Authority

We conceptualize authority and legitimacy as distinct but related entities. Whereas authority refers to an organization having the right to make decisions within a particular area, legitimacy refers to the perception that these rights are appropriately exercised (Tallberg and Zürn Reference Tallberg and Zürn2019, 586). This analytical separation of legitimacy from authority is well anchored in parts of social theory.Footnote ¹ Weber (Reference Weber1922/1978, 213), for instance, speaks of how every system of authority “attempts to establish and to cultivate a belief in its legitimacy.” At the same time, authority and legitimacy are related, in so far as legitimacy only becomes an issue once an institution possesses authority. In the absence of authority, the question of legitimacy becomes uninteresting.

Empirically, the authority of IOs is captured by three components (Hooghe et al. Reference Hooghe, Marks, Lenz, Bezuijen, Ceka and Derderyan2017; Zürn et al. Reference Zürn, Tokhi and Binder2021). First, IOs enjoy greater authority when they have been conferred greater policy-making competences in issue domains that previously were regulated at the domestic level or not at all (Zürn et al. Reference Zürn, Binder and Ecker-Ehrhardt2012). Second, IOs enjoy greater authority when the member states move away from intergovernmental cooperation by delegating increasing power to autonomous supranational bodies (Tallberg Reference Tallberg2002; Hawkins et al. Reference Hawkins, Lake, Nielson and Tierney2006). Third, IOs have greater authority when the member states pool power within intergovernmental bodies by shifting toward forms of majority voting that remove each state’s veto over decisions (Keohane and Hoffmann Reference Keohane and Hoffmann1991; Moravcsik Reference Moravcsik1998).

By these criteria, the authority of IOs has expanded considerably over recent decades (Hooghe et al. Reference Hooghe, Marks, Lenz, Bezuijen, Ceka and Derderyan2017; Zürn et al. Reference Zürn, Tokhi and Binder2021). States have empowered IOs with authority in more policy domains, delegated more authority to supranational bodies, and pooled more authority in collective decision-making. The growth in IO authority is particularly notable after the end of the Cold War. That said, IOs continue to vary in the authority they possess, ranging from greatly empowered organizations such as the EU, which scores high on all three components, to less empowered organizations such as NAFTA (and its successor, the United States–Mexico–Canada Agreement [USMCA]), which scores low on all components.

We expect information about the authority of IOs to matter for citizens’ legitimacy beliefs. Specifically, we anticipate that IOs with greater authority will have a harder time securing legitimacy from citizens, all else equal (Zürn Reference Zürn2018; Anderson et al. Reference Anderson, Bernauer and Kachi2019; Tallberg and Zürn Reference Tallberg and Zürn2019). When IOs enjoy extensive authority, they also have to meet demanding procedural and performance standards, or they will suffer from legitimacy deficits. When IOs enjoy less authority, the procedural and performance requirements they have to meet to be deemed legitimate are less demanding.

The EU is often said to offer an illustration of this logic (Banchoff and Smith Reference Banchoff and Smith1999; Hooghe and Marks Reference Hooghe and Marks2009; de Wilde and Zürn Reference De Wilde and Zürn2012). The greater the authority of the EU, the higher the demands on the organization to take decisions democratically and to solve problems effectively. As the EU often has fallen short of these expectations, despite more democratic procedures and effective performance than most IOs, legitimacy problems have arisen, reflected in low turnouts in European elections, rejections of new EU treaties in national referenda, and a decision on the part of the UK to leave the organization.

There is to date no systematic empirical evidence for a negative authority-legitimacy linkage, as the only empirical study so far on the relationship between authority and legitimacy finds no effect in either negative or positive direction. As Anderson et al. (Reference Anderson, Bernauer and Kachi2019, 663) conclude, based on a survey experiment in the context of global environmental governance: “[E]ven important shifts of authority from the national to the global level, such as majority decision making at the international level and automatic implementation of international decisions domestically, do not significantly affect citizens’ legitimacy perceptions on average.”

However, given that greater authority should set the bar higher for IOs to fare well in peoples’ perceptions, we expect a negative relationship between IO authority and legitimacy beliefs. In addition, we expect this negative relationship to be moderated by the degree to which citizens hold internationalist attitudes (Schlipphak et al. Reference Schlipphak, Schäfer and Treib2021). When citizens are more positive toward international cooperation in general terms, we expect the negative effect of authority on legitimacy to be weaker. Conversely, when citizens are more negative toward international cooperation in general terms, we expect this attitude to strengthen the negative effect of IO authority on legitimacy beliefs.

We advance two hypotheses on the basis of this argument. First, we formulate a general expectation about the effect of IO authority on people’s legitimacy beliefs. Second, we formulate a conditional expectation about heterogeneity in effects depending on the degree to which citizens hold internationalist attitudes.

Purpose

The notion that an organization’s social purpose would affect perceptions of its legitimacy is not novel, even if the logic has never been fully theorized or tested. The earliest considerations of purpose hark back to pioneers in the general study of political legitimacy. Easton (Reference Easton1975, 452), for instance, argued that political institutions may obtain legitimacy on the grounds of people’s ideological beliefs or moral convictions, next to their beliefs in the appropriateness of institutions and the personal qualities of rulers. Dowling and Pfeffer (Reference Dowling and Pfeffer1975, 126), in another seminal account, distinguished between the operation, output, and “the goals or domain of activity” of the organization as three sources of legitimacy. Scott (Reference Scott, Walter and DiMaggio1991, 169), similarly, speaks of how legitimacy for an institution may derive primarily from “societal evaluations of organizational goals.” In the study of global governance, Barnett (Reference Barnett1997, 539) offers an early discussion of the “substantive legitimacy” of IOs, understood as the “ends that are considered desirable,” to be distinguished from procedural legitimacy, or how IOs make decisions to reach those ends. Yet, despite these attempts, the idea of social purpose as an additional institutional source of legitimacy never truly took off. Possibly, such a development was stymied by Scharpf’s (Reference Scharpf1999) influential dichotomy between input (procedure) and output (performance).

In recent years, a number of contributions have again suggested that organizational purpose may present a driver of legitimacy beliefs in global governance. These accounts typically conceive of social purpose as an institutional quality on par with procedure and performance. Scholte and Tallberg (Reference Scholte, Tallberg, Tallberg, Bäckstrand and Scholte2018, 64) acknowledge that the common distinction between procedure and performance misses potential “substance-grounded” legitimacy beliefs. Lenz and Viola (Reference Lenz and Viola2017) explicitly speak of procedure, performance, and purpose as the three central organizational features of IOs that feed into assessments of their legitimacy. Nielson et al. (Reference Nielson, Hyde and Kelley2019, 692) suggest that “actors may assess organizations not merely on how they operate and whether they accomplish their goals, but on what the goals themselves are and whether these are normatively desirable.” Taken together, these contributions suggest that citizens would be sensitive to information about an IO’s social purpose when forming legitimacy beliefs vis-à-vis IOs.

Developing this intuition further, we expect the communicated purposes of IOs to matter for legitimacy beliefs because of how they activate citizens’ ideological priors. Political ideologies are systems of normative ideas that bundle ideological content in ways that help people to orient themselves on the political spectrum and to arrive at political choices (Hamill et al. Reference Hamill, Lodge and Blake1985; Sniderman et al. Reference Sniderman, Hagen, Tetlock and Brady1986). When citizens hold a particular ideological orientation, this offers them a shortcut to political positions on a whole range of issues interpretable in ideological terms (Jost et al. Reference Jost, Federico, Napier, Freeden and Stears2013).

We suggest that an IO’s social purpose often is perceived as inherently political or normative. Promoting free trade, combatting poverty, or protecting human rights may not be regarded by citizens as neutral exercises of international problem-solving, but as associated with the furthering of certain political ideals rather than others. In some cases, these ideals are closely linked to traditional political ideologies and cleavages in society, such as the left–right dimension. For instance, free trade and deregulation are often associated with market liberalism, while redistribution and social rights are associated with socialism or social democracy. When IOs promote purposes that are interpreted by citizens as political, we would expect citizens to use information about purpose when forming opinions about IOs. Organizations with purposes that accord more with a person’s political priors are more likely to be regarded as legitimate, while IOs that promote goals that diverge from a person’s ideological leanings are less likely to be seen as legitimate.

We advance two hypotheses on the basis of this argument. First, we formulate a general expectation about the effect of social purpose on people’s legitimacy beliefs. Second, we formulate a conditional expectation about heterogeneity in effects depending on people’s political priors.

Research Design

The conjoint experiment exposes participants to hypothetical IOs that differ with respect to authority, purpose, and other institutional qualities. Its primary objective is to test hypotheses about how important communicated levels of authority and social purposes are for citizens’ legitimacy beliefs toward IOs, and how internationalist attitudes and political ideology affect these relationships.

Survey Design

The experiment is embedded in an online survey with nationally representative samples of German (N = 2,044) and American respondents (N = 2,048).Footnote ² This cross-country design extends prior experimental studies on the effects of institutional qualities on legitimacy beliefs that have focused on a single country – the US (cf. Anderson et al. Reference Anderson, Bernauer and Kachi2019). An important rationale for selecting Germany and the US is the general differences between the two countries in our moderating factors – political ideology and internationalist attitudes (cf. Chapter 5). The US is a liberal market economy with a two-party system and strong public opinion polarization, as well as an ambivalent approach to international cooperation, alternating between isolationism and internationalism. Germany is a coordinated market economy with a multiparty system and less polarized public opinion, as well as a strong commitment to international cooperation. At the same time, the two countries are similar across several important contextual conditions, including their federal political systems, their high levels of economic development, and their political centrality in most IOs of which they are members, allowing us to hold potentially confounding context factors constant. The survey was implemented by YouGov in May 2019 (see Online Appendix A).

Experimental Design

We use a conjoint experiment to test our hypotheses about the effect of communication about an IO’s authority and purpose, respectively, on legitimacy beliefs. Conjoint analysis methods were developed in psychology and marketing, and have become increasingly common in political science in recent years (Hainmueller et al. Reference Hainmueller, Yamamoto and Hopkins2014). In a conjoint experiment, respondents typically receive two alternative descriptions of cases and are then asked to rank or rate these two hypothetical alternatives. These two alternative cases have multiple attributes with differing values. By systematically varying how these cases are described, analysts can estimate the importance of each attribute on respondents’ combined choices. In the context of global governance, scholars have used conjoint experiments to assess, for instance, which institutional qualities generate public support for international environmental agreements (Bechtel and Scheve Reference Bechtel and Scheve2013; Bernauer et al. Reference Bernauer, Mohrenberg and Koubi2020).

We devised a conjoint experiment in which each respondent is shown screens with two hypothetical IOs in comparison. Each IO has a set of distinct attributes. Respondents are then asked to rate their confidence in each IO. This design allows us to assess how information about different institutional features of IOs affects respondents’ legitimacy beliefs. Using hypothetical IOs allows us to estimate the effects of communicated IO attributes systematically and with great precision. While using real-world IOs would have added an element of realism to the experiment, it would have made it impossible to vary IO attributes systematically in the comparisons in such a way that effects could be established confidently. In addition, respondents may have been influenced by preexisting beliefs and knowledge about IOs when asked to choose between them.

This experimental component of the survey was introduced through a text providing context and instructions (Figure 7.1).

Figure 7.1 Example screen with survey instructions

After this introductory screen, each respondent received four randomly allocated screens. Each of these four screens compared two hypothetical IOs and asked the respondent to choose between them and to indicate its level of confidence in them. This comparison worked as follows. The order of the institutional qualities of the two IOs was randomly assigned across respondents, but consistent across the four binary comparisons for each respondent to avoid confusion. The values of the institutional qualities were fully randomized, with two exceptions. First, respondents were never given the same value on an institutional quality in a comparison across two IOs. Second, respondents were never confronted with the same screen twice.

The number of institutional qualities presented to the respondents in the experiment is well in line with the number of items respondents can meaningfully evaluate in the context of a conjoint experiment (Bansak et al. Reference Bansak, Hainmueller, Hopkins and Yamamoto2018). Authority is operationalized through a categorical variable capturing the power an IO exerts over member states. This measure includes both formal – codified – and informal – social – power (Barnett and Duvall Reference Barnett and Duvall2005). We focus on IO authority as degrees of power over member states, partly because this conceptualization captures the implications of delegation and pooling for individual states (Hooghe and Marks Reference Hooghe and Marks2015) and partly because it captures how IO authority typically is expressed in elite communication. Purpose is measured using descriptions of hypothetical core mandates of IOs that are relevant in global governance, such as the protection of human rights or poverty reduction. This measure captures the moral dimension of purpose (cf. Lenz and Viola Reference Lenz and Viola2017). When measuring procedures, we highlight two central aspects of IOs – transparency and participation. While we could have selected other procedural features, such as accountability and fairness, transparency and participation are two central procedural dimensions that have received much attention in prior studies in international relations. Similarly, we select two central aspects of performance – fair outcomes and problem-solving capacity (cf. Scholte and Tallberg Reference Scholte, Tallberg, Tallberg, Bäckstrand and Scholte2018). Table 7.1 summarizes the institutional features varied in the conjoint design. Figure 7.2 offers an example of what such a screen might look like.

Figure 7.2 Example screen with conjoint experiment

In operationalizing these institutional qualities, we faced a choice between nuanced and stark alternatives. We opted for more nuanced alternatives, since we believe those better capture the variation that exists in real-world global governance. For example, we chose to operationalize the dimension of authority by distinguishing between an IO having “limited,” “some,” or “extensive” power over member states, rather than simply “no” or “extensive” power.

The two outcome variables of interest tap into individuals’ confidence in IOs. Our preferred measure of legitimacy beliefs is the degree of confidence (see Chapter 3). We measure this degree on a scale from 1 (no confidence at all) to 9 (complete confidence) (Figure 7.2). For the purpose of robustness checks and to assess the sensitivity of experimental results across subgroups, we also ask respondents to indicate which of the two hypothetical IOs they would prefer. Answering this latter question only requires choosing between two, which is cognitively less demanding for respondents than indicating their confidence in either IO.

The experiment was preceded by indicators for the purpose of balance tests and additional robustness checks: intentional media consumption, cognitive mobilization, generalized trust, confidence in domestic government, and knowledge about global governance. The experiment was followed by an attention check. We use the information from this attention check to limit the sample in the main analysis to those respondents who correctly passed this test. Finally, YouGov provides demographic and political data on the respondents as background information, such as information on gender, age, education, and geographical region (see Online Appendix R for the entire questionnaire in English and German).

Measuring Internationalist Attitudes and Political Ideology

To explore H1b and H2b about a conditioning effect of attitudes toward international cooperation and political ideology on treatment effects, the survey assessed respondents’ opinions on international cooperation and partisan identification. The first indicator asked respondents to indicate if they think that international cooperation is a “good thing,” a “bad thing,” or “neither good nor bad.” The answers to this question reveal that similarly high shares of the population in both countries (between 71 and 74 percent) indicate that they think international cooperation is a good thing (Figure 7.3).

Figure 7.3 Internationalist attitudes in Germany and the US

Notes: Weighted percentage of those thinking that international cooperation is a bad thing, neither good nor bad, or a good thing.

The indicator for partisan identification is created based on a standard question about whether there is a particular political party they feel closer to than all the other parties. We described the patterns for this indicator in Chapter 5, which uses the same respondents for its two experiments. In short, partisan identification differs between Germany and the US. US public opinion is more polarized, since only about 18 percent of US citizens are estimated to be independents and the rest are either Democrats or Republicans. In Germany, about 22 percent are independents and the rest identify with a much larger number of political parties (Figure 5.1). That US opinion is more polarized than German public opinion can also be seen when looking at the distribution of left–right ideology (Figure 7.4).

Figure 7.4 Left–right ideology in Germany and the US

Notes: Weighted percentage.

Effects of Communication about Authority and Purpose

Figures 7.5 and 7.6 show our estimates of authority and purpose as well as the other institutional qualities on confidence. The dots represent the estimated average marginal component effects (AMCEs) of a given value for each quality on individual confidence toward the packaged IO profile, relative to a reference category or baseline. In other words, the AMCEs express the degree to which an IO feature increases (or decreases) citizen confidence in an IO. The bars indicate 95 percent confidence intervals, and the points without bars indicate the baseline for a given value of an institutional quality. The interpretation of each estimate is relative to the baseline for that dimension.

Figure 7.5 Effects of institutional qualities in the US

Notes: Sample includes attentive respondents only. AMCEs with their respective 95 percent confidence intervals. Weighted data. Vignette descriptions shortened for the sake of presentation.

To examine if authority and purpose matter, we need to decide on a baseline for each indicator. For all indicators except purpose, we use the lowest category as a baseline. That is, unlike in other experiments such as vignette experiments, there is no control group but a baseline. For example, for authority, we compare the effects of “some” and “extensive” power over member states to “limited” power. For the purpose, this logic is not applicable and we need another motivation. Here, we assume that there may be different understandings of social purpose among citizens, where ensuring peace and security arguable is one of the least contentious purposes of an IO. We thus use the protection of peace and security as a baseline.

The results clearly show that authority matters, thereby supporting H1a in both Germany and the US. Moving from an IO with limited power over its member states to one with extensive power over its member states decreases legitimacy beliefs by 0.221 in the US (p < 0.000, Figure 7.5) and 0.182 points in Germany (p < 0.000, Figure 7.6) on the 1–9 confidence scale. By contrast, moving from limited to some power does not have any effect on confidence. This finding suggests that respondents react to the formulation “extensive power” and not “some power.”

Figure 7.6 Effects of institutional qualities in Germany

Notes: Sample includes attentive respondents only. AMCEs with their respective 95 percent confidence intervals. Weighted data. Vignette descriptions shortened for the sake of presentation.

In addition, the results indicate that social purpose matters, thereby corroborating H2a in both countries. In the US (Figure 7.5), the purpose of combating climate change leads to a 0.309 point decrease on the confidence scale (p < 0.000). This is the strongest purpose effect on legitimacy beliefs when compared to the baseline of ensuring peace and security. We also find that moving from the baseline to the purpose of promoting free trade leads to a 0.177 point decrease (p < 0.007) – the second strongest effect.

In Germany, two effects are significant (Figure 7.6). The strongest effect is recorded for free trade. Moving from the baseline to the promotion of free trade represents a 0.401 point decrease on the confidence scale (p < 0.000). Similarly, moving from the baseline to the promotion of public health leads to a decrease of 0.267 points (p < 0.000). Taken together, these findings suggest similar results for Germany and the US in the sense that social purpose matters for legitimacy beliefs.

Communication of other institutional qualities also matters for legitimacy beliefs, further strengthening our confidence in the findings in Chapter 6. The effects are very similar in the two countries. We start with procedure-related qualities. In the case of transparency, moving from the baseline of a confidential organization to one that is public increases confidence by about 0.457 points (p < 0.000) in both the US and Germany. In the case of participation, moving from the baseline of an IO in which only powerful countries have a say to an IO in which citizens have a say increases confidence by an estimated 0.554 points in the US (p < 0.000) and 0.634 in Germany (p < 0.000). We then turn to performance-related qualities. In the case of fair outcomes, moving from the baseline of an IO that benefits some countries more than others to an IO that benefits all members equally raises confidence by 0.316 points (p < 0.000) in the US and 0.374 points (p < 0.000) in Germany. Finally, in the case of problem-solving, moving from the baseline of an IO that solves few important problems to an IO that solves the most important problems increases confidence by 0.293 points (p < 0.000) in the US and 0.393 points (p < 0.000) in Germany.

Can we expect information about the institutional design of IOs to lead to substantial shifts in confidence in the real world? We examine this by predicting levels of confidence for two hypothetical IOs (cf. Bechtel and Scheve Reference Bechtel and Scheve2013). By prediction, we mean the computation of levels of confidence for each of the hypothetical IOs based on 1,000 country-specific simulations using the results of respondents’ confidence ratings (see King Reference King, Michael and Wittenberg2000, for a discussion of the methodology). The first IO is one that has unattractive characteristics based on our experimental results: Its purpose is to combat climate change in the US (and promote free trade in Germany), it has extensive power over member states, it solves few important problems, it yields benefits for specific countries at the expense of others, it provides only powerful states with a say, and it is confidential. The second IO has all the features of an attractive design based on our experimental results: Its mandate is to ensure peace and security, it has limited power over member states, it solves most important problems, it yields equal benefits for all countries, it provides citizens with a say, and it is public.

This additional analysis suggests that information on how IOs are designed, from the most unpopular design to the most popular, may lead to noteworthy shifts in legitimacy beliefs. In Germany, average confidence is predicted to be 4.2 for an IO with an unattractive design on the 1–9 confidence scale, but as much as 5.9 in the case of an IO with popular design features. Similar results are found in the US, where we predict an average confidence level of about 4.3 for the unattractive IO design and about 6 for the attractive IO design.

Finally, we are interested in whether the findings on authority and purpose depend on respondents’ level of political awareness, as theorized in Chapter 3. To this end, we examine differences in effect sizes for subgroups that differ in their level of education and knowledge about global governance, respectively. Looking at differences in AMCEs, we do not find any systematic moderating effects of either indicator on authority and purpose effects (Online Appendices S1 and S2).

Interaction Analysis

Next, we examine H1b and H2b, which predict that the effects of authority and purpose on legitimacy beliefs are moderated by people’s preexisting political beliefs. To test H1b, we focus on attitudes toward international cooperation as a potential moderating factor. To test H2b, we concentrate on partisan identification.

For this analysis, we use a different way of calculating and comparing treatment effects, since each subgroup will have a different average value on the baseline quality used to identify effect strength. We, therefore, complement AMCEs (used to infer differences in causal effects within subgroups) with marginal means (MMs) (used to infer if subgroups differ in how they value specific institutional qualities) (Hainmueller et al. Reference Hainmueller, Yamamoto and Hopkins2014; Leeper et al. Reference Leeper, Hobolt and Tilley2020). MMs express the preferences of respondents for all institutional features; these quantities are column and row mean outcomes for each institutional feature level, averaging across all other features. We calculate differences between MMs to check whether observed differences between MMs of two subgroups are statistically significant with regard to a specific institutional feature. If they are, then that feature can be assumed to shape confidence more in a particular subgroup than in another. As suggested by Leeper et al. (Reference Leeper, Hobolt and Tilley2020), we use the discrete choice outcome variable to estimate AMCEs for each institutional feature separately and then compare those estimates to MMs to ascertain the sensitivity of the analysis (see Online Appendix S for detailed results).

Figure 7.7 shows the results for H1b in the US. We find that attitudes toward international cooperation, indeed, moderate the effects of authority on IO legitimacy beliefs. However, the effect is partly the opposite of what we theorized. The AMCEs suggest that Americans with positive or neutral attitudes toward international cooperation react with weaker confidence to information that IOs have extensive power compared to the baseline of limited power. This negative effect is not found among citizens with more nationalist attitudes. The MMs indicate that this negative effect on confidence is stronger among those people who are neutral toward international cooperation than among those with internationalist attitudes (p < 0.023).

Figure 7.7 Effects of authority in the US, by internationalism

Notes: Sample includes attentive respondents only. AMCEs and MMs with their respective 95 percent confidence intervals. Weighted data. Answers to the question: “Do you think international cooperation is: a bad thing, a good thing, or neither good nor bad?” Dependent variable: Discrete choice between two organizations. See Online Appendix S3 for detailed results.

This result is puzzling in view of our expectation that people with more internationalist attitudes would be more tolerant of greater IO authority. However, the finding makes more sense if we consider that people with negative attitudes toward international cooperation generally are less receptive to information about IOs than people with positive attitudes, as we discovered in Chapter 5. Those people who are negatively predisposed appear to already have strong opinions that are less malleable. That said, the results also provide some evidence consistent with our original expectation, since people with neutral attitudes toward international cooperation appear to react more negatively to information about IOs having extensive power than people with positive attitudes.

Figure 7.8 shows the corresponding results for Germany, where we do not find that internationalist attitudes systematically condition the relationship between authority and confidence, which contradicts H1b. While we find a negative effect of an IO having extensive authority among those neutral toward international cooperation (as in the US), differences in MMs among subgroups are not statistically significant, indicating that the three subgroups do not hold different preferences regarding the authority of IOs in Germany.

Figure 7.8 Effects of authority in Germany, by internationalism

Notes: Sample includes attentive respondents only. AMCEs and MMs with their respective 95 percent confidence intervals. Weighted data. Answers to the question: “Do you think international cooperation is: a bad thing, a good thing, or neither good nor bad?” Dependent variable: Discrete choice between two organizations. See Online Appendix S4 for detailed results.

Next, we are interested in whether partisan identification moderates the effects of an IO’s social purpose on citizens’ confidence in this organization (H2b). The results for the US are in line with this hypothesis (Figure 7.9). The AMCEs indicate that Democrats respond with greater confidence when informed that an IO engages in poverty alleviation compared to the baseline of ensuring peace and security, and with weaker confidence when the IO promotes free trade. Republicans, on the one hand, respond negatively to information about IOs fighting poverty and climate change compared to the baseline. The differences in MMs between Democrats and Republicans are statistically significant (except in the area of public health), indicating that partisan identification systematically moderates the effects of different social purposes on IO confidence.

Figure 7.9 Effects of social purpose in the US, by partisanship

Notes: Sample includes attentive respondents only. AMCEs and MMs with their respective 95 percent confidence intervals. Weighted data. Answers to the question: “Is there a particular political party you feel closer to than all the other parties?” Dependent variable: Discrete choice between two organizations. See Online Appendix S5 for detailed results.

In Germany, we examine H2b across partisans of the historically two largest parties, the CDU/CSU and SPD, which are also featured in Chapter 5, to make the analysis more comparable to the one of the US. The results do not support H2b (Figure 7.10). The AMCEs show that the effects of social purpose on confidence are quite similar across all subgroups, largely reflecting the aggregate pattern in Figure 7.6. The exception is the group of partisans who are neither SPD nor CDU/CSU, which appear to be more easily affected by an IO’s social purpose. In this group, all purposes except poverty alleviation lead to lower IO confidence compared to the baseline purpose of ensuring peace and security. However, the MMs for each purpose are not statistically different from each other across subgroups, suggesting that different groups of partisans in Germany do not have different preferences regarding the social purposes of IOs.

Figure 7.10 Effects of social purpose in Germany, by partisanship

Notes: Sample includes attentive respondents only. AMCEs and MMs with their respective 95 percent confidence intervals. Weighted data. Answers to the question: “Is there a particular political party you feel closer to than all the other parties?” Dependent variable: Discrete choice between two organizations. See Online Appendix S6 for detailed results.

Taken together, the evidence from the interaction analysis is mixed in that H1b and H2b are only supported in the US. We attribute this result partially to the level of polarization in public opinion and the political party landscape in the US. Against this backdrop, it is not so surprising that we do not find CDU and SPD partisans to react differently to information about the authority and purpose of IOs.

Validity and Robustness Checks

We perform several validity and robustness checks, which corroborate our findings about the effects of communicated authority and purpose on legitimacy beliefs (Figure 7.5). First, we conduct balance checks to assess whether the randomization produced well-balanced experimental groups (cf. Hainmueller et al. Reference Hainmueller, Yamamoto and Hopkins2014), which indicate that the attributes are jointly balanced (Online Appendices T1–T2).

Second, we replicate the analyses using the alternative dependent variable indicating whether respondents chose organization 1 or 2. Respondents were asked to make this discrete choice right after they were presented with the different institutional qualities, as described earlier. Respondents tended to be consistent in indicating relatively high levels of confidence in the organization they chose, and lower levels of confidence in the organization they did not choose. Interestingly, in Germany, a larger number of social purpose cues have effects when examining organization choice than when examining confidence in an IO, potentially because it is an easier task to choose between two organizations than to rate confidence in both. The social purposes of human rights, free trade, climate change, and public health all make an organization less likely to be chosen compared to the baseline security IO (Online Appendix T1).

Third, we checked whether our results in Figures 7.5 and 7.6 are conditional upon other individual characteristics than those we hypothesized, such as confidence in government. For this purpose, we use responses to a question about a respondent’s confidence in the government on a scale from no confidence at all (0) to complete confidence (10). Results suggest that there are no differences in AMCEs at different levels of confidence in government, so we do not investigate this further through MMs (Online Appendix T2).

Fourth and finally, we run the analysis from Figure 7.5 by including both attentive and nonattentive respondents. Our attention check asked respondents the following question after they had completed about 70 percent of the survey: “We are interested in learning about your preferences on a variety of topics, including colors. To demonstrate that you’ve read this much, just go ahead and select both red and green among the alternatives below, no matter what your favorite color is. Yes, ignore the question below and select both of those options. What is your favorite color?” Correct answers were coded as one and incorrect answers as zero (Bechtel and Scheve Reference Bechtel and Scheve2013). About 66 percent of US respondents were attentive, while only 54 percent of German respondents were attentive. When we test the robustness of the experimental results using the full sample, which also includes the nonattentive respondents, results vary slightly. In the US, a free trade purpose does not have the same negative effect when including nonattentive respondents as well; however, a climate change purpose continues to have a negative effect on confidence in IOs compared to the baseline security purpose. In Germany, the main results are robust, with the addition that also a human rights purpose has a negative effect of confidence on IOs compared to the baseline security purpose (Online Appendix T3).