Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

The psychology of persuasion in global politics : global image, source cues, and U.S. soft power Dragojlovic, Nicolas Isak 2010

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2010_fall_dragojlovic_nicolas.pdf [ 3.73MB ]
Metadata
JSON: 24-1.0071334.json
JSON-LD: 24-1.0071334-ld.json
RDF/XML (Pretty): 24-1.0071334-rdf.xml
RDF/JSON: 24-1.0071334-rdf.json
Turtle: 24-1.0071334-turtle.txt
N-Triples: 24-1.0071334-rdf-ntriples.txt
Original Record: 24-1.0071334-source.json
Full Text
24-1.0071334-fulltext.txt
Citation
24-1.0071334.ris

Full Text

The Psychology of Persuasion in Global Politics: Global Image, Source Cues, and U.S. Soft Power  by  Nicolas Isak Dragojlovic B.A., Yale University, 2003 M.A., The University of British Columbia, 2004  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in The Faculty of Graduate Studies (Political Science) THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) September 2010 ! Nicolas Isak Dragojlovic, 2010  ABSTRACT This dissertation applies the concept of source cue effects – drawn from the political psychology literature – to the study of transnational persuasion in International Relations. As a theoretical contribution, it elaborates a set of hypotheses about the conditions under which transnational source cue effects are most likely to occur. As a substantive contribution, it focuses on how foreign publics’ perceptions of the United States and of the U.S. President shape U.S. soft power. It explores how the U.S. President’s global popularity shapes the U.S. government’s transnational persuasiveness and influences the United States’ country image, and investigates the ways in which a negative U.S. country image (anti-Americanism) constrains its foreign policy. The empirical contribution of this dissertation consists primarily of a set of original computer-based source cue and priming experiments administered to undergraduate students at the University of British Columbia in the spring and winter of 2009. The results of these experiments show that foreign actors like Presidents Obama and Clinton and the United States can exert source cue effects on a Canadian audience. Furthermore, they point to three key moderators of transnational source cue effects: 1) the popularity of the source in question, 2) the audience’s familiarity with the source, and 3) the way in which the message source is covered in the news media. The experiments also show that priming a popular U.S. President like Barack Obama significantly improves participants’ overall attitudes towards the United States when compared to their attitudes if an unpopular president like George W. Bush is primed. In a separate empirical analysis, the dissertation also uses a cross-national dataset to show that aggregate attitudes towards the United States were associated with state decisions to participate in  ii  the U.S.-led invasion of Iraq and their subsequent decisions to withdraw their forces from the operation.  iii  PREFACE This dissertation is the sole work of Nicolas Dragojlovic. A version of Chapter 5 has received an invitation to revise and resubmit from the Journal of Conflict Resolution. All experimental research presented in this dissertation was approved by UBC’s Behavioural Research Ethics Board (BREB). Approval certificates for the original application and all subsequent amendments and extensions are included in Appendix E (UBC BREB number H08-02291).  iv  TABLE OF CONTENTS ABSTRACT ..........................................................................................................................................................ii PREFACE ............................................................................................................................................................iv TABLE OF CONTENTS .................................................................................................................................... v LIST OF TABLES .............................................................................................................................................vii LIST OF FIGURES ..........................................................................................................................................viii ACKNOWLEDGEMENTS...............................................................................................................................ix DEDICATION...................................................................................................................................................... x CHAPTER 1  INTRODUCTION ................................................................................................................ 1  1.1 ANTI-AMERICANISM AND U.S. SOFT POWER ........................................................................................... 4 1.2 ACTOR EVALUATIONS, SOURCE CUE EFFECTS, AND PERSUASION IN GLOBAL POLITICS ....................... 6 1.2.1 Source Cues ........................................................................................................................................ 8 1.3 DOES G LOBAL OPINION MATTER? ..........................................................................................................12 1.4 RESEARCH DESIGN ................................................................................................................................17 1.4.1 Advantages of the Experimental Design ........................................................................................18 1.4.2 Limitations of the Research Design................................................................................................19 1.5 STRUCTURE OF THE D ISSERTATION .........................................................................................................23 CHAPTER 2  BEYOND FRAMING: SOURCE CUES IN IR..............................................................25  2.1 INTRODUCTION .........................................................................................................................................25 2.2 PERSUASION IN INTERNATIONAL RELATIONS .........................................................................................26 2.2.1 Source Cues in Political Psychology .............................................................................................29 2.3 UNDER WHAT CONDITIONS DO SOURCE CUES MATTER? .....................................................................32 2.3.1 Factor #1: Source Popularity........................................................................................................32 2.3.2 Factor #2: Audience Familiarity with the Source.......................................................................34 2.3.3 Factor #3: Media Environment and Policy Salience ...................................................................36 2.4 RESEARCH D ESIGN AND EMPIRICAL A NALYSIS.....................................................................................38 2.4.1 Experiments #1 and 2: U.S. Presidents and Climate Policy .......................................................39 2.4.1.1 Presidential Source Cue Effects................................................................................................................42 2.4.1.2 Source Familiarity and Transnational Source Cue Effects ....................................................................46  2.4.2 Experiment #3: US-Canadian Cooperation on Climate Negotiations.........................................50 2.5 D ISCUSSION AND CONCLUSION ..............................................................................................................56 CHAPTER 3  THE GLOBAL BULLY PULPIT AND U.S. SOFT POWER.......................................59  3.1 INTRODUCTION .........................................................................................................................................59 3.1.1 The U.S. Presidency and Transnational Persuasion.....................................................................61 3.2 THEORIES OF PRESIDENTIAL I MAGE AND PRESIDENTIAL PERSUASIVENESS .........................................63 3.2.1 Information and Presidential Cues .................................................................................................65 3.2.2 U.S. Salience and Decreased Trust .................................................................................................66 3.2.3 Issue-Specific Credibility ................................................................................................................67 3.3 EMPIRICAL TESTING – TWO SURVEY EXPERIMENTS ..............................................................................70 3.3.1 Experiment #1: Mandatory Emissions Limits (MEL) ..................................................................71 3.3.1.1 Findings .......................................................................................................................................................72 3.3.2.1 Limits to Persuasiveness............................................................................................................................76 3.3.2.1.1 Information ........................................................................................................................................76  3.3.3 Experiment #2: The Afghanistan Experiment...............................................................................78 3.3.3.1 Issue-Specific Credibility ..........................................................................................................................80 3.3.3.2 U.S. Salience and Perceptions of Self-Interest.........................................................................................87  3.4 DISCUSSION AND CONCLUSION ...............................................................................................................90  v  CHAPTER 4  PRESIDENTIAL PRIMING AND ANTI-AMERICANISM........................................94  4.1 INTRODUCTION .........................................................................................................................................94 4.2 LEADER EFFECTS ON COUNTRY I MAGE ...................................................................................................96 4.2.1 A Model of Attitudes Towards the United States...........................................................................97 4.2.2 Hypotheses .......................................................................................................................................98 4.2.3 Presidential Transitions and the Presidential Priming Effect....................................................100 4.3 DATA AND RESULTS ..............................................................................................................................100 4.3.1 Experiment #1: Mandatory Emissions Limits.............................................................................102 4.4 TESTING FOR THE PERSISTENCE OF PRIMING EFFECTS ........................................................................114 4.4.1 Experiment #2: Climate Treaty and Presidential Primes...........................................................115 4.4.2 Findings..........................................................................................................................................117 4.4.3 Source Familiarity.........................................................................................................................119 4.5 DISCUSSION AND CONCLUSION .............................................................................................................124 CHAPTER 5  WHEN DOES ANTI-AMERICANISM MATTER? ....................................................127  5.1 INTRODUCTION .......................................................................................................................................127 5.2 ANTI-AMERICANISM AND IR THEORY ..................................................................................................130 5.2.1 What is Anti-Americanism? ..........................................................................................................131 5.2.2 Neo-realism....................................................................................................................................132 5.2.3 Structural Constructivism .............................................................................................................134 5.2.4 Domestic-Politics Approaches and Anti-Americanism ...............................................................136 5.3 EMPIRICAL TESTING – PARTICIPATION IN THE IRAQ WAR ..................................................................138 5.3.1 Participation in the Invasion and Post-War Stabilization..........................................................140 5.3.1.1 Independent Variables ............................................................................................................................141 5.3.1.1.1 Realist Controls ...............................................................................................................................143 5.3.1.1.2 Constructivist Controls ...................................................................................................................144 5.3.1.1.3 Elite Preferences..............................................................................................................................145 5.3.1.2 Data Analysis ...........................................................................................................................................145 5.3.1.2.1 How Much Did It Matter? ..............................................................................................................148 5.3.1.3 Testing for Contextual effects.................................................................................................................152 5.3.1.4 Discussion.................................................................................................................................................153  5.3.2 The Influence of Anti-Americanism on Withdrawal from Iraq...................................................154 5.3.2.1 Data and Variables...................................................................................................................................155  5.4 CONCLUSION ..........................................................................................................................................160 CHAPTER 6  CONCLUSION ..................................................................................................................163  6.1 TRANSNATIONAL SOURCE CUE EFFECTS .................................................................................................163 6.1.1 Source Familiarity.........................................................................................................................164 6.1.2 Communications Environment...................................................................................................166 6.2 ACTOR I MAGE AND U.S. SOFT POWER ...............................................................................................169 6.2.1 U.S. Country Image – Policy Consequences ............................................................................169 6.2.2 Presidential Image and U.S. Soft Power...................................................................................170 6.2.2.1 6.2.2.2  6.3  Presidential Popularity and Anti-Americanism ..................................................................................170 Presidential Popularity and Persuasion ...............................................................................................172  FUTURE RESEARCH .............................................................................................................................174  BIBLIOGRAPHY ............................................................................................................................................176 APPENDICES ..................................................................................................................................................184 APPENDIX A: QUESTION WORDING FOR SURVEY EXPERIMENTS ...............................................................184 APPENDIX B: MEASURES OF ANTI-A MERICAN ATTITUDES .........................................................................189 APPENDIX C: PROCEDURE FOR ESTIMATING FAVORABILITY SCORES FROM EUROBAROMETER POSITIVITY SCORES ...........................................................................................................................................................190 APPENDIX D: WITHDRAWAL DATES AND SOURCES FOR O PERATION IN I RAQ ...........................................192 APPENDIX E: BEHAVIOURAL RESEARCH ETHICS BOARD CERTIFICATES ...................................................193  vi  LIST OF TABLES Table 2.1 Presidential Favorability in MEL and Climate Treaty (Presidents) Samples 41 Table 2.2 Opposition to Mandatory Emissions Limits (by Presidential Cue) .............. 43 Table 2.3 Climate Treaty (Presidents) - Presidential Cues and Support for a Climate Treaty without Chinese and Indian Participation......................................... 44 Table 2.4 Effects of Source Familiarity on Obama and US Source Cue Effects .......... 49 Table 2.5 Effect of U.S. Cues by Policy Prime Condition – Climate Treaty (U.S.) ..... 53 Table 3.1 Opposition to Mandatory Emissions Limits (by Presidential Cue) .............. 72 Table 3.2 Presidential Favorability in MEL Experimental Sample.............................. 73 Table 3.3 Perceived Knowledgeability and Trust in Three Presidents......................... 75 Table 3.4 Presidential Credibility for MEL Experiment.............................................. 75 Table 3.5 Effect of Presidential Cues on Opposition to Mandatory Emissions Limits by Subject Information Levels......................................................................... 79 Table 3.6 Effect of Obama Cue in Afghanistan Experiment by Gender ...................... 81 Table 3.7 Effect of Obama Cue on Support for Extending Afghan Mission, by gender83 Table 3.8 Moderating Effect of Credibility on Afghan Support .................................. 84 Table 3.9 Effect of Presidential Cues on MEL Attitudes, by gender ........................... 86 Table 3.10 Perceptions of ISAF by U.S. Leadership Condition..................................... 88 Table 3.11 Moderating Effect of U.S. Leadership Cue in Afghanistan Experiment ....... 89 Table 4.1 Presidential Priming and U.S. Favorability in the MEL Experiment (Control Groups Comparison) ................................................................................ 106 Table 4.2 Presidential Priming and U.S. Favorability in the MEL Experiment (ObamaBush Comparison).................................................................................... 108 Table 4.3 Effect of Presidential Priming on U.S. Attitudes and Favorability in Two Experiments ............................................................................................. 117 Table 4.4 Presidential Familiarity as a Moderator of Presidential Priming in Two Experiments ............................................................................................. 122 Table 5.1 Countries Contributing to Operation Iraqi Freedom in 2003 ..................... 141 Table 5.2 Countries and Pre-war Aggregate U.S. Favorability.................................. 142 Table 5.3 Effect of U.S. Favorability on 2003 Participation in the Iraq MNF............ 146 Table 5.4 Effects of a Reduction in Aggregate U.S. Favorability on the Predicted Probability of Participation....................................................................... 149 Table 5.5 Full 2003 Model with Realist Variable Interactions .................................. 151 Table 5.6 Effect of Favorability on Decisions to Withdraw from Iraq....................... 156 Table 5.7 Effect of Reduction in Aggregate U.S. Favorability on Probabilities of Withdrawal (by casualty status) ................................................................ 160 Table C1 Predicting Country-Year Favorability....................................................... 190 Table C2 Predicted Favorability Scores .................................................................... 191 Table D1 Withdrawal Dates and Sources for Participants in the Iraq Multi-National Force (MNF) ............................................................................................ 192  vii  LIST OF FIGURES Figure 2.1 Figure 2.2 Figure 2.3 Figure 2.4 Figure 3.1 Figure 4.1 Figure 4.2 Figure 4.3 Figure 4.4 Figure 4.5 Figure 5.1 Figure 6.1  Presidential Source Cue Effects....................................................................45 Change in Probability of Strongly Supporting MELs in Canada................. 50 Change in Probability of Strong Opposition to Canadian Position.............. 55 Change in Probability of Strong Support to Canadian Position................... 56 Predicted Effects of Obama Cue ................................................................ 90 U.S. Favorability by Priming Condition................................................... 103 Estimated Priming Effects for President Obama ...................................... 108 Predicted Effects of Presidential Favorability, by Priming Condition....... 111 Obama Priming Effects, by Presidential Familiarity................................. 114 Media Salience of Presidents Bush and Obama........................................ 120 Hypotheses and Associated Empirical Tests ............................................ 138 The U.S. Global Image, 2002-2009 ......................................................... 171  viii  ACKNOWLEDGEMENTS I would like to thank, first and foremost, my supervisor, Richard Price, and the members of my committee, Fred Cutler, Alan Jacobs, and Paul Quirk, for their guidance and support from the beginning to the end of this dissertation project. I would also like to thank the following individuals for offering detailed comments on earlier drafts of the papers in this dissertation: Shane Barter, Clare McGovern, Ben Nyblade, Angel O’Mahoney, Lisa Sundstrom, and various conference discussants and anonymous journal reviewers. Finally, I’d like to thank my parents and Sophia for their tireless patience, support, and encouragement.  ix  DEDICATION  To my parents, and to Sophia.  x  Chapter 1  Introduction  An increasingly salient concept in constructivist theories of International Relations (IR) is the idea of “socialization,” which generally refers to the process by which states and other actors come to accept and abide by international norms originating somewhere else in the international system. This literature identifies two principal ideational mechanisms through which actors are socialized: social influence and persuasion (Johnston 2001).1 The former refers to practices like “backpatting” and “naming and shaming” (Hafner-Burton 2008) that involve the manipulation of a targeted actor’s sociological costs and benefits, and tend to result in the instrumental acceptance of norms rather than in their true “internalization” (Finnemore and Sikkink 1998). In contrast, persuasion tends to result in internalization, where "the gap or distance between actors' basic causal and affective understandings closes as a result of successful persuasion” (Johnston 2001, 499). I focus on persuasion in this dissertation precisely because it is the most transformative mechanism of socialization in that it changes actors’ beliefs and attitudes and not just their behavior.2 Moreover, in practice persuasion often underpins the other mechanisms of socialization. Assuming that a new idea originates with a specific actor or a small group of actors, for example, but that social influence can only operate effectively once a sufficient number of actors in a given system have adopted that idea (Finnemore and Sikkink’s (1998) “tipping point”), it follows that mechanisms other than social influence will dominate in the early stages of a norm's life  1  The manipulation of material incentives is, of course, also a means of inducing norm-compliant behavior. That said, some of the literature on socialization does make a point of arguing that the effect of ideational influences (such as persuasion) on actor behavior is strongly dependent on the political costs and benefits of norm compliance faced by decision-makers (Schimmelfennig 2005). 2  1  cycle (Finnemore and Sikkink 1998). In other words, persuasion and material influence must be deployed in order to gather the “critical mass” of supporters necessary to successfully use social influence. Persuasion is therefore likely to be particularly salient when material resources are scarce, such as when the early supporters of an advocacy campaign include only civil society groups and small countries. Moreover, a deployment of material power that relies on boycotts and mass political mobilization (Klotz 1995; Wapner 1995, for example) may itself rely on persuasion if entrepreneurs use persuasive communications to change the attitudes of the members of the mass public that constitute politicians' political constituencies and firms' consumers. In sum, persuasion is both a direct and indirect tool available to transnational norm and policy entrepreneurs, and is one of the key causal mechanisms at all stages of state socialization. An important question, then, is: under what conditions are transnational persuasive communications successful in changing public attitudes about a given policy? While some of the constructivist literature on persuasion tends to view the phenomenon from a Habermasian perspective – driven by a “logic of truth seeking” or “argumentative rationality” (Risse 2000) –, much of this work adopts a model of persuasion akin to that in social psychology (Eagley and Chaiken 1993), where advocates attempt to change their targets’ attitudes through the instrumental use of rhetorical arguments or frames (Payne 2001; Price 1998; Keck and Sikkink 1998; Krebs and Jackson 2007; and Price 2003, for example). In this context, attitude change does not depend on a shared commitment to discover the “truth,” but rather on the extent to which strategically deployed frames (Payne 2001) ‘resonate’ with the underlying values of the targeted  2  population (Sundstrom 2005).3 This focus on framing, however, ignores a second mechanism of persuasion – the source cue effect – that is just as relevant to global political communication. In his discussion of persuasion, Johnston (2001, 496-498) notes three types of situational attributes that influence the probability of successful persuasion: 1) contextual cues that link the message to the target's broader pre-existing values and schemas (framing), 2) the target's relationship to the persuader (the source cue effect), and 3) the cognitive characteristics of the target. The first – framing – has been extensively investigated in the international relations literature and in political psychology (Druckman 2001b; Chong and Druckman 2007b), while the last is more relevant to political psychologists than to those who hope to understand the dynamics of aggregate public attitudes as a potential cause of policy change. The second mechanism, however, is clearly relevant to international relations (for example, in the context of changes in global anti-Americanism), yet, with a few exceptions (Johnston 2001; Busby 2010) it has been ignored in the global politics literature on persuasion. My dissertation attempts to address this gap in the literature on both a theoretical and a substantive level. As a theoretical contribution, I draw on the literature on source cue effects in political psychology – which have typically been studied in the context of domestic political debates – to develop and test theories of how actors’ global images influence their transnational persuasiveness. I identify a number of moderators of the source cue effect that are particularly salient in a transnational context. Substantively, I focus on the United States’ global image and on the ability of the U.S. President to wield soft power in  3  These concepts also feature prominently in the social movements literature (Benford and Snow 2000; Smith 2002).  3  the global arena. I examine the relative persuasiveness of different U.S. presidents and the United States as a whole (Chapter 2), the limits to President Obama’s personal transnational persuasiveness (Chapter 3), the way in which President Obama’s personal popularity has improved the United States’ global standing since his inauguration (Chapter 4), and the effects of public anti-Americanism on countries’ decisions to participate in the U.S.-led invasion of Iraq in 2003 (Chapter 5). In this introductory chapter, however, I review two bodies of work that are particularly relevant to my substantive and theoretical objectives: the literature on antiAmericanism and soft power and the theoretical literature on source cue effects in political psychology. I then briefly describe the design of the survey experiments that provide the bulk of the empirical evidence presented in this dissertation. I conclude with an argument for why the study of public attitudes is relevant to substantive policy outcomes in international relations. 1.1 ANTI-AMERICANISM AND U.S. SOFT POWER The rapid decline in the standing of the United States among foreign publics during the past decade prompted a resurgence of the academic debate on antiAmericanism, which is ultimately a debate about both the factors that shape the U.S. global image (Nossal 2005; Ray and Johnston 2007; Katzenstein and Keohane 2007b; Nye 2004, 35-43; Nisbet et al. 2004; Carlson and Nelson 2008; Willnat et al. 2006; Graber 2009) and the effect that this image has on policy outcomes (Katzenstein and Keohane 2007a).4 This work was prompted, at least in part, by a concern that the decline  4  Note that I refer to actors’ global “image”, “standing,” and “popularity” interchangeably in this dissertation. The terms simply refer to the general evaluations of an actor by global publics, which are usually measured in cross-national surveys using items that ask respondents to report whether their general opinion of the object in question is “favorable” or “unfavorable” (Pew Global Attitudes Project) or  4  in the United States’ global standing would undermine its soft power (Nye 2004), thereby reducing its ability to achieve its foreign policy objectives. Goldsmith and Horiuchi (2009), for example, find that foreign visits by the U.S. President and Secretary of State (what they call “high-level public diplomacy”) can in fact be counterproductive and damage the U.S.’s global image when the U.S. government suffers from low credibility. Moreover, while initial work suggested the need for skepticism about the potential policy consequences of anti-Americanism (Katzenstein and Keohane 2007a; Jhee 2008), recent work has shown that aggregate public attitudes towards the United States can influence other states’ willingness to cooperate with the U.S. (Datta 2009, 2008; Goldsmith and Horiuchi n.d.), a finding that my own original research in Chapter 5 speaks to. Now, while the U.S.’s global image has improved dramatically following Barack Obama’s election as U.S. President, (Kohut et al. 2009), a number of questions remain about how the country’s corporate image interacts with the President’s global popularity and how it influences American soft power. First: how does the U.S. global image and the U.S. President’s global popularity translate into soft power? I explore this question by focusing on one component of soft power – the ability of actors to persuade foreign publics to support their preferred policy alternatives. Specifically, in Chapters 2 and 3, I consider the conditions under which the U.S. President is likely to be able to move global opinion on climate policy and on the war in Afghanistan, and whether the President possesses a personal soft power that is separate from that of the United States as a corporate entity. Second: how does the U.S. President’s personal image shape foreign publics’ general evaluations of the United States? I explore one mechanism for this  “positive” or “negative” (Program for International Policy Attitudes). For the purposes of this study, I operationalize “image” using a standard favorability item.  5  effect in Chapter 4, using the value-expectancy model of attitudes to outline what I call the presidential priming effect. Finally: Does the improvement in the U.S.’s country image caused by a popular president (such as President Obama) actually matter? Chapter 5 tests for the substantive effect of aggregate public opinion about the United States on countries’ willingness to participate in the U.S.-led invasion of Iraq and on their decisions to withdraw from the operation. This dissertation’s primary contribution, then, is to better our understanding of the ways in which the United States and U.S. Presidents can exert soft power. While many of the findings (such as the finding that Obama is more persuasive than Bush) are intuitive, some findings (such as the moderating effect of source familiarity on Presidential persuasiveness and the gender gap in responses to an Obama cue) are surprising, and point to the need for further study of the mechanisms of transnational persuasion. In the next section, I summarize the theoretical starting points for the substantive chapters in this dissertation. 1.2 ACTOR EVALUATIONS, SOURCE CUE EFFECTS, AND PERSUASION IN GLOBAL POLITICS The constructivist tradition in IR, with its focus on Social Identity Theory (SIT) (Wendt 1999; Flockhart 2005; Curley 2009), is a good starting point for thinking about how actor evaluations can influence transnational persuasiveness. A general expectation arising from this perspective is that the probability of a norm or policy being accepted by an actor increases if that norm or policy is associated with a group that the actor in question identifies with and values positively. Johnston (2001), for example, includes this SIT dynamic under his “social influence” mechanism. In as much as specific  6  applications of social influence, such as “naming and shaming” (Hafner-Burton 2008), are agentic phenomena, however, it follows that it is only by means of communication that actors become aware of what courses of action are supported by positively valued actors (which then sets the social expectation). Much of this communication, moreover, is likely to be persuasive in that it aims to change policy attitudes by highlighting who supports a given action.5 In short, persuasive communication can use social influence as a tool, even if social influence is not always a result of active persuasion. As used by constructivists, however, SIT is primarily a sociological rather than a psychological theory,6 and does not offer substantial guidance about when or under what conditions evaluations of a message source are likely to have a significant effect on the success of a persuasive communication. Instead, constructivist scholars have used SIT’s insights about the salience of in-group/out-group classifications to help explain the change of group identities in the international system (Flockhart 2005; Curley 2009). The literature on source cues in political psychology and political behavior, however, provides a wide array of well-specified theories to explain the influence of actor evaluations on persuasion and suggests a number of specific variables that should moderate the impact of a message source’s image on persuasion in a transnational context. I outline the relevant literature below, and lay out the broad expectations that I test in the substantive chapters.  5  Indeed, in Pelika’s (2007) “symbolic politics” framework, actors only become “opinion referents” when they are explicitly associated with a specific policy position in the public discourse. 6 Shannon (2000) and Curley (2009) do address the specific psychological motives underlying such behavior. Shannon (2000, 298), for example, argues that “[r]esearch on decision-making shortcuts and self-presentation offers insights into how agents deal with the psychological needs underlying constructivist logic: 1) organize and simplify reality for the purpose of effective action; 2) build and maintain social approval from one's peers, and 3) maintain and enhance one's self-image and esteem.” Busby’s (2010) and Johnston’s (2001) papers are, to my knowledge, the only work in the IR literature to directly address the moderating effect of actor image on persuasion.  7  1.2.1 Source Cues  The literature on source cues in political psychology provides us with a basis from which to integrate the concept into IR. Using a variety of methods – most prominent among which is the survey experiment (Gaines et al. 2007) –, scholars have documented source cue effects in a wide variety of settings. A source cue effect occurs when the characteristics of the source of a persuasive communication have an independent effect on the persuasiveness of her message that goes above and beyond the content of the argument. Source cue effects have been documented for U.S. Presidents (Sigelman and Sigelman 1981; Mondak 1993; Bailey et al. 2003; Mondak et al. 2004), partisan elites (Goren et al. 2009), countries (Ashmore et al. 1979), religious elites (Robinson and Goren 1997) and talk show hosts (Lupia 2000), among others. Explanations for source cue effects in political psychology usually rely on the proposition that individuals use environmental cues as a way to make decisions about political issues while expending the least amount of cognitive effort – that is, as “heuristics” or judgmental shortcuts (Sniderman et al. 1991). The key source attributes that moderate this effect include the “feeling” (positive or negative) towards the source (Mondak 1993; Pelika 2007) and the perceived “credibility” or expertise of the source (Lupia and McCubbins 1998; Lupia 2000; Druckman 2001a; Chong and Druckman 2007a) among the targeted population. One way of conceptualizing the impact of the U.S.’s country image and the U.S. President’s global image on American soft power, then, is to argue that a negative change in these images would change the affect towards these two actors as well as their perceived credibility, thereby reducing their persuasiveness. I investigate this dynamic in Chapters 2 and 3.  8  A separate literature, on framing, conceives of persuasion as resulting from the manipulation of the salience of specific considerations that individuals take into account when attempting to come to a reasoned position on an issue (Nelson et al. 1997; Chong and Druckman 2007b). This literature is grounded in the expectancy-value model of attitudes (Nelson et al. 1997, 225), and while I do not investigate framing effects (which are usually associated with changes in policy attitudes) in this dissertation, I do use this framework in Chapter 4 to investigate how the image of national leaders (in this case, the U.S. President) influences foreign publics’ general evaluations of the country in question (the United States). Specifically, I examine what I call the presidential priming effect, where a short-term increase in the salience of the U.S. President increases the weighting placed on individuals’ attitudes towards him when generating their overall opinion of the United States.7 I engage empirically with the literature described above in two ways. First, I test the extent to which these theoretical frameworks – which were developed almost exclusively in the context of domestic U.S. politics – apply to transnational communications. Second, I explore a dynamic that is particularly relevant in a transnational context, but remains largely unexplored in the existing literature. Namely, following from McGraw and Dolan’s (2007) finding that individuals tend to “personify” a state in the person of the state leader, we might wonder to what extent individuals respond to a persuasive communication from a U.S. President simply as a communication from the United States rather than from a specific individual.8 This is important because  7  This is the same process as in Nelson et al.’s (1997) model of framing effects, excepting the fact that the evaluated object is an actor rather than a policy attitude. 8 An analog in the domestic setting, of course, is to what extent a communication from the President is a presidential cue and to what extent it is a party cue.  9  individuals’ attitudes towards U.S. Presidents are often much more negative or much more positive than their general attitudes towards the United States (see Chapters 2 and 3 or (Kohut et al. 2008) and (Kohut et al. 2009); Chapter 4 discusses how the former can shape the latter). Whether an individual responds to the U.S. President as an individual actor or as the personification of the United States, then, should determine which set of attitudes moderates the presidential source cue effect, which could ultimately make the difference between a successful and a failed persuasive communication. In Chapters 2 and 3, I investigate two conditions that shape this dynamic. I begin by recognizing that levels of information about the U.S. President and about U.S. politics will vary dramatically across global publics (much more so than among the U.S. public). This is important to keep in mind because one implication of the heuristic model of source cue effects is that source cues will only influence attitudes when individuals have sufficient information to form a judgment about the source. While people may in general be able to use easily observable attributes of a message source – such as physical attractiveness (Pornpitakpan 2004, 248) – to evaluate its credibility or likeability, in a political context, an individual’s ability to form an opinion about the source and to use that opinion in evaluating the message will likely be at least in part a function of how familiar that individual is with the source. This is particularly true in a transnational context, where culturally-dependent similarity cues (such as the ability to infer the source’s gender from their name) or expertise cues (such as title or educational background) may not be as readily meaningful to individuals as they would be in a domestic setting. My general expectation, then, is that presidential source cues will have the strongest effect among individuals that have high levels of information about the  10  United States and about the U.S. political system in particular. Moreover, I expect that this moderating effect of source familiarity will be attenuated when the source in question is a country rather than an individual or a more specific corporate actor, since members of foreign publics should have a wider set of considerations available when asked to evaluate a country (particularly one as central to global politics as the United States) than when asked to evaluate a specific leader. A second consideration is the possibility that both the country leader and the country in question may be salient in a given communication (in fact, this is likely to be common in practice). To the extent that members of the targeted public have distinct opinions about the two actors, it is worth considering what type of effect the salience of the country has on the persuasiveness of its leader. On the one hand, it may be that when the country is also salient, it acts as a second source cue, thus activating general attitudes towards the country in question. To the extent that evaluations of the state and its leader diverge, then, this should reduce the leader’s individual persuasiveness. On the other hand (and this is a key argument in Chapter 3), I argue that when policy advocated by the U.S. President is explicitly tied to the United States, this depresses the U.S. President’s persuasiveness because his appeal is more likely to be perceived as self-interested than when he advocates for a policy that is not directly tied to the United States. I test this possibility in Chapter 3 in the context of the war in Afghanistan, which is a particularly relevant topic because media outlets across the world have a tendency to describe the NATO mission in the country as “U.S.-led” and the U.S. government has been engaged in a campaign to persuade its allies to maintain their force deployments in the longer term.  11  1.3 DOES GLOBAL OPINION MATTER? Before describing the empirical research design I use to test the theoretical expectations about transnational source cue effects that I summarize in the previous section, it is worth considering the question of why the study of public opinion matters in the context of global politics. I begin by exploring the issue of levels of analysis. A great deal of work in IR – including early constructivist work such as that inspired by the sociological institutionalist or “world polity” school of sociology (Finnemore 1996b) – focused on the state as the fundamental actor in the socialization dynamic. An example at the extreme end of the spectrum is Alexander Wendt’s (1999) “structural constructivism,” which quite explicitly anthropomorphizes the state. While it is reasonable to focus on the state at the level of outcomes (i.e., behavioral conformity usually requires policy change at the national level), it makes little sense to conceive of the state as the fundamental actor on which the mechanisms of socialization, such as persuasion, operate. For one, the “national interest” is rarely exogenously derived,9 and sociological mechanisms like social pressure and persuasion are, by definition, based on cognitive micro-processes that can only be reasonably theorized with respect to individuals and small groups of individuals (Beyers 2005, 900). Specifically, “[s]ocialization... refers to both individuals (that is, when and how they socialize) and groups (that is, the social aggregate's features and how interactions among individuals  9  The incorporation of domestic interests into IR theory has been a major development in the last twenty years, with examples ranging from Putnam’s (1988) two-level model to Moravcsik’s (1997) liberal theory to concepts from IPE like the Stolper-Samuelson Theorem, where national interests are a function of coalitions of domestic interests but where these interests themselves vary in response to changes in the international structure. These are examples of the “second image reversed” formulation proposed by Gourevitch (1978). On the other hand, for good examples of how the national interest can be influenced or even driven by the system even within a sociological paradigm, see (Finnemore 1996a, 1996b).  12  shape these aggregates” (Beyers 2005, 900). One can refer to these as the micro(individual) and meso- (group) levels of analysis. This insight has been increasingly accepted within the growing literature that aims to fully specify the mechanisms of socialization (Alderson 2001; Johnston 2001; and Checkel 2005, for example), and illustrated by a variety of “top-down” and “bottom-up” models of socialization that incorporate domestic actors and processes (see Checkel 1999, 87; Risse and Sikkink 1999 on the “Spiral” model; or Keck and Sikkink 1998 on the “Boomerang” model). Aggregate source cue effects among the domestic public are no exceptions, and in its capacity as a mechanism of socialization, it is clear that persuasion should be studied at both the elite and public levels. Now, while the nation-state or any political jurisdiction could in principle be conceptualized as a single “group” (Curley 2009, for example) with a single set of domestic values and identities that may or may not resonate, there is simply too much evidence for intra-jurisdictional cultural and ideological cleavages to accept this conceptualization in reference to most rhetorical frames or source cues. Flockhart (2005), for example, documents the substantively different responses to “Europeanization” norms among the Danish elite and mass public.10 Indeed, she specifically conceptualizes them as two distinct social groups with distinct group identities and associated norms. In terms of targeting groups for persuasive communications, then, we can follow Flockart (2005) in thinking of specific groups as belonging to one of two categories. On the one hand, groups may encompass national mass publics, or some subset thereof. This category can also include the “global” public,  10  This elite/public divide is an enduring theme in comparative politics and political behavior (Kull and Destler 1999; Sniderman et al. 1996)  13  where the target is to change public opinion at the regional or global level. On the other hand, targeted groups could be sets of political or policy elites, such as legislators, civil servants, or business leaders, which is what much of the existing literature on socialization tends to focus on (Busby 2007, for example). It is the former – the persuasion of foreign publics – than I deal with in this dissertation. Now, one might argue that the question of when global political actors can persuade foreign publics to support their preferred policies has received relatively little empirical attention in the IR literature11 because foreign public opinion simply does not matter to policy outcomes. Indeed, as I discuss in Chapter 5, Katzenstein and Keohane (2007a) make a strong argument against the influence of global anti-Americanism on policies related to the United States. If this is in fact the case, the study of public opinion would not be particularly interesting to students of global politics. There is, however, good reason to think that both public attitudes on specific policies and general attitudes towards global political actors can influence policy outcomes. To begin with, a significant body of evidence exists outside of the IR subfield in support of the contention that public opinion can influence policy under certain conditions. Wlezien and Soroka (2007) review this vast literature, distinguishing between studies that merely show that public opinion is consistent with changes in policy (consistency) and those that provide better evidence for a causal connection between opinion and policy (covariation and congruence). They also point to country attributes (electoral system, separation of powers) and issue attributes (issue salience) that can influence the degree to which policy is responsive to public opinion. This is consistent  11  With the notable exception of the work on public diplomacy and soft power (Nye 2004; Nye 2008; Snow 2007; Goldsmith and Horiuchi 2010).  14  with work by constructivist scholars, such as Risse-Kappen (1991), who argues that domestic political structures channel the way in which mass public opinion is represented by elites. Now, while the bulk of the evidence in the representation literature relates to domestic policy, this does not necessarily mean that it is irrelevant to the study of global politics. Indeed, many of the issues currently of interest to IR scholars and students of transnational advocacy involve changes in what is effectively domestic policy (abolition of the death penalty or domestic greenhouse gas emissions abatement, for example). Moreover, some of the literature does point to the role of public opinion in specifically foreign policies, like participation in the 2003 invasion of Iraq (Baum 2008a) or defense spending (Soroka 2003). In a related strand of research, a number of recent studies using cross-national quantitative evidence now suggest that country image can influence political outcomes. Datta (2008; 2009) and Goldsmith and Horiuchi (2010) both find that aggregate public attitudes towards the United States (i.e., public anti-Americanism) influence the extent to which countries support U.S. foreign policy initiatives.12 I pursue a similar line of inquiry in Chapter 5 of this dissertation, where I use a cross-national dataset to show that aggregate public attitudes towards the United States were associated with countries’ decisions to participate in the U.S.-led Coalition of the Willing in Iraq and with their decisions to withdraw their forces from Iraq. These findings are particularly relevant to this dissertation in light of the findings in Chapter 4 and in Goldsmith and Horiuchi (2009), which suggest that a key facet of presidential soft power is the ability of a popular U.S. President to improve the U.S.’s global image. 12  In related, though not directly relevant work, Krueger and Maleckova (2009) find that public attitudes in one country towards another country predict the number of terrorist acts committed in the second country by nationals of the first.  15  A complementary argument in favor of studying public opinion in IR is that the effect of source cues need not be limited to mass public attitudes. Much work in IR suggests that elites are persuadable (see (Busby 2007) on Bono and Jesse Helms, for example). Indeed, since elites are also people, there is no reason a priori to believe that source cues will not influence elites in the same ways that they influence members of the public. Of course, elites are probably more likely to have high levels of political sophistication, political information, or motivation - and may therefore be less likely to engage in the heuristic processing that underlies the source cue effect (Eagley and Chaiken 1993) –, but there is evidence that elites may not always be as informed as we would imagine. In a 2002 survey by the Chicago Council on Foreign Relations, for example, a sample of U.S. foreign policy elites was asked to estimate the proportion of the federal budget that was devoted to foreign aid. The mean response was 5%, which, while more accurate than the general public’s mean response of 31%, was still a large overestimation of the true value of around 1%.13 Moreover, the literature on heuristics in political psychology suggests that foreign policy (Jervis 1976) and congressional (Miller 2009) elites are susceptible to the use of judgmental short-cuts like the accessibility heuristic – another mechanism that has been investigated primarily among members of the general public. In short, while elites may be more resistant to persuasion, it appears likely that we can apply insights derived from the study of public attitudes to understanding elite-to-elite transnational persuasion, which is a core element of theories of state socialization in IR.  13  See Chicago Council (2002) for bibliographic information. Elites measured in this survey included members of Congress, administration officials, journalists, and religious leaders.  16  To sum up, while public opinion is only one of many variables that influence policy, it does matter to policy outcomes, and, consequently, studying how publics can be persuaded to change their policy attitudes and actor evaluations by global political actors is an important part of understanding the process of state socialization. As Alderson (2001, 418) puts it: “Attitude change on the part of judges, business leaders, politicians, students and members of the public is part of what we mean when we say that a state ‘internalizes’ norms arising elsewhere in the international system [italics added by the author]”. 1.4 RESEARCH DESIGN In this section, I briefly describe my methodological approach in investigating the dynamics of transnational source cue effects. The vast majority of the evidence presented in Chapters 2, 3, and 4 is drawn from a set of survey experiments conducted at the University of British Columbia (UBC) Political Opinion Lab in April-July and December of 2009. The participants were undergraduate students enrolled in political science classes whose instructors had agreed to offer extra credit in exchange for their participation (this was arranged through the UBC Political Science Subject Pool). All questionnaires and experiments were approved by UBC’s Behavioural Research Ethics Board (BREB), as demonstrated by the certificates included in Appendix E of this dissertation. The surveys were administered to participants on computer stations using Empirisoft’s Media Lab software, which automatically conducted all randomization. While detailed descriptions of question wording and experimental design are presented in the substantive chapters and in the appendices, I describe the experiments in general terms below.  17  I present data from four experiments in this dissertation: the Mandatory Emissions Limits Experiment (Chs. 2, 3, and 4), the Climate Treaty (U.S.) Experiment (Ch. 2), the Climate Treaty (Presidents) Experiment (Chs. 2, 3, and 4), and the Afghanistan Experiment (Ch. 3). The basic design of all these experiments involves varying the actor to which an argument supporting a specific policy position is attributed, with the argument remaining un-attributed in the control group. The source cue effect, then, is the difference in policy support observable between the treatment conditions and the control condition. 1.4.1 Advantages of the Experimental Design  A clear benefit of this type of survey experiment over an interpretive approach or the quantitative analysis of existing public opinion data is the clarity of the causal inferences that I am able to make about the impact of an actor’s image on their transnational persuasiveness. As discussed above, the substantive focus of my research is transnational persuasion in the context of U.S. soft power.14 Consequently, I test Presidents Obama, Bush, and Clinton as source cues in three out of four experiments. In the Climate Treaty (U.S.) experiment, I use the “United States” rather than a specific individual.15 I focus on climate policy (at both the domestic and international level) in three of the experiments, but also examine Obama’s persuasiveness in the context of the NATO operation in Afghanistan. Overall, this allows me to explore presidential source cue effects for a variety of specific actors and across a range of samples and policy  14  Note that I limit all my analyses to participants who were Canadian citizens or permanent residents, which excludes the possibility of biasing the results through the inclusion of American students. This ensures that what I am measuring is in fact transnational source cue effects. 15 I also use a U.S. source cue in the Mandatory Emissions Limits experiment, which allows me to directly compare the effect of presidential source cues with a general country cue.  18  contexts. Moreover, the use of both presidential and “U.S.” cues allows me to differentiate between the effects of a generic U.S. source cue and a specific presidential cue. The fact that these cues are substantially different lends strong support to my argument that U.S. Presidents can exert a personal soft power. Since it would be extremely difficult to make this distinction using observational data, this illustrates one of the key advantages of using an experimental design. Several of the experiments, moreover, include additional manipulations or features that allow me to test more nuanced hypotheses. The Climate Treaty (U.S.) and the Afghanistan experiments, for example, also manipulate the extent to which the United States’ Iraq policy and the U.S. leadership role in ISAF are salient to participants (a priming manipulation). Chapter 4 makes use of a separate feature of the Mandatory Emissions Limits experiment, where participants were asked to report their favorability towards the United States immediately following the experimental manipulation and policy question. This allows me to investigate to what extent priming a specific U.S. President (Bush vs. Obama, for example) shapes participants’ general attitudes towards the United States (in other words, the source attributions are treated as presidential primes). 1.4.2 Limitations of the Research Design  While these survey experiments provide an advantage over other methods in terms of causal inference, my experimental research design has at least three limitations: 1) the unrepresentativeness of the sample of participants, 2) the use of proper names in the experimental manipulations, and 3) the unrealistic setting in which lab experiments must be performed. All three of these issues call into question the external validity of the  19  experimental findings, and therefore point the need for an abundance of caution in generalizing from my U.S.-focused survey experiments both to other sorts of actors and to real-world settings in transnational politics. The first limitation derives from the fact that all the experiments presented in this dissertation were administered to domestic undergraduate students at the University of British Columbia. The worry is that this sample is unrepresentative of the population of interest in several ways. First, it is a student sample, and may therefore represent an extreme end of the general Canadian public in terms of political ideology, opinions of President Obama, or attitudes towards the use of force. Second, it is a Canadian sample, which is likely to have an atypical relationship to the United States and the U.S. government, given its unusual proximity to the United States in both geographical and cultural terms. While the limited representativeness of my experimental sample does substantially limit the generalizeability of the findings I present in this dissertation, this limitation is, for a number of reasons, not as severe as it might appear. First, student samples are in fact reasonably representative if the principal aim is to understand public responses to experimental stimuli rather than estimating the actual distribution of specific variables in the population of interest (Mintz et al. 2006; Gartner 2008, 100). Second, most of the key substantive findings in the survey research – the extreme divergence in popularity between Presidents Bush and Obama, for example – are consistent with the available cross-national survey data, as shown in Figure 6.1, for example. Moreover, while we should expect that a group of political science undergraduates would be more informed about global politics than the average citizen,  20  the effects of this difference should wash out. On the one hand, a higher level of political information might result in stronger source cue effects due to an increased familiarity with the sources in question (see Chapters 2 and 3 for an explanation of the moderating effect of familiarity with the message source on its source cue effect). On the other hand, a greater amount of policy-specific information should make students less likely to use the source cue heuristic in forming their policy attitudes, and should therefore lead us to expect weaker source cue effects in a student sample than in a sample of the general public. In short, it is unlikely that my student sample is dramatically more or less susceptible to the experimental effects discussed in this dissertation than a sample of the general public would be. Overall, then, while the issue of representativeness is a real limitation of my research design, it is not a fundamental problem since my focus is on studying the psychological underpinnings of transnational source cue and priming effects. Nevertheless, any generalization to other national and cultural contexts must remain tentative. One area where an abundance of caution is necessary, however, is in generalizing my results to the “global” public. Canadians (and particularly students in Vancouver) are likely to have a different relationship to the United States and U.S. Presidents than most other foreign publics, due to their cultural, linguistic, and geographical proximity. That said, some of the findings presented in this dissertation may actually facilitate generalization to different settings. The finding that presidential source cue effects are particularly strong among individuals who have a high level of information about U.S. politics, for example, may allow us to predict in what countries or sub-publics the U.S. President is likely to be most persuasive (assuming we have data on levels of U.S.  21  information for these publics, of course). Overall, then, these experiments should provide a reasonable empirical basis from which to theorize the scope and limits to transnational source cue effects and, therefore, transnational persuasion.16 A second limitation of the experimental design is the use of proper names in the experimental manipulations. Using the United States and its recent Presidents as the message sources does boost the substantive relevance of my findings, but makes it more difficult to be certain about what is driving the observed effects. For example, are participants responding to President Obama’s perceived knowledgeability and trustworthiness, or are they relying on their general affect towards him? I use a number of control variables in my analyses as a way to get at this question, but the ideal design would manipulate variables like knowledgeability, affect, or familiarity experimentally, which is very difficult to do when using real and well-known actors like the United States. Overall, this makes my conclusions more tentative than they might have been using a more controlled experimental design. That said, using substantively interesting sources allowed me to contribute to the literature on anti-Americanism and U.S. public diplomacy, which is a reasonable trade-off for the purposes of this dissertation. Finally, a third limitation to the generalizeability of my findings results from the unrealistic setting in which the experiments were administered and the potentially unrealistic content of the communications the participants encountered. Participants received cues and were asked to evaluate policies in the context of single computer-based 16  Note that Chapter 5’s focus is not on transnational persuasion per se, but rather on the effect of public attitudes on policy. To investigate this question appropriately, I carry out two cross-national quantitative analyses, where the dependent variables are participation in or withdrawal from the U.S. military operation in Iraq and the key independent variable is the aggregate level of favorability towards the United States in each country’s public (as measured by a set of existing multi-country public opinion surveys). Overall, I find that aggregate levels of favorability are associated with an increased probability of participating in the operation and a decreased probability of withdrawing from the operation.  22  experiments. This framework misses potentially important aspects of real-world political debates, including the opportunity to discuss issues with friends and family, the repetition of cues and communications over a period of time, and the possibility of becoming more informed on the issue. Moreover, real-world debates involve competitive communications by opposing political actors, an aspect which was also missing from my experimental design. Overall, then, while this research provides a valuable exploration of the transnational persuasiveness of U.S. presidents, any generalizations to other global political actors or to real-world political dynamics (such as specific instances of transnational advocacy) must remain tentative. 1.5 STRUCTURE OF THE DISSERTATION To conclude this introduction, I briefly summarize the logical progression of the substantive chapters of the dissertation. The first two chapters focus squarely on transnational source cue effects. In Chapter 2, I develop a theoretical framework through which to integrate the concept of source cue effects into IR theory, and propose some general expectations about when source image is likely to moderate transnational persuasion. In Chapter 3, I focus specifically on what variables influence the global persuasiveness of U.S. Presidents, and what factors limit the persuasiveness of popular presidents like Barack Obama. The last two substantive chapters switch their focus to the dynamics of country image. In Chapter 4, I investigate the mechanisms through which a popular U.S. President like Barack Obama is able to improve foreign publics’ general attitudes towards the United States. In Chapter 5, I demonstrate the policy effects of a change in general attitudes towards the United States (i.e., anti-Americanism), using two  23  cross-national datasets to test the extent to which aggregate anti-Americanism influenced state decisions to participate in the war in Iraq. In short, I show why the findings in Chapter 4 matter. I conclude the dissertation in Chapter 6, where I summarize the key findings presented in the substantive chapters, and consider both their implications and potential avenues for further research into transnational persuasion. As a final note, I should point out that this dissertation is written as a series of separate papers rather than a single scholarly manuscript. This has inevitably resulted in some repetition in the presentation of literature reviews and descriptions of the experiments. I have, however, aimed to minimize the amount of repetition by referring the reader to previous chapters as appropriate, and cutting out material whose repetition is not vital in the context of the dissertation. Some of this material will, of course, be restored when each chapter is submitted to journals for publication.  24  Chapter 2  Beyond Framing: Source Cues in IR  2.1 INTRODUCTION Over the last decade, constructivist work in International Relations (IR) has begun to focus on persuasion as a mechanism of state socialization and norm diffusion. As I note in Chapter 1, constructivist scholars have increasingly adopted a model of persuasion that involves the strategic use of persuasive messages by global political actors attempting to change their targets’ attitudes towards a specific policy or practice (Payne 2001; Price 1998; Keck and Sikkink 1998; Krebs and Jackson 2007; and Price 2003, for example). While framing (Chong and Druckman 2007b) has taken on a central role in constructivist IR theory through the concepts of “frame resonance” (Sundstrom 2005, 422) and “strategic framing” (Payne 2001), however, work on persuasion in IR has until now largely neglected a second mechanism of persuasion suggested by the social and political psychology literature – the source cue effect.17 While the importance of the identity of a message source to transnational persuasion is noted in some of the IR literature (Cass 2005; or Johnston 2001, for example), it has not yet been systematically theorized despite the fact that Social Identity Theory features so prominently in the discipline (Wendt 1999; Flockhart 2005; Curley 2009; Mercer 1995). This chapter makes the case for integrating source cue effects into IR theory, arguing that the concept provides a valuable new tool with which to study persuasion in global politics – a tool  17  Two exceptions are Johnston (2001) and Busby (2010), though the former is an exclusively theoretical treatment and the latter focuses on the influence of domestic sources on American attitudes towards transnational policies.  25  that allows us to elaborate hypotheses about the conditions under which transnational persuasion is likely to work and when it is likely to fail. This chapter has three parts. In Part 1, I trace the use of framing and related concepts in the constructivist literature in IR and in the literature on transnational advocacy, and introduce the concept of source cue effects as an alternative mechanism of persuasion. In Part 2, I develop the argument by generating three sets of hypotheses about when they are most likely to occur. Finally, in Part 3, I use data drawn from three original survey experiments administered to undergraduate students in Canada to illustrate these expectations empirically and to provide plausibility tests of the hypotheses. I conclude by pointing to several potential avenues for future research. 2.2 PERSUASION IN INTERNATIONAL RELATIONS Constructivist work in International Relations on state socialization has long relied on the concept of “resonance” – the way in which a norm’s prescriptions and an actor’s underlying values “match” – to help explain when actors are likely to be persuaded to support or adopt a norm or policy (see, for example, Cortell and Davis 1996, 2000; Legro 1997; Sundstrom 2005; Busby 2007; Checkel 1999).18 While resonance was originally a somewhat static concept, in recent years scholars have also begun to focus on the dynamic mechanisms that underpin the phenomenon. Chief among these is “strategic framing”, an active process through which entrepreneurs seek to maximize the resonance of the norms they advocate by linking them rhetorically and symbolically to widely-held values or beliefs (Payne 2001; for the same concept under other names, see Price 1998; Keck and Sikkink 1998; Krebs and Jackson 2007; Price 2003). The success of an attempt  18  Schimmelfennig’s (2005) concept of “liberal constellations” of political parties is also analogous to a cultural “match”.  26  at strategic framing, then, can be explained in reference to the concept of “frame resonance” (Sundstrom 2005, 422; Benford and Snow 2000, 619-22) – that is, when a rhetorical frame deployed by a political actor taps into a prominent set of underlying values or beliefs. For example, according to Busby (2007), the Jubilee 2000 campaign on debt relief succeeded in no small part because it was able to use religious framing to persuade Senator Jesse Helms (an important gatekeeper) to support its cause. The work on frame resonance can be better understood by breaking the phenomenon down into two components: the individual-level causal mechanism and the macro-level process. “Frame resonance” can be conceptualized as the change in policy attitudes among the public or among key policy-makers that results from an effort at strategic framing by norm entrepreneurs. This change in opinion in turn increases the probability that the advocated norm’s policy prescriptions will be adopted by decisionmakers, which is the macro-level outcome. This macro-level process, however, depends on an individual-level psychological mechanism – the framing effect19 - that has received increasing attention in political psychology and political communication over the past two decades. The work on message-based persuasion, however, points to another microlevel causal mechanism – the source cue effect – that also has the potential to influence socialization outcomes. As described in Chapter 1, a source cue effect occurs when the identity of the source of a persuasive communication has an independent effect on the persuasiveness of her message that goes above and beyond the content of the argument (that is, the frames used or the information presented in that argument). While the presumption that the identity of an actor can influence her transnational persuasiveness is intuitive, work in 19  See Chong and Druckman (2007b) for a comprehensive review.  27  constructivist IR has largely ignored this question. Some work in IR does provide empirical examples of source cue effects – Cass (2005), for example, describes how the lack of U.S. credibility on emissions trading influenced other actors’ positions on the issue – but does not develop the concept theoretically. Busby (2010) considers what he calls “messenger effects,” but focuses on the domestic politics of U.S. foreign policy rather than on direct transnational persuasion. Moreover, while Johnston (2001, 497) does point to an audience’s “affective relationship to the persuader” as one of the three avenues to persuasion, this insight has not been followed-up by IR scholars. Social Identity Theory (SIT), for its part (see Wendt 1999; Flockhart 2005; Curley 2009; Mercer 1995 for applications to IR), clearly suggests that whether a norm is accepted by an actor partly depends on whether the norm is associated with a group that the actor in question positively identifies with. SIT, however, focuses primarily on “social influence” – a more passive mechanism of attitude change – rather than on the more agent-driven mechanism of persuasion, though there is no reason to suspect that this insight would not also hold for message-based persuasion in international relations.20 To summarize, while the potential influence of a message source’s characteristics on persuasion is discussed in the IR literature, the phenomenon has not been systematically theorized to the same extent that strategic framing has. I propose to begin that process in this chapter, and argue that incorporating the concept of source cue effects into constructivist theories of persuasion can lead to novel hypotheses about when transnational persuasion – and by extension, socialization – is likely to be successful.  20  Note that when norm advocates tap into countries’ desire to “belong to a normative community of nations” (Keck and Sikkink 1998, 29), they are in effect incorporating “social influence” into persuasive messages, or, in a different formulation, deploying opinion referents strategically (Pelika 2007).  28  2.2.1 Source Cues in Political Psychology  The literature on source cues in political psychology provides us with a platform from which to begin integrating the concept into IR. Most of this literature has studied source cue effects in a U.S. context, examining the persuasiveness of a wide variety of actors, including presidents (Sigelman and Sigelman 1981; Mondak 1993; Bailey et al. 2003; Mondak et al. 2004), partisan elites (Goren et al. 2009), countries (Ashmore et al. 1979), religious elites (Robinson and Goren 1997) talk show hosts (Lupia 2000) and celebrities such as Colin Powell and Jerry Springer (Druckman 2001b). These and other studies identify two broad types of source attributes that moderate source cue effects. On the one hand, general “feeling” (positive or negative) towards the actor in question (Mondak 1993; Pelika 2007) is thought to structure responses to a source cue, with individuals either using their feeling towards the source as a heuristic with which to evaluate the message (Mondak 1993) or transferring their affect towards the source to the advocated policy (Pelika 2007). On the other hand, the more “credible” or “authoritative” (Lupia and McCubbins 1998; Lupia 2000; Druckman 2001b; Chong and Druckman 2007a; Pelika 2007) a source is perceived to be by the target, the more likely the target is to accept the source’s message.21 I choose to focus on the moderating influence of general affect or feeling towards a source in this chapter primarily because 21  The way in which credibility is operationalized varies widely in the literature. In the “low-information rationality” approach to persuasion (Lupia and McCubbins 1998), credibility is hypothesized to be a function of 1) the perceived “knowledgeability” of the source and 2) the extent to which the source is perceived to share the target’s interests. The second variable is itself thought to be a function of trust in the source (Lupia 2000) and of “costly” cues, where an actor advocates a position that goes against her own interests (Lupia and McCubbins 1998; Howell and Kriner Unpublished). The idea behind this perspective is that using this type of heuristic allows individuals to approximate the policy attitude they would arrive at if they expended the effort to make a considered evaluation of the policy. Credibility can, of course, also be thought from a non-rationalist perspective, where sanctioned “experts” or “authorities” in specific fields will prove more persuasive than laypersons (Pelika 2007). This fits in well with the constructivist approach, which has, for example, highlighted the role of expertise in empowering epistemic communities (Haas 1992).  29  measures of general affect towards other countries and their leaders – such as favorability (Pew Global Attitudes Project), feeling thermometers (Transatlantic Trends Survey), and positivity (Project on International Policy Attitudes) – are available for a wide variety of countries. The availability of this data makes it easier for scholars of global politics to make use of a theory of transnational persuasion that is based on general affect rather than on a more nuanced operationalization of credibility. That said, the theory of transnational source cue effects developed in this chapter is by no means incompatible with alternative specifications of the source attributes relevant to persuasion. What most of the studies on source cue effects cited above fail to consider is the possibility of transnational persuasion (where a foreign actor influences domestic opinion about a country’s domestic or foreign policies), which is the type of persuasion that is most directly relevant to IR.22 On the one hand, there is no a priori reason to think that the theories of source cue effects described above do not apply to all types of persuasion. After all, the target actors are people in both cases and should presumably experience similar psychological responses to persuasive communications, regardless of the nature of the source. Moreover, while domestic sources are likely to be most salient to individuals, all things being equal, the literature suggests that in some cases foreign sources come to dominate the media coverage of a certain political position, and should therefore be salient to members of the domestic public. Archetti (2007), for example finds that while news coverage of the war in Afghanistan was dominated by domestic sources in the U.S. media, foreign sources dominated media coverage in Italy, France,  22  Though, of course, much socialization occurs when domestic actors become advocates for an international norm that they have themselves internalized. The key dynamic is still transnational, however, because the domestic actors in question must themselves be persuaded to support the norm and policy at some point.  30  and Pakistan. Interestingly, Hayes and Guardino (2010) find that U.S. media coverage of the debate in the run-up to the invasion of Iraq tended to contrast domestic supporters of the war with foreign opponents of the war such as Saddam Hussein, the United Nations, or French officials, while domestic opponents of the war were largely absent from this coverage. This suggests that transnational source cue effects are a politically relevant phenomenon. There are, however, a number of reasons to expect that transnational source cue effects may not be as prevalent as those exerted by domestic political actors. While I leave a systematic investigation of the unique attributes of transnational persuasion to future research, I note two of these possible reasons below. First, as I discuss in the introductory chapter and in the discussion below, foreign publics should be quite a bit less familiar with global political sources than with domestic political actors. To the extent that source familiarity moderates source cue effects (which is one of the major findings of this dissertation), we should therefore expect foreign sources to have weaker aggregate effects than an equivalent domestic actor would. Second, foreign actors are likely to suffer from lower legitimacy than domestic actors when engaging in a domestic political debate. They would, for example, be vulnerable to counter-arguments of “foreign interference”, and the relevance of a foreign actor’s opinion on domestic policies, in particular, is more questionable than for a domestic political actor. These two dynamics suggest that we should be somewhat skeptical about the likelihood of foreign sources having source cue effects of a magnitude comparable to the effects of domestic sources. One of the contributions of the research presented in this dissertation, then, is to show that transnational source cue effects do occur, and that they can be as strong as the  31  type of effects detected in a domestic context (at least for popular actors like President Obama). 2.3 UNDER WHAT CONDITIONS DO SOURCE CUES MATTER? The primary contribution that the literature on source cue effects can make to IR is to help students and practitioners of global politics to better understand the conditions under which transnational persuasive communications are likely to be effective. In the three sub-sections below, I describe three distinct factors that influence the direction and magnitude of transnational source cue effects. These factors by no means exhaust the topic, but nevertheless provide a useful first step in exploring the role of source cue effects in a transnational political context. Note, moreover, that to the extent that these empirical implications of the source cue model of persuasion are borne out empirically, it should strengthen our confidence that the overall experimental effects presented in this chapter are in fact source cue effects driven by participants’ attitudes towards the source.23 For each factor, I state a set of specific hypotheses, which I then test empirically in Part 4. 2.3.1 Factor #1: Source Popularity  An intuitive hypothesis about when transnational source cue effects are likely to occur is that, all things being equal, a persuasive communication delivered by a popular actor is likely to be more successful than a communication delivered by a less popular actor (the relative popularity hypothesis). Indeed, this is the sometimes implicit expectation that underlies the concern in the United States about increasing global antiAmericanism during the Bush years (Kohut et al. 2008) and the hope that President 23  This is particularly relevant in light of the limitations to experimental control caused by my use of actual (and well-known) actors as sources. I discuss this “proper name” problem further in Chapter 6.  32  Obama’s popularity will restore the United States’ global “soft power” (Kohut et al. 2009; Nye 2004). The converse hypothesis is one of the more interesting findings in the source cue literature – namely, that unpopular actors can have a negative source cue effect. Mondak (1993, 189), for example, argues that if feelings about the message source are negative, the target may decide “to oppose the source’s policy position”. This dynamic opens the door to the possibility that an attempt at transnational persuasion will not only fail, but in fact backfire – with opposition to the advocated policy position increasing due to the identity of the message source. For example, attributing a policy position to North Vietnam (Ashmore et al. 1979), to religious elites (Robinson and Goren 1997) or to President Carter (Sigelman and Sigelman 1981) resulted in decreased aggregate support for the attributed positions.24 Thus, if a very unpopular actor is associated with a policy position, we should expect support for the advocated position to drop because the actor’s identity has resonated negatively among the target population (the backlash hypothesis). While we should expect more popular sources to be more persuasive than less popular sources, there are, as noted above, good reasons to be skeptical about the possibility of even extremely popular actors moving aggregate opinion in a transnational context, so the relative popularity hypothesis is not obvious. First, opponents of a foreign actor’s message in a targeted state would likely be able to mobilize opposition to the proposed policy using charges of “foreign interference” in the country’s domestic politics, though both the nature of the appeal and the openness of the targeted public would presumably moderate the extent to which this dynamic emerges. Second, foreign  24  Goldsmith and Horiuchi (2009) also find that public diplomacy by U.S. leaders when their credibility is low can lead to a backlash effect on the U.S.’s global image.  33  leaders and foreign actors in general are unlikely to be particularly salient opinion referents for members of the mass public, particularly for issues that touch on domestic politics (such as climate policy). Low-profile foreign actors may also not be particularly salient opinion referents for individuals that know little about them, since they are unlikely to have strong opinions about those actors. Finally, publics may mistrust the motives of foreign advocates, believing they may be promoting policies that are designed to promote their own country’s interests to the detriment of the targeted publics’ interests. What these reasons for skepticism suggest is that the specific context of an attempt at persuasion is likely to be just as important in determining whether we should expect transnational source cue effects to exist in a specific instance as the broad popularity of the source. I describe two conditioning factors below. The first – the level of audience familiarity with the source – refers to an attribute of the target audiences, while the second – the salience of the source’s global policies – refers to both audience attributes and the media environment. 2.3.2 Factor #2: Audience Familiarity with the Source  As I describe in Chapter 1, most theories of source cue effects rely on the idea that individuals use their evaluations of the source of a persuasive message as a cognitive shortcut in evaluating the policy position advocated by that source because they have a limited supply of attention and prefer to minimize the amount of cognitive effort they expend when making policy judgments. One implication of this heuristic model of source cues is that we should expect source cue effects to be most apparent when individuals have low levels of policy-specific information (Mondak 1993; Bullock 2008). A complementary implication of the heuristic model, however, is that source cues will  34  only influence attitudes when individuals have sufficient information to form a judgment about the source. While people may be able to use easily observable attributes of a message source, such as physical attractiveness (Pornpitakpan 2004, 248), to evaluate its credibility or likeability, in a political context an individual’s ability to form an opinion about the source and to use that opinion in evaluating the message will likely be at least in part a function of how familiar that individual is with the source. This is particularly true in a transnational context, where culturally-dependent similarity cues (such as the ability to infer the source’s gender from their name) or expertise cues (such as title or educational background) may not be as readily meaningful to individuals as they would be in a domestic setting. In addition, we should expect levels of information about specific global political actors to vary substantially across regions, countries, and sociodemographic groups, all of which should influence the likelihood of foreign source cues having an effect among a specific target population. This leads us to expect that, all things being equal, the higher an audience’s level of information about a message source, the larger the magnitude of the source cue effect is likely to be (source familiarity hypothesis). A further implication of this hypothesis is that the moderating effect of familiarity with the source’s political system will be attenuated when the source in question is a country rather than an individual or more specific corporate actor (country image hypothesis). This follows from the observation that individuals are likely to have a wider range of considerations available in forming their opinions about other countries (in the case of the U.S., for example, Hollywood movies, American tourists, goods imported from the United States) than in forming their opinions about specific individuals or groups. Familiarity with the U.S. government and its policies, then, is likely to be a  35  weaker moderator of the United States’ transnational source cue effect than is familiarity with President Obama’s policies and political orientation as a moderator of the Obama source cue effect. 2.3.3 Factor #3: Media Environment and Policy Salience  Another factor that can moderate a global political actor’s transnational source cue effect is the media environment in a given jurisdiction. The literature on media effects in communications focuses on three types of effects – framing, agenda-setting, and priming (Scheufele and Tewksbury 2007). Priming, in particular, is relevant to understanding transnational source cue effects because it suggests a mechanism through which media coverage can influence the manner in which the target audience of a transnational persuasive message evaluates the source of that message, thereby shaping its source cue effect. At the individual level, priming operates by increasing the weighting of a specific consideration (positive or negative) on an individual’s evaluation of a specific political actor. For example, if coverage of the United States in a jurisdiction’s news media is dominated by reporting on the war in Iraq (an extremely unpopular U.S. policy in a global context), that jurisdiction’s public is likely to have more negative views of the United States than it would hold otherwise (because their opinion of the war, which is negative, will receive a greater weighting), and should therefore react more negatively to a U.S. source cue. While priming effects have been documented primarily in the context of evaluating candidates for political office, media priming is particularly relevant in a transnational context precisely because audiences are less likely to have a fixed opinion of typical global political sources (countries or international organizations, for example) than sources in a purely domestic context. This  36  is particularly likely if the source is a country, because evaluations of corporate actors that have not been personified tend to be formed through memory-based processes (McGraw and Dolan 2007),25 and priming is a mechanism that emerges from memorybased models of information processing (Scheufele and Tewksbury 2007, 11). The general expectation, then, is that source attributes primed by the news media will influence the basis on which individuals evaluate potential message sources, and will therefore influence the magnitude and direction of those actors’ source cue effects. The question of what attributes determine country images has been most extensively studied in reference to the United States and to the causes of antiAmericanism (Katzenstein and Keohane 2007b), and is particularly important given the soft power component of U.S. global power (Nye 2008; Nye 2004). I therefore focus on the United States as a source cue in the following discussion, while noting that the implications I outline below are more broadly applicable. A key finding of this literature is that U.S. country image (favorable or positive opinion about the United States) is in part a function of foreign publics’ agreement with U.S. foreign policy (Nossal 2005) and the media coverage of U.S. policies (Nisbet et al. 2004). U.S. policies towards the Soviet Union (Isernia 2007), Taiwan (Johnston and Stockmann 2007), and Israel (Lynch 2007), for example, have been found to cause anti-Americanism in Europe, China, and the Arab world, respectively. We should therefore expect the U.S.’s transnational persuasiveness to be reduced when media coverage of the United States focuses on highly unpopular policies like the war in Iraq. In this context, the media coverage would increase the weighting of those policies in individuals’ evaluations of the United States’, thereby  25  Whereas evaluations of people (or corporate entitites that have been personified) tend to result in on-line evaluations (McGraw and Dolan 2007).  37  reducing affect towards the United States and the positivity of its source cue effect. It follows, then, that if a policy associated with a message source is salient, agreement or disagreement with this policy will moderate responses to the persuasive communication, whereas when this policy is not salient, other evaluative dimensions will moderate responsiveness to the source cue (policy salience hypothesis). 2.4 RESEARCH DESIGN AND EMPIRICAL ANALYSIS I present data from three survey experiments that were administered to undergraduate students at the University of British Columbia at the UBC Political Opinion Laboratory. Full question wording is provided in Appendix A. Students received extra course credit in exchange for participation. In all three experiments, I focused substantively on climate change because, on the one hand, it presents a “hard case” for transnational persuasion in that climate policy has significant domestic policy implications, and on the other because it is a topical issue of vital global political importance. Note that since my focus is on transnational persuasion and I use U.S. cues in all the experiments, I exclude data from participants who were not Canadian citizens or permanent residents to avoid the possibility that the presence of American students skews the results.26 This yields an N of 168 for the spring 2009 (April-June) wave, which includes the Mandatory Emissions Limits and Climate Treaty (U.S.) Experiments, and an N of 158 for the December 2009 wave, which includes the Climate Treaty (Presidents) Experiment. Before continuing, I should note once again that, as I argue in the Introduction to this dissertation (Chapter 1), I present this data with the usual caveats regarding external  26  I was not able to include non-American international students because the item used did not allow me to distinguish between American and non-American temporary residents.  38  validity that are inherent to student samples, so any generalization of the substantive findings must remain tentative. Inasmuch as the results presented in this chapter deal with the psychological mechanisms underlying transnational persuasion, however, they should provide a reasonable basis for general expectations and future investigation. 2.4.1 Experiments #1 and 2: U.S. Presidents and Climate Policy  The Mandatory Emissions Limits (MEL) Experiment begins with an introduction that familiarizes participants with the concept of mandatory greenhouse gas emissions limits for industry as a way to prevent climate change and provides both supporting and opposing arguments for the policy. The experimental manipulation involves varying the political actor – in this case, U.S. Presidents – to which an argument opposing mandatory emissions limits is attributed. In the control group, the opposing argument is simply attributed to “others”, while for the treatment conditions, it is attributed to “former U.S. President George W. Bush,” “former U.S. President Bill Clinton,” and “U.S. President Barack Obama”. Participants are then asked whether they agree or disagree that “[t]he Canadian government should pass laws that limit the amount of carbon dioxide that Canadian industries are allowed to emit.” Opposition to mandatory emissions limits is measured on a standard 5-point scale (with “strongly agree” coded as 1 and “strongly disagree” coded as 5). Note that because the experiment attributes opposition to mandatory emissions limits to the actors in question, higher values of this policy variable indicate greater agreement with the cued position (i.e., opposition to limits), so a positive difference between mean support in a treatment group and the control group would indicate a positive source cue effect.  39  The Climate Treaty (Presidents) Experiment deals with the same substantive topic (climate policy), but from a foreign policy angle, focusing on the international negotiations on a new climate treaty rather than on domestic emissions policy. As in the MEL Experiment, participants read a set of arguments both for and against demanding that large developing countries like China and India commit to immediately begin reducing their emissions reductions. They are then asked whether they agree that “Canada should join a new international treaty on climate change EVEN IF China and India refuse to sign it.” Support for joining the treaty is measured using a standard 5point scale (“strongly disagree” (1) to “strongly agree” (5)). The experimental manipulation involves attributing both the pro- and con- arguments to President Obama and to former President Bush, yielding four conditions with a single cue in each, and a control group where the arguments are not attributed to any specific actors. Note that each participant that was assigned to a treatment condition only received one cue (Obama pro, Obama con, Bush pro, or Bush con). These two source cue experiments allow me to test the first set of hypotheses about when transnational source cue effects are most likely to occur – the relative popularity and backlash hypotheses. As a measure of affect towards the source, I also asked participants to indicate whether they had a very favorable, somewhat favorable, somewhat unfavorable, or very unfavorable opinion of President Barack Obama and former Presidents Bush and Clinton. This is the same measure used by the Pew Global Attitudes Project surveys, which has been running large-N multi-country surveys since  40  Table 2.1: Presidential Favorability in MEL and Climate Treaty (Presidents) Samples Mandatory Emissions Limits  Climate Treaty (Pres)  Obama  Clinton  Bush  Obama  Bush  Very Favorable  55.7%  27.3%  1.8%  51%  0%  Somewhat Favorable  38.9%  58.8%  6.6%  38.2%  1.9%  Somewhat Unfavorable  3.6%  11.5%  26.8%  7.0%  22.9%  Very Unfavorable  1.8%  2.4%  64.9%  3.0%  75.2%  27  2001.  N 167 165 168 157 157 Table 2.1 shows the relative popularity of the three presidents. Obama is by far  the most popular, with over 50% of participants in both waves reporting that they had a very favorable opinion, and another 38% reporting a somewhat favorable opinion. As one might expect in a sample of Canadian undergraduate students, Bush is extremely unpopular, with between 65% and 75% of participants reporting a very unfavorable opinion, and less than 10% reporting a favorable opinion. President Clinton lies in between, with overall favorability at a similar level to Obama’s (86%), but with most respondents reporting a somewhat, rather than a very, favorable opinion. While this difference in favorability between Presidents Bush and Obama is extreme, the gap is consistent with cross-national survey data. The Pew Global Attitudes Project, for example, finds that in only one of 25 countries (Israel) surveyed was confidence in President Obama in 2009 lower than confidence in President Bush in 2008, with a 60 percentage point difference in Canada (Kohut et al. 2009). In short, this suggests that the student sample used in this study is not unrepresentative of the broader Canadian public, at least with respect to its attitudes towards Presidents Obama and Bush. 27  Available at: http://pewglobal.org/.  41  Any observed Presidential source cue effects observed for these samples, then, should be tentatively generalizeable to the broader set of global publics. 2.4.1.1 PRESIDENTIAL SOURCE CUE EFFECTS As an initial estimate of the presidential source cue effects in the two experiments, Table 2.2 presents differences of means in policy support for each treatment condition in the MEL Experiment relative to the control group. Baseline support for the cued position is quite low, with mean opposition to mandatory emissions limits at 1.53 on a 1 to 5 scale. Mean opposition in each of the treatment conditions is higher than in the control group, though opposition in the Bush condition is not significantly different (t=3.64, p=0.23). The differences of means for both President Obama and President Clinton are statistically significant, however, with a 0.73 increase in mean opposition to emissions limits attributable to the Obama cue and a 0.57 increase attributable to the Clinton cue. These effects represent changes of 18% and 14% of the range of the dependent variable, respectively. These results, then, show that the two popular presidents (Obama and Clinton) both had positive source cue effects, while Bush did not. Note that the significant effect for Clinton suggests that the difference in the magnitude of the effects for the Bush and Obama cues is in fact driven by the difference in popularity, rather than simply resulting from the fact that Obama is the current president and Bush is not. I conducted the same type of analysis for the Climate Treaty (Presidents) Experiment. Differences of means are presented in Table 2.3. The results are consistent with the findings presented above. Individually, none of the comparisons between the cued conditions and the control group are statistically significant, but the effects for the Obama cue are larger than for the Bush cue – as the relative popularity hypothesis would  42  Table 2.2: Opposition to Mandatory Emissions Limits (by Presidential Cue) Obama  Clinton  Bush  Control  Mean opposition to MELs  2.26  2.10  1.75  1.53  Difference of means relative to Control Group  0.73  0.57  0.22  n/a  t statistic of the difference  3.64  3.32  1.22  n/a  p-value (two-tailed)  0.0006  0.0015  0.2268  n/a  N  27  30  32  34  predict - and the effect for Obama’s argument against signing the treaty verges on statistical significance (p=0.11). A different approach to estimating source cue effects involves comparing the change in policy support when each president switches positions, and indeed, one might argue that this is a more realistic counterfactual against which to compare the effect of an actor publicly taking a policy position. For President Bush, aggregate support when he argues against joining the treaty is about the same as when he argues for joining the treaty (primarily due to the negative effect of the Bush (Con) cue). For President Obama, however, mean support for joining the treaty is 4.59 when he argues for it (more or less a consensus in favor), but 3.68 when he argues against it. The difference of 0.91 points represents about 23% of the range of the 1 to 5 scale for the dependent variable, and is statistically significant (p=0.025).  43  Table 2.3: Climate Treaty (Presidents) - Presidential Cues and Support for a Climate Treaty without Chinese and Indian Participation Obama (Pro Cue)  Obama (Con Cue)  Bush (Pro Cue)  Bush (Con Cue)  Control  Mean Support  4.59  3.68  4.43  4.55  4.29  Mean (Cue) – Mean (Control)  0.30  -0.61  0.14  0.26  -  t-statistic of the difference  1.08  1.62  0.493  1.08  -  p-value (two-tailed)  0.29  0.11  0.62  .29  -  Mean (Pro Cue) – Mean (Con Cue)  0.91  -  -0.12  -  -  t-statistic of the difference  2.33  -  0.55  -  -  p-value (two-tailed)  .025  -  0.59  -  -  17  25  28  31  24  N  44  I illustrate the presidential source cue effects graphically for these two experiments in Figure 2.1. Overall, the findings are strongly supportive of the relative popularity hypothesis. Source cues for Obama and Clinton – both of whom are extremely popular in these samples – almost uniformly resonated positively, with statistically significant increases in aggregate support for the cued positions. This was true even for the Climate Treaty (Presidents) Experiment, where the treatment is quite subtle, requiring participants to process the arguments and tie them to a Canadian decision on joining the new climate treaty with or without Chinese and Indian participation. Moreover, the effect for the Clinton cue in the MEL Experiment was of a smaller magnitude than the effect for the Obama cue, which is consistent with Obama’s greater relative popularity. In contrast, none of the Bush cues had statistically significant effects. Given Bush’s extreme unpopularity, then, all the effects are consistent with the 45  relative popularity hypothesis – namely, the more popular actors’ identities resonated positively to a greater degree than the less popular actors. The results of the two experiments do not, however, support the backlash hypothesis. Despite Bush’s extreme unpopularity, the Bush cue only appears to have caused a negative effect in one of three conditions (the anti-treaty argument in the first treaty experiment), and the differences of means were not statistically significant for any of the treatments. Moreover, a negative source cue effect should have been most likely in the Bush (Pro) position because the baseline support for this policy position was already extremely high (see (Mondak et al. 2004) for the effect of baseline support on the magnitude of aggregate source cue effects), but the direction of the effect was in fact positive. Overall, this is a surprising finding, given that large positive effects were observed for the popular cues. It suggests the potential for a degree of asymmetry in the relationship between source popularity and transnational source cue effects, but an explanation for this anomalous finding is left to future research. 2.4.1.2 SOURCE FAMILIARITY AND TRANSNATIONAL SOURCE CUE EFFECTS I now move on to test the expectations that the audience’s familiarity with the source and the communications environment moderate the transnational source cue effects of relatively popular actors. I begin by using data from the Mandatory Emissions Limits (MEL) Experiment to test the first set of expectations. The source familiarity hypothesis suggests that source cue effects will be greater when the audience is familiar with the source, while the country image hypothesis predicts that the moderating effect of source familiarity will be less evident when the source in question is a country instead of an individual or an organization. To test these  46  hypotheses, I examine President Obama’s source cue effect in the MEL experiment, and take advantage of the fact that I also included a fifth experimental group where opposition to mandatory emissions limits was attributed to “the United States” (N=51). This allows me to compare the effect of source familiarity separately for a country and for its leader. As a measure of source familiarity, I used an open-ended item in which participants were asked to type the name of the current Vice President of the United States (Joe Biden). Fifty-five percent of respondents were able to correctly identify the VP while 45% were not. I use this measure of familiarity for both sources. In both cases, familiarity with U.S. domestic politics (as measured by the item) should correlate with increased familiarity with (and therefore a more fixed opinion about) both President Obama and the U.S. government, which is the relevant source when a policy is attributed to the “United States”. Table 2.4 presents coefficients from a set of ordered logit models where opposition to emissions limits is regressed on dummy variables indicating the Obama and U.S. cues (the baseline is the control group). In the basic model (Model 1), the Obama cue has a statistically significant positive source cue effect, while the U.S. cue falls just short of significance. When a number of attitudinal and demographic controls are included (Model 2), both cues have positive, statistically significant effects on opposition to MELs. Note that this model includes a variable (GHG Information) that measures participants’ level of policy-specific information (the number of greenhouse gases they were able to volunteer). This ensures that any moderating effects of source familiarity do not simply reflect a participant’s overall level of political information. The model also  47  includes a variable measuring how important the issue of climate change is to participants. In order to test the hypotheses, Model 3 adds interaction terms between source familiarity (US VP Info) and the source cue dummies. As predicted by the source familiarity hypothesis, the coefficients for both interaction terms are positive, indicating that higher levels of source familiarity increase the magnitude of the source cue effect. Chi-squared tests for the joint significance of each cue dummy and its interaction term reveal that the Obama interaction is statistically significant (Chi-squared = 8.09, p=0.018), while the US interaction is significant at a 10% level (Chi-squared = 5.49, p=0.064). Figure 2.2 presents more substantively meaningful statistics, showing the predicted change in the probability of strongly supporting mandatory emissions limits in Canada caused by each cue, with separate results reported for participants with high and low source familiarity.28 The U.S. cue reduces the probability of strongly supporting emissions limits by about .25 for participants with both high and low source familiarity, though neither of these effects is statistically significant at the 95% level.29 For the Obama cue, however, while the effect for participants with low source familiarity is of about the same magnitude (-0.26), the effect for participants with high source familiarity is 65% greater at -0.43, and is statistically significant (the 95% confidence interval falls below zero). In sum, the predicted source cue effects for high familiarity participants were larger than those for low familiarity, which is consistent with the source familiarity  28  I computed first differences for Pr(Y=“Strongly Support”) using the Clarify package for STATA (King et al. 2000), holding continuous variables in the model at their mean and setting the gender as female. I focus on this value of the dependent variable (1) because, due to the overall popularity of emissions limits, most participants fell into the strongly support category. 29 Note that since the cued position was opposition to mandatory emissions limits, this represents a positive source cue effect.  48  Table 2.4: Effects of Source Familiarity on Obama and US Source Cue Effects Model 1  Model 2  Model 3  1.762*** (3.32)  1.797** (2.80)  1.423 (1.48)  0.789 (1.78)  1.245* (2.35)  1.209 (1.45)  US VP Info  -  -0.603 (-1.25)  -0.779 (-0.95)  Importance of Climate Change  -  -0.944*** (-3.90)  -0.945*** (-3.88)  Gender  -  -0.117 (-0.24)  -0.108 (-0.22)  GHG Information  -  -0.106 (-0.61)  -0.103 (-0.59)  Income  -  0.0157 (0.22)  0.0142 (0.2)  Favorability Obama  -  0.0525 (0.16)  0.0254 (0.08)  Favorability USA  -  0.817** (2.66)  0.808** (2.59)  Obama Cue * US VP Info  -  -  0.63 (0.53)  US Cue * US VP Info  -  -  0.0588 (0.06)  106  95  95  Obama Cue US Cue  N  Note: Coefficients were estimated using ordered logistic regression. The dependent variable measures opposition to mandatory emissions limits in Canada. Cut-off constants are not reported. t statistics in parentheses * p<0.05, ** p<0.01, *** p<0.001  49  hypothesis, and the fact that the difference in effect magnitude is much greater for the Obama cue than for the U.S. cue is consistent with the country image hypothesis. 2.4.2 Experiment #3: US-Canadian Cooperation on Climate Negotiations  The third experiment – the Climate Treaty (U.S.) Experiment – is similar to the Climate Treaty (Presidents) Experiment in that it also asks participants to express an opinion on Canada’s negotiating position on the successor treaty to the Kyoto Protocol, but instead of using U.S. Presidents as source cues, it attributes opposition to exempting China and India from the treaty to the “United States”. It then asks participants whether they approve or disapprove of Canada’s refusal “[o]ver the past two years” to sign any new climate treaty without Chinese and Indian participation. In the U.S. Cue condition,  50  participants are informed that this was the position of “Canada and the United States”, whereas in the control condition the policy is described simply as Canada’s position. Full question wording is provided in Appendix A. The dependent variable measures support for the Canadian position on a five point scale (strongly disapprove to strongly approve). I also measured participants’ opinions towards the United States using the same favorability scale as in the other experiments. The Climate Treaty (U.S.) Experiment has an additional component designed to test the extent to which the salience of U.S. policy in Iraq moderates its source cue effect in the context of climate policy. The expectation, as expressed in the policy salience hypothesis, is that responses to the U.S. source cue will be moderated by participants’ attitudes towards the war in Iraq and the Bush Administration’s foreign policy to a greater extent when the Iraq War is salient to participants than when it is not. Participants assigned to the U.S. cue condition were also randomly assigned to one of two priming conditions. The first group (Iraq Prime condition) received two questions immediately prior to the climate treaty question about support for the U.S.-led war in Iraq, whereas the other group (Non-Primed condition) did not received these questions. In order to test this hypothesis, I also measured participants’ policy agreement with U.S. policies associated with the War on Terror, including the use of military force to prevent Iran’s acquisition of nuclear weapons, the secret detention and coercive interrogation of terrorist suspects, and the assassination of terrorist leaders. 30 These individual attitudes were then combined into an additive “Policy Agreement” index (scaled from 0 for “low agreement” to 1 for “high agreement”). Note that, given the design of the experiment, the 30  I use these items instead of simply using participants’ expressed support for the war because I unfortunately did not measure support for the war in Iraq among participants who did not receive the Iraq Prime.  51  intervening variable – affect towards the United States when responding to the experimental question – remains unobserved, but the priming effect can be inferred from the difference in the structure of responses to the U.S. cue. Table 2.5 presents coefficient estimates for several ordered logistic regressions where support for the Canadian position is regressed on dummy variables indicating both of the U.S. cue groups (the baseline comparison is the control group) as well as two control variables and interaction terms between the policy agreement index and the U.S. cue dummies. The first thing to note is that all of the models show negative coefficients for the two U.S. cues, indicating that the U.S. cue reduced support for the Canadian negotiating position relative to the control group. This is somewhat surprising, since attitudes towards the United States in this sample were also generally favorable (55% favorable, 45% unfavorable), but the effects are not positive, as for the U.S. cue in the Mandatory Emissions Limits (MEL) Experiment. While it is impossible to determine the reasons for this discrepancy with any certainty, one might plausibly attribute the difference to the question wording, which might be having other effects in addition to the source attribution.31 If this interpretation is accurate, however, we should expect that it would actually be more difficult to detect the predicted pattern of moderation by level of policy agreement, since the complex manipulation would introduce noise into the signal. If the pattern nevertheless emerges from the data, we should view this as additional  31  Namely, the nature of the manipulation in this experiment implies that Canada has aligned its policy with the United States in the context of the climate negotiations, which participants may be reacting to negatively. Alternatively, we might view the experimental manipulation as overlaying a U.S. cue on an existing Canadian source cue, which is likely more positively viewed than the United States. Adding the U.S. cue would therefore reduce the persuasiveness of the combined cues. Regardless, the policy salience hypothesis should still hold, since at least part of the overall experimental effect is likely to be attributable to a U.S. source cue effect.  52  Table 2.5: Effect of U.S. Cues by Policy Prime Condition – Climate Treaty (U.S.) Model 1  Model 2  Model 3  Model 4  US + Iraq Prime Dummy  -0.788 (-1.91)  -0.74 (-1.73)  -2.122** (-2.62)  -2.214** (-2.67)  US + No Prime Dummy  -1.128** (-2.85)  -1.104** (-2.75)  -0.643 (-0.89)  -0.783 (-1.07)  Policy Agreement Index  3.259*** (4.44)  3.010*** (4.00)  2.896* (2.39)  2.386 (1.89)  Gender  -  0.129 (0.37)  -  0.0816 (0.23)  Climate Change Importance  -  -0.371* (-2.19)  -  -0.378* (-2.21)  US/Iraq Prime * Policy Agreement  -  -  3.548 (1.91)  3.927* (2.06)  US/No Prime * Policy Agreement  -  -  -1.436 (-0.88)  -1.015 (-0.61)  127  127  127  127  N t statistics in parentheses * p<0.05, ** p<0.01, *** p<0.001  Joint Significance of Interactions Iraq Prime  No Prime  Chi-squared =  7.29  7.26  p-value =  0.026  0.027  Chi-squared =  9.50  8.65  p-value =  0.009  0.013  53  confirmation that the salience of an actor’s foreign policies does indeed moderate its source cue effect in a transnational setting. In order to test for this moderating effect, I estimate a minimal interaction model (Model 3), and an interaction model with additional controls (Model 4). In both cases, the same pattern emerges (indeed, the pattern is remarkably robust to alternative specifications). For the primed condition, the U.S. cue has a negative effect for individuals with low policy agreement, but as agreement increases, the effect becomes less negative. For the unprimed condition, the U.S. cue has a negative effect for those with low agreement, but this effect becomes even more negative for those with higher levels of policy agreement. Both of these interactions are statistically significant, as is shown by the joint significance of the interactions. To better illustrate the magnitude and direction of these effects, I use Clarify to estimate the predicted changes in the probability of strongly opposing and strongly supporting the Canadian negotiating position caused by the U.S. cues, in both the primed and unprimed conditions for Model 3. Figures 2.3 and 2.4 show that the patterns of effects in the primed and unprimed conditions are markedly different, which is consistent with the policy salience hypothesis. Priming U.S. policy, in other words, appears to change the way in which participants respond to the U.S. cue. In the primed condition, the U.S. cue increases the probability of strong opposition to the Canadian position and decreases the probability of strong support among participants who had a low level of agreement with U.S. foreign policies. For people with high levels of policy agreement, the cue appears to have little effect, but if anything, increases the probability of strongly supporting the Canadian position (Fig. 2.4). In the non-primed condition, however, the  54  U.S. cue appears to slightly increase opposition to the Canadian position for all levels of the policy agreement variable, and reduces strong support, with the effect strengthening as policy agreement increases. The key point is that in the primed condition, the effect of the U.S. cue is, as expected, positively moderated by the audience’s level of policy agreement, whereas in the unprimed condition, the U.S. source cue effect appears to be either un-moderated or negatively moderated by policy agreement.  55  What this suggests, then, is that if the media coverage of a transnational actor focuses on its foreign policy, the audience’s evaluation of that actor as the source of a persuasive message, and subsequent responses to the attempt at persuasion will be more positive if the audience generally agrees with its policies, and negative if it disagrees with the policies. If coverage does not focus on the source’s policies, however, its source cue effect should be largely independent of the audience’s evaluation of its policies. 2.5 DISCUSSION AND CONCLUSION This chapter points to a phenomenon that has yet to be studied in the constructivist literature on persuasion – the role of a messenger’s identity in the success of her transnational persuasive communications – and proposes that the concept of source cue effects can be usefully imported into IR from political psychology. The theoretical predictions and empirical findings point to several clear expectations about when a 56  transnational advocate’s identity is likely to resonate positively. First, the target audience must have a positive opinion of the advocate. This may seem obvious, but as I argue in the first section, there are a number of good reasons to be skeptical about the transnational persuasiveness of even extremely popular global political actors like President Obama. One limiting factor is the degree to which global audiences are familiar with the advocate in question. If they do not know much about the actor – particularly if the actor is an individual leader rather than a country or other corporate entity –, then that actor’s identity is unlikely to resonate and her persuasiveness will be attenuated. A second limiting factor involves the audience’s communications environment. If global policies associated with the advocate are salient, then responses to the advocate’s communications are likely to be moderated by the audience’s attitudes towards those policies. This presents a problem for actors like the United States, which has an extremely popular president, but which is associated with extremely unpopular foreign policies (such as the War on Terror). The findings suggest that the United States will be most transnationally persuasive when policies like the war in Iraq are not salient in a targeted jurisdiction’s media environment. Looking forward, the concept of source cue effects suggests a number of further research questions relating to transnational persuasion. First, what other implications does the literature on source cues have for when transnational source cue effects are most likely to occur in global politics? Second, are there any features of source cue effects and political persuasion that are unique to transnational politics? For example, does the degree of nationalism in the target audience influence the reaction to a foreign source cue, perhaps resulting in a backlash against “foreign interference”? Third, under what  57  conditions are transnational advocates and their opponents able to instrumentally manipulate a target audience’s media environment (by, say, increasing their visibility or increasing the salience of the advocates’ foreign policies) as a means of increasing or constraining those advocates’ persuasiveness? To the extent that persuasion is a key mechanism underlying change in global politics, answering these questions would help us to better understand both the spread of specific international norms and policies and the broader processes of globalization. They would also help us to better understand the role of public diplomacy in global politics and provide more information to practitioners about when transnational advocacy campaigns are likely to work and when they are likely to fail, or even backfire.  58  Chapter 3  The Global Bully Pulpit and U.S. Soft Power  3.1 INTRODUCTION As I note in Chapter 1, the sharp increase in anti-Americanism in the last decade (Kohut et al. 2008) prompted concern that the decline in the United States’ global standing was undermining its soft power (Nye 2004), thereby reducing its ability to achieve its foreign policy objectives. This concern is lent validity by Datta’s (2009; 2008), Goldsmith and Horiuchi’s (n.d.), and my own findings in Chapter 5, all of which indicate that aggregate public attitudes towards the United States can influence other states’ willingness to cooperate with the U.S. The debate on the increase in antiAmericanism during the Bush Administration’s tenure gave rise to a number of proposals for reversing the decline in the U.S. global image, which fall into two broad categories. On the one hand, some scholars and practitioners – motivated by the belief that antiAmericanism partly reflects foreign publics’ lack of information about the U.S. and their asymmetrical exposure to hostile propaganda (Nisbet et al. 2004; Graber 2009) – have advocated a renewal of U.S. public diplomacy as a means of increasing American soft power (Rugh 2009; Nye 2004; Snow 2007). While public diplomacy includes a wide variety of components, one aspect involves international visits and direct persuasion by U.S. leaders such as the President and the Secretary of State (Goldsmith and Horiuchi 2009). A separate line of thinking focuses on this type of public diplomacy, putting great stock in the ability of then-Senator Barack Obama to single-handedly transform the U.S.’s global image on his ascension to the presidency. As Nicholas Kristof, the New  York Times columnist, put it: “We're beginning to get a sense of how Barack Obama's political success could change global perceptions of the United States, redefining the American "brand”... This… would help rebuild American political capital in the way that… John Kennedy's presidency did in the early 1960s.”32 The rationale for this expectation varied widely, ranging from a belief that majority Muslim publics would find it easier to identify with an African-American President who had a Muslim name to simply the fact that President Obama, by virtue of being a liberal Democrat and a multilateralist, would represent a clear break with the global image projected by President Bush. This perspective suggests that, in addition to Obama’s rhetorical skills (see Krebs and Jackson 2007 for the role of rhetoric in global politics), Obama’s identity and positive image would in themselves bolster American soft power by improving the country’s image and increasing the U.S.’s global persuasiveness, which is a key tool of public diplomacy. The dramatic improvement in the United States’ global image since Obama’s election in November of 2008 and his inauguration in January of 2009 (Kohut et al. 2009) lends support to this perspective. Together, these two lines of thinking raise a key research question for students of global politics: what effect does presidential image and identity have on U.S. soft power? We can break this question into two sub-questions, based on two commonly studied components of soft power. First: does the identity of the U.S. President influence foreign evaluations of the United States? That is, does a popular U.S. President reduce antiAmericanism? Second, and more directly related to soft power: Can U.S. Presidents be transnationally persuasive? That is, can they use their personal brand to move foreign public opinion in a direction that is more supportive of U.S. policy preferences? Existing 32  “Rebranding the U.S. with Obama,” The New York Times, October 23, 2008, A37.  60  evidence does suggest that attitudes towards country leaders play a role in shaping attitudes towards a country as a whole (McGraw and Dolan 2007), and I deal with this topic in Chapter 4. The results presented in Chapter 2 suggest that U.S. Presidents (and, in particular, President Obama) can be transnationally persuasive. In this chapter I deal with a more specific question: under what conditions are U.S. Presidents transnationally persuasive, and what limits exist to a popular President’s personal persuasiveness? 3.1.1 The U.S. Presidency and Transnational Persuasion  Theodore Roosevelt coined the term “bully pulpit” to describe the rhetorical opportunities afforded to him by his office – more specifically, it refers to the fact that the visibility of the Presidency gives the President a direct connection with the American public and an opportunity to persuade it to support his or her preferred policies. Along with a few other high-profile figures such as the U.N. Secretary General, U.S. Presidents enjoy a similar level of visibility on the global stage. Indeed, we might call the U.S. Presidency a “global bully pulpit”. Just because the U.S. President is heard, however, does not mean that he will necessarily be able to persuade foreign publics to support his objectives. While research in a domestic context has demonstrated the ability of the U.S. President to move opinion (Bailey et al. 2003l; Mondak et al. 2004; Mondak 1993)33 – while showing that these efforts can also backfire on occasion (Sigelman and Sigelman 1981) –, there are a number of reasons to be skeptical about the extent of the U.S. President’s transnational persuasiveness.  33  While Presidents do exert source cue effects in survey research, however, the literature is decidedly skeptical about the ability of Presidents to move the U.S. public through major speeches (see Edwards 2009, for example).  61  First, opponents of the President’s message in a targeted state would likely be able to mobilize opposition to the proposed policy using charges of “foreign interference” in the country’s domestic politics, though both the nature of the Presidential appeal and the openness of the targeted public would presumably moderate the extent to which this dynamic emerges. Second, foreign leaders and foreign actors generally are unlikely to be particularly salient opinion referents, particularly for issues, such as climate policy, that touch on domestic politics. The U.S. President may also not be a particularly salient opinion referent for individuals who have low levels of information about the American political system, since they are unlikely to have strong opinions about the President that are distinct from their opinions about the United States as a whole. Finally, publics may mistrust the motives of a U.S. President, feeling that as a foreign advocate, he may be promoting policies that are designed to promote U.S. interests to the detriment of the targeted publics’ interests. The contribution of this chapter, then, is twofold. On the one hand, I provide empirical evidence that U.S. Presidents can be transnationally persuasive, even in a relatively “hard” case. This suggests that the skepticism outlined above is not entirely warranted. On the other hand, I use existing theories of persuasion and political communication to develop a set of hypotheses about the conditions under which U.S. Presidents are likely to be transnationally persuasive and when their ability to move global opinion is likely to be constrained. I test these hypotheses empirically using two survey experiments administered to undergraduate students at the University of British Columbia in the spring and winter of 2009.  62  3.2 THEORIES OF PRESIDENTIAL IMAGE AND PRESIDENTIAL PERSUASIVENESS While President Obama is a gifted orator, and rhetorical strategies like framing have been carefully researched in the context of political persuasion in both domestic and international contexts (Payne 2001; Chong and Druckman 2007b), I focus on how Obama’s and other U.S. Presidents’ image and identity influence their persuasiveness, rather than focusing on their rhetorical skills. Instead of drawing on the literature on framing effects, I use the concept of the source cue effect, which is the second broadly accepted mechanism of message-based persuasion (Eagley and Chaiken 1993). In this formulation, individuals respond to the personal attributes of the source of a persuasive message, leading that source to have an independent effect on attitudes even when holding the types of arguments used constant. Once again, the literature on source cues in political psychology provides us with a useful platform from which to investigate how presidential image influences the transnational persuasiveness of U.S. Presidents, As I discuss in previous chapters, explanations for source cue effects in political psychology can be grouped into two broad categories based on the type of attributes of the source which they claim moderate the source cue effect. Generally speaking, theories focus on either 1) the audience’s general affect or opinion towards an actor (Mondak 1993; Pelika 2007) or 2) the credibility (variously defined) of that actor in a specific domain (Lupia and McCubbins 1998; Lupia 2000; Druckman 2001a; Chong and Druckman 2007a).34 Most work on source cue effects (with a few exceptions, such as (Pelika 2007)) relies on the proposition that individuals use environmental cues as a way to make decisions about political issues while expending the least amount of cognitive  34  As I argue below, this distinction is important when thinking of how presidential persuasiveness can vary across issue-types.  63  effort – that is, as “heuristics” or judgmental shortcuts (Sniderman et al. 1991).35 This “low-information rationality” perspective follows from dual-process models of attitudes in psychology (Eagley and Chaiken 1993), where individuals can either form an attitude by consciously processing all the information they have available on an issue (‘central’ or ‘systematic’ processing) or rely on judgmental shortcuts to arrive at an opinion with minimal effort (‘peripheral’ or ‘heuristic’ processing). This literature leads us to a number of hypotheses about when U.S. Presidents will be most persuasive in a transnational context. I begin with two reasonably intuitive hypotheses. Both strands of the literature reviewed above make it very clear that a prerequisite for persuasiveness is to either be liked (positive affect among the target population) or if one is not liked, to be perceived as credible, with credibility defined as the combination of knowledgeability and trustworthiness (Lupia 2000; Lupia and McCubbins 1998). Thus: H1 (favorability): On aggregate, presidents who are perceived more favorably will be more persuasive than presidents who are perceived unfavorably. H2 (credibility): On aggregate, presidents who are perceived to be more credible on a given issue will be more persuasive than presidents who are perceived to lack credibility. Based on existing cross-national surveys (Kohut et al. 2008; Kohut et al. 2009), then, we should expect President Obama to be more transnationally persuasive than former  35  Framing, on the other hand, operates by manipulating the salience of the considerations that individuals take into account when attempting to come to a reasoned position on an issue (Nelson et al. 1997; Chong and Druckman 2007b).  64  President Bush. We should also expect, all things being equal, that favorability and credibility moderate responses to presidential cues at the individual level. While existing theory provides clear expectations about the pre-requisites for presidential persuasiveness, it also provides expectations about the limiting factors on persuasiveness. Indeed, that is perhaps the more interesting question when considering a globally popular President like Barack Obama. The following sections outline three sets of constraints on presidential persuasiveness that are particularly relevant in a transnational context. 3.2.1 Information and Presidential Cues  Viewing presidential source cue effects as an example of heuristic processing suggests at least two hypotheses about the role of audiences’ levels of information in moderating these effects. On the one hand, we should expect source cue effects to be most apparent when individuals have low levels of policy-specific information (Bullock 2008). Low information is correlated with heuristic or peripheral processing because individuals do not have enough relevant information to engage in the considered analysis characteristic of systematic or central processing. Thus: H3 (Policy-Specific Information): Presidential cues will have weaker effects on individuals with high policy-specific information than on those with low information levels. This hypothesis is not original, though it does imply that U.S. Presidents will be more persuasive when discussing complex or low salience issues on the global stage, rather than ‘high politics’ issues like the war in Iraq. I propose a second way in which information can influence responses to presidential cues. Namely, if individuals have  65  low information about the U.S. political system, they are unlikely to have strong attitudes about the U.S. President, and indeed, are unlikely to have distinct attitudes about the President as opposed to a more general opinion about the United States as a whole. Low information individuals may therefore find it difficult to use the presidential cue in forming their attitudes, since they may not have a strong opinion about the President and, more likely, may not have enough information to evaluate her credibility. Thus: H4 (U.S. Information): Presidential cues will have weaker effects on individuals with a low level of information about the U.S. political system than on those with a high level of information. Note that this hypothesis is particularly relevant to a transnational context, since in a domestic environment one would expect very few members of the public to lack an opinion about the President. Moreover, to the extent that levels of policy information and information about the U.S. President are not collinear (and they are not), we can put these expectations together to yield the following hypothesis: H5 (Hi-Lo): Presidential cues will have the greatest effect on individuals with low policy-specific information AND high U.S.-specific information. 3.2.2 U.S. Salience and Decreased Trust  The literature on source cues suggests that individuals are particularly receptive to “costly” cues – where the source of a message takes on a counter-intuitive position or a position that appears to go against her interests – because this lessens concerns that the source does not have the audience’s interests in mind when making its recommendation and can therefore be more easily trusted (Howell and Kriner Unpublished; Lupia 2000; Lupia and McCubbins 1998). The converse of this proposition has been less explored,  66  however, and is particularly relevant to persuasion across jurisdictional and group boundaries, where mistrust may be the default position. Namely, if a source is perceived to be self-interested, does that reduce its persuasiveness? In the context of transnational presidential persuasiveness, we can easily imagine that global audiences will be less likely to trust a U.S. President’s policy recommendation if they think he is advocating a policy that is disproportionately in the United States’ national interest or that the U.S. has a particularly large stake in. The inference would be that the President’s communication is self-interested, and that he can therefore not be fully trusted to take other countries’ interests into account. Thus: H6 (self-interest): Presidential cues will have a weaker effect on policy attitudes when the policy appears to disproportionately favor U.S. national interest. 3.2.3 Issue-Specific Credibility  Credibility is a more nuanced moderator of source cue effects than favorability in that it is not an invariant attribute, but can vary across issues for any given source. Lupia and McCubbins’ (1998) model of persuasion, for example, allows the perceived knowledgeability of an actor to vary across issue-areas. We might consider, then, whether U.S. Presidents are likely to be systematically perceived by foreign publics as more credible on certain issues than on others. In this section I present one hypothesis about when this might happen. The starting point is a literature in American politics that suggests a different understanding of credibility than that espoused by the low-information rationality approach to persuasion (Lupia 2000). Namely, a number of scholars have pointed to the fact that the Republican and Democratic parties in the United States appear to “own”  67  certain issues (Petrocik 1996). For example, in recent decades the Republican Party has been consistently seen as a more credible manager of national security issues than the Democratic Party (Petrocik 1996, 832), which, according to the issue ownership literature, leads Republican candidates to emphasize national security issues on the campaign trail so as to maximize their electoral chances. Issue ownership has also been observed in other polities, including the Netherlands (Brug 2004) and Canada (Belanger 2003). It is possible that this difference in credibility simply reflects differences in perceived knowledgeability and trust on this issue between the parties, as the lowinformation rationality perspective would suggest. We can, however, also think of this as an example of symbolic credibility, where actors are perceived as credible or competent because their social position identifies them as an ‘authority’ (Pelika 2007), regardless of their actual expertise on a given topic. From this perspective, Democrats could be seen as lacking credibility on national security even if they are generally trusted and are perceived to be knowledgeable on the topic because it is the GOP that is the authoritative party on national security issues. One explanation for this gap in credibility is the common wisdom that the U.S. public tends to view the Democrats as the “Mommy” party, which it listens to on social issues, and the GOP as the “Daddy” party, which it listens to on national security issues (eg. Liebovich 2007; Duerst-Lahti 2005). This gendered view of national security makes it difficult for Democrats (particularly Democratic women) to be credible leaders in that arena (Swers 2007) and, according to Goldstein (2001), emerges in part from a “war system” culture, in which war is perceived as an experience where warriors must “dominate” their enemies, which excludes women – a traditionally dominated group –  68  from taking on this social role. Indeed, individuals’ comfort with inter-group dominance – as expressed in the social dominance orientation (SDO) trait – has been found to underlie the ‘gender gap’ in political attitudes (Pratto et al. 1997), a phenomenon that has been well documented in the area of national security (Eichenberg 2003). What this suggests, then, is that the gendering of national security helps to explain both the gender gap in national security attitudes and the way in which the Democratic Party, which is generally ascribed feminine traits, tends to suffer from a lack of credibility on national security issues. We might, then, wonder whether this type of gendering of national security influences how foreign publics respond to attempts by U.S. Presidents to persuade them on issues of war and peace. While we might observe some differences in Presidents’ domestic and global images due to differences in media coverage and we might expect that the political personas Presidents seek to project to a global audience may be different than those they use domestically, we should still expect some degree of consistency between foreign and domestic images. We should therefore expect that George W. Bush’s “macho man” image (Goldstein 2003) and Barack Obama’s “unisex” image (Cooper 2009) will extend, to some extent, to the global arena. We should also expect the gender stereotypes that appear to underlie the partisan differences in credibility in the United States to exist abroad. The gender gap in attitudes towards the use of force has certainly been documented in countries other than the U.S. (Eichenberg 2007), and Inglehart and Norris’ (2000) argument that the leftward movement of women’s political attitudes in post-industrial countries is in part a result of the rise of feminist political movements suggests that, at least in these countries, this gender gap results from  69  differences in the extent to which men and women view war as a gendered activity. All in all, this suggests the following hypotheses about President Obama’s transnational persuasiveness among publics in other post-industrial countries. Note that, as a visible minority President projecting a “unisex” image to avoid being seen as an ‘angry black male’ (Cooper 2009), Obama is unlikely to fit traditional male leadership stereotypes and would therefore be expected to have low symbolic credibility on war-related issues among people who hold a gendered view of warfare (mostly men). Thus: H7 (gender gap): For national security issues, an Obama cue will have a stronger effect on women’s attitudes than on men’s attitudes. H8 (non-security issues): For non-security issues, an Obama cue will have similar effects on men and women’s attitudes. 3.3 EMPIRICAL TESTING – TWO SURVEY EXPERIMENTS The data presented in this chapter were collected in two rounds of experiments conducted in April-June 2009 (N=198) and December 2009 (N=186). In each case, undergraduate volunteers that were enrolled in Political Science classes at the University of British Columbia in Vancouver, Canada took a computer-based survey at the UBC Political Opinion Laboratory within which were embedded a set of survey experiments, including the ones presented in this chapter. Participants received extra credit in one of their courses in exchange for participation. In this chapter, I present data from two survey experiments: the Mandatory Emissions Limits, or MEL, experiment (which was described in Chapter 2), and the Afghanistan experiment. Full question wording is provided in Appendix A. As in Chapter 2, I exclude data from participants who were not Canadian citizens or permanent  70  residents to avoid the possibility that the presence of American students in the sample would skew results. This yields an N of 168 for the spring 2009 wave and of 158 for the December 2009 wave. Note that while I have already presented some results from the MEL Experiment in Chapter 2, most of the analyses presented in this chapter are novel. 3.3.1 Experiment #1: Mandatory Emissions Limits (MEL)  As discussed in Chapter 2, the MEL Experiment begins by presenting arguments for and against imposing mandatory emissions limits in Canada, after which participants are asked whether they would agree that “[t]he Canadian government should pass laws that limit the amount of carbon dioxide that Canadian industries are allowed to emit.” The experimental manipulation involves attributing the arguments opposing mandatory limits to either Presidents Bill Clinton, George W. Bush, or Barack Obama, or simply to the United States, while the position remains un-attributed in the control group. Note that I have coded the dependent variable so that greater values indicate greater opposition to mandatory emissions limits in Canada. Since all presidential cues argue against emissions limits, this ensures that a positive statistical effect in fact represents a positive source cue effect. Note also that this experiment sets up a “hard” case for transnational persuasion by U.S. Presidents. Not only are the attributed sources not directly calling on Canada to change its emissions policies, but this experiment also involves a policy – emissions limits – that is as much a domestic as an international issue. We would therefore not expect foreign actors such as U.S. Presidents to be particularly salient opinion referents. The clear presidential source cue effects for Obama and Clinton that were presented in Chapter 2 therefore suggest that U.S. Presidents do have a significant ability to persuade foreign publics to support the policies they prefer.  71  3.3.1.1 FINDINGS Table 3.1 presents the same findings as in Table 2.2. Namely, using difference of means tests, I find clear positive source cue effects for both the Obama and Clinton cues, and a possible marginal effect for the United States. Measured on a 1-to-5 scale, Mean opposition to mandatory emissions limits was 1.53 in the control group (falling between the “Agree Strongly” and “Agree” values), whereas in the Obama and Clinton groups, mean opposition was 2.26 and 2.10, respectively (between “Agree” and “Neither agree nor disagree”). The differences of means with the control group for the two cue conditions are 0.73 (p=0.0006) and 0.57 (p=0.0015), respectively, which on a 1-5 scale represent effects of about 18% and 14% of the range of the dependent variable. These are robust source cue effects. The effects for the U.S. cue (0.34, p=0.058) and the Bush cue (0.22, p=0.23) are smaller, and marginally statistically significant at best. Table 3.1: Opposition to Mandatory Emissions Limits (by Presidential Cue) Obama  Clinton  Bush  U.S.  Control  Mean opposition to MELs  2.26  2.10  1.75  1.87  1.53  Difference of means relative to Control Group  0.73  0.57  0.22  0.34  n/a  t statistic of the difference  3.64  3.32  1.22  1.93  n/a  p-value (two-tailed)  0.0006  0.0015  0.2268  0.0576  n/a  N  27  30  32  45  34  What, then, do these main effects tell us about the hypotheses outlined in Part I? First and foremost, they support the favorability (H1) and credibility (H2) hypotheses,  72  since the source cue effects are ordered according to aggregate favorability and credibility. In order to measure general opinion towards the sources, I asked participants to report whether their opinion of the three presidents and the United States was very or somewhat favorable, or very or somewhat unfavorable. Table 3.2 presents aggregate favorability for the four actors in the MEL Experiment sample. President Obama was most popular, with a 56% “very favorable” rating, and an overall 95% favorable rating. He is followed by President Clinton, with 86% favorability, the United States, with 55% favorability, and President Bush, with only 8% favorability. This favorability ordering matches the persuasiveness of each actor, with Obama having the largest source cue effect, followed by Clinton, the United States, and Bush. Notably, only Obama and Clinton, who are extremely popular in this sample, have clear persuasive effects. Table 3.2: Presidential Favorability in MEL Experimental Sample Obama  Clinton  Bush  US  Very Favorable  55.7%  27.3%  1.8%  7.7%  Somewhat Favorable  38.9%  58.8%  6.6%  47.6%  Somewhat Unfavorable  3.6%  11.5%  26.8%  35.7%  Very Unfavorable  1.8%  2.4%  64.9%  8.9%  N 167 165 168 168 I measure the credibility of each of the presidents using Lupia’s (2000) approach, which involves measuring the perceived knowledgeability of each President on this issue, and the general alignment of participants’ interests with each President (i.e., trust). To measure knowledgeability, I ask participants how much they think the presidents know about “the economic effects of limiting carbon dioxide emissions.” To measure trust, I  73  use the following question: “How often would you say that you and the following political leaders agree on most political issues?” Distributions for these two variables for the three Presidents are presented in Table 3.3. As with favorability, we find that Obama has a slight advantage over Clinton in both knowledgeability and trust, while Bush is seen as less knowledgeable and is mistrusted. If credibility is defined as being seen to know some or a lot about the issue and being seen to share one’s political views most or almost all of the time, then Obama is clearly seen as credible by participants, with 76% finding him credible (Table 3.4). Clinton is also credible on aggregate, though less so, with 53% credible and 47% not credible. Finally, Bush has effectively no credibility on this issue in this sample, with only 2% of participants finding him credible. Again, this credibility ranking is consistent with the ordering of source cue effects by magnitude. In short, the main effects of the MEL Experiment are consistent with both the favorability and credibility hypotheses, suggesting that popularity and/or a perception of credibility on the issue in question are prerequisites for U.S. Presidential persuasiveness in a transnational context. Of course, Lupia (2000) would argue that any correlation between persuasiveness and favorability only exists because favorability influences perceptions of knowledgeability and common interests. While it is impossible for me to determine which factor is causally prior with the available measures, it is worth noting that the findings are consistent with existing theories of source cue effects. The clear implication of these findings, then, is that President Obama, given his widespread global popularity (Kohut et al. 2009) is likely to be much more persuasive to foreign publics than President Bush was. Moreover, the large effects for both Obama and Clinton in what is a hard case for transnational persuasion also suggest that popular U.S. Presidents  74  Table 3.3: Perceived Knowledgeability and Trust in Three Presidents (Knowledgeability) How much would you say that these political leaders know about the economic effects of limiting carbon dioxide emissions? Obama  Clinton  Bush  Knows nothing  0.61%  0.64%  35.76%  Knows a little  7.36%  17.20%  38.18%  Knows some  48.47%  55.41%  18.18%  Knows a lot  43.56%  26.75%  7.88%  163  157  165  N  (Trust) How often would you say that you and the following political leaders agree on most political issues? Obama  Clinton  Bush  Almost Never  2.47%  1.90%  55.15%  Only some of the time  18.52%  39.24%  40.00%  Most of the time  60.49%  51.90%  4.24%  Almost all of the time  18.52%  6.96%  0.61%  162  158  165  N  Table 3.4: Presidential Credibility for MEL Experiment OBAMA  CLINTON  BUSH  Credible  76%  53%  2%  Not Credible  24%  47%  98%  N  158  152  162 75  have a significant potential to exert soft power on behalf of the United States by means of public diplomacy. 3.3.2.1 LIMITS TO PERSUASIVENESS While it is a useful to confirm that President Obama can be persuasive in a transnational setting, this is not a particularly surprising finding. A more interesting endeavor is to attempt to define when popular U.S. Presidents like Barack Obama and Bill Clinton are likely to be most persuasive, and what limits exist on their persuasiveness and personal soft power. Hypotheses 3-8 provide some expectations about the limits of presidential persuasiveness. I begin by testing the information-related hypotheses using the MEL Experiment. 3.3.2.1.1 Information  Hypotheses 3 and 4 (the Policy-Specific Information and the U.S. Information hypotheses) predict that presidential source cue effects should be weakest for individuals that have a low level of information about U.S. politics and for individuals that have a high level of policy-specific information. In order to test these hypotheses in the context of the MEL Experiment I use two measures of information. First, to measure participants’ familiarity with the U.S. political system, I asked them to identify the U.S. Vice President (Joe Biden) in an open-ended answer.36 55% of participants were able to correctly answer this question. Second, to measure participants’ climate change-specific knowledge, I asked them to “list the names of as many different greenhouse gases as you can think of (up to a maximum of 5),” also in an open-ended format. A table of greenhouse gases from the IPCC’s Fourth Assessment Report was used to code correct answers, and I then created a variable indicating the number of correct GHGs listed by 36  The data was collected between April and July of 2009.  76  each participant. Overall, 41% of the sample was unable to report more than one greenhouse gas (mostly carbon dioxide), 26% of the sample correctly listed two, and the rest (33.3%) listed more than two. For the analyses below, I designated those who listed two or fewer GHGs as “low information” (coded 0) and those who listed two to five as “high information” (coded 1).37 Note that while the two information variables are related (r = 0.19, p=0.01), they are hardly identical. It therefore makes sense to think about these two variables separately. In order to test the effects of information levels I ran a set of ordered logistic regressions using the opposition to mandatory emissions limits variable as a dependent variable and the treatment dummies as independent variables. I broke the sample down by levels of information. Table 3.5 presents these results. The full sample model confirms the results of the difference of means tests presented in Table 3.1, with strong effects for Clinton and Obama, and non-significant effects for Bush and the U.S. cue. As a first step, I split the sample into low and high U.S. information sub-samples, and fit the same model. The results are consistent with the U.S. Information hypothesis. In the low U.S. information model, neither the Obama nor Clinton cues have a significant effect on policy attitudes, whereas in the high U.S. information model both effects become statistically significant. Similarly, splitting the sample into high and low policy information groups, we find that none of the cues has a statistically significant effect among the high policy information group, but both the Clinton and Obama cues have an effect in the low policy information group. Again, this is consistent with the Policy Information hypothesis. Note also that in one case the strongest effect is for the high  37  I use overlapping categories in order to ensure that the policy-specific information groups have approximately equal sample sizes.  77  information group (U.S. information) and in the other it is for the low information group (policy information). This is consistent with theoretical expectations, and should increase our confidence that there are two separate information effects in play. Indeed, Hypothesis 5 (Hi-Lo) predicts that presidential source cues will have the greatest effect on individuals that have both a high level of U.S. information and a low level of policy-specific information. To test this expectation, I further break down the sample into four sub-samples, filling out the possible combinations of the two types of information. I find that the cues have a statistically significant effect only in the high U.S. information/low policy information group, and that in this group, the U.S. source cue has a significant effect, in addition to the Clinton and Bush cues. Admittedly, the sample sizes (particularly for the low U.S. information and high policy information group) are small, but the pattern of results is clear, and consistent with all three information hypotheses. 3.3.3 Experiment #2: The Afghanistan Experiment  The Afghanistan Experiment begins by presenting participants with information about the NATO mission in Afghanistan, and reporting on how participants at “a recent NATO summit” called for NATO countries to “maintain or increase the number of combat forces they have deployed in Afghanistan”. In the control group, this appeal is attributed simply to “supporters of this mission,” while in the treatment condition it is attributed to President Obama. Participants are then asked whether they would support extending the Canadian military deployment in Afghanistan, with support measured on a standard 5-point scale ranging from “Strongly oppose” (1) to “Strongly support” (5). In a separate factor, I manipulated the extent to which the operation in Afghanistan is  78  Table 3.5: Effect of Presidential Cues on Opposition to Mandatory Emissions Limits by Subject Information Levels  Full Model  Lo US Info  Hi US Info  Lo Policy Info  Hi Policy Info  Lo US/Lo Pol  Low US/ Hi Pol  Hi US/ Lo Pol  Hi US/ Hi Pol  Obama Cue  1.777*** (3.41)  1.013 (1.35)  2.699*** (3.5)  2.165*** (3.44)  0.838 (0.85)  1.192 (1.35)  0.61 (0.41)  3.402*** (3.48)  1.197 (0.88)  Clinton Cue  1.507** (3.06)  1.276 (1.85)  1.627* (2.23)  1.811** (2.96)  1.012 (1.12)  1.175 (1.48)  1.853 (1.15)  2.355* (2.24)  0.566 (0.51)  Bush Cue  0.466 (0.97)  0.471 (0.67)  0.319 (0.48)  0.965 (1.67)  -0.99 (-1.06)  0.632 (0.81)  -39.21 (-0.00)  0.965 (1.10)  -0.81 (-0.73)  USA Cue  0.791 (1.79)  0.603 (0.87)  0.998 (1.71)  1.081 (1.93)  0.261 (0.34)  0.539 (0.65)  0.919 (0.65)  1.648* (2.06)  -0.0863 (-0.09)  168  76  92  112  56  59  17  53  39  N  t statistics in parentheses * p<0.05, ** p<0.01, *** p<0.001 Note: This table shows coefficients for ordered logistic regression on dummy variables indicating experimental conditions. Cut-off constants are omitted.  79  characterized as a American operation by highlighting the fact that the International Security and Assistance Force (ISAF) is a U.S.-led mission commanded by a U.S. general. This is meant to test how a communications and media environment in which a particular policy is strongly associated with the United States influences presidential persuasiveness. I use the Afghanistan Experiment to test my remaining hypotheses about the limiting factors on transnational presidential persuasiveness. 3.3.3.1 ISSUE-SPECIFIC CREDIBILITY In Part 1 I argue that U.S. Presidents may find it more difficult to persuade foreign publics to support their preferred policies in the context of national security than in other issue-areas. Specifically, I argue that President Obama will be more persuasive on national security issues among women than among men because he does not fit the traditional gender role of a military leader, and men tend to have a more gendered view of warfare (the gender gap hypothesis – H7). I test this hypothesis using the Afghanistan Experiment. Table 3.6 shows mean support for extending the Canadian deployment in Afghanistan in the control group and in the group that received the Obama cue, for both the full sample and for males and females separately. Note that, as in the sample for the MEL Experiment, Obama is extremely popular, with 89% of participants reporting a very or somewhat favorable opinion. The full sample is, on balance, mildly opposed to extending the mission, and the Obama source cue effect is positive (0.22), but not statistically significant. Breaking the sample down by gender, however, suggests that this null effect at the aggregate level is a function of a positive source cue effect among women, and a negative effect among men. Among men, I find higher baseline support for extending the mission than in the full sample, but find that support in the Obama cue  80  condition is actually lower than in the control group (the difference of means is -0.23), though this difference is not statistically significant. Among women, we find substantially lower baseline support for the policy, but a strong positive source cue effect for Obama (difference of means = 0.75, p = 0.005), which, as in the MEL Experiment represents about 19% of the range of the dependent variable. Table 3.6: Effect of Obama Cue in Afghanistan Experiment by Gender Full Sample  Males Only  Females Only  Support in Obama Cue Group N=  2.605 86  2.684 38  2.574 47  Support in Control Group N=  2.362 69  2.912 34  1.829 35  Difference of means  0.242  -0.228  0.746  T-statistic  1.22  -0.795  2.896  0.2257  0.4294  0.0049  P-value (two-tailed)  On first instance, then, these results appear to support the gender gap hypothesis. This gender difference in receptiveness to an Obama source cue could be due to a number of other variables that vary with gender, however, including an asymmetry in partisanship, ideology, or general attitudes towards the use of force. To test the hypothesis more systematically, I present the results of a set of ordered logistic regressions in Table 3.7, including a bivariate model and a multivariate model that controls for a set of potential confounds. Major Party ID is a dummy variable indicating whether participants identify with the federal Conservative or Liberal parties, which are the two governing parties in Canada, and both of which have been involved with  81  initiating and/or extending the mission in Afghanistan while in government. Ideology is a simple self-placement on a standard left-right scale of political ideology, where 0 is left and 10 is right. Finally, Pacifism is an ordinal variable measuring whether participants think that in general “Canada should use more military force,” are unsure about their position, or think that “Canada should use more diplomacy”. The bivariate models in Table 3.7 replicate the results from the comparison of means tests, with a positive, but non-significant effect of the Obama cue for the full sample, a negative but non-significant effect for males, and a positive, and statistically significant effect for females. When I control for potential confounds, I find a similar pattern, although the level of significance for the Obama cue among females is weaker, with a p-value of 0.065. Now, it is important to note that since I did not measure acceptance of gender stereotypes directly, I cannot conclusively state that the results confirm the mechanisms underlying the gender gap hypothesis, but the fact that the gender difference persists even when controlling for potential confounds does suggest that a more fundamental gender difference is driving the variation in effects. This is consistent with the gender role mechanism. The logic behind the gender gap hypothesis suggests that the negative-leaning response to an Obama cue among males is due to President Obama having a low “symbolic” credibility in the area of national security. A further indirect test of this hypothesis, then, involves taking into account the rational-choice measure of credibility. As in the MEL Experiment, I measured Obama’s perceived knowledgeability on “fighting the war on terrorism” and participants general agreement with Obama on political issues (perceived interest alignment), and created an Obama credibility variable by multiplying these two variables (yielding a variable with range 1-16). The gender role  82  Table 3.7: Effect of Obama Cue on Support for Extending Afghan Mission, by gender Bivariate Models  With Controls  Full Model  Males Only  Females Only  Full Model  Males Only  Females Only  0.346 (1.19)  -0.37 (-0.86)  1.244** (2.9)  0.204 (0.66)  -0.32 (-0.71)  0.896+ (1.84)  Major Party ID  0.123 (0.33)  0.32 (0.57)  0.264 (0.5)  Ideology (Self-placement)  0.17+ (1.77)  0.0574 (0.42)  0.307* (2.09)  -1.730*** (-5.14)  -1.591** (-3.23)  -1.468** (-2.92)  147  70  77  Obama Cue  Pacifism  N  155  72  82  t statistics in parentheses + p<0.10 * p<0.05, ** p<0.01, *** p<0.001 Note: Table presents estimated coefficients based on ordered logistic regressions. Constants are not reported.  83  Table 3.8: Moderating Effect of Credibility on Afghan Support Full Model  Males Only  Females Only  Obama Cue  -0.965 (-0.95)  1.423 (1.00)  -2.538 (-1.63)  Major Party Self-ID  0.128 (0.34)  -0.152 (-0.26)  0.412 (0.75)  Ideology (Self-ID)  0.175 (1.78)  0.0998 (0.69)  0.293* (1.96)  -1.692*** (-4.86)  -1.882*** (-3.48)  -1.535** (-2.87)  Credibility (Obama)  0.006 (0.07)  0.371** (2.85)  -0.246 (-1.91)  Obama Cue X Credibility  0.132 (1.18)  -0.208 (-1.35)  0.390* (2.29)  142  68  74  Pacifism  N  Joint Significance Test for Obama Cue and Interaction Term Chi squared =  1.7  2.52  7.78  p-value =  0.43  0.28  0.02  t statistics in parentheses * p<0.05, ** p<0.01, *** p<0.001  84  argument suggests that individuals that have a gendered view of war (mostly men) would tend to perceive Obama as lacking credibility on security issues even if they thought him knowledgeable and generally trustworthy. We should therefore expect that Obama’s level of rational credibility on terrorism should not differ substantially between men and women, but that this credibility variable should only moderate responses to the Obama cue among women (since symbolic credibility is more uniformly low among men). Results from my experiment support these expectations. Mean rational credibility among men is 8.42, and 8.97 among women, and the difference between these two means is not statistically significant. More interestingly, this rational credibility is an effective predictor of the magnitude and direction of the Obama source cue effect among women, but not among men. Table 3.8 shows the same ordered logit models as in Table 3.7, but with the rational credibility variable and an interaction between credibility and the Obama cue included. I test the significance of the interactions by running a joint significance test on the cue dummy and the interaction term, as suggested by Brambor et al. (2006), and find that the interaction is not significant for the full model or for the male sub-sample, but is significant for the female sub-sample. Among women, the effect of the Obama cue among participants who perceive him to have low knowledgeability and trustworthiness is negative, but becomes more positive as his credibility increases. Again, these results are consistent with the idea that a different mechanism is governing responses to the Obama source cue among men than among women, and that this mechanism is a different form of credibility based on a gendered understanding of war.  85  Table 3.9: Effect of Presidential Cues on MEL Attitudes, by gender Full Model  Males Only  Females Only  Obama Cue  1.765*** (3.36)  2.592** (2.73)  1.737** (2.61)  Clinton Cue  1.489** (3.01)  1.746* (2.36)  1.336+ (1.93)  0.464 (0.96)  0.581 (0.82)  0.406 (0.60)  123  53  70  Bush Cue  N  t statistics in parentheses + p<0.10 * p<0.05, ** p<0.01, *** p<0.001 Note: Table presents ordered logit coefficients. Constants are not reported. The complement to the gender gap hypothesis is the non-security issues hypothesis (H8), which predicts that this gender gap should not exist for non-security issues. To test this, Table 3.9 presents the estimated effects of presidential cues on opposition to mandatory emissions limits, for male and female sub-samples separately. We find that the overall effects for the Obama and Clinton cues exist for both men and women, and that the Bush cue does not have a significant effect on either gender. This pattern of effects is markedly different than in the Afghanistan experiment, and suggests both that the non-security issues hypothesis is correct, and that we should only expect to observe a significant gender difference in responses to President Obama for national security issues, and not for other political issues (like climate change).  86  3.3.3.2 U.S. SALIENCE AND PERCEPTIONS OF SELF-INTEREST The Afghanistan Experiment also allows me to test the self-interest hypothesis (H6), which argues that presidential cues will have weaker effects when the advocated policy appears to disproportionately favor U.S. national interests. In this case, I expect that the manipulation designed to increase the salience of U.S. leadership of ISAF should increase participants’ perceptions that the operation in Afghanistan is primarily a U.S. operation. This should reduce the extent to which individuals believe that they and President Obama share common political interests, since his appeal will appear to be selfinterested. We should therefore expect that the Obama source cue effect should be reduced in the U.S. leadership condition. Moreover, based on the findings in the previous section, we should expect to observe this reduction among women (because they appear to be responding to the cue based on the source’s perceived credibility, which is a function of the extent to which the audience perceives the source’s interests to align with their own – i.e., Lupia and McCubbins’ (1998) definition of trust), but not among men, who appear to be relying on a different cognitive mechanism. As a manipulation check, I asked participants whether they believed ISAF is “a NATO operation in which all the allied countries are doing their part” or “primarily a U.S. mission operating under a NATO flag.” Table 3.10 shows that among participants that received the U.S. leadership cue, 79% saw ISAF as being primarily a U.S. mission, while among those who did not receive the cue, only 62% had the same view. This difference of 17 percentage points suggests that the cue had the intended effect, and a difference of means test shows that it is statistically significant, although it is not  87  surprising that a majority of participants viewed the mission as U.S.-led regardless of the leadership cue condition. Table 3.10: Perceptions of ISAF by U.S. Leadership Condition No U.S. Leadership Cue  U.S. Leadership Cue  62%  79%  84  71  t-statistic (difference of means)  2.315  -  p-value (two-tailed)  0.0220  -  % Primarily U.S. Mission N  To test for the moderating effect of the U.S. leadership cue, I regress support for the mission in Afghanistan on the Obama cue, the U.S. leadership cue, and an interaction between the two, using ordered logistic regression. Table 3.11 shows this model fitted on the full sample, and on females and males separately. Again, I run tests for the joint significance of the Obama cue and the interaction term. I find that the interaction is not statistically significant for the full sample or for males, but is significant for the female sub-sample. As expected, the baseline effect of the Obama cue among women is positive, and is reduced in the U.S. leadership condition. Figure 3.1 shows predicted changes in the probability of each value of the dependent variable if participants receive the Obama cue for the female sub-sample.38 As we would expect, the probability of strongly opposing the mission decreases while the probability of supporting the mission  38  First differences were estimated using the Clarify software package for STATA (King et al. 2000).  88  increases. Note that for all but one of the values, the predicted effects are larger in magnitude when participants do not receive the U.S. leadership cue. Table 3.11: Moderating Effect of U.S. Leadership Cue in Afghanistan Experiment Full Sample  Females Only  Males Only  Obama Cue  0.294 (0.73)  1.374* (2.35)  -0.46 (-0.75)  U.S. Leadership Cue  -0.151 (-0.35)  -0.0213 (-0.03)  0.0141 (0.02)  Obama Cue X U.S. Leadership Cue  0.116 (0.20)  -0.31 (-0.37)  0.15 (0.17)  155  82  72  1.46  8.25  0.81  0.4808  0.0161  0.6656  N t statistics in parentheses * p<0.05, ** p<0.01, *** p<0.001  Joint Significance of Obama Cue and Interaction Term Chi squared = P-value =  Overall, then, the Afghanistan Experiment appears to strongly support the selfinterest hypothesis (H6). Admittedly, the magnitude of the reduction of the Obama source cue effect in the U.S. leadership condition is small, but this has to be balanced against the fact that the mission in Afghanistan is already widely perceived to be primarily a U.S. mission. Were we to perform a similar experiment using a policy that is not already associated with the U.S., we might expect a much larger effect on presidential persuasiveness when participants are induced to view the presidential appeal as selfinterested. 89  3.4 DISCUSSION AND CONCLUSION As might be expected, the two experiments reported in this chapter confirmed that popularity and/or credibility are crucial prerequisites for successful public diplomacy by U.S. Presidents. More specifically, I find that President Obama is likely to be more persuasive in the global arena than former President George W. Bush ever was. This suggests that those who expected Obama’s election to restore some measure of the United States’ global soft power were correct. That said, the results of these experiments also point to a number of important limitations to President Obama’s (and, indeed, other U.S. Presidents’) transnational persuasiveness. First, as suggested by dual-process models of attitude formation, I find that U.S. Presidents will be most persuasive among individuals who have a high level of information about the U.S. political system, but a low level of policy-specific information. Thinking about the implications of this finding in the real world, we might  90  expect, then, that U.S. Presidents will be most persuasive with respect to highly technical policies or with respect to newly emerging issues. They are likely to face significant limitations, however, in their ability to move global opinion on longstanding issues with which publics are familiar. Moreover, to the extent that the publics of other Western states and close allies of the United States are likely to be more informed about U.S. politics than publics in other parts of the world, we might speculate that presidential soft power will extend primarily to existing U.S. allies (like the NATO member states) and will be severely constrained among enemy or competitor states. I have also presented a wide range of circumstantial evidence in favor of the proposition that President Obama (and presumably other Democratic Presidents) will suffer from low symbolic credibility on security issues among individuals who espouse a gendered view of warfare. In post-industrial countries, this will include primarily men, but in other types of states may include the majority of the population. This suggests that President Obama may find it difficult to use his positive global image to persuade other countries to support his military initiatives, such as the recent troop “surge” in Afghanistan. Finally, the results of the Afghanistan Experiment appear to confirm that when a U.S. President is seen to be making a self-interested appeal, this will tend to reduce her credibility on that issue, thus limiting the persuasive impact of the communication. This finding has two important implications. On the one hand, it suggests that attempting to characterize military operations as multinational operations, either through token participation by allied countries (Coleman 2008) or by engaging in public diplomacy meant to shape media coverage of a mission can indeed have an impact on global support  91  for the operation. On the other hand, and more importantly, it suggests that Presidential soft power may be least effective when it is most needed – that is, when the United States’ national interest is directly at stake. These findings also point to several avenues for future research. First, it is important to confirm the findings of this study and investigate its implications in a wider context by replicating these or similar experiments on national probability samples in a wider range of countries (particularly in non-Western countries). This would provide more substantive evidence as to how the transnational persuasiveness of U.S. Presidents varies across different types of publics. Second, the information and self-interest hypotheses point to the need for scholars of political behavior and political psychology to consider how their existing theories apply to or need to be modified to operate in a transnational context. This is particularly important given the increasing political and scholarly relevance of transnational activism (Keck and Sikkink 1998) and transnational social movements (Tarrow 2001), where actors routinely attempt to use persuasion and other mechanisms of social and political influence in a transnational setting. Finally, the findings in this chapter point to the potential contribution that existing theories and methods in political psychology can make to the study of global politics and international relations. Constructivist work in IR, for example, focuses heavily on social psychological phenomena like persuasion (Payne 2001) or shaming (Hafner-Burton 2008), but typically (with notable exceptions like (Johnston 2001)) does not engage with the literature in political psychology that deals with these concepts. Among other problems, this has led constructivist scholars to largely ignore the influence of actor credibility and source cue effects in global politics (see Chapter 2 of this dissertation),  92  despite their heavy use of related concepts like framing. A particularly important area of research would involve investigating how the identity and attributes of country leaders in general influences the soft power (persuasion and social influence) of states and their ability to act as norm entrepreneurs and agents of socialization in the international system.  93  Chapter 4  Presidential Priming and Anti-Americanism  4.1 INTRODUCTION In the previous chapter, I argued that the sustained debate in academic and policy circles as to the causes of and the potential solutions to the declining U.S. global image (Willnat et al. 2006; Carlson and Nelson 2008; Katzenstein and Keohane 2007b; Nossal 2005; Ray and Johnston 2007) suggests that one way of reversing this decline – and thereby improving the United States’ soft power – was to elect President Obama, who has proved to be an extremely popular figure among global publics. I examined one way in which President Obama can improve U.S. soft power in Chapter 3 – namely, Obama’s popularity facilitates his personal ability to persuade foreign publics to support his preferred policies. In this chapter, I investigate a second mechanism: the extent to which President Obama’s personal popularity can improve the global image of the United States (a variable that is associated with foreign governments’ decisions to support U.S. policy initiatives, as I show in Chapter 5). Besides the dramatic improvement in U.S. favorability following Obama’s election (Kohut et al. 2009), the existing literature does suggest that U.S. Presidents can influence global attitudes towards the United States. Goldsmith and Horiuchi (2009), for example, examine what they call “high-level” public diplomacy, where high-ranking political leaders such as the President and the Secretary of State directly address foreign publics in an effort to improve the United States’ global image.39  39  As they note, this is an example of transnational political communication.  Goldsmith and Horiuchi’s (2009) finding raises two questions about the role of the U.S. President in shaping the global image of the United States. First, how does highlevel public diplomacy work at a psychological level? Specifically, how does increasing the short-term visibility of the U.S. President in a given jurisdiction (as might happen during a presidential visit) change the way in which the public in that jurisdiction views the United States as a whole? Like most of the literature on public diplomacy and international communication, Goldsmith and Horiuchi (2009) focus on the aggregate dynamics of public diplomacy, measuring the credibility of U.S. officials, for example, using a three-point time-period variable and not based on individual-level perceptions, and do not investigate the specific psychological mechanisms underlying the effects they uncover. One contribution of this chapter, then, is to identify and describe one such mechanism. Second, and following directly from the answer to the first question: among whom is the transition from a globally unpopular to a globally popular U.S. President (such as the shift from George W. Bush to Barack Obama) likely to enhance the United States government’s ability to engage in effective high-level public diplomacy? I begin to address these two questions in Part 2 by developing a theoretical framework within which to understand what I call the presidential priming effect of public diplomacy, which outlines a specific cognitive mechanism through which presidential popularity can shape global opinions of the United States. Using this model, I lay out a set of hypotheses about when and among what types of people high-level public diplomacy is likely to be most successful in improving the U.S.’s global image. In Parts 3 and 4, I test these hypotheses using two survey experiments administered to undergraduate students at a Canadian university.  95  4.2 LEADER EFFECTS ON COUNTRY IMAGE In order to theorize the effect of presidential public diplomacy on global attitudes towards the United States, I begin by drawing on the concept of media priming, which Goldsmith and Horiuchi (2009, 865) also use to set up their argument. As Scheufele and Tewksbury (2007, 11) note, priming is one of the three primary mechanisms of media influence (the others being framing and agenda-setting), and operates by increasing the salience of specific considerations in people’s minds (in other words, it increases the accessibility of those considerations), which in turn increases the weighting that those considerations receive when the audience is asked to make a political evaluation (Iyengar and Kinder 1987). Presidential-level public diplomacy as described by Goldsmith and Horiuchi (2009, 865), then, is likely to influence opinion about the United States at least in part as a result of presidential priming. That is, we might hypothesize that presidential public diplomacy increases the weighting of individuals’ attitudes towards the President of the United States in their evaluations of the U.S. as a whole.40 To the extent that the president is popular, then, this presidential priming effect should lead to more positive general attitudes towards the United States. It is important at this point to distinguish between the presidential priming effect – which captures the short-term spike in the visibility of the U.S. President that results from a presidential visit – and the longer-term effect that a transition from an unpopular to a popular U.S. President is likely to have on global opinion. This latter effect – which, as 40  An alternative hypothesis about how high-level public diplomacy could influence public attitudes towards the United States refers to the content of the President’s communications. Either increased information about the United States or its policies, or a new way of framing them could lead to change in attitudes towards U.S. policies, and consequently a change in general attitudes towards the United States. These effects, however, do not directly depend on foreign publics’ evaluations of the U.S. President, and would therefore be hard-pressed to explain the huge effect that the election of Barack Obama has had on global attitudes towards the United States. As such, I choose to focus on the priming effect in this paper, and leave informational and framing effects to future research.  96  noted, has been observed in the Bush to Obama transition (Kohut et al. 2009) – is likely to be a function of both the new president’s global image and, just as important, actual or predicted changes in U.S. policies. The effect of U.S. foreign policy on global antiAmericanism has been studied elsewhere (Lynch 2007; Nossal 2005; Nisbet et al. 2004; Isernia 2007; Johnston and Stockmann 2007), however, and a rigorous analysis of the changes in policy between the Bush and Obama Administrations lies beyond the scope of this chapter. I therefore focus exclusively on investigating 1) the psychological dynamics that underlie the short-term presidential priming effect of high-level public diplomacy, and 2) how the transition from the Bush to the Obama presidency is likely to change the magnitude and direction of the presidential priming effect, and as a consequence, the U.S.’s global image. 4.2.1 A Model of Attitudes Towards the United States  I begin by elaborating a more formal definition of the presidential priming effect. The expectancy value model of attitudes conceives of attitudes as “summary evaluations based on a weighted average of a sample of beliefs about the attitude object” (Nelson et al. 1997, 225). Using this framework, which follows from the memory-based model of attitudes, we can conceive of an individual’s general attitude towards the United States (AUS in Eq. 1) as a weighted sum of “n” considerations,41 including the individual’s opinion of the U.S. President. Since this chapter’s substantive focus is on the effects of the transition between the Bush and Obama Administrations, I include individual terms for the considerations relating to people’s opinions of both President Obama (vO) and former-President Bush (vB). 41  vi are the considerations, which would either negatively or positively influence the individual’s evaluation of the United States, while wi are the non-negative weights associated with each consideration.  97  [Eq. 1]  AUS = ! (vi * wi) + vOwO + vBwB , for i = {1 to n-2}  The presidential priming effect, then, involves an increase in the weight for the consideration involving the individual’s opinion of the President in question (wO or wB), which thereby strengthens the overall impact of those presidential evaluations (vO or vB) on the overall evaluation of the United States (AUS), and changes AUS from its unprimed state. In the remainder of this section, I develop a set of hypotheses about when, and among whom, this presidential priming effect is likely to be strongest. 4.2.2 Hypotheses  The most basic hypothesis is that if the U.S. President is popular among the audience – that is, if he is evaluated positively by a majority of the audience –, priming the President will improve the audience’s aggregate opinion of the United States (H1 popular president hypothesis).42 A more general hypothesis is that – at the individual level – people’s attitudes towards the U.S. President will moderate the presidential priming effect, causing them to hold a less favorable opinion of the United States if they dislike the President and to hold a more favorable opinion of the U.S. if they like him (H2 – favorability moderator hypothesis). A complementary hypothesis is that when the U.S. President is primed, attitudes towards the President will be better predictors of general attitudes towards the United States than when the President is not as salient (H3: favorability priming hypothesis). 42  Aggregate attitudes towards the United States is how public anti-Americanism is usually measured in the literature and is the most widely available measure in the growing set of regular multi-country surveys (Kittilson 2007). In fact, aggregate anti-Americanism has been shown to influence other countries’ policies towards the United States (Datta 2009; Goldsmith and Horiuchi n.d.).  98  Returning to Eq. 1, note that the three hypotheses outlined in the previous paragraph result from the change in the weightings of considerations (wO or wB) caused by the increased salience of the U.S. President. It follows, then, that priming individuals’ attitudes towards the U.S. President can only have an impact on their overall opinion about the United States (AUS) if their attitudes towards the U.S. President (vO or vB) are distinct from their overall attitudes towards the United States. Indeed, if AUS = vO or AUS = vB, then priming the presidential consideration will only reinforce people’s existing attitudes towards the United States, and should not result in an observable change in levels of anti-Americanism. Existing research suggests that global publics do in fact sometimes hold different attitudes towards different U.S.-related attitude objects (the American people and government, for example – (Isernia 2007, 61)), but the extent to which attitudes towards the U.S. President and attitudes towards the United States are independent is likely to vary substantially across individuals (and indeed, across countries and sub-publics). A variable that is particularly likely to influence the extent to which people hold distinct attitudes is the extent to which individuals have access to information about the U.S. President (the more information, the less likely they are to simply transpose their general opinion of United States to the President). From a general perspective, then, we should expect the presidential priming effect to be strongest among individuals who are most familiar with U.S. politics and weaker among those who are less familiar with U.S. politics (H4: presidential familiarity hypothesis).43  43  Put another way, individuals with lower levels of information about U.S. politics will have few political considerations in memory when evaluating the U.S. When the U.S. President is primed, then, they are more likely to form their consideration about the President based on their general opinion of the United States, which will therefore not change AUS. High information individuals are more likely to have an existing attitude towards the U.S. President in memory as part of a larger set of considerations. When the  99  4.2.3 Presidential Transitions and the Presidential Priming Effect  A second set of substantive hypotheses relate to the change in the presidential priming effect of American high-level public diplomacy that has resulted from the transfer of power from George W. Bush to Barack Obama. The major effect of moving from one president to the other is that the transition changes the object that is primed by a high-level presidential visit (this is obvious, since it would be a different president undertaking the trip). Increasing the weighting of a popular president (Obama) through a prime, then, should lead to a sharp increase in positive attitudes towards the United States when compared to the effect of priming his unpopular predecessor (Bush) (H5: the Obama effect hypothesis). Moreover, to the extent that President Obama promised a shift in foreign policy and – at least in the period immediately following his inauguration – could not be as closely linked to recent U.S. foreign policy as could President Bush, we should expect an audience’s opinion of U.S. foreign policies to be less predictive of general attitudes towards the United States when President Obama is primed than when President Bush is primed (H6: policy un-priming hypothesis). 4.3 DATA AND RESULTS The data presented in this chapter were collected in two rounds of experiments conducted in April-June 2009 (N=198) and December 2009 (N=186). In each case, undergraduate volunteers that were enrolled in Political Science classes at the University of British Columbia in Vancouver, Canada took a computer-based survey at the UBC Political Opinion Laboratory, within which were embedded a set of survey experiments,  President is primed, the increased weighting on the presidential consideration is therefore more likely to lead to a change in AUS.  100  including the ones presented in this chapter. Participants received extra credit in one of their courses in exchange for participation. In this chapter, I present data from two survey experiments: the Mandatory Emissions Limits, or MEL, experiment, and the Climate Treaty experiment. Full question wording is provided in Appendix A. The primary purpose of both experiments is to test for transnational source cue effects when attributing policies to U.S. Presidents (these results are described in Chapters 2 and 3 of this dissertation). A secondary aspect of the experimental design, however, uses the presidential source attributions as shortterm presidential primes (that should, in theory, simulate the short-term presidential priming effect caused by high-level public diplomacy). Priming effects are estimated by comparing the reported favorability towards the United States (which was measured after the experimental questions) across different presidential attribution conditions. In the next sub-section, I use the data from the first experiment to analyze the priming effect on U.S. attitudes immediately after the administration of the experimental manipulation. I then test for the robustness of this effect by combining data from the two experiments and examining the extent to which this priming effect persists when U.S. attitudes are measured later in the survey. As in previous chapters, because my focus is on how foreign attitudes towards the United States are affected by presidential priming, I exclude data from participants who were not Canadian citizens or permanent residents to avoid the possibility that the presence of American students in the sample would skew results. This yields an N of 138 for the spring 2009 wave and of 158 for the December 2009 wave.  101  4.3.1 Experiment #1: Mandatory Emissions Limits  As described in Chapters 2 and 3, the Mandatory Emissions Limits (MEL) Experiment begins with a vignette that familiarizes participants with the concept of mandatory greenhouse gas emissions limits for industry as a way of preventing climate change and provides arguments both for and against the policy. The experimental manipulation involves varying the U.S. President to which the argument opposing mandatory emissions limits is attributed. In the two control groups, the opposing argument is simply attributed to “others” or to “the United States”, while for the treatment conditions, it is attributed to “former U.S. President George W. Bush” or to “U.S. President Barack Obama” (or to “President Clinton”, but this condition is not used in this chapter). Participants are then asked whether they support mandatory emissions limits in Canada. As noted earlier, the effect of the experimental manipulation on policy support represents the main effect of the experiment. For the purposes of this chapter, however, I focus on the secondary dependent variable. Immediately following the policy question, participants were asked to report whether they had a “very favorable,” “somewhat favorable,” “somewhat unfavorable”, or “very unfavorable” opinion of the United States. This is the favorability question used in a wide variety of global public opinion surveys, most notably as part of the Pew Global Attitudes Project. The expectation is that the source attributions to Bush and Obama will act in much the same way as the short-term spike in presidential media salience caused by highlevel diplomatic visits, thereby exerting a short-term presidential priming effect. Note that the only difference between conditions is the object primed (whereas the policy  102  position attributed and the arguments used remain the same). Figure 4.1 presents mean levels of favorability towards the U.S. in each of the four experimental groups.  The first thing to note is that aggregate opinion of the United States in the sample is more or less neutral, though slightly positive. Mean favorability in the control condition is 2.53, which on a 1 to 4 scale is just above the mid-point, and for the full sample (N=168), 44.6% have an unfavorable opinion of the United States, while 55.4% have a favorable opinion. Opinions of Obama and Bush, however, are much more polarized, with 92% of the sample holding an unfavorable view of Bush and 95% of the sample holding a favorable view of Obama. We should therefore expect that priming Bush would depress U.S. favorability, while priming Obama would increase it. Figure 4.1 confirms this expectation, with mean U.S. favorability at 2.74 in the Obama condition and 2.34 in the Bush condition, compared to the 2.53 in the control condition and in the United States attribution condition. While the differences between the means of the presidential conditions and the control conditions are not statistically significant using a t-  103  test, they approach significance (with single-tailed p-values of between 0.10 and 0.15).44 The direction of effects is as expected, however, and the difference in mean favorable opinion of the United States between the condition in which President Bush is primed and when President Obama is primed is statistically significant (0.4, or about 10% of the range of the dependent variable p<0.05). Overall, then, the main effects of this priming experiment are consistent with the popular president hypothesis (H1) and the Obama effect hypothesis (H5).45 I next proceed to analyze this same data using ordered logistic regression, but since the non-presidential conditions show no aggregate difference, I pool them into a single control group from now on. Table 4.1 presents a range of regression models, where opinion of the United States is regressed on dummy variables indicating the two presidential priming conditions (the baseline condition is the pooled control group). I also include four control variables in half of the models. These are the respondent’s gender (the expectation being that women tend to be less favorable towards the United States), whether the participant identifies with the federal Conservative Party (which tends to be more in favor of closer relations with the United States than the other parliamentary parties), and participants’ opinions of Presidents Bush and Obama, which were measured later in the survey. Table 4.2 presents a similar set of models, but compares the Obama prime condition directly to the Bush prime (excluding the control data), with the Bush prime condition as the baseline.  44  Single-tailed t-tests are justifiable because there is a strong theoretical expectation about the direction of the presidential priming effect for both Bush and Obama. 45 Note, moreover, that the design of the experiment makes it particularly difficult to find an effect for the Obama prime because a highly unpopular (and unexpected) policy position is attributed to him. This could plausibly have temporarily reduced his favorability among participants.  104  The first thing to note is that the effects observed in the comparison of means tests are replicated in the regression analyses. Namely, the Obama prime has a positive and significant effect on attitudes towards the United States, both when compared to the unprimed control conditions (Model 2) and when compared to the Bush condition (Models 9 and 10). The Bush prime, however, does not appear to have a statistically significant effect on attitudes towards the United States, though it is worth noting that the direction of the Bush priming effect is negative in all eight models in Table 4.1, which is consistent with President Bush’s overall unpopularity in the sample. These findings strongly support the expectations that a popular U.S. President will improve general opinion about the United States (H1: the popular president hypothesis) and that the transition from Bush to Obama is likely to have improved the United States’ global image (H5: the Obama effect hypothesis). Moreover, these are substantively meaningful effects. Figure 4.2 shows the predicted changes in the probability of a typical participant holding a very unfavorable through very favorable opinion of the United States as a result of the Obama prime.46 All three models show a substantial decrease in the probability of individuals holding an unfavorable opinion of the United States and an increase in the probability of favorable opinions. The largest effects occur for Model 10, where the Obama prime results in a decrease of 0.30 in the probability of holding a somewhat unfavorable opinion and an increase of similar magnitude in the probability of holding a somewhat favorable opinion.  46  Probabilities were computed using the first differences function of the Clarify software package for STATA (King et al. 2000).  105  Table 4.1: Presidential Priming and U.S. Favorability in the MEL Experiment (Control Groups Comparison) Model 1  Model 2  Model 3  Model 4  Model 5  Model 6  Model 7  Model 8  Obama Prime  0.574 (1.34)  0.983* (2.14)  -0.473 (-0.21)  0.373 (0.6)  0.688 (1.06)  0.393 (0.88)  1.632 (1.69)  1.472 (1.47)  Bush Prime  -0.451 (-1.14)  -0.277 (-0.68)  -0.606 (-0.69)  -0.314 (-0.54)  -0.328 (-0.55)  -0.313 (-0.78)  -0.301 (-0.43)  -0.321 (-0.44)  Bush Favorability  1.249*** (4.16)  1.181*** (3.45)  1.255*** (4.13)  1.227*** (3.87)  Obama Favorability  0.862** (3.23)  0.771** (2.62)  0.836** (3.11)  0.828** (3.03)  0.177 (0.42)  0.165 (0.39)  0.163 (0.39)  0.175 (0.41)  -0.876* (-2.40)  -0.853* (-2.32)  -0.825* (-2.13)  -0.822* (-2.21)  Tory Party ID  Female  Bush Prime X Bush Fav  0.234 (0.41)  Obama Prime X Obama Fav  0.431 (0.68)  Bush Prime X VP Info  -0.221 (-0.28)  0.106 (0.13)  106  Table 4.1: Presidential Priming and U.S. Favorability in the MEL Experiment (Control Groups Comparison) [Continued] Model 1  Model 2  Model 3  Model 4  Model 5  Model 6  Model 7  Model 8  US VP Info  0.324 (0.75)  0.0252 (0.05)  Obama Prime X VP Info  0.469 (0.54)  0.603 (0.67) 1.890* (2.54)  2.423* (2.5)  0.623 (0.57)  Obama Prime X PA  -2.85 (-1.46)  -1.322 (-0.65)  Bush Prime X PA  0.0934 (0.05)  0.246 (0.13)  136  135  Policy Agreement (PA)  N  138  t statistics in parentheses , * p<0.05, ** p<0.01, *** p<0.001  137  137  138  137  136  Note: Cut-off constants are not reported.  107  Table 4.2: Presidential Priming and U.S. Favorability in the MEL Experiment (Obama-Bush Comparison)  Obama Prime  Model 9  Model 10  Model 11  Model 12  1.001* (1.96)  1.757** (2.77)  0.664 (0.94)  1.412 (1.74)  Bush Favorability  1.790*** (3.4)  1.762*** (3.31)  Obama Favorability  1.449*** (3.49)  1.415*** (3.37)  0.0249 (0.04)  -0.0307 (-0.04)  Female  -1.996** (-2.73)  -2.088** (-2.81)  US VP Info  -0.0327 (-0.05)  Tory Party ID  Obama Prime X VP Info  N  59 t statistics in parentheses , * p<0.05, ** p<0.01, *** p<0.001  58  0.0983 (0.15)  -0.381 (-0.47)  0.699 (0.70)  0.785 (0.68)  59  58  Note: Cut-off constants are not reported.  108  The difference in the directions of the effects for the Bush and Obama primes also provides preliminary support for the expectation that the effect of a presidential prime will depend on the audience’s opinion of the President (H2: favorability moderator hypothesis). Model 3 tests H2 more rigorously by including interactions between the presidential dummy variables and individuals’ opinions of the two presidents (presidential favorability). In both cases, the coefficients for the interactive terms have the expected sign (positive). This suggests that the presidential priming does indeed have a stronger positive effect for individuals who have a more favorable opinion of the President in question. Neither of these interactions is statistically significant, however, though the Obama prime and the interactive term may be on the margin of joint statistical  109  significance (chi-squared=4.99, p=0.0824).47 Overall, this provides very tentative evidence in favor of H2. Model 3 also allows us to test the favorability priming hypothesis (H3) – namely, the expectation that the presidential primes will strengthen the relationship between people’s opinion of the president in question and their opinion of the United States. Interestingly, the presidential favorability terms and the interactive terms are jointly statistically significant in both cases (chi-squared=10.71, p=0.0047 for Obama and chisquared=17.49, p=0.0002 for Bush). Figure 4.3 illustrates the substantive magnitude of this relationship. It shows the change in the predicted probabilities of a typical participant holding a given opinion of the United States as a result of a significant improvement in that participant’s opinion of President Obama or Bush (moving from “very unfavorable” to “very favorable”). The effect of a change in presidential favorability on U.S. favorability is estimated separately when the president is primed (red bars) and when the president is not primed (blue bars). In all cases, the predicted effect of an improved presidential image on public attitudes towards the U.S. is greater when that president is primed than when the president in question is unprimed. Most of these differences are small, however, relative to the magnitude of the predicted effect of presidential favorability on general opinions of the United States – the only exception being the increased probability of a participant reporting a “very favorable” opinion of the United States when her opinion of Obama improves. In that case (rightmost bars in Figure 4.3), the overall effect of presidential favorability when Obama is primed is approximately three times the magnitude of the effect when no president is primed. This  47  I test the joint significance of the presidential dummies and the interaction terms as suggested by (Brambor et al. 2006).  110  does provide support for the favorability priming hypothesis (H3), but the small magnitude of the favorability priming effect together with the robust significance of the presidential favorability variables across all models suggests that people’s attitudes towards the U.S. President are important determinants of general attitudes towards the United States regardless of the short-term salience of the President in question. This is not surprising, of course, seeing as participants were political science undergraduates who likely had well-formed pre-existing attitudes towards both the two presidents in question and towards the United States government.  Models 6, 7, and 8 in Table 4.1 allow me to test the policy un-priming hypothesis (H6), which predicts that priming President Obama should reduce the extent to which agreement or disagreement with U.S. foreign policy is associated with overall opinion of 111  the United States.48 Conversely, this link should strengthen when President Bush is primed. As a measure of participants’ attitudes towards U.S. foreign policy, I constructed the Policy Agreement variable, which is an additive index (scaled from 0 for “low agreement” to 1 for “high agreement”) constructed from a set of items asking participants whether they agreed with: 1) the use of military force to prevent Iran’s acquisition of nuclear weapons, 2) the secret detention and 3) coercive interrogation of terrorist suspects, 4) and the assassination of terrorist leaders. These are all policies associated with the U.S. War on Terror, which was President Bush’s signature foreign policy initiative (and arguably the most controversial), and the global publics’ agreement with these policies should therefore be expected to influence their opinion of the United States. In Model 6, policy agreement is included as a control variable, and shows a positive and statistically significant association with attitudes towards the United States. In order to test H6, this policy agreement variable is interacted with the Obama and Bush dummies in Models 7 and 8. The effects of the interactions are in the expected direction, with the Obama prime reducing the magnitude of the effect of Policy Agreement on U.S. opinion, and the Bush prime strengthening this link. While both of these interactions are statistically significant in Model 7,49 the effect of Policy Agreement weakens considerably when partisanship, gender, and attitudes towards the two Presidents are included as control variables (Model 8), and the overall effect of the variable is no longer statistically significant. What this suggests, then, is that agreement with U.S. foreign policy is “priced in” to individuals’ attitudes of the U.S. President, and 48  Note that this data was collected in March of 2009, less than two months after his inauguration. The unpriming effect, if it exists, should steadily disappear as Obama takes ownership of U.S. foreign policy. Unfortunately, the data presented in this paper do not allow me to test this expectation. 49 Testing for joint significance of the presidential priming dummies and the interaction terms, I find chisquared = 6.29, p=0.043 for the Obama prime and chi-squared = 8.50, p=0.014 for the Bush prime.  112  that any link between policy (dis)agreement and anti-Americanism would flow through individuals’ opinions of U.S. leaders. Overall, then, the policy un-priming hypothesis (H6) is not supported by the evidence. In order to test the presidential familiarity hypothesis (H4), I use two different measures of familiarity. On the one hand, I measured participants’ familiarity with the U.S. political system directly using an open-ended item where individuals were asked to report the name of the current U.S. Vice-President, with the US VP Info variable coded as “1” if participants could identify the VP and “0” if they could not. On the other hand, to the extent that foreign publics’ primary source of information about U.S. politics is the news media, we can estimate participants’ familiarity with the U.S. President based on the extent to which media coverage of the United States features the U.S. President and the extent to which they pay attention to the news. I use the first, direct, measure in this section, and the second, indirect measure in Part 3. Models 4 and 5 in Table 4.1 and Models 11 and 12 in Table 4.2 use the US VP Info variable to test the presidential familiarity hypothesis (H4) by including interactions between the presidential primes and the information variable. Five of the six interactive term coefficients are in the expected, positive direction. This suggests that the presidential priming effects are stronger among those familiar with U.S. political leaders than among those who are less familiar. The interactions are not statistically significant, however, except for the Model 12 interaction (chi-squared = 8.05, p = 0.018). To gauge the magnitude of this apparent moderating effect, I estimated predicted priming effects for individuals with low and high familiarity using Clarify. These estimates are presented for Models 5 and 12 in Figure 4.4, which shows that for all values of the dependent  113  variable, the change in predicted probabilities caused by the Obama prime is larger for high familiarity individuals (red bars) than for low familiarity individuals (blue bars), though the differences in effect sizes are not dramatic. Overall, then, the MEL experiment provides support for the presidential familiarity hypothesis (H4).  4.4 TESTING FOR THE PERSISTENCE OF PRIMING EFFECTS The three most robust findings to emerge from the Mandatory Emissions Limits Experiment described above are: 1) that presidential primes can indeed have an effect on general attitudes towards the United States, 2) that the presidential priming effect is moderated by the extent to which individuals are familiar with the president in question, and 3) that the link between peoples’ opinions of U.S. Presidents and their general opinion of the United States appears to be strengthened when the figure of the President is more salient. In this section, I test these three findings using a separate set of data. 114  First, I continue to use the MEL experiment, but I use an alternative dependent variable – namely, I use a second measure of U.S. favorability that was taken between 5 and 14 items after the experimental question (an average of 9.5 items later).50 Since priming and framing effects have often been found to be fleeting, this presents a more difficult test for the presidential priming effect. Secondly, I use data collected in the December 2009 round, which featured a different source cue experiment (described below). This new data is useful not only because its dependent variable (U.S. favorability) was also measured some time after the experimental question, but also because it represents an independent test of the presidential priming effect – using an experiment that features a much more subtle presidential cue – and was conducted when President Obama was well into his first term and the presidential transition was completed. As I describe below, it provides evidence that the short-term presidential priming effect persists even outside the context of a high-profile presidential transition. In order to check the robustness of any findings, I conduct the analyses described below separately for both experiments and in a pooled sample, whose larger size should allow us to detect more subtle effects. First, however, I describe the second experiment, fielded in December of 2009. 4.4.1 Experiment #2: Climate Treaty and Presidential Primes  The Climate Treaty Experiment deals with the same substantive topic (climate policy), but from a foreign policy angle, focusing on the international negotiations on a new climate treaty rather than on domestic emissions policy.51 As in the MEL 50  The second U.S. favorability item was part of a battery of favorability measures where the order of countries and organizations was randomly rotated. 51 In fact, this data was collected at the same time as the Copenhagen Climate Change Conference (COP15) was underway, so its substantive policy focus was likely to have been a particularly salient issue for participants, particularly given UBC’s high-profile focus on reducing its overall GHG emissions and the ensuing salience of climate change policy at that university.  115  Experiment, participants read a set of arguments both for and against the policy in question, which is whether Canada and the other industrialized countries should demand that large developing countries like China and India commit to immediately begin reducing their emissions reductions. They are then asked whether they agree that “Canada should join a new international treaty on climate change EVEN IF China and India refuse to sign it.” The experimental manipulation involves attributing both the proand con- arguments to President Obama and to former President Bush, yielding four conditions with a single cue in each, and a control group where the arguments are not attributed to any specific actors. Additionally, two-sevenths of the sample instead received an un-related question dealing with policy aimed at preventing the spread of HIV/AIDS. These conditions are treated as part of the control group. Once again, these source attributions can be seen to have a presidential priming effect. Participants’ general opinion of the United States is measured with the same favorability item used in the MEL experiment, in this case appearing between 7 and 19 items after the experimental question (13 questions later, on average), which is comparable to the delay in receiving the second US favorability item experienced by participants in the MEL experiment. Opinions of Presidents Bush and Obama are also measured in the questionnaire.  116  Table 4.3: Effect of Presidential Priming on U.S. Attitudes and Favorability in Two Experiments Model 13  Model 14  Model 15  Model 16  Model 17  Model 18  Obama Prime  1.067* (2.32)  0.593 (1.44)  0.801** (2.65)  0.434 (0.20)  3.064 (1.79)  2.18 (1.59)  Bush Prime  0.216 (0.53)  -0.00192 (-0.01)  0.0979 (0.36)  -0.175 (-0.19)  -0.0838 (-0.09)  -0.141 (-0.22)  Obama Favorability  1.162*** (4.22)  0.621** (2.78)  0.836*** (4.82)  1.114*** (3.64)  0.810** (3.12)  0.933*** (4.72)  Bush Favorability  1.495*** (4.96)  0.631 (1.84)  1.131*** (5.11)  1.420*** (4.18)  0.63 (1.56)  1.090*** (4.31)  Female  -0.525 (-1.45)  -0.700* (-2.14)  -0.614* (-2.56)  -0.523 (-1.43)  -0.704* (-2.15)  -0.634** (-2.63)  Tory Party ID  0.289 (0.69)  0.667 (1.49)  0.413 (1.36)  0.295 (0.70)  0.708 (1.57)  0.425 (1.4)  Climate Experiment Dummy  -0.26 (-1.09)  -0.249 (-1.04)  Obama Prime X Obama Fav  0.189 (0.31)  -0.722 (-1.48)  -0.404 (-1.04)  Bush Prime X Bush Fav  0.291 (0.46)  0.0811 (0.11)  0.19 (0.42)  137  153  290  N  137  t statistics in parentheses * p<0.05, ** p<0.01, *** p<0.001  153  290  Note: Cut-off constants are omitted.  4.4.2 Findings  The effects of the presidential primes are estimated in the same manner as in the previous section by running ordered logistic regressions in which general U.S. favorability is regressed on dummy variables indicating which cue was received and on a set of control or moderating variables. The first three models in Table 4.3 (13 through  117  15) estimate the effect of the presidential primes without including any interactions. As with the analysis in the previous section, we find that the Obama prime has a positive and significant effect on U.S. favorability in the MEL Experiment sample (Model 13).52 While this positive effect for Obama is not statistically significant in the Climate Treaty Experiment sample (Model 14), the effect in the pooled model (Model 15) is. Overall, this suggests that while the effect may have weakened, the presidential priming effect for Obama did persist beyond the immediate aftermath of the prime, which represents added evidence in favor of the popular president hypothesis (H1). Models 16 to 18 include interactions between the presidential primes and the reported favorability for each president. Model 16 replicates the results reported in Part 2 for the MEL Experiment, with positive coefficients for both interactive terms. This indicates that the presidential priming effects are more positive for individuals who hold a favorable impression of the president in question (as suggested by the favorability moderator hypothesis (H2)), and that the association between presidential favorability and general attitudes towards the United States is strengthened by the presidential primes (as suggested by the favorability priming hypothesis (H3)). The first interactive effect approaches statistical significance for the Obama prime (chi-squared=5.48, p=0.0647) and the second is clearly significant for both presidents (chi-squared=17.59, p=0.0002 for Obama and chi-squared=23.97, p<0.00005 for Bush). Models 17 and 18, however, are inconsistent with the previous findings. For both the Climate Treaty data and the pooled data, the coefficient for the Obama interactive terms are negative (though the interactive term coefficient for the Bush prime is also positive and statistically significant in the  52  Note that the dummy variable indicating which experiment the data were collected in is not a significant predictor of attitudes towards the United States.  118  pooled model: chi-squared=26.45, p<0.00005). This suggests that the Obama presidential prime is weaker among those who hold a more favorable opinion of President Obama (chi-squared=8.06, p=0.0178) and that the association between participants’ attitudes towards Obama and their attitudes towards the United States is weaker among participants that receive the prime (chi-squared=24.36, p<0.00005). While this effect is quite unexpected, a potential explanation lies in the fact that the Climate Treaty Experiment was conducted in December of 2009, a full 11 months into President Obama’s term. One might speculate that by that point, people’s opinion of Obama had already been “priced in” to their general opinion of the United States, and as a result the effect of the experimental question is to prime the United States’ climate policy (which is plausible given that the data were collected during the 2009 Copenhagen Conference on Climate Change), thus increasing the influence of participants’ climate policy attitudes on their general opinion of the United States. Nevertheless, the principal interpretation is that evidence in favor of the favorability moderator (H2) and favorability priming (H3) hypotheses is mixed, and at best tentative. 4.4.3 Source Familiarity  In Part 2, I tested the presidential familiarity hypothesis (H4) by interacting the presidential primes with the US VP Info variable. In this section, I use the alternative approach mentioned earlier, focusing on media coverage. Before discussing the analysis and results, it is worth establishing that Presidents Obama and Bush did in fact receive  119  120  significant news coverage in Canada within this study’s time-frame. Figure 4.5 shows the proportion of stories in The Globe and Mail mentioning the United States that also mentioned Presidents Bush or Obama in a given month, starting with June of 2009 (the month in which Obama clinched the Democratic nomination).53 It shows that while the overall salience of President Bush and then-Senator Obama in the Globe and Mail’s coverage of the United States was about equal in the second half of 2008, coverage of Obama spiked upon his election, and particularly his inauguration. From that point forward, Obama maintained a high (if reduced) level of media salience, while former President Bush’s salience (which was not high to begin with) began a rapid decline to between one and two percent. Given the fact that President Obama did feature prominently in Canadian news coverage of the United States, we should expect that participants who reported a high level of attention to the news would be most familiar with Obama, and therefore most influenced by the presidential prime. I operationalize this attention variable in two different ways. First, I simply use an Attention to News variable that codes self-reported time spent following the news in the last few days (ranging from 1 for “Haven’t spent any time” to 6 for “More than two hours”). Full question wording is presented in Appendix A. Second, I combine this variable with the actual amount of media coverage of each president by multiplying the attention variable with the media salience of each president reported in Figure 4.5 for the month preceding the date of administration of the 53  I used Lexis-Nexis Academic (accessed April 21, 2010) to measure the relative salience of both Presidents Bush and Obama to the Globe and Mail’s coverage of the United States. I use the Globe and Mail because it is Canada’s largest newspaper with national distribution, and since my experimental sample consists of Canadian undergraduates, is therefore a relevant source for coverage of U.S. Presidents. I searched for the following terms anywhere in the document: United States ("United States" OR "U.S." OR "USA"), Bush (("United States" OR "U.S." OR "USA") and ("George W. Bush" OR "President Bush")), Obama (("United States" OR "U.S." OR "USA") and ("Obama" OR "Barack Obama" OR "President Obama")).  121  Table 4.4: Presidential Familiarity as a Moderator of Presidential Priming in Two Experiments Model 19  Model 20  Model 21  Model 22  Model 23  Model 24  0.97 (0.71) (0.05) 0.2 (0.18)  -1.461 (-1.33) (1.71) -1.538 (-1.58)  -0.431 (-0.52) (1.12) -0.606 (-0.86)  0.802 (0.62)  -1.18 (-1.07)  -0.348 (-0.44)  0.0532 (0.05)  -1.51 (-1.56)  -0.402 (-0.65)  Obama Favorability  1.148*** (4.15)  0.554* (2.45)  0.812*** (4.65)  1.153*** (4.10)  0.619** (2.67)  0.809*** (4.62)  Bush Favorability  1.470*** (4.84)  0.715* (2.04)  1.138*** (5.11)  1.457*** (4.76)  0.628 (1.79)  1.125*** (5.04)  Female  -0.602 (-1.58)  -0.798* (-2.15)  -0.685** (-2.62)  -0.595 (-1.55)  -0.836* (-2.24)  -0.697** (-2.65)  Tory Party ID  0.294 (0.7)  0.597 (1.32)  0.414 (1.36)  0.361 (0.85)  0.603 (1.33)  0.427 (1.40)  Attention to News  -0.1 (-0.61)  -0.422* (-2.13)  -0.228 (-1.84)  Obama Prime X Attention  0.035 (0.10)  0.608* (2.05)  0.36 (1.65)  Bush Prime X Attention  0.014  0.455  0.207  Obama Prime  Bush Prime  122  Table 4.4: Presidential Familiarity as a Moderator of Presidential Priming in Two Experiments [Continued] Model 19  Model 20  Climate Treaty Dummy  Model 21  Model 22  Model 23  -0.285 (-1.18)  Model 24 -0.406 (-1.21)  Obama Individual Salience  4.49 (0.78)  -14.11* (-2.05)  -0.561 (-0.16)  Bush Individual Salience  -27.4 (-1.08)  58.25 (1.44)  -9.478 (-0.54)  Obama Prime X Obama Salience  0.56 (0.16)  5.739 (1.78)  3.552 (1.57)  Bush Prime X Bush Salience  2.733 (0.22)  29.74 (1.72)  8.578 (0.96)  137  153  290  N t statistics in parentheses * p<0.05, ** p<0.01, *** p<0.001  137  153  290  123  experiment for each participant. The rationale is that individuals paying high attention would be more familiar with Presidents Obama and Bush than individuals reporting low attention, but only when actual media coverage of the two presidents is high. These two familiarity variables are interacted with the presidential primes in Models 19 through 24 (Table 4.4). The results are clear and robust to different model specifications. For both the Bush and the Obama primes, the interactive terms are positive in all models, indicating that the presidential priming effect is strongest for high familiarity individuals. That said, none of the interactions for the Bush prime were statistically significant. For the Obama prime, however, the moderating effect of familiarity is robust, with the joint significance of the prime and the interaction terms at p<0.05 in Models 20, 21, and 24 and p<0.10 in Models 19 and 23 (Model 22 has p=0.1065). Of particular note is that the results are consistent in both the Mandatory Emissions Limits Experiment and the Climate Treaty Experiment samples, separately, and in the pooled sample. Note also that the moderating effect of attention does not appear to simply be a proxy for political sophistication, since the effects persist even when domestic political information variables are included in the regressions (results are not shown). Overall, then, these results provide strong support for the presidential familiarity hypothesis (H4). 4.5 DISCUSSION AND CONCLUSION The empirical analyses in Parts 2 and 3 of this chapter point to some clear conclusions. First, short-term presidential priming effects for a popular president like President Obama can improve foreign publics’ general opinion of the United States (H1), particularly when compared to people’s opinion of the U.S. when an unpopular president  124  is primed (H5). While the evidence is mixed, these presidential primes appear to increase the association between people’s opinions of the presidents in question and their general opinion of the United States (H3), though this effect may be reduced as the Obama administration becomes more firmly established as the global face of America. Finally, the most robust finding is that the presidential priming effect is strongest among those who have a high degree of familiarity with the president in question (H4). If individuals are not familiar with the U.S. President, the priming effect is likely to be negligible. These results have a number of clear implications for the future of high-level U.S. public diplomacy (Goldsmith and Horiuchi 2009). First, due to President Obama’s personal popularity across the globe, presidential public diplomacy under his administration is likely to be much more effective at improving the U.S.’s global image and attenuating global anti-Americanism than it was during the Bush Administration. That said, because of his “star” status and the increasing link between President Obama and U.S. foreign policy as his first term progresses, it is possible that Obama may face a reduction in his ability to boost the U.S.’s global image as a result of his personal popularity, since global publics’ opinion of Obama will become “priced in” to their image of the U.S. as a whole.54 The evidence in favor of this possibility is tentative, however, and the unique attributes of the transitional period between presidential administrations deserve further study. Finally, it is clear that the effects of short-term increase in media coverage of President Obama are likely to be strongest among the subsets of national or global publics that are most familiar with the President. Familiarity is likely to vary both with individual characteristics (political interest, education, access 54  This can result, for example, from the personification (McGraw and Dolan 2007) of the United States in the person of the President, which should become increasingly established as the memory of former President Bush fades.  125  to media) and with the substantive content of media coverage in a particular jurisdiction. Given the findings presented in this paper, further investigation of how knowledge of the U.S. political system varies across global publics is warranted, since it would allow practitioners and scholars of public diplomacy to better predict when and where highlevel public diplomacy during the Obama administration is likely to be effective.  126  Chapter 5  When Does Anti-Americanism Matter?  5.1 INTRODUCTION The rapid decline in the standing of the United States among foreign publics during the past decade has prompted a resurgence in concern about anti-Americanism, both in academia and in popular discussion. In response, a sizeable literature has emerged across a range of disciplines and sub-fields of political science that aims to understand the nature and causes of the phenomenon (Nossal 2005; Ray and Johnston 2007; Isernia 2007; Chiozza 2007; Nye 2004; Nisbet et al. 2004; Carlson and Nelson 2008; Willnat et al. 2006; Graber 2009). Very little of this research, however, has explored what should arguably be the key question in the study of anti-Americanism: does it matter? As Katzenstein and Keohane (2007a) put it, “[t]o persuade Americans that [anti-Americanism] should not be so easily shrugged off, one would have to provide evidence that anti-Americanism either has direct political effects that damage U.S. interests and values, or that its indirect effects would be harmful” (273). Indeed, if it does not have a clear effect on actor behavior – particularly on countries’ foreign policies – it arguably should not be a major concern of the U.S. foreign policy establishment. In the only work to consider this question from a broad perspective, Katzenstein and Keohane (2007a) review several “most likely” cases in which attitudes towards the United States would be expected to have had an effect on policy.55 Prominent among 55  Their cases, ranging from the bilateral agreements regarding extradition of U.S. citizens to the International Criminal Court, to the War on Terrorism, to the invasion of Iraq, are all major U.S. foreign policy initiatives, and all represent policies that are inescapably tied to the United States. Unlike some of the policies examined in this dissertation (like climate change policy), the primary source of the policy should be clear to global publics and governments, and anti-American opinion should therefore be  127  these is an analysis of participation in the U.S. “Coalition of the Willing” during the invasion of Iraq. They find that “the evidence… does not support the view that mass opinion toward the United States in democratic countries affected decisions on whether to join the coalition” (Katzenstein and Keohane 2007a, 291). More broadly, having reviewed a wide array of literature and consulted a variety of issue specialists, they conclude that their “analysis suggests that even high levels of expressed antiAmericanism do not translate readily into government or individual action” (Katzenstein and Keohane 2007a, 303). In the only other recent works that directly address the question of whether attitudes towards the United States effect political behavior, ByongJuen Jhee (2008) finds that attitudes towards the United States had little if any effect on vote choice in the 2002 South Korean presidential elections, but Datta (2009) finds a relationship between aggregate public attitudes towards the United States and countries voting alignment with the United States in the United Nations General Assembly.56 Taken together, then, what research exists on the issue suggests that antiAmericanism has only a weak or occasional effect on political behavior. Katzenstein and Keohane (2007a) provisionally resolve the question by stating that “[t]he burden of proof is on those who claim that anti-American opinion, as distinguished from opposition to particular U.S. foreign policies, has had significant policy effects” (304). The problem, of course, is that if the policy effects of anti-Americanism are so hard to identify, it becomes difficult to justify both the growing concern over anti-Americanism during the Bush years and the body of research that accompanies it. particularly relevant in these contexts. With regards to Iraq in particular, Katzenstein and Keohane (2007a, 290) note that “[i]t seems that pervasive anti-Americanism should most strongly affect responses to highly controversial American policies, such as U.S. policy in Iraq.” 56 Some work in marketing research has looked into “country-of-origin” effects on consumer behavior. See Amine (2008) for a review.  128  This chapter takes on Katzenstein and Keohane’s challenge by focusing in greater detail on one of their flagship cases – the U.S.-led war in Iraq. Unlike their analysis, which only controls for states’ Freedom House scores when examining the relationship between public anti-Americanism and participation in the Coalition of the Willing, I use data on country contributions to the U.S.-led effort in Iraq for the years 2003-2009 and a collection of aggregate public opinion data to expand on their work by using regression analysis to test for the ceteris paribus influence of anti-Americanism.57 I find that when other factors are controlled for, aggregate public attitudes towards the United States are associated with state decisions to participate in the war and with decisions to withdraw from the operation. I also focus on what are arguably the more interesting questions: when does anti-Americanism matter and how much does it matter? I accomplish this, first, by deriving hypotheses about when anti-Americanism should matter using the major theoretical perspectives in International Relations, and second, using regression analysis to test these hypotheses empirically in the context of the Iraq War. Overall, I find that anti-Americanism is most likely to influence the foreign policy of powerful states – those that are least constrained by anarchy and the distribution of power – and is most likely to exert its effect when the policy in question (the war in Iraq, in this case) has high domestic salience in a given state. These findings speak not only to the literature on anti-Americanism, but also to the growing literature on global public opinion. A number of recent articles have taken advantage of the growing availability of cross-national surveys (see Kittilson 2007 for an  57  Datta (2008) deals with the same topic, and finds similar results to those presented below for decisions to participate in the Multinational Force (MNF) in Iraq, but he does not examine the effect of antiAmericanism on decisions to withdraw from the mission in Iraq, which I also address in this paper. Goldsmith and Horiuchi (n.d.) also address this topic.  129  overview), to study the determinants of global opinion on U.S. foreign policy (Goldsmith et al. 2005, for example) and the influence of global public opinion on policy behavior (Baum 2008b). Moreover, the finding that underlying public attitudes towards the United States influence countries’ foreign policy has clear implications for U.S. public diplomacy. Proponents of soft power, such as Joseph Nye (2004), argue that a decrease in the popularity of the United States can reduce its soft power, thus reducing “the ability of the American government to achieve its desired policy outcomes” (36). This ties in to the literature focused on the practice and effects of U.S. public diplomacy (Goldsmith and Horiuchi 2009; or Nye 2004). The findings in this study suggest that decreasing favorability towards the United States does sometimes decrease support for U.S. policy initiatives among foreign governments. To the extent that this lack of support constrains the U.S. government’s ability to act on the world stage, the results presented here highlight the need for the U.S. to make a sustained effort to maintain its global image. 5.2 ANTI-AMERICANISM AND IR THEORY I begin my discussion by setting anti-Americanism in a broad theoretical context. Different IR theories provide persuasive arguments for both why we should expect antiAmericanism to influence countries’ foreign policies and why we should not. Viewing anti-Americanism through these theoretical lenses, then, provides a number of predictions about when we should expect to observe the influence of anti-Americanism on policy. Before tackling the IR theories, however, it is worth considering what anti-Americanism is and how we can measure it, since different interpretations will shape expectations about when it will influence policy.  130  5.2.1 What is Anti-Americanism?  Generally speaking, attitudes towards the United States can be thought of as an ideational variable. Katzenstein and Keohane (2007b, 12-15), for example, differentiate between both the “cognitive,” “emotional,” and “normative” components of antiAmericanism, and between anti-Americanism as an individual attitude and antiAmericanism as a “collective belief”. These distinctions are both definitional (what do we mean when we talk about anti-Americanism?) and methodological (how do we measure anti-Americanism?). For the purposes of this chapter, however, such nuanced definitions are not directly relevant. The key question is to consider the conditions under which anti-Americanism, as measured by mass public opinion surveys, influences the policy behavior of states.58 Thus, the extensive debate on the nature of anti-Americanism (which is chronicled by the various contributions to Katzenstein and Keohane’s (2007b) volume is only relevant in as much as it speaks to the question of when and how it will influence policy behavior, such as decisions to participate in the war in Iraq and decisions to withdraw from the Multinational Force. Three questions about our measures of anti-Americanism are relevant to this analysis: First, are attitudes towards the United States simply measuring opinion about the Iraq War? If so, finding that these attitudes influence participation in the war would be interesting, but would not genuinely address this chapter’s research question. Part of 58  I focus on this measure because it is the only widely available direct measure of anti-Americanism, and as such, the best means of researching its effects on policy. It is worth noting, however, that I am not making the assumption that such surveys necessarily measure the independent levels of public antiAmericanism. Indeed, to the extent that public opinion is often a function of elite leadership (Zaller 1992, for example), it is equally possible that cross-national variance in these surveys reflect variance in the attitudes of elites towards the United States and towards U.S. policies. In this paper, I treat this as an empirical question and attempt to test it as far as possible, but, again, this matters only in as much as this distinction would change our expectations about when anti-Americanism is likely to have an impact on policy.  131  the empirical task, then, is to exclude this possibility. Second, does anti-Americanism have an independent influence on policy? As I argue below, a number of theoretical perspectives in IR would likely view any association between anti-Americanism and policy outcomes as a spurious relationship. Finally, and to the extent that any relationship between anti-American public attitudes and policy is not spurious, are we measuring the independent effect of public opinion, or are public attitudes towards the United States simply a proxy for the policy preferences of national elites? In the sections that follow, I examine three broad theoretical perspectives. Two are structural – neorealism and structural constructivism -, and both suggest that antiAmericanism is epiphenomenal to deeper structural variables. The other perspective – ideational liberalism and/or liberal constructivism – takes into account the role of antiAmericanism in domestic politics, and how this might shape policy outcomes. 5.2.2 Neo-realism  Given that anti-Americanism as defined above is an ideational variable, a realist perspective would suggest that it is unlikely to be an important determinant of state behavior in the international system. Neo-realist theory as laid out by Waltz (1979) suggests that state behavior in the international arena, particularly in the realm of security, is likely to be overwhelmingly determined by the distribution of relative power in the international system. This systemic variable shapes both the interests of and the options available to individual states, and fundamentally constrains their foreign policy behavior. Put more generally, when decision-makers make a decision on whether to go to war, they will pay foremost attention to exogenously determined material interests; thus their behavior is unlikely to reflect the influence of ideational variables like anti-  132  Americanism, regardless of whether we conceive of anti-Americanism as an attribute of the mass public or as an attitude held by decision-makers themselves. In short, antiAmericanism by and large should not matter.59 Moreover, much as Kagan (2003) argues that differences in strategic culture between Europe and the United States are merely a function of each entity’s relative power, or Mearsheimer (1994/5) questions the independent causal influence of international institutions, many realists would likely argue that anti-Americanism is largely epiphenomenal to power, and its apparent effects on policy are therefore spurious. That is, even if we were to observe an association between anti-Americanism and policy behavior, this would likely be a result of omitted variable bias, with the real causal variable – relative power – causing both policy behavior and resentment towards the current global hegemon (the United States). The neo-realist perspective, then, provides four hypotheses. First, neo-realism suggests that we are unlikely to find any systematic correlation between antiAmericanism and policy behavior, especially in the realm of international security (the materialism hypothesis). Moreover, because anti-Americanism may emerge out of the resentment that is borne of powerlessness, any bivariate relationship we do observe should disappear if we control for relative power in the international system (the material primacy hypothesis). The neo-realist perspective also provides us with expectations about when anti-Americanism is most likely to influence policy – specifically, when decision-makers do not feel constrained by their material circumstances. On the one  59  As Katzenstein and Keohane (2007a) put it: “A long tradition of political realism emphasizes that in the effort to be successful leaders need to subordinate their emotional reactions and ideological predispositions to strategic calculations. Such calculations should preclude the direct translation of anti-American views into policy” (287).  133  hand, decision-makers should feel a greater freedom to act based on personal attitudes or domestic incentives when their state’s power in the international system is greater, since they are then more able to chart an independent foreign policy (the power hypothesis). On the other hand, states are likely to feel more constrained the more dependent they are on their relationship with the United States for material benefits like trade, security, or aid, and should therefore be most influenced by anti-Americanism in conditions of low dependence (the dependence hypothesis). These two variables measure different aspects of material constraint, and while they are unlikely to be orthogonal, they are by no means fully collinear (think of Canada’s trade dependence with the U.S. compared to, say, Spain’s). 5.2.3 Structural Constructivism  While materialist theories like structural realism are likely to be skeptical of the policy relevance of anti-Americanism, ideational IR theories would not necessarily accommodate anti-Americanism as an independent causal variable. From a structural constructivist perspective (Wendt 1999), for example, aggregate public attitudes towards the United States can be viewed as a reflection of a state’s social identity – or more specifically, of the extent to which a state shares a common social identity with the United States. Again, the elite/public distinction is not particularly important here because social identity would be an attribute of states, not of subnational actors. The degree to which a state shares a common social identity with the United States, moreover, would be indicative of the extent to which these two states share common interests, a situation that facilitates cooperation. Thus, we would expect more pro-American states to support U.S. policy initiatives more frequently because they share a common identity  134  and, therefore, common interests. That said, structural constructivism would tend to view public attitudes towards the United States simply as an indicator of shared social identity with the United States, and not as an independent concept or causal variable. This perspective leads to two hypotheses. One the one hand, structural constructivism would certainly expect to find a bivariate association between public attitudes towards the United States and states’ participation in the Iraq War (the shared identity hypothesis). On the other hand, because anti-American attitudes do vary with events - suggesting that they are “noisy” indicators of shared identity - it would also expect their effect to disappear or at least be greatly attenuated when more stable measures of shared identity are controlled for (the underlying identity hypothesis). Unfortunately, the structural constructivist perspective does not provide clear expectations about when anti-Americanism should matter. On the one hand, we might expect to observe the greatest impact of anti-American attitudes on policy among states that have a lower degree of shared identity with the United States, because in this case anti-American public attitudes would be a useful measure of the residual variance in shared identity that is not captured by the more stable measures of state identity. On the other hand, we might expect residual variation in general attitudes towards the United States to be more salient to policy among states that already share extensive social and regime similarities with the United States (Canada, for example) than those that have very different domestic attributes (North Korea).60  60  As I discuss earlier, structural constructivism would see anti-Americanism as indicative of a state’s deep social identity. Other constructivist approaches would not necessarily discount this hypothesis, but would also likely focus on more proximate instantiations of the phenomenon, such as its impact of country images  135  5.2.4 Domestic-Politics Approaches and Anti-Americanism  Up to now, I have considered only theories that view anti-Americanism as an attribute of unitary states (both due to the theories’ ontological assumptions and because more fine-grained distinctions would not actually change the theories’ predictions about when anti-Americanism would be likely to influence policy). A number of approaches in IR, however, do consider how domestic politics influences foreign policy behavior, and in this context, the question of what is being measured by public opinion surveys becomes an important one. Specifically, it is obvious that public opinion surveys are in fact measuring mass public attitudes towards the United States61, but is mass public opinion likely to be an independent variable that influences foreign policy decisionmakers, or does mass public opinion simply reflect the preferences of national elites? There is ample evidence in the literature on public opinion to suggest that mass public preferences on foreign policy are heavily influenced by elite discourse (Zaller 1992, for example). Taking this finding further, we might speculate that elites may also be able to influence more general attitudes about external actors (countries or international organizations, for example) in addition to attitudes about specific foreign policies. If this is the case, finding an association between public attitudes towards the United States and policy outcomes would simply be telling us that policy outcomes follow elite preferences - a rather uninteresting finding.62 One observable implication of this perspective is that if we were to directly control for elite preferences, we should expect the observed  on persuasion and socialization (Johnston 2001; Payne 2001). The data used in this study, however, are not fine-grained enough to test these more nuanced mechanisms. 61 Within the limitations of sampling constraints and instrument design, of course. 62 Although, as argued above, this would be contrary to structural realist expectations.  136  relationship between public anti-Americanism and policy to disappear (the elite preferences hypothesis). There is a large body of research, however, that has detected the independent effects of public opinion on policy outcomes (see Wlezien and Soroka 2007 for a review). Moreover, a number of approaches in IR, such as Moravcsik’s (1997) “ideational liberalism” or Risse-Kappen’s (1996) “liberal constructivism” can accommodate the possibility of public anti-Americanism influencing policy outcomes. Indeed, Risse-Kappen (1991) is one of the few articles in IR to theorize the impact of public opinion on foreign policy in a non-U.S. context. This same literature, however, also offers a number of reasons to be skeptical about the likelihood of public opinion influencing foreign policy behavior. A key expectation is that the impact of public opinion on policy tends to be moderated by an issue’s electoral salience, and foreign policy – with some notable exceptions such as major military engagements – tends to be of lower salience to voters than domestic “bread and butter” issues.63 In the context of the Iraq War, then, this discussion provides the following testable hypothesis about when public anti-Americanism is likely to matter: if public anti-Americanism is in fact an independent variable, we would expect its influence on policy outcomes to be greatest when the policy in question has high domestic salience (the issue salience hypothesis). All the hypotheses described in this section are presented in Figure 5.1. They are categorized by the type of statistical technique that is necessary to test for them and by the theoretical perspective from which they are derived. The materialism and shared  63  That said, the concept of latent opinion in political behavior (Zaller 2003) suggests that political leaders may sometimes act based on anticipated public opinion about their policies, rather than on existing opinion. In this context, politicians might anticipate that a specific issue could become politically salient, and therefore allow the content of present public opinion on the issue to shape their decision-making.  137  identity hypotheses, for example, simply posit the existence or non-existence of a bivariate relationship between public anti-Americanism and policy outcomes,  FIGURE 5.1: HYPOTHESES AND ASSOCIATED EMPIRICAL TESTS TEST  NEO-REALISM  STRUCTURAL CONSTRUCTIVISM  DOMESTIC POLITICS  BIVARIATE  H(MATERIALISM)  H(SHARED IDENTITY)  N/A  CONTROLS  H(MATERIAL PRIMACY)  H(UNDERLYING IDENTITY)  H(ELITE PREFERENCES)  H(POWER)  N/A  H(ISSUE SALIENCE)  INTERACTIONS  H(DEPENDENCE)  whereas the power and issue salience hypotheses propose that the effects for public antiAmericanism are conditional on state power and issue salience, respectively. The next two sections present empirical analyses of the link between public anti-Americanism and: 1) decisions to join the U.S.-led Multi-National Force in Iraq, and 2) decisions to withdraw from the operation. All the hypotheses in Figure 5,1 are tested in at least one context. 5.3 EMPIRICAL TESTING – PARTICIPATION IN THE IRAQ WAR As I note earlier, my decision to focus on the Iraq War as an empirical case follows from Katzenstein and Keohane (2007a, 290-293), who deem this to be one of the “most likely” cases in which to detect the influence of anti-Americanism on policy outcomes. As they put it, “[i]t seems that pervasive anti-Americanism should most strongly affect responses to highly controversial American policies, such as U.S. policy  138  in Iraq” (290). They attempt an initial test of this hypothesis by cross-tabulating aggregate data on favorability towards the U.S. drawn from the 2002 Pew Global Attitudes Project survey and data on coalition participation drawn from a U.S. Senate resolution that listed members of the “coalition of the willing” in March of 2003. They find that there is no clear evidence for a relationship between aggregate attitudes and policy. I build on this analysis by, on the one hand, using a somewhat expanded dataset, and, more importantly, using regression analysis instead of cross-tabulation. Crucially, this allows me to control for the influence of other factors on participation and withdrawal, and estimate the marginal effect of anti-American opinion on policy decisions. To the extent that different countries are likely to have had very different underlying probabilities of participating in the operation, techniques that do not control for other influences on participation would be unlikely to detect the influence of antiAmericanism. Moreover, regression analysis makes it possible to investigate systematically the more nuanced hypotheses I lay out in Part I about when antiAmericanism is likely to matter, as well as allowing me to estimate how much it mattered.64 Contrary to Katzenstein and Keohane’s (2007a) analysis, I find substantial effects of anti-Americanism on the probability of participation, although only for some countries and under certain conditions.  64  Katzenstein and Keohane (2007a) do, of course, address these more nuanced questions about antiAmericanism, but their primary mode of analysis is comparison across policies or critical junctures. The advantage of the method in this paper is that by using the same policy decision, it allows me to control for potential unobservable confounds that might vary across policy issues and domains – this allows me to focus on the marginal effect of the moderating variables I investigate. Of course, a more comprehensive study would apply the insights gained in this paper to a broader, cross-issue analysis. That objective, however, is left to future research.  139  My empirical analysis has two parts. First, I use data on whether countries participated either in the invasion or in the immediate post-invasion mission in Iraq (which I define as a contribution made in 2003). I use cross-sectional country-level data for 2003 to test for the influence of attitudes towards the United States on country decisions to contribute to the U.S.-led force in Iraq. Second, because anti-Americanism might influence the degree of commitment to a deployment in Iraq, I test for the impact of attitudes towards the United States on country decisions to withdraw from the MultiNational Force in Iraq. In both cases, I use a series of regression models to test the hypotheses listed in Figure 5.1. 5.3.1 Participation in the Invasion and Post-War Stabilization  For this part of the empirical analysis, I begin with a cross-sectional dataset containing observations for 166 countries. The core of the data is drawn from a dataset on national contributions to multilateral military operations compiled by Katharina Coleman at the University of British Columbia, using annual country reports by the Military Balance on states’ foreign military deployments.65 Using the Military Balance data, I created a dummy variable for each country, indicating whether it participated in Operation Iraqi Freedom in 2003 (during the invasion or the immediate post-invasion period). Any type of participation – including military observers, aircraft, or any number of troops (though not civilian police) – was coded as an instance of participation in the operation. This resulted in 30 contributing countries for 2003, as shown in Table 5.1. I cross-checked this participation data online, using the Wikipedia entry on the 65  I thank Professor Coleman for granting me use of this data. Professor Coleman’s dataset also contained a number of useful control variables, including defense budget, number of active forces, GDP, Freedom House scores, and voting affinity at the U.N. The availability of this data made this project much more manageable than it would otherwise have been.  140  Multinational Force (MNF) in Iraq.66 Among countries for which I have pre-war U.S. attitudes data (which I describe below), the only country for which participation data is inconsistent between sources is South Korea.67 Consequently, I exclude this case from the analyses below and in the duration analyses of withdrawal in the second empirical analysis. This leaves 29 contributors for 2003. Table 5.1: Countries Contributing to Operation Iraqi Freedom in 2003 Albania Fiji Macedonia Portugal Australia Honduras Mongolia Romania Bulgaria Hungary Netherlands Slovakia Czech Republic Italy New Zealand Spain Denmark Japan Nicaragua Thailand Dominican Republic Latvia Norway Ukraine El Salvador Lithuania Philippines United Kingdom Estonia Poland 5.3.1.1 INDEPENDENT VARIABLES The key independent variable is, of course, anti-Americanism. As Katzenstein and Keohane (2007a) note, in any analysis of anti-Americanism it is crucial to minimize the possibility that the measure of anti-Americanism is actually capturing public attitudes towards the war, rather than underlying, general attitudes towards the United States. Their solution is “ to examine public opinion data prior to the war, since [they] wish to test the inference that general public sentiment shaped policy toward the Iraq war, rather than vice-versa” (2007a, 290, footnote #54). I adopt the same approach in this chapter, using, as they do, favorability data from the 2002 Pew Global Attitudes Project (PGAP) survey.68 I collected additional data from two other surveys, however: the 2002  66  http://en.wikipedia.org/wiki/Multinational_Force_Iraq, accessed May 15, 2009. The Military Balance shows South Korea participating as of 2006, whereas news articles linked to by the Wikipedia entry show that South Korea had deployed some personnel in 2003. 68 Data is available at: http://pewglobal.org/. The Pew Global Attitudes Project is not responsible for the use and interpretation of its data made in this article. 67  141  Latinobarometro69 and the Fall 2002 Eurobarometer (EB58). The questions asked and countries surveyed are shown in Appendix A. It bears noting that the questions for the Eurobarometer survey are quite different than the standard favorability item used in both the PGAP and Latinobarometro surveys. It asks respondents to rate (positive, negative, or neutral) the role the U.S. plays in five different policy areas. In order to estimate comparable aggregate favorability scores from this data, I create average “positivity” Table 5.2: Countries and Pre-war Aggregate U.S. Favorability Country Angola Argentina Austria Bangladesh Belgium Bolivia Brazil Bulgaria Canada Chile Colombia Costa Rica Czech Republic Denmark  Favorability 54.5 34.2 43.3 52.0 53.5 56.7 50.9 71.0 73.0 69.3 79.2 84.0  Ecuador Egypt El Salvador Finland France Germany Ghana  87.1 5.8 85.0 53.5 62.9 58.8 83.2  Country Greece Guatemala Honduras India Indonesia Ireland Italy Ivory Coast Japan Jordan Kenya Lebanon  Favorability 33.5 81.8 80.2 63.2 60.8 67.1 69.7 84.6 71.2 24.4 80.1 35.3  70.6 Luxembourg 55.0 Mali Mexico Netherlands Nicaragua Nigeria Pakistan Panama Paraguay  Country Peru Philippines Poland Portugal Russia Senegal Slovakia South Africa Spain Sweden Tanzania Turkey  Favorability 71.3 91.1 77.8 52.8 59.3 64.7 59.6 64.9 50.0 48.6 53.3 30.9  49.6 Uganda 73.5 Ukraine United 64.6 Kingdom 48.9 Uruguay 84.4 Uzbekistan 77.2 Venezuela 12.9 Vietnam 89.1 69.7  74.4 82.0 75.1 48.8 85.3 81.7 67.9  Note: 2003 Participants in Iraq MNF are italicized  69  I obtained the original individual-level data and cross-tabulated responses to the U.S. favorability question described in Appendix A with the country variable.  142  scores for each country using all five items, and then predict favorability for the Fall 2002 Eurobarometer countries using coefficient estimates obtained by regressing favorability scores on positivity scores for countries for which both are available. Details of this procedure are provided in Appendix B. Overall, this yields 61 countries for which I have favorability data (excluding South Korea) of which 16 contributed to the Multi-National Force in Iraq in 2003. Table 5.2 shows each country and the favorability score used in this analysis. In addition to this key independent variable, I use a number of control variables to test the hypotheses listed in Figure 5.1. 5.3.1.1.1 Realist Controls  As I argue in Part I, the neo-realist perspective would view any observed bivariate relationship between anti-Americanism and policy as a spurious correlation caused by the fact that a state’s relative power in the international system determines both its foreign policy behavior and its level of anti-Americanism. To account for this possibility, I use three control variables. First, I include the Correlates of War (COW) Project’s Composite Index of National Capability (CINC) score, which takes into account population, industrial production, and military capabilities.70 This variable measures each state’s overall power in the international system. I also include two variables that measure countries’ relative dependence on the United States, which should control for the relative power of each country in the specific context of its relationship with the United States. I include a measure of trade dependence, calculated as total bilateral trade (imports + exports) in 2002 as a percentage of the state’s GDP, and a measure of aid  70  I use 2001 data for each country (the most recent available). This data is available online at: http://www.correlatesofwar.org/COW2%20Data/Capabilities/nmc3-02.htm.  143  dependence, measured as the total amount of U.S. bilateral foreign assistance to a given country in 2002 as a percentage of that country’s GDP.71 5.3.1.1.2 Constructivist Controls  To test the structural constructivist argument that anti-Americanism is simply a proxy measure of deeper state identity, I include four variables meant to measure both the general social identity of a state in the international system and the degree to which it shares a social identity with the United States. First, I use the percentage of a country’s population that is Muslim.72 The United States being a secular and/or Christian nation, we would expect higher percentages to indicate a lower degree of shared identity with the U.S. Similarly, more democratic nations would presumably feel more in common with the United States than less democratic countries, so I include the Freedom House score for each country in the sample. Because the Freedom House score varies inversely with the degree of freedom, a greater Freedom House score should also be associated with a lower degree of shared identity with the United States.73 Finally, I include two dummy variables indicating whether a country was “Western” (i.e., situated in the “EuroAtlantic” area)74 and whether it had a formal mutual defense alliance with the United States.75  71  Trade data is obtained from the International Trade Administration at the U.S. Department of Commerce. It is available online at: http://tse.export.gov. Bilateral foreign assistance data is drawn from the U.S. Department of States’ annual “Voting Practices in the United Nations” report, which is available online at: http://www.state.gov/plio/conrpt/vtgprac. GDP data is from the Military Balance country reports. 72 This data is obtained from the Quality of Governance Project cross-sectional dataset, which is available online at: http://www.qog.pol.gu.se/data/data.htm. 73 This data was already included in Professor Coleman’s dataset, and is also available online at: http://www.freedomhouse.org/template.cfm?page=439. 74 The following countries are coded as “Western” in the N=61 dataset: Austria, Belgium, Bulgaria, Canada, Czech Republic, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, Netherlands, Poland, Portugal, Slovakia, Spain, Sweden, the United Kingdom. 75 Alliance data is drawn from the Correlates of War Project. Data is available at: http://www.correlatesofwar.org/COW2_Data/Alliance/alliance.htm. Note that this variable might also be  144  5.3.1.1.3 Elite Preferences  Finally, to control for the possibility that our measures of public attitudes towards the United States are simply proxies for elite preferences and are not actually measuring public anti-Americanism (the elite preferences hypothesis), I use an “affinity” score that captures how often each country votes with the United States in the U.N. General Assembly (coded as a percentage of all UN votes cast).76 Because these scores are based on votes in a variety of issue-areas over an entire year, this should provide an adequate measure of the extent to which the general policy preferences of a country’s elites are aligned with those of the United States. 5.3.1.2 DATA ANALYSIS In order to begin testing for the effect of public attitudes towards the United States on countries’ decisions to join the U.S.-led coalition in Iraq, I estimated a bivariate logistic regression regressing the participation variable on the aggregate favorability variable. Table 5.3 shows the results (Model 1). The coefficient for favorability is positive, as would be expected (with greater favorability being associated with a higher probability of participation in the U.S.-led mission in Iraq). The relationship is not statistically significant at conventional levels, however, though the p-value approaches significance at 0.077. While the relationship is not clearly significant, it is not clearly absent, either, so the evidence of the bivariate model does not speak clearly to either the materialism nor  plausibly grouped with realist controls. Regardless, it does not have a conditioning effect on either participation or withdrawal, as I describe below. 76 These scores were calculated based on Erik Voeten and Adis Merdzanovic’s UN General Assembly Voting Data, available at: http://dvn.iq.harvard.edu/dvn/dv/Voeten/faces/study/StudyPage.xhtml?studyId=38311andstudyListingInde x=0_dee53f12c760141b21c251525332. Details on the procedure used to calculate them are available on request.  145  shared identity hypotheses. In Models 2-4 (also shown in Table 5.3), I run the same regression model, but include the realist, constructivist, and elite preferences control Table 5.3: Effect of U.S. Favorability on 2003 Participation in the Iraq MNF  US Favorability  Model 1  Model 2  Model 3  Model 4  Model 5  0.0346 (1.77)  0.101** (2.62)  0.0339 (1.53)  0.0973* (2.44)  0.112* (2.28)  US Affinity  0.180** (3.26)  Power (COW)  0.268 (1.38) 7.616 (0.34)  -10.04 (-0.19)  Bilateral Trade (% GDP)  0.00764 (0.39)  -0.00923 (-0.36)  Aid Dependence (% GDP)  -0.52 (-0.68)  1.013 (0.98)  Western Dummy  2.925* (2.07)  -1.5 (-0.39)  US Ally  0.544 (0.56)  0.117 (0.11)  Muslim (% Population)  -0.106 (-0.85)  -0.133 (-0.85)  Freedom House Score  -0.0145 (-0.06)  0.00535 (0.02)  Constant N  -3.311* (-2.41)  -12.34** (-3.19)  -3.282* (-2.15)  -8.868* (-2.50)  -14.55** (-2.59)  61  61  61  61  61  t statistics in parentheses * p<0.05, ** p<0.01, *** p<0.001  146  variables, separately. Model 2 controls for elite preferences towards cooperation with the United States by including the UN voting affinity variable. When this variable is controlled for, the magnitude of the coefficient for US favorability increases, and becomes statistically significant at the 0.01 level.77 It is also interesting to note that the affinity variable is also a significant predictor of participation in the operation (greater affinity increases the probability of participation). This is strong evidence against the elite preferences hypothesis, since controlling for elite preferences not only fails to make the relationship between favorability and policy disappear – it actually reveals a stronger relationship, which suggests that the effect of aggregate favorability on participation is largely independent of elite preferences. Model 3 controls for the three realist variables described above, which measure systemic power and power relative to the United States. Including the realist controls does not appear to change the estimated relationship between favorability and participation – the favorability coefficient remains about the same as in the bivariate model, as does its t-statistic. Moreover, none of the power variables appear to influence the probability of participation. This suggests that the effects of power and antiAmericanism on participation are also largely independent, providing little evidence in favor of the material primacy hypothesis. Finally, in Model 4 I include the four constructivist shared identity variables described above. While the “Western” dummy is the only one that appears to have an effect on participation (Western states are more likely to participate), the favorability coefficient’s magnitude increases relative to the bivariate model, and is statistically significant at the 0.05 level. This is not good evidence in support of the underlying identity hypothesis, since controlling for deep state 77  More rigorous analysis of the magnitude of predicted effects follows in the next section.  147  identity actually strengthens the estimated relationship between anti-Americanism and policy. In Model 5, which includes all the controls together, the US favorability variable remains statistically significant at the .05 level, while none of the other independent variables achieves significance. Overall, this series of analyses provides consistent evidence that antiAmericanism was associated with decisions to participate in the U.S.-led mission in Iraq. Given Katzenstein and Keohane’s (2007a) skepticism, it is particularly noteworthy that this relationship remains even when controlling for all three sets of variables in Model 5. Moreover, none of the hypotheses presented in the first two rows of Figure 5.1 are supported by the results. Overall, this suggests that anti-Americanism, at least as measured by aggregate public attitudes towards the United States, had an independent effect on policy outcomes – in short, it was not simply epiphenomenal to state power, identity, or elite preferences.78 5.3.1.2.1 How Much Did It Matter?  While finding that US favorability is a significant predictor of participation in the mission in Iraq is interesting, perhaps the more interesting question is: how much did it matter? That is, what was the magnitude of the estimated effect on the probability of participation. I attempt to answer this question by estimating predicted changes in the probability of participation using the Clarify software package for STATA (King et al. 2000). Using Model 5, I estimated the predicted change in the probability of participation when reducing each country’s U.S. favorability score by one standard  78  As a robustness test, I ran the same models excluding the favorability data estimated from the Eurobarometer data (N=51) and using aggregate unfavorability (N=50), which could be argued to be a more valid (if less widely available) measure of anti-Americanism. The pattern of results was largely the same as that presented in Table 3.  148  deviation (18.6 percentage points), while setting all other independent variables at the actual values for each country in the dataset. The mean predicted change in probability for all countries was -0.17, which is a substantial effect for a single variable. That said, the effects for individual countries (show in Table 5.4) varied widely, ranging from a negligible -0.008 (Argentina) to a more impressive -0.405 (the Czech Republic).  Table 5.4: Effects of a Reduction in Aggregate U.S. Favorability on the Predicted Probability of Participation Country Czech Republic Italy El Salvador France Canada Germany Nicaragua Bulgaria Denmark Ukraine Ecuador Japan Slovakia Belgium Ireland Luxembourg Costa Rica Portugal Panama Honduras Finland Guatemala Netherlands Uganda Bolivia Poland Philippines Colombia Spain  Change in Prob. -0.405 -0.401 -0.38 -0.37 -0.37 -0.361 -0.359 -0.328 -0.326 -0.316 -0.310 -0.305 -0.298 -0.292 -0.287 -0.286 -0.282 -0.269 -0.256 -0.250 -0.241 -0.219 -0.210 -0.202 -0.200 -0.197 -0.196 -0.183 -0.164  Baseline Prob. 0.751 0.765 0.544 0.792 0.740 0.592 0.681 0.583 0.483 0.498 0.415 0.551 0.422 0.433 0.420 0.435 0.403 0.388 0.351 0.428 0.346 0.299 0.310 0.292 0.323 0.912 0.264 0.249 0.241  149  Table 5.4 [Continued] Country Peru Sweden Venezuela Chile Paraguay United Kingdom Ghana Austria Russia Kenya Angola Ivory Coast Mexico India Greece Vietnam Uzbekistan South Africa Nigeria Mali Senegal Turkey Indonesia Tanzania Jordan Uruguay Brazil Bangladesh Lebanon Argentina  Change in Prob. -0.143 -0.137 -0.125 -0.122 -0.118 -0.106 -0.097 -0.071 -0.069 -0.060 -0.057 -0.051 -0.050 -0.044 -0.043 -0.037 -0.036 -0.033 -0.032 -0.027 -0.025 -0.022 -0.022 -0.021 -0.020 -0.019 -0.018 -0.017 -0.015 -0.008  Baseline Prob. 0.194 0.199 0.169 0.173 0.167 0.947 0.152 0.109 0.137 0.084 0.103 0.116 0.083 0.098 0.080 0.067 0.216 0.052 0.123 0.167 0.153 0.160 0.090 0.058 0.129 0.039 0.039 0.119 0.051 0.026  Note: Changes in predicted probabilities of participation are based on a one standard deviation reduction in pre-war aggregate U.S. favorability (18.6 percentage points). Overall, then, these basic models point to two principal conclusions. First, the evidence is consistent with the view that aggregate public favorability towards the U.S. influenced countries’ decisions to participate in the war in Iraq. Moreover, the results suggest that this influence was not negligible – across a range of countries, the magnitude of the effect  150  Table 5.5: Full 2003 Model with Realist Variable Interactions Model 5  Model 6  Model 7  Model 8  US Favorability  0.112* (2.28)  0.0695 (1.22)  0.0566 (1.00)  0.086 (1.71)  US Affinity  0.268 (1.38)  0.147 (0.71)  0.322 (1.35)  0.333 (1.62)  Power (COW)  -10.04 (-0.19)  -749.8 (-1.53)  24.45 (0.40)  -16.67 (-0.31)  -0.00923 (-0.36)  0.00207 (0.07)  -1.313* (-2.25)  -0.0176 (-0.66)  Aid Dependence (% GDP)  1.013 (0.98)  1.356 (1.28)  1.321 (0.74)  -28.66 (-1.09)  Western Dummy  -1.5 (-0.39)  1.287 (0.30)  0.232 (0.05)  -2.664 (-0.67)  US Ally  0.117 (0.11)  0.857 (0.73)  -0.523 (-0.43)  -0.122 (-0.11)  Muslim (% Population)  -0.133 (-0.85)  -0.0702 (-0.51)  -0.241 (-1.28)  -0.145 (-0.91)  Freedom House Score  0.00535 (0.02)  -0.00549 (-0.02)  0.233 (0.64)  0.0602 (0.18)  Trade Dependence (% GDP)  Favorability X Power  11.15 (1.54)  Favorability X Trade Dep.  0.0167* (2.26)  Favorability X Aid Dep.  0.389 (1.16)  Constant  N Joint Significance Tests  Chi-squared p-value  -14.55** (-2.59)  -10.73 (-1.75)  -13.71* (-2.06)  -13.97* (-2.46)  61  61  61  61  n/a n/a  6.41 0.0405  8.00 0.0183  4.84 0.0891  t statistics in parentheses * p<0.05, ** p<0.01, *** p<0.001  151  was reasonably large. The pattern of predicted effects presented in Table 5.4, however, suggests that the degree to which anti-Americanism influences policy may substantially depend on context. The next section, then, addresses the question of when public antiAmericanism is most likely to influence policy. 5.3.1.3 TESTING FOR CONTEXTUAL EFFECTS Three of the hypotheses presented in Figure 5.1 suggest that the impact of antiAmericanism may vary systematically either with a state’s power or with the domestic salience of the policy in question. I test the issue salience hypothesis in the next section, which deals with decisions to withdraw from the mission, both because comparable measures of salience are more easily obtained in that context, and because we would expect greater variation in salience in subsequent years (the Iraq War was likely highly salient in most countries around the world in 2003, as suggested by the large-scale demonstrations before the war). The power and dependence hypotheses, however, can be tested in this context, by interacting the U.S. favorability variable with the power, trade dependence, and aid dependence variables, respectively. According to the power hypothesis, more powerful states should be most influenced by public favorability towards the U.S. when making policy decisions, leading us to expect a positive coefficient for the interactive term. The dependence hypothesis suggests that states that are more dependent on the United States should be less influenced by anti-Americanism, leading us to expect negative coefficients for the interactive terms. Table 5.5 presents results for versions of Model 5 that include the relevant interactive terms.  152  The first thing to note is that all the interactive terms are positive. That is, while the effect of favorability on participation appears to be greater for more powerful states, it also appears to be greater for more dependent states. When testing the joint significance of the favorability term and the interactive terms79, I find that the power and trade dependence interactions are both significant at the 0.05 level, while the aid interaction is significant at the 0.10 level. These results provide mixed evidence in support of the realist hypotheses about when anti-Americanism should matter. On the one hand, it appears that greater systemic power does reduce constraints on state behavior, and allows their foreign policy decisions to be influenced by their general attitudes towards the United States. On the other hand, more dependent states – that is, those that have lower relative power and are therefore presumed to be more constrained – also appear to be more influenced by their attitudes towards the United States. Overall, then, the evidence is mixed, but does appear to moderately support a structural realist interpretation of when anti-Americanism should matter (particularly to the extent that overall systemic power is a true independent variable, whereas dependence on the United States and general favorability towards the United States may be co-determined). 5.3.1.4 DISCUSSION The findings presented in this empirical section lead to several conclusions. While the weak bivariate relationship between aggregate public attitudes towards the United States and participation in the war in Iraq is consistent with Katzenstein and Keohane’s (2007a) skepticism, a strong and statistically significant effect emerges when other potential influences on participation are controlled for. Moreover, the empirical results clearly suggest that public anti-Americanism is an independent phenomenon, and 79  As recommended in Brambor et al. (2006).  153  is not simply a function of state identity, state power, or elite preferences. That said, while the estimated effect of anti-Americanism on participation is of sizeable magnitude on average, the degree to which it influences state policy appears to vary substantially across countries. Specifically, the realist hypothesis that more powerful states are most likely to be influenced by this type of variable is supported by the evidence. A final conclusion is that anti-Americanism as measured by mass public favorability towards the United States is indeed an attribute of public opinion and not simply an indirect measure of elite preferences. The fact that controlling for UN voting affinity actually increased the estimated effect of aggregate favorability on policy suggests that it was public attitudes towards the United States that influenced policy. As argued above, an implication of this conclusion is that the effect of public antiAmericanism on policy should vary with the domestic salience of the policy in question. The next section tests this hypothesis using data on withdrawal from the Iraq operation. 5.3.2 The Influence of Anti-Americanism on Withdrawal from Iraq  While the evidence presented in the previous section does suggest that public attitudes towards the United States influenced state decisions to participate in the war in Iraq, the question remains as to whether this is a generalizable finding. After all, the spring of 2003 was a highly charged and unusual moment in the global political context that prompted an unusually high level of global attention on a highly controversial U.S. initiative. One might speculate that anti-Americanism has little relevance to everyday global politics – that is, for “less likely” cases. While this chapter cannot fully address this question, it is possible to test for the influence of anti-Americanism on policy in a more “normal” political context by examining whether underlying pre-war attitudes  154  towards the United States influenced states’ decisions to withdraw from the mission. These decisions would arguably have taken place after the international focus on the operation had dissipated. 5.3.2.1 DATA AND VARIABLES In order to test for the influence of anti-Americanism on states’ decisions to withdraw from the operation, I created a dataset using a country-month unit of analysis, in which each of the sixteen 2003 participants used in the previous section’s analysis was coded as a 0 for each month in which they did not withdraw and a 1 for the month in which they withdrew, after which I ceased observing the country in question. As Beck, Katz and Tucker (1998) argue, this type of binary time-series cross section (BTSCS) data “are identical to grouped duration data” (1260), and can be analyzed as a survival model using logistic regression if steps are taken to account for temporal dependence in the data.80 Using this framework, then, I am able to test for the influence of pre-war attitudes towards the United States on the probability of withdrawal, while accounting for the fact that the probability of withdrawal (the hazard rate) is also likely to be a function of the length of time since the country’s forces were deployed. Withdrawal dates for the 16 countries are listed in Appendix C.81 This yields a dataset with 617 country-month observations, of which 16 are withdrawals.  80  Standard survival analysis techniques, like the Cox proportional-hazards model, are appropriate when temporal data are measured in a continuous manner, but is not appropriate in analyzing grouped or discrete duration data. 81 I obtained the month of withdrawal for each country in the dataset online. I started with the previously mentioned Wikipedia article on the Multinational Force in Iraq. I then followed up the linked news articles to confirm that the withdrawal month was in fact accurate. Where these links were inactive, I found alternative news reports indicating withdrawal for the country in question. Note that countries’ contributions to the Iraq mission varied substantially over time, and, in many cases, the withdrawal of the main force preceded withdrawal of the final troops. For the purpose of this research, I treat withdrawal as withdrawal of all the remaining forces (no matter the number or type) affiliated with the Multinational Force (MNF-I). Moreover, because the exact month of initial deployment proved to be more difficult to  155  Table 5.6: Effect of Favorability on Decisions to Withdraw from Iraq Model 9  Model 10  Model 11  Model 12  -0.0247 (-0.93)  -0.194** (-3.20)  -0.193** (-3.18)  -0.185** (-2.87)  -0.000176** (-2.86)  -0.000176** (-2.86)  -0.000176** (-2.76)  Log (Active Forces)  3.364** (3.1)  3.374** (3.12)  3.390** (3.03)  U.N. Affinity with U.S.  -0.149 (-1.83)  -0.148 (-1.79)  -0.149 (-1.70)  U.S. Ally  2.955 (1.94)  2.931 (1.94)  2.817 (1.88)  Aid Dependence  4.447** (3.04)  4.386** (2.99)  4.547** (2.93)  Trade Dependence  0.0898** (2.65)  0.0884** (2.6)  0.0898* (2.57)  -0.286 (-0.28)  2.632 (0.56)  U.S. Favorability  Defense Budget  3-Month Casualties Dummy  Casualties X Favorability  Deployment Months  -0.0451 (-0.62) -0.0152 (-0.10)  0.5 (1.9)  0.491 (1.86)  0.559 (1.87)  Spline 1  0.000384 (0.41)  0.00157 (1.31)  0.00156 (1.3)  0.0018 (1.4)  Spline 2  -0.000138 (-0.15)  -0.000306 (-0.28)  -0.000297 (-0.27)  -0.000474 (-0.41)  Spline 3  -0.000211 (-0.44)  -0.000703 (-1.10)  -0.000709 (-1.11)  -0.000643 (-1.00)  Constant  -1.744 (-0.90)  -30.24** (-2.79)  -30.21** (-2.80)  -31.28** (-2.80)  617  617  N  617  t statistics in parentheses * p<0.05, ** p<0.01, *** p<0.001  617 Joint Significance  Chi-squared p-value  9.46 0.0088  ascertain than the month of withdrawal, I begin observing all countries in January of 2004, at which point all countries in the dataset had made an initial deployment and were therefore “at risk”. An added advantage of Beck et al.’s (1998) method is that it “allows left censoring so long as all observations are equally left censored,” in which case “left censoring causes literally no problems” (Beck et al. 1998, 1272).  156  To conduct this analysis, I ran a logistic regression of the withdrawal dummy on the same aggregate US favorability variable used in the previous section. As Beck et al. (1998) recommend, I use temporal splines82 in order to model the temporal dependence of the data (the probability of withdrawal is likely to increase with the length of deployment).83 The bivariate model is shown as Model 9 in Table 5.6. It shows that while the effect of an increase in US favorability was negative, as expected (greater aggregate favorability reduces the probability of withdrawal from the U.S.-led mission), the effect was not statistically significant. Once again, however, it may be that when I control for other variables, the influence of US favorability will become clearer. I therefore ran a series of other models, also presented in Table 5.6. As control variables, I include, first, the defense budget of each country in a given year and the natural logarithm of its active forces – a state’s overall military capabilities should influence the duration for which it is able to sustain a deployment. These variables are both obtained from Professor Coleman’s dataset, and were originally listed in the Military Balance country entries. I also included several variables measuring each country’s overall strategic alignment with the United States (the UN affinity variable, and the US alliance dummy) and the extent to which each country is dependent on the United States (aid and trade dependence).84 Model 10, which includes these control variables, finds that pre-war U.S. favorability does indeed have a statistically significant effect (p is 0.003) on the probability of withdrawal. As expected, greater aggregate favorability decreases the probability of withdrawal at any given time.  82  I used Richard Tucker’s btscs.ado package for STATA to create 3 temporal splines. Joint significance tests for the deployment spell variable and the three temporal spline showed p-values of 0.0001 or smaller for all models reported in Table 6. 84 Note that all these control variables are measured at the country-year level. 83  157  Put another way, we would expect that more anti-American countries would, ceteris paribus, be more likely to withdraw from the mission at any given time than more proAmerican countries. This withdrawal model does not simply provide evidence for the impact of antiAmericanism on policy in a less unique context than the 2003 invasion of Iraq – it also allows me to test the final hypothesis in Figure 5.1. The issue salience hypothesis suggests that the impact of aggregate public attitudes towards the United States on policy should be greatest when the policy in question is salient to the mass public. There is no standardized method of measuring salience, or, indeed, much empirical research on how salience moderates the impact of public opinion on policy (Wlezien and Soroka 2007, 807-808), especially when data is needed across a wide range of countries. I therefore opt to use a reasonable proxy for issue salience. I added data on military fatalities among coalition forces in Iraq, creating a dummy variable that indicates whether a country suffered casualties in Iraq in the three months prior to the time-period under observation.85 The reasoning is that casualties lead to increased domestic media coverage of the war, both because the event provides natural story possibilities (interviewing the families, covering the return of the remains, for example) and because opponents of the war are empowered and are able to gain a greater voice in the public debate. Overall, this should increase the salience of the war in the country. Using this casualties variable, then, I can examine the extent to which the influence of U.S. favorability on the probability of withdrawal is increased when the country in question has sustained casualties in the preceding time-periods.  85  This data is available online at: <http://icasualties.org/Iraq/prdDetails.aspx>, accessed May 26, 2009.  158  Models 11 and 12 show the withdrawal model with, respectively, the casualties dummy variable on its own and with the dummy and its interaction with the U.S. favorability variable. As we can see, the casualties variable on its own does not have a statistically significant effect on the probability of withdrawal. Admittedly, the threemonth dummy may not be the best specification for the influence of casualties on withdrawal (a scaled variable, for example, might work better). Nonetheless, estimating the effect of casualties on withdrawal is not the goal of the paper, and, given the generally low level of casualties among U.S. allies in Iraq, a simple dummy variable is arguably a better measure for the salience-increasing effect of casualties, and its subsequent moderation of the influence of anti-Americanism on withdrawal decisions.86 Model 12 attempts to model this moderating effect using an interaction between casualties and U.S. favorability. Two findings stand out. First, the U.S. favorability variable and the interactive term are jointly significant. Second, the coefficient for the interactive term is negative, which indicates that the effect of increasing U.S. favorability on the probability of withdrawal (which is negative) is strengthened when a country has suffered casualties in the 3 months prior to the time-period under observation. That is, anti-Americanism has a greater influence on policy under high-salience conditions. This finding lends support to the issue salience hypothesis. To further investigate this finding, I used Clarify to calculate the predicted changes in the probability of withdrawal for each country in the sample caused by a 20 percentage point decline in pre-war U.S. favorability, for both the casualty and no casualty conditions.87 On average, the predicted change in probability of withdrawal was  86 87  Using a casualties dummy with a six-month lag produces similar substantive results. The temporal variables were set at their means.  159  0.22 for the no casualties condition and 0.27 in the presence of casualties. Table 5.7 presents the results, as well as a column showing the percent difference between the two casualties conditions in the predicted effects of a decline in U.S. favorability on the probability of withdrawal. For all but three countries, the predicted effects of a decrease in favorability were larger if countries were assumed to have sustained casualties in the preceding three months than if they were assumed not to have sustained casualties. Table 5.7: Effect of Reduction in Aggregate U.S. Favorability on Probabilities of Withdrawal (by casualty status)  Country Denmark Slovakia Nicaragua Czech Republic Japan Honduras Italy Portugal Bulgaria United Kingdom Netherlands Poland El Salvador Philippines Ukraine Spain  Change in Prob. (No Casualties in last 3 months) 0.056 0.042 0.125 0.043 0.036 0.276 0.175 0.378 0.223 0.007 0.514 0.162 0.059 0.530 0.463 0.434  Change in Prob. (Casualties in last 3 months) 0.18 0.125 0.237 0.075 0.058 0.401 0.249 0.535 0.297 0.009 0.584 0.173 0.062 0.527 0.451 0.385  Percent Change Due to Casualties 221 198 90 74 61 45 42 42 33 29 14 7 5 -1 -3 -11  Note: Changes in probabilities of withdrawal are calculated based on a 20 percentage point reduction in pre-war aggregate U.S. favorability.  5.4 CONCLUSION The goals of this chapter were twofold: 1) to begin meeting the burden of proof for the policy impact of anti-Americanism laid out by Katzenstein and Keohane (2007a), and, as part of that effort, 2) to explore when anti-Americanism is most likely to influence  160  policy and how big its effect is likely to be. Overall, the analyses presented above suggest that anti-Americanism did have an influence on states’ decisions to participate in the U.S.-led war in Iraq, and that this effect was of large magnitude for some countries, though dependent on contextual factors. From a theoretical perspective, none of the skeptical perspectives appears to be strongly supported – hypotheses ascribing the effect of anti-Americanism to some other, unobserved factor were unsupported by the evidence. In fact, the effect of anti-Americanism on policy became clearer once these factors were controlled for. One exception is that the realist perspective may be useful to the extent that anti-Americanism appears to be most likely to influence policy when states are relatively unconstrained by anarchy – that is, when they possess significant material capabilities (the power hypothesis). The domestic politics perspective, however, appears to be the most useful. All of the evidence presented in this chapter suggests that anti-Americanism as measured by mass public attitudes towards the United States is indeed measuring an independent public opinion that has its own effect on policy. When elite preferences were controlled for (the elite preferences hypothesis), the effect of public anti-Americanism actually appeared to increase and its effect was moderated by the domestic salience of the war, as would be expected when estimating the effect of public opinion on policy (the issue salience hypothesis). In short, the American public and U.S. leaders appear to have been right to be concerned about the increasingly negative global sentiment towards the United States that developed during the Bush Administration’s tenure. While this chapter cannot claim to have established a causal relationship between public attitudes towards the United States  161  and other countries’ decisions to participate in or withdraw from the operation in Iraq, the association between the two variables was consistent across a variety of scenarios and robust when controlling for a variety of other potential influences on participation or withdrawal. Together, this empirical evidence suggests that U.S. public favorability did play a role in countries deployment decisions. Whether this sort of effect will be detected for other military operations or in policy domains where the U.S. is less closely associated with the policy, however, remains to be seen, and is left to future research. The findings presented in this chapter highlight the relevance of anti-American opinion to global political behavior, and the need for more systematic research into when general public attitudes towards the United States – or, indeed, towards other countries or organizations – are likely to influence responses to foreign policy initiatives associated with those actors. Future work should also investigate the mechanisms through which these attitudes are likely to influence policy outcomes, whether indirectly through public opinion or directly by shaping elite opinion about specific policy initiatives. Together, the findings presented in this chapter and the future research suggested above should help scholars and practitioners to better understand when and how public diplomacy and country image branding can be a useful tool of foreign policy.  162  Chapter 6  Conclusion  This dissertation sought to accomplish two principal goals: 1) to evaluate the conditions under which source cues influence transnational persuasion and 2) to better understand how anti-Americanism and the global image of the U.S. President constrain and enable U.S. soft power. Chapters 2 and 3 focus primarily on the first objective, while Chapters 4 and 5 deal with the second. In this concluding chapter, I outline a number of conclusions that emerge from the findings described in the substantive chapters, and consider the policy implications of this research. In the next two sections, I consider my findings with reference to each of the two goals listed above. In the final section, I discuss several avenues for future research. 6.1 TRANSNATIONAL SOURCE CUE EFFECTS As I argue in the introduction to this dissertation, the dynamics of source cue effects in the context of transnational persuasion have rarely been studied in both the political psychology and the IR literatures. One of this dissertation’s principal contributions, then, is the identification of a set of variables and conditions that moderate transnational source cue effects. Some of these are consistent with the existing literature on domestic source cue effects, while others are particularly relevant to a transnational context, and have not been studied in the political science literature. Overall, my experiments point to two principal conclusions: 1) that the level of an audience’s familiarity with a message source is a key moderator of transnational source cue effects, and 2) that the environment in which a transnational persuasive communication occurs can shape its effectiveness. I discuss findings relevant to both of these conclusions in the following subsections. 163  6.1.1 Source Familiarity  One of the more robust results presented in Chapters 2 and 3 is that the magnitude of a cross-border presidential source cue effect is greatest among individuals who are most familiar with U.S. politics, as measured by their ability to name the current U.S. Vice-President. My explanation for this finding is that people can only use an actor’s position on a policy as a heuristic device if they actually have a well-defined opinion about that actor. If they do not, the source attribution will not provide any new information other than the fact that someone supports or opposes the policy in question. While it is possible that this finding simply reflects the moderating effect of political sophistication, two more nuanced findings suggest that this is not the case and that the key variable is in fact familiarity with the source. First, as I report in Chapter 3, the moderating effect of source familiarity is distinct from the moderating effect of policy-specific information on the same source cue effect. Specifically, the presidential source cue effects appear to be strongest among individuals who have high source familiarity, but low levels of policy-specific information. Second, as reported in Chapter 2, the moderating effect of source familiarity on persuasiveness appears to be attenuated when the source in question is a country (in this case, the United States) rather than that country’s leader or former leader (the U.S. President). My explanation for this finding is that individuals are more likely to have clear opinions about a corporate actor (particularly a high-profile actor like the United States) than about an individual who is temporarily representing that actor. I argue that this is because, in general, individuals will have access to a wider array of sources of information about the corporate actor (such as personal visits, interactions with citizens or members of that organization, and  164  information about the country’s history, cultural or economic products) than about a specific political leader. Note, moreover, that political sophistication would be expected to moderate transnational source cue effects irrespective of the actor type, which suggests that these results are indeed driven by variation in source familiarity. These two findings have a number of implications for the dynamics of political persuasion in a global context. First, while my argument and results suggest that countries are likely to be more well-known to foreign publics than their respective leaders, and therefore less susceptible to a source familiarity effect, not all countries will be as well known as the United States. McGraw and Dolan (2007), for example, use Moldova as the subject of research in their investigation of state personification precisely because the American public is unlikely to have any opinion of the country in the first place (thus rendering their experimental manipulations more controlled). This dissertation does not investigate what types of countries are most familiar to global publics, but we might speculate, for example, that more powerful states will have a greater global profile than those at the margins of the international system. If this is true, an implication of my research is that countries that are removed from the global “mainstream” (particularly those that are less powerful) are less likely to have transnational source cue effects because potential target publics will not be as familiar with them. A question that follows directly from this expectation is whether actor type influences levels of global familiarity, thereby influencing that actor’s transnational persuasiveness. Are states, on average, more familiar to global publics than nongovernmental organizations (NGOs), inter-governmental organizations (IGOs), or  165  organized activist campaigns? Moreover, is the relationship between a corporate actor and its representatives’ familiarity the same for non-state actors? For example, to the extent that a campaign uses a global celebrity – like Bono (2007) – to spread its message, we would expect that celebrity would be more familiar to global audiences than the organization he is speaking on behalf of. Thus, messages coming from the celebrity would be less susceptible to the effects of familiarity with key global political actors in that issue-area than messages coming from the campaign (because the celebrity, by definition, will be familiar to global publics outside the context of the campaign in question, whereas the campaign will only be familiar to those who know more about that issue). This is precisely the reverse of the relationship I illustrate between the U.S. and its president. Overall, this issue merits more investigation in future work. 6.1.2  Communications Environment  A second major conclusion that can be drawn from this dissertation research is that the effectiveness of transnational persuasive communications will be moderated by the context in which the communication takes place. Specifically, I identify two contextual factors – the way in which the advocated policy is characterized and the attributes of the message source that are salient to the audience – that moderate transnational source cue effects. Crucially, the experimental manipulations that defined these two factors in my research can be seen as close analogues to well-known media effects – framing and priming effects, respectively (Scheufele and Tewksbury 2007). My findings, then, speak to the way in which the news coverage of a policy and of a global political actor can influence efforts by that actor to engage in transnational persuasion.  166  In Chapter 3, I report on the results of the Afghanistan Experiment. Aside from the observed gender gap in responses to the Obama source cue effect (which I discuss below), a key finding of the experiment is that Obama’s persuasiveness was reduced when the International Security and Assistance Force (ISAF) was portrayed as a U.S.-led mission compared to when it was simply presented as a NATO mission. An implication of this finding is that, to the extent that foreign media outlets tend to characterize operations like ISAF as “U.S.-led”, this media framing effect would tend to reduce the ability of the U.S. government to use persuasion as a tool of foreign policy. More broadly, it points to a potentially serious limitation to the effectiveness of transnational persuasion. While the low-information rationality model of persuasion (Lupia and McCubbins 1998; Lupia 2000) argues that the perceived self-interest or altruism of a message source is a key moderator of source cue effects, this finding suggests that it may be particularly difficult for foreign advocates to avoid appearing self-interested. Specifically, one might speculate that it may be easier for opponents of the advocated policy to portray a given persuasive communication as self-interested (or, more precisely, as not having the audience’s interest in mind) when the policy is advocated by an external actor than in a domestic context, where advocates may more credibly argue that the policy is in their and their audience’s collective interest. This research cannot speak to this question, but points to the need for further research on the particular constraints of transnational communications. A second finding that points to the importance of contextual effects on transnational persuasion is the U.S. source cue experiment presented in Chapter 2, which shows that when the Iraq War is primed, the persuasiveness of “the United States” on  167  global climate policy is reduced. Again, this experimental manipulation is analogous to the priming effect that continued coverage of the war in the targeted jurisdiction’s media would have on the targeted public’s perceptions of the source in question. The finding suggests that the media salience of policies associated with a global political actor will moderate that actor’s global persuasiveness – specifically, if a policy that a majority of the targeted audience disagrees with is featured prominently, then the actor’s favorability and persuasiveness will be attenuated. One might also speculate that the moderating effect of policy salience on transnational persuasiveness should be greatest for corporate actors like countries, precisely because they will tend to be more familiar. Less familiar actors, like foreign leaders, will in general have smaller source cue effects (due to the lower levels of aggregate familiarity), but those individuals who are familiar with the leaders should be aware of their policies in the first place, and should therefore be less susceptible to policy priming effects (this is supported by the finding in Chapter 4 that disagreement with U.S. foreign policy appears to be “priced in” to individuals’ opinions of the U.S. President). Overall, both of these findings point to the likely existence of significant media effects on the effectiveness of transnational persuasive communications. Together with the findings on the importance of source familiarity, they point to three potentially generalizeable moderators of transnational persuasion.88 This represents the primary theoretical contribution of this dissertation.  88  I discuss the limitations of the research design and the generalizeability of my findings in the final section of this conclusion.  168  6.2  ACTOR IMAGE AND U.S. SOFT POWER In addition to the theoretical insights described above, this dissertation also has a  number of substantive implications for U.S. foreign policy and the potential for the United States to use soft power. I describe the most robust conclusions below. 6.2.1  U.S. Country Image – Policy Consequences  I begin with Chapter 5, which provides evidence that the U.S.’s global standing (i.e., the level of global anti-Americanism) does matter to policy outcomes. My analyses show that aggregate public favorability towards the United States was associated with an increased probability of joining Operation Iraqi Freedom in 2003 and a decreased probability of withdrawal in subsequent years. This association was particularly strong when the country in question’s involvement in the war was domestically salient (i.e., when it had suffered casualties) and among powerful states, who are presumably less constrained by the U.S.’s hegemony and therefore freer to follow domestic political considerations when making foreign policy decisions. Together with other recent research (Datta 2008, 2009; Goldsmith and Horiuchi n.d.), this work goes a long way towards meeting the “burden of proof” outlined by Katzenstein and Keohane (2007a) on those who argue that anti-Americanism has political consequences. Overall, the findings suggests that the United States government should aim to improve the United States’ global image because a positive global image will facilitate the pursuit of its foreign policy objectives.  169  6.2.2  Presidential Image and U.S. Soft Power  A follow-up question to the finding in Chapter 5, then, is what, if anything, the U.S. President can do – as head of state and of government – to maximize U.S. soft power. In particular: how does the President’s personal global image influence antiAmericanism? This is a particularly relevant question given the recent transition between the extremely unpopular President Bush to the extremely popular (in a global context) President Obama, and the simultaneous dramatic improvement of the U.S.’s global image. Figure 6.1, which uses aggregate data drawn from the Pew Global Attitudes Survey (PGAP) shows the huge gap in popularity between Bush and Obama and the spike in U.S. favorability following Obama’s election and inauguration.89 What, then, is the nature of the Obama effect on U.S. favorability? 6.2.2.1 PRESIDENTIAL POPULARITY AND ANTI-AMERICANISM I address this question in Chapter 4, where I evaluate the effect that priming Presidents Bush or Obama has on participants’ general opinion of the United States. The main effects of the experiment are very clear. Priming Obama increases U.S. favorability, while priming Bush makes participants’ opinion of the United States more negative. The Obama priming effect persists, moreover, when controlling for other potential influences on U.S. favorability, and the predicted magnitude of this effect is substantively meaningful. Other than the overall effect, the most robust finding in Chapter 4 is that, as with presidential source cue effects, familiarity with U.S. politics has  89  The red line represents aggregate favorability scores for the United States averaged across the following countries (for which consistent data was available): Britain, France, Germany, Jordan, Pakistan, Russia, and Turkey. The yellow line shows mean confidence in the U.S. President averaged across the same countries.  170  a strong moderating effect on the Obama priming effect – an effect that persists even when general levels of political information are controlled for.  What does this imply for U.S. foreign policy, then? First, it provides clear evidence that the Obama effect is real,90 and is driven at least in part by a priming effect. This has direct implications for U.S. public diplomacy. As shown by Goldsmith and Horiuchi (2009), high-level visits by U.S. government officials – including the President – can improve foreign attitudes towards the United States and opinion of U.S. policy, but only if the official in question is seen as credible. They find that visits by Bush administration officials after it became clear that Iraq had not been in possession of WMD stockpiles actually backfired, worsening foreign opinion of the United States. This is consistent with the negative priming effect for Bush in my experiments. The 90  Though, of course, the overall Obama effect is likely to involve a number of mechanisms in addition to priming effects, not least being the perceived or actual change in U.S. foreign policy under the new administration.  171  strong positive priming effect for Obama, moreover, suggests that, going forward, foreign visits by Obama are likely to be an effective tool of public diplomacy and (combined with his transnational persuasiveness, which I discuss in the next sub-section), are likely to improve both general attitudes towards the United States and possibly attitudes towards specific U.S. policies. Finally, while the results also suggest that this type of high-level public diplomacy is likely to be most effective in countries where publics already are reasonably familiar with Obama, it also seems likely that high-level visits increase levels of familiarity themselves, since the media coverage accompanying a visit is likely to provide more information about the President than is usually available. This would prompt the visited publics to develop a more distinct opinion about President Obama, thus not only boosting U.S. favorability in the short-term, but also increasing his ability to use his personal popularity to persuade that public in the future (due to an increase in source familiarity). 6.2.2.2 PRESIDENTIAL POPULARITY AND PERSUASION While a popular president can improve the United States’ global image, presidential image also matters because a positive image can provide the U.S. President with significant personal soft power. As we might expect, I find that attributing a policy position to President Obama has a consistent positive source cue effect across a range of issues. My results suggest, however, that President Obama’s personal soft power also faces significant constraints. First, depending on the distribution of information about Obama across the global public, the moderating effect of source familiarity may or may not be a politically relevant limitation to his personal persuasiveness. For example, if allied publics are more familiar with Obama, this would imply that he would be most  172  persuasive when trying to gain actual material support from traditional U.S. allies – say, on financial reform, or for the mission in Afghanistan – but would be least persuasive when trying to convince the publics of potential competitors or enemy countries to be less critical of U.S. foreign policy. Overall, however, it appears unlikely that the U.S. President – one of the most visible global figures – is unknown to most publics, so the moderating effect of familiarity on persuasiveness is likely less relevant to President Obama’s soft power (particularly as his term in office progresses) than to the general study of transnational persuasion. Two other limitations uncovered by my research are, however, more important. First, the Afghanistan Experiment (Chapter 3) suggests that Obama’s personal popularity will be least effective in moving global opinion when he is dealing with an issue – such as the War in Afghanistan – that is perceived to be primarily a U.S. initiative. This would suggest that Obama’s personal soft power is greatest when advocating for a common policy prescription to a common problem (say, a global bank tax as a preventive measure for future financial crises) and weakest when requesting support for a U.S. initiative – such as the war in Afghanistan. Second, the results of this experiment also suggest a gender gap in Obama’s ability to persuade global publics to support his security policies. While he exerted a strong source cue effect on women, the Obama source cue effect among men was non-existent, and even potentially negative. I attribute this (with appropriate caveats) to his low symbolic credibility on security issues due to his nongendered leadership style. If this interpretation is correct, it would imply that Obama will find it particularly difficult to convince the publics most directly affected by U.S. security policy – say, in the Middle East – to support his security policy initiatives, since these  173  happen to be areas of the world where men (and, indeed, women) are particularly likely to value traditional gender roles and hold gendered views of political leadership. Given that one of the selling points of the Obama brand was the President’s alleged ability to communicate in a more effective fashion with Muslim publics, this is, if correct, a rather significant limitation to his personal soft power.91 Overall, then, the results of this research suggest that the U.S.’s country image (anti-Americanism) matters to U.S. foreign policy, that the global image of the U.S. President can improve the U.S.’s country image (thus facilitating its foreign policy) and that, if positive, it can provide the U.S. President with a significant (though limited) personal soft power – namely, the ability to persuade global publics to support U.S. policy. Both of these findings help to explain the mechanisms responsible for and the implications of the “Obama Effect” illustrated in Figure 6.1, and point to the very real advantages to the United States of having a globally popular President in office. 6.3  FUTURE RESEARCH To conclude, I discuss three additional avenues for future research suggested by  my findings that are informed, in particular, by the significant limitations of my experimental research design, which I discussed in Chapter 1. First, the findings in this dissertation should be tested in a wider variety of samples. This would include a national sample in Canada, but, perhaps more importantly, probability or convenience samples in other countries, including countries that are not traditional allies of the United States. The findings regarding anti-Americanism and U.S. soft power would, in particular, benefit from being tested for robustness in this manner. A second set of follow-up 91  Though admittedly, this implication must remain very tentative, given the sample of participants used in these experiments.  174  research would use a wider variety of actors as source cues. It would test the extent to which the expectations derived from this dissertation research apply to global actors that are less prominent than the United States and than the U.S. President. Given the effects of source familiarity described above, the dynamics for unknown actors might be substantially different. Finally, future experiments could aim to exert a greater degree of experimental control on key variables as a means of confirming the expectations suggested by this dissertation. For example, if using relatively unknown actors, participants could be provided with information about the sources prior to receiving the experimental question. This would allow us to manipulate source familiarity and rigorously demonstrate both its moderating effect on transnational source cue effects and its distinctiveness from the general effect of political sophistication.  175  Bibliography Alderson, Kai. 2001. "Making sense of state socialization." Review of International Studies 27:415-33. Amine, Lyn Suzanne. 2008. "Country-of-origin, animosity and consumer response: Marketing implications of anti-Americanism and Francophobia." International Business Review 17:402-22. Archetti, Christina. 2007. "A Multidisciplinary Understanding of News: Comparing Elite Press Framing of 9/11 in the US, Italy, France, and Pakistan." Journal of International Communication 13 (1):86-118. Ashmore, Richard D., David Bird, Frances K. Del Boca, and Robert C. Vanderet. 1979. "An Experimental Investigation of the Double Standard in the Perception of International Affairs." Political Behavior 1 (2). Bailey, Michael, Lee Sigelman, and Clyde Wilcox. 2003. "Presidential Persuasion on Social Issues: A Two-Way Street?" Political Research Quarterly 56 (1):49-58. Baum, Matthew. 2008a. "A Coalition of the Unrestrained: Mass Media, Electoral Institutions and the Constraining Effect of Public Opinion Regarding Iraq." In American Political Science Association. Boston, M.A. Baum, Matthew. 2008b. "Coalition of the Unconstrained: Mass Media, Electoral institutions and the Constraining Effect of Global Opinion Regarding Iraq." In Annual Meeting of the International Studies Association. San Francisco, California. Beck, Nathaniel, Jonathan N. Katz, and Richard Tucker. 1998. "Taking Time Seriously: Time-Series-Cross-Section Analysis with a Binary Dependent Variable." American Journal of Political Science 42 (4):1260-88. Belanger, eric. 2003. "Issue Ownership by Canadian Political Parties 1953-2001." Canadian Journal of Political Science 36 (3):539-58. Benford, Robert, and David Snow. 2000. "Framing Processes and Social Movements: An Overview and Assessment." Annual Review of Sociology 26:611-39. Beyers, Jan. 2005. "Multiple Embeddedness and Socialization in Europe: The Case of Council Officials." International Organization 59:899-936. Brambor, Thomas, William Roberts Clark, and Matt Golder. 2006. "Understanding Interaction Models: Improving Empirical Analysis." Political Analysis 14:63-82. Brug, Wouter van der. 2004. "Issue ownership and party choice." Electoral Studies 23:209-33. Bullock, John. 2008. "Party Cues and Policy Information: Real and Perceived Effects." In Comparative-Canadian Workshop. Vancouver, B.C.: Department of Political Science, University of British Columbia. Busby, Joshua William. 2007. "Bono Made Jesse Helms Cry: Jubilee 2000, Debt Relief, and Moral Action in International Politics." International Studies Quarterly 51 (2):247-76. ———. 2010. "From God's mouth: messenger effects and donor responses to HIV/AIDS." In Moral Movements and Foreign Policy. Cambridge, U.K.: Cambridge University Press.  176  Carlson, Matthew, and Travis Nelson. 2008. "Anti-Americanism in Asia? Factors shaping international perceptions of American influence." International Relations of the Asia-Pacific 8:303-24. Cass, Loren. 2005. "Norm Entrapment and Preference Change: The Evolution of the European Union Position on International Emissions Trading." Global Environmental Politics 5 (2). Checkel, Jeffrey T. 1999. "Norms, Institutions, and National Identity in Contemporary Europe." International Studies Quarterly 43:83-114. ———. 2005. "International Institutions and Socialization in Europe: Introduction and Framework." International Organization 59:801-26. Chicago, Council on Foreign Relations, and German Marshall Fund of the United States. 2002. "American Public Opinion and U.S. Foreign Policy." Ann Arbor, MI: InterUniversity Consortium for Political and Social Research [distributor]. Chiozza, Giacomo. 2007. "Disaggregating Anti-Americanism: An Analysis of Individual Attitudes toward the United States." In Anti-Americanisms in World Politics, ed. P. Katzenstein and R. Keohane. Ithaca, NY: Cornell University Press. Chong, Dennis, and James N. Druckman. 2007a. "Framing Public Opinion in Competitive Democracies." American Political Science Review 101 (4). ———. 2007b. "Framing Theory." Annual Review of Political Science 10:103-26. Coleman, Katharina. 2008. "Building International Legitimacy: The Significance of Token Forces in Contemporary Military Interventions." In International Studies Association. San Francisco, CA. Cooper, Frank Rudy. 2009. "Our First Unisex President?: Black Masculinity and Obama's Feminine Side." Denver University Law Review 86. Cortell, Andrew P., and James W. Davis. 1996. "How Do International Institutions Matter? The Domestic Impact of International Rules and Norms." International Studies Quarterly 40 (4):451-78. ———. 2000. "Understanding the Domestic Impact of International Norms: A Research Agenda." International Studies Review 2 (1):65-87(23). Curley, Tyler M. 2009. "Social Identity Theory and eu expansion." International Studies Quarterly 53 (3):649-68. Datta, Monti. 2008. "Anti-Americanism and Support for the U.S.-led War in Iraq." In Annual Meeting of the Western Political Science Association. San Diego, California. ———. 2009. "The Decline of American Soft Power in the United Nations." International Studies Perspectives 10 (3):266-85. Druckman, James N. 2001a. "On the Limits of Framing Effects: Who Can Frame?" Journal of Politics 63 (4):1041-66. ———. 2001b. "The Implications of Framing Effects for Citizen Competence." Political Behavior 23 (3). Duerst-Lahti, Georgia. 2005. "Mommy Party, Daddy Party: The Gendering of Parties in the 2004 Presidential Election." In Midwest Political Science Association. Chicago, IL. Eagley, Alice H., and Shelly Chaiken. 1993. The Psychology of Attitudes. Orlando, Florida: Harcourt Brace Jovanovich, Inc.  177  Edwards, George C. III. 2009. The Strategic President: Persuasion and Opportunity in Presidential Leadership. Princeton, NJ: Princeton University Press. Eichenberg, Richard C. 2003. "Gender Differences in Public Attitudes towards the Use of Force by the United States, 1990-2003." International Security 28 (1):110-41. ———. 2007. "Gender Differences in Support for the Use of Military Force in CrossNational Perspective: The War System, Modernization, and the Universal Logics of Military Action." In Midwest Political Science Association. Chicago. Finnemore, Martha. 1996a. National Interests in International Society. Ithaca, NY: Cornell University Press. ———. 1996b. "Norms, Culture, and World Politics: Insights from Sociological Institutionalism." International Organization 50 (2):325-47. Finnemore, Martha, and Kathryn Sikkink. 1998. "International Norm Dynamics and Political Change." International Organization 52:887-917. Flockhart, Trine. 2005. "Critical Junctures and Social Identity Theory: Explaining the Gap between Danish Mass and Elite Attitudes to Europeanization." Journal of Common Market Studies 43 (2):251-71. Gaines, Brian J., James H. Kuklinski, and Paul J. Quirk. 2007. "The Logic of the Survey Experiment Reexamined." Political Analysis 15:1-20. Gartner, Scott Sigmund. 2008. "The Multiple Effects of Casualties on Public Support for War: An Experimental Approach." American Political Science Review 102 (1):95-106. Goldsmith, Benjamin E., and Yusaku Horiuchi. 2009. "Spinning the Globe? U.S. Public Diplomacy and Foreign Public Opinion." The Journal of Politics 71:863-75. ———. 2010. "U.S. Foreign Policy and Global Public Opinion." ———. n.d. "Does Foreign Public Opinion Matter for U.S. Foreign Policy?" University of Sydney; Australian National University. Goldsmith, Benjamin, Yusaku Horiuchi, and Takashi Inoguchi. 2005. "American Foreign Policy and Global Opinion: Who Supported the War in Afghanistan?" Journal of Conflict Resolution 49. Goldstein, Joshua S. 2001. War and Gender: How Gender Shapes the War System and Vice Versa. Cambridge, UK: Cambridge University Press. Goldstein, Richard. 2003. "Neo-Macho Man: Pop Culture and Post-9/11 Politics." The Nation, March 24, 2003, 16-9. Goren, Paul, Chrisopher M. Federico, and Miki Caul Kittilson. 2009. "Source Cues, Partisan Identities, and Political Value expression." American Journal of Political Science 53 (4):805-20. Gourevitch, Peter. 1978. "Second Image Reversed: The International Sources of Domestic Politics." International Organization 32:881-911. Graber, Doris A. 2009. "Looking at the United States Through Distorted Lenses: Entertainment Television Versus Public Diplomacy Themes." American Behavioural Scientist 52 (5):735-54. Haas, Peter. 1992. "Introduction: Epistemic Communities and International Policy Coordination." International Organization 46 (1):1-35. Hafner-Burton, Emilie M. 2008. "Sticks and Stones: Naming and Shaming the Human Rights Enforcement Problem." International Organization 62:689-716.  178  Hayes, Danny, and Matt Guardino. 2010. "Whose Views Made the News? Media Coverage and the March to War in Iraq." Political Communication 27 (1):59-87. Howell, William, and Douglas Kriner. Unpublished. "Political Elites and Public Support for War." Chicago, Illinois: University of Chicago, Harris School of Public Policy. Inglehart, Ronald, and Pippa Norris. 2000. "The Developmental Theory of the Gender Gap: Women's and Men's Voting Behavior in Global Perspective." International Political Science Review 21. Isernia, Pierangelo. 2007. "Anti-Americanism in Europe during the Cold War." In AntiAmericanisms in World Politics, ed. P. Katzenstein and R. Keohane. Ithaca, NY: Cornell University Press. Iyengar, Shanto, and Donald Kinder. 1987. The News That Matters. Jervis, Robert. 1976. Perception and Misperception in International Politics. Princeton, NJ: Princeton University Press. Jhee, Byong-Kuen. 2008. "Anti-Americanism and electoral Politics in Korea." Political Science Quarterly 123 (2):301-18. Johnston, Alastair Iain, and Daniela Stockmann. 2007. "Chinese Attitudes toward the United States and Americans." In Anti-Americanisms in World Politics, ed. P. J. Katzenstein and R. O. Keohane. Ithaca, NY: Cornell University Press. Johnston, Alistair Iain. 2001. "Treating International Institutions as Social Environments." International Studies Quarterly 45 (4):487-515. Kagan, Robert. 2003. Of Paradise and Power: America and Europe in the New World Order. New York: Alfred A. Knopf. Katzenstein, Peter J., and Robert O. Keohane. 2007a. "The Political Consequences of Anti-Americanism." In Anti-Americanisms in World Politics, ed. P. J. Katzenstein and R. O. Keohane. Ithaca, NY: Cornell University Press. ———, eds. 2007b. Anti-Americanism in World Politics. Edited by P. J. Katzenstein. Ithaca, NY: Cornell University Press. Keck, Margaret, and Kathryn Sikkink. 1998. Activists Beyond Borders. Ithaca, NY: Cornell University Press. King, Gary, Michael Tomz, and Jason Wittenberg. 2000. "Making the Most of Statistical Analyses: Improving Interpretation and Presentation." American Journal of Political Science 44 (2):341-55. Kittilson, Miki Caul. 2007. "Research Resources in Comparative Political Behavior." In The Oxford Handbook of Political Behavior, ed. R. J. Dalton and H.-D. Klingemann. New York, NY: Oxford University Press. Klotz, Audie. 1995. "Norms reconstituting interests: global racial equality and U.S. sanctions against South Africa." International Organization 49 (3):451-78. Kohut, Andrew, Richard Wike, erin Carriere-Kretschmer, and Kathleen Holzwart. 2008. "Global Public Opinion the Bush Years (2001-2008): America's Image, Muslims and Westerners, Global economy, Rise of China." Washington, D.C.: The Pew Global Attitudes Project, Pew Research Center. Kohut, Andrew, Richard Wike, erin Carriere-Kretschmer, Kathleen Holzwart, and Jacob Poushter. 2009. "Most Muslim Publics Not So easily Moved: Confidence in Obama Lifts U.S. Image Around the World, 25-Nation Pew Global Attitudes Survey." Washington, D.C.: The Pew Global Attitudes Project.  179  Krebs, Ronald R., and Patrick Thaddeus Jackson. 2007. "Twisting Tongues and Twisting arms: The Power of Political Rhetoric." European Journal of International Relations 13 (1):35-66. Krueger, Alan B., and Jitka Maleckova. 2009. "Attitudes and Action: Public Opinion and the Occurence of International Terrorism." Science 325:1534-. Kull, Steven, and I.M. Destler. 1999. Misreading the Public: The Myth of a New Isolationism. Washington, D.C.: Brookings Institution Press. Legro, Jeffrey W. 1997. "Which Norms Matter? Revisiting the "Failure" of Internationalism." International Organization 51 (1):31-63. Liebovich, Mark. 2007. "Leftward, Ho?" The New York Times, February 18, 2007. Lupia, Arthur. 2000. "Who Can Persuade Whom?: Implications from the Nexus of Psychology and Rational Choice Theory." In Thinking About Political Psychology, ed. J. H. Kuklinski. Cambridge, United Kingdom: Cambridge University Press. Lupia, Arthur, and Mathew D. McCubbins. 1998. The Democratic Dilemma: Can Citizens Learn What They Need to Know? New York: Cambridge University Press. Lynch, Marc. 2007. "Anti-Americanisms in the Arab World." In Anti-Americanisms in World Politics, ed. P. J. Katzenstein and R. O. Keohane. Ithaca, NY: Cornell University Press. McGraw, Kathleen M., and Thomas M. Dolan. 2007. "Personifying the State: Consequences for Attitude Formation." Political Psychology 28 (3):299-327. Mearsheimer, John J. 1994/5. "The False Promise of International Institutions." International Security 19:5-49. Mercer, Jonathan. 1995. "Anarchy and identity." International Organization 49 (2):22952. Miller, Kristina C. 2009. "The Limitations of Heuristics for Political Elites." Political Psychology 30 (6):863-94. Mintz, Alex, Steven B. Redd, and Arnold Vedlitz. 2006. "Can We Generalize from Student Experiments to the Real World in Political Science, Military Affairs, and International Relations?" Journal of Conflict Resolution 50:757-76. Mondak, Jeffery J. 1993. "Source Cues and Policy Approval: The Cognitive Dynamics of Public Support for the Reagan Agenda." American Journal of Political Science 37 (1):186-212. Mondak, Jeffrey J., Christopher J. Lewis, Jason C. Sides, Joohyun Kang, and J. Olyn Long. 2004. "Presidential Source Cues and Policy Appraisals, 1981-2000." American Politics Research 32 (2):219-35. Moravcsik, Andrew. 1997. "Taking Preferences Seriously: A Liberal Theory of International Politics." International Organization 51 (4):513-53. Nelson, Thomas E., Zoe M. Oxley, and Rosalee A. Clawson. 1997. "Toward a Psychology of Framing Effects." Political Behavior 19 (3):221-46. Nisbet, Erik C., Matthew C. Nisbet, Dietram A. Scheufele, and James E. Shanahan. 2004. "Public Diplomacy, Television News, and Muslim Opinion." Press/Politics 9 (2):11-37.  180  Nossal, Kim Richard. 2005. "Anti-Americanism in Canada." In Anti-Americanism Working Papers. Budapest, Hungary: Center for Policy Studies, Central European University. Nye, Joseph. 2004. Soft Power: The Means to Success in World Politics. Public Affairs: New York, NY. Nye, Joseph S. 2008. "Public Diplomacy and Soft Power." ANNALS of the American Academy of Political and Social Science 616:94-109. Payne, Rodger A. 2001. "Persuasion, Frames and Norm Construction." European Journal of International Relations 7 (1):37-61. Pelika, Stacey L. 2007. Picking Teams and Choosing Sides: Opinions As Symbols in the Policy Process, Political Science, University of Wisconsin - Madison, Madison, Wisconsin. Petrocik, J.R. 1996. "Issue ownership in presidential elections, with a 1980 case study." American Journal of Political Science 40 (3):825-50. Pornpitakpan, Chanthika. 2004. "The Persuasiveness of Source Credibility: A Critical Review of Five Decades' Evidence." Journal of Applied Social Psychology 34 (2):243-81. Pratto, F., L.M. Stallworth, and J. Sidanius. 1997. "The gender gap: differences in political attitudes and social dominance orientation." British Journal of Social Psychology 36 (1):49-68. Price, Richard. 1998. "Reversing the Gun Sights: Transnational Civil Society Targets Land Mines." International Organization 52 (3):613-44. ———. 2003. "Transnational Civil Society and Advocacy in World Politics." World Politics 55 (4):579-606. Putnam, Robert. 1988. "Diplomacy and Domestic Politics: The Logic of Two-Level Games." International Organization 42 (3):427-60. Ray, Leonard, and Gregory Johnston. 2007. "European Anti-Americanism and Choices for a European Defense Policy." Political Science & Politics XL (1):85-91. Risse, Thomas. 2000. "Let's Argue: Communicative Action in World Politics." International Organization 54:1-39. Risse, Thomas, and Kathryn Sikkink. 1999. "The Socialization of International Human Rights Norms into Domestic Practices: Introduction." In The Power of Human Rights: International Norms and Domestic Change, ed. T. Risse, S. C. Ropp and K. Sikkink. Cambridge, U.K.: Cambrdige University Press. Risse-Kappen, Thomas. 1991. "Public Opinion, Domestic Structure, and Foreign Policy in Liberal Democracies." World Politics 43 (4):479-512. ———. 1996. "Collective identity in a democratic community: The case of NATO." In The Culture of National Security: Norms and Identity in World Politics, ed. P. J. Katzensteion. New York: Columbia University Press. Robinson, Lynn, and Paul Goren. 1997. "Religious Conservative Elites as Source Cues." Social Science Quarterly 78 (1):223-33. Rugh, William A. 2009. "The Case for Soft Power." In Toward a New Public Diplomacy: Redirecting U.S. Foreign Policy, ed. P. Seib. New York, NY: Palgrave Macmillan.  181  Scheufele, Dietram A., and David Tewksbury. 2007. "Framing, Agenda Setting, and Priming: The Evolution of Three Media Effects Models." Journal of Communication 57:9-20. Schimmelfennig, Frank. 2005. "Strategic Calculation and International Socialization: Membership Incentives, Party Constellations, and Sustained Compliance in Central and Eastern Europe." International Organization 59:827-60. Shannon, Vaughn P. 2000. "Norms Are What States Make of Them: The Political Psychology of Norm Violation." International Studies Quarterly 44:293-316. Sigelman, Lee, and Carol K. Sigelman. 1981. "Presidential Leadership of Public Opinion: From "Benevolent Leader" to "Kiss of Death"?" Experimental Study of Politics 7 (3):1-22. Smith, Jackie. 2002. "Bridging Global Divides? Strategic Framing and Solidarity in Transnational Social Movement Organizations." International Sociology 17 (4):505-28. Sniderman, Paul, Richard Brody, and Philip E. Tetlock. 1991. Reasoning and Choice: Explorations in Political Psychology. New York: Cambridge University Press. Sniderman, Paul, Joseph F. Fletcher, Peter Russell, and Philip E. Tetlock. 1996. The Clash of Rights: Liberty, Equality, and Legitimacy in Pluralist Democracy. New Haven, CT: Yale University Press. Snow, Nancy. 2007. The Arrogance of American Power: What U.S. Leaders Are Doing Wrong and Why It's Our Duty to Dissent. Oxford, UK: Rowman & Littlefield Publishers, Inc. Soroka, Stuart N. 2003. "Media, Public Opinion, and Foreign Policy." Press/Politics 8 (1):27-48. Sundstrom, Lisa McIntosh. 2005. "Foreign Assistance, International norms, and NGO Development: Lessons from the Russian Campaign." International Organization 59:419-49. Swers, Michele. 2007. "Building a Reputation on National Security: The Impact of Stereotypes Related to Gender and Military experience." Legislative Studies Quarterly 32 (4):559-. Tarrow, Sidney. 2001. "Transnational Politics: Contention and Institutions in International Politics." Annual Review of Political Science 4:1-20. Waltz, Kenneth. 1979. Theory of International Politics. New York, NY: McGraw-Hill. Wapner, Paul. 1995. "Politics Beyond the State: Environmental Activism and World Civic Politics." World Politics 47:311-40. Wendt, Alexander. 1999. Social Theory of International Politics. Cambridge, U.K.: Cambridge University Press. Willnat, Lars, Annette Aw, Naila Nabil Hamdy, Zhou He, Victor Menayang, Maria Teresa La Porte, Karen Sanders, and Ezhar Tamam. 2006. "Media Use, AntiAmericanism and International Support for the Iraq War." International Communication Gazette 68:533. Wlezien, Christopher, and Stuart N. Soroka. 2007. "The Relationship between Public Opinion and Policy." In The Oxford Handbook of Political Behaviour, ed. R. J. Dalton and H.-D. Klingemann. Oxford, U.K.: Oxford University Press.  182  Zaller, John. 2003. "Coming to Grips with V.O. Key's Concept of Latent Opinion." In Electoral Democracy, ed. G. Rabinotwitz and M. B. MacKuen. Ann Arbor, MI: University of Michigan Press. Zaller, John R. 1992. The Nature and Origins of Mass Opinion. Cambridge and New York: Cambridge University Press.  183  APPENDICES APPENDIX A: QUESTION WORDING FOR SURVEY EXPERIMENTS Mandatory Emissions Limits (MEL) Experiment The next few questions are about government policies to deal with the problem of global warming. Some governments have passed laws that limit the amount of carbon dioxide that companies are allowed to emit. If a company's emissions are over this limit, the government can then fine it or even shut it down. Opponents of this approach, however, like former U.S. President George W. Bush, like former U.S. President Bill Clinton, like U.S. President Barack Obama, like the United States, say that these laws will only end up hurting the economy. Instead, they former-President Bush former-President Clinton President Obama the United States propose to encourage companies to limit their carbon dioxide emissions voluntarily.  Please tell us if you agree or disagree with the following statement: The Canadian government should pass laws that limit the amount of carbon dioxide that Canadian industries are allowed to emit. ___ Agree strongly ___ Agree ___ Neither agree nor disagree ___ Disagree ___ Disagree strongly  184  Climate Treaty (Presidents) Experiment There is currently a disagreement about what countries should join the new international treaty on limiting carbon emissions, which are the main cause of climate change. Some political leaders [,like U.S. President Barack Obama,] {like former U.S. President George W. Bush} argue that developing countries like China and India should be allowed to focus on increasing the living standards of their citizens before they pledge to limit their carbon emissions. Others {,like U.S. President Barack Obama, } {like former U.S. President George W. Bush} argue that in order to stop global warming, ALL countries must commit to limit their carbon emissions in the next treaty on climate change.  Do you agree or disagree with the following statement?: Canada should join a new international treaty on climate change EVEN IF China and India refuse to sign it. ___ Strongly agree ___ Somewhat agree ___ Neither agree nor disagree ___ Somewhat disagree ___ Strongly disagree  Climate Treaty (U.S.) Experiment Now, let's return to global warming. There is currently a disagreement about which countries should make the greatest effort to reduce their carbon dioxide emissions.  185  Some argue that developing countries like China and India should focus on increasing the living standards of their citizens before they try to limit their carbon dioxide emissions.  On the other hand, the United States argues some argue that in order to stop global warming, all countries must begin to limit their carbon dioxide emissions as soon as possible.  Over the past two years, Canada and the United States have Canada has refused to sign any new international treaty on reducing carbon dioxide emissions unless China and India also agree to sign it. Do you approve or disapprove of Canada's position? ___ Strongly approve ___ Somewhat approve ___ Neither approve nor disapprove ___ Somewhat disapprove ___ Strongly disapprove Iraq Prime Questions Q1. Moving on to other topics: Do you approve or disapprove of the United States' decision to invade Iraq in 2003? Q2. And do you approve or disapprove of how the United States is handling the military situation in Iraq today?  186  Afghanistan Experiment [Introduction] As you may know, Canada has deployed approximately 3,000 combat troops in Afghanistan since 2001 as part of a NATO mission led by the United States. The Canadian general leading this force reports to NATO’s commanding general in Afghanistan, U.S. Army General Stanley McChrystal.  At a recent NATO summit, supporters of the mission, like U.S. President Barack Obama, called on the NATO allies to maintain or increase the number of combat forces they have deployed in Afghanistan. They argue President Obama argues that a stable Afghanistan is crucial to European and North American security. [Policy Question] Would you support or oppose continuing the deployment of Canadian troops in Afghanistan for another 3 years? Control Variables Policy Agreement Battery [Participants were asked if they supported or opposed the following policies, with support measured on a five point scale from “strongly oppose” to “strongly support”. Responses were used to create a “policy agreement” additive index variable.] Q (Iran): Using military force against Iran to prevent it from developing nuclear weapons.  187  Q (Detention): Setting up a special prison system to hold terrorist suspects indefinitely without pressing criminal charges. Q (Torture): Using coercive techniques to interrogate persons suspected of planning terrorist attacks. Q (Assassinate): Assassinating individual terrorist leaders. Attention to News Q: Some people spend a lot of time following the news; others do not. In the last few days, how much time would you say you've spent each day reading, watching, or listening to political and governmental news? ___ Haven’t spent any time ___ 1-10 minutes ___ 11-30 minutes ___ 31-60 minutes ___ 61-120 minutes ___ More than two hours  188  APPENDIX B: MEASURES OF ANTI-AMERICAN ATTITUDES Survey Organization  Question  Countries Surveyed  Pew Global Attitudes Project  Please tell me if you have a very favorable, somewhat favorable, somewhat unfavorable, or very favorable opinion of... the United States?  Latinobarometro  And do you have a very good, good, bad or very bad opinion of the United States? Also don't know and no answer. U.S. plays a positive or negative role in...? 1) fight against terrorism, 2) growth of the world economy, 3) peace in the world, 4) fight against poverty in the world, 5) protection of the environment.93  Angola, Argentina, Bangladesh, Bolivia, Brazil, Bulgaria, Bulgaria, Canada, Czech Republic, Egypt, France, Germany, Ghana, Guatemala, Honduras, India, Indonesia, Italy, Ivory Coast, Japan, Jordan, Kenya, Lebanon, Mali, Mexico, Nigeria, Pakistan, Peru, Philippines, Poland, Russia, Senegal, Slovakia, South Africa, South Korea, Spain,92 Tanzania, Turkey, Uganda, Ukraine, United Kingdom, Uzbekistan, Venezuela, Vietnam Chile, Colombia, Costa Rica, Ecuador, El Salvador, Nicaragua, Panama, Paraguay, Uruguay Austria, Belgium, Denmark, Finland, Greece, Ireland, Luxembourg, Netherlands, Portugal, Sweden  Eurobarometer  92  Spain’s favorability score is not strictly from the 2002 PGAP survey, but rather is drawn from a list of pre-2002 aggregate favorability scores sourced to the U.S. State Department that are presented in the 2002 PGAP survey report, also available on the PGAP website. 93 Aggregate positivity data is obtained by averaging aggregate positivity scores for the five questions for each country, as described.  189  APPENDIX C: PROCEDURE FOR ESTIMATING FAVORABILITY SCORES FROM EUROBAROMETER POSITIVITY SCORES The averaged positivity scores computed from the Eurobarometer survey are similar to the more general sentiment question asked in Program on International Policy Attitudes (PIPA) surveys.94 I collected PIPA positivity data for the years 2002-2007, adding Eurobarometer data for countries and years in which PIPA data was not available. I then regressed country-year favorability scores (drawn from PGAP and Latinobarometro surveys) on the positivity scores for the same country-years. Table C1 shows the results.  Table C1: Predicting Country-Year Favorability Coefficient Aggregate U.S. Positivity  1.217 (0.256)  Constant  12.853 (7.489)  N= Adjusted R-squared =  24 0.48  The positivity scores are a significant predictor of aggregate favorability, with a reasonably large R-squared value. Interestingly, positivity scores are consistently smaller than favorability scores. Using this estimated equation, I then calculated the predicted aggregate favorability scores for the ten European countries for which Eurobarometer positivity data is available for 2002, but for which direct favorability data from PGAP is 94  Reports are available at < http://www.worldpublicopinion.org/>. The standard sentiment question is: “Is the United States playing a mainly positive or mainly negative role in the world?”  190  not. Table C2 shows the Eurobarometer positivity scores and the predicted favorability scores. This method, while admittedly imperfect, provides a better estimate of aggregate favorability than would simply using the averaged positivity scores, and, crucially, increase the sample size from 51 countries to 61 countries.  Table C2: Predicted Favorability Scores  Country Belgium Denmark Netherlands Portugal Austria Finland Greece Ireland Luxembourg Sweden  Eurobarometer 2002 Positivity Score 33.4 34.6 29.6 32.8 25.0 33.4 17.0 44.6 30.2 29.4  Predicted Favorability Score 53.50 54.96 48.88 52.77 43.28 53.50 33.54 67.13 49.61 48.63  191  APPENDIX D: WITHDRAWAL DATES AND SOURCES FOR OPERATION IN IRAQ Table D1: Withdrawal Dates and Sources for Participants in the Iraq Multi-National Force (MNF) Country Bulgaria Czech Republic Denmark  Final Withdrawal Date December 2008 December 2008 December 2008  Dominican Republic El Salvador Honduras Italy Japan Netherlands Nicaragua Philippines Poland Portugal Slovakia Spain Ukraine United Kingdom  May 2004 February 2009 May 2004 December 2006 December 2008 March 2005 February 2004 July 2004 October 2008 February 2005 December 2007 May 2004 December 2008 Still Deployed  Sources http://www.thenews.com.pk/updates.asp?id=62711 http://www.globalsecurity.org/wmd/library/news/iraq/2008/12/iraq-081205-mnfi01.htm 1) http://www.icenews.is/index.php/2009/01/18/last-danish-soldiers-return-home-fromiraq/ 2)http://www.diis.dk/graphics/Publications/Books2008/Yearbook08/DIIS_Yearbook_2008 .pdf http://english.people.com.cn/200503/16/eng20050316_177085.html http://www.channelnewsasia.com/stories/afp_world/view/407533/1/.html http://www.chinadaily.com.cn/english/doc/2004-05/19/content_331956.htm http://www.forbes.com/feeds/afx/2006/12/01/afx3219440.html http://www.npr.org/templates/story/story.php?storyId=98385367 http://www.accessmylibrary.com/coms2/summary_0286-19012187_ITM http://english.people.com.cn/200503/16/eng20050316_177085.html http://english.people.com.cn/200407/20/eng20040720_150208.html http://www.unian.net/eng/news/news-281051.html http://news.bbc.co.uk/2/hi/europe/4254279.stm http://news.bbc.co.uk/2/hi/europe/country_profiles/1870906.stm http://news.bbc.co.uk/2/hi/middle_east/3734751.stm http://www.army.mil/-news/2008/12/11/15056-ukrainians-complete-mission-in-iraq/ (as of May 2009)  192  APPENDIX E: BEHAVIOURAL RESEARCH ETHICS BOARD CERTIFICATES  193  194  195  196  197  198  199  200  201  202  203  204  205  206  207  208  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0071334/manifest

Comment

Related Items