent?

Notes: OLS estimations in Models 1–3. Robust standard errors clustered on Dispute Combined in parentheses. + p < 0.10, **p < 0.05, ***p < 0.01. Model 3 includes cubic spline for time trend. Model 4 shows Bayesian estimates of draws from the posterior distribution, and top and bottom quantiles of marginal posterior distribution, implemented through Zelig using a Gibbs sampler.

In Model 2, I control for a number of additional possibly confounding variables. Systemic Interest Indicator is a dummy coded as 1 if any of the third parties in the dispute joined by invoking “systemic,” rather than “material” interests, and 0 otherwise. The invocation of systemic interests tends to be indicative of issues of law that have a far-reaching impact on WTO jurisprudence. We might thus expect it to be positively related to a dispute’s precedential value.

Additionally, I include Ruling Outcome, a measure of what proportion of the claims ruled on were decided in favor of the complainant. This measure thus varies from 0 to 1, with a sample average of 75.1%. Along these lines, I also account for whether a case has been appealed. Appealed cases potentially contain more material apt to be cited. I disaggregate this measure into two variables: an Indicator of Complainant Appeal, and an Indicator of Defendant Appeal, both relating to appeals of the panel findings. These also serve as useful ways of getting at the degree of satisfaction of each party with the panel ruling, which may indicate how close to the desired precedent the judges’ reasoning was.

Model 2 also changes the way time is included, by replacing the simple Panel Year variable with a cubic spline with four knots, as per the line drawn in Figure 4, where the nonlinearity of the relationship between time and citations comes across. Though less efficient, the inclusion of the cubic spline serves as a potentially better control of such nonlinearity.

Model 3 in Table 1 includes the same right-hand side variables as model 2. The difference here is the dependent variable, which is now a simple measure of in-degree—that is, the number of direct citations to the dispute under evaluation. I include this third model as a means of comparing the weighted, complainant-specific measure of Precedential Value to a simpler, unweighted, system-wide measure of direct citations. The time parameter in Model 3 returns to the simple Panel Year measure.

Finally, Model 4 features estimates from a Bayesian model. Including network measures into regression equations generates a number of potential challenges that political scientists are only starting to recognize. Fortunately, the application in this article—which amounts to placing a centrality measure on the left-hand side of the equation—is simple enough that there is little concern about bias. However, it is worth running an additional test of the relationship using an equivalent Bayesian model, which is more flexible in terms of its underlying assumptions, and which also fares better with small samples. The mean of random draws from the posterior distribution is shown in the 4th column, and columns 5 and 6 show bottom and top quintiles.

The findings are consistent across all estimations. In all four models, disputes’ commercial value is negatively related to their subsequent precedential value, offering support for the article’s main hypothesis: countries appear to strategically file low commercial value disputes which go on to build favorable precedent for subsequent cases. And the effect is substantively significant: looking at Model 2, increasing the commercial value of the average dispute by one standard deviation reduces the subsequent precedential value of that case (from the complainant’s standpoint) by nearly 48%.Footnote ³⁷ Though it is difficult to compare coefficients across the two dependent variables, it is apparent that the Dispute’s Commercial Value has a more significant impact on the complainant-specific precedential value than on disputes’ in-degree. This accords with intuition, and provides support for the view that complainants are strategically pursuing precedential value for their own benefit, in a way that has limited spillovers on the membership as a whole. By comparison, nonmerchandise disputes are associated with significantly lower precedential value. This, too, accords with intuition: these politically sensitive disputes tend to be one-offs, filed to challenge specific legislation. As a result, they are seldom invoked subsequently, either by the initial complainant or by others.

By contrast, the effect of market size is ambiguous. While positive in all the estimations of precedential value, it is only significantly positive in the Bayesian model. It has an insignificantly negative effect on the simple in-degree dependent variable, in Model 3. It is worth pointing out, however, that the reference category in this case is not made up of the entire membership, but of the set of 43 countries that have initiated a WTO dispute that ended with a ruling. These countries constitute a subset of members with high average legal capacity. To assess this relationship further, I replace the Logged Complainant GDP with indicator variables for the two superpowers, the United States and the European Union. When I do so for Models 1–3 (not shown to save space), both superpowers are associated with higher precedential value, but only the dummy for the United States is statistically significant at the 5% level. Overall, while there is some indication that wealth may play a role, the effect thus remains tenuous, and seems limited to the case of the United States.

There are interesting comparisons to be drawn between Models 1, 2, and 4, which use the precedential value dependent variable, and Model 3, which uses the in-degree as the dependent variable. Recall that one of the main distinctions between the two dependent variables is that the precedential value only considers citations in subsequent cases filed by the same complainant. By contrast, the in-degree measure treats invocations of a dispute by all countries alike. This distinction is reflected in the results over appeals. When considering precedential value from the complainant’s standpoint, an appeal by the complainant indicates a complainant’s dissatisfaction with the ruling, suggesting the dispute may be of limited value in setting the ground for subsequent filings. Accordingly, complainant appeals have an insignificant effect in Model 2 and a negative effect in the Bayesian estimation in Model 4. By contrast, even complainant appeals should increase the amount of legal reasoning to be drawn on by the membership as a whole. This is what Model 3 shows: complainant appeals have a positive effect on a dispute’s in-degree. Defendant appeals, in turn, should have a similar positive effect both on precedential value from a complainant’s perspective and on a dispute’s in-degree, since they indicate both a finding favorable to the complainant, and increase the amount of reasoning apt to be cited. So they do: all coefficients for Defendant Appeal are significantly positive.

Finally, and as expected, the year of the panel ruling is statistically and substantively significant. The earlier a case is filed, the more time it has to acquire precedential value by being cited by other commercially significant cases.

Evidence of Contestation?

One may ask, what about other countries? Are those states strategically filing test cases uniquely capable of anticipating the implications of the precedent they seek to set? In other words, if the European Union is strategic enough to file against Korean powdered milk to reify its preferred interpretation of the Agreement on Safeguards, it stands to reason that others might be both able to take the long-term view and willing to invest in countering the European Union’s efforts. Korea may have little at stake in the matter—that is why the European Union targets it—but why would those countries with more at stake not take Korea’s side and oppose the European Union’s claims?

Theory offers support for these questions. Insights from the legal interpretivist literature teach us that interpretations can be contested, and that political actors will invest resources in doing so (Johnstone Reference Johnstone2003; Venzke Reference Venzke2009). The precise legal status of precedent notwithstanding, once an interpretation has been laid down, it becomes more difficult to dislodge it. As Venzke (Reference Venzke2012, 58) has it, “if an actor manages to implement new reference points in legal argument that are aligned with its interest, others face an uphill battle.” If some countries invest in precedent, others should invest in contesting their efforts. Do we see signs of such contestation?

In the case of the WTO, contestation can come from only one place. While countries may join a complainant’s side in a dispute by becoming a co-complainant and thus bolster the case against the defendant, there exists no equivalent means of formally supporting the defendant’s case—with the exception of third party submissions. Third parties are countries other than the litigants that join a dispute from its start, and that seek to influence the outcome by submitting their views alongside the litigants (Johns and Pelc Reference Johns and Pelc2014). The average WTO dispute counts just under four third parties, and their submissions make up an integral part of the final panel report. Third parties’ submissions matter, not least because judges may be swayed by their arguments (Busch and Reinhardt Reference Busch and Reinhardt2006). The question becomes, do we see activity among third parties hinting at systematic contestation of precedent?

The case of the European Union and safeguards certainly suggests so. The United States joins as the only third party in Korea—Dairy. The majority of the US submission, recorded in the panel report, is spent denying the importance of the “unforeseen developments” requirement. In doing so, the United States clearly takes Korea’s side, refuting every EU argument in turn.Footnote ³⁸ The United States then takes the very same stance in Argentina—Footwear, which it also joins, this time as one of five third parties. Here too, the United States takes the defendant’s side, refuting all the European Union’s arguments.Footnote ³⁹

The United States is also a vocal third party during the appeal stage in both disputes. Interestingly, after restating its legal arguments, the United States shifts tactics during the appeal process, looking for support in state practice: “State practice has also treated the question of “unforeseen developments” as marginal, legally non-binding or subsumed by other aspects of the safeguards process.”Footnote ⁴⁰ The United States argues that the criterion should not be a requirement, since most countries do not treat it as such. By relying on state practice, the United States is actively seeking to delegitimize a particular interpretation of the text, were its legal argument to fall short.

In this instance, the United States interpretation loses out to the European Union’s. Yet the very fact that we observe the United States vehemently contesting the European Union’s efforts in these two minor disputes gives additional credence to the notion that countries are able to see far enough into the future to both invest in precedent and oppose attempts by others to do so if the resulting precedent is unfavorable to them. All this, while precedent has strictly no formal legal authority.

The question, once again, is whether this happens on average. If it does, then we should expect states to challenge complainants disproportionately more in the test cases that go on to acquire greater precedential value. To find out, I examine the partisanship of all third party submissions in every WTO dispute. I then verify whether prodefendant submissions are associated with greater subsequent invocations. To start, I use data from Busch and Reinhardt (Reference Busch and Reinhardt2006), which code all third party submissions as procomplainant, prodefendant, or mixed. I update the data for the disputes that took place between 2005 and 2012 following their methodology.

As in Busch and Reinhardt, most (71%) third parties are procomplainant; a minority (22%) are prodefendant; and an even smaller share (7%) are mixed submissions. The bivariate correlations between the partisanship of submissions and disputes’ precedential values, shown in Table 2, speak for themselves. What these correlations suggest is that contestation in the Korea—Dairy case is representative of the average among precedentially important disputes, which attract a disproportionate number of prodefendant submissions. By comparison, precedential influence is not positively correlated with either procomplainant or mixed submissions. The same split is apparent in correlations between partisanship and the simpler in-degree measure of precedential impact, shown in the second column of Table 2.

TABLE 2.

Correlations Between Precedential Value and Partisanship of Third Party Submissions

To be sure, the same relationship emerges in regression analysis. A poisson count model reveals a statistically significant positive relationship with prodefendant submissions, and a negative relationship with procomplainant submissions (not shown to save space). This holds once we add the other third party count (procomplainant and prodefendant, respectively) on the right-hand side, to control for the systemic interest of the dispute. Nonetheless, because the right-hand-side variable (precedential value) corresponds to events that took place later in time than the dependent variable (third party submissions), I choose not rely on these estimations too strongly. The assumption would need to be that the precedential values are subsequently realized proxies of a latent precedential value that is observable by third parties. While this assumption does in fact correspond to the theory, I nonetheless prefer to rely on the simpler bivariate correlations shown in Table 2, which are less prone to misinterpretation.

In this respect, it is unlikely that the causality runs in the opposite direction, i.e., that pro-defendant submissions are themselves producing greater subsequent citations. Prodefendant submissions represent a disagreement with the complainant’s claims. Since legal citations invariably serve to approve rather than to negate, prodefendant submissions should themselves lead to fewer, rather than more, citations. We can thus be confident that countries’ expectations over potential precedent-setting cases is what produces contestation, as evidenced in the number of prodefendant submissions. These cases then stand out through their high invocation in subsequent disputes. In sum, efforts by countries to shape precedent are observed by other strategic actors, who systematically seek to contest these interpretations when they appear unfavorable.

Distributional Consequences: Who Gains?

The empirical findings not only suggest that countries’ behavior reflects a shared view of precedent as binding, but also that certain countries, such as the United States, may be better able to exploit the implications of this view than others. It is worth thinking through these distributional consequences. Who gains more?

The question of distributional effects is especially interesting given how the judicialization of international politics, as a general phenomenon, is usually taken as a gain for small countries. The added “stability and predictability”—an explicit goal of the WTO’s legal systemFootnote ⁴¹ —that comes from the delegation of decision-making to courts is usually thought to benefit poor countries. Legalization is by construction a move away from power politics, and thus a reduction of the returns to power. Yet as I discuss, a commitment to predictability also means that the interpretations proffered by judges become “definitive.” In a decentralized regime such as the WTO, this forward-looking aspect of precedent presents major powers with opportunities to affect the meaning of the rules to their advantage.

Indeed, the meaning of the rules does not evolve in a random fashion. This shared meaning is less a function of the court itself than of far-sighted countries strategically initiating cases. In this respect, the concerns of observers warning against judicial rule-making seem overblown. Indeed, states are likely to retain the final trump card.Footnote ⁴² Were a court to regularly impose interpretations of the rules at odds with states’ interests, it would quickly be subject to political correction (Steinberg Reference Steinberg2004). The normative concern lies elsewhere. Insofar as some countries are better able to initiate multiple disputes than others, they will also have a greater say over the meaning of the rules, as seen through the prism of precedent. In this respect, differences in countries’ legal capacity, and thus their ability to file strings of related disputes, may hold even more significant implications than previously imagined (Busch, Reinhardt, and Shaffer Reference Busch, Reinhardt and Shaffer2009). At stake is not only the enforcement of one’s rights under the treaty; it is also about shaping that treaty’s meaning to one’s favor. The final irony, then, as highlighted by Ginsburg (Reference Ginsburg2005), is that those powerful countries best able to exploit precedent are also the ones to oppose it most often. The United States, in particular, stands out as opposing the authority of precedent more than any other WTO member, even as the analysis suggests that it is especially good at exploiting the precedents it sets. Consider a recent statement in US—Civil Aircraft: “The United States requests that the first sentence of paragraph 7.248 of the Interim Report be modified to remove any suggestion that reports of the Appellate Body are binding except with respect to resolving the particular dispute between the parties to that dispute.”Footnote ⁴³ Such explicit challenges, however, remain rare: WTO members, including the major powers, have tacitly accepted the growing authority of precedent.

Of course, this is no zero-sum game. Even if countries with high legal capacity have the greatest ability to initiate cases and thus shape case law, in so doing they also clarify the rules to everyone’s benefit. Political actors will often delegate decision-making to courts in politically sensitive areas. Any agreement may be preferable to no agreement, yet sometimes no single outcome is politically ratifiable by all parties. Courts provide a partial solution. This, after all, is the case made by students of European integration (Burley and Mattli Reference Burley and Mattli1993). Studies have shown that faced with legislative deadlock, countries grew more tolerant of the rising authority of precedent in European courts (Bouwen and McCown Reference Bouwen and McCown2007; McCown Reference McCown2003). This had similar distributional consequences to the ones identified in this article, whereby private actors endowed with high legal capacity (the ECJ allows private standing) grew “able and likely to pursue long planned litigation strategies in an attempt to reshape rules in a direction beneficial to them” (McCown Reference McCown2003). Despite the fact that this represented considerable delegation of power to the courts, European member states did not act to block precedential reasoning in the ECJ. Then as now, governments may realize that courts have the ability to converge expectations on politically fraught issues in a manner that benefits all members.

As is often the case, this public benefit—clarification of the rules—conceals a distributional conflict: some gain more than others. This is a familiar story. The existence of rules favors everyone, and the weak most of all. Yet in exchange for binding themselves, powerful states often seek to extract rents. They influence the shape of the rules to their advantage. Precedent in international law is one more instance of such travel along the Pareto frontier (Krasner Reference Krasner1991).