Elite legitimation and delegitimation of international organizations in the media: Patterns and explanations

Legitimacy communication in the media reveals when elites become attentive to international organizations’ (IOs) legitimacy and whether they support or question their legitimacy. The intensity and tone of this communication results in communicative support or legitimacy pressures on IOs. Extant research gives few insights into the scope and nature of elite legitimacy communication and the factors that shape it. This article offers a comparative and longitudinal analysis of the patterns of elite communication in the media. It maps and explains variation in the intensity and tone of legitimacy communication based on a quantitative content analysis of roughly 6500 legitimacy evaluations of the EU, the G8, and the UN in the quality press of four established democracies. A multinomial logistic regression analysis yields three key results. First, in contrast to conventional expectations, there is no clear shift from low intensity and positive tone to high intensity and negative tone. Second, communication intensity is considerably higher for powerful IOs. Third, political events, including security crises and institutional reform, are important drivers of the ebbs and flows of western elites’ communicative support and pressure on major IOs.


Introduction
Why and when do societal elites contest the legitimacy of international organizations (IOs) in the media? In June 2005, just a few weeks after the failed European https://doi.org/10.1007/s11558-018-9320-9 Electronic supplementary material The online version of this article (https://doi.org/10.1007/s11558-018-9320-9) contains supplementary material, which is available to authorized users.
2 Elite communication about the legitimacy of international organizations in the media: Theories and hypotheses I refer to elite legitimacy communication as the entirety of public statements evaluating the normative appropriateness of IOs as well as their basic institutions and principles made by actors whose strategic positions in society enable them to potentially influence political decision-making (Higley and Moore 2001, p. 176). These elites, including national executives and political parties, international bureaucrats, economic actors, and civil society organizations, are central actors in the political struggle about IO legitimacy as they Bmay be able to make their positive or negative support count more than high levels of support from unorganized millions^ (Easton 1965, p. 167). Elites use this ability to evaluate IO legitimacy in the media for at least three reasons. First, to facilitate international cooperation IOs require a modicum of positive legitimacy perceptions among their members and their governance targets. Hence, IO supporters have incentives to provide communicative support, whereas opponents may use public legitimacy evaluations to discredit IOs. Second, both governments and private actors tend to use IOs for camouflage. Scapegoating IOs in the media to deflect criticism for unpopular policies and outcomes is likely to trigger negative legitimacy evaluations. By contrast, laundering the Bdirty work^of governments and business (Abbott and Snidal 1998, p. 167;Vaubel 1986, p. 49), such as the provision of foreign aid in exchange for cooperation (Lim and Vreeland 2013) or decisions to go to war (Chapman 2009), should go hand in hand with positive legitimacy evaluations. Third, in many western democracies, IOs have become an important issue in party competition. Since international integration has become politicized (Hooghe and Marks 2009;Zürn et al. 2012) political parties have incentives to present their positions on IOs by publicly evaluating their legitimacy.
In mass democracies, the privileged arena for these legitimacy struggles is the news media. They constitute the public sphere and link its various arenas. As such, news media serve as direct communication channels between citizens, elites, and IOs, and give voice to positive legitimacy evaluations of IO supporters and criticism from their opponents (Baum and Potter 2008;Bennett and Entman 2001). Consequently, elites engaged with IOs and their legitimacy try to gain public support through the news media and to position their legitimacy evaluations in this highly visible format (Kriesi et al. 2012, p. 39). Exposure to such evaluations in the news media has the potential to affect citizens' perceptions of IO legitimacy, to mobilize their political engagement, and consequently to shape national agendas (King et al. 2017).
I conceptualize elite legitimacy communication in the media as having two dimensions, namely (a) the intensity of communication and (b) its tone. Intensity indicates the frequency of elite legitimacy evaluations in the media. It reveals that elites are attentive to IO legitimacy and that they feel the need to take communicative action. More intensive communication makes IOs publicly visible, provides them with focality, and eventually helps them governing by communicative means (Barnett and Finnemore 2004). In addition, intensive communication informs IOs about the legitimacy standards they are expected to meet and signals the importance of individual IOs to citizens. This focus on intensity breaks with previous scholarship on legitimacy evaluations, which assumes that IO legitimacy matters to elites and focuses exclusively on tone and sources of legitimacy (for instance, Dellmuth and Tallberg 2015;Schlipphak 2015). Yet, if we do not know whether legitimacy questions are at all pertinent to elites, answers to these latter questions are difficult to interpret.
The tone of legitimacy evaluations informs us about the most critical content of elite communication in the media. Do they question or support IO legitimacy? Positive tone signals normative support and adds to IOs' focality. Negative tone indicates that elites perceive a mismatch between their legitimacy standards and IOs' institutional design, purpose, or policymaking. It may foreshadow a legitimacy crisis which forces adaptation or even disempowerment (Zürn 2018). Due to the importance of intensity and tone I consider both dimensions jointly and distinguish four patterns of elite legitimacy communication.
Low-intensity legitimation (bottom-right in Table 1) and high-intensity delegitimation (top-left) mark the extremes in both dimensions. The former corresponds to Lindberg and Scheingold's (1970, p. 41) notion of a Bpermissive consensus^frequently used to capture how International Relations scholars have tended to describe the legitimacy of IOs in the past. It indicates that elites take IOs for granted and therefore do not engage much in legitimacy communication (Suchman 1995, p. 583). To the extent that elites evaluate IO legitimacy in the media, tone is positive. This pattern is unlikely to create legitimacy pressure on IOs as their representatives rarely confront demand for change and normative adaptation in public debates. The opposite pattern of high-intensity delegitimation matches the notion of Bconstraining dissensus,^Hooghe and Marks (2009) use to describe the growing contestation of European integration. This pattern is more likely to put IOs under normative stress as they have to deal with mounting media attention, face various demands for change, and come under pressure to reform procedures and to improve performance (Lenz and Viola 2017). These pressures may also follow from low-intensity delegitimation albeit to a lesser extent because legitimacy challenges are less frequent and can be ignored more easily. Finally, high-intensity legitimation describes a situation in which proponents of IOs are willing to support IOs actively by voicing their positive legitimacy evaluations in the media. This pattern is likely to encourage IOs to maintain their course.
Little is known about the determinants of these patterns. What drives elites to discuss IO legitimacy in the media? Why do elites publicly legitimate or delegitimate IOs? To address these questions, I draw on three bundles of explanatory factors highlighted by Tallberg and Zürn (2017): the formal authority of IOs, national political cultures, and political events. The varying levels of IO authority give an impulse for legitimacy communication. Yet, authority is socially constructed and therefore likely to be viewed differently across national public spheres and concrete political contexts. Hence, I expect the impact of authority levels on elite communication to be mediated by national The authority explanation is rooted in the Weberian notion that political authority is the analytical starting point for questions of legitimacy (Weber 1978). Following this tradition, legitimacy is often defined as an actors' normative belief that a rule or institution ought to be obeyed (Hurd 1999, p. 381). This implies that demand for legitimacy and legitimation grows, with authority (Zürn 2018). This causal link was a key driver for the formation of the modern nation-state, where the accumulation of the authority to make and enforce collectively binding decisions by the state sparked elite demands for participation and representation (Tilly 1990). More recently, research on the politicization of international institutions (Zürn et al. 2012) has demonstrated a similar relationship between international authority and public legitimacy evaluations. Increasing IO authority drives the growing intensity of public debates about IOs and shapes their tone (Rauh and Zürn 2017;Rixen and Zangl 2013). This research defines authority as a Bsocial contract in which a governor provides a political order of value to a community in exchange for compliance by the governed with the rules necessary to produce that order^ (Lake 2010, p. 587). For IOs, this social contract is found in the formal rules that specify (a) the issue scope in which the IO is granted authority, (b) the tasks of political order making delegated directly to international bodies, and (c) the decision-making capacity of member states pooled in joint international bodies (Hooghe et al. 2017, p. 22). Three mechanisms link this authority to elite's legitimacy communication in the media. Most fundamentally, authority makes IOs more newsworthy. IOs that operate in many issue areas and have a high level of pooled and delegated authority create more occasions for the media to report on their activities and, thus, for elites to discuss legitimacy questions. Elites seize these opportunities for two basic reasons. First, systems of authority and their representatives attempt to establish and cultivate the belief in their legitimacy (Barker 2001;Claude 1966). Consequently, IO representatives and government actors seek to justify their organizations in the media (Zaum 2013). Second, those subject to IO authority put forward their normative demands in exchange for compliance. They are likely to question IO legitimacy in the media because international public authorities have the capacity to restrict freedom, which makes justification necessary. Overall, this combination of motive and opportunity implies that: (H 1a ) the intensity of elite legitimacy communication is likely to be higher for IOs with high political authority.
IO authority should also have an effect on tone. Since IO policy-making does no longer stop at national borders but tackles core functions of sovereign government, national elites are more frequently confronted with the directoften redistributiveeffects of IO decision-making (Barnett and Finnemore 2004, p. 1;Kreuder-Sonnen and Zangl 2015). Over the course of the past decade, this empowerment of IOs has been met with opposition from two fronts. First, not only since the election of Donald Trump and the British vote to leave the EU, we witness a new integration-demarcation cleavage, mobilizing nationalist backlash against international integration (Burgoon 2009;Kriesi et al. 2012). Because economic globalization has produced losers (Fordham and Kleinberg 2012) and because IO authority threatens national political identities (Hooghe and Marks 2009) nationalist elites have come to challenge the legitimacy of globalization and its political institutions. Second, beginning in the 1990s, the global justice movement emerged as a powerful contester of IO legitimacy (della Porta 2007;O'Brien et al. 2000). Focusing on the redistributive and ecological consequences of IO policy-making, civil society and left-wing parties have challenged IOs' role in enforcing global capitalism and US imperialism (Woods 2006). In sum, these right and left-wing critics of IOs' growing authority constitute a strong opposition, which should shape the overall tone of elite legitimacy communication in the media (Zürn 2018). Therefore, I hypothesize that: (H 1b ) the tone of elite legitimacy communication is likely to be less positive for IOs with high political authority.
While IO legitimacy may even engage observers in non-member states, I expect particularly strong effects of IO authority on the legitimacy communication of elites based in IO member states. Consequently, high-intensity delegitimation is most likely for constituencies of IOs with high authority. Low-intensity legitimation is most likely for observers of IOs that exercise little authority.
My second explanation builds on the literature highlighting the role of political cultures and discursive institutionalism (Schmidt 2010). It assumes that the media as the primary arena of elite legitimacy communication have largely conserved their national character (Risse 2010;Schneider et al. 2010). Consequently, the effect of IO authority on elite communication should be mediated by national political cultures. The literature on the Europeanization and transnationalization of public spheres has provided ample evidence for the tenacity of such cultures (Peters 2005;Risse 2010). Another pertinent literature explores the relationship between legitimacy evaluations of domestic political systems and IOs. It reveals that individuals often draw on their experience with domestic political authority as cues for their evaluation of international authority (Dellmuth and Tallberg 2015;Torgler 2008). This effect has been confirmed for the EU. Individuals with higher levels of trust in domestic political institutions are also more likely to trust European institutions (Armingeon and Ceka 2014;Harteveld et al. 2013). I expect this mechanism to shape national variation in elite legitimacy communication. Political cultures, which tend to regularly scrutinize and debate the political authority of the state and, thus, examine the issue of legitimacy more intensely in the media, should also pay more attention to IO legitimacy. Thus, I hypothesize that: (H 2a ) the intensity of elite legitimacy communication is likely to be higher where legitimacy communication on domestic political systems is more intensive.
Theoretical perspectives and empirical findings on tone are more mixed. On the one hand, survey research has demonstrated that IO legitimacy evaluations are based on actors' general confidence in political institutions. Premised on cue theory, this research shows that actors use their evaluation of domestic institutions' legitimacy as short-cuts to form opinions about IO legitimacy. In this mechanism, a general predisposition, that is, the appreciation or rejection of political authority, rather than the characteristics of IOs shapes the tone of elites' legitimacy communication (Armingeon and Ceka 2014;Dellmuth and Tallberg 2015). On the other hand, research on the politicization of international institutions has shown that political identities shape cross-country variance in IO legitimacy evaluations (Zürn et al. 2012). In this mechanism, growing IO authority constitutes a significant loss to actors Bwho possess a strong sense of identification with their national community,^ (Kriesi et al. 2012, p. 15) whereas actors with a cosmopolitan identity evaluate international authority as legitimate (Ecker-Ehrhardt 2012). In aggregate, this mechanism creates more positive elite communication about IOs in the media of societies which see domestic institutions critically due to their cosmopolitan outlook and more critical communication about IO legitimacy in countries which value domestic institutions highly due to their nationalist culture. As both mechanisms of upward extrapolation and identification may plausibly affect elite legitimacy communication in the media, it is an open empirical question whether legitimacy evaluations of domestic political systems have a positive or negative effect on the tone of communication about IO legitimacy. Hence, I hypothesize more openly that: (H 2b ) the tone of elite legitimacy communication is likely to be shaped by the tone of communication about domestic political systems.
My third explanation focuses on political events as discursive opportunities for the mobilization of support or challenges to IO legitimacy in the media (Chong and Druckman 2007;Snow and Benford 1988). The argument draws on the notion that the development of ideational constructs about standards of appropriateness is a collective, self-reinforcing process (Pierson 2004, p. 39), which is generally marked by path-dependence. Yet, this process can be interrupted by critical junctures, that is, political events and crises that create uncertainty as to the future of institutional arrangements and, thus, allow for political agency and choice to play a decisive role in setting paths for future developments (Capoccia 2015, p. 148). Since authority relationships in global governance are less deeply entrenched and taken for granted than at the national level, elites are more prone to react to such events and to engage in more volatile legitimacy communication (Zürn 2018, p. 62). Consequently, the political events, in which legitimacy communication in the media is embedded, should shape both the intensity and tone of elites' legitimacy evaluations (Imerman 2018). When a specific event, such as the Great Recession or the wars in Afghanistan and Iraq, shift to the front burner of media agendas, they are likely to focus elite attention on particular legitimacy resources or challenges of IOs and to invite affirmative or critical evaluations.
Two types of events are particularly likely to affect legitimacy communication: First, I expect international security crises to trigger a Brally around the flag effect^among political elites. As IOs are widely credited with comparative advantages in providing the public good of international security (Cottrel 2016), international security crises should create support for IO legitimacy, including more intensive and positive elite evaluations in the media. Second, political events highlighting institutional features and IO authority rather than the policy responsibilities of IOs are likely to foreground potential legitimacy challenges. Following up on the above discussion on how growing IO authority shapes legitimacy communication, I expect this effect to unfold during periods of actual authority transfer. When IO treaty reforms highlight institutional features and IO authority, elite communication in the media should become more intensive and less positive. In sum, I hypothesize that: (H 3a ) high-intensity legitimation is more likely during times of international security crises. (H 3b ) high-intensity delegitimation is more likely during times of institutional reform.

Research design
This section explains the IO and country sample, outlines the content-analytical method used to map elite legitimacy communication in the media, and operationalizes dependent and independent variables.

Measuring elite legitimacy communication
The empirical analysis focuses on elite legitimacy communication about the EU, the G8, and the UN in the media of Germany, the United Kingdom, Switzerland, and the US from 1998 to 2013. While this sample cannot prove the causal relationships discussed above, it may point the way to further research, building a comprehensive understanding of elite legitimacy communication, which is until now largely uncharted territory.
The IO sample represents the full spectrum of formal IO authority, ranging from low-authority informal clubs to highly integrated, formal IOs (Vabulas and Snidal 2013). The EU is clearly the IO with the highest level of formal authority in the sample. The UN represents classical post-Second World War IOs with a medium level of authority. The G8 is an informal club that has only policy coordinating functions and hardly any formalized authority. While this sample helps to vary levels of IO authority, it is not representative of the entire IO population. Since all three cases are highly important general-purpose IOs, my empirical results are limited to this segment of the population and cannot be generalized directly to less influential, task-specific IOs.
The period of analysis spans both a set of important international events, including security crises, and a set of institutional reforms in all three selected IOs. Each year is represented by a ten-day window around the major summits of the sampled IOs. During these meetings, the normative foundations of IOs are regularly reevaluated and if necessary institutional reforms to bring IOs in line with elite legitimacy demands are put on the agenda. Hence, IO summits are, among other things, key windows of opportunity to discuss IO legitimacy in the media and to create normative support or pressure (see Online Appendix A2 for details on sampling). 1 The country sample focuses on western democracies because these states largely dominate the selected IOs. Elite legitimacy communication in these public spheres is essential. My empirical results cannot be generalized beyond this type of states, especially not to countries on the receiving end of international cooperation. In countries dominated by IOs, elite legitimacy evaluations are likely to follow different logics. This is, for instance, indicated by the observation that states not represented on the UN Security Council evaluate the legitimacy of this body more critically than its members (Binder and Heupel 2015, p. 244). Furthermore, the focus on western democracies is also premised on the forum in which elite legitimacy communication is studied. Freedom of press and expression are important scope conditions to analyze elite communication in the media. Consequently, the results of this analysis can also not be generalized to autocratic regimes in which these freedoms are restricted.
Since I expect constituencies in IO member states and observers in non-member states to evaluate IO legitimacy with different intensity and tone, the sample includes member states of the EU (Germany and United Kingdom), the G8 (Germany, United Kingdom, and the United States), and the UN (which Switzerland joined in 2002) along with non-member states. The four countries, moreover, represent different types of democratic government and political cultures. Elite evaluations of domestic political systems vary substantively across countries but relatively little over time. In the UK, high-intensity delegitimation dominates elite debates, whereas high-intensity legitimation is the modal pattern in the US. In Germany and Switzerland, elites evaluate the legitimacy of their domestic system less intensely. Tone is more positive (low-intensity legitimation) in Switzerland and more critical (low-intensity delegitimation) in Germany (see Appendix A4.2). I expect these differences to shape the intensity and tone of elite communication on IO legitimacy.
I use quality newspapers, that is, newspapers which in comparison to tabloids report political matters more extensively, in more serious tone, and in higher level language (Cole and Harcup 2009, p. 31) as a source for tracing legitimacy communication. The rationale for this focus on the quality press is twofold. First, the quality press is a key political venue through which elites compete with one another to foster new ideas and to promote their evaluations of IO legitimacy. Second, the quality press tends to give voice to actors perceived to be influential and reliable sources due to their position in the political system (Binderkrantz et al. 2017, p. 315). The ebbs and flows these legitimacy evaluations are likely to be consequential as they convey legitimacy pressures to international bureaucrats and because they provide cues to citizens (King et al. 2017).
A potential risk of focusing on the quality press is that access is not equally distributed and that elites might have limited control over which legitimacy evaluations are covered. Due to limited capacity, journalists often rely on a set of routines to select their sources. Existing literature offers useful guidance on this issue. On the one hand, the media has been shown to grant preferential access to high-ranking state actors and resources-rich business organizations (Danielian and Page 1994;Tresch 2009). This focus on actors with high political and economic status could bias my results on legitimacy communication. Because quality newspapers do not give equal access to all elites, the voice of less prominent but potentially very active actors might be underrepresented. As a result, my data might underestimate communication intensity and provide a skewed picture of communication tone. On the other hand, more recent research, unpacking the concepts of status and prominence, demonstrates that media access is most strongly shaped the elites' activity and engagement. Actors that often publish press releases, hold press conferences, and interact regularly with journalists are more likely to have media access (Andrews and Caren 2010;Oehmer 2017). This implies that actors who invest more time and resources to make their voices heard because this is important to them are more likely to get access to the media. This effect of elites' public relations efforts on their news access appears to be even stronger in the context of international news coverage. Because journalists often lack the necessary expertise about international politics to select sources according to status, they are compelled to rely on less prominent but proactive sources, including civil society (De Bruycker and Beyers 2015; Van Leuven and Joye 2014).
On balance, there is reason to believe that the quality press gives relatively equal access to those elites most strongly determined to have their evaluations of IO legitimacy heart. This includes actors often disadvantaged in the context of national reporting. Especially civil society actors and IOs themselves have greatly developed their central capacities for public communication. They use these tools extensively to gain access to the media, mobilizing and shaping public opinion (Ecker-Ehrhardt 2017; Tallberg et al. 2018). My data support this expectation. While journalists (50% of all legitimacy evaluations) and national executives (almost 20% of all legitimacy evaluations) dominate in all four countries, both civil society actors (15% of all legitimacy evaluations) and international bureaucrats (almost 10% of all legitimacy evaluations) are also important voices in elite legitimacy communication. Journalists and civil society actors are highly critical (approximately 20% positive evaluations). National executives (almost 50% positive evaluations) and international bureaucrats (approximately 40% positive evaluation) are more supportive.
Moreover, media research shows that newspapers differ ideologically and can be ordered on a left-right spectrum (Hallin and Mancini 2004;Lichter 2017) because they give preferential access to actors and present issues in ideologically biased ways. To avoid this potential bias, I analyze the content of one center-left and one center-right quality newspaper per country. For Germany and the UK, I follow the classification by Koopmans and Statham (2010, p. 52) and examine the German Frankfurter Allgemeine Zeitung (center-right) and Süddeutsche Zeitung (center-left) and the British Times (center-right) and Guardian (center-left), respectively. The selection of the Swiss Neue Zürcher Zeitung (center-right) and Tagesanzeiger (center-left) builds on the classification by Tresch (2009, p. 76). For the US, I deviate slightly from these criteria. I select one centrist (Washington Post) and one center-left (New York Times) newspaper because the Wall Street Journal, which is the only center-right quality newspaper in the US focuses more on business and financial news than the other newspapers in the sample (Budak et al. 2016;Gentzkow and Shapiro 2010). Empirically, communication intensity is, with the exception of British newspapers, highly similar for the two selected newspapers per country. In the case of Germany, the Frankfurter Allgemeine Zeitung contributes 51% of legitimacy evaluations. In Switzerland, the picture is similar. The Neue Zürcher Zeitung contributes 54% of legitimacy evaluations. For the US case, the New York Times contributes 56% of legitimacy evaluations, whereas the British debate is dominated by the Times which contributes 63% of legitimacy evaluations. The tone of legitimacy communication in these newspapers is also relatively similar. With 32% positive evaluations the Washington Post is most positive, whereas the Times is most negative (20% positive evaluations). With the exception of the UK, where the center-left Guardian is more positive than the center-right Times, center-right newspapers are slightly more supportive of IO legitimacy than center-left newspapers. Overall, the newspaper sample can be considered to provide a balanced picture of elite legitimacy communication in the selected countries.
The coding procedure identifies individual legitimacy evaluations in newspaper articles. These evaluations are defined as Ba generalized perception or assumption that the actions of an entity are desirable, proper, or appropriate within some socially constructed norms, values, beliefs, and definitions^ (Suchman 1995, p. 574 Easton 1965, pp. 436-453 for this distinction) because such statements of specific support are not rooted in shared norms or social purpose that underlies legitimacy. 2 Legitimacy evaluations are coded in three steps. First, all articles mentioning the selected IOs were downloaded from the electronic newspaper archive Factiva. 3 Second, articles in which the IOs were not evaluated were discarded because they can, by definition, not reveal elites' legitimacy evaluations. Third, the remaining articles were searched for legitimacy evaluations. This coding was done by a team of six researchers based on a formal codebook. 4 It zeros in on individual evaluative propositions. A proposition qualifies as a legitimacy evaluation if it explicitly assesses one of the selected IOs (legitimacy object), drawing on a normative criterion (legitimacy standard). Only positive or negative evaluations (tone) generalizing beyond policies and actors were coded because the analysis focuses on the overall legitimacy of IOs. Other than journalists' evaluations, the coding distinguished government speakers, international bureaucrats, and civil society actors, including academics, economic actors, and NGOs. Speakers were identified according to the manifest text of the newspaper article. 5 In addition, coding determined the event context of the paragraph in which legitimacy evaluations are embedded. Table 2 gives further examples of legitimacy evaluations (further details on text retrieval and coding procedures are presented in Online Appendix A1).
The coding procedure yielded a dataset of 6658 legitimacy evaluations. To map and explain the patterns of elite legitimacy communication, I first aggregated the data on evaluations into a Bbalanced panel^of 192 (3 × 4 × 16) BIO-country-year^cases, my unit

Security event
Elite legitimation and delegitimation of international... of analysis. Second, intensity and tone were considered together to classify these cases into the four patterns of elite communication (see Online Appendix A3). My measure of intensity is the number of propositions, evaluating the EU, the G8, or the UN in a given national public sphere and year. The measure has an absolute minimum of zero, indicating the absence of legitimacy communication, but no absolute maximum. Cases were classified as high-intensity if the number of evaluations was higher than the mean number of evaluations across the 192 IO-country-years (34.7 evaluations). A value lower than the mean indicates low-intensity. My indicator of tone is the percentage of positive evaluations of the selected IOs in each national public sphere and year. The measure has a minimum of zero and a maximum of 100%. Cases were identified as legitimation if the share of positive evaluations was above the mean across all cases (27.3% positive evaluations). Values lower than the mean indicate delegitimation. 6 Consider elite communication in the Swiss media about the 1998 EU summit as an example. This meeting attracted 29 elite evaluations of which 13.8% were positive. As both intensity and tone are below the sample mean, I classified this case a low-intensity delegitimation. I have chosen mean values over absolute classification thresholds for theoretical and methodological reasons. Theoretically, legitimacy research (irrespective whether it uses survey data, behavioral indicators, or public communication) has long suffered from the difficulty to define a convincing threshold between legitimacy and illegitimacy (Gilley 2006). What share of positive legitimacy evaluations constitutes legitimation cannot be answered theoretically without reference to context. Consequently, I base my analysis on a relative measure, which takes context into account. When intensity and tone are above average, they are already more intensive and positive than usual and in that sense high-intensity legitimation. Methodically, the relative measure has the advantage to tease out more clearly variation in legitimacy communication, which is the focus of this paper. An absolute cutoff point of, for instance, 50% of positive evaluations would have resulted in a concentration of cases at the delegitimation-end of the continuum.

Independent variables
I apply a multinomial logistic regression model to probe the determinants of elite communication patterns. I analyze the effect of IO authority with the help of data on the level of formal authority pooling in intergovernmental IO-bodies and authority delegation to IO bureaucracies, judicial bodies, and parliamentary assemblies provided by Hooghe et al. (2017). The data provide time series information on the annual development of authority for 76 IOs, including the EU and the UN, from 1950 to 2010. Based on the coding of the formal rules in IO treaties, constitutions, and conventions, the dataset measures how much formal authority states have granted to IOs and their individual bodies (delegation) and how much authority states have transferred to collective IO member state bodies (pooling). I use the sum of pooled and delegated authority, as I expect both types of authority to shape patterns of legitimacy communication. Since the G8 is not included in the dataset, I set a value of zero for this informal IO which has not been substantively empowered during the observation period. As the time series by Hooghe and colleagues do not provide data for the most recent years of my observation period, I assume the authority of the EU and the UN to remain constant from 2011 to 2013. In addition, I employed an indicator variable for IO membership, with a value of one, if the country in the sample is a member of the selected IOs, and a value of zero, if it is not.
To probe the effect of political cultures, I draw on a text corpussimilar to my dataset on IO legitimacy communicationof legitimacy evaluations in newspaper articles of the Swiss, German, British, and US political systems in the years 1998-2013 (Haunss et al. 2015). My measure of national intensity is the number of legitimacy evaluations in a given country-year. My measure of national tone is the share of positive evaluations in a given country-year.
The empirical basis of my measure for event contexts is the dataset on elite legitimacy communication. I draw on the coding of the political events highlighted in the paragraph in which a legitimacy evaluation is embedded. A security event was, for instance, coded, when Kofi Annan asserted that Bwhen states decide to use force to deal with broader threats to international peace and security, there is no substitute for the unique legitimacy provided by the United Nations^(New York Times, 12. September 2002). An institutional event was, for example, coded when Gerhard Schröder, commenting on the failed Dutch referendum on a European Constitution Bblames fear of the future in the face of globalisation for the growing public disenchantmentperhaps even disgust tinged with racismwith the EU^(The Guardian, 18 June 2005). Both variables measure the percent share of these event contexts in a given IO-countryyear. I excluded frames referring to domestic issues as a residual category. This operationalization has the advantage over event count data that it reveals which events out of many possible actually played a role in the media. Finally, to control for the legitimation strategies of national political actors and IO representatives (Ecker-Ehrhardt 2017), I include a variable taken from the content analysis. It measures the percent share of evaluations contributed by government speakers, that is, national political establishment and international bureaucrats. To facilitate the interpretation of the multivariate analysis, I z-standardize all continuous explanatory variables. For details on independent and dependent variables, see Online Appendix A3, A4, and A5.

Empirical analysis
How intensive is elite legitimacy communication in the media? Is its tone negative or positive? Finally, what explains these patterns of elite legitimacy communication? The empirical analysis of these questions proceeds in two steps: First, I present descriptive findings on patterns of elite legitimacy communication. Second, I employ multinomial logistic regression to probe my explanations of these patterns.

Patterns of elite legitimacy communication
Descriptively, I make three core observations. First, IO legitimacy is an enduring topic for elite communication in the media. The IO summits in the sample constantly confront the evaluation of IO legitimacy. Despite this continuous flow of elite legitimation and delegitimation, intensity varies considerably across IOs and over time. Second, on average the tone of elite evaluations in the media is highly critical. Third, variance in intensity and tone results in strong temporal fluctuations of elite communication patterns. Instead of a clear trend towards high-intensity delegitimation, which would signal nationalist backlash, all IOs in the sample have to deal with different patterns in the media of the analyzed countries at different points in time. Consequently, IOs are, at times, confronted with a complex combination of different levels of intensity and tone (see Online Appendix A3 for further descriptive data and explanations). Figure 1 illustrates this complexity, displaying the distribution of elite legitimacy communication patterns across IO-country-years. The figure shows that elite legitimacy communication in the media features all four patterns. Over time, the three IOs in the sample confront all four patterns. In contrast, to the literature that agrees on the sui generis nature of the EU (Phelan 2012) or the specificities of informal IOs (Vabulas and Snidal 2013), such as the G8, western elites do not generally differentiate between the three IOs in their legitimacy evaluations. Furthermore, individual IOs are frequently subject to different patterns of elite communication in the media of the four countries at a given point in time. A case in point is the 2004 UN summit which triggered highintensity delegitimation in the UK, high-intensity legitimation in Germany, and lowintensity delegitimation in the US and Switzerland. In contrast, only eight out of the 48 analyzed summits show a consistent pattern across countries. Consequently, IOs face complex normative environments. They have to maneuver different combinations of communication intensity and tone at a given point in time. While some elites may provide communicative support others exert legitimacy pressures, thus signaling conflicting demands and expectations.
This complex mix of patterns varies across IOs. The EU is most frequently confronted with high-intensity delegitimation (in 37.5% of the cases) whereas highintensity legitimation is least frequent (17.2%). The G8 most often faces low-intensity delegitimation (45.3%) and least frequently high-intensity legitimation (3.1%). The UN is most frequently subject to low-intensity legitimation (43.8%), whereas high-intensity   IO-country-years, 1998IO-country-years, -2013 delegitimation is least frequent (9.4%). Consequently, the idea that IOs might benefit from a Bpermissive consensus,^in which elites are generally less attentive but supportive when evaluating IO legitimacy in the media, finds little support as low-intensity legitimation is not the dominant pattern in elite communication.
Communication patterns also vary over time. While all three IOs in the sample confront high-intensity legitimation and delegitimation during the first half of the observation period, the picture changes during the second half. Elite communication on the UN shifts, with the exception of British high-intensity delegitimation in 2009, completely to the two low-intensity patterns. Similarly, elite attention to the legitimacy of the G8 fades between 2010 and 2013, when high-intensity patterns are no longer present. By contrast, elite attention to the legitimacy of the EU remains high over the entire observation period, as high-intensity delegitimation still occurs during the second part of the observation period. These temporal fluctuations show that elite legitimacy communication in the media is volatile and shaped by country-specific political contexts. In contrast to current debates on a new era of Bconstraining dissensus( Hooghe and Marks 2009) marked by populist nationalism, high-intensity delegitimation is not the dominant pattern. Rather, periods of public elite support alternate with periods of legitimacy pressure.

Multivariate analysis
What explains this variation in elite legitimacy communication in the media? To predict the occurrence of the four patterns, I specify a multinomial logistic regression with this four-category outcome variable (Agresti 2002) as: 3 nat tone i þ β HID 4 ev sec i þ β HID 5 ev inst i þβ HID 6 speaker gov i The model estimates the risk of low-intensity delegitimation (LID), high-intensity legitimation (HIL), and high-intensity delegitimation (HID), relative to low-intensity legitimation (LIL) as a baseline, based on the values of the explanatory variables authority, membership (member), national intensity (nat_int), national tone (nat_tone), security event (ev_sec), institutional event (ev_inst), and government speakers (speaker_gov). 7 I use a multinomial instead of an ordinal logistic regression because the four patterns cannot be ordered in a theoretically meaningful way. Given the hard to interpret nature of multinomial logistic regression coefficients, I present the results as exponentiated coefficients. Each exponentiated coefficient represents the relative risk (analogous to an odds ratio) of changing from low-intensity legitimation to the other three patterns, given a one-unit change in the explanatory variable. To illustrate the interpretation of the results from Table 3, consider the effect of a one-unit change in the authority variable on the relative risk of high-intensity legitimation. Look at the authority row of coefficients associated with the occurrence of high-intensity legitimation. A one-unit change in this variable, that is, an increase of one standard deviation (0.38) increases the relative risk of high-intensity legitimation in comparison to low-intensity legitimation by a factor of 3.122. In other words, the relative risk of high-intensity legitimation in comparison to low-intensity legitimation is 212.2% higher for IOs with one standard deviation more authority.
Global tests of fit confirm that the model represents a significant improvement on intercept-only models. The overall explanatory power of the model is indicated by the McFadden pseudo-R 2 (0.21) and the high percentage (59.4) of correctly assigned cases (see Online Appendix A6). 8 The model is particularly successful in assigning lowintensity legitimation, low-intensity delegitimation, and high-intensity delegitimation cases but fails with regard to the rare high-intensity legitimation pattern.
The analysis grants no support to the argument that the intensity and tone of elite legitimacy evaluations of domestic political systems shape western elites' legitimacy communication about IOs. This result suggests that the findings from research on the national character of public spheres and on the effects of national contexts on citizens' perceptions of IO legitimacy do not extend to elites. It seems that western elites have developed a shared understanding of legitimacy in global governance and discuss IO legitimacy in the media on a common ground. The strongest results are found for the authority and political events explanations. Authority and membership have a strong positive and significant statistical effect on communication intensity. Similarly, institutional events have a positive and significant statistical effect on communication intensity, whereas security events have a unique negative and significant statistical effect on high-intensity delegitimation.
Turning to individual parameter estimates, I find, first, that the level of IO authority significantly affects the risk of the four communication patterns. Both measures -IO 7 Multinomial logistic regression maintains two basic assumptions. First, it assumes that the introduction of another alternative will not affect the relative probabilities of the other patterns of legitimacy communication. I test this independence of irrelevant alternatives assumption (IIA) with the Hausman test (Hausman and McFadden 1984). The results suggest that IIA is violated (see Online Appendix A7.3). Yet, as Cheng and Long (2007) demonstrate, these tests perform poorly for small sample sizes and the violation of the IIA assumption does not cast doubt over the reliability of the parameter estimates. Second, non-perfect separation is assumed. Both parameter estimates and the respective standard errors suggest that this assumption is not violated. 8 The relevant benchmark is not 50% (as in a coin toss), but 25% (as the outcome variable has four categories) ormore conservativelythe 35.4% share of low-intensity delegitimation (the modal category of the outcome variable). authority and membershiphave highly significant and strong positive effects on the relative risk of the two high-intensity patterns. All else equal, the relative risk of highintensity legitimation in comparison to low-intensity legitimation is 212.2% higher for IOs with one standard deviation more authority. The relative risk of high-intensity delegitimation is 132.1% higher. The effects are even stronger for IO-membership. The relative risk of high-intensity legitimation increases by 402.3% and the relative risk of high-intensity delegitimation grows by 413% for elite communication in the media of IO member states. Based on my sample of major multi-purpose IOs, these results grant strong support for hypothesis H 1a . Higher levels of IO authority increase the relative risk of high-intensity communication in the media. A country's membership in important IOs and exposure to their authority also increases communication intensity. However, there are two important limitations to the explanatory power of these variables. First, IO authority cannot fully account for the observed variation over time, as IO authority and membership is less dynamic than the observed patterns. Second, the explanatory power of both variables for tone is limited. This suggests that western elites are more inclined to engage with the legitimacy of powerful IOs but that the tone of evaluations does not depend on authority levels. Rather, elites provide intensive communicative support and exert more legitimacy pressure on IOs with more authority. McFadden pseudo-R 2 0.212 All coefficients are exponentiated to represent risk ratios, relative to low-intensity legitimation as a baseline. The 95% confidence intervals are in brackets *p < 0.1; **p < 0.05; ***p < 0.01 Second, elite legitimacy evaluations of domestic political systems in the media of western democracies do not affect elite communication about IOs. Both variables on national intensity and tone are non-significant. This result contradicts much of the literature on the effects of national public spheres and identities on individual legitimacy evaluations (Armingeon and Ceka 2014;Dellmuth and Tallberg 2015). While this research demonstrates that citizens draw cues on IO legitimacy from their national contexts, my analysis suggests that elites in western democracies are better informed and consequently more sensitive to IOs' objective features and political events than citizens.
Finally, the results for political events support this conclusion. Institutional events have a strong positive effect on both high-intensity patterns, highlighting that changes in prominent IOs' institutional design are controversial among western elites. An increase of 28.9% in the share of institutional events shaping IO summit agendas increases the relative risk of high-intensity legitimation by 258.6% and the relative risk of high-intensity delegitimation grows by 151.7%. This suggests that elites of western democracies do not generally perceive institutional changes negatively but evaluate them more specifically, granting either communicative support or exerting normative pressure. A typical case of strong delegitimation in the media is the evaluation of the 2000 EU summit by British elites. During this meeting, EU legitimacy is evaluated 82 times in 34 newspaper articles. Of these evaluations, 84% are negative and largely related to institutional reform. Many negative evaluations explicitly criticize the pooling of decision-making authority at the EU level. William Hague, for instance, accuses BEU leaders of taking three big steps towards a European 'superstate'^and attacks Tony Blair for giving up BBritain's veto on 23 areas, allowing other members to impose further integration^ (Times, 12 December 2000). A typical example of communicative support in the media for institutional change is the evaluation of the 2004 EU summit by German elites. Out of 83 evaluations in 21 articles, 35% are positive and frequently tied to the positive role the constitutional treaty could play for European integration. A German journalist argues, for instance, that a European constitution Bwould prescribe the EU quasi-state more democracy and more control by independent judges^ (Süddeutsche Zeitung, 25 March 2004). 9 In line with H 3a , security events have a more specific effect, as western elites evaluate the legitimacy of prominent IOs in this context less intensely and less negatively. While security events do not increase the risk of high-intensity legitimation, a one-unit change in their share (an increase of 28.1%) significantly decreases the relative risk of high-intensity delegitimation by 60.2%. This suggests that western elites are less inclined to challenge the legitimacy of prominent IOs in times of security crises. Although elites do not necessarily rally around the flag, they do also not exert strong legitimacy pressures during these periods. A typical example of this effect is the evaluation of the 2001 UN summit in the media. Just two months after the 9/11 terrorist attacks, elites rarely evaluate UN legitimacy. With a maximum of 29 evaluations by US elites and minimum of just one evaluation by Swiss elites, this meeting generates low communication intensity in the media of all four countries, with a more positive tone in Switzerland and Germany and a more negative tone in the US and UK. Elites from all four countries tie their evaluations almost exclusively to terrorism as an international 9 Own translation. security threat. George W. Bush notes, for instance, Bwe will defend ourselves and our future against terror and lawless violence. The United Nations was founded in this cause^ (Washington Post, 11 November 2001).
Finally, the results show that the government speaker variable has significant negative effects on the relative risks of high-intensity delegitimation and lowintensity delegitimation. A strong presence of western government speakers in elite communication decreases the relative risk of delegitimation. This shows that in contrast to the literature on national executives' attempts to shift blame for unpopular decisions to IOs (Daugbjerg and Swinbank 2007;Hobolt and Tilley 2014), western political elites and international bureaucrats are more likely to engage in legitimation strategies to support the work of their important IOs (Ecker-Ehrhardt 2017; Gronau and Schmidtke 2016).
In sum, the multivariate analysis suggests that elite communication in the media of established western democracies about the legitimacy of important IOs is driven by a combination of two bundles of explanatory factors. Formal IO authority and specific political contexts shape a complex normative environment for IOs. Elites are considerably more likely to engage in legitimacy communication about powerful IOs of which their home country is a member. Yet, intensity and tone fluctuate over time due to the political context in which IO summits take place and thus make normative support and legitimacy pressures in the media difficult to predict for IOs.
Due to the restricted sample of IOs and countries, these results cannot be generalized to the larger population of organizations and elites from other world regions. To demonstrate that my results areat leastrobust under these scope conditions I estimate a set of alternative models (see Online Appendix A7.4). First, I use median instead of mean communication intensity and tone as classification thresholds to account for the potential effects of outliers which are more likely to affect mean values. With the exception of a less significant effect of security events on high-intensity delegitimation, the results of this model are almost identical to those presented in Table 3.
Second, to examine the potential effects of the clustered data structure, I estimate the model with robust, clustered standard errors. While these models produce statistically significant effects of national communication intensity and tone on the risk of high and low-intensity delegitimation, they are similar to results for the other variables presented in Table 3.
Third, to account for potential secular trends in the data I rerun the model for two time periods (1998-2005 and 2006-2013). Taking into account the very low number of cases on which these estimations are based, the result of the model for the early time period are comparable to those presented here. Especially, the coefficients for IO authority and membership are similar. The low number of cases makes the model for the latter period difficult to interpret. As the number of high-intensity legitimation cases is low for this period the model produces very high coefficients and standard errors, pointing to the instability of the model. While this test is not fully conclusive, it suggests that secular trends do not fundamentally shape my results.
Finally, to account for spatial biases in the data I rerun the model based on different subsamples of the data, excluding individual IOs and countries (jackknife resampling). Despite the very low number of cases used in these estimations, the coefficients are largely similar to those presented in Table 3. This indicates that neither specific IOs nor countries critically drive my results.

Conclusion
This article examined elite communication in the media of established democracies about the legitimacy of major multi-purpose IOs. This legitimacy communication matters for three reasons. First, elites' support in the media is important for the effectiveness of IOs and for national governments engaging in international cooperation. Second, citizens draw on elite cues presented in the media to form legitimacy beliefs. Third, the legitimacy pressure resulting from high-intensity delegitimation might drive IOs to reconcile their institutional design and policy-making with the legitimacy standards held dear by elites. Based on a dataset that captures the intensity and tone of elite legitimacy communication on three IOs in the media of four western democracies between 1998 and 2013, I make descriptive and causal inferences on the contours and determinants of patterns of elite legitimacy communication.
My argument can be summarized in terms of two principal conclusions. First, major IOs cannot rely on a permissive elite consensus on global governance and its normative foundations. Instead, they face continuous elite communication about their legitimacy in the media. This communication is generally critical and puts, at times, firm legitimacy pressure on IOs. Yet, the strength of this pressure varies across IOs and alternates with communicative support. While the EU frequently confronts high-intensity delegitimation, the UN more often experiences communicative support.
Second, for the limited sample of IOs and countries presented here, variation in patterns of elite communication is best explained by a combination of two principal factors. Formal IO authority accounts for variance in communication intensity across IOs and countries as more powerful IOs attract more elite communication. Political contexts and events help to explain the ebbs and flows of intensity and tone. Together, these factors account for elites' normative support for and pressure on IOs in the media. Further research, extending the data basis to other types of IOs, countries, and public arenas will be needed to establish the generalizability of these findings.
Beyond these results, the article generates implications for three areas of research on IOs. The first is the literature on IOs' indirect modes of governance (Abbott et al. 2016;Andonova 2017) highlighting that IOs can enhance their governance capacity by drawing on the capabilities of third parties. When IOs and elites have common goals, they can join forces and use their complementary capabilities. In this context, elites can offer access to relevant constituencies, material and personnel resources, and legitimacy (Abbott et al. 2015, p. 21). Regarding the latter, I observe that elites' communicative support in the media is rather rare and volatile. Consequently, western elites' communicative support in the media appears to be a governance resource of limited use for major IOs.
Second, the article carries implications for research on citizens' legitimacy perceptions of global governance. Elite cues published in the media and national political cultures are identified as central determinants of individual legitimacy beliefs (Schlipphak 2015;Torgler 2008). More specifically, citizens tend to have more negative beliefs about IO legitimacy, when they regularly receive negative evaluations of IO legitimacy by domestic political elites and civil society actors (Dellmuth and Tallberg 2017). My findings suggest that in western societies the effect of elite cues is likely to be stronger for high-authority IOs and for citizens in IO member states as high-intensity delegitimation, which should have the strongest effect on citizens, occurs more often in the media if IOs are powerful. Furthermore, as western elites' legitimacy communication in the media is not shaped by perceptions of domestic systems' legitimacy but by a common understanding of legitimacy in global governance, elite cues should help to level strong cross-country differences in citizens' beliefs over time.
Finally, my results have implications for the literature on institutional development in global governance (Fioretos 2011;Lipscy 2015). While both historical and rational choice institutionalism tend to highlight the reasons for IOs' resistance to change (Hawkins et al. 2006;Koremenos et al. 2001), more recent studies have shown that firm legitimacy pressures drive IOs to adapt their institutional designs to normative demands (Zürn 2018;Lenz and Viola 2017). My findings suggest that such periods of high-intensity delegitimation in the media, potentially forcing IOs to reconcile their institutions with normative demand are rather rare. Clearly, high-intensity delegitimation is a common pattern in elite legitimacy communication. Yet, elites from different countries rarely agree on their evaluations at a given point in time. During the period of analysis, there are only two years in which all four national spheres exert high legitimacy pressures on an IO. In all other years, high-intensity delegitimation by some national elites is countered by highintensity legitimation by others. Consequently, western elites' communication in the media, more often than not, signals conflicting evaluations and demands to IOs. Furthermore, high-intensity delegitimation, like the three other patterns of elite communication, fluctuates strongly over time. Assuming that institutional adaptation is often a slowmoving process, which requires continuous pressure over an extended period (Grigorescu 2015), my results indicate that legitimacy pressure in the media is likely to shape institutional development only under specific conditions, including a high level of IO authority and conductive political events that keep IO legitimacy on the media agenda.