Erosion of International Organizations’ Legitimacy under Superpower Rivalry: Evidence on the International Court of Justice

Abstract This article investigates how superpower rivalry affects public perceptions of international organization (IO) legitimacy in the hegemon. We argue that the representation of a superpower rival state at an IO in the form of its key decision maker's nationality can dampen the IO's perceived legitimacy within the rival power. We test this argument using a survey experiment in the United States under President Trump, where we manipulate the nationality of the International Court of Justice (ICJ) judge who casts a tie-breaking vote against the United States. Our results show that when the judge is Chinese, there is a strong and robust dampening of Americans’ perceptions of the ICJ's legitimacy, with no comparable effect arising when the judge is from other countries, including Russia. Replication of the experiment in the United States under President Biden offers external validity for our findings, which may have important implications for the future of the liberal international order.


Introduction
Recent international relations literature has expanded significantly on the legitimacy of international organizations (IOs), which is considered fundamental for global governance (Binder and Heupel, 2015;Hooghe et al., 2017;Hurd, 2010;Zürn, 2018).Scholars have explored the sources of such legitimacy, which often revolve around a dichotomy between procedural-and performance-based conceptualizations (Dellmuth et al., 2019;Dellmuth and Tallberg, 2015;Hurd, 2007;Scharpf, 1999;Tallberg and Zürn, 2019).For those who emphasize procedural legitimacy, what matters is how IOs function-that is, whether or not the rules are transparent, democratic and fair (Dellmuth et al., 2019;Johnson, 2011).In contrast, accounts of performance-based legitimacy look at whether IOs are achieving their purposes and are fit for such purposes, as well as at the outcomes of their actions (Edwards, 2009;Gabel, 1998;Hooghe and Marks, 2005).
Nevertheless, few existing studies have explicitly investigated how superpower rivalry 1 affects IO legitimacy.As China rises to superpower status, it increasingly poses a challenge, or even creates an "existential struggle" (Freking and Knickmeyer, 2023), for the United States as the current hegemonic power.This dynamic leads to a situation where both superpowers compete for the dominant position within the existing international order (Onea, 2014).Most scholarly and public discourses on this topic today are suffused with accounts of intense competitions between the two rival superpowers in the economic and security spheres, yet another important arena that has received less attention is their influence upon global governance, as exemplified by contestation for leadership roles in IOs (Chhabra et al., 2021).
Indeed, there are already growing tensions between the two about staffing at IOs.Although US nationals continue to hold the most United Nations (UN) executive leadership positions, China is quickly catching up.As of April 2020, among senior leaders in the UN, there were 26 American nationals compared to 3 Chinese nationals, in addition to Chinese nationals heading the four specialized UN agencies (Feltman, 2021).China's expanded staffing presence in the UN has prompted the current US ambassador to the UN, Linda Thomas-Greenfield, to openly state that the United States will "resist China's efforts to overfill key UN positions with Chinese citizens" (Fung andLam, 2021: 1146).More recently, US Secretary of State Antony Blinken announced his government would aim to ensure that US and partner countries won elections in key international organizations and pushed back against those countries (clearly referring to China) who are undermining the integrity of the international system that the United States helped to build (Blinken, 2021).
Leadership positions in IOs are an integral component of a country's soft power and undoubtedly worth vying for in the larger context of geopolitical competition, but they also have important implications for the legitimacy of IOs in the court of public opinion.In particular, given that many of the existing IOs were established and endorsed by the United States, would the fact that a superpower rival state is heading one or more of these IOs lead citizens in the hegemonic state to view these organizations as illegitimate and unfit for their original purposes?
In this article, we put forward an argument that the representation of a superpower rival state at an IO in the form of its key decision maker's nationality can undermine the acceptance of the IO's mandate by citizens of the hegemonic state and that the effect will increase as the perceived threat of the rival state rises to the point of posing an "existential threat" to the hegemonic state.Furthermore, when the rival state is considered undemocratic and untrustworthy, the public in the hegemonic state is more likely to view the IO as undemocratic and its decisions as unfair.Consequently, the IO's perceived legitimacy in the hegemonic state will suffer because the performance and procedural sources of the IO's institutional legitimacy will be damaged.
Following existing studies on public reaction to international organizations (Chapman and Chaudoin, 2020;Chilton, 2014;Wallace, 2013), we examine this argument empirically using a survey experiment on Americans' perception of the legitimacy of the International Court of Justice (ICJ), a crucial international legal organization, in the context of superpower competition between the United States and China.In the survey experiment, implemented in August 2020 toward the end of the Trump presidency, we presented respondents with a hypothetical scenario in which the United States was sued at the ICJ for breaking international law, and the ICJ president cast the tie-breaking vote against the United States.Leveraging the increasingly confrontational bilateral relations between the United States and China at the time of the survey, we randomly assigned some respondents to view the additional information that the ICJ president was from China or from a number of other countries.
Our results show that when the judge is Chinese, there is a strong and robust dampening of Americans' perceptions of the ICJ's legitimacy, with no comparable effect arising when the judge is from other countries, including Russia.To assess the generalizability of these results, and in particular whether anti-Chinese sentiments could be driving the reduced support for an ICJ presided over by a Chinese judge, we conducted a replication study in the United States nine months later, following the inauguration of President Biden, and in the United Kingdom, a country that is not involved in superpower rivalry with China, despite mounting negative public sentiment toward China.Our results indicate that the effect size observed in the United States was similar to that of the original study, despite increasing anti-China sentiment, while the study in the United Kingdom produced null results.These findings provide further support for our theoretical argument and highlight how escalating superpower competition can erode public perceptions of IO legitimacy within the dominant state.They also demonstrate possible negative consequences of representational politics in IOs during superpower competition.
The article proceeds as follows.First, it reviews the different dimensions of IO legitimacy and presents the main argument that the perceived legitimacy of the IO by the public in the hegemonic state can be damaged by superpower rivalries.
The article then introduces the ICJ and its relations with the United States and explains why this prominent IO is relevant for studying the legitimacy of IOs more broadly.Next, the survey experimental design and results are presented and discussed.The article concludes with a reflection on the future of the international order in the context of intensifying superpower rivalry between the United States and China.

Different Dimensions of International Organizations' Legitimacy
The legitimacy of international institutions is fundamental to global governance.Whether an IO is perceived as legitimate has implications for how its rules and decisions are accepted or rejected (Hurd, 2019: 717).IOs with legitimacy find it easier to develop new rules and norms, regulate compliance, co-ordinate policies and solve global governance problems.As key dispute resolution/mediation venues, international institutions also play a crucial role in regulating relations among states (Hurd, 2007;Finnemore, 2009;Franck, 1990).For our purpose, we treat the legitimacy of an IO as the public's belief about whether the said international institution has the right to rule, or whether that authority is appropriate (Tallberg and Zürn, 2019: 585).We are particularly interested in what factors influence perceptions of IO legitimacy.We believe that such perceptions are crucial because they create the domestic conditions for national governments to comply with the rules and decisions of these IOs (Buchanan and Keohane, 2006;Dai, 2005).
There are different approaches to conceptualizing the sources of IO legitimacy, but much of the existing literature tends to emphasize the dichotomy between procedural and performance legitimacy for IOs (Dellmuth et al., 2019).Many scholars contend that the presence or absence of proper procedure influences the public's perception of IOs' legitimacy; IOs that function on the basis of more democratic principles (Johnson, 2011) and with more participation from civil society are viewed as more legitimate (Bernauer and Gampfer, 2013).The key variables here relate to the IO's procedures, including transparency, democracy and the scope and inclusiveness of participation.On the other hand, accounts of performancebased legitimacy emphasize that IOs should be evaluated by what they do-by how much they contribute to global welfare, help solve governance issues, and promote human rights and democracy around the world (Dellmuth et al., 2019;Edwards, 2009;Rohrschneider and Loveless, 2010).In this formulation, what IOs do-either in principle or in practice-matters most for their perceived legitimacy.
The conceptualizations of IO legitimacy based on these two dimensions of procedure versus performance understate other potential sources of legitimacy.Some scholars contend that the legitimation patterns of different IOs depend on their types of authority (Tallberg and Zürn, 2019: 593).Others have explored how elite communication by national governments, civil society organizations and IOs themselves can affect the public's perception of IOs' legitimacy (Dellmuth and Tallberg, 2021).Moreover, many international legal scholars have pointed out that international legitimacy can have normative foundations, as well as descriptive and sociological dimensions (Bodansky, 2012: 327).All of these arguments suggest that states and their representatives at IOs can also meaningfully shape perceptions of IO legitimacy.

Superpower Rivalry and IO Legitimacy
The liberal international order in place since the end of the Cold War, with the United States as the reigning hegemonic state, has allegedly been in crisis (Ikenberry, 2018;Mearsheimer, 2019).With China's rise, there have been perceptible anxieties within the United States about the challenges China poses to the existing international order and whether America's hegemonic position will endure (Acharya, 2014;Beckley, 2011;Brooks and Wohlforth, 2016;Johnston, 2019;Layne, 2012).Changing from a unipolar order to a bipolar one would have real consequences for the functioning of international institutions.In the not-so-distant past, for instance, the intense rivalry between the United States and the Soviet Union during the Cold War meant that activities of the UN, the only IO where both superpowers coexisted, were seriously hampered (Haviland, 1965;Mosely, 1965).Thus, if the current liberal international order is going to evolve into one that resembles the Cold War, the competitive superpower rivalry between the United States and China will inject a strong sense of uncertainty within existing IOs, which could affect their perceived legitimacy.
One area on which IO legitimacy hinges is the representation of countries in key leadership positions; these have been considered crucial in shaping global governance, as well as in how the public views who "controls" prominent IOs (Fung and Lam, 2021;Stone 2011).In other words, the nationalities of an IO's key decision makers matter for perceptions of the institution's legitimacy, either independent of or in addition to traditional conceptualizations of descriptive representation in domestic institutions.This is because the politics of national representation convey valuable information for audiences.Who is present at IOs on behalf of which countries transmits different messages.Indeed, studies have found that the United Nations Security Council (UNSC) can function as an information agent because of its diverse membership (Chapman, 2007;Fang, 2008;Thompson, 2006).As the decision-making process and voting record of most, if not all, IOs are made public, which states do what under what circumstances can have significant implications for how the outcomes are perceived.
We argue that superpower rivalry would influence public perceptions of IO legitimacy through an information mechanism transmitted by the nationalities of IO leaders.Because ordinary citizens generally lack knowledge about the workings of an IO, they take the nationality of the IO representative as an information shortcut in their evaluation of the IO.In contrast to an IO represented by like-minded states, when a superpower rival state is a key player in an IO decision-making body, citizens from the opposing country will likely question whether decisions made by the IO are intended to fulfil the interests of the rival state at the expense of their country's power and status rather than to achieve the stated goal of the IO.For example, it has been found that the US public is less receptive to vetoes at the UNSC by Russia and China over authorization of the use of force by the United States than vetoes made by US allies (Matsumura and Tago, 2019;Tago and Ikeda, 2015).Such effects on the IO's performance legitimacy will be stronger when the perceived threat from the rival state is higher, in which case, decisions made by the IO may be viewed as unjust and the mandate of the IO repudiated.
Two features of the rival state represented in the IO can further undermine perceived procedural legitimacy, especially for democratic audiences.First, when the rival state is an authoritarian one, the decision-making process of the IO may be called into question as undemocratic.The increased influence of authoritarian states in IOs has been identified as diluting the human rights components of such institutions, Interpol being one example (Lemon, 2019).Regional organizations founded and dominated by authoritarian states, such as the Gulf Cooperation Council and the Shanghai Cooperation Organization, have also been shown to help stabilize authoritarian rule and insulate against pressures for democratization (Libman and Obydenkova, 2018).
Second, when the rival state has a reputation for being untrustworthy, this may cast doubt on the fairness of a decision rendered by the IO, especially if the decision is unfavourable.Studies on US Supreme Court justices have shown that the ideological and political values of the justices can negatively affect how the court and its decisions are perceived by people with opposing values (Bartels and Johnston, 2012;Scheb and Lyons, 2001).Similarly, at the international level, studies have shown that the presence of many authoritarian states on the UN Human Rights Council affects how controversial proposals are presented and has led to its polarization (Hug, 2016;Hug and Lukács, 2014).Indeed, trust and mistrust have been pointed out as core features that defined the Cold War relational dynamic between the United States and the Soviet Union (Kydd, 2007).
The above theoretical discussions lead to one testable hypothesis: When IOs are represented by a superpower rival state that is undemocratic and untrustworthy, the IO's perceived legitimacy will likely suffer in the eyes of the hegemonic state's citizens, because the performance and procedural sources of the institution's legitimacy will be damaged.

Testing the Argument on the ICJ
We use a prominent international organization-the International Court of Justice-and its relationship with the United States to test our theoretical argument.Often called the World Court, the ICJ was founded in 1946 as part of the UN's mandate.Although the court was relatively quiet during the Cold War, it has become increasingly more active since then (Johns, 2015: 80).More importantly, the ICJ has been engaged by developing countries targeting more powerful states (Posner and de Figueiredo, 2005).Since the ICJ's founding, the United States has been sued by more than 10 nations.Currently, the United States is involved in two pending cases at the ICJ: (1) Palestine suing the US government for relocating the US embassy to Jerusalem and (2) Iran suing the US government for alleged violation of the 1955 Treaty of Amity, Economic Relations, and Consular Rights between the two countries.While the general public may not be familiar with the ICJ, major news outlets covered both cases and introduced them into the public discourse.
The ICJ has a panel of 15 judges elected through the United Nations General Assembly (UNGA) and the UNSC.There are specific rules about the judges' nationality (Posner and de Figueiredo, 2005: 603).Customarily, the five permanent members of the UNSC each have one slot, although currently, the United Kingdom does not have a judge at the ICJ.The rest of the slots are filled by continents/ civilizations, the idea being that the ICJ represents people of different races and different legal traditions (Lachs, 1992;Schwebel, 1999).As we discussed earlier, however, the nationalities of the judges should be a politically relevant factor for the ICJ's perceived legitimacy.
In principle, elected judges at the ICJ should act neutrally and interpret laws independently.Nevertheless, there is a suspicion that their decisions can be influenced by the countries they represent (Smith, 2004).As Joan Donoghue, the current US judge and president of the ICJ, aptly stated, "nationality has shaped the way we think because of the schooling we have, the kind of legal training we have, certainly our perspectives on issues are influenced by our nationality" (Donoghue, 2011).Although few studies have systematically tested this proposition, there is at least one that specifically addressed the issue of biased rulings at the ICJ.According to Posner and de Figueiredo (2005), ICJ judges do not decide on cases impartially.The authors found strong evidence that judges favour the states that appoint them and those whose wealth level is close to their own, in addition to weak evidence for the effect of regime types and cultural affinity.
When it comes to public perceptions of the ICJ, Americans have a generally positive view.According to a report by the Council on Foreign Relations, in the 2000s, a narrow majority of Americans (57 per cent) believed that the ICJ would rule fairly and impartially in cases involving the United States.The same majority (57 per cent) also believed the United States should grant compulsory jurisdiction to the ICJ (Council on Foreign Relations, 2009: 9).However, attitudes toward the court are heavily shaped by partisanship.Analyzing World Values Survey data, for example, Voeten (2013) found Democrats are nearly twice as likely as Republicans to be confident that an ICJ judgment on the United States would be fair and impartial.Voeten concluded that attitudes toward the court may be derivative of attitudes toward domestic institutions more generally.Our argument suggests that superpower rivalry may also drive such attitudes, including about the court's perceived legitimacy.

Experiment
The key observable implication of our theory is that representation of a rival superpower state at an IO will undermine the public's perception of its legitimacy in the hegemonic state.For an American audience, currently the most salient superpower rival with the greatest potential to challenge the United States is China, which has dominated policy and public debates in the recent past.Since the start of the trade war in 2019, bilateral relations between the two countries have been in free fall, entering a new phase "characterized by the primacy of competition and an augmented risk of conflict and confrontation" (Medeiros, 2019: 113).More recently, Mike Gallagher, chair of the newly created special House committee dedicated to countering China, framed the competition between the United States and China as "an existential struggle over what life will look like in the 21st century" (Freking and Knickmeyer, 2023).
More Americans are now viewing China as the number one threat to the United States.According to a recent Pew survey, the majority of Americans have unfavourable views of China and consider China's power and influence a major threat (Silver et al., 2020). 2 The perception of China's threat to American well-being has also been identified as a strong indicator of white Americans' support for Trump during the 2016 election (Mutz, 2018).
In addition, China is an authoritarian state with a ruling communist party, and the country has often been portrayed in America as untrustworthy.Titles such as "When China Cheats, America Loses" are bountiful in the media, and politicians across the political spectrum often emphasize China's tendency not to play by the rules.Commenting on how China's entry into the World Trade Organization (WTO) has affected the United States, for example, Donald Trump said, "I will not say the word 'cheated,' but nobody's cheated better than China, I will say that" (Fox, 2019).Mistrust toward China is thus a prominent feature in American political rhetoric and public sentiment.
All in all, for the American public, an ICJ with a judge from a country with these combined characteristics-a rival superpower state led by an authoritarian government perceived to be untrustworthy-will be viewed as less legitimate.The combination of these features, especially as an existential threat to the United States' hegemonic position, separates China from other traditional adversaries of the United States such as Iran, North Korea, and arguably even Russia.
We implemented the survey experiment in the United States in August 2020 through Prolific Academic. 3A total of 1,451 respondents completed the survey.Consistent with existing studies using opt-in samples (Huff and Tingley, 2015), our respondents were younger and more educated than the national average.While this is by no means a representative sample, according to the American Association for Public Opinion Research's guideline on nonprobability samples (Baker et al., 2013), it is appropriate for our purpose in establishing causal relationships between treatment conditions and outcome measures.Furthermore, public opinion researchers have demonstrated the external validity of drawing causal inferences from such online convenience samples (Berinsky et al., 2012;Boas et al., 2020;Coppock and McClellan, 2019).
The survey 4 began by introducing respondents to the ICJ and how it works, with the following paragraph: As you may know, when there is a dispute about whether a country is abiding by international law, the case is often tried in front of the International Court of Justice (ICJ), the principal judicial organ of the United Nations.It is comprised of fifteen justices from around the world, including one President and one Vice President.
Here is a picture of the current panel of judges (see SI-1 in Supplementary Material for the picture).
The court decides cases by majority rule.When there is a tie vote, the President of the court is entitled to cast the decisive vote to break the tie.Since 1945, when the ICJ was established, 178 cases have been entered onto the General List for consideration before the court.The United States has appeared, as either applicant or respondent, in 24 of these cases.comprehension checks to ensure that the respondents had properly processed the information presented to them. 5 After reading this background information, respondents answered two questions-one about the number of judges at the ICJ and the other about procedures for resolving tied votes (see SI-1 for the full survey questionnaire).Respondents who failed either question were given the same information as on the previous page to read one more time before proceeding.
In the next part of the survey, respondents were randomly assigned to one of two groups.Each group contained the same scenario, in which the United States was sued by another country at the ICJ for breaking international law, the court vote was tied at 7-7, and the president of the court cast the deciding vote against the United States.This scenario of a tie is realistic, as the president of the ICJ is mandated with "a casting power in the event of votes being equally tied" (ICJ, n.d.).While tie votes are uncommon, they have occurred multiple times in the ICJ's history. 6The eventual loss of the United States is necessary because "legitimacy is for losers"-that is, winners ordinarily accept decisions and thus have no reason to view the court as illegitimate (Gibson et al., 2014).
The difference between the control and treatment groups was the national identity of the ICJ president.In the control group, it was not mentioned, and respondents saw a picture of the entire ICJ panel. 7In the treatment group, respondents were told that the president of the court was a judge from China and were provided with a picture of the current Chinese judge serving at the ICJ.We also included a "placebo" group, in which the ICJ president identified and pictured was randomly drawn from five current judges from Jamaica, Japan, Russia, Uganda and the United Kingdom, representing a range of regime types and bilateral relationships with the United States. 8 Our outcome measure is respondents' perceptions of ICJ legitimacy.Here, we follow one of the more recent efforts to construct such a measure, proposed by Anderson et al. (2019).Adapting their measure of the legitimacy of the UN Climate Change Conferences, we asked survey participants whether they agreed, somewhat agreed, somewhat disagreed, or disagreed with the following four statements (the order of which was randomized): • The ICJ should continue to make rulings in the future.
• I believe the ICJ is necessary.
• I think the ICJ serves an important role in the world.
• I sympathize with the goals of the ICJ.
These statements build on the conceptual and theoretical literature on legitimacy and capture both dimensions of the definition of legitimacy described earlier: deference to the decisions of the institution (that is, the ICJ has the right to rule) and respondents' social affinity with the institution (that is, the goal or purpose of the ICJ is inherently desirable).
We then aggregated the responses for the four statements using item response theory (IRT), a common method to evaluate the psychometric properties of survey questionnaires with categorical responses (Raykov and Marcoulides, 2010).Since the answers are on an ordinal scale, we employed a graded response model to calculate the IRT scores, which represent respondents' perceptions of the ICJ's legitimacy (see SI-2 for details).Finally, we rescaled the resulting measure to be between 0 (lowest legitimacy) and 1 (highest legitimacy) for ease of interpretation.

Results
We start by reporting the descriptive statistics for the IRT legitimacy measures and the average treatment effects (ATEs).We then examine heterogeneous treatment effects, explore potential causal mechanisms, and perform additional robustness checks, including using alternative measures of legitimacy, restricting the sample to respondents who passed the comprehension questions, applying sampling weights, and accounting for censoring and truncation in the data with a limited dependent variable model.Finally, we report two additional replication studies conducted in the United States and the United Kingdom in April 2021 to assess the generalizability of the findings.

Main findings
Figure 1 shows the distribution of the IRT measure for the entire sample and for the three experimental groups.Focusing first on the full sample, we can see a large peak at close to 1, which indicates the highest level of perceived legitimacy.The 499 respondents in this group agreed with all four statements.A second, much smaller peak at around 0.5 represents respondents who somewhat agreed with all four statements.Values to the left of this second peak were respondents who disagreed with some or all of the statements.In other words, the majority of American respondents perceived the ICJ to be legitimate despite being told that the United States was sued for breaking international law and lost the case.
When looking at the effect of the ICJ president's nationality on perceived legitimacy, we observe no significant differences between the control and placebo groups.The shapes of the distributions are very similar to each other and to the one in the full sample.In the treatment group, however, the distribution of perceived legitimacy shifts to the left.The proportion of respondents in the high legitimacy group is down by nearly half.At the same time, more respondents view the ICJ as not legitimate, indicated by the fatter tail to the left.
Given the randomized design of the experiment, we rely on ordinary least squares (OLS) regressions with robust standard errors to estimate the ATEs.To increase the precision of the estimates and to account for potential imbalance across the experimental groups (see SI-3 for balance check), we control for the following pretreatment covariates: a binary female/male gender variable, age in years, education level on a 12-point scale, and income on a 12-point ordinal scale.
The first two lines in Figure 2 display the ATEs of the treatment and placebo groups on perceived ICJ legitimacy.While the difference in perceived legitimacy between the control and placebo groups is indistinguishable from 0, the Chinese judge treatment reduces perceived legitimacy by 7.4 percentage points.The difference is statistically significant ( p < .001),although the substantive effects differ depending on the baseline level of perceived legitimacy.For respondents who already view the ICJ as highly legitimate (that is, they agree with all four statements), the treatment does not change their beliefs.For those respondents in the second peak of Figure 1, who somewhat agree with all four statements, knowing that the Chinese judge cast the deciding vote against the United States will flip their responses to one or two of the four statements from "somewhat agree" to "somewhat disagree."In other words, the effect is stronger for those less convinced of the court's legitimacy to begin with.
We can further disaggregate the placebo group with respect to the other five judges.The results in Figure 2 indicate no statistically significant differences among these judges or relative to the control group.In particular, perceived ICJ legitimacy does not suffer when the judge is from Russia, a country often associated with China as another external threat to the United States and that to varying degrees shares many of the same features with China.Arguably, though, Russia is missing the superpower status capable of challenging the United States' hegemonic position.The null effect of the Russian judge thus lends support to our argument that IO legitimacy may suffer only in the case of a superpower rival state.

Heterogeneous treatment effects
While we do not have a priori expectations regarding heterogeneous treatment effects, it could be of interest to see whether the treatment has an effect on some subpopulation defined by observed characteristics.More practically, we can more confidently claim the external validity of the findings if observable features that distinguish our online sample from the rest of the US population do not moderate the treatment effects on perceived legitimacy.
To see whether this is indeed the case, we estimate a series of OLS regression models that include multiplicative interaction terms between the treatment and pretreatment socio-demographic variables, including age, gender (female), education, income, and party affiliation (see SI-10 for summary statistics).Table 1 shows that none of the six interaction terms between the treatment and the potential moderators is statistically significant, indicating no evidence of treatment-effect heterogeneity.This also suggests that the main findings are likely to generalize to the broader population.

Causal mechanisms
In the survey, immediately after the legitimacy questions, we followed up with three questions that tapped into respondents' perceptions of the performance and procedural sources of ICJ's legitimacy. 9As with the questions on legitimacy, we asked respondents whether they agreed, somewhat agreed, somewhat disagreed or disagreed with the following statements (the order of these statements was randomized): • The ICJ should not be allowed to tell the United States what to do.
• I find the ICJ ruling to be fair.
• The ICJ is undemocratic.
We used these questions to explore the causal mechanisms through which the knowledge of the United States' defeat at the ICJ at the hands of a Chinese judge reduced the perceived legitimacy of the court.We use causal mediation analysis to decompose the total effect of the experimental treatment on perceived legitimacy into the average direct effect (ADE)-the portion of the treatment effect not mediated through the mediatorand the average causal mediation effect (ACME).The mediation models combine the control and placebo groups into a single category and include the aforementioned pretreatment covariates (Imai et al., 2011).The results are presented in Figure 3.
We focus on the ACME, which is the portion of the treatment effect attributable to the change in the value of the mediating variable stemming from the treatment.The statistically significant estimates of the ACME can be interpreted as the extent to which perceived legitimacy would change in the control and placebo groups if we could set the average value of the mediator in question to the average value of what we would have observed in the treatment group.For example, if we could set the perceived fairness of the ICJ ruling among respondents in the control and placebo groups to what we would have observed in the treatment group, perceived legitimacy would have decreased by about 2.5 percentage points (−0.025 in the range of 0 to 1 in Figure 4).
The ACMEs of "fairness" and "democracy" are statistically significant and account for 34 per cent and 25 per cent of the total effects, respectively.The ACME of "mandate" is statistically significant at the 0.10 level (see SI-4 for more details) and accounts for 24 per cent of the total effect.Overall, the results of the causal mediation analyses provide evidence that the treatment effects are mediated by the erosion of respondents' faith in the performance and procedure of the court, the two mechanisms suggested by our theory.
An important identification assumption of the causal mediation analysis is sequential ignorability, which may or may not hold for the mediators, as they were not randomly assigned in the experiment.To test the sequential ignorability assumption, we performed the sensitivity analysis proposed by Imai et al. (2011), which evaluates the robustness of the findings from causal mediation analysis to the possible existence of unmeasured pretreatment variables that confound the relationship between the mediator and the outcome variable.The results of the sensitivity analysis, detailed in SI-4, indicate that the conclusion is plausible even in the presence of fairly large departures from the ignorability of the mediator.

Robustness checks
We performed a number of additional robustness checks.First, we replaced the IRT measure of legitimacy with alternative measures, using a simple additive method and principal component analysis.The Cronbach's alpha among the three measures is 0.9, indicating a high level of internal consistency.Second, we excluded respondents who failed the comprehension questions.Slightly over 70 per cent of the respondents answered both questions correctly, which reduces the sample size to 1,028.As detailed in SI-5 and SI-6, alternative measures of legitimacy and reduced sample both yield results similar to those in the main finding.Third, we re-estimated the treatment effects using poststratification weighting to remedy the differences between our sample and the national census on gender, age, education and race.The results, reported in SI-7, are also similar to those presented in Figure 2. Finally, because the IRT measure is bounded by 0 and 1 and is left-skewed (see Figure 1), we estimated a limited dependent variable model (Tobit) to account for the censoring and truncation in the data.The results can be found in SI-8.Once again, our conclusion remains unchanged.

External validity
Because our survey experiment was launched in August 2020 toward the end of the Trump presidency, at the height of the COVID-19 global pandemic and when anti-Chinese sentiments were high, one might wonder whether our findings would hold if we ran similar experiments at a different time or in a different country.In a bid to answer this question, we replicated the original experiment in the United States and the United Kingdom simultaneously in April 2021. 10While only nine months separated the two surveys in the United States, the contexts had shifted quite dramatically.The country was under a new administration, with President Joe Biden having pledged to take action against anti-Asian bias and violence (The White House, 2021).
The US experiment was fielded to 659 Americans, who were recruited by Lucid (Coppock and McClellan, 2019) using proportional sampling on gender, age, race and geographical location.The UK experiment was administered to 667 adults, who were recruited by Prolific Academic.We used the same experiment design in the US survey.In the UK experiment, we changed the wording to indicate that the United Kingdom was being sued for breaking international law (see SI-9).
Figure 4 summarizes the ATEs in both experiments.Focusing on the US experiment first, we can see that compared to the control group, the Chinese judge significantly reduces the ICJ's perceived legitimacy by the same margin as in Figure 2. The results are robust to the inclusion of pretreatment covariates and a reduced sample with attentive respondents only.Remarkably, the effect size remained after nine months, despite rising anti-Asian crimes (Yam, 2022).This helps boost our confidence that the findings may not be driven by racial prejudice.
In the UK experiment, we do not see any differences in perceived legitimacy between the control and treatment groups.This is expected, as our theory posits that the public will find the ICJ less legitimate if it is presided over by a judge from a superpower rival state.Even though public unfavourability toward China has similarly reached new heights in Britain, the bilateral relationship between the two countries is not the same as the superpower competition between China and the United States.In fact, a recent study shows that while Britons consider China's rise disconcerting, they are also pragmatic in their understanding of how these bilateral relations should be managed and are supportive of expanding trade and investment between the two countries (Chow et al., 2019).

Conclusion
This article makes a novel contribution by bringing together the literature on IO legitimacy with superpower rivalry.Using a survey experiment on the ICJ, we show that a superpower rival state-in this case, China-presiding at the ICJ undermines the American public's perception of the court's legitimacy.Nevertheless, there are a few caveats when interpreting our findings.First, it is perhaps going to be much easier to shift the American public's attitudes by having a Chinese judge who breaks a tie to rule against the home country than by simply having a Chinese national sitting on the board of a major IO.Furthermore, the American public in general is probably not particularly familiar with the workings of the ICJ, and their perceptions of different IOs are also subject to particular media narratives within the United States at a given time.Therefore, one needs to exercise caution when extrapolating our findings to other IOs.Still, there are reasons to believe that the stakes could be higher for IOs in the security and economic spheres, particularly those with voting rules that grant veto power to the rival superpower.
It is important to note that the dynamics of superpower rivalry between the United States and China are likely unique and thus may not be generalizable to other geopolitical rivalries.On the other hand, we can expect the argument to apply to the Chinese public concerning the United States' control over key IOs, such as the World Bank, which could potentially undermine the legitimacy of these institutions in the eyes of Chinese citizens.In fact, one of the Chinese government's common complaints is how "biased" the US-led international order is, especially in those Bretton Woods institutions where Washington holds a significant amount of agenda-setting power (Wade, 2011).For example, a recent document titled "US Hegemony and Its Perils," published by China's Ministry of Foreign Affairs in February 2023, claimed that the United States has "established institutional hegemony in the international economic and financial sector by manipulating the weighted voting systems, rules and arrangements of international organizations" and "using its control over international organizations [to] coerce other countries into serving America's political and economic strategy" (Ministry of Foreign Affairs, 2023).
While Chinese people in general have a high degree of trust in international organizations such as the UN and the World Health Organization (WHO), ranking the highest among 27 countries in a recent survey (Edelman, 2022), there is also evidence of a growing skepticism toward the United States.In fact, 66 per cent of Chinese respondents in a recent study stated that the United States is playing a less important role in the world compared to 10 years ago.Furthermore, nearly half of the respondents expressed a preference for a world led by China in the future (Fang et al., 2022).Therefore, it is reasonable to assume that the legitimacy of certain IOs in the eyes of the Chinese public might be influenced by their association with the United States, especially in the context of the escalating superpower rivalry between the United States and China.Investigating this aspect could be a valuable direction for future research.
More broadly, what we find in this study prompts us to ponder the consequences of an intensifying China-US competition for the existing international order.As Weiss and Wallace (2021) point out, how the liberal international order can survive depends on domestic consensus within the United States on preserving the existing order.But such a consensus is currently crumbling in the context of the perception that China is benefiting from the system at the expense of the United States.The fact that Americans view an ICJ presided over by a Chinese judge as less legitimate does not bode well for other international institutions-many of which include China as the key member-or the liberal international order built upon these institutions.
Indeed, as witnessed during the COVID-19 pandemic, the Trump administration's response to that international public health crisis was to announce the withdrawal of the United States from the WHO, accusing it of being a "puppet" of the Chinese government.In a similar vein, partly fuelled by the widely held belief that China manipulates WTO rules to promote its state capitalism, the United States has blocked the appointment of new Appellate Body judges to the WTO, resulting in the paralysis of its dispute settlement system.
Even under the new Biden administration, tensions between the two countries have not shown any clear signs of easing, and Americans continue to view China with hostility, demanding a tougher approach toward handling the bilateral relationship (Silver et al., 2021).With discussions of a looming new Cold War between the two countries, much is at stake, and the future depends upon how both China and the United States manage the challenges and pressures they face from each other and from their own populaces.
8 These judges were selected for mundane realism.Countries such as Iran and North Korea, while theoretically relevant for our argument, are unlikely to have a judge at the ICJ because such an appointment is unlikely to be approved by the UNGA and the UNSC.The vignette in its entirety, including the picture and identity of the judge, constitutes a hypothetical scenario written in the subjunctive mood.As such, while none of the judges used in the experiment was the president of the ICJ at the time of the survey, the alteration was deemed to be of minimal risk to the survey participants based on feedback from and consultation with the institutional review board.9 We did not implement the recommendation by Chaudoin et al. (2021) to randomize the order of the mediation questions to be either before or after the outcome questions, because we did not have a priori expectations with respect to the distribution of the treatment effects or the relative influence of the mediators, which may be affected by the survey design (Chaudoin et al., 2021).Importantly, they show that the overall treatment effects remain unaffected regardless of the order of the mediation and outcome questions. 10 Due to resource constraints, we did not include the placebo groups in the replication studies.

Figure 1 .
Figure 1.Perceptions of ICJ legitimacy Note: Distribution of perceived legitimacy in the full sample and the subsamples.

Figure 2 .
Figure 2. Average treatment effects on ICJ legitimacy Note: Results based on an OLS regression model with robust standard errors adjusted with pretreatment covariates on age, gender, education and income.The dependent variable is the IRT measure of perceived ICJ legitimacy.Vertical lines indicate 95 per cent confidence intervals.Point estimates represent ATEs relative to the control condition.The number of observations in the underlying model is 1,390 due to missing values in the covariates.

Figure 3 .
Figure 3. Causal mediation analysis Note: Dots indicate the coefficients on the total, mediation and direct effects of the treatment on perceived legitimacy, with 95 per cent confidence intervals.The three panels each represent one mediator variable regarding the institutional source of ICJ legitimacy.

Figure 4 .
Figure 4. Average treatment effect in the replication studies Note: Point estimates based on an OLS regression model represent ATEs relative to the control condition.The dependent variable is the IRT measure of perceived ICJ legitimacy.Horizontal lines indicate 95 per cent confidence intervals.The sample sizes without covariate adjustment for the control and treatment groups are 333 and 326 in the US experiment, and they are 346 and 321 in the UK experiment.