Race, Representation, and the Legitimacy of International Organizations

Abstract

This study explores how race impacts the legitimacy of international organizations (IOs). Specifically, we examine whether the representation of Black people in IO leadership positions influences perceptions of IO legitimacy among Black and white individuals. To do so, we fielded seven survey experiments in two racially diverse countries, South Africa and the United States, and three experiments in one predominantly Black country, Kenya. Our experiments were designed to distinguish the effects of an IO leader's race from their region of origin. We find that Black IO leadership enhances perceptions of institutional legitimacy among Black citizens, but does not strongly influence the legitimacy perceptions of their white counterparts. Our findings suggest that improving the representation of historically marginalized racial groups within IOs can enhance their popular legitimacy.

International organizations (IOs) face criticism for underrepresenting marginalized racial groups.¹ Activists assert that the near-complete exclusion of Black-majority countries from the G20 reflects “institutional racism.”² The International Monetary Fund and World Bank are disparaged for maintaining weighted voting structures that contribute to the underrepresentation of Black-majority countries.³ The exclusion of certain racial groups from IO leadership positions has also stoked discontent. In 1981, when a popular Black candidate for the position of UN secretary-general was rejected in favor of a white candidate, a prominent African-American newspaper concluded that the UN “should be disbanded.”⁴

As these examples illustrate, race is one dimension through which people evaluate IOs. Here, we examine whether Black representation in IO leadership positions influences the popular legitimacy of these organizations. Between 1945 and 2005, across twenty-five major worldwide IOs, just four out of 145 leaders were Black (see Appendix A, in the online supplement). However, since 2017, Black individuals have ascended to top positions at many prominent IOs, including the International Criminal Court, International Labour Organization (ILO), World Health Organization (WHO), and World Trade Organization (WTO).

Building from research on descriptive representation in domestic institutions, we posit that the racial identity of an IO's leader influences the popular legitimacy of that IO. We hypothesize that Black leadership improves IO legitimacy among Black citizens, regardless of their country or region of origin. By contrast, the race of an IO leader is expected to have a weaker, and potentially even detrimental, effect on IO legitimacy perceptions for white individuals.

To test these claims, we fielded seven survey experiments that randomly assigned information on IO leaders’ race to self-identified Black and white respondents in South Africa and the United States. We also conducted three experiments in Kenya, a country with a predominantly Black population. The experiments, which cover the WTO, WHO, ILO, and one hypothetical IO, manipulate respondents’ perceptions of the IO leader's race without altering beliefs about the leader's region of origin. We find that providing information that an IO leader is Black significantly enhances perceptions of institutional legitimacy among Black individuals in all three countries; it does not strongly influence the perceptions of their white counterparts.

Our study responds to calls from international relations scholars to examine the role of race,⁵ and to more specific suggestions to “analyze when and to what extent racial inclusiveness improves the [liberal international order's] legitimacy.”⁶ Studies in the emerging field of race and international relations demonstrate that race impacts nations’ foreign policy actions⁷ and public attitudes on a range of foreign policy issues.⁸ Research in this area is beginning to examine whether IOs’ behavior depends on the racial composition of their member states and staff.⁹ Most closely related to our study, Chow and Han establish that Americans have more positive views of IO decisions made by racially diverse (rather than all-white) panels.¹⁰ Our study's contributions are to highlight the importance of IO leaders’ racial identity; empirically disentangle the effect of leaders' race from that of leaders' nationality; show that leaders’ race shapes the legitimacy of important international institutions; and demonstrate that this effect varies across racial groups.

These findings also contribute to scholarship on international institutions more generally. Research shows that public perceptions of IO legitimacy are shaped by individual respondent characteristics such as gender, age, and education,¹¹ as well as IO attributes.¹² Our evidence advances this literature by revealing that an IO leader's ascriptive features are an important, but previously overlooked, attribute that can influence their legitimacy. We also demonstrate that the effects of IO-level features (such as who leads the institution) and individual-level attributes (such as a citizen's race) on legitimacy perceptions are contingent on one another. Finally, we provide new insights on the importance of IO leaders. While prior studies establish that leaders affect IOs’ policies and influence,¹³ we show that leaders also matter because they shape how people perceive IOs.

Racial Representation and International Organization Legitimacy

International institutions are legitimate when they are widely believed to have the “right to rule”—the authority to craft, promote, and enforce rules within an issue area.¹⁴ We extend the scholarship on descriptive representation within domestic institutions to IOs. The central claim we test is that an IO leader's race influences perceptions of the organization's legitimacy, particularly among individuals from historically underrepresented racial groups. We focus here on the representation of Black people in IO leadership positions.

Race is a social construct in which human beings are classified by their phenotypical features, such as skin color and hair texture.¹⁵ Racial differences are not based on essential biological differences between groups but reflect social structures at the global and national levels. Because social structures differ between countries and over time, racial identities are not consistent across all social contexts.¹⁶ On the other hand, the global and historical nature of anti-Black racism, from the transatlantic slave trade to modern-day discrimination, has generated a fairly high degree of consensus on who is and is not considered “Black.”¹⁷

Many Black voters in the United States and other racially diverse countries form their opinions about politics by considering how an issue impacts the well-being of their racial group.¹⁸ Concerns about the group's status generate a preference for Black representation. A large body of research shows that racial minorities in the United States favor “descriptive representation” within domestic institutions. For example, Black voters are more supportive of Black than non-Black legislators in the United States¹⁹ and Brazil.²⁰ The impact of descriptive representation extends beyond assessments of individual representatives: Black representation within domestic political institutions improves Black Americans’ assessments of these institutions as a whole,²¹ and increases political trust and efficacy.²²

There are reasons to expect that Black citizens will also perceive international institutions with Black leaders as more legitimate, given the transnational nature of racial identity. For example, leading Black intellectuals and activists—like W.E.B. Du Bois and Malcolm X in the United States, Marcus Garvey in Jamaica, and African independence leaders like Kwame Nkrumah and Sékou Touré—have long emphasized that global white supremacy has produced similar experiences for people of African descent regardless of their country of residence.²³ Similarly, the Black Lives Matter movement pledges “solidarity with our international family against the ravages of … anti-Black racism.”²⁴

Consistent with these elite positions, many Black citizens view their relevant identity group as extending beyond national borders. Survey-based research from the United States indicates that many Black Americans feel close to their co-racials in Africa and are concerned about the welfare of Black communities around the world.²⁵ Ethnographic research from Ghana suggests that Ghanaians feel a “personal connection to diaspora Blacks that they insisted were long-lost ‘brothers,’ ‘sisters,’ or ‘cousins.’”²⁶ Given this sense of shared racial identity, Black people are likely to prefer Black-led IOs, even if the leader does not share their national identity.

There are at least three reasons that we expect members of the Black community to view IOs with Black leaders as more legitimate. First, scholarship on representation at the domestic level finds that the presence of representatives that share one's racial heritage enhances the perception that an institution is procedurally fair, inclusive, and trustworthy.²⁷ Thus, members of marginalized racial groups may also view international institutions as fairer organizations if they are led by in-group members. Second, an IO leader's race may influence public perceptions of the competence of those running the organization and, therefore, its ability to achieve its objectives. Black Americans believe that Black congressional candidates are more knowledgeable and intelligent and possess stronger leadership skills than white congressional candidates,²⁸ and they may view Black IO leaders similarly. Third, the race of an IO leader may influence public beliefs about whether that organization serves their interests. Research on descriptive representation within domestic institutions finds that Black representatives are perceived as more likely to work to promote the interests of the Black community.²⁹ Likewise, citizens should anticipate that Black IO leaders will shift their organization's policies in a direction that benefits communities of color. In short, Black individuals are likely to view Black-led IOs as more legitimate than IOs headed by individuals belonging to other racial groups because these organizations will be viewed as fairer, more competently run, and more likely to deliver positive outcomes for members of their racial group.

Anecdotal evidence conforms to these expectations. For example, in 1981, a majority of UN member countries favored appointing Salim Salim, a Black diplomat from Tanzania, as the institution's secretary-general. Prominent African American leaders also endorsed Salim.³⁰ When US president Ronald Reagan vetoed his appointment, some charged that the UN had “shown itself to be systematically white.”³¹ A decade later, when the position was vacant again, African countries insisted it was their turn to fill the post and nominated six candidates. Egypt's Boutros Boutros-Ghali, the only non-Black candidate of the six, was selected. This outcome left African diplomats “seething furiously” because they “want[ed] to have a black African Secretary-General.”³² Similarly, the recent appointment of Ngozi Okonjo-Iweala as the first Black director-general of the WTO was widely and positively covered by Black American media outlets.³³ It generated significant public interest in Black-majority countries, as evidenced by a substantial spike in Google search volume about the WTO within these countries following the announcement (see Figure B1 in the online supplement).

While we expect Black citizens to respond favorably to Black IO leadership, it is less obvious how white citizens—as the most politically dominant racial group globally—will respond.³⁴ One possibility is that Black leadership reduces an IO's perceived legitimacy among white individuals because they view Black IO leaders as a threat to their group's status or interests,³⁵ or believe Black IO leaders are less competent than their white counterparts.³⁶ On the other hand, an IO leader's race may have little effect on their perceptions of IO legitimacy. Since white people are less likely than Black individuals to believe their personal fates are connected to their racial group's status, their evaluations of IOs may not be sensitive to the race of these organizations’ leaders.³⁷ A third possibility is that white people will view IOs with Black leaders as more just organizations, giving them more legitimacy. In sum, different theories provide contrasting expectations for how white individuals will respond to Black representation in IOs.³⁸ Overall, we anticipate that the presence of a Black leader will not increase the perceived legitimacy of an IO as strongly for white people as for Black people.

This discussion implies two testable hypotheses:

H1: Black citizens perceive IOs with Black leaders as more legitimate than IOs with non-Black leaders.

H2: Black leadership in IOs has a stronger positive effect on IO legitimacy for Black citizens than for white citizens.

Research Design

We use survey experiments to examine how Black representation influences IO legitimacy. Our first battery of experiments was fielded in two racially diverse countries—one with a majority-Black population in Africa (South Africa), and one outside Africa where Black citizens constitute a minority of the national population (the United States).³⁹ The two main surveys, conducted in August to September 2021, sample two groups: self-identified white and self-identified Black individuals. Our main samples, which include around 2,000 respondents in South Africa and 3,000 in the United States, are split nearly evenly between these two racial groups to ensure sufficient statistical power. We also fielded two additional follow-up surveys, one on a smaller sample of Black Americans in January 2022 (N = 1,200), and one on a sample of Black and white Americans in February 2023 (N = 3,050). Across these four surveys, we fielded seven experiments that covered three different real-world IOs and one hypothetical IO. Finally, we repeated three of these experiments in Kenya in February 2024 (N = 1,000). Table 1 summarizes the timing and location of our experiments.

Table 1. Summary of experiments

Dynata, an international survey research firm, administered the surveys. Dynata sent out invitations to its pool of respondents to complete the survey online in return for a small compensation. Respondents who failed more than one attention-check question or other quality check were excluded from the sample. Each survey included demographic quotas to help attain a broad and representative sample. Given the relatively low rates of internet penetration in South Africa, despite including quotas for gender, age, and education, the sample over-represents young, highly educated, and female individuals (see Appendix C).

The main US survey included quotas for age, gender, education, and whether the respondent lives in an urban, suburban, or rural area. Our US sample is obviously not representative of the national population. But within each racial group, it reflects the broader population from which it is drawn: both the white and Black American samples largely match the population of white and Black people in the United States along these dimensions (see Appendix C).⁴⁰

Each of our experiments randomly assigned information about the race of an IO leader to respondents. Our research design draws on the experimental approach commonly used in studies of descriptive representation within domestic institutions.⁴¹ However, analyzing the effects of representation at the international level faces a distinctive challenge: the difficulty of disentangling the effects of an IO leader's race from their region of origin. Most of the world's Black population lives in Africa, and virtually all Black IO leaders are African. Hence, informing survey respondents of a leader's race runs the risk of unintentionally altering their perceptions of the leader's region of origin as well. This “information equivalence” problem could bias our estimate of the effect of race.⁴² For instance, it is possible that Black South Africans would respond positively to a leader described as “Black” not because they care about the leader's race but because they infer that he or she is African and they favor increased African regional representation.

We adopt two complementary strategies to address this empirical challenge. First, our conjoint experiment randomly assigns information on the race and nationality of a hypothetical IO leader independently. Second, we focus on real-world IOs with Black leaders who have dual citizenship, allowing us to more cleanly distinguish the effect of Black leadership from that of African leadership. Focusing on actual IOs permits respondents to consider both the experimental stimuli and their priors about an IO, which enhances external validity.⁴³ The next section describes our focal experiment in detail; subsequent sections present the other experiments.

The WTO Experiment

In 2021, Ngozi Okonjo-Iweala became the first Black leader of the WTO, the main IO responsible for promoting an open global trading system.⁴⁴ Beyond its substantive importance, the WTO case enables us to design an experiment that distinguishes the effect of race from that of nationality. Okonjo-Iweala has both Nigerian and American citizenship. Her dual nationality, and the stark differences in the share of governing elites in Nigeria and the United States that identify as Black, allow us to better manipulate subjects’ perceptions of the WTO leader's race (our central focus) without changing perceptions of her nationality (a likely confounder).

Experimental Design

This experiment includes four conditions. The control group sees a brief description of the WTO that does not mention its leader's race or nationality: “The World Trade Organization (WTO) is an international institution that deals with the rules of trade between nations. Recently, the WTO appointed a new leader.” For the second group, “who is American” is added to the end of this description. In the third condition, the leader is “a Black American”; and in the fourth, “a Black African.”⁴⁵

This design permits us to examine the effect of Black leadership while holding constant (perceptions of) the leader's region of origin. We test the effect of race by comparing responses in the “American” and “Black American” conditions. We anticipate that many respondents in the “American” condition will not perceive the WTO leader as Black; thus, perceived race should be the only difference between these two conditions. To test the effect of regional representation, we compare the “Black American” and “Black African” conditions.

To measure IO legitimacy perceptions, we ask, “How much confidence do you personally have in the WTO?”—with answers on an eleven-point scale from “no confidence” (0) to “complete confidence” (10).⁴⁶ This question helps capture peoples’ general faith in an organization and whether they view it as a legitimate authority.⁴⁷ And it has become the standard question in scholarship on IO legitimacy,⁴⁸ which helps in comparing our results with previous findings.

Manipulation Checks

The treatments in the WTO experiment manipulated perceptions of race and nationality in the manner expected (Figure 1). On the survey page following the main experiment, respondents were asked, “From what you understand about the World Trade Organization (WTO), which of the following attributes do you think describe the WTO's leader? Please do not seek out new information to answer these questions. We are interested in your best guess based on what you already know.” (See Appendix C for details.) The top panels of Figure 1 plot the proportion of respondents that believe the WTO leader is Black across the four experimental conditions. This proportion is much higher in the “Black American” and “Black African” conditions than in the “American” and control conditions. Perceptions of race are very similar in the “Black American” and “Black African” conditions.

Figure 1. Manipulation checks for WTO experiment

The bottom panels of Figure 1 display the proportion of respondents in each country that believed the WTO leader is co-regional (Nigerian in the South Africa survey, American in the US survey). Perceptions of nationality differ dramatically in the “Black African” versus the “Black American” and “American” conditions. They are nearly identical across the latter two groups.

These patterns in perceptions of the race and region of origin of the WTO leader across experimental conditions suggest that our treatments worked as intended. Since subjects in the “American” and “Black American” conditions hold similar views on the WTO leader's nationality, but vastly different beliefs about her race, comparing the two groups provides an effective test of the impact of an IO leader's racial identity. Yet, it is important to acknowledge that a non-negligible share of respondents who received information about race and region did not accurately recall it. As we discuss later and in Appendix D, this is largely explained by a lack of attentiveness from parts of the subject pool.

Average Treatment Effects

This section tests our two main hypotheses about how Black IO leadership shapes public confidence in IOs. Figure 2 presents the mean levels of confidence in the WTO, along with 95% confidence intervals, across the four experimental conditions. Recall that we compare the “American” and “Black American” conditions to assess the effect of the WTO leader's race while fixing perceptions of the leader's nationality. The patterns are consistent with our theoretical expectations. Among Black respondents in South Africa and the United States, mean levels of confidence in the WTO are noticeably higher in the “Black American” condition than in the “American” condition. The difference is smaller in the white subsamples. It is also notable that mean levels of confidence in the “Black African” and “Black American” conditions are very similar for all four subgroups, suggesting that legitimacy perceptions are more closely connected to the race of an IO leader than to their region of origin (see also Figure C2 in the online supplement).

Figure 2. Confidence in the WTO

To more rigorously test our two main hypotheses, following our pre-analysis plan, we fit ordinary least squares regression models in which the regressors are indicators of each respondent's experimental condition, self-identified race, and the interaction between the two. Based on this output, Figure 3 displays the effect of Black IO leadership by comparing the difference between mean levels of WTO confidence in the “Black American” and “American” conditions.

Figure 3. Effect of race on confidence in WTO

The evidence in Figure 3 provides considerable support for H1. In both countries, Black individuals’ confidence in the WTO is substantially higher when its head is described as a “Black American” versus just “American.” The difference is about 0.6 points (out of 11) among Black South Africans and 0.7 points among Black Americans. This effect is statistically distinguishable from zero with high confidence in both samples (p < 0.01). We are also able to conclude, based on our experimental design, that these differences reflect information about race rather than region of origin.

We find some support for H2, that an IO leader's race matters less for white individuals than for Black individuals. White South Africans report similar levels of confidence in the WTO in the “Black American” and “American” conditions (the mean difference is 0.13). Among white Americans, average confidence in the WTO in the “Black American” condition is moderately higher than in the “American” condition, and this difference is marginally significant (p < 0.10). However, this effect (0.34) is just half the size of the effect among Black Americans. The bottom row in Figure 3 plots the difference in effect size across the two racial groups (the effect among white subjects minus the effect among Black subjects). This difference is negative, as expected, but falls short of conventional significance levels (p < 0.10 in South Africa; p = 0.16 in the United States).

The positive effect of Black IO leadership among white Americans is likely driven by at least two factors. First, it may reflect social desirability bias among white survey respondents, an issue we revisit later. Second, the positive effect of the “Black American” treatment appears to be driven by the reactions of one subgroup of white Americans—those on the political left. In exploratory analyses presented in Appendix D, we find that white Americans with more left-wing, or progressive, political views respond favorably to the presence of a Black IO leader, while conservative white Americans do not.⁴⁹

In sum, the evidence strongly supports our hypothesis that the race of an IO's leader influences Black citizens’ confidence in the institution.⁵⁰ We also find suggestive evidence for our hypothesis that a leader's race has more muted effects among white individuals, though the differences between Black and white respondents were not as large or precisely estimated as anticipated.

Addressing Potential Threats to Validity in the WTO Experiment

We next subject our main findings to a range of robustness checks. The results of these tests increase our confidence in the internal and external validity of our initial results.

Non-compliance with Experimental Treatments

If some respondents did not comply with our treatments, the “intent-to-treat” effects just presented might provide biased estimates of the average treatment effects.⁵¹ Two forms of noncompliance may be present. First, a non-negligible fraction of respondents who were told that the WTO has a Black leader did not accurately answer our manipulation-check question about the WTO leader's race (see Figure 1). This suggests that some subjects assigned to the treatment condition did not receive the treatment. Second, respondents not assigned to the treatment condition may have had prior knowledge that the WTO's leader is Black, thus receiving the treatment nonetheless.

To address these concerns, we estimate complier average causal effects (CACEs), which represent the average treatment effect for the subset of respondents known as compliers, defined as those who received the treatment because they were assigned to the treatment condition. Following Angrist, Imbens, and Rubin, we use instrumental variable regression to estimate CACEs, using treatment assignment as an instrument for a respondent's beliefs about the leader's race.⁵² This approach quantifies how a treatment-induced change in beliefs about the WTO leader's race influences confidence in the WTO. Appendix F provides more details on this analysis and presents our results: learning that the WTO leader is Black has a very large positive effect on Black citizens’ confidence in the WTO. We estimate CACEs of 1.2 in South Africa and 3.0 in the United States. These effects are large in both an absolute sense and relative to the effects of other variables. In each country, learning that the WTO leader is Black has a larger impact on Black peoples’ attitudes to the WTO than any other variable in the model, including education, age, gender, income, and partisanship. The CACEs are much smaller for self-identified white respondents.⁵³

Potential Confounders

Another concern is that we may not be capturing the moderating effect of a respondent's race but rather that of some other variables that correlate with race. While the information on IO leaders’ race is randomly assigned, the racial identity of our respondents is not. To address this concern, we estimate models controlling for several demographic variables that correlate with race and potentially influence how an individual responds to our treatments (age, education, income, gender, prior vote choices, and regional fixed effects). We also employ a more conservative specification that includes interactions between these demographic covariates and treatment status. The results of these specifications do not meaningfully differ from the main results (see Appendix G).

These multivariate models also enable us to compare the relative importance of an IO leader's race with other drivers of IO legitimacy perceptions (see Table G1 in the online supplement). In both countries, partisanship is the strongest predictor of individuals’ confidence in the WTO. However, the effect of a co-racial IO leader is similar in magnitude to a shift from the minimum to maximum level of educational attainment in South Africa, and it has a larger effect than all other demographic variables in the model. In the United States, the effect is equivalent to moving from the lowest to the highest income bracket, and it is stronger than the effects of education, age, and gender identity. This suggests that an IO leader's race is at least as important as education and other individual-level attributes that have received considerable attention in prior scholarship.⁵⁴

Representativeness of the Sample

Turning to external validity, some might wonder whether our sample's average treatment effects provide valid estimates of the underlying population's average treatment effects. While our US sample closely approximates the Black and white US populations along key demographic dimensions, our South African sample over-represents younger and more educated/affluent individuals. To help determine whether the non-representativeness of our sample impacts our findings, we examine whether age, income, education, and other demographic variables moderate our treatment effects. Appendix H presents the results of separate models for Black and white respondents that include interaction terms between the treatments and a range of pretreatment variables. We find no evidence that demographic variables moderate the effect of our race treatment among either racial group. The absence of treatment-effect heterogeneity provides reassuring evidence that our findings are likely to apply to the broader populations of Black and white people in these countries.⁵⁵

Replication with Additional Contextual Detail

The brevity of the vignette in our WTO experiment may have led us to exaggerate the importance of racial representation. Experimental vignettes that contain additional contextual detail have been found to produce weaker treatment effects; hence, more contextually rich designs can help determine whether treatments have strong effects in more realistic settings outside of a survey experimental context.⁵⁶

In January 2022 we conducted a follow-up experiment with 1,200 Black Americans to determine whether a lengthier vignette containing additional detail about the WTO reduces the effect of racial representation on Black Americans’ legitimacy perceptions (see Appendix I for details). To maximize statistical power, we included only two conditions in this experiment—the “American” treatment and the “Black American” treatment. Two sentences were added before the sentence describing the WTO leader: “The WTO provides technical and legal support to member countries, monitors developments in world trade, and organizes conferences between member countries. The WTO's annual budget is around $200 million.” This updated vignette is about twice as long as the original experiment (fifty-nine words versus twenty-eight words).

Racial representation continues to have a strong effect (see Figure I1 in the online supplement). In this experiment, average confidence in the “Black American” condition is 0.79 points higher than in the “American” condition, which is similar to the difference in the previous experiment (0.71). The large treatment effect here provides reassuring evidence that co-racial leaders are likely to enhance IO legitimacy among Black individuals even in more realistic, information-rich settings.

Evidence from Other IOs

To explore whether leaders’ racial identity has similar effects on public perceptions of other institutions, we draw on three additional experiments, focused on the WHO, the ILO, and a hypothetical IO. These supplementary experiments are useful for several reasons. First, by including evidence from the WHO and ILO, the study covers the three most prominent global IOs that had Black leaders at the time this research was conducted.⁵⁷ Second, these experiments enable us to explore whether the results of our WTO experiment generalize to IOs with differing levels of public salience. At the time of our survey, the WHO was arguably the most prominent IO in the world due to its role in managing the COVID-19 pandemic; the ILO is more obscure than either the WHO or WTO.⁵⁸ A third advantage of the ILO and hypothetical-IO experiments is that they make it possible to assess how individuals respond to leaders who are Black but not American—a useful complement to our WTO design, which estimates the effect of race in the context of an American IO leader.

Across these three additional experiments we consistently find that having a Black IO leader boosts support for the organization among Black people, but has a smaller effect among white individuals. The treatment effect size varies across IOs, with stronger effects for less salient IOs.

The WHO Experiment

Our initial surveys in South Africa and the United States included a second experiment on the WHO.⁵⁹ The WHO experiment included three conditions: a control condition, where no information was given about the WHO leader's race or region of origin; a “Black” treatment that mentioned his being Black; and a “Black African” treatment that noted both his region of origin and his race. This design includes only one treatment that refers to national or regional origin because the director-general, Tedros Adhanom Ghebreyesus, does not have dual citizenship.

The results largely mirror those from the WTO experiment, though with smaller effect sizes (Figure 4, top; see also Appendix J). The “Black” treatment increases confidence in the WHO among both Black Americans and Black South Africans (though it is only significant at the 90 percent level in South Africa). In both countries, the estimated treatment effect is smaller among white subjects than among Black respondents; however, this difference is not statistically significant. These findings suggest that our main results are externally valid. Yet, internal validity is lower in this experiment: the “Black” treatment in this experiment altered respondents’ perceptions of both the race and nationality of the WHO leader (see supplemental Figure J1), which limits our ability to distinguish the effects of these two attributes.

Figure 4. Effect of race on IO legitimacy in South Africa and the US

ILO Experiment

The February 2023 survey included an experiment focused on yet another IO that had elected a Black leader a few months prior—the ILO. Gilbert Houngbo is originally from Togo, a majority-Black African country. However, he is also a citizen of Canada, which is predominantly white. We used this fact to design another experiment capable of disentangling the effects of race and region of origin.

This design largely mirrors that of the WTO experiment. Subjects in the control condition read that the ILO leader is a “Canadian man” but did not receive any information about race. The second condition noted that he is a “Black man from Canada”; and the third, a “Black man from the African country of Togo.” Appendix K describes the experiment in more detail and presents the full set of results. The treatments altered perceptions of race and nationality as intended (Figure K1). Respondents in the “Canadian” and “Black Canadian” conditions reported similar perceptions of the ILO leader's nationality but vastly different perceptions of his racial identity.

Confidence in the ILO is more than one full point higher among Black Americans who were told the ILO leader is a Black Canadian than among those who were told only that he is Canadian (Figure 4, lower left). The treatment effect among white Americans is less than half as large as it is among Black Americans; this difference is statistically significant. The consistency in results across the WTO and ILO experiments is important because it suggests that Black Americans do not simply favor IOs that are led by fellow Black Americans (as in the WTO experiment). Co-racial leaders from other countries—Canada, in this case—have similar effects. In fact, race has a larger effect in this experiment than in the others, presumably because most study participants did not have strong pre-existing opinions about the ILO and even fewer had prior knowledge of the race of its leader.

Conjoint Experiment

The final test was a conjoint experiment that randomly assigned information about six features of a hypothetical IO: whether its leader is white or Black, the leader's country of origin, the number of member countries, the age of the institution, the number of staff members, and staff gender compositon. Notably, the outcome question in this experiment differs from previous ones. Rather than asking about confidence in the institution, we ask respondents whether the United States should increase, decrease, or maintain its funding, using an eleven-point scale. Appendix L provides more details on the design and analysis.

The conjoint design has two key advantages. First, it allows us to randomly assign race and nationality independently. Therefore, this design disentangles these factors in a different, but complementary, way compared to other experiments. Second, conjoint designs mitigate social desirability bias,⁶⁰ which is potentially important because white respondents may have exaggerated their support for Black-led IOs in our other experiments.

Yet again, we find that an IO leader's race matters for our Black American respondents. We estimate that the presence of a Black leader increases support for funding among Black Americans by 0.37 points (Figure 4, lower right). By contrast, the race of the IO leader has a closely estimated null effect (–0.01) among white respondents. Our finding that Black IO leadership has the smallest effect on white peoples’ attitudes in this experiment raises the possibility that the modest positive effects observed among this subgroup in some other experiments reflected social desirability bias, and that the true effects are smaller than estimated.

Evidence from Kenya

To this point, all experiments were conducted in countries where racial differences are highly salient due to a history of racial discrimination. It is reasonable to ask whether our findings generalize to countries where racial identities are less politicized. To address this, we consider the case of Kenya, a country where ethnicity, rather than race, is the principal identity marker.⁶¹

We fielded three experiments in Kenya in February 2024 with 1,000 self-identified Black/African Kenyans. The first is a simplified version of the WTO experiment containing just the two main experimental conditions of interest: “American” and “Black American.” Similarly, the second is a streamlined version of the ILO experiment we conducted in the United States, focusing again on the two key conditions for testing the effect of an IO leader's race: describing him as a “Canadian man” and describing him as a “Black man from Canada.” Third, we also fielded the conjoint experiment, replicating the US design verbatim. Appendix M reports the complete experimental protocol.

Figure 5 presents the main results from these experiments. Each panel displays the effect of Black IO leadership among Black Kenyans. In all three experiments, information that an IO leader is Black has a positive and statistically significant effect on opinion, replicating our earlier findings. This demonstrates that the race of an IO leader can influence institutional legitimacy in countries where racial identities are not highly salient domestically. However, the effects are smaller in this experiment than in the Black American samples. In countries where racial identities are less salient, the race of IO leader may have a substantively smaller effect on the public legitimacy of IOs.

Figure 5. Effect of race on IO legitimacy in Kenya

Conclusion

Using a series of ten survey experiments in three countries, we find strong and consistent evidence that Black people are more confident in IOs when informed they have Black leaders. Black IO leadership had a positive and statistically significant effect on Black individuals’ perceptions of legitimacy in nine of the ten experiments, and the effect was nearly significant (p < 0.1) in the tenth. White individuals were not as strongly affected. In five of the six experiments that included self-identified white respondents, the effect of Black IO leadership on white people's attitudes was either statistically insignificant or significantly weaker than the effect for the Black subsample. The main takeaway is that historically underrepresented racial groups have more positive feelings toward IOs led by those who share their racial identity—a finding that holds across experiments focused on multiple IOs in three different countries.

The consistency in results suggests that our main finding, that IOs with Black leaders are more legitimate in the eyes of Black citizens, is likely to generalize far beyond the specific IOs, countries, and time period considered in this study. The fact that we observe little heterogeneity in how Black Americans or Black South Africans respond to our treatments (see Appendix H) further suggests that our findings have a high degree of external validity.⁶² While this effect should hold across a wide array of settings, our findings imply that the magnitude of this effect may depend on the salience of racial identities relative to other considerations. Our treatment effects are stronger in countries, such as the United States and South Africa, where racial identities are more politicized, than in countries like Kenya, where other identities are more prominent. The treatments also have stronger effects on attitudes about obscure IOs like the ILO than for better-known IOs such as the WTO and WHO, presumably because individuals’ assessments of the latter organizations consider a wider array of factors, reducing the relative importance of race. In sum, while our results are likely to apply broadly, the importance of IO leaders’ racial identity is likely to be weakest in institutional and national contexts where racial identities are less salient.

Our study sheds new light on the importance of race for the popular legitimacy of IOs, but a number of questions on this critical topic remain unanswered. Our focus on two racial groups—self-identified white and Black people—does not address how members of other racial groups respond to Black IO leadership. Similarly, our study cannot say whether members of other historically underrepresented racial groups also view IOs with co-racial leaders more positively. It is also unclear whether the effect of an IO leader's race depends on other identities of the leader, such as their gender. Future work should address these questions.

For policymakers seeking to boost IOs’ popular legitimacy, our study suggests there are clear upsides and few downsides to a racially diverse leadership. Improving the representation of disadvantaged racial groups at top levels in these organizations appears to be one promising way for global-governance institutions to enhance their public image and fend off growing challenges to their legitimacy.

Still, it would be wrong to conclude, based on our research, that Black leadership is a cure-all for global-governance institutions. For example, the presence of a Black leader at the International Criminal Court did not protect it from charges that its patterns of prosecution are racially biased.⁶³ Having a Black president may have helped the court limit the fallout from these concerns, but it did not eliminate them. Diversifying IO leadership, then, is only one element of the solution for enhancing the legitimacy and effectiveness of these bodies.

Finally, it is also important to ensure that the appointment of racial-minority IO leaders does not constitute a new form of tokenism at the global level. Representational gains not accompanied by improvements in the welfare of historically marginalized groups might come to be viewed as performative acts divorced from substantive change. Whether the elevation of Black leaders actually advances the cause of racial equality remains an open but crucial question—one that scholars will need to address in the future.